Outstanding Improvement of the Advanced Lipoprotein Prole in Subjects with New-Onset Type 1 Diabetes Mellitus After Achieving Optimal Glycemic Control

Background: An altered lipoprotein prole is often hidden in normolipidemic subjects with uncontrolled type 1 diabetes mellitus (T1D). Conceivably, its amelioration after glycemic control may be overlooked in controlled T1D subjects. In this regard, we investigated the inuence of glycemic optimization on lipoprotein subfraction parameters in normolipidemic new-onset T1D subjects. Methods: Twelve subjects at onset of T1D (5 women and 7 men) were studied. Serum lipid and advanced lipoprotein proles were determined by routine laboratory methods and 1H-NMR spectroscopy shortly after diabetes diagnosis (baseline), and after achieving optimal glycemic control (HbA1c<7%) at a follow-up of 11.3 to 25.6 weeks. Results: Advanced lipoprotein analysis revealed a signicant reduction from baseline in plasma concentrations of triglycerides (TG) and cholesterol (C), and in apolipoprotein (Apo)B-containing lipoproteins of treated subjects (VLDL-TG: -21%, LDL-TG: -36%, IDL-TG: -32%, LDL-TG: -36%, P<0.05; VLDL-C: -46%, IDL-C: -46%, LDL-C: -16%; P<0.05). Additionally, neither VLDL-TG and LDL-TG nor LDL-C of T1D subjects at follow-up differed from those of non-diabetic subjects. Decreased VLDL and LDL lipids were mainly attributed to concomitant reductions in the concentration of medium-sized VLDL (-39%) and medium-sized LDL (34%) and, to a lesser extent, to large-sized LDL (13%), respectively. Notably, proatherogenic IDL characteristics and related surrogates of atherogenicity were resolved upon intensive glycemic control. However, neither the concentration of HDL-P, nor the relative proportion of HDL sizes was inuenced by the glucose-lowering

Hyperglycemia is a hallmark of T1D and a major independent risk factor for CVD in this type of diabetes [2]. The contribution of other hidden additional risk factors further elevates CVD risk in a synergistic manner in T1D [2], and their management has been prioritized in recent updates of the American Diabetes Association (ADA) and the European Association (EASD) recommendations [3,4]. In this regard, accumulating evidence suggests that glycemic control has been strongly associated with improved CVD outcomes in T1D [2,5,6]. Although the underlying mechanisms remain insu ciently understood, a signi cant part of the cardioprotective effect of intensive glucose-lowering is likely related to the favorable effects on lipoprotein metabolism [7,8].
Additionally, normal LDL-C values may coexist with increased intermediate-density lipoproteins (IDL), and small and dense LDL (sdLDL) in T1D subjects [12,13]. Likewise, HDL-C values may be normal or even moderately elevated in T1D subjects [11,14,15], but HDL in diabetes may display poorer antiatherogenic properties [16,17], due at least in part to the cumulative associated glycoxidative modi cations of HDL proteins [14].
Advanced lipoprotein analysis using nuclear magnetic resonance (1H-NMR) spectroscopy allows the screening and characterization of circulating lipoprotein classes [11,18], and the identi cation of speci c lipoprotein characteristics, such as size and composition, provided by this advanced analysis may be potentially useful in timely and accurately predicting CVD risk in subjects with T1D.
The use of advanced lipoprotein analysis to more accurately evaluate parameters of lipid metabolism beyond the classical clinical lipid pro le may yield relevant information of how glycemic regulation is in uencing lipid metabolism in T1D. Supporting this view, optimization of glycemic control with intensive therapy signi cantly decreases sdLDL in subjects with poorly controlled T1D with a mean disease duration of 1.7 ± 3.5 (0-16) years [15]. In a recent cross-sectional study that included different cohorts, lipoprotein analysis by 1H-NMR spectroscopy revealed reduced serum concentrations of apolipoprotein (Apo)B-containing lipoproteins, both very-low-density lipoprotein (VLDL) and LDL particles (VLDL-P and LDL-P), and concomitant lower levels of VLDL-TG and non-HDL-C in subjects with T1D [18]; however, in the latter study, T1D participants were on intensi ed insulin therapy for > 1 year and/or ≥ 10 years, and thus the favorable impact, if any, of the improvement of glycemic control on their lipoprotein pro le in the same individual could not be assessed with this cross-sectional design. Therefore, we tested the hypothesis that optimization of glycemic control would improve the extended serum lipoprotein pro le of subjects with new-onset T1D by assessing the advanced lipoprotein pro le before and after the optimization glycemic control.

Participants
This was a two-center before-and-after study performed at the University Hospital Arnau de Vilanova (Lleida) and the University Hospital Germans Trias i Pujol (Badalona) in northwestern Spain. The study protocol was approved by the Ethics Committees of both participating centers (CEIC-1571 and PI-15-147, respectively), and all participants were informed and gave their consent to participate in the study.
T1D subjects included in the present study were part of a before-and-after study primarily designed to assess the metabolomic changes induced by hyperglycemia in new-onset T1D and by its intensive treatment. The inclusion criteria were as follows: adults (≥ 18 years) with a new clinical diagnosis of T1D according to the ADA criteria who provided written informed consent. Importantly, serum sampling in new-onset T1D subjects was done after full recovery from ketosis/ketoacidosis (i.e., absence of ketonemia was de ned as < 0.5 mmol/L) to prevent further metabolic masking and, thus, assess the direct effect of previous exposure to severe hyperglycemia on the lipoprotein phenotype. Additionally, a pre-speci ed HbA1c target of < 7% at follow-up was established; thus, subjects not ful lling this criterion after a maximum time period of 26 weeks were excluded from the study. Consequently, a small subset of T1D subjects (N = 12) was eventually included to analyze the impact of intensive glycemic control on lipoprotein subclasses. According to this rationale, this study should be considered as an exploratory pilot sub-study. Exclusion criteria at onset were: presence of dehydration or ketosis, and previous history of dyslipidemia, hypertension, chronic kidney disease or any form of cardiovascular disease.
As a control group, we selected 12 non-diabetic subjects (HbA1c < 5.7% and fasting glucose < 100 mg/dL) matched for gender, age, and body mass index (BMI) to the T1D subjects after optimization of glycemic control. This control group was used as a reference group to compare the bene ts of intensive therapy on the lipoprotein pro le in this subset of T1D subjects (Additional le 1). Exclusion criteria for the control subjects were the same as those considered for the T1D subjects.

Clinical and conventional biochemical determinations
A thorough clinical assessment was performed, including the collection of relevant parameters including weight, waist, and BMI. Fasting blood samples were obtained for all biochemical measurements. Glucose and glycated hemoglobin (HbA1c) were determined using standard procedures. The glomerular ltration rate (GFR-CKD-EPI) was calculated according to Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [19]. Serum samples were stored at -80ºC until analysis. Standard laboratory analysis included total cholesterol and triglycerides (TG), and LDL-C and HDL-C. ApoB and ApoA-I, and Lp(a) analyses were performed enzymatically using commercial kits (Roche Diagnostics SL, Sant Cugat del Vallès, Spain), and non-esteri ed fatty acids (NEFA) were determined using reagents from Wako Diagnostics (Wako Chemicals, Osaka, Japan) adapted for a COBAS c501 autoanalyzer (Roche Diagnostics SL, Sant Cugat del Vallès, Spain). Both were considered to further con rm favorable lipoprotein changes and insulin therapy in intensi ed T1D subjects.

Advanced serum lipoprotein pro le
Advanced lipoprotein analysis was performed in serum samples using the Liposcale® test (Biosfer Teslab, Reus, Spain), a 2D diffusion-ordered 1H-NMR spectroscopy-based method, as previously described [20]. The advanced analysis included the determination of lipid (TG and cholesterol) concentrations, size of three different classes of lipoproteins (very-low-density lipoprotein (VLDL), LDL and HDL) as well as the particle concentration of nine subclasses (large, medium and small VLDL, LDL and HDL). The method also included the evaluation of IDL lipid concentrations. Brie y, 2D 1H-NMR spectra were recorded on a Bruker Avance III 600 spectrometer, operating at a proton frequency of 600 MHz at 310 K (Bruker BioSpin, Rheinstetten, Germany). The methyl signal was surface-tted with nine functions associated with speci c lipoprotein subclasses according to its NMR size. The particle number of each main lipoprotein fraction was calculated by dividing the lipid volume by the particle volume of a given class and weighted average VLDL, LDL and HDL particle sizes (in nm diameter units) were determined [21]. In relation to the atherogenic indexes, non-HDL-P, the sum of all ApoB-containing lipoprotein classes, the total-P/HDL-P ratio, and the LDL-P/HDL-P ratio were evaluated. In addition, the following were also determined: atherogenic particles (VLDL-P + LDL-P); the atherogenic particles/HDL-P ratio, atherogenic cholesterol (as VLDL-C + IDL-C and LDL-C); the atherogenic cholesterol/HDL-C ratio; and remnant cholesterol (VLDL-C + IDL-C).

Statistical analysis
Sample size (n = 12 per group) was determined by the availability of subjects with con rmed intensive optimization of glycemic control in a sub-cohort of T1D subjects selected from a previous study. The number of subjects included in our study was su cient to assess the changes in NMR-assessed lipoprotein subclass pro les.
Data are presented as percentage for categorical variables, and as mean and standard deviation (SD) for continuous variables. Standardized differences (effect size) were computed to assess the magnitude of changes between clinical and lipoprotein parameters using the Cohen's d function of effect-size package (version 0.4.1) using the free R statistical software version 3.6.0. (https://www.r-project.org/). A test signi cance of global changes was performed for each table (Clinical, laboratory, lipid content, etc.) using non-parametric Sign test. P values of each speci c sub variable were computed using post hoc t-test corrected for Benjamini & Hochberg (false discovery rate). Differences between groups were analyzed using independent t-test between groups. A p-value < 0.05 was considered as statistically signi cant at a family-wise level. Pearson correlation coe cients were used to assess the relationship between changes in NMR-assessed lipoprotein subclass pro les and changes in conventional clinical/biochemical variables in treated T1D subjects. A heat map was created to show the correlations of changes, and a forest plot was generated to show the relative changes in NMR-assessed lipoprotein subclass pro les. Calculations and plots were made using the above-mentioned version of the free R statistical software.
Data of this study is available via open code at the following cloud domain: https://github.com/jrealgatius/LIPOS_DEBUTS.

Results
A total of 20 subjects with T1D who ful lled the inclusion criteria were initially recruited after signing the informed consent. However, 8 subjects were excluded, 6 for not achieving the pre-de ned target HbA1c at follow-up, and 2 who were lost to follow-up. For the remaining 12 subjects, all were assessed in the outpatient clinic within 7 days after the diagnosis of T1D.
Clinical and conventional lipid pro le analysis in subjects with T1D The clinical characteristics of the study T1D subjects are shown in Table 1. Optimization of glycemic control in subjects with T1D was shown by HbA1c levels of 6.2±0.5%, being the highest HbA1c value at follow-up 6.7%. Body mass index was increased after intensi ed optimization of glycemic control.
Conventional biochemical pro les did not differ among groups, except for plasma glucose and HbA1c. Lipid parameters determined in subjects with onset T1D were within established normal ranges, except for a female with a TG concentration of 239 mg/dL and 2 males with lower baseline HDL-C (32 and 35 mg/dL, respectively) (Additional le 2). Overall, the conventional lipid pro le normalized after the intensive management of glycemic control except in one subject who had a mild increase in the concentration of TG (163 mg/dL) (Additional le 2). The only meaningful change in the conventional lipid pro le produced by intensively optimized glycemic control was a signi cant decrease in the LDL-C (-16%; p=0.011) ( Table 1). ApoB was consistently lower at follow-up (-21%, p=0.002) in T1D (Additional le 3). Despite this, total cholesterol was just marginally reduced in subjects after achieving optimal glycemic control (p=0.054) compared with baseline concentrations (Table 1). Additionally, serum concentrations of both TG and NEFA did not signi cantly differ between baseline and follow-up (Table1, Additional le 3).
Of note, HbA1c values showed a close-to-signi cant trend to be increased (p=0.056) in T1D subjects (6.2%) compared with non-diabetic subjects (5.2%), even with intensi ed optimization of glycemic control (Additional le 1).
Advanced NMR analysis of the serum lipoprotein pro le of subjects with T1D Advanced 1H-NMR analysis yielded important changes in the lipoprotein pro le in T1D subjects after intensive glycemic control (Table 2). Indeed, the serum concentration of non-HDL-P, which is representative of total ApoB-containing lipoproteins, was signi cantly decreased (-17%, p=0.005), as was the sum of atherogenic particles (-16%, p=0.006) (Additional le 4).
Total serum levels of VLDL-P did not change after glycemic optimization. However, the concentration of medium VLDL-P (representing about 10% of the three VLDL subclasses) was signi cantly reduced after achieving optimal glycemic control (-39.5%, p=0.033) ( Table 3). This nding was accompanied by a concomitant decrease in the serum levels of total VLDL-TG and VLDL-C (-22%, p=0.021 and -23%, p=0.045, respectively) and in the IDL-C (-46%, p<0.001) and IDL-TG concentrations (-30%, p<0.001) ( Table  2). Of note, the ratio IDL-C/IDL-TG was 23% lower in T1D subjects after glycemic control optimization (Additional le 5). Taken together, total circulating atherogenic remnants, which may be inferred by the sum of both VLDL and IDL classes, were favorably in uenced (-35%, p=0.002) by optimization of glycemic control (Additional le 4).
Regarding HDL, serum levels of HDL-TG were signi cantly decreased in T1D subjects (-18%, p=0.045) after reaching optimal glycemic control, but did not signi cantly in uence serum concentrations of HDL-C (Table 2). In line with this, the HDL-TG-to-HDL-P ratio was concomitantly decreased after optimization (-39%, p=0.003) whereas that of HDL-C-to-HDL-TG increased compared with baseline values (Additional le 5).
Correlation between the changes produced after optimization of glycemic control on changes in serum lipid pro le with anthropometric and clinical variables in subjects with T1D Favorable changes of most VLDL characteristics, including concentrations and composition of circulating VLDL subclasses, were positively associated with improved HbA1c values (Additional le 6). Though moderately, the IDL lipids and circulating remnants (non-HDL-P) were also associated with favorable changes in HbA1c. Interestingly, large-sized (buoyant) LDL-P was inversely associated with leukocyte counts, a surrogate of in ammation. No additional signi cant relationships among lipoprotein characteristics and clinical or biochemical variables were seen, except for those between the HDL-C-to-HDL-TG ratio and liver gamma-glutamyl transferase.

Forest plot
This analysis showed changes in advanced lipoprotein characteristics during glycemic control in subjects with T1D (Additional le 7). Interestingly, the effect size of intensi ed glycemic control optimization was signi cantly larger in LDL and IDL than that observed in the VLDL.

Discussion
Current evaluation of lipid abnormalities often fails to uncover the full lipoprotein atherogenicity in normolipidemic subjects with T1D under conventional metabolic management [15,[22][23][24][25][26][27]. We herein show for the rst time the potential of 1H-NMR spectroscopy to unveil hidden favorable changes in the whole spectrum of circulating ApoB-containing lipoprotein classes and lipoprotein atherogenicity in new onset T1D subjects after the achievement of optimal glycemic control.
Our data is consistent with previous studies from our group [18,27] and others [15,[23][24][25] showing favorable reductions in the serum levels of ApoB-containing lipoproteins, VLDL and LDL, and elevations in HDL in T1D subjects compared with matched non-diabetic subjects. However, a favorable, less proatherogenic change in serum IDL in T1D subjects after optimized glycemic control in T1D subjects may have been overlooked in the previous studies [18 , 27].
Serum concentrations of different classes of ApoB-containing lipoproteins were overall decreased in T1D subjects after optimal glycemic control (Additional le 1). Consistently, this nding was accompanied by a concomitant reduction of serum concentrations of ApoB (Additional le 3), which is a surrogate of the total number of circulating non-HDL particles. In line with previous data [28], signi cant reductions in cholesterol transported by ApoB-containing lipoproteins ( Table 2, Additional le 4) would reveal an improved atherogenicity of these lipoproteins. Additionally, our data also showed a signi cant delipidization in all the ApoB-containing lipoproteins. Increased TG content in these lipoprotein classes is a marker for several types of atherogenic lipoproteins [29], and elevations of TG-rich lipoproteins, including VLDL and IDL (VLDL remnants), promote atherogenesis independently of LDL [30]. TG reductions are accompanied by a favorable reduction of cardiovascular risk in T1D subjects after glycemic control optimization.
IDL are proatherogenic cholesterol-laden lipoproteins contributing to remnant cholesterol and may frequently accumulate in the serum of several dyslipidemic disorders associated with accelerated atherosclerosis [31]. Additionally, elevated IDL may coexist with apparently normal levels of serum LDL levels and is considered an independent risk factor for aortic atherosclerosis in subjects on hemodialysis [32]. As occurs in the current study subjects, higher serum levels of IDL-C increase the atherogenic risk in normolipidemic T2D subjects [33,34]. Importantly, glycemic optimization produced a favorable decrease in IDL-C and IDL-TG and thus atherogenic remnants, which in turn improved lipoprotein atherogenicity indexes in controlled T1D subjects. Because the lipoprotein pro le of T1D subjects used in our previous study were analyzed in subjects with a reasonable diabetic control [11], IDL could plausibly be within the normal range de ned by the non-diabetic subjects and thus not detected even in the case that serum IDL-C could have been eventually measured. Of note, these data highlight the importance of measuring serum IDL-C and IDL-TG to further characterize the atherogenic lipoprotein pro le in subjects with T1D and thus their contribution to elevated CVD risk.
Some VLDL characteristics were inversely related to waist circumference in optimized T1D subjects.
Intriguingly, these same VLDL parameters have been positively related to components of metabolic syndrome, i.e., BMI, waist circumference, in ammatory state, and fatty liver index in a cohort of wellcontrolled T1D subjects [27]. The reason for this discrepancy is unknown, but it could, at least partly, be due to differences in the study design. Indeed, our study provides data on dynamic changes in serum medium and large VLDL-P levels and its negative relationship with the magnitude of change in the waist circumference of our T1D subjects in response to glycemic control optimization, whereas the previous study shows a steady-state relationship between both parameters [27].
Hyperglycemia also profoundly affects HDL metabolism [35]. Indeed, glycated ApoA-I is a main determinant of the atheroprotective actions of HDL. Moreover, glycated ApoA-I also reduces the clearance of HDL from circulation, which could conceivably explain the elevated trend towards increased serum concentrations of HDL-C in T1D subjects even after glycemic control optimization compared with nondiabetic subjects (Additional le 1). On the other hand, TG enrichment of HDL, a hallmark of T1D at onset [35], was reversed with optimal glycemic control in our T1D subjects (Additional le 5). The latter could be attributed to increased hepatic lipase (HL) activity, among other factors. HL hydrolyzes TG from IDL/LDL and also HDL [36]. Interestingly, its release is positively in uenced by insulin [37][38][39][40], and hence it could possibly be induced in optimized T1D subjects in our study, and potentially underlie the decreased TG content in both IDL/LDL and, to a lesser extent, HDL.
The strengths of this study include the fact that it is focused on intensi ed optimization of glycemic control of T1D, and its longitudinal nature. T1D subjects were apparently su ciently insulin treated, as revealed by well-resolved ketosis/ketoacidosis (showing normal serum levels of TG at baseline) and an absence of changes in the serum concentrations of NEFA (Additional le 3). Limitations include the relatively small sample size used, contributing to the pilot nature of the study. Only T1D participants reaching glycemic intensi cation criteria (with HbA1c values below 6.5%) were included, and thus the results may not be representative of the entire T1D population. The latter may limit the interpretation of the observations and should therefore be considered with caution. Thus, the favorable changes in ApoBcontaining lipoproteins, particularly the proatherogenic IDL, should be con rmed in future studies with a larger number of subjects.

Conclusion
The current work reinforces the value of determining the advanced lipoprotein pro le for the study of lipid abnormalities in T1D. Optimal glycemic control in very poorly controlled T1D subjects favorably in uences lipoprotein disturbances. Importantly, the net effect of hyperglycemia in new-onset T1D subjects was analyzed without the contribution of other metabolic confounders and the 1H-NMR approach unveiled a relevant improvement in the atherogenic characteristics of lipoproteins, mainly involving ApoB-containing lipoproteins, which often remain undetected using conventional methodologies. Remarkably, the advanced lipoprotein analysis also unveiled favorable changes in the atherogenic IDL characteristics after glycemic optimization. Further studies are warranted to validate the atheroprotective implications of these conclusions.