Serum apolipoproteins and apolipoprotein-defined lipoprotein subclasses: a hypothesis-generating prospective study of cardiovascular events in type 1 diabetes

: Apolipoproteins APOB, APOC3, and APOE, and apolipoprotein-defined lipoprotein subclasses (ADLS; based on qualitative apolipoprotein complement) have been associated with dyslipidemia and cardiovascular disease (CVD). Our main objective was to define associations of serum apolipoproteins and ADLS with ‘any CVD’ and ‘major atherosclerotic cardiovascular events’ (MACE) in a prospective study of T1D. Serum apolipoproteins and ADLS (14 biomarkers in total) were measured in sera (obtained 1997-2000) from a subset (n=465) of the Epidemiology of Diabetes Interventions and Complications (EDIC) cohort. Prospective associations of ‘any CVD’ (myocardial infarction, stroke, confirmed angina, silent MI, revascularization, or congestive heart failure) and MACE (fatal or nonfatal myocardial infarction or stroke), over 5942 and 6180 patient-years follow-up respectively, were investigated using Cox proportional hazards models, unadjusted and adjusted for risk factors. During 15 years follow-up, 50 ‘any CVD’ and 24 MACE events occurred. Nominally significant positive univariate associations with ‘any CVD’ were APOB, APOC3 and its sub-fractions [heparin precipitate (HP), heparin soluble (HS)], and ADLS-defined Lp-B. In adjusted analyses, APOC3-HS remained nominally significant. Nominally significant positive univariate associations with MACE were APOC3 and sub-fractions, and Lp-B:C; those with total APOC3 and APOC3-HS persisted in adjusted analyses. However, these associations did not reach significance after adjustment for multiple testing. There were no significant associations of APOA1, APOA2, APOE, or other ADLS with either ‘any CVD’ or MACE. These hypothesis-generating data suggest that total serum APOC3 and APOC3 in HDL are potentially important predictive biomarkers for ‘any CVD’ and MACE in T1D adults.


INTRODUCTION
The incidence and prevalence of type 1 diabetes (T1D) is increasing globally, and despite modern management of associated risk factors, it remains associated with greatly increased morbidity and mortality from cardiovascular disease (CVD) (1-3). Serum lipids and apolipoprotein concentrations are important biomarkers of CVD, and several studies have identified the clinical utility of serum apolipoprotein levels and measurements of apolipoprotein-defined lipoprotein subclasses (ADLS) in predicting vascular complications of Type 2 diabetes (4)(5)(6)(7)(8), complementing known associations of CVD with conventional lipids; however, such data are limited in the T1D population.
In studies of apolipoprotein concentrations in plasma in non-diabetic cohorts, APOB has been positively associated with CVD events (9,10). Similar associations have been observed with APOC3, which is involved in the transport and catabolism of triacylglycerols (11). The distribution of APOC3 between APOB-and nonAPOB-containing particles can be elucidated by heparin precipitation. Thus, APOC3 in APOB-containing lipoproteins can be precipitated using heparin (APOC3-HP), while that in APOA1-containing lipoproteins remains in the supernatant, i.e. is 'heparin-soluble' (APOC3-HS). A higher APOC3-HS: APOC3-HP ratio ('APOC3 ratio') is considered a useful index of peripheral catabolism of triacylglycerol (TG)-rich lipoproteins (6). Serum APOE is less clearly associated with atherogenic risk (12): it enhances uptake of triglyceride-rich particles by remnant receptors primarily in the liver (13). APOB-containing lipoproteins include LDL, IDL and VLDL, while APOA1-containing lipoproteins are predominantly associated with HDL.
Apolipoprotein-defined lipoprotein subclass (ADLS) analysis relates particle function and metabolism: it defines particles according to their qualitative apolipoprotein complement, and the subclass names reflect the apolipoproteins present on the particle (14). ADLS thus provides information about the distribution of apolipoproteins among particle classes, in contrast to 'simple' quantification of and generally has pro-atherogenic effects (15). The Monitored Atherosclerosis Regression Study (MARS) and the Cholesterol Lowering Atherosclerosis Study (CLAS) demonstrated that high levels of Lp-B and Lp-B:C particles predict coronary events even when adjusted for conventional lipid profiles (16,17). Prospective data on the associations of serum apolipoproteins and ADLS and with CVD events in T1D are lacking and are the subject of our current analysis.
In 1993, the Diabetes Control and Complications Trial (DCCT, 1983(DCCT, -1993) demonstrated that intensive diabetes management for an average of 6.5 years, with an emphasis on glycemic control, dramatically reduced the onset and progression of the microvascular complications of diabetes (18). Few CVD events had occurred by the time of DCCT close-out in 1993 because the study cohort was still young (34 ± 7 years, mean ± SD) and with relatively short duration of T1D (12±4 years, mean ± SD) (19).
Furthermore, risk of early events was reduced by the original DCCT entry criteria: these mandated an absence of CVD events or CVD risk factors such as hypertension and hyperlipidemia. In the continuing follow-up phase of DCCT, the Epidemiology of Diabetes Interventions and Complications (EDIC) study, implemented in 1994, HbA1c levels converged and became almost identical between the former DCCT randomization groups, but despite this, over ensuing years carotid intima media thickness (IMT) progressed more slowly in the former intensive vs 'conventional' management group, and there were fewer clinical CVD events (20,21), a phenomenon dubbed 'metabolic memory' (19). We previously reported crosssectional and some longitudinal associations between novel lipoprotein-related risk factors and carotid intima-media thickness (IMT) in the DCCT/EDIC cohort (5,22,23), but the longitudinal associations of these detailed metrics with macrovascular disease events in T1D have not been explored. Now with 27-at UCLA DIGITAL COLLECTIONS SVCS, on June 28, 2019 www.jlr.org Downloaded from years follow-up from DCCT baseline, we report longitudinal associations of serum ADLS and individual apolipoproteins with 'any CVD' and Major Adverse Cardiac Events (MACE) in this T1D cohort.

METHODS
The DCCT/EDIC study cohort and related methods have been previously described (18). Briefly, starting in 1983, the DCCT was a randomized study designed to compare the rates of microvascular complications between participants assigned to receive intensive therapy (n=711) aimed at lowering glycemic values to near the non-diabetic range, and participants assigned to conventional therapy (n=730) aimed at maintaining clinical well-being with no specific glucose targets. At baseline, the DCCT study cohort consisted of a primary prevention cohort (1-5 years of diabetes duration, no retinopathy based on fundus photography and albumin excretion rate (AER) less than 40 mg/24 h) and a secondary intervention cohort (1-15 years of diabetes duration, minimal to moderate non-proliferative retinopathy and AER < 200 mg/24 h). In 1993, at the end of DCCT, all participants were instructed in intensive therapy and referred back to their healthcare providers. Then 97% of the surviving DCCT cohort enrolled in the EDIC study, the DCCT follow-up observational study, and 94% of the surviving DCCT participants were still actively participating after an additional 20 years follow-up.
In 1996, a collaboration between the Medical University of South Carolina (MUSC) and DCCT/EDIC was initiated to identify vascular risk factors. Twenty-five of 28 DCCT/EDIC clinical centers participated, and in 1997-2000 (EDIC Years 4-6), serum samples were shipped overnight on dry ice to MUSC; on arrival, aliquots were promptly prepared and stored at -70°C until analysis. The study, which meets Declaration of Helsinki guidelines, was approved by the Institutional Review Boards at MUSC and all participating DCCT/EDIC centers. Each participant gave written informed consent. Of the 1,441 DCCT participants, 968 agreed to participate in the MUSC study, but cost and resource considerations precluded determination of ADLS and apolipoprotein concentrations in all of these participants. The present study therefore utilized a previously-described subset (n=465) (24)(25)(26). Briefly, all those with abnormal albuminuria (albumin excretion rate (AER) >40 mg/24 hours), increased ETDRS retinopathy score (53/<53 at UCLA DIGITAL COLLECTIONS SVCS, on June 28, 2019 www.jlr.org Downloaded from or higher), or elevated carotid atherosclerosis (≥25% stenosis at a carotid lesion) were included (i.e., all available cases meeting one or more of these definitions were sampled), together with a larger group of subjects free of all of these complications. The three disease categories (albuminuria, retinopathy, carotid stenosis ≥25%) were combined and reweighted to reflect the demographic and vascular disease status of the entire EDIC cohort at EDIC Year 6 (see Statistical Analysis, below).

CVD Risk Factors. DCCT/EDIC study visits included a detailed medical history including demographic
and behavioral risk factors, medical outcomes, and a physical examination which included measurements of height, weight, sitting blood pressure, and pulse rate. Blood samples were collected at each visit and assayed centrally at the DCCT/EDIC Central Biochemistry Laboratory (University of Minnesota) for HbA1c, using high-performance ion-exchange liquid chromatography. Fasting lipids (triglycerides, LDL and HDL cholesterol) were measured annually during DCCT and in alternate years during EDIC and were evaluated centrally (27). Total cholesterol, triglyceride, and HDL-C levels were determined using previously reported enzymatic methods (23). LDL-C was estimated according to the Friedewald equation Variables were fixed (e.g. sex) or time-dependent (e.g. HbA1c), the latter captured either as the current (most recent) value, or as the updated mean from baseline. The updated mean is the weighted average of prior values using weights proportional to the time interval between the measurements.

Apolipoproteins and Apolipoprotein-Defined Lipoprotein Subclasses Measurements
We evaluated a subset (n=465) of the DCCT/EDIC cohort using fasting samples collected at EDIC Years 4-6 (1997-2000), with participants selected based on retinopathy, nephropathy and carotid stenosis status.
For all measures, sampling weights were calculated based on the relative distribution of retinopathy, nephropathy and carotid stenosis in this sub-cohort and in the full EDIC cohort, and were further adjusted for sex and original DCCT cohort (primary prevention vs. secondary intervention). Apolipoproteins were quantified by electro-immunoassays for APOA1, APOA2, APOB, APOC3, and APOE (14). APOC3 was measured in whole serum before and after heparin precipitation. The precipitation step enables determination of APOC3 bound to APOA1-containing lipoproteins (heparinsoluble, APOC3-HS) and APOC3 bound to APOB-containing lipoproteins (heparin precipitate, APOC3-HP) (28). We also report APOC3-HS/HP or "APOC3 ratio" (APOC3-R) as a useful index of catabolism of TG-rich lipoproteins (29).
For determination of APOB-containing ADLS, 100µL of whole plasma was mixed with buffer solution and then sequentially treated with polyclonal antisera to APOA2, followed by antisera to APOE, and finally with antisera to APOC3, with overnight incubations at each step followed by centrifugation to separate the precipitates and supernatants (30). Determination of APOB in the first precipitate and all supernatants  (30). Lp-A1 and Lp-A1:A2 were measured by differential turbidimetry and defined according to APOA1 content (31). ADLS assays were conducted in the Lipid and Lipoprotein Laboratory at the Oklahoma Medical Research Foundation using previously described procedures (30,32).

Cardiovascular outcomes
CVD events were ascertained based on medical history, electrocardiogram and available medical records, and were adjudicated by a committee masked to DCCT treatment group and HbA1c levels. The primary CVD outcome ("any CVD") was defined as the time to the first occurrence of CVD death, non-fatal myocardial infarction, non-fatal stroke, subclinical myocardial infarction on ECG, angina confirmed by ischemic changes with exercise tolerance testing or by clinically significant obstruction on coronary angiography, revascularization (with angioplasty or coronary artery bypass) or congestive heart failure (paroxysmal nocturnal dyspnea, orthopnea or marked limitation of physical activity caused by heart at UCLA DIGITAL COLLECTIONS SVCS, on June 28, 2019 www.jlr.org Downloaded from disease) (33)(34)(35)(36). The secondary CVD outcome, major atherosclerotic cardiovascular events (MACE) included only the time to fatal or nonfatal myocardial infarction or stroke, whichever occurred first.
Participants free of a CVD event were administratively censored as of 12/31/2013 (the date of the last CVD data lock). Participants with 'any CVD' events (n=11) and MACE events (n=6) prior to the study sample collection time-point were excluded from these analyses, while all subsequent incident CVD events until censoring were included in the analyses.

Statistical analysis
Summary statistics (counts and percentages for binary variables, and medians and quartiles for continuous variables) of the baseline (i.e., EDIC Year 4-6) characteristics were used to describe the participants and to assess whether, after adjustment for the sampling weights, the 465 participants with available apolipoprotein and ADLS measurements included in these analyses were representative of the entire DCCT/EDIC cohort. All statistical analyses employed the sampling weights computed based on retinopathy, nephropathy and carotid stenosis status, further adjusted for sex and the original DCCT cohort.
Summary statistics were also used to describe the lipoprotein measures both overall and separately by the initial DCCT treatment group, while the Wilcoxon test allowed for the sampling weights as employed to test for differences between the two DCCT treatment groups. Kaplan-Meier survival curves were used to describe the time-to-event outcomes (i.e. any CVD and MACE). Cox proportional hazards models (separately for any CVD and MACE) were employed to assess the association between each lipoprotein measure and the risk of CVD. The power to detect associations in time-to-event analyses is dictated by the number of events. Given the relatively small number of incident any CVD events among the 465 participants in our study, and informed by our previous work on risk factors for CVD in the DCCT/EDIC cohort (35), a set of pre-specified models was considered: Model 1 (unadjusted); Model 2 (age and mean at UCLA DIGITAL COLLECTIONS SVCS, on June 28, 2019 www.jlr.org Downloaded from HbA1c); Model 3 (age, mean HbA1c and log triglycerides); Model 4 (age, mean HbA1c and LDL); and Model 5 (age, mean HbA1c, log triglycerides, systolic BP and pulse rate). HbA1c, LDL, pulse rate and systolic blood pressure (SBP) were defined using the updated mean values, with weights proportional to the time between visits. For triglycerides, we used the most recent value, as this has previously been shown to be most strongly associated with CVD events (35), and data were log-transformed. Given the lower number of MACE events, only models 1 to 4 were considered for the MACE outcome. All analyses were performed using R, and p-values ≤0.05 were considered nominally significant. Given the exploratory nature of our analyses, the reported p-values were computed without adjustment for multiple testing. Table 1 presents the characteristics of the 465 participants at EDIC Years 4-6 (1997-2000), with values presented after adjustment for sampling weights, and compares these to the characteristics of the full DCCT/EDIC cohort at the same time-point. For the 465 participants, sampling weights were computed based on retinopathy, nephropathy, and carotid stenosis status, and further adjusted for sex and original DCCT cohort (primary or secondary prevention). After weight adjustment, 47% of the participants were from the original DCCT intensive treatment group, 50% were from the original primary prevention cohort, 53% were males, and 17% were current smokers, and the study sub-set was representative of the full cohort.

RESULTS
Among the 465 participants in this study, by the end of 2013, there were 50 incident 'any CVD' events over a total follow-up of 5,942 patient-years (rate of 8.4 events (any) per 1,000 individuals at risk for one year), and 24 MACE events over a total follow-up of 6,180 patient-years (rate of 3.9 events per 1,000 individuals at risk for one year). Figure 1 shows the survival probability curves for the time to 'any CVD' event and time to MACE in our cohort.
Summaries of the apolipoprotein and ADLS biomarkers stratified by initial DCCT randomization groups (intensive vs. conventional) are presented in Table 2. Only Lp-A1 was borderline nominally different, being higher in the former conventional vs. intensive group (p=0.049), and APOA1 also tended to be higher in the former conventional vs. intensive group (p=0.06). Table 3  events, these potential risk factors were included separately as additional covariates in the models described in Table 3, Model 5 for 'any CVD event' and in Table 4, Model 4 for MACE. The results were little different from those above.

DISCUSSION
Overall, our analyses revealed that APOC3 and its subfractions showed the most consistent and nominally significant prospective associations with risks of 'any CVD' event and MACE in models adjusted for traditional cardiovascular risks and conventional lipids. Among the other apolipoproteins measured in our cohort, APOB showed a positive association with 'any CVD' event, but not with MACE, after adjustment for age and mean HbA1c. Among the ADLS biomarkers, only Lp-B showed a positive association with 'any CVD event', but none was associated with MACE after adjustment for age and mean HbA1c. The findings add new data on the role of APOC3 in predicting CVD risks in individuals with T1D.
Apolipoprotein C3 (APOC3) is synthesized in the liver. It is found on the surface of VLDL, LDL and HDL particles, and has been independently associated with hypertriglyceridemia and cardiovascular events. Various mechanisms for APOC3 atherogenicity have been proposed: it leads to decreased clearance of APOB from the circulation (37), formation of small dense LDL (38), and stimulation of hepatic formation of triglyceride-rich VLDL (39). In a meta-analysis of eleven studies that included 2,832 cases with cardiovascular events, each 5mg/dL increase in total APOC3 was associated with a 33% increase in risk of cardiovascular events (40). In our study, we observed risk to increase by approximately 10% for every mg/dL increase in total APOC3 for 'any CVD' event and for MACE. In a recent report of 4,659 participants with CVD risk factors in the Multi-Ethnic Study of Atherosclerosis (MESA), HDL particles containing APOC3 were positively associated with coronary artery calcification in women (41).
Our findings are consistent with these previous observations, as we observed nominally significant at UCLA DIGITAL COLLECTIONS SVCS, on June 28, 2019 www.jlr.org Downloaded from positive associations of the HDL-containing sub fraction of APOC3 (APOC3-HS) with 'any CVD' event and with MACE: this persisted in our most rigorously-adjusted model which adjusted for systolic blood pressure, and pulse, both pre-specified variables in our analytic plan. Recently, again in DCCT/EDIC, we reported prospective associations of APOC3-HP with carotid IMT in men (17); similar positive associations of APOC3 with carotid intima-media thickness (IMT) have been reported in a healthy population (42), and with coronary heart disease in high-risk adults, including those with type 2 diabetes (43).
We did not observe any association of APOC3 ratio with CVD events or MACE, possibly due the absence of associations of 'APOC3 in LDL and VLDL' (i.e. APOC3-HP) and APOB with cardiovascular outcomes in adjusted models. Our findings in this study add to the existing literature, and suggest that APOC3, as well as APOC3 content of HDL, may represent a therapeutic target in people with Type 1 diabetes.
Among the other apolipoproteins measured, we observed a nominally significant positive association of APOB with 'any CVD' event, but not with MACE in models adjusted for age and HbA1c.
We previously found a positive association of APOB with IMT in DCCT/EDIC (17), and APOB has been associated with risk for stroke and increased IMT in other populations with and without diabetes (44,45).
No significant associations of HDL-containing APOA1, A2, or APOE with either 'any CVD' or MACE were found in our cohort; this may be explained by the overall low prevalence of dyslipidemia at baseline and the low number of events in the follow-up period.  (46), as well as with macrovascular disease in subjects with type 2 diabetes (47). Elevated Lp-B:C was shown to be highly correlated with the progression of atherosclerosis, reflected by coronary artery calcium (CAC) scores, in patients with rheumatoid arthritis (48). We previously showed that Lp-B was prospectively associated with increased IMT in men in DCCT/EDIC, but no previous studies have reported associations with CVD events and MACE in a T1D population. In our study, in models adjusted for blood pressure and/or conventional lipids, ADLS were not significantly associated with future CVD, perhaps because of strong collinearity among atherogenic lipids that contribute to CVD.

Competing interests
The authors declare no competing interests.

Funding
This project was funded by grants from the National Institutes of Health (R01DK080043) and the American Diabetes Association (7-12-CT-46). DCCT/EDIC has been supported by U01 Cooperative