Metabolic labeling unveils alterations in the turnover of HDL-associated proteins during diabetes progression in mice

Abstract

Several aspects of diabetes pathophysiology and complications result from hyperglycemia-induced alterations in the structure and function of plasma proteins. Furthermore, insulin has a significant influence on protein metabolism by affecting both the synthesis and degradation of proteins in various tissues. To understand the role of progressive hyperglycemia on plasma proteins, in this study, we measured the turnover rates of high-density lipoprotein (HDL)-associated proteins in control (chow diet), prediabetic [a high-fat diet (HFD) for 8 wk] or diabetic [HFD for 8 wk with low-dose streptozotocin (HFD + STZ) in weeks 5–8 of HFD] C57BL/6J mice using heavy water (²H₂O)-based metabolic labeling approach. Compared with control mice, HFD and HFD + STZ mice showed elevations of fasting plasma glucose levels in the prediabetic and diabetic range, respectively. Furthermore, the HFD and HFD + STZ mice showed increased hepatic triglyceride (TG) levels, total plasma cholesterol, and plasma TGs. The kinetics of 40 proteins were quantified using the proteome dynamics method, which revealed an increase in the fractional synthesis rate (FSR) of HDL-associated proteins in the prediabetic mice compared with control mice, and a decrease in FSR in the diabetic mice. The pathway analysis revealed that proteins with altered turnover rates were involved in acute-phase response, lipid metabolism, and coagulation. In conclusion, prediabetes and diabetes have distinct effects on the turnover rates of HDL proteins. These findings suggest that an early dysregulation of the HDL proteome dynamics can provide mechanistic insights into the changes in protein levels in these conditions.

NEW & NOTEWORTHY This study is the first to examine the role of gradual hyperglycemia during diabetes disease progression on HDL-associated protein dynamics in the prediabetes and diabetic mice. Our results show that the fractional synthesis rate of HDL-associated proteins increased in the prediabetic mice whereas it decreased in the diabetic mice compared with control mice. These kinetic changes can help to elucidate the mechanism of altered protein levels and HDL dysfunction during diabetes disease progression.

INTRODUCTION

The incidence of diabetes mellitus (DM) continues to be a widespread health concern. It is estimated that 34.2 million or 10.5% of the United States population has diabetes (1). Worldwide, the diabetic population was estimated at 463 million in 2019, with an anticipated increase to 700 million by 2045 due to the obesity epidemic (2). DM is a multisystem disease affecting various organs and systems. Type 2 diabetes and associated insulin resistance have a reciprocal relationship with nonalcoholic fatty liver disease (NAFLD) and other conditions associated with obesity-induced metabolic syndrome (3). Interestingly, both DM and NAFLD increase the risk of cardiovascular disease (CVD) (4). CVD accounts for significant mortality associated with DM (5).

Prediabetes is recognized as a condition that often progresses to DM and warrants clinical vigilance. The prediabetic state can persist for several years before the onset of diabetes (6). Although the glycemic abnormalities of prediabetes [HbA1_c 5.7%–6.4%, 2 h glucose of 141–199 mg/dL in an oral-glucose tolerance test (OGTT), and/or fasting blood glucose (FPG) of 100–126 mg/dL] are of lower severity compared with diabetes (HbA1_c 6.5%–11%, 2 h glucose ≥200 mg/dL, and/or fasting blood glucose ≥126 mg/dL), they can still lead to deleterious consequences. Among these is the observation that liver enzymes are elevated in patients with prediabetes (7). Recently, a population-based study using magnetic resonance imaging identified a strong association between prediabetes and hepatic steatosis (8).

The liver is a key source of plasma proteins with >80% of proteins synthesized in the liver exported into the systemic circulation (9). These include proteins such as those involved in inflammation, coagulation, lipid homeostasis, hormone transport, and acute-phase reactions, among others. The manifestation of diabetes pathophysiology in the liver can potentially modulate hepatic proteostasis and can have a profound systemic impact. Although type 2 diabetes is not characterized by the catabolic state typically seen in type 1 diabetes patients, several chronic complications of diabetes are a direct result of alteration in protein structure and function (10). The onset of liver abnormalities in prediabetic state suggests that liver protein synthesis and degradation processes could be dysregulated early in the pathophysiology of diabetes.

Identification of a reliable plasma protein signature would greatly assist in diagnosing the transition from normal to prediabetes and prediabetes to diabetes in addition to helping guide the treatment and management of this chronic metabolic disorder. Conventional proteomics has been used to characterize diabetes-related changes in plasma proteome composition and identify potential protein biomarkers of CVD. Recently, more comprehensive plasma proteome profiling in patients with diabetes was achieved using a SOMAscan technology with the detection reagent for each individual protein referred to as SOMAmers selected based on chemically modified DNA libraries (11, 12). These methods have been useful in identifying protein biomarker signatures and proteins playing a causal role in diabetes. However, the static measurements of protein levels offer no information on the cause of altered levels and are less sensitive to small changes in protein homeostasis.

Importantly, insulin is a key hormonal regulator of proteostasis. Insulin regulates both protein synthesis and protein degradation in skeletal muscles as well as the splanchnic tissues (13). However, the effects of insulin resistance or deficiency during diabetes disease progression on the homeostasis of liver-secreted proteins are poorly understood. Several animal models have been used to model diabetes including both genetic and nongenetic models. Mice fed a fat-enriched diet (≥ 60% of total calories) for 12 wk or longer develop an obese phenotype with mild hyperglycemia and underlying insulin resistance, a hallmark of prediabetes. The administration of streptozotocin (STZ), a β-cell toxin, results in the destruction of the pancreas and is frequently used to model type 1 diabetes. However, it has been shown that the combination of a high-fat diet (HFD) with frequent low-dose STZ injections can generate animal models of type 2 diabetes characterized by severe hyperglycemia, insulin resistance, obesity, and pancreatic β-cell dysfunction, all features of the human clinical diabetic disease (14, 15). In this study, we used a heavy (deuterated) water (²H₂O)-based metabolic labeling approach to study the effect of gradual hyperglycemia on plasma proteome dynamics in mice by examining the turnover rates of high-density lipoprotein (HDL)-associated proteins. We used 8 wk of HFD feeding coupled with the final 4 wk of the vehicle or low-dose STZ administration to model the prediabetic and diabetic states in mice, respectively. We hypothesized that the pathophysiological changes associated with diabetes disease progression influence the turnover rates of HDL-associated proteins.

METHODS

Materials

High-performance liquid chromatography (HPLC) grade solvents for nanospray chromatography and sample preparation were purchased from Fluka (Milwaukee, MO). All other reagents were purchased from Sigma-Aldrich.

Animals.

All animal procedures were approved by the Institutional Animal Care and Use Committee at the Northeast Ohio Medical University and were performed in accordance with National Institutes of Health (NIH) guidelines. Our study design combined the use of nongenetic mouse models of prediabetes and diabetes (14) with an HDL proteome dynamics approach to enable the evaluation of the effect of continuous hyperglycemic changes on plasma protein dynamics (Fig. 1).

Figure 1.

Experimental design and the workflow of the study. Mice fed either a chow diet (n = 6 mice) or a high-fat diet (HFD; n = 10 mice). After 4 wk, HFD-fed mice were randomized into two groups. The first group continued their HFD for an additional 4 wk to model prediabetes (n = 4/group). The second group of mice on HFD was treated weekly with a low dose of streptozotocin (HFD+STZ) to mimic type 2 diabetes in humans (n = 6/group). Prediabetic, diabetic, and control mice (control) received 30 μL/g bolus dose of ²H₂O followed by 8% ²H₂O in their drinking water during the last 7 days of the study. Blood samples were drawn at different time points during 1 wk of the study. HDL proteins were isolated and analyzed by LC-MS/MS after delipidation and trypsin-digestion. High-resolution mass spectra were collected and processed to 1) identify proteins and their posttranslational modifications and 2) assess protein fractional synthesis rates (k) based on ²H-incorporation.

Six-eight-week-old male C57BL/6J mice were fed a regular chow diet (Control) or HFD (D12492, Research Diets, New Brunswick, NJ). After 4 wk, mice on HFD were randomized into two groups, and then they were weekly injected intraperitoneally with a low dose of STZ (90–100 mg/kg body wt, diabetic group, n = 6) or vehicle (0.05 M citric acid, pH 4.5, prediabetic group, n = 4) for the following 4 wk. Control mice (n = 6) continued their chow diet for the remaining 4 wk. Postprandial blood glucose and body weights were measured weekly. The weight gain in each group was computed as the difference in average weights at the end of the study period and the initial average weight for each group. The insulin sensitivity was assessed based on an insulin tolerance test (ITT) in nonfasted animals 3 days before the termination of the experiment. A glucose tolerance test (GTT) was performed in fasted control and prediabetic (HFD mice) animals on the last day of the experiment. Mice with fasting serum blood glucose >250 mg/dL were classified as diabetic.

After the prediabetic and diabetic phenotypes were established, the kinetic study was performed on individual animals from each group. This study design minimizes the biological variability through the collection of small blood samples from the same animal at selected time points (total ∼300–350 μL of blood during the 7 days of tracer study).

The ²H₂O-based kinetic study was performed in age-matched diabetic, prediabetic, and nondiabetic control mice during the last week of the study, as we described previously (16). Briefly, a bolus dose of ²H₂O (30 μL of 99.9% ²H-labeled saline per gram body weight) was followed by free access to drinking water enriched with 8% ²H₂O. Small blood samples (∼60 µL) were collected at 4, 8, 24, and 48 h. After 7 days of ²H₂O exposure, the terminal blood sample was collected through the cardiac puncture. Plasma was separated immediately and saved at −80°C until the analyses.

Analytical procedures.

Metabolic measurements.

Blood glucose was measured using a glucometer. Fasting triglycerides (TGs) and total cholesterol (TC) were measured using triglyceride and cholesterol assay kits (Infinity, Thermo Fisher Scientific, Cat. No. TR13421 and Cat. No. TR22421, respectively). HDL cholesterol (HDL-C) was measured after the precipitation of APOB-containing particles using a commercially available kit (Stanbio Laboratory, Boerne, TX). The low-density lipoprotein cholesterol (LDL-C) was calculated based on TC, HDL-C, and TG levels using the traditional Friedewald equation (17).

GTT and ITT.

Glucose tolerance was measured in 6-h fasted control and prediabetic mice. Briefly, after the baseline blood sampling from the tail vein, a bolus of glucose (20% w/w, 1 mg/g body wt) was injected intraperitoneally. Blood samples were taken at 15, 30, 60, and 120 min to measure glucose levels to assess glucose tolerance.

The ITT was performed in fed prediabetic and diabetic mice. After taking the baseline glucose measurements, human insulin (Humulin, Lilly; 0.75 units/kg body wt) was administered by intraperitoneal injection and blood glucose was measured every 15 min for 60 min. Glucose utilization and insulin tolerance were determined based on the areas under the blood glucose response curves.

Measurement of body water deuterium enrichment.

The enrichment of plasma with deuterated water (²H₂O) was measured using the acetone exchange method (18). This method is based on the isotopic exchange between the hydrogens of water and acetone in an alkaline medium (pH 12–13) and headspace analysis of acetone by gas chromatography-mass spectrometry (GC-MS). See the online supplement for the details of the assay (all Supplemental material is available at https://doi.org/10.6084/m9.figshare.20231991).

Proteomics sample preparation.

HDL-associated proteins in the plasma were analyzed after the depletion of APOB-containing particles. APOB depletion with magnesium chloride/dextran sulfate reagent (Stanbio Laboratory, Boerne, TX) allows reproducible quantification of HDL proteins (16, 19). In addition, this approach enables simultaneous quantification of HDL-C, a known risk factor of CVD. Briefly, after the precipitation of APOB-containing lipoproteins from 30 μL of plasma, the supernatant was recovered and used for the proteomic analysis. Proteins were precipitated with 1 mL of cold acetone (−20°C) and the pellets were dried. The residue was then dissolved in 200 μL of 50 mM ammonium bicarbonate solution containing 1% sodium deoxycholate. BCA assay was performed, and 10 µg of proteins were taken. The thiol groups of proteins were reduced with the excess of dithiothreitol (DTT) solution at 45°C for 30 min and then alkylated with iodoacetamide in the dark at room temperature for 60 min. Proteins were digested overnight in solution with an excess of Promega sequencing grade trypsin (10 μL of 0.1 μg/μL trypsin solution in 100 mM pH 8 ammonium bicarbonate) at 37°C. Samples were desalted by solid-phase extraction using the OMIX C18 pipette tips (Agilent, Cat. No. A5700310). The peptides were eluted with 70% acetonitrile/0.1% trifluoroacetic acid solution. Dried samples were reconstituted in 30 μL 2% acetonitrile/0.1% formic acid solution, and 5 μL of the solution was injected for the LC-MS analyses.

LC-MS/MS analysis.

The tryptic peptides were analyzed by Ultimate 3000 UHPLC (Thermo Fisher Scientific, CA) coupled online to Q Exactive Plus Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific) as described. Briefly, peptides were introduced to an Acclaim PepMap RSLC reverse-phase nanocolumn (75 μm × 15 cm, C18, 2 μm, 100 Å, Thermo Fisher Scientific) using mobile phase A (0.1% formic acid in water) and B (20% water in acetonitrile with 0.1% formic acid) at a flow rate of 300 nL/min. Peptides were fractionated using a stepwise gradient with an initial 2% of mobile phase B, which was linearly increased to 40% in 100 min, ramped to 90% in 5 min, and then held at 90% B for 10 min. At the end of the run, the column was equilibrated for 13 min with 2% of phase B.

Mass spectrometry analysis was performed in data-dependent acquisition mode with a full profile MS scan at 70,000 resolution (200 m/z) between 380 and 1,300 m/z. MS/MS spectra were collected for the 12 most abundant product ions with an isolation window of 1.8 and offset of 0.2 m/z and 17,500 resolution (200 m/z) at a normalized collision energy of 25%. The precursor ion masses were dynamically excluded from MS/MS analyses for a duration of 17 s. Ions with the charge state 1 and greater than 6 were excluded from MS/MS analyses. MS and MS/MS spectra were acquired for 100 ms with the automatic gain control (AGC) target set at 1.0 × 10⁶ and 2.0 × 10⁴ ions for MS and MS/MS scans, respectively.

Protein identification.

To identify proteins, all the MS/MS spectra were transformed into MZML file format using ProteoWizard MSconvert Version 3.0.18116 (https://www.proteomesoftware.com/proteowizard/windows-3.0.21229-x64). The MZML files were searched using Mascot software version 2.3 (Matrix Science, London, UK) against the National Center for Biotechnology Information mouse subset of the SwissProt protein database released in October 2016 (containing 552,884 entries). The search was performed using cysteine carbamidomethylation as the fixed modification and methionine oxidation and lysine glycation as variable modifications. A maximum of two missed cleavages per peptide were allowed. The mass tolerance of the parent and product ions was set at 10 ppm and 0.1 Da, respectively. A Mascot score of greater than 35 was applied as the cutoff threshold along with 95% identification confidence. To minimize the false discovery rate (FDR), the Mascot search was performed with the decoy option. Blast search (http://blast.ncbi.nlm.nih.gov/Blast.cgi) was used to confirm the uniqueness of quantified peptides. Protein identifications were accepted if they could achieve greater than 99% probability, FDR of less than 1%, and contained at least two unique peptides.

Proteome dynamics analysis.

The turnover rates of HDL proteins were quantified using a ²H₂O-based metabolic labeling approach (20). According to this method, when body water is enriched with ²H₂O, ²H-incorporation during protein synthesis results in time-dependent change(s) of mass isotopomers of peptides. Mass isotopomer is defined as a family of isotopic isomers that have the same chemical composition but different heavy isotopic atoms. The monoisotopic peak (M₀) is composed of the light isotopes of each chemical element constituting the molecular formula of a peptide. Additional mass isotopomer peaks (M₁ − M_j) result from the probabilistic combinations of heavy isotopes. ²H labeling of the peptides was extracted from high-resolution full scan (MS1) spectra (Supplemental Table S1) using a specialized software (21). The relative isotope abundance of the monoisotopic signal (I₀(t)) at each time point was calculated based on the intensity of all measured isotopomers:

$$I_0\left(t\right)=M_0(t)/\sum _{j=0}^n{M}_j\left(t\right),$$ (1)

where M_j is the intensity of measured individual isotopomers. The total heavy isotope labeling (²H plus baseline natural enrichment of peptides (I(t)) was calculated as:

$$I\left(t\right)=1-I_0\left(t\right)=\sum _{j=1}^n{M}_j\left(t\right)/\sum _{j=0}^n{M}_j\left(t\right).$$ (2)

The turnover rate constants (k, pool/day) for tryptic peptides were determined using one-compartmental modeling of total labeling values. For this purpose, the data were fitted into the exponential curve using the Prism (GraphPad, La Jolla, CA) software (v. 5):

$$I\left(t\right)=I_0+\left[I_{\text{plateau}}-I_0\right]\times \left(1-e^{-\mathit{kt}}\right),$$ (3)

where I₀ and I_plateau are the baseline and plateau enrichments of a peptide. The turnover rates of quantified peptides were aggregated to derive the averaged turnover rate of a protein (Supplemental Table S2). To improve the accuracy of protein turnover measurements, the outlier peptides were eliminated from the analysis if the coefficient of variation of a rate constant was higher than 25%.

The fraction of newly synthesized protein molecules was calculated as the ratio of the observed ²H labeling to the predicted plateau labeling of the tryptic peptide in the same experiment (22).

Data Presentation and Statistical Analysis

In all graphs, error bars represent standard deviations based on biological variability between individual animals in each group. For each animal, the aggregated turnover rates were derived from the average rate constants of protein turnover based on multiple unique peptides of each protein. The statistical significance of differences between the biochemical and metabolic parameters of the three groups and the rate constants of the same protein in different groups was tested using a one-way ANOVA. Post hoc analysis was performed using Tukey’s test to determine the statistical significance of differences between a pair of group means. When protein turnover was quantified in only two groups, a Student t test with a two-tailed distribution was used to analyze the differences between groups. When the newly made fraction of individual proteins were compared based on a selected individual peptide (Fig. 3), the differences were evaluated using the extra sum of squares F test, which compares the means between two populations after regression analysis (23). Thus, an F test determines if a “group” of variables is jointly significant. P < 0.05 was considered statistically significant. Proteins with significantly different kinetics were identified with FDR < 0.05.

Figure 3.

The effect of high-fat diet (HFD) alone or in combination with streptozotocin (HFD+STZ) on the fractional synthesis of the selected HDL proteins assessed with the ²H₂0 metabolic-labeling technique. The newly made fractions of APOAI (A) and KNG1 (B) were calculated based on ²H incorporation into the peptides TQVQSVIDK and DIPVDSPEK, respectively. The fractional synthesis rate is denoted by the rate constant k. Data are presented as means ± SD. Statistical significance, *P < 0.05 compared with HFD.

Bioinformatics Analysis

To functionally characterize proteins with altered turnover rates, we focused on proteins that exhibited a statistically significant change in their turnover rates relative to either control and/or prediabetic groups. These proteins with significant changes in turnover rates were annotated with Gene Ontology (GO) Biological Process (BP) terms, to identify overrepresented terms, using The Database for Annotation, Visualization and Integrated Discovery (DAVID Knowledgebase v2022q2) (24), and mouse plasma proteome of 1461 proteins described by Stocks et al (25) as a background. To examine the effect of HFD (prediabetes) and HFD + STZ (diabetes) on HDL-associated proteins turnover, the turnover rates of proteins annotated with the top 3 GO BP terms, defined as those with FDR values < 0.05, were compared. Tukey’s test was used to determine the statistical significance of differences between pairs of group means. Protein-protein interaction networks of proteins with related functions were analyzed by (PPI) STRING (string-db.org). Default STRING parameters with coexpression and cooccurrence were used as sources for active interactions. The statistical assessment of network enrichment and significance was calculated by STRING taking into account probabilities from different types of evidence and correcting for probabilities of random observations (26), and pathways were ranked by the FDR-corrected P values generated from this test. The smaller P value means that it is less likely that the association is due to random chance.

RESULTS

Metabolic Parameters

Table 1 shows the observed weight and plasma lipid profile in the three groups of mice at the end of the study. Mice received either the chow diet (control), HFD with vehicle (HFD), or HFD with STZ (HFD + STZ). Although all three groups of mice gained weight, the highest weight gain was observed in the HFD group compared with their baseline weight. Although at the end of the study, the body weights of diabetic and control animals were similar, diabetic animals exhibited a higher increase in body weight during the 8-wk study. The HFD + STZ group exhibited a 10% weight loss after starting STZ injections, which is consistent with reported findings in the literature (27). The measured values of TC and TG and the calculated values of LDL-C were significantly higher in the HFD group versus the control group (TC: 202.48 ± 34.02 mg/dL vs. 80.39 ± 11.32 mg/dL, TG: 63.60 ± 0.04 mg/dL vs. 32.94 ± 9.75 mg/dL, and LDL-C: 151.21 ± 28.63 mg/dL vs. 42.81 ± 12.08 mg/dL).

Table 1.

Comparisons of the metabolic profile between the control, high-fat diet (HFD) and the high-fat diet and streptozotocin (HFD+STZ) groups

	Control	HFD	HFD + STZ	P Value
Body weight, g	25.2 ± 0.8	32.9 ± 1.5*	24.4 ± 1.0#	<0.0001
Increase in body weight, g	2.3 ± 0.9	10.45 ± 0.43*	4.6 ± 1.2*#	<0.0001
Plasma TG, mg/dL	32.94 ± 9.75	63.63 ± 10.04	163.03 ± 60.54*#	0.0002
Plasma TC, mg/dL	80.39 ± 11.32	202.48 ± 34.02*	224.70 ± 43.24*	<0.0001
Plasma HDL-C, mg/dL	30.99 ± 1.09	38.55 ± 4.93	39.11 ± 6.09*	0.0315
Plasma LDL-C, mg/dL (calculated)	42.81 ± 12.08	151.21 ± 28.63*	152.98 ± 35.24*	<0.0001
Hepatic TG, µg/mg liver tissue	11.32 ± 0.36	116.69 ± 51.41*	45.50 ± 22.53#	0.0004
Plasma TG/HDL-C ratio	0.52 ± 0.11	1.71 ± 0.33*	3.42 ± 0.67*#	<0.0001
Fasting plasma glucose, mg/dL	106.5 ± 18.40	218.5 ± 18.6*	490 ± 88.4*#	<0.0001

P values for comparing Control, HFD, and HFD+STZ groups were obtained using one-way analysis of variance (ANOVA). TC, total cholesterol; TG, triglycerides.

*P < 0.05 vs. control;

#P < 0.05 vs. HFD determined after post hoc Tukey’s multiple comparison test.

Compared with controls, the HDL-C levels increased significantly in the HFD + STZ group (39.11 ± 6.09 mg/dL vs. 30.99 ± 1.09 mg/dL). The HFD + STZ mice had a significant increase in plasma TGs compared with the HFD group (163.04 ± 60.54 mg/dL vs. 63.63 ± 10.04 mg/dL), whereas TC, HDL-C and LDL-C were not significantly different from the HFD group. The TG/HDL-C ratio was significantly higher in the HFD + STZ group (3.42 ± 0.67) as compared with the HFD (1.71 ± 0.33) and control groups (0.52 ± 0.11).

The fasting plasma glucose was elevated in the HFD and the HFD + STZ groups as compared with the control group and reached the reported ranges for prediabetes (200–249 mg/dL) and diabetes (>250 mg/dL) for rodents (28). The GTT revealed exaggerated glycemic excursion in the HFD mice compared with control suggesting glucose intolerance and a prediabetic state (Fig. 2). Blood glucose levels remained significantly elevated in the HFD + STZ mice 1 h after an intraperitoneal insulin injection (0.75 units/kg body weight) compared with HFD mice suggesting insulin resistance and the presence of a diabetic state in HFD + STZ mice (Fig. 2).

Figure 2.

Glucose and insulin tolerance tests. A and B: the time-course and the area under the curve (AUC) of the glucose tolerance test (GTT) in the mice after intraperitoneal injection of glucose following an 8-h fast. C and D: the time-course and the area under the curve (AUC) of the insulin tolerance test (ITT) in nonfasted mice after intraperitoneal injection of insulin (0.75 units/kg). Data are presented as means ± SD.

HDL Protein Turnover Measurements

Proteome analysis identified 207 proteins in at least one mouse from all three groups. Of those, 158 proteins were shared by all mice from each group. The majority of these proteins (129/158) have previously been characterized as components of the HDL proteome (19). The turnover rates of HDL proteins were traced with ²H₂O, which rapidly labels proteogenic amino acids (29). A bolus dose of pure ²H₂O with free access to drinking water enriched with 8% ²H₂O resulted in 4%–5% body water enrichment (Supplemental Fig. S1) without any adverse effects as determined based on food intake (data not shown) and body weight (Table 1). The time course ²H labeling of tryptic peptides was used to assess the turnover rates and fraction of newly synthesized HDL proteins. For example, Fig. 3 shows progressive increase in ²H labeling of TQVQSVIDK [apolipoprotein A1 (ApoAI)] and DIPVDSPELK (kininogen1, KNG1) peptides due to protein synthesis over 1 wk of treatment with ²H₂O. Although during the first 4 days of labeling almost the entire pools of APOAI and KNG1 turned over in prediabetic mice, only ∼70% pool of APOAI and KNG1 was renewed in the diabetic animals, reflecting reduced fractional synthesis rates (FSRs) of these proteins in a diabetic state (P < 0.05). Using this approach, the kinetics of 40 proteins were quantified in mice from at least two groups and 37 proteins were quantified in all three groups (Supplemental Table S3). Overall, HFD-induced prediabetes was associated with increased turnover rates of HDL-associated proteins compared with the control group (Fig. 4A and Supplemental Table S4). Remarkably, the turnover rates of the majority of analyzed HDL-associated proteins declined in diabetic (HFD + STZ) group (Fig. 4B and Supplemental Table S4). Thus, STZ treatment not only abolished HFD-induced protein synthesis rate (Fig. 4C and Supplemental Table S4), but it also further reduced the turnover rates of proteins relative to the controls (Fig. 4B). Out of 40 quantified proteins (Supplemental Table S3), 10 proteins experienced significant turnover rate change due to both HFD and HFD + STZ relative to controls (Supplemental Table S5). Additional eight proteins changed in both control versus HFD + STZ and HFD versus HFD + STZ. Other 10 proteins changed in both control versus HFD and HFD versus HFD + STZ that discriminated all three groups. The turnover rates of three proteins changed due to diabetes progression (HFD vs. HFD + STZ comparison). This shows that although HFD treatment affects turnover rates of 22 out of 40 quantified proteins, progression to diabetes through STZ treatment further expands the list of affected proteins by additional 11 (8 + 3) proteins, for a total of 33 proteins that experience significant turnover change in at least one of the three pairwise comparisons (Supplemental Fig. S2). A detailed description of these proteins with altered turnover rates is provided in the Supplemental Material (https://doi.org/10.6084/m9.figshare.20231991).

Figure 4.

Volcano plot summarizing prediabetes and diabetes-induced changes in HDL proteins turnover. A: prediabetes (HFD) vs. control mice. B: diabetes (HFD+STZ) vs. prediabetes (HFD). C: diabetes vs. control. Red dots represent proteins that changed significantly in turnover rate (adjusted P value < 0.05). Each dot represents a protein. HFD, high-fat diet; STZ, streptozotocin. Full protein names and accession numbers are provided in Supplementary Table S3.

To assess the overall effect of prediabetes and diabetes on HDL proteins’ turnover, we averaged the rate constants of all proteins in individual animals from each group and compared the turnover rates of quantified proteins in different groups using one-way ANOVA. This analysis revealed that overall turnover rates of proteins were increased in the prediabetic group (0.92 ± 0.41 pool/day in prediabetes vs. 0.78 ± 0.34 pool/day in control, P < 0.05), which declined in the diabetic group (0.51 ± 0.20 pool/day) relative to both prediabetic and control groups (P < 0.05; Fig. 5A). The turnover rates of proteins from prediabetic and diabetic groups were significantly associated (R² = 0.799, P < 0.001) with the descending unity line. The slope of the regression curve (0.43) of the turnover rates in diabetes as a function of values in prediabetes shows that the average turnover of plasma proteins in diabetic animals was reduced about twofolds (Fig. 5B).

Figure 5.

Comparison of proteome dynamics in control, HFD, and HFD+STZ groups. A: the average turnover rates for all proteins identified in each animal of the control, HFD, and HFD+STZ groups. Data are presented as means ± SD. The mean values indicate the average of total protein turnover for each group (Control: n = 6, HFD: n = 4, and HFD+STZ: n = 6). The P value indicating significant differences is based on each animal as biological replicate. B: the correlation between the turnover rates observed for proteins identified in the HFD and HFD+STZ groups. The equation for the regression curve fitting is presented and indicates a >50% decrease in the turnover of proteins in the HFD+STZ group than the HFD group. HFD, high-fat diet; STZ, streptozotocin.

Pathways Associated with the Prediabetes- and Diabetes-Induced Alterations in HDL Proteins Turnover

To understand the biological significance of continuous hyperglycemia-induced changes in protein turnover in the context of HDL functions, the protein turnover results were further analyzed using Gene Ontology (GO) enrichment and STRING network analysis. Proteins with altered turnover rates were annotated with various GO terms and were compared based on their functions in different biological processes (BP). The functional grouping showed that the proteins with increased turnover rates in prediabetic animals are involved in acute-phase response and negative regulation of peptidase and endopeptidase activities (Fig. 6A and Supplemental Table S6). Consistent with our previous study in mice fed HFD for 4 wk (30), we found that the turnover rates of the several key proteins of the complement pathway were significantly increased in prediabetic mice (Supplemental Table S3). Diabetes reversed HFD-induced increase in the turnover rates of several acute-phase proteins, including transferrin and ceruloplasmin, and further decreased their synthesis as compared with control animals (Fig. 6B). These proteins are involved in iron transport and CVD pathophysiology. Similarly, the diabetes-induced decline in turnover rates was observed for coagulation cascade proteins in the prediabetic state (Fig. 6C). Diabetes also reduced the turnover rates of APOAI and apolipoprotein AIV (APOAIV), which are involved in reverse cholesterol transport (RCT) and lipid absorption, respectively (Fig. 6D). APOAI and APOAIV deficiencies are associated with CVD and diabetes (31, 32).

Figure 6.

Diabetes progression affects HDL proteome dynamics. A: relationships between the turnover rates for proteins overrepresented (FDR < 0.05) compared with the mouse plasma proteome, with functionally related gene ontology (GO) categories in control, prediabetic (HFD), and diabetic (HFD+STZ) mice. Proteins within a pathway exhibit coordinated increased and decreased protein turnover in prediabetic and diabetic states, respectively. Differences in turnover rates are shown for the proteins from the top 3 GO Biological Process (BP) categories that experienced significant changes between at least one of the experimental conditions (ANOVA tests, FDR-corrected P < 0.05). The list of individual proteins in each of these and other GOBP categories is available in the Supplemental Table S6. ANOVA results are shown for each GO BP category. *,#P < 0.05 vs. Control and HFD, respectively. B–D: differences in the turnover rates between the control, high-fat diet (HFD) and the high-fat diet and streptozotocin (HFD+STZ) groups for proteins belong to the acute-phase response (B), coagulation (C), and reverse cholesterol transport pathways (D). *,#P < 0.05 vs. Control and HFD, respectively. ANT3, antithrombin III; APOH, β-2-glycoprotein 1; APOA1, apolipoprotein A-1; APOA2, apolipoprotein A-II; APOA4, apolipoprotein A-IV; ALBU, albumin; CERU, ceruloplasmin; CFAI, complement factor I; CO3, complement C3; FDR, false discovery rate; FIBA, fibrinogen α-chain; HRG, histidine rich glycoprotein; KNG1, Kininogen-1; KLKB1, plasma kallikrein; PLMN, plasminogen; STZ, streptozotocin; THRB, prothrombin; TRFE, serotransferrin; VTDB, vitamin D binding protein.

To understand the biological significance of hyperglycemia-induced changes in HDL proteome dynamics in the context of diabetic CVD, we further examined relationships among proteins with significantly altered turnover rates using STRING network analysis (26). Notably, all three groups of proteins with significantly altered turnover rates due to HFD and/or HFD + STZ showed networks with significant enrichment of interactions among studied proteins (PPI enrichment P values < 1.0e-16). Similar to GO results, STRING analysis demonstrated that many proteins with altered turnover rates are involved in acute-phase response, coagulation, and transport functions. Specifically, STRING analysis (Supplemental Fig. S3A) revealed that overrepresented Reactome pathways (33) in prediabetic group are related to acute-phase response and coagulation (FDR <0.01), suggesting that HFD-induced insulin resistance in prediabetic state is associated with proinflammatory and prothrombotic changes in the HDL proteome. The STRING-based network analysis also demonstrated that, in addition to changes in coagulation and innate immunity cascades, the diabetes progression resulted in changes in proteins involved in lipid metabolism (Supplemental Fig. S3, B and C). Furthermore, diabetes progression was associated with the alterations in the turnover rates of proteins involved in hemostasis, insulin action, platelet degranulation, innate immunity, and regulation of complement cascade (Supplemental Fig. S3D). Interactions networks further identified hub proteins that have multiple interactors, including fetuin-A, pro-thrombin, kininogen 1, anti-thrombin, among others, known to be involved in the pathogenesis of diabetes and diabetic CVD (34–36).

Role of Glycation in Diabetes-Induced Changes in Protein Turnover

Our published work demonstrated that HDL dysfunction is related to alterations in HDL proteins, including an increase in the amount of glycation of APOAI, the key HDL protein involved in cholesterol transport (37, 38). Therefore, we investigated proteomics data for glycation. The posttranslational modifications (PTM) analysis revealed that multiple HDL-associated proteins were glycated. However, due to the low glycation stoichiometry, we were able to quantify the kinetics of only Amadori glycated albumin in prediabetic and diabetic mice based on its glycated peptide k^GluQTALAELVK at Lys-600 residue. Interestingly, the glycated form of albumin in both groups had slower turnover rates compared with nonglycated native protein (Fig. 7). Glycation-induced alterations were associated with the slower turnover rate of albumin in the diabetic mice relative to both control and prediabetic animals (Supplemental Table S1).

Figure 7.

The effect of hyperglycemia-induced glycation on albumin turnover. Slow ²H-incorporation into glycated albumin peptide k^GluQTALAELVK at Lys-600 suggests that glycation of albumin inhibits its degradation and contributes to oligomerization.

DISCUSSION

Diabetes pathophysiology has a widespread systemic impact. The deleterious systemic effects of diabetes are triggered even before the disease symptomology is expressed. Especially, the evidence suggests that cardiovascular pathophysiology starts to manifest even in the prediabetic state. Since the changes in protein turnover precede changes in protein expression, here we used the ²H₂O-metabolic labeling approach to assess the dynamics of plasma proteins in mice. Since there is a linear relationship between glycemia and CVD (39–41), and HDL is involved in CVD protection, we focused on HDL-associated proteins during the disease progression from prediabetes to diabetes in mice. Because high blood glucose results in glycation-induced alterations in protein stability (37, 42), we also searched for the glycated proteins and quantified the turnover of glycated albumin, which exhibited higher glycation levels. The findings uncovered in this study suggest distinct HDL proteome dynamics in the prediabetic and diabetic states. Specifically, we found increased protein turnover during disease onset in the prediabetic state, which declined in the mature diabetic state. Because of the importance of HDL in CVD prediction, the changes in HDL proteome dynamics have the potential to provide unique insights into the early development of prediabetes-related CVD risk and further diabetic complications.

Our study indicates a divergence in the effect on protein turnover based on whether the mice were in a prediabetic or diabetic state. Although prediabetes was characterized by an increase in FSR of several HDL-associated proteins compared with control conditions, diabetic mice showed a general decrease in protein synthesis rate compared with prediabetic mice and in some cases even lower rates than the control mice. Several potential mechanisms may explain these observed differences. For example, the changes in protein turnover rate could be attributed to the effect of changes in insulin levels during the progression of diabetic pathophysiology. The prediabetic state is characterized by hyperinsulinemia, a finding that is replicated in the high-fat feeding regimens such as that used in the present study. In contrast, a diabetic state is characterized by a progressive decline in insulin levels which is reproduced in the STZ-based nongenetic diabetes model used in this study. Insulin regulates all components of proteostasis, including protein synthesis, folding, posttranslational modifications, and degradation. The tracer studies demonstrated that in normal healthy conditions, insulin increases protein synthesis but decreases protein degradation (13). Insulin infusion decreases protein degradation in nondiabetic fasting men (43). In contrast, reduced insulin sensitivity in diabetes is associated with enhanced protein catabolism (44). Interestingly, insulin deficiency also increases whole body protein synthesis, however, the increase in synthesis rate is much less than that of the degradation rate (45, 46). Insulin can have differential effects on the turnover rates of individual plasma proteins that largely originate from the liver (13). Importantly, high-fat feeding leads to insulin resistance in mice. Increased FSR of HDL-associated proteins in prediabetic mice in our study suggests that the resistance to the insulin’s effect on protein synthesis develops later than the resistance to the effect on cellular glucose uptake. In diabetic mice, the decrease in protein turnover is likely explained by the decrease in insulin levels, which leads to increased protein catabolism with unmatched protein synthesis.

Furthermore, protein glycation in the diabetic condition is likely an important determinant of protein turnover. A widely understood mechanism of diabetic complications is the nonenzymatic glycation of extracellular proteins. Such reactions lead to the formation of Amadori glycation adducts and their degradation products, advanced glycation end products (AGEs) that are causative in several pathological changes associated with diabetes. AGEs formed due to excess glucose have also been linked to impairment of the RCT function of HDL and lipid accumulation in macrophages (20). Recently, we found that hyperglycemia-induced glycation in diet-controlled patients is associated with increased degradation of APOAI and serotransferrin involved in CVD protection (37, 38). Furthermore, we demonstrated overall altered HDL proteome dynamics in mildly hyperglycemic diabetic patients (47). Interestingly, glycation contributes to the increased degradation of HDL proteins that is likely an important factor in HDL dysfunction. In contrast, our results in this study show that glycation was associated with the reduced turnover rate of albumin. It has been shown that glycation results in the aggregation of albumin and its reduced lipid-binding and anti-oxidant properties (48). In this study, the glycation status of other proteins could not be determined due to the low occupancy of glycation sites. Nevertheless, based on other reported data (37, 42), we speculate that glycation could be involved in altered turnover rates and functions of proteins in diabetes. The data from the current study suggests that glycation of albumin could be a contributing factor towards a slower turnover of albumin. Liver pathology is another factor that can affect the turnover rates of plasma proteins. Although the fat deposition in the liver as measured by hepatic TG content was higher in prediabetic mice than the diabetic mice, the levels of hepatic inflammation were not evaluated. Diabetes is characterized by increased liver inflammation and ballooning fibrosis which occurs in the late stage of NAFLD. Furthermore, diabetes is one of the risk factors for the progression of NAFLD to nonalcoholic steatohepatitis (NASH). Therefore, differences in the extent and type of liver pathology may explain the observed differences in the turnover rates of HDL-associated proteins. In addition, STZ is known to cause liver cell death in a dose-dependent manner (49). The effect of STZ to cause hepatocellular toxicity is also likely to influence the synthesis of HDL-associated proteins in the liver.

The present study identified changes in the turnover rates of several HDL proteins implicated in the pathogenesis of diabetes and diabetic complications. Complement factor H, α-2-HS-glycoprotein, β-2-glycoprotein-I, and hemopexin have been shown to be elevated in diabetes (35, 50–54). Complement factor H levels are inversely related to insulin sensitivity; α-2-HS-glycoprotein mediates inflammatory signaling leading to insulin resistance; β-2-glycoprotein-I is associated with atherogenic and thrombotic complications of diabetes; hemopexin’s heme scavenging property protects against atherosclerosis suggesting that its elevation in diabetes is protective in nature. The deficiency of α-1 antitrypsin is associated with an increased risk of type 2 diabetes (55), and it has been shown to be inactivated by glycation (56, 57). Ceruloplasmin promotes Fe²⁺ to Fe³⁺ conversion whereas serotransferrin is a transport protein for iron (38). Since excessive iron levels induce β-cell dysfunction and insulin resistance, whereas iron deficiency can impair the function of β-cell enzymes and proteins involved in glucose sensing and oxidation (58), changes in ceruloplasmin and serotransferrin turnover rates may play a role in diabetes pathophysiology. Vitamin D-binding protein and afamin regulate vitamin D and E levels respectively. Vitamin D deficiency is associated with decreased insulin release and insulin resistance and an increase in inflammatory markers (59, 60). Afamin levels have been shown to relate to hepatic lipid content and insulin resistance and the incidence of prediabetes, diabetes, and metabolic syndrome (61, 62). Diabetes is associated with not only a reduction in HDL-C levels but also qualitative changes in HDL function including reduced cholesterol efflux capacity and anti-inflammatory activity (63–65). Therefore, we speculate that hyperglycemia could affect the cardio-vascular protective functions of HDL during diabetes progression via altered turnover rates of HDL proteins.

It is important to acknowledge the limitations of the present study. First, we were unable to quantify the kinetics of low-abundant HDL proteins. Second, although steatotic changes in the liver secondary to diabetes development are a likely cause of alterations in plasma protein profile, the liver histology and phenotype were not characterized in the present study. However, the hepatic TG content was elevated in both HFD and HFD + STZ groups, although not to the levels observed in long-term HFD feeding studies. Third, the present study relies solely on the measurement of kinetic changes in protein turnover and did not investigate protein expression levels. Plasma protein expression profiling, as reported in other dietary investigations (66), would complement the findings of turnover data in this study. Fourth, due to sample limitations, the present study did not assess whether changes in protein turnover rates correlated with the changes in the insulin levels during diabetes progression. However, based on published reports characterizing the HFD + STZ mouse model, insulin levels are expected to be higher than control in the prediabetes group and lower in the diabetes group (due to the STZ-mediated 50% reduction of pancreatic β-cell mass) (67, 68). In addition, the measured apparent FSR values represent the net result of intracellular protein synthesis and secretion. Thus, we could not differentiate and evaluate the hyperglycemia-mediated relationship between protein synthesis and secretion. Despite these limitations, this study for the first time demonstrated the role of diabetes progression on the turnover rates of multiple HDL proteins implicated in pathogenesis and complications of diabetes.

In conclusion, our study uncovered an early dysregulation in HDL-associated plasma proteome dynamics during type 2 diabetes development. Since many of these proteins are involved in diabetes and its complications, these findings may provide mechanistic insights into the changes in protein levels and can help to identify early changes in diabetes disease progression.

SUPPLEMENTAL DATA

Supplemental Figs. S1–S3 and Supplemental Tables S1–S6:https://doi.org/10.6084/m9.figshare.20231991.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grant 1R01HL129120-01A1 (to T. Kasumov), National Institute of Diabetes and Digestive and Kidney Diseases Grant 3U24DK097771-07S1 (to T. Kasumov), National Institute on Alcohol Abuse and Alcoholism Grant 2P50AA024333-06 (to T. Kasumov), and National Institute of General Medical Sciences Grant 5R01GM112044-02 (to T. Kasumov). The Q Exactive Orbitrap mass spectrometer was purchased via Leppo/Nicely Research Fund to College of Pharmacy, NEOMED.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.A.P., S.K., and T.K. conceived and designed research; M.E., M.A., A.A.-A., E.C., S.I., and T.K. performed experiments; M.E., M.A., A.A.-A., E.C., S.I., and H.P. analyzed data; M.E., M.A., A.A.-A., E.C., S.I., and T.K. interpreted results of experiments; M.A., A.A.-A., E.C., and S.I. prepared figures; P.S. and T.K. drafted manuscript; P.S., J.A.P., S.K., and T.K. edited and revised manuscript; P.S., M.E., M.A., A.A.-A., E.C., S.I., H.P., J.A.P., S.K., and T.K. approved final version of manuscript.