The separation of N
-linked glycans by RP chromatography affords several advantages including an increase in separation efficiency, an increase in the number of glycans detected, and practical advantages such as the ability to use a more widely employed LC-MS technique (e.g. proteomic analyses). In order to demonstrate these advantages, two samples, maltodextrin and N-
linked glycans cleaved from pooled human plasma, were used to compare the effectiveness of HILIC and RP chromatography on the separation of both native and hydrazide derivatized glycans. As HILIC has been shown to effectively retain and separate native and derivatized glycans,10, 13
this was used as the control separation method. The N
-linked glycans cleaved from plasma and the maltodextrin samples were first analyzed by HILIC coupled online with LTQ FT-ICR MS in order to insure sample integrity. Aliquots of the same samples (stored frozen at ‒20°C) were then analyzed by online RP chromatography coupled to FT-ICR MS using the same LC instrument platform.
contains the base peak chromatograms for both native and derivatized maltodextrin samples on HILIC (1a
, respectively) and RP chromatography (1c
, respectively). As shown previously, both native and derivatized glycans are retained and separated by HILIC.10, 13-15, 22
In contrast, shows that there is no retention of native maltodextrin on the RP column, as is expected due to the polar nature of the glycans and the non-polar nature of the C18
stationary phase. Upon derivatization of maltodextrin, shows that maltodextrin is capable of being separated and retained on a RP C18
column. It must be mentioned that these are the optimized HILIC and RP gradients for derivatized glycans. Though the gradient for RP analysis of native glycans begins at >20% B, studies were also performed in RP starting at 2% B in order to optimize the native glycan retention. However, there were no conditions in which native glycans were retained on RP chromatography.
The base peak chromatograms of both native and derivatized (bold with *) maltodextrin (Hexn) oligosaccharides analyzed by HILIC (a and b, respectively) and RP chromatography (c and d, respectively).
Several additional important observations are made from the base peak chromatograms in . The HILIC base peak chromatograms show that the separation efficiency for both native and derivatized glycans varies as a function of the glycan retention time. This is in contrast to the separation by RP chromatography, where the peak widths are uniform throughout the chromatogram. It is important in the analysis of N
-linked glycans to have as narrow peak widths as possible in order to reduce the amount of possible co-elution and peak suppression due to the more easily ionized or more abundant glycans. In the HILIC chromatograms, the glycans that elute over a large time window are the more hydrophobic and more abundant glycans, meaning that these glycans will likely elute at the same time as, and out-compete, the more hydrophilic, less abundant glycans. Figure S1 in the supplemental material
contains example extracted ion chromatograms (EIC) for both HILIC and RP chromatography to further demonstrate the difference in chromatographic efficiency between the two approaches.
Though the fact that HILIC is capable of retaining both native and derivatized N-linked glycans is often viewed as an advantage, shows that these two species often have overlapping analytes, and this is a disadvantage when one is only interested in analyzing the derivatized N-linked glycans. During derivatization, ≤5% of each glycan remains untagged. Though this seems like a small number, when the dynamic range of the sample comprises orders of magnitude, 5% of an abundant analyte is capable of out-competing lower abundant analytes in the electrospray droplet (e.g. Hex7 and *Hex14 in ). In contrast, the RP chromatography of derivatized glycans has no contamination by the un-reacted glycan. This is a significant advantage in comparison to HILIC and decreases the competition in the electrospray droplet and AGC ion trap capacity for the low abundant glycans.
The final important piece of data from is the total number of unique glycan compositions detected in each chromatographic system. There are 12 glycan compositions detected in , 16 in , 0 in , and 23 in . This shows that independent of the chromatographic technique, it is always better to derivatize the N
-linked glycans with hydrophobic hydrazide reagents than to analyze their native counterparts. However, also shows that it is significantly more productive (~50% more compositions detected) to separate derivatized glycans by RP chromatography than HILIC. There are several differences in the two chromatography techniques that are possible reasons for this outcome. First, as mentioned above (Figure S1
), the efficiency of RP chromatography is more uniform throughout the glycan retention window, and in general, the peak widths are narrower. The oligosaccharide peak widths are compared in the supplemental material
for the same 15 glycans (only oligosaccharides that were detected in both HILIC and RP techniques). The HILIC peak widths have a range of 0.34-10.37 min with a mean of 2.23 min and standard deviation of 2.87 (Table S1
). In contrast, the RP peak widths have a range of 0.34-1.1 min with a mean of 0.49 min and a standard deviation of 0.18 (Table S1
). Because peak broadening spreads the analytes out over a longer retention time, the detection limit will be lower for chromatographic techniques with narrower peak widths (i.e
., higher concentration). Additionally, the elution overlap of un-reacted glycans with lower abundant, derivatized glycans in HILIC causes competition in the electrospray droplet. This hinders the ionization of the lower abundant glycans, causing fewer glycans to be detected in HILIC.
Furthermore, the efficiency of HILIC is often increased and the peak tailing minimized by addition of an acidic buffer such as ammonium acetate or ammonium formate.19, 68
However, the addition of ammonium salts into the buffer often has an adverse affect on detection due to the ammonium adduction in ESI. shows the difference in mass spectra between HILIC (top/blue) and RP chromatography (bottom/red) for three different glycan compositions ( from maltodextrin and cleaved from human plasma – vide infra
). It must be noted that all the glycans are detected in light and heavy pairs due to the derivatization with both the native and stable-isotope labeled glycan reagents (as published previously for the relative quantification of N
). This was necessary to confirm that the light and heavy tagged glycans are affected identically in RP chromatography with no chromatographic shift due to the incorporation of 13
Figure 2 Example mass spectra for (a) a hexose oligosaccharide in the maltodextrin sample and (b and c) two glycans cleaved from pooled human plasma. The top spectra (blue) were analyzed using HILIC-MS and the bottom spectra (red) were analyzed with RP-MS. The (more ...)
The absence of ammonium acetate in the RP solvent system allows for only the detection of the [M+nH+]n+ peak, whereas ammonium adduction is detected in the HILIC spectra due to the buffer system containing ammonium acetate. This negatively impacts the profiling of N-linked glycans because the ammonium adduction spreads the signal of one glycan composition into multiple m/z channels. Additionally, show that for certain types of glycans in HILIC, the [M+nH+]n+ peak is not detected in the mass spectrum, making detection and identification more difficult.
The observations discussed above have led to several conclusions on the mechanism of N
-linked glycan separation by HILIC and RP chromatography. summarizes the observations made in the separation of N
-linked glycans, both native and derivatized, in HILIC () and RP chromatography (). The mechanisms of both HILIC4, 21, 69-72
and RP chromatography40, 73
are frequently debated, and often, the compromise is that the separation techniques both involve a combination of partitioning and adsorption depending on the molecular properties of the analytes. These properties permit the native glycans to interact significantly with the HILIC stationary phase and allow them to be retained on the HILIC column (). However when derivatized, the glycan molecules become more hydrophobic, and the hydrophobic tag causes the glycan to partition more into the organic mobile phase, which decreases the retention time (as seen in ). In the case of RP chromatography, the native glycan interacts very little with the hydrophobic C18
stationary phase (), but when derivatized (), the glycan molecule becomes hydrophobic enough to interact with the C18
stationary phase, and the derivatized glycans are retained and separated (as seen in ).
Figure 3 The observed retention behavior of native and derivatized N-linked glycans in HILIC and RP chromatography. The magnitude of the arrows indicates the observed relative amount of interaction of the glycan with the stationary phase. The N-linked glycan shown (more ...)
The magnitude of the arrows in are used to portray the distribution coefficient (CS
) of the analyte and it's dependence on the physical properties of the molecule. Molecules that are hydrophobic will tend to interact more with the C18
stationary phase, whereas hydrophilic molecules will interact more with the mobile phase. A molecule that has both hydrophobic and hydrophilic character (such as the derivatized glycans) will have a competing interaction between the mobile phase and stationary phase. implies that there is little to no interaction between the native glycans and the RP stationary phase because no retention is observed. However, because separation of derivatized N
-linked glycans is observed on a RP column, the glycans must have some interaction with the stationary phase in order to have different retention times. If the glycan molecules are assumed to have no interaction with the stationary phase, then all possible retention would be due to the hydrophobic tag interaction, and since all the glycans are derivatized with the identical tag, one would assume that there would be no separation of the glycans. Since this was not observed, the partitioning of derivatized glycans was estimated by calculating the ratio of the hydrophobic surface area (NPSA) and the hydrophilic surface area (MW) (equation development in supplemental Equation S1
). A plot of both retention time and
vs. the number of hexose monosaccharide units is shown in . It is seen that the estimated hydrophobicity:hydrophilicity ratio predicts a similar trend as experimentally observed for the retention of derivatized glycans. This supports a partitioning mechanism for derivatized glycans based on both the hydrophobic and hydrophilic character of the molecule. However, eventually (based on the extrapolation of ), the glycans will be so large, adding an additional hexose monosaccharide (Hexn+1
) will not afford any chromatographic separation from the previous oligosaccharide (Hexn
). Thus, a limitation of the N
-linked glycan RP separation will be at the most hydrophilic (largest) glycan molecules, where separation of the analytes will be limited, and competition in the electrospray droplet could occur. HILIC could be a possible alternative to RP for these large glycans; however, due to the competition in the ESI droplet with underivatized glycans (vide supra)
, HILIC does not perform better than RP for glycans derivatized by hydrazone formation.
Figure 4 The retention times (left y-axis) of the maltodextrin ladder with respect to the size of the maltooligosaccharide. The calculated ‘interaction’ of derivatized glycans with the stationary phase is modeled by the ratio of the NPSA/MW and (more ...)
In order to apply and study the advantages of RP separation over HILIC to complex mixtures, N-linked glycans cleaved from pooled human plasma were used to further compare the two chromatographic techniques. The maltodextrin oligosaccharide ladder is composed only of hexose monosaccharide residues; thus, it was necessary to test and compare more complex glycans (containing fucose and sialic acid residues) and glycan mixtures on the RP system. The base peak HILIC and RP chromatograms for derivatized N-linked glycans are shown in , respectively (the native glycan chromatograms were not shown due to the results presented above – vide supra). These chromatograms are a complex mixture of glycans. However, shows that several glycans in HILIC are dominating the glycan retention window with large abundances and retention times on the order of minutes. This is not seen in the RP chromatogram, implying less spectral competition due to the most abundant analytes. In , it is seen that all glycans that were detected in HILIC (42 unique compositions) were also detected in RP. Moreover, 18 additional N-linked glycans were detected using RP separation rather than HILIC. The majority of these glycans were of the large, highly sialylated, complex glycan type that are typically low abundant and hydrophilic, two properties that make detection by ESI-MS difficult. By increasing the chromatographic resolution, reducing ammonium adduction, and reducing the competition in the ESI droplet, the effective detection limit of the instrument was decreased so that one is able to delve deeper into the N-glycome of human plasma. In a few cases (mostly sialylated glycans), it was observed that the glycan compositional mass was detected at two distinct retention times. This could possibly be due to isomeric separation, which would allow for quantification of glycan isomers. However, this needs to be investigated more thoroughly in order to determine the cause and utility of this peak splitting.
The (a) HILIC and (b) RP chromatography base peak chromatograms for derivatized N-linked glycans. (c) Venn diagram of the total number of unique N-linked glycan compositions cleaved from pooled human plasma and analysed by online HILIC- and RP-MS.
Table S2 in the supplemental material
details the glycan compositions that were detected and their respective molecular weights that were detected using both HILIC and RP chromatographic techniques. It is seen that numerous large complex glycans are only detected with RP separation. The charge state has also been listed in Table S2
for each glycan, and it was observed that glycans separated and analyzed in the RP solvent system are detected in higher charge states. This could be due to the elution of the glycans in different organic:aqueous solvent ratios and the difference in acidic buffer (formic acid or ammonium acetate). In comparison to other studies, the 60 unique N
-linked glycan compositions detected using this method is on par with current literature. Reinhold and coworkers were able to detect 53 unique compositions,74
and this was increased to 106 by using plasma depletion methods. Lebrilla and coworkers report up to 64 unique compositions,67
and Ruhaak and coworkers report 47 unique compositions.75
However, utilizing the hydrazide quantification strategy presented previously,22
each sample consisted of a 1:1 combination of a light and heavy derivatized sample, and only the glycans that were detected with a light and heavy pair in the spectra were determined to be identified. Thus, each of the glycans that was detected is also capable of being relatively quantified using stable-isotope labeled derivatization.