MALDI-TOF-MS and HILIC-UPLC based characterization of rhesus macaque serum and IgG
We characterized the N-glycans isolated from total serum glycoproteins of a pool of seven rhesus macaques and those from purified IgG using a combination of techniques including MALDI-TOF-MS and HILIC-UPLC. First, N-glycans were released from rhesus macaque serum glycoproteins, labeled with 2-AA and analyzed by MALDI-TOF-MS. This allowed the initial identification of 45 different serum N-glycan structures (Supplementary Fig. S1.A online). An additional 38 sialylated N-glycan structures, mostly in the higher mass range, were detected with a similar analysis including a linkage-specific derivatization and stabilization of sialic acid using ethyl esterification[53;54] (Fig. 1.A and Supplementary Fig. S1.B online). In total, 83 different N-glycan structures could be characterized. This number included many isomeric forms differing only by the linkages of the terminal sialic acids, as revealed using the specific derivatization that permits discrimination of α(2–3) and α(2–6) linkages. These 83 structures were represented by 65 different N-glycan compositions. Structural assignment was performed by digesting sequentially and individually the rhesus macaque serum N-glycans using the broad Neuraminidase A, Fucosidase O, β1–4 Galactosidase S and β-N-Acetylglucosaminidase (Supplementary Fig. S2 and S3 online). Using these enzymes sequentially, we reduced the N-glycans to their tri-mannosylated core and demonstrated the presence of a large variety of structures ranging from short, truncated N-glycans to multi-antennary complex N-glycans. Structures with up to four antennae (A1 to A4) were confirmed upon digestion with a mix of Neuraminidase A, Fucosidase O and β1–4 Galactosidase S. These antennae were extended with β-linked galactoses and sialic acids (Supplementary Fig. S2.A online).
In parallel, we used a previously validated HILIC-UPLC profiling workflow to analyze PNGase F-released and procainamide-labeled N-glycans[32] (Fig. 1.B). Forty-three distinct integratable peaks (referred to as “glycopeaks” in the following sections) were identified and correlated with the serum profile (Supplementary Fig. S1.C online). N-glycan structures comprising the glycopeaks were characterized using MS data obtained from the QDa mass detector linked to the UPLC instrument (see Methods) in combination with glycosidase digestions (Supplementary Fig. S2.B online). Data from these orthogonal approaches were highly complementary, with only two minor glycopeaks from the HILIC-UPLC spectra not matching any MALDI-TOF-MS data. Similarly, 11 structures detected by MALDI-TOF-MS (mostly minor ions generated by short, truncated or high-mannose N-glycans) were not observed in HILIC-UPLC profiles.
Peak areas were quantified from HILIC-UPLC spectra and the relative abundance of each glycopeak compared to the total peak area of each spectrum. The most abundant N-glycan observed in rhesus macaque serum was biantennary A2G2SGc2 which accounted for almost 48% of the total N-glycan pool. It was followed by the core-fucosylated version of the same structure (F(6)A2G2SGc2) which accounted for over 10% (Supplementary Tab. S1 online). Due to the instability and ionization bias experienced by sialylated glycan species during MALDI-TOF-MS analysis[53–55], the major ion in the MALDI-TOF-MS spectra of underivatized N-glycans corresponded to A2G2SGc1 (m/z 2067.927 [M – H]−) (Supplementary Fig. S1 online). This was corrected using the linkage-specific derivatization method, where A2G2SGc2 (2432.090 [M – H]−) appeared as the most abundant ion species. Notably, this analysis allowed for identification of N-acetylneuraminic acid (Neu5Ac) in many N-glycans (e.g. m/z 2372.012, 2488.125, 3103.509 [M – H]−, Fig. 1), although this form of sialic acid was less prominent than N-glycolylneuraminic acid (Neu5Gc). Using HILIC-UPLC, we estimated the proportion of N-glycans carrying Neu5Ac to be ~ 6% of the total glycan pool while 81% was sialylated with Neu5Gc. About 4% were complex N-glycans with high m/z (2800–4000 range) and late HILIC-UPLC retention times that possessed both sialic acid forms (e.g. m/z 3103.509, 3279.513 [M – H]−, and glycopeaks 35 at 17,4 min and 38 at 19,1 min, Fig. 1).
In addition to the sequential glycosidase digestion panels (Supplementary Fig. S2 online), complementary digestions with individual enzymes were also performed (Supplementary Fig. S3 online) to confirm serum N-glycan structural assignment. Notably, comparison of digestions with the specific α2–3 Neuraminidase S with broad-specificity α2–3,6,8,9 Neuraminidase A using HILIC-UPLC analysis showed that although the majority of the sialic acids were α2–6 linked, a significant proportion were α2–3 linked, supporting the assignments made by MALDI-TOF-MS analysis with sialic acid linkage-specific derivatization. Thus, by combining the glycosidase digestion results with HILIC-UPLC and MALDI-TOF-MS analysis, a comprehensive structural assignment of rhesus macaque serum N-glycans was obtained (Fig. 1). Detailed structural assignments and MALDI spectra annotated with glycopeak numbering are presented in Supplementary Fig. S1 and Tab. S1 online.
An identical analysis was performed on IgG purified from the rhesus macaque pool of sera (Fig. 2). Using MALDI-TOF-MS, 26 N-glycan structures were identified that were further extended to 43 with the additional step of ethyl esterification of sialic acids (Supplementary Fig. S4 online). In addition, 19 glycopeaks were observed in the HILIC-UPLC profile that correlate with 30 of the N-glycans identified using MALDI-TOF-MS, with the exception of glycopeak #16 (Supplementary Tab. S2 online). This resulted in a total of 13 N-glycan structures being identified by MALDI-TOF-MS analysis that were not detected by HILIC-UPLC profiling (due to their lower abundance).
These N-glycans were further characterized using sequential and individual glycosidase digestion. This analysis revealed that IgG N-glycans were comprised of structures bearing terminal sialic acids, terminal β-linked galactoses, α-linked fucoses, or were mannosylated structures (Supplementary Fig. S5 and S6 online). Finally, MALDI-TOF-MS/MS confirmed the presence of bisecting GlcNAcs on some structures, consistent with prior observations on rhesus macaque IgG N-glycans[28;37]. Fragmentation profiles of selected ions with theoretical m/z 1436.53 and 1639.61 [M – H]− from IgG N-glycans both indicate the presence of bisecting GlcNAc (Supplementary Fig. S6.B online).
N-glycan profiles of serum and IgG from human and rhesus macaques exhibit differences
Exoglycosidase digestions combined with HILIC-UPLC analysis permitted quantification of the following N-glycan categories in rhesus serum: galactosylated (N-glycans bearing at least one terminal unsubstituted galactose), fucosylated (fucose-containing N-glycans), sialylated (N-glycans having at least one antenna terminated by sialic acid) and mannosylated (oligomannosidic or hybrid N-glycans). Analysis of human serum N-glycans was previously conducted using the same workflow[32], allowing a direct comparison. Most rhesus macaque serum N-glycans are sialylated, with over 91% of the UHPLC profile area being sensitive to α2–3,6,8,9 Neuraminidase A treatment (Fig. 3.A). This percentage appeared higher than observed for humans (~ 87%). The majority of the sialic acids in rhesus macaque serum N-glycans are Neu5Gc. This is consistent with the presence of a putative functional CMP-N-acetylneuraminic acid hydroxylase (CMAH) gene – the only enzyme capable of synthesizing Neu5Gc by hydroxylation of Neu5Ac[56–58] – in rhesus macaques[59]. This gene is deactivated in a number of lineages including that leading to humans[60]. Thus, while the major structure in human is A2G2S2, the major structure in rhesus macaques is A2G2SGc2. Using Fucosidase O digestion, we determined that around 30% of the total N-glycan serum pool is fucosylated in rhesus macaques while it was found to be slightly less than 24% in humans. The abundance of FA2G2SGc2 in rhesus macaques largely contributes to this overall higher fucosylation, since the ratio between the A2G2SGc2 and its fucosylated variant (FA2G2SGc2) in rhesus macaque is considerably lower than between A2G2S2 and FA2G2S2 in human. We also noted a smaller amount of agalactosylated N-glycans in rhesus macaque serum compared to human serum, which was particularly evident for the F(6)A2-containing glycopeak, while the proportion of galactosylated N-glycans was similar in both species at 20% of the total N-glycan pool. Digestions with β1–4 Galactosidase showed that all galactoses were β1–4 linked and this was confirmed by the absence of sensitivity to the broad-specificity α1–3,4,6 Galactosidase (Supplementary Fig. S3 online), consistent with the absence of α1,3 galactosylation in the common ancestor of old world primates due to the loss of α1,3 galactosyltransferase[61]. Finally, a small portion (around 3%) of mannosidic and hybrid N-glycans, which have also been found in humans, was revealed using Endoglycosidase H (Supplementary Fig. S3 online).
In accordance with previous studies[28;37], we also found the IgG N-glycans of rhesus macaque to be substantially different from human IgG (Fig. 3.B). Although IgG profiles from both species share many, largely core-fucosylated, N-glycan structures, the relative proportion of these N-glycans differed considerably. Indeed, the rhesus macaque IgG profile is dominated by the N-glycan structure F(6)A2G2SGc1, while the agalactosylated (F(6)A2) and galactosylated (F(6)A1-2G1-2) structures, which constitute the majority of human IgG N-glycosylation, appear in markedly lower amounts on rhesus macaque IgG. Thus, rhesus macaque IgG exhibits a significantly higher abundance of larger sialylated N-glycans compared to human, which differentially impacts the contribution of IgG N-glycans to the total serum N-glycoprofile. Indeed, in human, IgG N-glycans are nearly all smaller than A2G2S2 and, thus, do not contribute to the most abundant complex N-glycan structures[1], while in rhesus macaque, IgG N-glycans contribute to FA2G2SGc2 and FA2(B)G2SGc2 (glycopeaks #25 and 26), two major sialylated N-glycans of the HILIC-UPLC serum profile. As for serum N-glycans, the vast majority of the IgG sialic acids were Neu5Gc. However, we detected a minor amount of the Neu5Ac form of sialic acid in several glycopeaks (Fig. 2 and Supplementary Tab. S2 online) which has not been reported previously for rhesus macaques[28;37].
Finally, we also found bisecting GlcNAc which is a common feature in both species, but has been determined to be more prominent in rhesus macaques compared to humans[37]. Although we did not perform any quantitative analysis of bisecting GlcNAc-containing N-glycans, we detected them in a significant amount on rhesus macaque IgG (Supplementary Fig. S5 online). Presence of bisecting GlcNAc explains the incomplete digestion of the structure FA2B (m/z 1639.506 [M – H]−) using N-acetylglucosaminidase (residual peak m/z 1395.347 [M – H]−).
Changes in serum N-glycosylation during the course of B. malayi infection
To study if infection with B. malayi affects serum N-glycosylation, we used the HILIC-UPLC workflow to monitor the profiles from a longitudinal set of rhesus serum samples over the course of their infection with this filarial nematode. Firstly, serum samples from 7 different healthy animals were analyzed and yielded highly similar HILIC-UPLC profiles resulting in a stable baseline for the serum N-glycan profile of healthy rhesus macaques (Supplementary Tab. S4.A online). Secondly, serum was additionally sampled from four rhesus macaques at 5, 12 and 15 weeks post-infection (wpi) with B. malayi infective larvae (Methods section and Supplementary Tab. S3 online). Using statistical analysis based on linear-mixed effect models[62], we highlighted an alteration of the N-glycan serum profile during the course of B. malayi infection. Changes arose as early as 5 wpi when 17 glycopeaks showed significant differences in their relative peak area (in percentages) between infected and baseline time-points (adjusted p-value < 0,05). Alteration of serum N-glycosylation increased as the infection progressed, with 24 glycopeaks showing statistically significant changes in abundance at 12 wpi and 27 glycopeaks at 15 wpi. Many altered glycopeaks were abundant ones in the HILIC-UPLC profile with, for instance, the major glycopeak (#23 in the HILIC-UHPLC profile) that significantly decreased at 15 wpi or glycopeak #25 that showed a significant reduction as early as 5 wpi and continued until 15 wpi when compared to baseline. Interestingly, many of the early changes - either a decrease or increase in relative abundance - lasted up to 15 wpi as this was observed for 12 out of the 17 glycopeaks (Fig. 4 and Supplementary Tab. S4 online). In addition, some noticeable changes became evident at 15 wpi, with a clear increase for many shorter glycan structures that appear early in the HILIC-UPLC profile (e.g. glycopeak #1, 2, 3, 5, 6, 7, 9 11, 12, 15, 16, 20) and a significant decrease for the most abundant serum N-glycan (glycopeak #23). Overall, the directionality of changes appeared to be consistent over the course of the infection, since only glycopeak #2 fluctuated (decrease at 12 wpi and increase at 15 wpi), while the relative abundance of all other glycopeaks that showed changes were consistently increased or decreased throughout the longitudinal study.
We also analyzed N-glycosylation of purified IgG from serum of all 4 infected monkeys at each time-point. In contrast, total IgG glycosylation in infected rhesus macaques did not show any significant changes in N-glycosylation over the course of infection (data not shown).
We next aimed to investigate trends within the main N-glycan classes present in rhesus macaque serum during infection. Structural assignment performed on the basis of exoglycosidase digestions allowed us to estimate the relative abundance of sialylated, fucosylated, galactosylated and mannosylated (oligomannosidic or hybrid) N-glycans at each time-point (Fig. 5 and Supplementary Tab. S5 online). Unlike changes in individual N-glycan structures, the first significant changes in these broad N-glycan classes appear only at 12 wpi with an increase in both galactosylation and mannosylation. However, at 15 wpi this increase in galactosylated and mannosylated N-glycans appears to become more substantial (with p-values < 0.001) and is accompanied by a decrease in sialylated structures. This finding is in line with changes in individual glycans described above since structures eluting at the earliest retention times are mainly galactosylated and mannosylated. Thus, the increased abundance of these structures at 15 wpi and the decreased abundance of the major sialylated N-glycan likely explain the observed trends for the N-glycan classes.
Given that the main goal of performing such glycoprofiling studies is to identify potential glycan biomarkers of infection or disease, we decided to focus on changes in the most abundant N-glycan structures which constitute more promising targets for diagnostic application. Since we noticed a slightly higher technical variation for the smallest peaks in the profile and for those with the later retention times, we selected the 20 glycopeaks with areas larger than 0.5% of the total profile area and with a retention time less than 18 minutes in the HILIC-UPLC profile. The volcano plots summarizing the results of the statistical analysis performed on these peaks are shown in Fig. 6. As previously seen when studying changes for the overall HILIC-UPLC profile, the number of glycopeaks showing statistical differences compared to healthy serum increased during infection. Thus, in this glycopeak selection, 7 peaks are significantly altered at 5 wpi, 10 at 12 wpi and 17 peaks at 15 wpi. Moreover, this was accompanied by a noticeable increase in the negative log2 of the adjusted p-values. Interestingly, 6 out of the 7 glycopeaks showing differences as early as 5 wpi were either decreased (#13, 25, 34, 35) or increased (#3, 22) throughout the 3 time-points, supporting consideration of these glycan structures as potential biomarker candidates.
Finally, taking advantage of the use of the exact same workflow for both studies, we compared the changes observed in rhesus macaque N-glycans during infection with B. malayi with the changes in canine serum N-glycans during infection with the dog heartworm Dirofilaria immitis (D. immitis)[51]. We compared the alteration of serum N-glycosylation observed at 27 wpi for the dogs and at 15 wpi for the rhesus macaques since these time-points coincide with the microfilariaemic stage of infection for the animals in both studies. Both filarial infections impact the serum N-glycosylation and we focused on N-glycan structures that were identified in both dog and rhesus macaque serum (Supplementary Tab. S6 online). Some similarities could be observed, for instance the abundance of F(6)A2, F(6)A2G1, F(6)A2G2 and F(6)A2G2SGc1 is increased in both infections, while abundance of F(6)A2G2SGc2 decreased in both cases. However, several glycan structures were differentially impacted in both studies. As an example, abundance of A2G2SGc2 decreased in rhesus macaque serum upon microfilariaemia while it increased in dog serum. Thus, we could not highlight a very clear common trend for both filarial infections and our data tends more toward a parasite- and species-specific alteration of the serum N-glycome.