Serum Aberrant N-Glycan Profile as a Marker Associated with Early Antibody-Mediated Rejection in Patients Receiving a Living Donor Kidney Transplant

We determined if the serum N-glycan profile can be used as a diagnostic marker of antibody-mediated rejection (ABMR) in living donor kidney transplant (LKTx) recipients. Glycoblotting, combined with mass spectrometry, was used to retrospectively examine N-glycan levels in the postoperative sera of 197 LKTx recipients of whom 16 recipients had ABMR with or without T-cell-mediated rejection (TCMR), 40 recipients had TCMR, and 141 recipients had no adverse events. Multivariate discriminant analysis for prediction of ABMR was performed by inputting an ABMR event as an explanatory variable and sex, age, and serum N-glycan level as objective variables. The N-glycan score was calculated by multiplying the level of candidate objective variables by objective function values. The ABMR predictive performance of the N-glycan score was assessed by receiver operator characteristic curve and Kaplan–Meier curve analyses. The N-glycan score discriminated ABMR with 81.25% sensitivity, 87.85% specificity, and an area under the curve (AUC) of 0.892 that was far superior to that of preformed donor-specific antibody status (AUC, 0.761). Recipients with N-glycan-positive scores >0.8770 had significantly shorter ABMR survival than that of recipients with N-glycan-negative scores. Although the limitations of our study includ its small sample size and retrospective nature, the serum N-glycan score may contribute to prediction of ABMR.

. Thirty-six types of N-glycans that showed good quantitative reproducibility in all samples and could be analyzed statistically.

The ABMR Predictive Performance of N-Glycan Score Based on the Aberrant Serum N-Glycan Profile Was Far Superior to That of Preformed Donor-Specific Antibody Status
The N-glycan scores one day before LKT (Bfr LKTx), at postoperative day 1 (POD1) and at POD7 were significantly higher in the recipients who developed ABMR (Figure 4a,b,d). The N-glycan score of HLT was not significantly different to the N-glycan score of non-ABMR groups at Bfr LKTX, at POD1 and POD7. Longitudinal follow-up at POD28 showed that there was no significant difference in the N-glycan score between the non-ABMR and ABMR groups. ROC curves were then used to compare the diagnostic performance between preformed DSA status and N-glycan score for ABMR prediction (Figure 4c,e). The area under the curve (AUC) of preformed DSA status and N-glycan score before LKTx and at POD1 for the prediction of ABMR were determined (preformed DSA AUC, 0.7619; N-glycan score before LKTx, 0.7975; N-glycan score at POD1, 0.8916). At the cutoff N-glycan score (0.8770 points) on POD1 for prediction of ABMR, the diagnostic accuracy was 86.29%, the positive predictive value was 81.25%, and the negative predictive value was 86.74%. The positive predictive value was much higher than that of preformed DSA (56.25%) ( Table 4), which suggests that the N-glycan score can detect ABMR recipients with preformed DSA-negative status. The N-glycan scores at POD1 in the recipients who developed ABMR and those who did not; (c) Receiver operating characteristic (ROC) curve analysis of N-glycan score at POD1 and preformed DSA status for detection of ABMR; (d) The level of the N-glycan score before LKTx (Bfr LKTx) in the recipients who developed ABMR and those who did not; (e) ROC curve analysis of the N-glycan score at Bfr LKTx and preformed DSA status for detection of ABMR.   Furthermore, the recipients with an N-glycan score greater than the cutoff value (0.8770) had significantly worse ABMR-free survival (log-rank test, p < 0.0001) (Figure 5a), but there was no significant difference in TCMR-free survival (log-rank test, p = 0.0836) (Figure 5b). Furthermore, the recipients with an N-glycan score greater than the cutoff value (0.8770) had significantly worse ABMR-free survival (log-rank test, p < 0.0001) (Figure 5a), but there was no significant difference in TCMR-free survival (log-rank test, p = 0.0836) (Figure 5b).

N-Glycan-Carrying Serum Immunoglobulin (Igs) Levels Were Not Significantly Different between the ABMR Group and Non-ABMR Group
Serum N-glycomics revealed that the level of sialyl hybrid-type and sialyl bisecting-type Nglycans were significantly lower in the ABMR group. Serum N-glycomics may detect N-glycancarrying glycoproteins in serum, such as Igs (IgGs, IgA, and IgM), which are major N-glycosylated proteins in serum [12]. Therefore, we analyzed serum Igs levels in all samples. Although serum Igs levels of all recipients were much lower than the benign level because of the administration of immunosuppressants, longitudinal follow-up of serum Igs levels before LKTx and on POD1, POD7, and POD28 after LKTx were not significantly different between the ABMR group and non-ABMR groups (Figure 6a-f).
To characterize the N-glycan profile of serum Igs, non-Igs proteins were eliminated by Melon Gel column chromatography (Figure 7a, lanes 5-8) and then compared with whole serum (lanes [1][2][3][4]. N-glycomics of the Igs fraction showed that the levels of N-glycans in the Igs fraction were significantly lower than those in the whole-serum samples ( Figure 7b) and Igs levels in LKTx patients were significantly lower than that of HLT. Although three ABMR-related terminal sialylated Nglycans (hybrid type: m/z 2033; bisecting type: m/z 2728; complex biantennary type: m/z 1709) in the whole serum levels were significantly lower in the recipients who developed ABMR than in those who did not, it is noteworthy that these three N-glycans in the Igs fraction was not significantly changed between non-ABMR and ABMR groups. Furthermore, these three N-glycans in the Igs fraction was significantly lower than in those of whole serum. The level of other ABMR-related sialylated N-glycans (bisecting-type: m/z 1810 and 2728; biantennary-type: m/z 2058) in the Igs fraction was not significantly changed between non-ABMR and ABMR groups. The levels of biantennary Nglycans (m/z 1591, 1607, 1753, and 1915), which were not selected as ABMR-related N-glycans, were not significantly different between the Igs fractions and whole-serum (Figure 7c). These results suggest that ABMR-related N-glycan change did not mainly originate from aberrant N-glycosylation of Igs.

N-Glycan-Carrying Serum Immunoglobulin (Igs) Levels Were Not Significantly Different between the ABMR Group and Non-ABMR Group
Serum N-glycomics revealed that the level of sialyl hybrid-type and sialyl bisecting-type N-glycans were significantly lower in the ABMR group. Serum N-glycomics may detect N-glycan-carrying glycoproteins in serum, such as Igs (IgGs, IgA, and IgM), which are major N-glycosylated proteins in serum [12]. Therefore, we analyzed serum Igs levels in all samples. Although serum Igs levels of all recipients were much lower than the benign level because of the administration of immunosuppressants, longitudinal follow-up of serum Igs levels before LKTx and on POD1, POD7, and POD28 after LKTx were not significantly different between the ABMR group and non-ABMR groups (Figure 6a-f).
To characterize the N-glycan profile of serum Igs, non-Igs proteins were eliminated by Melon Gel column chromatography (Figure 7a, lanes 5-8) and then compared with whole serum (lanes 1-4). N-glycomics of the Igs fraction showed that the levels of N-glycans in the Igs fraction were significantly lower than those in the whole-serum samples ( Figure 7b) and Igs levels in LKTx patients were significantly lower than that of HLT. Although three ABMR-related terminal sialylated N-glycans (hybrid type: m/z 2033; bisecting type: m/z 2728; complex biantennary type: m/z 1709) in the whole serum levels were significantly lower in the recipients who developed ABMR than in those who did not, it is noteworthy that these three N-glycans in the Igs fraction was not significantly changed between non-ABMR and ABMR groups. Furthermore, these three N-glycans in the Igs fraction was significantly lower than in those of whole serum. The level of other ABMR-related sialylated N-glycans (bisecting-type: m/z 1810 and 2728; biantennary-type: m/z 2058) in the Igs fraction was not significantly changed between non-ABMR and ABMR groups. The levels of biantennary N-glycans (m/z 1591, 1607, 1753, and 1915), which were not selected as ABMR-related N-glycans, were not significantly different between the Igs fractions and whole-serum (Figure 7c). These results suggest that ABMR-related N-glycan change did not mainly originate from aberrant N-glycosylation of Igs.

Discussion
N-glycomics is a promising methodology, and several studies have shown that differences in glycan profiles between diseased and benign states may be useful in the diagnosis or prognosis of diseases [8][9][10]13,14]. In the present study, serum N-glycomics was used for recipients who developed ABMR within one month after LKTx. To the best of our knowledge, this is the first report to identify serum-aberrant N-glycosylation profiles as predictive biomarkers for ABMR in LKTx. Our results revealed that the N-glycan scores based on the whole aberrant N-glycosylation profile in serum one day before LKTx, at POD1, and at POD7 were significantly higher in the recipients who developed ABMR (Figure 4a,b,d). Longitudinal follow-up at POD28 showed that the N-glycan scores were not significantly different between the two groups. This finding suggests that the N-glycan score may reflect the early phase of the ABMR reaction in recipients who develop ABMR. Especially, we

Discussion
N-glycomics is a promising methodology, and several studies have shown that differences in glycan profiles between diseased and benign states may be useful in the diagnosis or prognosis of diseases [8][9][10]13,14]. In the present study, serum N-glycomics was used for recipients who developed ABMR within one month after LKTx. To the best of our knowledge, this is the first report to identify serum-aberrant N-glycosylation profiles as predictive biomarkers for ABMR in LKTx. Our results revealed that the N-glycan scores based on the whole aberrant N-glycosylation profile in serum one day before LKTx, at POD1, and at POD7 were significantly higher in the recipients who developed ABMR (Figure 4a,b,d). Longitudinal follow-up at POD28 showed that the N-glycan scores were not significantly different between the two groups. This finding suggests that the N-glycan score may reflect the early phase of the ABMR reaction in recipients who develop ABMR. Especially, we demonstrated that serum sialyl hybrid type and bisecting type N-glycans (m/z 1709, 2033, and 2728) on POD1 in the ABMR group were significantly lower than those in the non-ABMR group (Figure 3). One study showed a higher level of sialylated antibodies on the day of the transplant and at first DSA detection in patients who had good transplant outcomes [7]. Hess et al. reported that T-cell-independent B-cell activation was associated with the production of immunosuppressive sialylated serum IgGs, which inhibit B-cell activation and immune reactions, independent of FcγRIIB [15]. In addition, IgG molecules can perform pro-and anti-inflammatory effector functions depending on the composition of the fragment crystallizable (Fc) domain glycan. Quast et al. reported that IgG Fc sialylation of human monoclonal IgG1 molecules impaired their ability to induce complement-mediated cytotoxicity [16]. They also reported that the presence of sialic acid abrogated the increased binding of C1q to Fc-galactosylated IgG1 and resulted in decreased levels of C3b deposition on the cell surface [16]. Several reports have suggested that Fc-sialylated IgGs affect B-cell activation and complement-mediated cytotoxicity. Several previous reports and the present study results suggest that decreased amounts of sialylated N-glycans on serum glycoproteins may be associated with ABMR in LKTx. It remains unclear why ABMR-associated N-glycan downregulation on serum glycoprotein occurs and what kind of carrier proteins are involved in ABMR-associated changes in the serum N-glycan pattern. Igs are major N-glycosylated proteins in serum [12]. Thus, we hypothesized that serum N-glycan profiles might reflect an N-glycosylation change of Igs in ABMR patients. However, in our study, total Igs levels of LKTx patients were significantly lower than healthy people and ABMR-associated N-glycans (m/z 1566, 1709 and 2033) were not detected in Igs fractions (Figure 6b). In contrast, concentrations of biantennary N-glycans (m/z 1591, 1607, 1752, and 1915) did not differ significantly between the Igs fractions and whole serum and between the ABMR and non-ABMR groups (Figure 6c). These findings suggest that the major carrier protein(s) of ABMR-related sialyl hybrid-type N-glycans do not originate from Igs. This result was not reconciled in the previous study of Malard-Castagnet et al. They focus on higher levels of terminal sialylated IgG in DSA-positive patients which were detectable on the day of the transplant in those who did not develop ABMR, and also did not identify which sialylated Nor O-glycan structure on IgG was associated with ABMR. In this study, we focused our comprehensive N-glycan analysis of whole serum and Igs fractions (mixture of IgG, IgM, and IgA) in both DSA-positive and DSA-negative patients who developed ABMR or not. The patient background and method of our study is completely different from Malard-Castagnet et al. Thus, we hypothesized that both the reduced sialylation of IgG and whole aberrant N-glycan profile changes of serum glycoproteins, except for Igs, may occur in recipients who develop ABMR.
The other possibility is that the serum levels of free N-glycans change in ABMR. Recently, Seino et al. demonstrated that levels of disialylated free N-glycans in serum samples were higher in patients with hepatocellular carcinoma than in healthy controls [17]. Nonetheless, their results showed 100-fold lower amounts of disialylated free N-glycans in serum than those shown by our glycoblotting method, and their free N-glycan analysis did not detect sialyl hybrid-type N-glycans in serum samples. This observation suggests that our high-throughput N-glycomics detects not only free N-glycans in serum, but also N-glycans derived from serum glycoproteins.
Another possible carrier serum protein is α-1-acid glycoprotein (AGP), which is secreted from the liver into plasma [18]. AGP has been studied as an acute-phase serum glycoprotein that possesses five N-linked complex type heteroglycan side chains, which may be present as biantennary, triantennary, or tetra-antennary structures. Additionally, AGP has been studied in association with inflammation, autoimmune diseases, and cancer [19]. Although the origin and clinical implications of serum N-glycans remain unclear, our ongoing studies address these issues and show potential clinical utility. Several previous reports and the present study results suggest that decreased amounts of sialyl hybrid-type N-glycans on serum glycoproteins, except for Igs, may be associated with ABMR in LKTx. These results suggest that the use of N-glycomics may provide insights into new factors predicting ABMR.
We also demonstrated that the positive predictive value of the N-glycan score (81.25%) for detection of ABMR was significantly higher than that of preformed DSA status (56.25%). Although, preformed DSA was a powerful indicator of recipients who did not develop ABMR, some with preformed DSA-negative recipients developed ABMR. In the present study, we demonstrated that the N-glycan score could identify preformed DSA-negative recipients who developed ABMR. Thus, the N-glycan score may be a complement to preformed DSA status.
This was a small study in only 16 ABMR recipients, so the findings should be considered preliminary. To validate the proposed predictive biomarker of ABMR, a study with a greater number of patients is required. Despite this sample limitation, the results suggest that the serum aberrant N-glycan profile can reflect a systemic immunogenic reaction in the early ABMR state. Future studies should determine whether these alterations are a direct result of antibody-mediated allograft injury in LKTx recipients.

Ethics Statement
This study was performed in accordance with the ethical standards of the Declaration of

Serum Samples and Diagnosis of ABMR
A total of 753 recipients underwent LKTx at Akita University Hospital, St. Marianna University of Medicine, Tokyo-Woman's Medical University, Sapporo City General Hospital, or Hirosaki University Hospital between 2007 and 2016. Of those serum available 197 recipients underwent LKTx were retrospectively selected from our serum bank. Healthy controls (HLT, n = 135) selected from community-dwelling volunteers in the health maintenance programme of Iwaki Health Promotion Project. Serum samples were collected one day before LKTx and on POD1, POD7, and POD28 and stored at −80 • C until use. Of those 197, 16 recipients with biopsy-proven ABMR with or without TCMR, 40 recipients with biopsy-proven TCMR, 141 patients without any adverse events, and 135 healthy controls were subjected to serum N-glycomic analysis using the glycoblotting method and matrix-assisted laser desorption ionization-time-of-flight mass spectrometry (MALDI-TOF MS) analysis. Clinical ABMR and TCMR were diagnosed according to Banff classifications by protocol and/or episode biopsy when we observed exacerbation of renal function [20]. Preformed DSA detection was evaluated by using a FlowPRA Single antigen kit (Veritas Corp., Tokyo, Japan) before LKTx. We also examined DSA by using a FlowPRA Single antigen kit when we found exacerbation of renal function and/or pathological abnormality by protocol or episode biopsy after LKTx.
Quantitative reliability was then evaluated on the basis of the following parameters: outliers were allowed <3 points, slope of <3.0, and the significance level of the correlation coefficient r was <0.05. Glycan peaks were assumed to be useful when the above-mentioned criteria of the assay were met, and the resulting glycans were used for statistical analysis.

Statistical Analysis
All calculations for clinical data were performed by using SPSS software, ver. 21.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 6.03 (GraphPad Software, San Diego, CA, USA). Intergroup differences were statistically analyzed by performing Student's t test for normally-distributed variables or by performing the Mann-Whitney U test for non-normally distributed models. Differences with p < 0.05 were considered to be significant. Multivariate discriminant analysis for Figure 8. The general protocol for the integrated glycoblotting technique and workflow for glycoblotting-based high-throughput clinical glycan analysis. (a) Ten-microliter serum samples are applied to SweetBlot TM for glycoblotting; (b) After enzymatic cleavage from serum protein, total serum N-glycans released into the digestion mixture were directly mixed with BlotGlyco H beads to capture N-glycans; (c) After the beads are separated from other molecules by washing, sialic acid is methyl esterified; (d) These processed N-glycans are then labeled with BOA and released from BlotGlyco H beads; (e) Mass spectra of BOA-labeled N-glycans are acquired by using an Ultraflex III instrument.

Statistical Analysis
All calculations for clinical data were performed by using SPSS software, ver. 21.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 6.03 (GraphPad Software, San Diego, CA, USA). Intergroup differences were statistically analyzed by performing Student's t test for normally-distributed variables or by performing the Mann-Whitney U test for non-normally distributed models. Differences with p < 0.05 were considered to be significant. Multivariate discriminant analysis for prediction of ABMR was performed by inputting an ABMR event as an explanatory variable and sex, age, and N-glycans level as objective variables. The ABMR predictive N-glycan score was calculated by multiplying objective variables by objective function values. The ABMR predictive performance of the N-glycan scoring method was evaluated by ROC curve analysis. ROC curves were developed by using the library "rms" in R (http://www.r-project.org/) [23], and statistical differences between AUCs were calculated by using the same program. ABMR-free and TCMR-free survivals were evaluated by using Kaplan-Meier curves, and differences between groups were assessed by performing the log-rank test. Differences with p < 0.05 were considered to be significant.

Conclusions
The N-glycan score, based on the whole aberrant N-glycosylation profile, including downregulation of sialylated N-glycans in serum glycoprotein, was shown to be a promising predictive biomarker of early ABMR in this study. Future large-scale prospective validation studies may definitively determine the clinical utility of these carbohydrate biomarkers for ABMR prediction.