Carbamylated HDL and Mortality Outcomes in Type 2 Diabetes

Nonenzymatic posttranslational modifications are chemical reactions that contribute to protein molecular aging (1). Carbamylation is one of these chemical processes and involves nonenzymatic binding of isocyanate to free amino groups of proteins. The carbamoyl moiety (-CONH₂) is added to the amino terminus residues of protein-like lysine, forming ε-carbamyllysine (homocitrulline). The source of isocyanate can be derived from urea dissociation or from myeloperoxidase-mediated catabolism of thiocyanate (2,3). Carbamylation leads to changes in both the structural and functional properties of proteins, and carbamylated proteins have been implicated in the pathogenesis of cardiovascular and renal diseases as well as immune system dysfunction (2–6).

The magnitude of carbamylation burden in an individual can be determined by measuring the overall levels of carbamylated plasma proteins (e.g., carbamylated albumin and protein-bound homocitrulline) or by quantifying specific carbamylated proteins (3). The prognostic value of measuring systemic carbamylation burden has been best characterized in patients with end-stage renal disease (ESRD), because carbamylation is markedly increased in these individuals as a result of uremia (7). Carbamylation is an independent risk factor for adverse outcomes in individuals with ESRD. Several studies have shown that protein carbamylation levels can predict overall survival in hemodialysis patients with and without diabetes, and high degree of carbamylation burden is associated with cardiovascular and all-cause mortalities (8–10). Moreover, it has been shown that advanced kidney disease is not a prerequisite for excessive carbamylation. Recent studies have demonstrated that measures of increased systemic protein carbamylation burden are associated with severity of coronary disease, cardiovascular events, and mortality in patients without renal impairment (5,11). Carbamylation is also increased in patients with type 2 diabetes and normal renal function, and carbamylation is mainly driven by elevated myeloperoxidase activity in these individuals (12).

The pathophysiological effects of carbamylation depend on the molecules being modified. Lipoproteins are subjected to carbamylation, and we have shown that carbamylated HDL is associated with progression of diabetic kidney disease in type 2 diabetes independently of traditional risk factors (13). Carbamylation can lead to HDL dysfunction (14), and whether carbamylated HDL can predict mortality outcomes in individuals with diabetes not receiving dialysis has not been investigated. In the current study, we aimed to evaluate the value of carbamylated HDL as a prognostic indicator of survival in patients with type 2 diabetes and relatively well-preserved kidney function and the association with mortality outcomes.

The association between plasma carbamylated HDL and mortality outcomes was investigated in a clinic-based cohort of patients with type 2 diabetes. The cohort was recruited from the Diabetes Clinic from 1996 to 2014 to study the pathogenesis of and progression of complications in type 2 diabetes in Chinese patients, and details of the cohort have been reported previously (13). Individuals with type 2 diabetes were invited to participate on their first clinic visit, and major exclusion criteria were non-Chinese descent, type 1 diabetes, malignancy or major illness with limited life expectancy, any hospitalization in the preceding 3 months, or unwillingness to return for regular follow-up. Only patients who fulfilled all the inclusion criteria were invited to participate, and the overall participation rate was ∼40%. Fasting blood samples were taken at baseline, and clinical data of the cohort were collected during longitudinal follow-up. Mortality outcomes were ascertained from review of electronic medical records/death certificates. The primary outcome was all-cause mortality. Secondary outcomes were cause-specific deaths based on the principal diagnoses and further defined according to the ICD-9. The diagnoses were reviewed and adjudicated by two physicians independently, and any disagreement was resolved by a third. Length of follow-up was calculated as the time from baseline examination to the date of death or last follow-up as per the censoring date of 30 June 2020, whichever was earliest. Patients receiving renal replacement therapy (dialysis or renal transplantation) at baseline were excluded from the analysis in the current study. Informed consent was obtained from all participants, and the study was approved by the Ethics Committee of the University of Hong Kong.

Plasma levels of lipids, carbamylated HDL, glucose, HbA_1c, and creatinine were measured in fasting blood samples taken at baseline. Serum creatinine was measured by the Jaffe method, and estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation. Plasma total cholesterol and triglyceride were determined enzymatically, and HDL cholesterol (HDL-C) was measured using a homogenous method with polyethylene glycol–modified enzymes and α-cyclodextrin. LDL cholesterol (LDL-C) was calculated by the Friedewald equation or measured directly if plasma triglyceride was >399 mg/dL. Plasma carbamylated HDL concentration was measured by an in-house sandwich ELISA as previously described (13), using a polyclonal rabbit anti-human carbamylated HDL antibody as the capture antibody and polyclonal goat anti-human ApoA1-horseradish peroxidase antibody (K45252P; Biodesign, Memphis, TN) as the secondary detecting antibody. To raise polyclonal antibody against carbamylated HDL, rabbits were immunized with purified human carbamylated HDL synthesized by Alfa Aesar (Lancashire, U.K.). The polyclonal rabbit antibody detected carbamylated adduct present on HDL, and the specificity was confirmed by Western blotting and indirect ELISA. The quantity of carbamylated adduct in test samples was determined by comparing its absorbance with that of a purified human carbamylated HDL calibration curve, and the results were expressed as µg/mL carbamylated HDL protein. Each sample was assayed in triplicate, and the intra- and interassay coefficients of variation were 6.8 and 9.6%, respectively.

Results were expressed as mean and SD or as median and interquartile range if the data were not normally distributed. Skewed data were logarithmically transformed before analyses were performed. Comparisons between two groups were done using independent sample Student t test. Multivariable Cox regression analysis was used to estimate the hazard ratios (HRs) and 95% CIs. Competing risk analysis was performed to obtain a better estimate of the absolute risk of each cause-specific mortality outcome. All models were adjusted for variables that were statistically significant in univariate analysis or were biologically relevant. Receiver operating characteristic (ROC) analysis was performed to evaluate models including traditional risk factors with or without plasma carbamylated HDL. Comparisons of C index (the areas under the ROC curves) were made by using the DeLong method. The improvement in reclassification of risk of death was evaluated using category-less net reclassification index (NRI) and integrated discrimination improvement (IDI) using the R package “Hmisc” (15).

The summary data that support the findings of this study are available from the corresponding author on reasonable request.

A total of 1,679 participants with type 2 diabetes were recruited. After excluding those receiving renal replacement therapy and those with missing follow-up data or baseline samples, 1,517 patients were available for analysis. The proportions of patients according to stage of chronic kidney disease at baseline were as follows: stage 1, 20.5%; stage 2, 54.5%; stage 3, 21.0%; stage 4, 3.6%; and stage 5, 0.4%. During follow-up, 301 patients died. Nine deaths occurred within the first 12 months of recruitment, and these patients were excluded from the analysis to minimize the possibility of reverse causation. The median duration of follow-up was 14 years (interquartile range 9.5–17.4 years), and the mortality rate was 14.5 per 1,000 person-years. Thirty-one percent of deaths resulted from cardiovascular causes (n = 91), 21% from renal causes (n = 60), 20% from malignancy (n = 59), 19% from infection (n = 55), and the remaining from miscellaneous causes like trauma or chronic lung disease or uncertain etiologies. Baseline demographic and clinical characteristics of the patients, as well as their laboratory measurements, are shown in Table 1. Patients with a fatal outcome were older (P < 0.01), but there was no significant difference in the duration of diabetes. There was a significantly higher proportion of smokers and a higher prevalence of hypertension and cardiovascular disease at baseline in those patients with adverse outcomes. They also had lower eGFR, worse glycemic control, and higher LDL-C. There was no significant difference in plasma HDL-C between the two groups of patients, and plasma HDL-C was not associated with all-cause mortality on Cox regression analysis (HR 1.04; 95% CI 0.93–1.15).

Despite similar concentrations of plasma HDL-C in the two groups of patients, plasma carbamylated HDL was elevated in those with a fatal outcome (46.1 ± 17.8 µg/mL vs. 32.9 ± 10.7; P < 0.01), and the difference remained significant even after adjusting for age and sex (Table 1). Kaplan-Meier analysis was performed to evaluate the association of plasma carbamylated HDL and all-cause mortality (Fig. 1). Baseline plasma carbamylated HDL was stratified into quartiles (Q1, <28.5 µg/mL; Q2, ≥28.5 and <32.8 µg/mL; Q3, ≥32.8 and <41.0 µg/mL; and Q4, ≥41.0 µg/mL), and there was a graded association between increasing quartiles of plasma carbamylated HDL and all-cause mortality (log-rank test P < 0.001). The unadjusted HRs of the third and fourth quartiles versus the first quartile were 2.11 (95% CI 1.40–3.17; P < 0.001) and 6.55 (95% CI 4.67–9.77; P < 0.001), respectively.

Multivariable Cox regression analysis was performed to adjust for potential confounding factors, including traditional cardiovascular risk factors and baseline clinical characteristics and renal function. In the crude model, the unadjusted HR per 1 SD change in plasma carbamylated HDL was 1.72 (Table 2). In the fully adjusted model including age, sex, BMI, duration of diabetes, smoking, systolic blood pressure, HbA_1c, LDL-C, HDL-C, baseline cardiovascular disease, renin-angiotensin system blockers, lipid-lowering therapy, and eGFR, the association between plasma carbamylated HDL and all-cause mortality remained significant. We further analyzed whether plasma carbamylated HDL was associated with cause-specific mortality outcomes, including cardiovascular-, renal-, infection-, and cancer-related deaths, and the results are shown in Table 2. In the fully adjusted model, plasma carbamylated HDL was independently associated with cardiovascular- (HR 1.36), renal- (HR 1.37), infection- (HR 1.53), and cancer-related deaths (HR 1.39). Adjusting for HDL-C did not change the result in any analysis, and there was no interaction between carbamylated HDL and HDL-C with regard to mortality outcomes. In competing risk analysis, plasma carbamylated HDL also independently predicted cardiovascular- (HR 1.31; 95% CI 1.15–1.50; P < 0.01), infection- (HR 1.34; 95% CI 1.11–1.60; P < 0.01), and cancer-related deaths (HR 1.26; 95% CI 1.06–1.49; P < 0.01), but such prediction was only of borderline significance for renal-related deaths (HR 1.19; 95% CI 1.00–1.40; P = 0.05).

ROC analysis was performed to evaluate the predictive power of plasma carbamylated HDL for mortality outcomes. For all-cause mortality, the C index for the model including traditional risk factors, duration of diabetes, HbA_1c, eGFR, baseline cardiovascular disease, renin-angiotensin system blockers, and lipid-lowering therapy was 0.78. The predictive power of the model improved when plasma carbamylated HDL was added; the C index significantly increased to 0.80, NRI was 27.6%, and IDI was 1.74% (P < 0.001) (Table 3). For cause-specific mortalities, the increments in C index were significant but only modest for both cardiovascular- and renal-related deaths, and no improvement was seen in NRI or IDI (Table 3). In contrast, plasma carbamylated HDL enhanced the prognostic performance of the models for infection- and cancer-related deaths, with significant increments in C index and improvements in NRI and IDI.

Our study is the first prospective study to investigate the association between carbamylation and mortality outcomes in patients with type 2 diabetes not receiving dialysis. Berg et al. (8) demonstrated that serum carbamylated albumin was a risk factor for mortality in patients with diabetes with ESRD in the German Diabetes and Dialysis study. We studied a different population of patients with type 2 diabetes. A majority of participants in our study cohort had relatively well-preserved kidney function at baseline, and individuals receiving dialysis were excluded. Instead of examining the overall levels of carbamylated plasma proteins, we chose to evaluate a specific carbamylated protein, carbamylated HDL, as a prognostic marker. This was based on our previous findings that plasma carbamylated HDL is linked to progression of diabetic kidney disease and that carbamylated HDL is a better biomarker than plasma HDL-C (13). The results from the current study show that plasma carbamylated HDL is independently associated with all-cause mortality as well as cardiovascular-, renal-, infection-, and cancer-related causes of death in patients with type 2 diabetes.

For all-cause mortality, the primary outcome of our study, patients in the top quartile of plasma carbamylated HDL had a 6.5-fold increase in mortality risk compared with those in the lowest quartile. There were significant improvements in C index, NRI, and IDI values when plasma carbamylated HDL was added to predictive models including traditional risk factors and potential confounding parameters. Previous epidemiological studies have shown that HLD-C is associated mortality in the general population (16,17). In our clinic-based cohort of patients with type 2 diabetes, plasma carbamylated HDL was a much better predictor of overall survival than HDL-C. There was no association between plasma HDL-C and mortality, and the relationship between carbamylated HDL and mortality outcomes was independent of HDL-C. Our data further show that plasma carbamylated HDL is independently associated with deaths related to cardiovascular or renal causes. Experimental studies have suggested mechanistic links between carbamylation of HDL and atherosclerosis as well as renal disease. Carbamylation of HDL causes HDL dysfunction and reduces the antiatherogenic properties of HDL (14,18). Not only does it lead to a loss of anti-inflammatory, antioxidative, and vasoprotective properties of HDL (14,18–20), carbamylation renders HDL particles ineffective in promoting reverse cholesterol transport and facilitates cholesterol accumulation in cells and tissues like the vasculature and kidneys (19,21,22). However, the C index increments were only modest when plasma carbamylated HDL was added to predictive models including traditional cardiovascular risk factors and measures of renal function. Therefore, plasma carbamylated HDL may not provide additional prognostic value beyond traditional conventional risk factors.

We further explored the association between plasma carbamylated HDL and infection- and cancer-related deaths. Plasma carbamylated HDL significantly improved the predictive power for infection-related deaths, as shown by the significant changes in C index, NRI, and IDI values. This may be because carbamylated HDL captures information related to immune function not reflected by traditional risk factors, including HDL-C, which does not reflect HDL function. HDL particles and their components have been linked with protection against infections because HDL can bind and neutralize lipopolysaccharide and lipoteichoic acid from gram-negative and gram-positive bacteria, respectively (23). Furthermore, it has been shown that HDL can modulate immune response and activation of the complement system (24,25). Thus, HDL dysfunction caused by carbamylation of HDL may potentially impair the anti-infectious properties of HDL. In addition, our data suggest that plasma carbamylated HDL also predicts cancer-related deaths. The mechanisms behind the relationship between carbamylated HDL and cancer-related deaths are unclear. There were insufficient events for us to undertake any detailed analyses of cancer subtypes. Apolipoprotein A1 (apoA1), the major structural protein component of HDL, has been demonstrated to inhibit tumor development in animal models of malignancy, including ovarian cancer and melanoma (26,27). ApoA1 mimetic suppressed vascular endothelial growth factor signaling pathways and inhibited tumor angiogenesis in an experimental study (28). Whether the formation of carbamyllysine–apoA1 in carbamylated HDL affects the antitumor effect of apoA1 warrants further investigation.

The main strengths of our study are the long duration of follow-up and the use of all-cause mortality as the primary outcome, which is a robust ultimate outcome measure. Our study has a number of limitations. We adjusted for traditional risk factors, but we cannot discount the possibility of residual unmeasured confounding factors. The number of outcome events was small for cause-specific deaths, and the results should be interpreted with caution. These analyses were exploratory to shed light on the potential relationship of carbamylated HDL with cause-specific mortality, and additional studies are required. In addition, our data are observational, and we cannot infer causality. HDL is known to have multiple functions (18,29), and we have not determined how carbamylation of HDL affects the different aspects of HDL function. We therefore cannot address whether carbamylated HDL plays a role in the development of adverse outcomes. We did not have serial measurements of carbamylated HDL over time, and we could not completely account for progression of disease or any changes in treatment. However, our results are unlikely to have been influenced by new treatments like sodium-glucose cotransporter 2 inhibitors or glucagon-like peptide 1 receptor agonists, because these two classes of agents only became available in the drug formulary of our public health care system in 2018. During follow-up, the proportions of patients initiating statin therapy were fairly similar (39 and 43% in those who did and did not have a fatal outcome, respectively). Another limitation is that our cohort of patients was recruited from a secondary/tertiary referral center. They were likely to have more complex disease cases than those individuals treated in primary care or in the general population. We also cannot completely exclude selection bias in the recruitment process. Therefore, our results may not be generalized to other patient populations. Missing baseline samples and follow-up data reduced the sample size for analysis and may have also biased our results, but there were no significant differences between those with complete data and those with missing data.

In conclusion, carbamylation of proteins is part of the aging process, and carbamylation is increased in individuals with type 2 diabetes. Carbamylation of HDL renders HDL dysfunctional, and carbamylated HDL is independently associated with mortality outcomes in patients with type 2 diabetes. Additional mechanistic studies are warranted to determine whether carbamylated HDL is simply a disease biomarker or also a mediator of adverse mortality outcomes.

Funding. This study is supported by an endowment fund established for the Sir David Todd Professorship in Medicine awarded to K.C.B.T.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.T.W.L. researched the data and drafted the manuscript. C.-L.C. contributed to study design and analyzed the data. A.C.H.L. and Y.W. recruited the patients and collected clinical data. S.W.M.S. performed the laboratory assays and analyzed the data. K.C.B.T. designed and oversaw the study, data collection, and analysis and critically reviewed and edited the manuscript. All authors approved the final version of the manuscript. K.C.B.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.