N-glycosylation proteome enrichment analysis in kidney reveals differences between diabetic mouse models
© The Author(s) 2016
Received: 5 May 2016
Accepted: 23 August 2016
Published: 15 October 2016
Diabetic nephropathy (DN) is a late complication in both type 1 diabetes mellitus (T1DM) and T2DM. Already at an early stage of DN morphological changes occur at the cell surface and in the extracellular matrix where the majority of the proteins carry N-linked glycosylations. These glycosylated proteins are highly important in cell adhesion and cell–matrix processes but not much is known about how they change in DN or whether the distinct etiology of T1DM and T2DM could have an effect on their abundances.
We enriched for the N-glycosylated kidney proteome in db/db mice dosed with insulin or vehicle, in streptozotocin-induced (STZ) diabetic mice and healthy control mice dosed with vehicle. Glycopeptides were analyzed with label-free shotgun mass spectrometry and differential protein abundances identified in both mouse models were compared using multivariate analyses.
The majority of the N-glycosylated proteins were similarly regulated in both mouse models. However, distinct differences between the two mouse models were for example seen for integrin-β1, a protein expressed mainly in the glomeruli which abundance was increased in the STZ diabetic mice while decreased in the db/db mice and for the sodium/glucose cotransporter-1, mainly expressed in the proximal tubules which abundance was increased in the db/db mice but decreased in the STZ diabetic mice. Insulin had an effect on the level of both glomerular and tubular proteins in the db/db mice. It decreased the abundance of G-protein coupled receptor-116 and of tyrosine-protein phosphatase non-receptor type substrate-1 away from the level in the healthy control mice.
Our finding of differences in the N-glycosylation protein profiles in the db/db and STZ mouse models suggest that the etiology of DN could give rise to variations in the cell adhesion and cell–matrix composition in T1DM and T2DM. Thus, N-glycosylated protein differences could be a clue to dissimilarities in T1DM and T2DM at later stages of DN. Furthermore, we observed insulin specific regulation of N-glycosylated proteins both in the direction of and away from the abundances in healthy control mice.
KeywordsN-linked glycosylation N-glycosylation Diabetic nephropathy Proximal tubules Glomerulus Mass spectrometry Insulin STZ db/db
The prevalence of both T1DM and T2DM is increasing and together they affect over 340 million people worldwide. The insulin-dependent T1DM covers 5–10 % of the cases where as T2DM comprise 90 % of the disease and the prevalence increases more rapidly [1–3]. Despite different etiology, both T1DM and T2DM result in a 20–30 % occurrence of DN. Type 2 DM patients often go undiagnosed for several years, which means that DN often is diagnosed at a later stage in development. The larger heterogeneity of T2DM patients is reflected in larger structural differences in the kidneys of patients with DN compared to T1DM patients with DN [4, 5]. The structural changes of T1DM and T2DM during progression of DN are however roughly similar [5, 6] with tubular and glomerular basement membrane (BM) thickening, extracellular matrix (ECM) and mesangial expansion and changes in the proteoglycan structure both at the cell surface and in the extracellular space .
In the kidney-specific processes of blood filtering, protein retention and reabsorption of fluid and small molecules, the extracellular matrix and its connection to cellular surfaces are regions of high functional importance. In the glomeruli the podocytes guard the slit diaphragm and prevent the loss of larger molecules from the blood to the primary urine and in the tubules solute and water reabsorption is managed . A large proportion of the extracellular and cell surface proteins involved in those processes, including proteoglycans, carry glycosylation on asparagine (N) residues . The N-linked glycosylation pattern is controlled by enzymatic processes as compared to the non-enzymatic, unspecific glucose dependent glycation of proteins, measured by the level of glycated hemoglobin-A1c (HbA1c%) [10, 11].
Addressing changes in the extracellular space and the cell surface can be a challenge since this sub-proteome may be difficult to map due to technical aspects in a regular proteome analysis or in a non-biased way in antibody based analysis . Changes in the N-glycosylated proteome have repeatedly been shown in various diseases and cell-types [12–15]. Since N-linked glycosylation is a hallmark of cell membrane and extracellular proteins [16, 17] and early kidney changes in DN are observed at those locations  we enriched for glycosylated peptides with hydrazide capture  to target this sub proteome. By using label free shotgun mass spectrometry, we compared the N-glycosylated fraction of the kidney proteomes from two regularly used diabetic mouse models; the STZ-induced diabetic mice with reduced pancreatic beta cell mass and the obese db/db mice with deficient leptin receptor signaling leading to obesity, insulin resistance and finally overt diabetes. Both mouse models are characterized by having early signs of DN in the form of increased albumin excretion rates (AER)  driven by hyperglycemia.
Hyperglycemia has been shown to promote DN by increasing oxidative stress [20, 21] and maintaining stable normal blood glucose (BG) levels has a paramount position in diabetes care [22, 23]. In T1DM, insulin is the only treatment option adequately lowering BG levels. Type 2 DM can be managed with several different compounds in the initial phases in order to rescue the reduced insulin sensitivity. At later stages in T2DM insulin administration can be required due to decreased beta-cell mass and impaired insulin secretion . The effects of insulin on the N-glycosylation pattern in DN are not known but changes in proteoglycans are reported in DN and altered glycosylation levels are well known features from diseases like cancer . In the kidneys of STZ diabetic rats, the glycosylation pattern was recently shown to be altered over time , however, it is not known if the different etiologies of diabetes give rise to variations in the glycosylation pattern in DN.
We sought to compare the N-glycosylated protein profiles in the kidneys of the db/db and STZ induced diabetic mouse models since the two models represent separate etiologies of diabetes and both exhibit traits of early stage DN . In addition, the effects of insulin on N-glycosylation was investigated in the db/db mouse model kidney to elucidate its influence on the extracellular and cell surface proteome as previously described by Bausch-Fluck et al. . Insulin has earlier been reported to contribute to podocyte survival which is important for maintaining proper glomerular function [26, 27].
Experimental design and mouse parameters
The STZ diabetic mice reached 33.4 ± 1.45 mM in BG 10 weeks after diabetes induction relative to a fairly stable level of 6.06 ± 1.28 mM throughout the study in the healthy NoSTZ control group. Body weight in the STZ mice did not change much from baseline and was ~5 g lower than controls at termination (25.3 ± 1.2 vs. 30 ± 0.5 g).
The AER was ~tenfold higher in the STZ mice at 578 ± 117 µg/24 h compared with 55.7 ± 12.0 µg/24 h in the NoSTZ controls at the final sampling time point.
Mouse parameters in the db/db and STZ mouse models
db/db mouse model
STZ mouse model
Control db/+ (n = 5)
db/db (n = 5)
Insulin db/db (n = 5)
Control (STZ) (n = 6)
STZ (n = 6)
10 weeks old, baseline (insulin dosing start)
10 weeks old, baseline, 2 weeks after STZ
8.85 ± 0.48
19.27 ± 2.29
24.34 ± 1.28
6.50 ± 0.27
22.1 ± 1.60
4.06 ± 0.08
6.44 ± 0.48
7.88 ± 0.44
AER (μg/24 h)
42.5 ± 9.8, n = 4
420.8 ± 111.9
529.3 ± 115
40.8 ± 16.2, n = 5
408 ± 114
27.1 ± 0.9
46.66 ± 0.5
47.92 ± 0.2
25.2 ± 1.1
23.6 ± 0.4
16 weeks old, 6 weeks after insulin dosing start
14 weeks old, 6 weeks after STZ
8.4 ± 0.33
22.88 ± 1.86
16.14 ± 1.61
5.63 ± 0.25
30.9 ± 3.19
4.16 ± 0.10
7.94 ± 0.62
7.32 ± 0.23
3.62 ± 0.031
6.95 ± 0.1
AER (μg/24 h)
59.4 ± 13.3, n = 4
855.3 ± 272
488.6 ± 61.2
55.6 ± 5.39, n = 5
424 ± 71
31 ± 0.8
54 ± 3.2
57.2 ± 1.1
28.1 ± 1.0
24.8 ± 0.5
22 weeks old, 12 weeks after insulin dosing start
18 weeks old, 10 weeks after STZ
8.91 ± 0.28
23.12 ± 2.54
21.23 ± 1.49
8.57 ± 1.21
33.4 ± 1.45
4.3 ± 0.07
8.52 ± 0.70
6.78 ± 0.23
3.75 ± 0.06
7.60 ± 0.13
AER (μg/24 h)
112.2 ± 54.5, n = 4
930.2 ± 184.8
795.1 ± 183.8
55.7 ± 12.0, n = 5
578 ± 117
32.7 ± 0.7
53.7 ± 3.8
62.3 ± 0.9
30 ± 0.5
25.3 ± 1.2
Visualizing the STZ and db/db mouse models with multivariate analyses and clustering
Quantitative comparison of the N-glycosylated proteomes of the mouse models
Selected proteins identified in both mouse models with opposite or similar protein regulation
STZ versus NoSTZ
db/db versus db/+
db/db insulin versus dbdbB
Protein abuncance level
Protein abundance level
G = H, T = L
Sodium/glucose cotransporter 1
T = M
T = H
Alkaline phosphatase, unreviewed
T = M
Lysosomal membrane glycoprotein 1, unreviewed
G = H, T = H
G = L, T = L
Integrin alpha-3, isoform 2
G = M, T = H
Family with sequence similarity 151, member A
T = H
Cadherin-related family member 5, unreviewed
T = H
Meprin A subunit beta, isoform 2
ND, RNA ND
Solute carrier organic anion transporter family, member 1a6, unreviewed
G = L, T = L
Ingenuity pathway analysis
Proteins in the top rated networks in IPA in both mouse models conform
Molecules in network
Top diseases and functions
AQP2, ATP1B1, CTSV, DPEP1, EMILIN1, FAM151A, HSPA8, ILK, ITGA3, KLK3, MEP1A, MEP1B, PITX2, PKD1, SLC5A10, STUB1, SYNE1, TREH
Cell-to-cell signaling and interaction, renal and urological system development and function, organ morphology
AQP11, BICC1, CA3, CLCN5, CLU, CUBN, EGF, EGFR, ITGA1, ITGB1, LGMN, LRP2, SLC34A2, SOD1
Organ morphology, organismal development, renal and urological system development and function
ACE2, COL4A3, CTGF, FN1, LAMA1, LAMA5, LAMB1, LAMB2, LAMC1
Tissue development, decreased levels of albumin, tissue morphology
ADIPOQ, ATP2B1, BGN, CALB1, COL18A1, COL1A1, COL1A2, COL4A1, COL4A3, CTGF, CYP27B1, DCN, FAS, FBN1, FCGR2B, ICAM1, IFNG, ITGAV, KL, LEP, LUM, PARP1, SLC34A1, SLC9A3, SLC9A3R1, SMAD4, SMAD7, SPARC, TGFB1, TNF, TNFRSF1B, TRPV5, UMOD, VCAM1, VDR
Cellular movement, hematological system development and function, immune cell trafficking
AK2, ALDOB, AQP1, ATP1B1, AVPR2, DPEP1, EMILIN1, FAM151A, KLK3, MEP1A, MEP1B, MIOX, NAPSA, PITX2, PKD1, SLC5A10, TREH, TTR
Small molecule biochemistry, hematological system development and function, tissue development
ACE2, ACTA2, AQP11, BICC1, CLU, COL1A1, COL1A2, COL4A1, COL4A3, CTGF, Ccl2, EGF, EGFR, FCGR2B, FN1, ICAM1, ITGA1, ITGAV, ITGB1, LAMA1, LAMA5, LAMB1, LAMB2, LAMC1, PARP1, PKD1, RALBP1, SMAD3, SMAD4, SMAD7, TGFB1, TNF, TNFRSF1B, UMOD, VEGFA
Organismal injury and abnormalities, cellular movement, hematological system development and function
BGN, COL6A3, DCN, FAS, FBN1, LUM
Hair and skin-, skeletal and muscular system-development and function, organ morphology
CA2, CALB1, CYP27B1, EZR, KL, MSN, S100G, SLC34A1, SLC9A3R1, TRIM24, TRPV5, VDR
Drug metabolism, lipid metabolism, molecular transport
The effect of insulin on the db/db mice
In addition to the differences and similarities between the N-glycosylated proteomes of the STZ and db/db mouse models, we investigated the effect of insulin on the db/db mouse model. The db/db insulin group was poorly separated from the db/db vehicle mice in the unsupervised PCA (Fig. 2c) although a 3D projection of the PCA showed that the db/db insulin and db/db vehicle groups were separable (see Additional file 2: Panel C). The db/db insulin group was clearly separated from the db/db vehicle in the supervised OPLS analysis (Fig. 2d). When comparing the db/db insulin mice to the db/db vehicle mice, 173 of the 395 identified proteins had a VIP score above 1 in the OPLS model and a hierarchical clustering resulted in 37 proteins with P < 0.05.
Insulin effect on protein abundance in the db/db mouse model
db/db versus db/+
db/db insulin versus db/db
Protein abundance level
Protein abundance level
Away from healthy control
Laminin subunit alpha 1
G = M, T = H
Adhesion G protein-coupled receptor 116
ND, RNA T = H, G = L
CD166 antigen, activated leukocyte cell adhesion molecule, unreviewed
Insulin increases protein abundance towards healthy control
Tripeptidyl peptidase I, PUP
G = L, T = H
Low-density lipoprotein receptor-related protein 2
Cleft lip and palate associated transmembrane protein 1
ND, RNA T = H, G = L
Insulin reduces protein abundance towards healthy control
Hypoxia up-regulated protein 1
ND, RNA T
Sodium channel beta4 subunit, unreviewed
G = L, T = M
Acid phosphatase 2, lysosomal, PUP
ND, RNA T
Integrin alpha-V light chain, unreviewed
G = H, T = M
Alanyl (membrane) aminopeptidase, PUP
Inducible T-cell co-stimulator ligand
G = L, T = M
In this study we focused on the N-linked glycosylated proteome of the kidney in two well-known mouse models for diabetes with early DN; the STZ model where diabetes is induced by repeated intermediate doses of STZ resulting in a reduction of the beta cell mass and the obese db/db model with insulin resistance and later beta cell failure . In both the STZ and the db/db mice we saw increased levels of HbA1c% indicative of diabetes and increased albuminuria compared to the healthy control mice, implying early traits of DN. There was a comparable increase in HbA1c% between the healthy control and the diabetic groups in the two mouse models indicating similar degrees of diabetes despite the separate etiology in the two mouse models. Diabetes nephropathy is a late complication that in T1DM develops a decade or later after diabetes is diagnosed . It is therefore noteworthy that no mouse has a lifespan that is long enough to allow them to reach the later stages of renal failure seen in human subjects with DN .
We detected a higher number of significant protein changes within the db/db mouse model compared to the STZ mouse model between the shared proteins. An explanation could be that some of the proteins with the largest differences within the STZ mouse model not were identified in the db/db mouse model and were not included in the model-to-model comparison. Five of these STZ mouse model specific proteins are shown in Additional file 2: Panels A–B. Proteins not detected in one mouse model could be present at lower levels in the mouse model or could have lower levels of N-glycosylation which, with the N-glycosylation capture-technique we use, would result in a lower yield of peptides. In the comparison of the two mouse models we found that most protein abundances were unchanged or similarly regulated. However, the regulation of some proteins was inverted in the db/db vehicle and STZ mice compared to the healthy control mice in the respective mouse model, with increased protein abundance in the diabetic mice in one model while decreased abundance was observed in the other. Some of the proteins with differentiated protein abundances are summarized with their possible relation to kidney damage or diabetes in Additional file 1e.
Most extracellular, secreted and cell surface proteins are N-glycosylated  and the majority of the proteins we identified were cell surface proteins or associated with the extracellular space. Interestingly, three of the four proteins that we identified with the inversed regulation in the STZ and db/db mouse models mainly reside in the proximal tubules in human according to The Human Protein Atlas . The three mainly tubular proteins SGLT1, PROM1 and ALPL all had decreased protein abundances in the STZ and increased abundances in the db/db vehicle mice compared to the healthy control mice in the respective mouse model. The sodium/glucose co-transporters (SGLT) reabsorb glucose from the primary urine to the blood  and different expression pattern of SGLT1 in the two mouse models indicates that there could be DM type-specific differences in the protein abundance of this sodium glucose co-transporter. Prominin-1 is a cholesterol binding protein in the proximal tubules. No connection between DN and altered levels of PROM1 has previously been reported to our knowledge, but in a STZ mouse model of retinal vasculopathy and neuropathy, also secondary complications to diabetes, PROM1 in the photoreceptors was shown to be destroyed by metalloprotease (MMP)-9 whereas no retinopathy or neuropathy was seen in MMP-9 knockout mice . Those results indicate that the presence of PROM1 could be involved in reducing micro vascular damage in retinopathy since the lack of PROM1 is associated with retinopathy. Similar effects could take place in the proximal tubules in the STZ mice where PROM1 abundances were decreased compared to the NoSTZ healthy control mice. The tubular protein ALPL, also diametrically regulated in the STZ and db/db mouse models is proposed as a urinary biomarker for tubular damage in DN . Tubular changes has long been known to be correlated to alterations in the glomerular filtration rate  and these changes could have a closer connection to the initial states of progression in DN compared to glomerular changes .
Several of the proteins involved in the four top-rated networks in IPA were integrin and laminin subunits. Integrins are involved in cell adhesion, cell–matrix interactions and serve as receptors for several laminin subunits . Here an inverse regulation of ITGB1 (β1) protein abundance was seen in the two mouse models (Fig. 3c) whilst a similar regulation of both ITGA3 (α3) and LAMA1 abundances were observed in the mouse models [36, 37]. The changes we see in the integrin and laminin proteins indicate that there are structural changes taking place in the kidney at a very early stage of DN and some of them could be etiology specific.
Insulin is probably the most important tool used in diabetes care to control hyperglycemia and thereby to a large degree reduce the risk of late complications like DN. We saw that insulin had a clear effect on several of the renal proteins, in general tubular , in the db/db mice, mainly in the direction of the healthy db/+ control mice but also in the opposite direction. The level of the glomerular protein SIRPA was already decreased in the db/db vehicle mice compared to the healthy db/+ vehicle mice and insulin decreased the abundance even further. It has been shown to interact with nephrin in the slit diaphragm of the podocyte  and could be involved in podocyte injury . Insulin also significantly decreased the abundance of GPR116 similar to SIRPA in the db/db mice (Fig. 5a) and it has been shown in mouse adipose tissue that deletion of Gpr116 could cause impaired insulin tolerance . The unchanged protein abundance of GPR116 in the STZ mouse model compared to the decreased abundances in the db/db mice indicate mouse model specific differences that could be connected to an altered body weight in the db/db mouse.
In the majority of the proteins affected by insulin, the protein levels in the db/db mice were changed towards the levels seen in the healthy db/+ mice. The effects we observe of insulin on the protein abundance levels in the db/db mice are indicative of a protective effect. However, in the cases where the protein levels are further away from the healthy db/+ mice than the levels in the db/db vehicle mice, the effect of insulin could either be detrimental or compensatory, this we do not know.
We focused on ECM and cell surface kidney proteomics using N-linked glycosylated peptide enrichment and showed that there were significant protein differences between the STZ and db/db mouse models. Our data suggests that DM type specific protein differences in the kidney could precede later shared morphological alterations in DN.
We also show that insulin changes the N-glycosylated protein abundances in the db/db mouse model both in the direction of the levels in the healthy control mice and in the opposite direction. These insulin induced changes in protein abundance in the db/db mouse kidney could be adipose tissue related, caused by reduced hyperglycemia or other systemic effects of insulin in the mice, by insulin signaling in the kidney or could possibly compensate for the adverse effects of diabetes in the animals. Insulin administration, crucial for survival in type 1 DM but to a large extent replaceable in type 2 DM, has a distinct effect on the N-glycosylated proteome in the db/db mice that could be of high interest to compare with the effect of other type 2 DM intervention strategies.
The handling and use of all animals in the present study was approved by The Danish Animal Experiments Inspectorate and was carried out according to the guidelines of “The Council of Europe Convention for the protection of vertebrate animals used for experimental and other scientific purposes”.
Male 129SV mice 6 weeks of age and male db/db and db/+ mice 8 weeks of age were purchased from Charles River, Germany and fed Altromin 1324 and tap water ad libitum from arrival. The animals were housed in the animal unit facility at Novo Nordisk, Denmark (SV129 mice 10/cage and db/db and db/+ mice 5/cage) in controlled temperature (19–21 °C for SV129 and 23–24 °C for db/db) and a 12 h light, 12 h dark cycle. After 2 weeks of acclimatization one group of randomly selected 129SV mice were given STZ injections intraperitoneally twice at a dose of 125 mg/kg with 3 days between doses. Two weeks post STZ intervention, BG measurements were conducted and mice in the STZ group (STZ) with elevated BG above 16 mM were included in the experiment together with the group of SV129 mice that was not given STZ (NoSTZ). Concomitantly vehicle dosing (s.c. 4 mL/kg QD, vehicle composition: pH 7.4; 20 mM phosphate; 130 mM sodium chloride; 0.05 % polysorbate 80) was initiated of the STZ induced diabetic and healthy NoSTZ groups of mice.
After 2 weeks of acclimatization of the db/+ and db/db mice, BG measurements were conducted and db/db mice with BG above 16 mM were included in the experiment. The db/db mice were divided into dosing groups given vehicle (db/db vehicle) or insulin (db/db insulin) while the healthy db/+ control mice (db/+) were given vehicle. Vehicle composition and dosing was as with the STZ mice except for dosing volume (2 mL/kg) due to the larger size of the db/db mice. The first dose of insulin was 2 LinBits per 20 g mouse plus 1 LinBit (LinShin Inc, Scarborough, ON, Canada) for each additional 5 g of mouse. The insulin dose was thereafter adjusted according to BG profiles with insulin glargine (Nomeco A/S, Copenhagen, Denmark) at a dose of 10 U/kg twice a day. During the study the animals were weighed once weekly on a digital scale.
In vivo measurements
Blood glucose was analyzed once weekly on a Biosen S-line/5040 (EKF-diagnostics, Magdeburg, Germany) and HbA1c% was analyzed on a Cobas 6000 autoanalyzer (Roche Diagnostics Ltd, Rotkreuz, Switzerland) once weekly for the db/db mice and at 6 and 10 weeks after STZ dosing for the STZ mice. Blood was taken at all time points from non-fasted mice before dosing.
Metabolic cages (Techniplast S.p.A., Buguggiate, Italy) for individual collection of urine was used at baseline, 6 and 12 weeks (age 10, 16 and 22 weeks) for the db/db mice and at 2, 6 and 10 weeks (age 10, 14 and 18 weeks) after the STZ intervention for the STZ mice. Urine albumin excretion determination (AER) was determined using a sandwich ELISA (Bethyl Labs, Montgomery, TX, USA).
The animals in the db/db mouse model were sacrificed after 12.5 weeks of insulin dosing and the animals in the STZ mouse model were sacrificed 10 weeks after the STZ intervention by perfusion under isoflurane anesthesia with 20 mL 0.9 % NaCl with heparin (10 U/mL). Kidneys were weighed individually after having the surrounding fat removed and were snap frozen in liquid nitrogen. The left kidney from the db/db and the right kidney from the STZ animals were studied.
In the STZ mouse model, the total number of analyzed mice in the proteomics study was n = 12 with n = 6 STZ induced diabetic SV129 and n = 6 NoSTZ SV129 healthy controls, both groups dosed with vehicle. The total number of analyzed animals in the db/db mouse model part of the study were n = 15 with n = 5 db/db diabetic vehicle dosed, n = 5 db/db diabetic insulin dosed and n = 5 healthy db/+ control vehicle dosed mice (Fig. 1a, b).
Snap frozen kidney tissue was transferred to a Denator Stabilizor T1  (Denator, Gothenburg, Sweden). From the animals in the STZ mouse model, 50 mg of tissue was used resulting in one sample/animal and for the animals in the db/db mouse model 100 mg of tissue, divided into two technical replicates, were used. The purification was done as described in Kurbasic et al.  based on the original paper by Zhang et al. . The protein concentration was determined using the Micro Lowry assay kit Peterson’s Modification (Sigma-Aldrich, Stockholm, Sweden). In brief, the homogenized tissue was degraded with 10 μg/mL sequencing grade modified Trypsin (Promega, Madison, WI, USA) oxidized with sodium periodate (Serva, Heidelberg, Germany) to a final concentration of 8 mM and coupled to hydrazide Affi-gel (Bio-Rad, Hercules, CA, USA). Unbound peptides were washed off the hydrazide Affi-gel, where after the N-glycosylated peptides were cleaved off overnight by 5 U (1 U/μL) PNGase F (Roche Gmbh, Mannheim, Germany), dried on a SpeedVac (Thermo Scientific, Waltham, MA, USA), cleaned up by reverse-phase C18 chromatography (Waters, Milford, MA, USA) and dried again.
All solvents for high performance liquid chromatography (HPLC) were from Sigma-Aldrich and percentages are reported as (v/v). Samples were dissolved in 5 % acetonitrile (ACN) and 0.1 % formic acid (FA) for analysis on an linear trap quadropole (LTQ) Orbitrap XL mass spectrometer (Thermo Electron, Bremen, Germany). Peptide separation was carried out using an Eksigent 2D NanoLC system (Eksigent Technologies, Dublin, CA, USA) where the mobile phase A was water/0.1 % FA and the mobile phase B was ACN/0.1 % FA. Loading, washing and separation was performed as in Kurbasic et al. . Peptides were eluted from the analytical column using a linear gradient of mobile phase B developed from 3–35 % B during 60 min. The gradient was followed by 20 min column washing with 90 % ACN, 0.1 % FA and a 15 min re-equilibration with 3 % B. Peptides were analyzed using data-dependent acquisition, simultaneously scanning a mass range between 400 and 2000 Da in the Orbitrap and MS/MS spectra in the LTQ. Four MS/MS spectra were collected per second using collision-induced dissociation (CID) in the LTQ ion trap. The normalized collision energy was set to 35 %. Each Orbitrap MS scan was acquired at 60000 FWHM nominal resolution settings using the lock as mass option (m/z 445.120025) for internal calibration. A dynamic exclusion list restricted to 500 m/z values was used for 2 min with a repeat count of 2. Data was acquired using Xcalibur software, version 2.0.7 (Thermo Fischer Scientific, Hägersten, Sweden).
Each mouse model was examined as an individual project and all data comparisons were made within the project that the samples belonged to. Raw mass spectrometric data were analyzed with the in-house developed software Proteios SE (version 2.19.0)  and also independently with Progenesis QI v. 1.0.5156.29278 (Nonlinear Dynamics, Waters). In Progenesis, raw data was automatically aligned, manually inspected and adjusted. There is a built-in quality assessment in Progenesis where a 3-step color code for the spectrum alignment is used (green = good, beige = ok and red = needs revision). For both the STZ and db/db datasets, alignments were in good or ok agreement (green or beige). All detected features were exported to Mascot (MatrixScience, London, UK) version 2.4.1 and searched against the UniprotKB mouse 2015.08 database with equal number of reversed sequences. The parent ion mass tolerance was set to 5 ppm and to 0.8 Da for the fragment ions and the MS/MS ion charge was set to 2+/3+/4+. One missed protease cleavage was allowed. Cys carbamidomethylation was set as fixed modification and Met oxidation and Asn to Asp deamidation as variable modification (gain of 0.984 Da) as a consequence of PNGase F cleavage. Peptide identifications were propagated between runs and both peptides and proteins were filtered at 5 % FDR. Peptides with reverse sequences and no Asn to Asp modification were removed. Only proteins with a Mascot score > 25 and at least one unique peptide were kept. Protein grouping was employed in Progenesis. No missing value imputation was done. The results from Proteios and Progenesis were comparable.
The mass spectrometry data have been deposited to the ProteomeXchange Consortium [43, 44] via the PRIDE partner repository with the dataset identifier PXD003196 and 10.6019/PXD003196 for the STZ project and PXD003349 and 10.6019/PXD003349 for the db/db project. A guide to the file names is listed in an Additional file 1f. After peptide identification and protein assembly in Progenesis, the processed non-normalized data intensities from both mouse models were run through the software Normalyzer 1.1.1 [45, 46]. Normalyzer assesses the optimal normalization method for omics data sets compared to log2 transformation and the data sets are validated from both quantitative and qualitative perspectives, since the optimal normalization method is dependent on the intrinsic characteristics of the data set [45, 46]. Both the STZ and db/db data sets were normalized using Loess-G . Mean values were used for the statistical analysis of technical replicates. Of the 153 proteins identified in both mouse models, peptides with the highest scores were manually examined for the NXS/T motif.
Multivariate data analysis
Unsupervised principal component analysis (PCA) and supervised orthogonal partial least square discriminant analysis (OPLS-DA) were conducted in SIMCA v. 14 (Umetrics, Umeå, Sweden) with univariate scaling. Both analysis methods reduce the dimensionality in the dataset by introducing components that describe the joint behavior of several variables. The major difference between the unsupervised and the supervised methods is that in the PCA the first component is the one describing the largest co-variance within the dataset, regardless of direction  and in the OPLS-DA, the first component is class dependent and describes the major difference between the included groups, separating them on the x-axis . In the PCA, the second component is composed of the variables with the largest orthogonal co-variance to the first one . In the OPLS-DA analysis, the orthogonal difference takes place on the y-axis, identifying differences within the groups [49, 50]. The PCA and OPLS model statistics R2X(cum), R2Y(cum), Q2(cum)  and PCA components are summarized in an Additional file 2: Panel A.
In this study, OPLS-DA was used to compare the N-glycosylated proteomes of the diabetic to the healthy mice within the STZ and db/db mouse models and to compare the db/db insulin to the db/db vehicle group. The most important protein variables for the separation of the mouse groups in the OPLS-DA can be seen in the Variable Importance for the Projection (VIP)-plot. Values above 1 indicate importance for the OPLS model. The VIP plot of the STZ mouse model was compared to the VIP plot of the db/db mouse model not including the db/db insulin group. Proteins identified both in the STZ and db/db mice with a VIP score above 1 were used in the comparison of the two mouse models. Proteins separating the db/db insulin from the db/db vehicle group were also examined. An overview of all the models is shown in Additional file 2: Panel A.
Hierarchical clustering of both the variables and mouse samples was done in Qlucore v. 3.2 (Qlucore AB, Lund, Sweden) where the cut-off level was set to P < 0.05. Proteins with significant expression were evaluated using univariate analyses.
Statistics for univariate data
For univariate comparisons between the two groups in the STZ mouse model, two-tailed Student’s t test for equal standard deviations (SD) was performed in GraphPad Prism 6 (La Jolla, CA, USA), which simultaneously calculates the F-test for equal SD. For multiple group comparisons within the db/db mouse model one-way ANOVA and Tukey post hoc test was used simultaneously with Brown-Forsythe (BF) test to compare intra group variance. In all tests P < 0.05 was considered significant. In the text, mouse parameters and protein data are presented as mean ± SD and in the graphs mouse parameters are presented with mean ± standard error of mean (SEM) and protein data with means and 95 % confidence intervals (CI). Peptide data for the 153 shared proteins are listed in Additional files 1a–b and protein mean intensities and SD for the proteins listed in Tables 2, 4 and Additional file 1d are reported in Additional file 1c. All calculations were done in GraphPad PRISM 6, and AmberBio (Amber BioScience, Lund, Sweden).
Protein network analysis was performed with Ingenuity Pathway Analysis (IPA) (Ingenuity Systems, Mountain View, CA, USA). In IPA, individual protein relationships obtained from the literature are gathered into hypothetical interaction networks [52, 53]. The MS obtained normalized intensities were converted to fold change for all proteins comparing the diabetic to the healthy control mice. The two mouse models were entered separately into IPA and core analysis was performed for kidney tissue in mice, allowing the maximum default of 35 proteins per network. Scores of 2 or above have 99 % confidence or more of not being generated by chance. The acquired top rated networks and canonical pathways were compared between the two mouse models.
albumin excretion rate
leptin receptor deficient
glycated hemoglobin A1c
liquid chromatography tandem mass spectrometry
orthogonal partial least square discriminant analysis
principle component analysis
type 1 or 2 diabetes mellitus
LL performed all experimental work, data analysis and drafted the manuscript. LL, PJ and JM contributed to the overall study design. MP and JN designed the animal study, collected the preclinical data and MP edited the manuscript. All authors have read and approved the final manuscript.
We thank Bidda Rolin and Lisbeth Fink for providing the possibility to carry out this collaboration, Helene Dyhr, Jeanett Raun Lott and Andreas Engler for carrying out the in vivo part of the experiments, Karin Hansson for running the Orbitrap LTQ MS, Sofia Waldemarson and Céline Fernandez for contributions to the discussion, Aakash Chawade for help on dataset normalization and Fredrik Levander for assistance on data processing. LL has received research funding from the Nordforsk foundation and PJ from Vetenskåpsrådet, Sweden.
The authors declare that they have no competing interests.
Availability of data and materials
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium [43, 44] via the PRIDE partner repository (http://proteomecentral.proteomexchange.org/) as two separate datasets: The “N-glyco STZ mouse kidney” project has accession PXD003196 and project DOI 10.6019/PXD003196 and the “N-glyco db/db mouse kidney” project has accession PXD003349 and project DOI 10.6019/PXD003349.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Ma RCW, Chan JCN. Diabetes: incidence of childhood type 1 diabetes: a worrying trend. Nat Rev Endocrinol. 2009;5:529–30.View ArticlePubMedGoogle Scholar
- Maahs DM, West NA, Lawrence JM, Mayer-Davis EJ. Epidemiology of type 1 diabetes. Endocrinol Metab Clin North Am. 2010;39:481–97.View ArticlePubMedPubMed CentralGoogle Scholar
- World Health Organization. Global status report on noncommunicable diseases 2014. Geneva: WHO; 2014.Google Scholar
- Dalla Vestra M, Saller A, Bortoloso E, Mauer M, Fioretto P. Structural involvement in type 1 and type 2 diabetic nephropathy. Diabetes Metab. 2000;26(Suppl 4):8–14.PubMedGoogle Scholar
- Osterby R, Gall MA, Schmitz A, Nielsen FS, Nyberg G, Parving HH. Glomerular structure and function in proteinuric type 2 (non-insulin-dependent) diabetic patients. Diabetologia. 1993;36:1064–70.View ArticlePubMedGoogle Scholar
- Osterby R. Glomerular structural changes in type 1 (insulin-dependent) diabetes mellitus: causes, consequences, and prevention. Diabetologia. 1992;35:803–12.View ArticlePubMedGoogle Scholar
- Kolset SO, Reinholt FP, Jenssen T. Diabetic nephropathy and extracellular matrix. J Histochem Cytochem. 2012;60:976–86.View ArticlePubMedPubMed CentralGoogle Scholar
- Scott RP, Quaggin SE. The cell biology of renal filtration. J Cell Biol. 2015;209:199–210.View ArticlePubMedPubMed CentralGoogle Scholar
- Hynes RO. Integrins: bidirectional, allosteric signaling machines. Cell. 2002;110:673–87.View ArticlePubMedGoogle Scholar
- Schwarz F, Aebi M. Mechanisms and principles of N-linked protein glycosylation. Curr Opin Struct Biol. 2011;21:576–82.View ArticlePubMedGoogle Scholar
- Rahbar S. The discovery of glycated hemoglobin: a major event in the study of nonenzymatic chemistry in biological systems. Ann N Y Acad Sci. 2005;1043:9–19.View ArticlePubMedGoogle Scholar
- Deeb SJ, Cox J, Schmidt-Supprian M, Mann M. N-linked glycosylation enrichment for in-depth cell surface proteomics of diffuse large B-cell lymphoma subtypes. Mol Cell Proteomics. 2014;13:240–51.View ArticlePubMedGoogle Scholar
- Kurbasic E, Sjöström M, Krogh M, Folkesson E, Grabau D, Hansson K, Rydén L, Waldemarson S, James P, Niméus E. Changes in glycoprotein expression between primary breast tumour and synchronous lymph node metastases or asynchronous distant metastases. Clin Proteomics. 2015;12:13.View ArticlePubMedPubMed CentralGoogle Scholar
- Pan S, Chen R, Tamura Y, Crispin DA, Lai LA, May DH, McIntosh MW, Goodlett DR, Brentnall TA. Quantitative glycoproteomics analysis reveals changes in N-glycosylation level associated with pancreatic ductal adenocarcinoma. J Proteome Res. 2014;13:1293–306.View ArticlePubMedPubMed CentralGoogle Scholar
- Bausch-Fluck D, Hofmann A, Bock T, et al. A mass spectrometric-derived cell surface protein atlas. PLoS ONE. 2015;10:e0121314.View ArticlePubMedPubMed CentralGoogle Scholar
- Gahmberg CG, Tolvanen M. Why mammalian cell surface proteins are glycoproteins. Trends Biochem Sci. 1996;21:308–11.View ArticlePubMedGoogle Scholar
- Apweiler R, Hermjakob H, Sharon N. On the frequency of protein glycosylation, as dedu…: full Text Finder. Biochim Biophys Acta. 1999;1473:4–8.View ArticlePubMedGoogle Scholar
- Zhang H, Li X-J, Martin DB, Aebersold R. Identification and quantification of N-linked glycoproteins using hydrazide chemistry, stable isotope labeling and mass spectrometry. Nat Biotechnol. 2003;21:660–6.View ArticlePubMedGoogle Scholar
- Breyer MD, Böttinger E, Brosius FC, Coffman TM, Harris RC, Heilig CW, Sharma K. Mouse models of diabetic nephropathy. J Am Soc Nephrol. 2005;16:27–45.View ArticlePubMedGoogle Scholar
- Giacco F, Brownlee M. Oxidative stress and diabetic complications. Circ Res. 2010;107:1058–70.View ArticlePubMedPubMed CentralGoogle Scholar
- Giacco F, Du X, D’Agati VD, Milne R, Sui G, Geoffrion M, Brownlee M. Knockdown of glyoxalase 1 mimics diabetic nephropathy in nondiabetic mice. Diabetes. 2014;63:291–9.View ArticlePubMedGoogle Scholar
- Brownlee M. The pathobiology of diabetic complications: a unifying mechanism. Diabetes. 2005;54:1615–25.View ArticlePubMedGoogle Scholar
- Aronson D. Hyperglycemia and the pathobiology of diabetic complications. Adv Cardiol. 2008;45:1–16.View ArticlePubMedGoogle Scholar
- Fonseca VA, Stratton I, Adler A, et al. Defining and characterizing the progression of type 2 diabetes. Diabetes Care. 2009;32(Suppl 2):S151–6.View ArticlePubMedPubMed CentralGoogle Scholar
- Ravidà A, Musante L, Kreivi M, Miinalainen I, Byrne B, Saraswat M, Henry M, Meleady P, Clynes M, Holthofer H. Glycosylation patterns of kidney proteins differ in rat diabetic nephropathy. Kidney Int. 2015;87:963–74.View ArticlePubMedGoogle Scholar
- Welsh GI, Hale LJ, Eremina V, et al. Insulin signaling to the glomerular podocyte is critical for normal kidney function. Cell Metab. 2010;12:329–40.View ArticlePubMedPubMed CentralGoogle Scholar
- Coward R, Fornoni A. Insulin signaling: implications for podocyte biology in diabetic kidney disease. Curr Opin Nephrol Hypertens. 2015;24:104–10.View ArticlePubMedPubMed CentralGoogle Scholar
- Uhlen M, Fagerberg L, Hallstrom BM, et al. Tissue-based map of the human proteome. Science. 2015;347:1260419.View ArticlePubMedGoogle Scholar
- Kussman MJ. The clinical course of diabetic nephropathy. J Am Med Assoc (JAMA). 1976;236:1861.View ArticleGoogle Scholar
- Varki A, Esko JD, Colley KJ. Chapter 3, Cellular organization of glycosylation. Cold Spring: Cold Spring Harbor Laboratory Press; 2009.Google Scholar
- Harada N, Inagaki N. Role of sodium-glucose transporters in glucose uptake of the intestine and kidney. J Diabetes Investig. 2012;3:352–3.View ArticlePubMedGoogle Scholar
- Mohammad G, Vandooren J, Siddiquei MM, Martens E, Abu El-Asrar AM, Opdenakker G. Functional links between gelatinase B/matrix metalloproteinase-9 and prominin-1/CD133 in diabetic retinal vasculopathy and neuropathy. Prog Retin Eye Res. 2014;43:76–91.View ArticlePubMedGoogle Scholar
- De Carvalho JAM, Piva SJ, Hausen BS, Bochi GV, Kaefer M, Coelho AC, Duarte MMMF, Moresco RN. Assessment of urinary γ-glutamyltransferase and alkaline phosphatase for diagnosis of diabetic nephropathy. Clin Chim Acta. 2011;412:1407–11.View ArticlePubMedGoogle Scholar
- Schainuck LI, Striker GE, Cutler RE, Benditt EP. Structural-functional correlations in renal disease. II. The correlations. Hum Pathol. 1970;1:631–41.View ArticlePubMedGoogle Scholar
- Thomas MC, Burns WC, Cooper ME. Tubular changes in early diabetic nephropathy. Adv Chronic Kidney Dis. 2005;12:177–86.View ArticlePubMedGoogle Scholar
- Kreidberg J, Donovan M, Goldstein S, Rennke H, Shepherd K, Jones R, Jaenisch R. Alpha 3 beta 1 integrin has a crucial role in kidney and lung organogenesis. Development. 1996;122:3537–47.PubMedGoogle Scholar
- Ning L, Kurihara H, de Vega S, Ichikawa-Tomikawa N, Xu Z, Nonaka R, Kazuno S, Yamada Y, Miner JH, Arikawa-Hirasawa E. Laminin α1 regulates age-related mesangial cell proliferation and mesangial matrix accumulation through the TGF-β pathway. Am J Pathol. 2014;184:1683–94.View ArticlePubMedPubMed CentralGoogle Scholar
- Kajiho Y, Harita Y, Kurihara H, et al. SIRPα interacts with nephrin at the podocyte slit diaphragm. FEBS J. 2012;279:3010–21.View ArticlePubMedGoogle Scholar
- Takahashi S, Tomioka M, Hiromura K, et al. SIRPα signaling regulates podocyte structure and function. Am J Physiol Renal Physiol. 2013;305:F861–70.View ArticlePubMedGoogle Scholar
- Nie T, Hui X, Gao X, et al. Adipose tissue deletion of Gpr116 impairs insulin sensitivity through modulation of adipose function. FEBS Lett. 2012;586:3618–25.View ArticlePubMedGoogle Scholar
- Svensson M, Boren M, Sköld K, Fälth M, Sjögren B, Andersson M, Svenningsson P, Andren PE. Heat stabilization of the tissue proteome: a new technology for improved proteomics. J Proteome Res. 2009;8:974–81.View ArticlePubMedGoogle Scholar
- Häkkinen J, Vincic G, Månsson O, Wårell K, Levander F. The proteios software environment: an extensible multiuser platform for management and analysis of proteomics data. J Proteome Res. 2009;8:3037–43.View ArticlePubMedGoogle Scholar
- Vizcaíno JA, Deutsch EW, Wang R, et al. ProteomeXchange provides globally coordinated proteomics data submission and dissemination. Nat Biotechnol. 2014;32:223–6.View ArticlePubMedPubMed CentralGoogle Scholar
- Vizcaíno JA, Côté RG, Csordas A, et al. The PRoteomics IDEntifications (PRIDE) database and associated tools: status in 2013. Nucleic Acids Res. 2013;41:D1063–9.View ArticlePubMedGoogle Scholar
- Chawade A, Alexandersson E, Levander F. Normalyzer: a tool for rapid evaluation of normalization methods for omics data sets. J Proteome Res. 2014;13:3114–20.View ArticlePubMedPubMed CentralGoogle Scholar
- Chawade A, Sandin M, Teleman J, Malmström J, Levander F. Data processing has major impact on the outcome of quantitative label-free LC-MS analysis. J Proteome Res. 2015;14:676–87.View ArticlePubMedGoogle Scholar
- Smyth GK. Limma: Linear models for microarray data. In: Gentleman R, Carey VJ, Huber W, Irizarry RA, Dudoit S (eds) Bioinformatics and computational biology solutions using R and bioconductor. New York: Springer; 2005. p. 397–420.
- Ringnér M. What is principal component analysis? Nat Biotechnol. 2008;26:303–4.View ArticlePubMedGoogle Scholar
- Bylesjö M, Rantalainen M, Cloarec O, Nicholson JK, Holmes E, Trygg J. OPLS discriminant analysis: combining the strengths of PLS-DA and SIMCA classification. J Chemom. 2006;20:341–51.View ArticleGoogle Scholar
- Wiklund S, Johansson E, Sjöström L, Mellerowicz EJ, Edlund U, Shockcor JP, Gottfries J, Moritz T, Trygg J. Visualization of GC/TOF-MS-based metabolomics data for identification of biochemically interesting compounds using OPLS class models. Anal Chem. 2008;80:115–22.View ArticlePubMedGoogle Scholar
- Wiklund S. Multivariate analysis in Omics. Umetrics. Umeå: Sweden metabolomics.se; 2008. p. 1–228.Google Scholar
- Krämer A, Green J, Pollard J, Tugendreich S. Causal analysis approaches in ingenuity pathway analysis. Bioinformatics. 2014;30:523–30.View ArticlePubMedGoogle Scholar
- Overgaard AJ, Thingholm TE, Larsen MR, Tarnow L, Rossing P, McGuire JN, Pociot F. Quantitative iTRAQ-based proteomic identification of candidate biomarkers for diabetic nephropathy in plasma of type 1 diabetic patients. Clin Proteomics. 2010;6:105–14.View ArticlePubMedPubMed CentralGoogle Scholar