Laparoscopic sleeve gastrectomy alters 1H-NMR-measured lipoprotein and glycoprotein profile in patients with severe obesity and nonalcoholic fatty liver disease

Patients with morbid obesity frequently present non-alcoholic fatty liver (NAFL) and non-alcoholic steatohepatitis (NASH) associated with pro-atherogenic alterations. Laparoscopic sleeve gastrectomy (LSG) is an effective treatment for weight reduction, and for the remission of hepatic alterations. Using 1H-nuclear magnetic resonance (1H-NMR), we investigated the effects of LSG on lipoprotein and glycoprotein profile in patients with morbid obesity and liver disease. We included 154 patients with morbid obesity (49 non-NASH, 54 uncertain NASH, 51 definite NASH). A blood sample was obtained before surgery and, in patients with definite NASH, one year after surgery. Patients with NASH had increased concentrations of medium and small VLDL particles, VLDL and IDL cholesterol concentrations, IDL, LDL, and HDL triglyceride concentrations, and elevated glycoprotein levels. These changes were more marked in patients with type 2 diabetes mellitus. LSG produced significant decreases in the concentration of VLDL particles, VLDL cholesterol and triglycerides, an increase in the concentration LDL particles and LDL cholesterol concentrations, and a decrease in protein glycation. We conclude that patients with obesity and NASH had significant alterations in circulating levels of lipoproteins and glycoproteins that were associated with the severity of the disease. Most of these changes were reversed post-LSG.

www.nature.com/scientificreports/ rapid, reliable and detailed analyses of circulating lipoproteins. 1 H-nuclear magnetic resonance ( 1 H-NMR) has the advantage of simultaneously quantifying the number, size and composition of lipoprotein particles, thus providing a better understanding of modifications associated with metabolic disturbances such as NAFLD 9 . In addition, 1 H-NMR spectroscopy allows measurement of different classes of glycoproteins, and this is clinically relevant in patients because increased protein glycation is a marker of the extent of type 2 diabetes mellitus. Also, increased plasma glycoprotein concentrations have been associated with insulin resistance and obesity 10,11 . Laparoscopic sleeve gastrectomy (LSG) is a widely-used surgical procedure for the treatment of morbid obesity and its associated comorbidities 12 . Patients who benefit from LSG not only reduce weight, but also improve insulin resistance and histological features of NAFLD 13,14 . For example, recent studies have reported that patients with morbid obesity treated with LSG have significant decreases in liver volume and hepatic steatosis at 6 months [15][16][17] and improvement of other serious comorbidities, including but not limited to type 2 diabetes, hypertension, and obstructive sleep apnea 18 . A study by our research group showed that the improvement in the histology and liver function of these patients after LSG was associated with mechanisms that involve the reduction of oxidative stress and inflammatory processes 3 . Hence, the present study investigated the effect of LSG on 1 H-NMR-lipoprotein and glycoprotein profile in patients with morbid obesity, and with several degrees of hepatic alterations.

Methods
Study design and participants. This post hoc retrospective cohort study includes new objectives derived from a previous prospective longitudinal study investigating the molecular mechanisms associated with liver injury in morbid obesity and searching for plasma biomarkers of obesity-associated liver disease 3,19 . In this previous prospective study, we included 436 patients with severe obesity (body mass index, BMI > 40 kg/m 2 ) who underwent LSG. All subjects provided 12-h fasting blood samples immediately before surgery together with an intraoperative wedge-liver biopsy. Patients that were diagnosed as having definite NASH were asked to have a second blood extraction and an additional liver biopsy at 12 months post-surgery 3 . In the present study we selected 154 of those patients who were matched for age, sex, BMI, and incidence of diabetes mellitus, hypertension and dyslipidemia, and who were representative of the three most frequent degrees of liver injury: Patients with mild hepatic lesions without NASH (n = 49), patients with uncertain NASH (n = 54), and patients with definite NASH (n = 51). Samples were stored at − 80 °C in the Biobank of our Institution (Banc de Mostres Biològiques, Institut d'Investigació Sanitària Pere Virgili). Clinical indication for LSG was according to guidelines currently used in pre-operative evaluation 20 . We excluded patients with current, or past, history of daily alcohol abuse (≥ 30 g for men and ≥ 20 g for women), long-term consumption of hepatotoxic drugs, and liver disease of infectious origin. LSG was performed under general anesthesia with the patient in the Lloyd-Davies position. A five-port technique was used in all patients. The greater gastric curvature was dissected, separated from the gastroepiploic arcade of the greater omentum, and continued to the His angle. The gastric transection was performed under the guidance of a 38-Fr Faucher bougie. Three cm was the distance from the pylorus to the first section point (measured intraoperatively with a ribbon). The suture line was reinforced using polycarbonate derivatives of polyglycolic acid (SEAMGUARD, W.L. GORE & ASSOCIATES, USA) in order to avoid hemorrhagic processes and leaks. A methylene blue leak was always performed before closing abdominal wall 21 .
All experimental protocols were approved by the Ethics Committee (Institutional Review Board) of Hospital Universitari de Sant Joan (OBESPAD/14.07-31proj3), and patients provided fully informed, signed consent (OM-NAFLD, ESO3/18012013 project). All methods were carried out in accordance with relevant guidelines and regulations.
Histological analysis. Liver biopsies were obtained from the same site in all patients, and were examined by a pathologist blinded to clinical data. Samples were processed conventionally for diagnostic purposes and histological grading and staging, as described 22 . Steatosis was graded in four categories, depending on whether fat droplets occupied < 5%, 6-33%, 34-66%, or ≥ 67% of the total microscopic field; inflammation was graded as: no foci of lobular inflammation observed, or < 2, 2-4, and > 4 foci per field; fibrosis was classified as: absence of fibrosis (Stage 0), mild to moderate fibrosis (Stages 1 and 2), and bridging fibrosis (Stage 3). The presence or absence of NASH was estimated using the NAFLD activity score (NAS score) and defined as the sum of steatosis, inflammation and hepatocyte ballooning. Patients were classified in three categories: non-NASH (n = 49; NAS ≤ 2), uncertain NASH (n = 54; NAS 3-4) and definite NASH (n = 51; NAS ≥ 5) 23 .
Lipoprotein and glycoprotein analyses by 1 H-NMR spectroscopy. Whole blood was centrifuged at 2500×g and 4 °C and serum was aliquoted and stored at − 80 °C until analyses were performed. Lipoproteins and glycoproteins were analyzed by the 1 H-NMR-based LIPOSCALE test, as previously reported [9][10][11] . Cholesterol and triglyceride concentrations, particle size and concentration of the four main classes of lipoproteins (verylow-density lipoproteins or VLDL, low-density lipoproteins or LDL, intermediate-density lipoprotein or IDL, and high-density lipoproteins or HDL), as well as particle concentration of nine lipoprotein subclasses (large, medium and small VLDL, LDL, and HDL) were analyzed. The methyl signals of the serum 2D 1 H-NMR spectra were derived from deconvolution analysis using 9 lorentzian functions to determine the lipid concentration of each lipoprotein, and its diffusion coefficient (Z); these are estimations of particle diameter. Finally, information on lipid concentrations and particle volumes derived from the diffusion coefficients were combined to quantify the number of lipoprotein particles required to transport the measured lipid concentration of each lipoprotein subclass.
Glycoproteins were analyzed at the 2.15-1.90 ppm region of the 1 H-NMR spectrum i.e. where the glycoproteins resonate. We determined the total area (proportional to concentration), height (H), position and bandwidth Statistical analysis. The Kolmogorov-Smirnov test was used to assess the normality distribution of the variables. Wilcoxon rank-sum test or Kruskal-Wallis test (non-parametric) were used to compare independent quantitative variables, and the-square test was used to compare categorical variables. Wilcoxon signed-rank test was employed to compare dependent variables. Correlations between quantitative variables were analyzed with Spearman's Rho test. Multivariate analysis was applied to pattern recognition, including the supervised partial least squares discriminant analysis (PLS-DA). The relative magnitude of observed changes was evaluated using the variable importance in projection (VIP) score 24 . Statistical significance was set at p ≤ 0.05. Statistical analyses were performed with the SPSS 22.0 package and the R program version 3.4. METABOANALYST 4.0 program (available on the web http://www.metab oanal yst.ca/) was used to generate scores and loading plots.

Results
Patient characteristics. Serum aminotransferase activities and the HOMA-IR index were higher in patients with definite NASH compared to those without NASH. We did not find any significant differences in age, sex, BMI, incidence of type 2 diabetes mellitus, hypertension, dyslipidemia, or lipid profile as analyzed by standard clinical laboratory tests (Table 1). However, the alterations observed were associated with the highest degrees of steatosis and fibrosis, but not with lobular inflammation (Supplementary Table S1).

Hepatic alterations and type 2 diabetes mellitus influences serum 1 H-NMR lipoprotein and glycoprotein profiles. Patients with definite NASH had significantly higher values of medium and small
VLDL particles, VLDL and IDL cholesterol concentrations, and IDL, LDL, and HDL triglycerides than those with uncertain NASH. Glycoproteins were significantly elevated in the LMWM2 area ( Fig. 1a and Table 2). Also, patients with the most severe degrees of steatosis had significantly higher values of medium and small VLDL particles, VLDL and IDL cholesterol concentrations, triglycerides in all lipoprotein fractions, and LMWM2 area than patients with less severe steatosis ( Fig. 1b and Supplementary Table S2). Patients with advanced fibrosis had significantly elevated values of medium and small LDL particles, IDL, LDL, HDL cholesterol concentrations, LDL triglycerides, and LMWM1 area that patients with less severe steatosis ( Fig. 1c and Supplementary  Table S3). Patients with more inflammation had significantly higher medium VLDL particles and HDL triglyceride concentrations ( Fig. 1d and Supplementary Table S4).
Type 2 diabetes mellitus was associated with elevated VLDL and IDL cholesterol concentrations, VLDL, IDL and HDL triglyceride concentrations, LMWM2 and Glyc-A areas (Supplementary Table S5).

LSG modifies serum 1 H-NMR lipoprotein and glycoprotein profiles. One year after LSG, patients
showed a significant decrease in BMI, insulin resistance and aminotransferase activities. The 1 H-NMR lipoprotein profile showed that these patients had: a significant decrease in the concentration of small, medium and large VLDL particles, and decreased VLDL cholesterol and triglyceride concentrations; an increase of small, medium and large LDL particles and LDL cholesterol concentrations; an increase in small HDL particles and cholesterol concentrations; an increase in LDL particle diameter ( Fig. 2 and Table 3). The glycoprotein profiles showed a decrease in Glyc-A, B, and F areas, Glyc-A and B width, H/W ratios, and LMWM 1 and 2 areas ( Table 3).
The score plot of the PLS-DA analysis of pre-surgery values of serum lipoproteins and glycoproteins showed a considerable overlap such that a clear distinction between non-NASH, uncertain NASH, and NASH patient groups could not be made. To identify the lipoproteins and glycoproteins that showed the most relevant alterations, we calculated the VIP scores. This score is a measure of the variable's degree-of-alteration associated with the disease i.e. a higher VIP score is considered more relevant in disease status classification. The VIP analysis identified IDL cholesterol concentration as the lipoprotein presenting the most relevant alterations between groups (Fig. 3a). Conversely, the score plot of the PLS-DA analysis clearly distinguished between pre-and postsurgery values because the components had a very slight overlap. The VIP analysis identified LMWM1 area as the parameter showing the most relevant pre-surgery and post-surgery differences (Fig. 3b).

Discussion
Despite the conventional lipid panel not showing any significant differences between groups, we did find significant changes in the 1H-NMR-analyzed lipoprotein profile in patients with morbid obesity and NASH, when compared to patients without NASH. The key findings are that NASH was associated with a greater amount of total VLDL particles due to an increase in medium and small VLDL, and this change was related to higher concentrations of VLDL cholesterol. These alterations were more pronounced in patients with type 2 diabetes Table 1. Clinical, biochemical, and histological variables segregated with respect to the NAS score. Values are shown as number of cases and percentages, or medians and interquartile ranges. ACEIs angiotensinconverting-enzyme inhibitor; ALT alanine aminotransferase; AST aspartate aminotransferase; ARA-II angiotensin II receptor antagonists; BMI body mass index; GGT γ-glytamyl transferase; HOMA-IR homeostatic model assessment of insulin resistance; HDL high-density lipoproteins; LDL low-density lipoproteins; NAS nonalcoholic fatty liver disease activity score; NASH nonalcoholic steatohepatitis; T2DM type 2 diabetes mellitus. Superscript letters indicate significant (at least p < 0.05) differences between: a non-NASH vs. definite NASH; b uncertain NASH vs. definite NASH; *Global P value using the Kruskal-Wallis one-way analysis of variance. www.nature.com/scientificreports/ mellitus. In addition, patients with the most severe degrees of fibrosis had a greater amount of all subclasses of LDL particles. At present, there is a paucity of studies evaluating changes in lipoprotein metabolism in patients with liver disease, using 1H-NMR methods. For example, Siddiqui et al. 7 did not find any significant differences in the lipoprotein profile between NASH and simple steatosis in non-diabetic patients with a moderate degree of obesity. Conversely, Männistö et al. 25 observed an increase in VLDL and LDL concentrations in patients with morbid obesity and NASH. Overall and considering the variety of populations studied the results obtained suggest that alterations in the lipoprotein profile are aggravated as the patient's clinical severity increases in terms of hepatic alterations, the degree of obesity, or the presence of type 2 diabetes mellitus. Our study showed, as well, that patients with advanced fibrosis had the greatest alterations in the lipoprotein profile. A recent study has related fibrosis and lipoprotein metabolism via a receptor termed Recepteur d'Origine Nantais (Ron) 26 . This is a MET proto-oncogene receptor tyrosine kinase, the deficiency of which in mice results in the increase in concentrations of circulating VLDL and LDL, upregulation of collagen synthesis, and downregulation of matrix metallopeptidase-9 (an enzyme with collagenase activity). The possibility that there is a decreased activity of Ron in patients with NASH and fibrosis is an interesting hypothesis, and one that deserves further research.
There is a dearth of information regarding changes in glycoprotein values in patients with morbid obesity and liver disease. We did not find any significant association between these parameters and the presence of NASH, but we did find an increase in the areas of Glyc-A (which is an estimation of the protein-bound N-acetylglucosamine and N-acetylgalactosamine) and Glyc-F (which is an estimation of N-acetylglucosamine, N-acetylgalactosamine and N-acetylneuraminic acid not bound to proteins) in patients with diabetes. These latter results would be expected since protein glycosylation is an essential feature of diabetes. High Glyc-A and Glyc-B levels have been found in diabetes 10 , cardiovascular disease 27,28 , and other inflammatory diseases 11,29 . N-glycosylation of proteins can be a useful biomarker for the diagnosis of type 2 diabetes mellitus as well as for the identification and prognosis in groups at high risk of future diabetes 30,31 . If N-glycosylated proteins are confirmed as biomarkers, the results of the present study would add valuable information since our results indicate that N-glycoprotein levels are not altered with the degree of liver injury and, as such, do not constitute a confounding variable in the associated diabetes. 1 H-NMR glycoprotein analysis adds further details in the detection albeit with low sensitivity www.nature.com/scientificreports/ and specificity of the levels of some soluble metabolites not directly related to glycosylation. In the present study we found an association between the LMWM 1 and 2 areas and the presence of NASH. These areas correspond, mainly, to glutamate and glutamine. These results are consistent with targeted metabolomics studies conducted in our research group in which high concentrations of these metabolites were noted in patients with morbid obesity and liver disease 19,32 .
The treatment-of-choice for morbid obesity is bariatric surgery. There are several techniques in performing this type of surgery, but the most commonly used are LSG and Roux-en-Y gastric bypass. To the best of our knowledge, the effects of LSG on 1 H-NMR lipoprotein and glycoprotein profiles have not been reported in the literature, to date. The present study showed that one year after LSG, the patients had a significant decrease in the concentration of VLDL particles of all sizes, a decrease in VLDL cholesterol and triglyceride concentrations as well as Glyc-A, B, and F, and LMWM 1 and 2 areas. This was accompanied by an increase in the concentration of LDL particles of all sizes, in LDL particle diameter, in LDL cholesterol concentrations, and in small HDL particles and HDL cholesterol concentrations. These results differ from those obtained by Männistö et al. 25 in patients with morbid obesity treated with Roux-en-Y gastric bypass; the authors reported a normalization of VLDL, IDL, and LDL levels and an increase in HDL one year after surgery. However, our results are similar to those of a study in patients with mild obesity receiving pharmacological treatment to induce NASH remission 33 . The authors reported that patients with NASH had high concentrations of small LDL and large VLDL subfractions, and that treatment was associated with an increase in LDL particle diameter and a decrease in VLDL concentrations. We are not sure of the cause of the increase in the concentration of LDL particles in our patients post-LSG, nor of the clinical consequences with respect to the long-term cardiovascular risk that this phenomenon may have. The increase in LDL concentrations together with the decrease in VLDL concentrations suggests a normalization of VLDL metabolism. This may be related to the improvement of insulin resistance and the normalization of liver function with weight loss 34 . Another possibility is an increase in the direct secretion of LDL by the liver. That the increase in LDL particles occurs, essentially, at the expense of large and medium-sized particles (i.e. those with less atherogenic capacity than the smaller particles), and with no increase in LDL cholesterol concentrations suggests that these changes are not necessarily detrimental to the patient's cardiovascular disease status. Monitoring Table 2. Serum cholesterol and triglyceride concentrations in lipoprotein fractions, lipoprotein particle diameter and glycoprotein variables measured by 1 H-NMR segregated with respect to the NAS score. Values are shown as medians and interquartile ranges. HDL high-density lipoproteins; IDL intermediate-density lipoproteins; LMWM low molecular weight molecules; LDL low-density lipoproteins; NAS nonalcoholic fatty liver disease activity score; NASH non-alcoholic steatohepatitis; VLDL very low-density lipoproteins. Superscript letters indicate significant (at least p < 0.05) differences between: a non-NASH vs. uncertain NASH, b non-NASH vs. definite NASH and c uncertain NASH vs. definite NASH. *Global P value using the Kruskal-Wallis one-way analysis of variance. www.nature.com/scientificreports/ these patients for longer periods would clarify whether these changes in lipoproteins (concentrations and/or composition) persist and what effect they would have on the patient's cardiovascular status.
In conclusion, our study reveals important alterations in the characteristics of the different classes of lipoproteins and glycoproteins in patients with obesity, especially in those with definite NASH and/or diabetes. It also indicates that most of these abnormalities can be reversed with LSG.  www.nature.com/scientificreports/ www.nature.com/scientificreports/ LDL low-density lipoprotein; LSG laparoscopic sleeve gastrectomy; P particle concentration; VLDL very-lowdensity lipoproteins; Z particle diameter.