Embed Size (px)
Citation preview
Identification of Novel UrinaryBiomarkers for Predicting RenalPrognosis in Patients With Type 2Diabetes by Glycan Profilingin a Multicenter ProspectiveCohort Study: U-CARE Study 1Diabetes Care 2018;41:1765–1775 | https://doi.org/10.2337/dc18-0030
OBJECTIVE
Because quantifying glycans with complex structures is technically challenging,little is known about the association of glycosylation profiles with the renalprognosis in diabetic kidney disease (DKD).
RESEARCH DESIGN AND METHODS
In 675 patients with type 2 diabetes, we assessed the baseline urinary glycan signalsbinding to 45 lectins with different specificities. The end point was a decrease ofestimated glomerular filtration rate (eGFR) by ‡30% from baseline or dialysis forend-stage renal disease.
RESULTS
During a median follow-up of 4.0 years, 63 patients reached the end point. Coxproportional hazards analysis revealed that urinary levels of glycans binding to sixlectins were significantly associated with the outcome after adjustment for knownindicators of DKD, although these urinary glycans, except that for DBA, were highlycorrelated with baseline albuminuria and eGFR. Hazard ratios for these lectinswere (+1 SD for the glycan index) as follows: SNA (recognizing glycan Siaa2-6Gal/GalNAc), 1.42 (95% CI 1.14–1.76); RCA120 (Galb4GlcNAc), 1.28 (1.01–1.64); DBA(GalNAca3GalNAc), 0.80 (0.64–0.997); ABA (Galb3GalNAc), 1.29 (1.02–1.64); Jacalin(Galb3GalNAc), 1.30 (1.02–1.67); and ACA (Galb3GalNAc), 1.32 (1.04–1.67). Addingthese glycan indexes to a model containing known indicators of progression improvedprediction of the outcome (net reclassification improvement increased by 0.51 [0.22–0.80], relative integrated discrimination improvement increased by 0.18 [0.01–0.35],and the Akaike information criterion decreased from 296 to 287).
CONCLUSIONS
The urinary glycan profile identified in this study may be useful for predictingrenal prognosis in patients with type 2 diabetes. Additional investigation of gly-cosylation changes and urinary glycan excretion in DKD is needed.
1Department of Nephrology, Rheumatology, En-docrinology and Metabolism, Okayama Univer-sity Graduate School of Medicine, Dentistry andPharmaceutical Sciences, Okayama, Japan2Diabetes Center, Okayama University Hospital,Okayama, Japan3Department of Chronic Kidney Disease andCardiovascular Disease, Okayama UniversityGraduate School of Medicine, Dentistry andPharmaceutical Sciences, Okayama, Japan4Center for Innovative Clinical Medicine,Okayama University Hospital, Okayama, Japan5GlycoTechnica Ltd., Yokohama, Japan6Division of Clinical Research of New Drugs andTherapeutics, Center for Innovative Clinical Med-icine, Okayama University Hospital, Okayama,Japan
Corresponding author: Koki Mise, [email protected], or Jun Wada, [email protected].
Received 6 January 2018 and accepted 14 May2018.
Clinical trial reg. no. UMIN000011525, www.umin.ac.jp/ctr.
This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc18-0030/-/DC1.
© 2018 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.
Koki Mise,1 Mariko Imamura,1
Satoshi Yamaguchi,1 Sanae Teshigawara,1
Atsuhito Tone,2 Haruhito A. Uchida,3
Jun Eguchi,1 Atsuko Nakatsuka,1
Daisuke Ogawa,1 Michihiro Yoshida,4
Masao Yamada,5 Kenichi Shikata,6 and
Jun Wada1
Diabetes Care Volume 41, August 2018 1765
PATH
OPHYSIO
LOGY/CO
MPLIC
ATIO
NS
Several biomarkers have been demon-strated to predict renal prognosis in pa-tients with diabetes. Blood levels of tumornecrosis factor receptors 1 and 2 are well-known prognostic indicators for diabetickidney disease (DKD) (1,2). Similarly, markersof tubulointerstitial injury, inflammation,and filtration have been reported to pre-dict the progression of DKD (3–6). Thesebiomarkers partly allow us to predict renalprognosis at an early stage of DKD inde-pendently of established clinical factors,such as the estimated glomerular filtrationrate (eGFR) and albuminuria. However,despite early detection of high-risk patientsand recent new protective treatments forDKD (7–9), the management of patientswith DKD and rapid deterioration of renalfunction remains difficult. Thus, new bio-markers are needed to identify the path-ogenesis of DKD and new therapeutictargets.The role of glycans and their enzymatic
modification (glycosylation) lately haveattracted much attention in relation to re-search on cancer and metabolic diseases,including diabetes and DKD. Ohtsubo et al.(10) reported that human pancreatic b-cell–specific GnT-4a glycosyltransferase,which generates the core b1–4 GlcNAclinkage in N-glycans, has a protective effectagainst diabetes by increasing GLUT2 ex-pression and maintaining insulin secretion.Differences of glycosylation, includingO-GlcNAcylation, were reported to con-tribute to the progression of DKD in rats(11,12). There have been few reportsabout differences of glycosylation in pa-tients with DKD because of technicalobstacles to glycan analysis as a resultof the complicated structures of thesemolecules and the time-consuming pro-cesses required for mass spectrometry(MS). However, the evanescent-fieldfluorescence-assisted lectin microarraythat we reported previously enables high-throughput quantification of glycan bind-ing to 45 specific lectins (13,14). In thepreliminary analysis of our previous study(14), the urinary glycosylation patternvaried considerably, whereas the changesin serum glycosylation were barely detect-able among patients with various kidneydiseases, including DKD, associated withalmost equivalent proteinuria and eGFR.Moreover, studies of the serum N-glycanprofile have revealed that differences ofN-glycosylation are associated with dia-betic complications and glycemic control inpatients with type 1 and 2 diabetes (15,16),
but the association between the urinaryglycosylation profile and renal prognosishas not been investigated in patients withdiabetes to the best of our knowledge.
Accordingly, we investigated the re-lationship between urinary excretion ofO- and N-glycans binding to 45 lectinsand the renal prognosis in patients withtype 2 diabetes. In addition, we assessedthe incremental predictive value of add-ing promising glycans to a model thatcontained established clinical variables,including albuminuria and eGFR.
RESEARCH DESIGN AND METHODS
Study DesignThis prospective cohort study was initi-ated in 2012. Among 688 patients withtype 2 diabetes admitted to eight hospitalsin Japan from June 2012 to March 2013,675 patients were eligible for enrollment.Exclusion criteria were a diagnosis of slowlyprogressive type 1 diabetes during follow-upor abaselineeGFR,15mL/min/1.73m2
(Supplementary Fig. 1). The diagnosisof diabetes was based on Japanese Di-abetes Society criteria (17). In addition,134 volunteers who underwent medicalcheckups at Okayama Health Foundationin March 2016 and were confirmed to haveneither diabetes nor chronic kidney dis-ease (CKD) (17,18) were enrolled as con-trol subjects to compare differences inglycosylation. The protocol for this studywas approved by the ethics committee ofOkayama University Hospital in June 2012(identification number: H24-003). This studywas registered with the University Hospi-tal Medical Information Network in June2012. Written informed consent wasobtained from all patients with diabetes,whereas comprehensive anonymous con-sent was obtained from control subjects.
Laboratory Parameters and DefinitionsUrine samples collected in the earlymorning and stored at baseline (patientsin 2012–2013, control subjects in 2016)were used to measure urinary glycansand urinary albumin levels in 2015–2016.All specimens were aliquoted, stored at280°C until measurement, and thawed forthe first time to perform this study. The av-erage storage period until measurementwas 2.16 0.1 years for patients and 0.960.0 years for control subjects. As shown inSupplementary Table 1, the influence ofthe difference in frozen storage duration onmeasurement of glycosylation is negligible.
Urinary glycans were measured by theevanescent-field fluorescence-assistedlectin microarray, which is a new methodof glycan profiling (13,19–21). In brief, wemeasured urinary levels of Cy3-labeledglycoproteins that bound to 45 lectinswith different specificities, and urinaryglycan intensity can be measured in 300samples in 3 working days (SupplementaryFig. 2). Cy3 binds to primary amine inprinciple. Urinary albumin was labeled byCy3, although it lacks glycan modification,resulting in minimal background reactivity.Urinary creatinine also was labeled by Cy3,causing background reactivity like albu-min. In preliminary experiments, bothurinary albumin and creatinine weresignificantly associated with the inten-sity of background reactivity (r betweenlog[urinary albumin concentration] andlog[background intensity (BG-I)] = 0.697,r between log[urinary creatinine con-centration (UCr)] and log[BG-I] = 0.166)(Supplementary Fig. 3). We also ob-served that the intensity of glycan re-activity (glycan intensity) in 24-h urineand spot urine samples was more closelycorrelated with the net glycan intensity(Net-I = raw glycan intensity [Raw-I] – BG-I)than with the Net-I/UCr ratio or the Raw-I/UCr ratio (spot urine/24-h urine Net-I ratio1.206 0.74 [mean6 SD], Net-I/UCr ratio1.776 2.69, Raw-I/UCr ratio 1.686 2.00)(Supplementary Fig. 4). Furthermore, com-parison of glycan intensity between orig-inalurinesamplesand10-folddilutedurinesamples showed that all the Raw-I ratios(ratio = original urine intensity/10-fold di-luted urine intensity) and most of the Net-Iratios were ,10, suggesting that the uri-nary glycan index did not increase linearly,but quadratically (Supplementary Table 2).On the basis of these results, we per-formed all analyses by using the glycanindex appropriately transformed fromthe Net-I according to its distribution.
GFR was estimated by using the Japa-nese coefficient–modified Chronic KidneyDisease Epidemiology Collaboration equa-tion (22). The baseline urinary albumin/creatinine ratio (UACR) in milligrams pergram creatinine was measured in a spoturine specimen, and normoalbuminuria,microalbuminuria, and macroalbuminuriawere defined as UACR ,30, $30 and,300, and$300 mg/gCr, respectively (15).HbA1c data are presented as National Gly-cohemoglobin Standardization Programvalues according to the recommendationsof the Japanese Diabetes Society and the
1766 Novel Glycan Biomarkers of DKD Diabetes Care Volume 41, August 2018
International Federation of Clinical Chem-istry (23). BMI was calculated as weight inkilograms divided by the square of heightin meters. Mean arterial pressure (MAP)was calculated as two-thirds of diastolicpressure plus one-third of systolic pressure(mmHg). Hypertension was defined as abaseline blood pressure (BP) $140/90mmHg or use of antihypertensive drugs.The average annual values of clinical pa-rameters, including systolic BP (SBP), di-astolic BP (DBP), MAP, and HbA1c plusthe use of an ACE inhibitor (ACE-I) or an-giotensin receptor blocker (ARB) duringfollow-up were compared between all pa-tients with and without outcome and thosestratified according to baseline eGFR cate-gories. The grade of diabetic retinopathywas determined by an ophthalmologist atbaseline (24). In this study, cardiovasculardisease (CVD), stroke, and peripheral arte-rial disease (PAD) were defined as eventsrequiring admission for treatment, cerebralbleeding or infarction requiring admissionfor treatment, and PAD requiring admis-sion for intervention or surgery, respec-tively. Cardiovascular events were definedas any CVD, stroke, or PAD event.
Study End PointThe primary study end point was definedas a decrease of eGFR by at least 30%from baseline or commencement of di-alysis for end-stage renal disease (ESRD).None of the patients received a kidneytransplant during follow-up.
Statistical AnalysisData are summarized as percentages oras the mean 6 SD, as appropriate. Allskewed variables were subjected to log-transformation to improve normality be-fore analysis. Correlations among glycanindexes were evaluated by Pearson cor-relation analysis. Univariate and multi-variate linear regression analyses wereused to explore the association of urinaryglycan index with baseline HbA1c and age.In both regression models, the depen-dent variable was glycan index. In eachmultivariate model, the b-coefficient forHbA1c was adjusted for baseline age, sex,and duration of diabetes, whereas theb-coefficient for age was adjusted for sexand duration of diabetes. The cumulativeincidence rate of the primary outcomewas estimated by Kaplan-Meier methodfor urinary glycan quartiles in all patients,and incidence rates were comparedwith the log-rank test. In addition, we
compared Kaplan-Meier curves betweengroups of patients with normoalbuminuriaor microalbuminuria stratified according tourinary glycan quartiles (Q1–3 vs. Q4 or Q1vs. Q2–4) and compared groups of patientswith macroalbuminuria and higher orlower glycan indexes than the medianvalue. TheCoxproportional hazardsmodelwas used to calculate the hazard ratio (HR)and 95% CI for the death-censored endpoint. In themultivariatemodel, HRs wereadjusted for age, sex, MAP, HbA1c, eGFR,and log-transformed UACR at baseline.These covariates were selected as potentialconfounders on the basis of biological plau-sibility and metabolic memory (25,26). Wealso examined another multivariate modelwithout baseline HbA1c to evaluate theeffect of baseline HbA1c levels on the out-come. Improvement in discriminating therisk of the study outcome at the medianfollow-up time (4.0 years) was assessed byanalysis of category-free net reclassifica-tion improvement (NRI) and absolute/relative integrated discrimination improve-ment (IDI) as reported elsewhere (27–29).The median follow-up time was selectedas the cutoff point for analysis because itwas previously used in similar biomarkerstudies (3,30). We also used the Akaikeinformation criterion (AIC) to compare thefit of various models. Moreover, we com-pared Harrell’s concordance index (C-index)between multivariate Cox proportionalhazards models with or without glycanbiomarkers. The 95% CIs for category-free NRI and IDI and the differences ofthe C-index were computed from 5,000bootstrap samples to adjust for optimismbias. Two-tailed P , 0.05 was consideredto indicate statistical significance. Analysesand creation of graphs were performedwith Stata/SE version 14.0 (StataCorp) andOrigin version 2017 (OriginLab) software.
RESULTS
Follow-up Period and Incidenceof the OutcomeThe median follow-up period was 4.0years (interquartile range [IQR] 3.9–4.0years). During follow-up, theprimary endpoint occurred in 63 (9%) patients, and12 (2%) patients died as a result of causesother than ESRD after refusing dialysis ortransplantation.
Patient CharacteristicsThe baseline characteristics of the subjectsare shown in Table 1. Theirmean6 SDage
was 636 11 years, 61%weremen, and themedian known duration of diabetes was11.0 years (IQR 6.2–17.7 years). The base-line SBP, DBP, andMAPwas 131.06 17.1,74.8 6 10.9, and 93.5 6 11.7 mmHg,respectively. One-third of the patientshad diabetic retinopathy (any type), andtheir mean baseline HbA1c was 7.16 1.1%(54.3 6 12.0 mmol/mol). In addition, themean baseline eGFR was 71.46 17.1 mL/min/1.73 m2, and median UACR was 17.3mg/gCr (IQR 7.8–71.1 mg/gCr); 594 pa-tients had normoalbuminuria (64%) or mi-croalbuminuria (24%). With respect to BPand glycemic control during follow-up, theaverage SBP, DBP, MAP, and HbA1c werenot significantly different between patientswith and without outcome. As well, theuseof anACE-I orARBduring follow-upwasnot significantly different between the pa-tients with and without outcome in groupsstratified according to baseline eGFR cate-gories (Supplementary Table 3).
Associations of Urinary Glycan Indexwith HbA1c and AgeUnivariate and multivariate regressionanalysis in all patients and patients withbaseline eGFR $60 mL/min/1.73 m2 re-vealed that lower HbA1c was significantlyassociated with glycan complex struc-tures that have higher GlcNAc [recog-nized by PHA(L), PHA(E), DSA, LEL, STL,and WGA], higher Siaa3Galb4GlcNAc(recognized by MAL_I and ACG), higherGalNAcb4GlcNAc(recognized by WFA),and higher GalNAca3GalNAc (recog-nized by DBA), whereas higher HbA1cwas significantly associated with higherMana3Man (recognized by GNA). Mostof these significant associations were atten-uated in the patients with baseline eGFR,60 mL/min/1.73 m2 (SupplementaryTable 4). Similar regression analysis forage showed that older age was associatedwith higher Fuca2GalbGlcNAc (recognizedby UEA_I), higher (Galb4GlcNAc)n (poly-lactosamine, recognized by LEL), higher(GlcNAcb4MurNAc)n (peptidoglycan back-bone, recognized by STL), and higherSiaa3Galb3(Siaa6)GalNAc (recognizedby MAH) (Supplementary Table 5).
Relation Between the Renal Outcomeand Glycan Binding to the Lectin PanelUnadjusted and adjusted HRs for glycanbinding to the panel of 45 lectins withdifferent specificities and the reportedstructure of the glycan binding to eachlectin are shown in Fig. 1.Anumberofurinary
care.diabetesjournals.org Mise and Associates 1767
glycans were significantly correlated withthe renal outcome in the univariate Cox re-gressionmodel, whereas the urinary glycansbinding to six specific lectins (Sambucusnigra [SNA], Ricinus communis [RCA120],Dolichos biflorus [DBA], Agaricus bisporus[ABA], Artocarpus integrifolia [Jacalin], andAmaranthus caudatus [ACA]) were signifi-cantly correlatedwith renal outcome in boththe univariate and multivariate models. Inthe multivariate model adjusted for knownindicators of DKD progression, includingbaseline eGFR and UACR, the HR for positiveglycan binding to SNA (+1 SD for the glycanindex) was 1.42 (95% CI 1.14–1.76), whereasthe HR for glycan binding to RCA120 was 1.28(1.01–1.64); DBA, 0.80 !(0.64–0.997); ABA,1.29 (1.02–1.64); Jacalin, 1.30 (1.02–1.67);and ACA, 1.32 (1.04–1.67) (Fig. 1A). Theseassociations remained largelyunchangedwhen average MAP and/or HbA1c duringfollow-up were incorporated into the mul-tivariate model and when baseline HbA1cwaseliminatedfromthemultivariatemodel(Supplementary Tables 6 and 7). As shownin Fig. 1B, the glycans Siaa2-6Gal/GalNAc,Galb4GlcNAc, and GalNAca3GalNAc werereported to bind with SNA, RCA120, andDBA, respectively, whereas Galb3GalNAc wasreported to bind with ABA, Jacalin, and ACA.
Glycan Binding to SNA, RCA120, DBA,ABA, Jacalin, and ACA in the ControlGroup and Patients With DiabetesStratified According to the CKDHeat MapWe stratified the patients with diabe-tes into four CKD heat map groups and11 categories by using the baselineUACR and eGFR (31). Comparisons ofglycan binding to SNA, RCA120, DBA,ABA, Jacalin, and ACA among the controlgroup and patients with diabetes strat-ified according to the CKD heat mapgroups and 11 categories are shown inSupplementary Fig. 5. Overall, the glycanindex values were higher in the severecategories of the CKD heat map, albu-minuria, and eGFR, except for the glycanindex of DBA. Of note, the glycan indexof SNA was significantly higher in thegreen heat map group than in the con-trol group, even though both groupswere defined by normoalbuminuria andeGFR.60mL/min/1.73m2. Correlationsamong positive binding to ABA, Jacalin,and ACA were extremely strong (r =0.952 between ABA and Jacalin, r = 0.931between ABA and ACA, and r = 0.942between ACA and Jacalin) as was the
Table 1—Baseline clinical parameters
Clinical parameter All patients (n = 675)
Male sex 61
Age (years) 63 6 11
BMI (kg/m2) 25.6 6 4.6
Duration of diabetes (years) 11.0 (6.2–17.7)
SBP (mmHg) 131.0 6 17.1
DBP (mmHg) 74.8 6 10.9
MAP (mmHg) 93.5 6 11.7
Hypertension* 70
Diabetic retinopathyNondiabetic 37Simple 17Preproliferative 6Proliferative 10
Serum creatinine (mg/dL) 0.85 6 0.35
eGFR (mL/min/1.73 m2) 71.4 6 17.1
CKD GFR category (grade)†1 102 703a 123b 64 3
UACR (mg/gCr) 17.3 (7.8–71.1)
Normoalbuminuria 64
Microalbuminuria 24
Macroalbuminuria 12
HbA1c (%) 7.1 6 1.1
HbA1c (mmol/mol) 54.3 6 12.0
Triglycerides (mg/dL) 116 (81–162)
Total cholesterol (mg/dL) 180.5 6 32.1
LDL cholesterol (mg/dL) 100.0 6 25.4
Uric acid (mg/dL) 5.3 6 1.4
Any type of antihypertensive agents 62
ACE-I or ARB 53
Calcium channel blocker 38
Number of antihypertensive agents 1 (0–2)
Treatment for diabetesDiet regimen only 4Oral hypoglycemic agent 64Insulin 32
Drug treatment for hyperglycemiaSulfonylurea 32Meglitinide analogs 10Biguanide (metformin) 35a-Glucosidase inhibitor 28Thiazolidinedione 15Dipeptidyl peptidase 4 inhibitor 49Glucagon-like peptide 1 7
Drug treatment for dyslipidemia 65
Drug treatment for hyperuricemia 10
Prior CVD 17
Prior stroke 10
Prior PAD 2
Prior cardiovascular event 28
Data are mean6 SD, %, or median (IQR) unless otherwise indicated. *Hypertension was defined asBP $140/90 mmHg or any antihypertensive drug treatment. †Grade 1, $90 mL/min/1.73 m2;grade 2, 60–90 mL/min/1.73 m2; grade 3a, 45–59 mL/min/1.73 m2; grade 3b, 30–44 mL/min/1.73 m2; and grade 4, 15–29 mL/min/1.73 m2.
1768 Novel Glycan Biomarkers of DKD Diabetes Care Volume 41, August 2018
correlation between SNA and RCA120(r = 0.921) (Supplementary Table 8).
Cumulative Incidence Rate of thePrimary Outcome in Urinary GlycanQuartilesKaplan-Meier curves stratified accordingto quartiles for baseline urinary glycanbinding to SNA, RCA120, DBA, ABA, Jacalin,and ACA are shown in Fig. 2. The cumu-lative incidence rate of the renal outcomewas significantly higher in the highest quar-tile for urinary glycan binding to SNA,RCA120, ABA, Jacalin, or ACA than in theother quartiles, whereas it was significantlyhigher in the lowest quartile for glycanbinding to DBA than in the other quartiles(Fig. 2A). Similar results for glycans bindingto SNA, ABA, Jacalin, and ACA were ob-tained in patients with normoalbuminuria/microalbuminuria (Fig. 2B). Among patientswith macroalbuminuria, the cumulative
incidence rate was significantly higherin those with higher glycan index valuesthan in those with lower glycan indexvalues, except in the case of the glycanindex for DBA (Fig. 2C).
Incremental Predictive Power ofUrinary Glycan Binding to SNA,RCA120, DBA, ABA, Jacalin, and ACAThe category-free NRI, absolute/relativeIDI, and AIC for predicting the primaryoutcome at the median follow-up time(4.0 years) obtained by adding the glycanindexes as well as the difference of the C-index between Cox regression modelswith or without the urinary glycan in-dexes are summarized in Table 2. Theaddition of any single glycan index to themultivariate model did not improve pre-diction, whereas adding all six glycanindexes significantly improved risk clas-sification, integrated discrimination, and
AIC (category-free NRI 0.51 [95% CI 0.22–0.80], relative IDI 0.18 [0.01–0.35], AICdecrease from 296 to 287). Similarly,when four glycan indexes (on the basisof the reported glycan specificities) wereadded, risk classification, integrated dis-crimination, andAIC alsowere significantlyimproved (Table 2). However, the C-indexdid not increase significantly when theseglycan indexes were added to the multi-variate Cox regression model (Table 2).
CONCLUSIONS
Glycans are involved in various biologicalprocesses, including development, immu-nity, infection, hormone actions, cell ad-hesion, and oncogenesis (19,32,33). One ofthe most prominent features of glycansis structural heterogeneity, and this distin-guishes glycans from other major biopoly-mers, such as nucleic acids and proteins.Such heterogeneity is largely attributable to
Figure 1—Univariate and multivariate Cox proportional hazard models of the renal outcome and reported glycans binding to 45 lectins. A:Cox proportional hazard models. In the multivariate model, HR was adjusted for age, sex, MAP, HbA1c, eGFR, and log-transformed urinary albuminexcretion at baseline. Renal outcome was defined as 30% eGFR decline or dialysis as a result of ESRD. B: Preferred glycan structures binding to45 lectins with different specificity.
care.diabetesjournals.org Mise and Associates 1769
theglycan fabricationprocessand dependson structural modification by glycosyltrans-ferases,which isknownasglycosylation(19).Podocalyxin is the molecule in the
kidneys for which the functional effects
of changes in glycosylation have beenmost thoroughly investigated (34). It isa glycoprotein expressed by podocytesthat plays an essential role in the main-tenance of podocyte slit pore integrity
and effective glomerular filtration (35).Podocalyxin is decorated by O-glycans,including an abundance of sialic acid, andremoval of sialic acid leads to efface-ment of the podocyte foot processes and
Figure 1dContinued.
1770 Novel Glycan Biomarkers of DKD Diabetes Care Volume 41, August 2018
proteinuria in mice secondary to loss ofthe negative charge on the glomerularbasement membrane (36). Similarly,doxycycline-induced global deficiencyof glycosyltransferase core 1 synthaseand glycoprotein-N-acetylgalactosamine3-b-galactosyltransferase 1 (C1galt1)leads tomarked reduction of SNA-bindingSiaa2-6Gal/GalNAconpodocalyxin inmicefollowed by albuminuria, rapid progres-sion of glomerulosclerosis, effacement ofpodocyte foot processes, and thicken-ing of the glomerular basement mem-brane (37). These results might suggestthat expression of mature O-glycans mod-ified by a key glycosyltransferase in thekidneys, especially on podocytes, is cru-cial to the maintenance of the normalrenal architecture. If abnormalities ofrenal O-glycosylation occur, parts ofmature O-glycans could undergo reduc-tion andbeexcreted in the urine. AlthoughC1galt1 is not involved in the modificationof Siaa2-6Gal/GalNAc (Supplementary
Fig. 6A), deletion of C1galt1 induces adecrease of Siaa2-6Gal/GalNAc and an in-crease of Tn antigen in mouse podocalyxin(37), suggesting that impairment of oneO-glycosyltransferase also affects otherO-glycosylation pathways and leads tothe creation of abnormal O-glycans. Thecurrent study findings might supportthese suggestions as follows.
We found that urinary levels of glycansbinding to SNA, ABA, Jacalin, and ACAwere significantly higher in the severeCKD heat map group and that the cu-mulative incidence rate of the renal out-come was significantly higher in thehighest glycan quartiles than in the otherglycan quartiles not only in all patientsbut also in patients with normoalbumi-nuria/microalbuminuria (SupplementaryFig. 5 and Fig. 2A and B). Of note, urinaryexcretion of Siaa2-6Gal/GalNAc bindingto SNA was significantly higher in non-CKD patients with diabetes than in non-CKD control subjects without diabetes,
suggesting that a change in the glycosyl-ation of Siaa2-6Gal/GalNAc could occur inthe early stage of DKD before the onset ofalbuminuria or detectable deterioration ofrenal function. On the basis of these resultsand the above-mentioned hypothesis thatchanges of glycosylation may lead torenal structural damage, it might be rea-sonable that indexes for these glycansmay be significantly associated with therisk of the renal outcome independentlyof baseline albuminuria and eGFR (Fig. 1).
Galb4GlcNAc binds to RCA120 and isinvolved in N- and O-glycosylation. Dur-ing O-glycosylation, Galb4GlcNAc playsa role in the extension of cores 2 and 4,which are finally modified by sialic acid(32) (Supplementary Fig. 6A). Therefore,if impairment of the enzyme adding sialicacid to cores 2 and 4 exists, urinary ex-cretion of Galb4GlcNAc could increaseby the same mechanism as that whichincreases urinary levels of Siaa2-6Gal/GalNAc and Galb3GalNAc. In this study,
Figure 2—Cumulative incidence rate of the renal outcome. A: Cumulative incidence rate in all patients stratified by urinary glycan quartiles. Theestimated 4-year renal failure rate was 25%, 22%, 25%, 23%, and 26% in patients from the highest glycan quartiles for SNA, RCA120, ABA, Jacalin,and ACA, respectively, whereas it was 15% in patients from the lowest glycan quartile for DBA. The cumulative incidence rate was significantly higherin the highest glycan quartiles for SNA, RCA120, ABA, Jacalin, and ACA than in the other quartiles (P, 0.001), although it was significantly higher inthe lowest glycan quartile for DBA than in the other quartiles (P , 0.05). B: Cumulative incidence rate in the highest urinary glycan quartiles andother glycan quartiles combined in patients with normoalbuminuria or microalbuminuria. The cumulative incidence rate was significantly higher in thehighest glycan quartile (Q4) for SNA, ABA, Jacalin, and ACA than in the other quartiles combined (Q1–3 vs. Q4: P = 0.026 for SNA, P = 0.028 for ABA,P = 0.019 for Jacalin, and P = 0.002 for ACA). On the other hand, the difference of the cumulative incidence rate between Q4 and Q1–3 was not significantfor RCA120 and DBA (Q1–3 vs. Q4: P = 0.433 for RCA120, P = 0.270 for DBA). C: Cumulative incidence rate in patients with macroalbuminuria andhigher or lower glycan indexes than the median. The cumulative incidence rate was significantly higher in patients with higher glycan indexes thanin those with lower glycan indexes (P, 0.001), except for the glycan index for DBA (P = 0.186). Outcome:$30% decline of eGFR or dialysis as a resultof ESRD. The log-rank test was used for failure analysis.
care.diabetesjournals.org Mise and Associates 1771
the glycan index for RCA120 was stronglycorrelated with that for SNA (r = 0.921)(Supplementary Table 8), which suggeststhat the common underlying mechanismis impairment of sialylation. In addition,Ravida et al. (12) demonstrated thatglycans binding to RCA120 gradually
show a significant decrease in the renalcortex of diabetic rats with progressionof DKD, whereas these glycans increasein nondiabetic rats. Therefore, increasedurinary levels of Galb4GlcNAc might re-flect abnormal O-glycosylation in kidneytissues.
With regard to GalNAca3GalNAc, wecould not reasonably explain why urinaryexcretion of this glycan binding to DBAwas negatively correlated with the renaloutcome. DBA is well known for recog-nizing the epitope of glycans on thesurface of type A red blood cells (33).
Figure 2dContinued.
1772 Novel Glycan Biomarkers of DKD Diabetes Care Volume 41, August 2018
It has also been reported to bind toglomerular lesions in human DKD andto the distal tubules in human kidneytissue, but further investigations havenot been performed (38,39). DBA recog-nizes core 5 (GalNAca3GalNAc), which isa rare component of O-glycans, and therelevant glycosyltransferase has not yetbeen reported (32). Therefore, additionalinvestigation of the role of this glycan inthe kidneys and DKD is needed.
Taken together, changes of glycosyla-tion in DKD seem to be strongly associatedwith renal prognosis because abnormalglycosylation might be involved in theprogression of DKD. Supplementary Fig.6 shows a scheme of our hypothesis re-garding the mechanism by which abnor-malities of glycosylation may be associatedwith progression of DKD.We also speculatethat the urinary glycosylation differencewell reflects the local glycosylation changesin kidney tissues rather than alterations ofsystemic glycosylation on the basis of thediscrepancies of glycan profiles betweenurine and serum/plasma samples (15,16,40).Serum protein N-glycan profiling has beenshown to lower levels of fucosylated bian-tennary glycans, and higher levels of com-plexes that have more GlcNAc and heavilygalactosylated and sialylated glycans havebeen associated with higher HbA1c levelsin patients with diabetes (15,16), whereasthose associations were not observed inthe current urinary glycan profiling. Simi-larly, plasma IgG N-glycan profiling revealedthat higher levels of bisecting GlcNAc andlower levels of sialylation were correlatedwith older age, especially .60 years (40),whereas these results were not seen in thecurrent study with urine samples. In thepreliminary data of our previous study (14),we found completely different glycosyla-tion patterns between serum and urinesamples as well as different urinary glyco-sylation among patients with biopsy-provenkidney diseases, including DKD, and similardegrees of proteinuria and eGFR, whichmight support our speculation. These hy-potheses are based on in vivo and clinicalstudies of human DKD, so additional in-vestigation of glycosylation changes in hu-man kidney tissue is required.
Potential limitations of this study werethat the current lectin microarray systemdoes not allow complete determinationof glycan structures as can MS, and un-known preferred glycan structure to eachlectin might have biased the results.However, this technique is useful for
Table
2—Category-fre
eNRI,ID
I,andAIC
forpredictin
gthe4-yearoutco
mewith
glyca
nindexdata,anddiffe
rence
ofC-in
dexbetw
eenCoxregressio
nmodels
with
orwith
outglyca
nindexdata
Catego
ry-freeNRI
(95%CI)
Pvalu
eAbsolute
IDI
(95%CI)
Pvalu
eRelative
IDI
(95%CI)
Pvalu
eAIC
C-in
dex
(95%CI)
Differen
ceofC-
index
(95%CI)
Pvalu
e
Only
covariates
295.90.89
(0.84–0.93)
Glycan
toSN
A(Siaa
2-6Gal/G
alNAc)
0.27(2
0.13to
0.66)0.184
0.02(2
0.01to
0.05)0.184
0.06(2
0.04to
0.17)0.232
292.50.89
(0.84–0.94)
0.00(2
0.01to
0.01)0.958
Glycan
toRCA120
(Galb
4GlcN
Ac)
0.02(2
0.25to
0.29)0.891
0.00(2
0.01to
0.02)0.712
0.01(2
0.03to
0.05)0.725
297.60.89
(0.84–0.93)
20.00
(20.01
to0.00)
0.491
Glycan
toDBA(GalN
Aca
3GalN
Ac)
0.20(2
0.11to
0.50)0.203
0.01(2
0.01to
0.02)0.488
0.02(2
0.04to
0.07)0.529
295.80.89
(0.85–0.94)
0.00(2
0.00to
0.01)0.402
Glycan
toABA(Galb
3GalN
Ac)
0.26(2
0.10to
0.61)0.156
0.01(2
0.01to
0.04)0.309
0.04(2
0.04to
0.12)0.336
295.10.89
(0.84–0.94)
0.00(2
0.01to
0.01)0.801
Glycan
toJacalin
(Galb
3GalN
Ac)
0.11(2
0.18to
0.40)0.439
0.01(2
0.01to
0.04)0.375
0.04(2
0.05to
0.12)0.383
295.70.89
(0.84–0.94)
0.00(2
0.01to
0.01)0.834
Glycan
toACA(Galb
3GalN
Ac)
0.13(2
0.17to
0.44)0.388
0.01(2
0.01to
0.04)0.286
0.04( 2
0.04to
0.13)0.311
294.40.89
(0.84–0.94)
0.00(2
0.01to
0.01)0.975
Combinatio
noffourtyp
esofglycan
(SNA,RCA120,
DBA,andABA)
0.39(0.11
–0.67)0.006
0.06(0.01
–0.10)0.009
0.17(0.01
–0.33)0.037
284.90.90
(0.85–0.94)
0.01(2
0.01to
0.03)0.351
Combinatio
noffourtyp
esofglycan
(SNA,RCA120,
DBA,andJacalin
)0.42
(0.15–0.69)
0.0020.06
(0.01–0.10)
0.010.17
(0.01–0.33)
0.033284.9
0.90(0.85
–0.94)0.01
(20.01
to0.03)
0.441
Combinatio
noffourtyp
esofglycan
(SNA,RCA120,
DBA,andACA)
0.46(0.20
–0.71),0.001
0.06(0.01
–0.10)0.009
0.17(0.02
–0.33)0.027
284.50.90
(0.85–0.94)
0.01(2
0.01to
0.03)0.402
Combinatio
nofall
glycans
0.51(0.22
–0.80)0.001
0.06(0.02
–0.10)0.008
0.18(0.01
–0.35)0.036
287.00.90
(0.85–0.94)
0.01(2
0.01to
0.03)0.447
Covariates
were
age,sex,
MAP,
HbA1c ,eG
FR,andlog-transfo
rmed
urin
aryalb
umin
excretionat
baselin
e.
care.diabetesjournals.org Mise and Associates 1773
differentiating urinary glycan profiling inindividuals, and it enabled us to measurea wide range of urinary glycan intensity inhigh throughput without a liberation pro-cess, although conventional methods, in-cluding MS, require both prior liberationand subsequent labeling (19). Anotherlimitation was that this study includedpatients with various stages of CKD, andrenal biopsy was not performed in allsubjects. Therefore, we cannot excludethe possibility that the decline of eGFRin some patients resulted from kidneydiseases other than DKD. However, inprevious well-known cohort studies oftype 2 diabetes, DKD was not always con-firmed by renal biopsy (1,3). Moreover, inthe patients with diabetes, renal biopsy isconsidered when hematuria, granular casts,sudden onset of nephrotic syndrome, andrapidly progressive glomerulonephritis areindicated, whereas such conditions werenot observed during follow-up in the currentstudy. We hope that an ongoing researchbiopsy study of DKD at our institution willsolve this issue in the future.In conclusion, we demonstrate that
urinary excretion of glycans binding toseveral lectins, including SNA (recognizingSiaa2-6Gal/GalNAc), RCA120 (Galb4GlcNAc),DBA (GalNAca3GalNAc), ABA, Jacalin, andACA (Galb3GalNAc), are significantly asso-ciated with renal outcome in patients withtype 2 diabetes. The addition of the com-bined glycan index to a model with stan-dard risk factors significantly improvesprediction of renal outcome, suggestingthat these urinary glycans may be novelpredictors of renal prognosis in patientswith type 2 diabetes. The findings couldprovide new insights into changes of gly-cosylationrelatedtoDKD.Themechanismsthat underlie differences of urinary glycanexcretion and changes of glycosylationin DKD should be investigated further.
Acknowledgments. The authors thank Dr.Kazuyuki Hida (National Hospital OrganizationOkayama Medical Center), Dr. Tatsuaki Nakato(Okayama Saiseikai General Hospital), Dr. TakashiMatsuoka (Kurashiki Central Hospital), Dr. IkkiShimizu(TheSakakibaraHeart InstituteofOkayama),Dr. Tomokazu Nunoue (Tsuyama Chuo Hospital),Dr. Katsuhiro Miyashita (Japanese Red CrossOkayama Hospital), Dr. Shinichiro Ando (OkayamaCity General Medical Center), and Dr. Akiho Seki(Okayama Health Foundation) for providing dataand for patient treatments. The authors alsothank Drs. Ryo Kodera and Satoshi Miyamoto(Okayama University Hospital), Drs. AkihiroKatayama, Chigusa Higuchi, Hidemi Takeuchi,
Yusuke Shibata, Ichiro Nojima, Yuzuki Kano,Yuriko Yamamura, Nozomu Otaka, Chieko Kawakita,Keigo Hayashi, and Yasuhiro Onishi (Okayama Uni-versity Graduate School of Medicine, Dentistryand Pharmaceutical Sciences, Okayama) for collect-ing data. The authors greatly appreciate Dr. HirofumiMakino (Okayama University) for critical sugges-tions in the design of the current investigation.Funding. This work was partly supported by theMinistry of Health, Labour and Welfare in Japan(Health Labor Sciences Research Grant 201413003),the Japan Agency for Medical Research and De-velopment (AMED) (grant 17ek0210095h0001),Novo Nordisk Pharma Ltd. (Junior Scientist Devel-opment Grant [2016–2017]) from the OkinakaMemorial Institute for Medical Research (a grantin 2017), and The Yukiko Ishibashi Foundation (agrant in 2016).Duality of Interest. M.Ya. was a former em-ployee of GP BioSciences Co., Ltd., and is cur-rently an employee of GlycoTechnica, Ltd. Noother potential conflicts of interest relevant tothis article were reported.Author Contributions. K.M. contributed todesigning the research, analyzing and interpret-ing data, measuring urinary glycan levels, collect-ing and summarizing clinical data, and writing themanuscript. M.I. contributed to collecting, sum-marizing, and assessing clinical data. S.Y. con-tributed to measuring urinary glycan levels andcollecting and summarizing clinical data. S.T.,A.T., H.A.U., J.E., A.N., and D.O., contributed tomanaging patients and assessing data. M.Yo.contributed to interpreting data and performingstatistical analyses. M.Ya. contributed to measuringurinary glycan levels; interpreting data, especiallyurinary glycan data; and writing themanuscript. K.S.contributed to managing patients and assessingdata. J.W. was responsible for the study design,supervised data collection and data analysis, andcontributed to drafting and editing themanuscript.J.W. is the guarantor of this work and, as such, hadfull access to all the data in the study and takesresponsibility for the integrity of the data and theaccuracy of the data analysis.Prior Presentation. Part of this work was pre-sented at the American Society of NephrologyScientific Oral Sessions, New Orleans, LA,31 October–5 November 2017, and the Interna-tional Society of Nephrology Frontiers Meetings,Tokyo, Japan, 22–25 February 2018.
References1. Niewczas MA, Gohda T, Skupien J, et al.Circulating TNF receptors 1 and 2 predict ESRDin type 2 diabetes. J Am Soc Nephrol 2012;23:507–5152. Gohda T, Niewczas MA, Ficociello LH, et al.Circulating TNF receptors 1 and 2 predict stage3 CKD in type 1 diabetes. J Am Soc Nephrol 2012;23:516–5243. Fufaa GD, Weil EJ, Nelson RG, et al.; ChronicKidney Disease Biomarkers Consortium Investi-gators.AssociationofurinaryKIM-1, L-FABP,NAGand NGAL with incident end-stage renal diseaseand mortality in American Indians with type 2diabetes mellitus. Diabetologia 2015;58:188–1984. Nowak N, Skupien J, Niewczas MA, et al.Increased plasma kidney injury molecule-1 sug-gests early progressive renal decline in non-proteinuric patients with type 1 diabetes. KidneyInt 2016;89:459–467
5. Nadkarni GN, Rao V, Ismail-Beigi F, et al.Association of urinary biomarkers of inflamma-tion, injury, and fibrosis with renal functiondecline: the ACCORD trial. Clin J Am Soc Nephrol2016;11:1343–13526. Foster MC, Inker LA, Hsu CY, et al.; CKDBiomarkers Consortium. Filtration markers aspredictors of ESRD and mortality in SouthwesternAmerican Indians with type 2 diabetes. Am J KidneyDis 2015;66:75–837. Wanner C, Inzucchi SE, Lachin JM, et al.;EMPA-REG OUTCOME Investigators. Empagliflo-zin and progression of kidney disease in type 2diabetes. N Engl J Med 2016;375:323–3348. Mann JFE, Ørsted DD, Brown-Frandsen K,et al.; LEADER Steering Committee and Inves-tigators. Liraglutide and renal outcomes in type 2diabetes. N Engl J Med 2017;377:839–8489. Mise K, Hoshino J, Ueno T, et al. Clinical andpathological predictors of estimated GFR declinein patients with type 2 diabetes and overt pro-teinuric diabetic nephropathy. Diabetes MetabRes Rev 2015;31:572–58110. Ohtsubo K, Chen MZ, Olefsky JM, Marth JD.Pathway to diabetes through attenuation ofpancreatic beta cell glycosylation and glucosetransport. Nat Med 2011;17:1067–107511. Gellai R, Hodrea J, Lenart L, et al. Role ofO-linked N-acetylglucosamine modification indiabetic nephropathy. Am J Physiol Renal Physiol2016;311:F1172–F118112. Ravida A, Musante L, Kreivi M, et al. Glyco-sylation patterns of kidney proteins differ in ratdiabetic nephropathy. Kidney Int 2015;87:963–97413. Kuno A, Uchiyama N, Koseki-Kuno S, et al.Evanescent-field fluorescence-assisted lectin mi-croarray: a new strategy for glycan profiling. NatMethods 2005;2:851–85614. Inoue K, Wada J, Eguchi J, et al. Urinaryfetuin-A is a novel marker for diabetic nephrop-athy in type 2 diabetes identified by lectin micro-array. PLoS One 2013;8:e7711815. Testa R, Vanhooren V, Bonfigli AR, et al.N-glycomic changes in serum proteins in type 2diabetes mellitus correlate with complicationsand with metabolic syndrome parameters. PLoSOne 2015;10:e011998316. Bermingham ML, Colombo M, McGurnaghanSJ, et al.; SDRN Type 1 Bioresource Investigators.N-glycan profile and kidney disease in type 1 di-abetes. Diabetes Care 2018;41:79–8717. Seino Y, Nanjo K, TajimaN, et al.; Committeeof the Japan Diabetes Society on the Diagnos-tic Criteria of Diabetes Mellitus. Report of thecommittee on the classification and diagnosticcriteria of diabetes mellitus. J Diabetes Investig2010;1:212–22818. Chapter 1: definition and classification ofCKD. Kidney Int Suppl (2011) 2013;3:19–6219. Hirabayashi J, Yamada M, Kuno A, Tateno H.Lectin microarrays: concept, principle and ap-plications. Chem Soc Rev 2013;42:4443–445820. Safina G. Application of surface plasmonresonance for the detection of carbohydrates,glycoconjugates, and measurement of thecarbohydrate-specific interactions: a compar-ison with conventional analytical techniques.A critical review. Anal Chim Acta 2012;712:9–2921. Cheng CI, Chang YP, Chu YH. Biomolecularinteractions and tools for their recognition: focuson the quartz crystal microbalance and its diverse
1774 Novel Glycan Biomarkers of DKD Diabetes Care Volume 41, August 2018
surface chemistries and applications. Chem SocRev 2012;41:1947–197122. Horio M, Imai E, Yasuda Y, Watanabe T,Matsuo S. Modification of the CKD epidemiologycollaboration (CKD-EPI) equation for Japanese:accuracy and use for population estimates. Am JKidney Dis 2010;56:32–3823. Kashiwagi A, Kasuga M, Araki E, et al. In-ternational clinical harmonization of glycatedhemoglobin in Japan: from Japan Diabetes Societyto National Glycohemoglobin Standardization Pro-gram values. J Diabetes Investig 2012;3:39–4024. Fong DS, Aiello L, Gardner TW, et al.; Amer-ican Diabetes Association. Retinopathy in dia-betes. Diabetes Care 2004;27(Suppl. 1):S84–S8725. Holman RR, Paul SK, Bethel MA, MatthewsDR, Neil HA. 10-year follow-up of intensive glu-cose control in type 2 diabetes. N Engl J Med2008;359:1577–158926. Writing Team for the Diabetes Control andComplications Trial/Epidemiology of DiabetesInterventions and Complications Research Group.Effect of intensive therapy on the microvascularcomplications of type 1 diabetes mellitus. JAMA2002;287:2563–256927. Pencina MJ, D’Agostino RB Sr., D’AgostinoRB Jr., Vasan RS. Evaluating the added predictiveability of a new marker: from area under the ROC
curve to reclassification and beyond. Stat Med2008;27:157–172; discussion 207–11228. Pencina MJ, D’Agostino RB Sr, SteyerbergEW. Extensions of net reclassification improve-ment calculations to measure usefulness of newbiomarkers. Stat Med 2011;30:11–2129. Pencina MJ, D’Agostino RB, Pencina KM,Janssens AC, Greenland P. Interpreting incre-mental value of markers added to risk predictionmodels. Am J Epidemiol 2012;176:473–48130. Mise K, Hoshino J, Ueno T, et al. Prognosticvalue of tubulointerstitial lesions, urinaryN-acetyl-b-d-glucosaminidase, and urinary b2-microglobulin in patients with type 2 diabetesand biopsy-proven diabetic nephropathy. Clin JAm Soc Nephrol 2016;11:593–60131. Levin A, Stevens PE. Summary of KDIGO2012 CKD Guideline: behind the scenes, needfor guidance, and a framework for moving for-ward. Kidney Int 2014;85:49–6132. Gill DJ, Clausen H, Bard F. Location, location,location: new insights into O-GalNAc proteinglycosylation. Trends Cell Biol 2011;21:149–15833. Belicky S, Katrlık J, Tkac J. Glycan and lectinbiosensors. Essays Biochem 2016;60:37–4734. Kerjaschki D, Sharkey DJ, Farquhar MG.Identification and characterization of podoca-lyxin–the major sialoprotein of the renal glomer-ular epithelial cell. J Cell Biol 1984;98:1591–1596
35. Dekan G, Gabel C, Farquhar MG. Sulfate
contributes to the negative charge of podocalyxin,
the major sialoglycoprotein of the glomerular fil-
tration slits. Proc Natl Acad Sci U S A 1991;88:
5398–540236. Gelberg H, Healy L, Whiteley H, Miller LA,
Vimr E. In vivo enzymatic removal of alpha 2–.6-
linked sialic acid from the glomerular filtra-
tion barrier results in podocyte charge alter-
ation and glomerular injury. Lab Invest 1996;74:
907–92037. Song K, Fu J, Song J, et al. Loss of mucin-type
O-glycans impairs the integrity of the glomerular
filtration barrier in the mouse kidney. J Biol Chem
2017;292:16491–1649738. Holthofer H, Pettersson E, Tornroth T. Di-
abetes mellitus associated changes in glomerular
glycocompounds: a fluorescence microscopical
study. Histochem J 1987;19:351–35639. Truong LD, Phung VT, Yoshikawa Y, Mattioli
CA. Glycoconjugates in normal human kidney.
A histochemical study using 13 biotinylated
lectins. Histochemistry 1988;90:51–6040. Ruhaak LR, Uh HW, Beekman M, et al. De-
creased levels of bisecting GlcNAc glycoforms
of IgG are associated with human longevity.
PLoS One 2010;5:e12566
care.diabetesjournals.org Mise and Associates 1775
