Body Weight and Prandial Variation of Plasma Metabolites in Subjects Undergoing Gastric Band-Induced Weight Loss

Abstract

BACKGROUND:

Bariatric procedures are safe and effective treatments for obesity, inducing rapid and sustained loss of excess body weight. Laparoscopic adjustable gastric banding (LAGB) is unique among bariatric interventions in that it is a reversible procedure in which normal gastrointestinal anatomy is maintained. Knowledge regarding how LAGB effects change at the metabolite level is limited.

OBJECTIVES:

To delineate the impact of LAGB on fasting and postprandial metabolite responses using targeted metabolomics.

SETTING:

Individuals undergoing LAGB at NYU Langone Medical Center were recruited for a prospective cohort study.

METHODS:

We prospectively analyzed serum samples from 18 subjects at baseline and 2 months after LAGB under fasting conditions and after a 1-hour mixed meal challenge. Plasma samples were analyzed on a reverse-phase liquid chromatography time-of-flight mass spectrometry metabolomics platform. The main outcome measure was their serum metabolite profile.

RESULTS:

We quantitatively detected over 4,000 metabolites and lipids. Metabolite levels were altered in response to surgical and prandial stimuli, and metabolites within the same biochemical class tended to behave similarly in response to either stimulus. Plasma levels of lipid species and ketone bodies were statistically decreased after surgery whereas amino acid levels were affected more by prandial status than surgical condition.

CONCLUSIONS:

Changes in lipid species and ketone bodies postoperatively suggest improvements in the rate and efficiency of fatty acid oxidation and glucose handling after LAGB. Further investigation is necessary to understand how these findings relate to surgical response, including long term weight maintenance, and obesity-related comorbidities such as dysglycemia and cardiovascular disease.

Introduction

Obesity is a growing epidemic in the United States with significant medical and economic consequences¹. It is defined as having a body mass index (BMI) of 30 or greater, and is a major risk factor for cardiovascular disease, stroke, and type 2 diabetes mellitus, as well as certain types of cancer². Bariatric procedures have revolutionized the treatment of obesity as they have proven to be extremely effective at eliciting sustained weight loss in individuals with obesity, thereby decreasing their morbidity and mortality³. While most bariatric surgeries result in the permanent alteration of gastrointestinal anatomy, laparoscopic adjustable gastric banding (LAGB) is unique in that it is a reversible procedure in which normal gut anatomy is maintained. As a result, the weight loss that is induced in patients who undergo this procedure is largely the result of restricted food intake, as occurs with dieting^4,5. This is in contrast to other bariatric surgeries which elicit metabolic change, at least in part, due to the endocrinologic changes that come from anatomic alteration^6,7.

LAGB is reported to cause excess body weight (EBW) loss of 42% in the first year, and 55.2% at 5 years⁸. It remains controversial whether changes in the hormonal milieu occur after LAGB and, if they do exist, how they affect weight loss and overall metabolic health^9–13. A recently published pilot study from our medical center aimed to address this question by examining the resting metabolic rate, orexigenic hormone levels, and gastrointestinal hormone levels in 18 patients undergoing LAGB¹⁴. This study reported an LAGB-induced rapid and clinically significant total body weight loss of 9.7 ± 3.5% at 2 months. This weight loss resulted in decreased resting metabolic rate, as well as decreased fasting leptin and ghrelin levels¹⁴. However, there was no significant change in the gastrointestinal hormone response profile, which includes hormones such as amylin, peptide YY (PYY), gastric inhibitory polypeptide (GIP), pancreatic polypeptide (PP), glucagon like peptide-1 (GLP-1), and insulin¹⁴. This is in contrast to what is seen after Roux-en-Y gastric bypass (RYGB) ^6,7, likely for the aforementioned reasons. We hypothesized that the altered leptin and ghrelin levels following LAGB would impact plasma metabolites in a manner dependent on body fat or prandial status respectively.

The detailed characterization of hormonal changes with LAGB sets a framework by which to integrate the metabolite responses to LAGB-induced weight loss with the physiology of this restrictive bariatric procedure. In a post-hoc analysis of the aforementioned study, we used reverse phase LC-TOF-MS to quantitatively detect 4,474 unique metabolites and lipid species from the plasma of study subjects at baseline and 2 months postoperatively, under both fasting conditions and after a 1-hour mixed meal test. We highlight the levels of target amino acids (AA), ketones, lipid species, and free fatty acids (FFAs) and observe that fluctuations in these metabolites follow either prandial or body weight related patterns. Furthermore, we integrated this post-hoc analysis with the previously published clinical and incretin data to develop a linear mixed effects model that might help predict response to weight loss as a function of the variables of interest at baseline. This work couples our understanding of the physiologic changes in LAGB with metabolomic biomarkers that may serve as predictors of weight loss response and provides a framework for understanding the metabolic transformation that occurs after clinically significant weight loss.

METHODS

Subjects and Clinical Study

Twenty-five patients aged 18 to 65 with a BMI > 35 were followed at baseline and 2 months after LAGB. Of these, eighteen patients completed the study. Subjects with history of prior metabolic surgery, thyroid disease, lung disease, active smoking, or concurrent use of medications causing weight loss or weight gain were excluded from the study. The Institutional Review Board approved the study and written consent was obtained from all participants. No participant received a stipend for participation in the study. Sample size in the initial study was determined by power calculations to determine a 5,000 pg/mL (5 ng/mL) difference in leptin levels from baseline to 2 months after LAGB with 5% type I error and 80% power. Patient demographics and metabolic profiles both pre- and post-surgery have been previously published¹⁴.

Data and Sample Collection

Plasma sample collection is previously described¹⁴. Briefly, patients served as their own controls, with blood samples collected prior to LAGB surgery and 2 months post-operatively. Baseline testing was done prior to initiation of the pre-operative liquid or restricted diet. Blood samples were obtained twice at each visit: on arrival, after at least ten hours of fasting, and then one hour after consumption of a standardized liquid meal (Ensure Original: 220 kcal, 6 g fat, 32 g carbohydrates, 15 g sugars, 9 g protein). A butterfly needle was used to procure blood samples. Samples were isolated and mixed with a protease inhibitor to ensure that plasma hormones were not degraded. They were then stored at −80°C until analysis; multiple freeze thaw cycles were avoided as much as possible.

Plasma Metabolomic Post-hoc Analysis

Briefly, plasma samples were defrosted and deproteinated with cold methanol prior to extraction and 1 uL of the LipidoMix standard mixture (Avanti Lipids) was added. Non-polar lipids and polar metabolites were extracted using a modified Bligh-Dyer chloroform/methanol/water extraction method. Polar metabolites and non-polar metabolites were separated to two layers post centrifugation. The top layer is a polar water/methanol layer, while the bottom layer is nonpolar chloroform. The non-polar chloroform layer was centrifuged in vacuum at 40°C and re-suspended in a 4:3:1 (v/v) isopropanol/acetonitrile/water solution. Analysis of polar and nonpolar fractions was completed using an Agilent Infinity II 1290 liquid chromatography system coupled to an Agilent 6230B Time-of-flight mass spectrometry platform (Santa Clara, CA). Metabolite and lipid separation during chromatography was performed through a Zorbax C18 4.6um, 100mm length column. Column temperature was programmed at 50°C with a biphasic gradient change in mobile phase using a 5:1:4 solution of isopropanol/methanol/water and 5mM ammonium acetate and 0.1% acetic acid buffers on pump A and a solution of 99:1 (v/v) of isopropanol/water and 5mM ammonium acetate and 0.1% acetic acid buffer on pump B. Predefined AA/polar metabolites were run as external standards at the beginning and end of each batch of sample runs.

Plasma metabolite profiles were analyzed using the proprietary Agilent platform MassHunter and Mass Profiler Pro (Santa Clara, CA). Briefly, raw metabolomic spectra were log-2-normalized and median centered. Metabolites present in at least 50% of samples were included in further analyses. Statistical significance was determined by paired t-tests. Graphing was performed in MassProfiler Pro and GraphPad Prism.

Statistical Methods

Patient demographics and metabolic profiles are presented as mean ± standard deviation (SD) for continuous variables and count for categorical variables in Table 1. Metabolite values were log2-transformed and compared at pre-surgery (fasting and postprandial) vs 2 months post-surgery (fasting and postprandial) using paired t-tests. These data are presented as mean ± standard error of the mean (SE). P values ≤ 0.05 are statistically significant. A linear mixed-effects model was used to evaluate the change between pre- and post-surgical samples in metabolite variables of interest for fasting and postprandial samples (Table 3). Data is adjusted for age, gender, and race. Subject ID is regarded as a random effect. Next, we used linear regressions to assess for relationships between weight loss (lb and %) and the variables of interest at baseline, adjusting for age, race and gender (Table 4). In all analyses above, race is described by two categories – either white or non-white. Metabolite variables were log2-transformed, and median normalized. Benjamini-Hochberg (BH) procedure is used for multiple comparisons within each category. Finally, we explored the possibility of classifying samples based on only metabolites. However, no significant classifier was found. All statistical analyses were performed using the R software environment (Version 3.6) at an alpha level of p ≤ 0.05, selected a priori as the threshold for significance. P values > 0.05 and <0.100 were considered as statistical trends.

Table 1 –

Patient demographics and metabolic profiles both before and after LAGB. Data is shown as mean ± standard deviation for continuous variables and count for categorical variables.

Variables	Baseline (n=25)	Post-LAGB (n=18)	P Value
Gender (Female/Male)	18/7	13/5
Race (white/non-white)	21/5	14/4
Age (years)	42.5±9.4	44.7±9.0	0.55
Height(m)	1.676±0.076	1.663±0.061	0.59
Weight(kg)	129.9±20.6	116.1±19.7	0.032
BMI	46.3±6.8	42.0±7.4	0.055
Hip Circumference (cm)	124.7±15.5	111.8±10.7	0.037
Waist Circumference (cm)	138.7±12.7	134.9±13.7	0.49
Resting Energy Expenditure	1985.8±414.3	1757.8±353.0	0.067
Systolic Blood Pressure	120.5±13.4	118.9±14.0	0.707
Diastolic Blood Pressure	74.7±8.4	73.9±9.8	0.77
A1C	6.2±1.2
Total Cholesterol (mg/dL)	208.7±31.7	189.3±36.9	0.24
Total Triglycerides (mg/dL)	125.8±48.1	127. 0 ±46.9	0.91
HDL (mg/dL)	68.5±27.6	46.6±10.2	0.11
LDL (mg/dL)	115.1±11.8	117.3±37.4	0.28
%EBMIL		22.6±9.9

Table 3 –

P-values based on linear mixed-effects model to evaluate changes in the variables of interest after LAGB, adjusted for age, race, and gender.

Variable	Fasting			Post-prandial
	Estimate	P value	Adjusted P value	Estimate	P value	Adjusted P value
Ghrelin	−7.776	0.033*	0.149	−8.176	0.083#	0.249
GIP	40.512	0.361	0.650	−0.506	0.990	0.990
GLP-1	1.841	0.551	0.650	6.924	0.242	0.455
Insulin	278.141	0.578	0.650	367.035	0.564	0.725
Leptin	−9612.18	0.001*	0.009	−10034.8	0.001*	0.009
PP	19.312	0.398	0.650	16.671	0.474	0.711
PYY	2.294	0.855	0.855	21.529	0.253	0.455
Amylin	−7.765	0.423	0.650	−2.859	0.832	0.936
Glycine	−0.631	0.500	0.650	−2.48	0.037*	0.167
Metabolites
L-Alanine	−2.891	0.092#	0.660	−3.21	0.034*	0.419
L-Arginine	−1.259	0.503	0.871	0.131	0.921	0.997
L-Asparagine	0.48	0.719	0.871	−0.3	0.471	0.670
L-Aspartic Acid	−1.365	0.306	0.733	0.501	0.254	0.577
L-Cysteine	−0.883	0.464	0.871	1.608	0.171	0.575
L-Cystine	−0.463	0.691	0.871	−2.206	0.062#	0.574
L-Glutamate	0.735	0.554	0.871	0.292	0.795	0.979
L-Glutamine	−1.712	0.210	0.660	−0.085	0.953	0.997
L-Histidine	−0.333	0.772	0.871	−0.186	0.590	0.780
L-Leucine	−1.395	0.190	0.660	−0.33	0.343	0.577
L-Lysine	0.218	0.852	0.923	1.125	0.464	0.670
L-Methionine	0.295	0.873	0.923	−1.549	0.340	0.577
L-Phenylalanine	−0.273	0.681	0.871	−0.004	0.997	0.997
L-Proline	−1.595	0.186	0.660	−1.647	0.392	0.604
L-Serine	1.803	0.232	0.660	0.831	0.297	0.577
L-Threonine	−0.025	0.982	0.982	0.642	0.189	0.575
L-Tryptophan	0.305	0.727	0.871	−0.253	0.820	0.979
L-Tyrosine	−0.093	0.955	0.982	0.291	0.863	0.997
L-Valine	0.743	0.688	0.871	0.06	0.972	0.997
Beta-Hydroxybutyrate	−1.746	0.020*	0.389	−0.684	0.337	0.577
Beta-Hydroxy-Beta-Methylbutyric Acid	−4.262	0.021*	0.389	−4.314	0.006*	0.222
Alpha-Ketobutyrate	−2.333	0.285	0.733	−2.778	0.202	0.575
Methylglyoxyl	−0.926	0.688	0.871	−1.843	0.363	0.584
3-Deoxyglucosone	−3.029	0.042*	0.518	−3.783	0.018*	0.333
Cholesterols	−0.716	0.224	0.660	−0.248	0.245	0.577
Aldosterone	−0.661	0.731	0.871	2.045	0.304	0.577
Cholesterols	1.625	0.223	0.660	−0.83	0.563	0.772
Estradiol	−2.016	0.317	0.733	0.461	0.773	0.979
Progesterone	−0.897	0.177	0.660	−1.19	0.337	0.577
Oleic Acid	−0.114	0.611	0.871	−0.817	0.091#	0.575
Glycerol	0.087	0.777	0.871	−0.572	0.180	0.575
Stearic Acid	−0.311	0.155	0.660	−0.719	0.119	0.575
Palmitic Acid	−0.138	0.507	0.871	−0.328	0.196	0.575
Linoleic Acid	−0.103	0.619	0.871	−0.723	0.137	0.575
Triglycerides	−0.067	0.760	0.871	−0.341	0.254	0.577
Diglycerides	−0.071	0.725	0.871	−0.323	0.197	0.575
Ceramides	−0.311	0.155	0.660	−0.044	0.892	0.997

Median normalization and log 2 transformation method are used for all these metabolites (from L-Alanine to Ceramides) before using the linear mixed model.

Table 4 -.

P-values based on linear model of weight loss (%) regressed by the variables of interest at baseline, adjusted for age, race and gender.

Variable	Fasting			Post-prandial
	Estimate	P value	Adjusted P value	Estimate	P value	Adjusted P value
Ghrelin	−0.164	0.260	0.741	0.138	0.268	0.654
GIP	−0.015	0.488	0.741	0.008	0.465	0.654
GLP-1	0.026	0.868	0.926	0.049	0.509	0.654
Insulin	−0.001	0.394	0.741	0.001	0.322	0.654
Leptin	0.001	0.190	0.741	−0.001	0.157	0.654
PP	−0.003	0.926	0.926	0.001	0.961	0.961
PYY	−0.027	0.494	0.741	0.026	0.329	0.654
Amylin	−0.041	0.341	0.741	0.012	0.728	0.819
Glycine	0.164	0.835	0.926	−0.462	0.476	0.654
Metabolites
L-Alanine	0.196	0.152	0.834	−0.133	0.305	0.966
L-Arginine	0.074	0.628	0.834	0.007	0.966	0.966
L-Asparagine	0.103	0.578	0.834	−0.81	0.058#	0.966
L-Aspartic Acid	−0.616	0.374	0.834	−0.095	0.893	0.966
L-Cysteine	−0.222	0.533	0.834	−0.145	0.344	0.966
L-Cystine	0.23	0.396	0.834	0.13	0.588	0.966
L-Glutamate	−0.293	0.306	0.834	0.066	0.794	0.966
L-Glutamine	−0.166	0.379	0.834	0.037	0.795	0.966
L-Histidine	0.98	0.104	0.834	−0.91	0.128	0.966
L-Leucine	0.42	0.456	0.834	−0.273	0.699	0.966
L-Lysine	−0.281	0.132	0.834	0.247	0.066#	0.966
L-Methionine	−0.153	0.249	0.834	−0.243	0.177	0.966
L-Phenylalanine	1.768	0.026*	0.574	0.173	0.332	0.966
L-Proline	−0.191	0.247	0.834	−0.098	0.462	0.966
L-Serine	0.095	0.573	0.834	0.06	0.828	0.966
L-Threonine	0.093	0.633	0.834	0.087	0.788	0.966
L-Tryptophan	0.251	0.381	0.834	0.051	0.865	0.966
L-Tyrosine	−0.177	0.266	0.834	0.133	0.318	0.966
L-Valine	0.049	0.716	0.834	0.13	0.462	0.966
Beta-Hydroxybutyrate	−0.224	0.324	0.834	−0.896	0.093#	0.966
B-Hydroxy-B-Methylbutyric Acid	−0.023	0.894	0.920	−0.794	0.560	0.966
Alpha-Ketobutyrate	−0.108	0.617	0.834	0.102	0.555	0.966
Methylglyoxyl	0.137	0.359	0.834	−0.038	0.820	0.966
3-Deoxyglucosone	0.23	0.431	0.834	0.156	0.798	0.966
Cholesterols	0.09	0.752	0.843	0.388	0.790	0.966
Aldosterone	−0.131	0.318	0.834	−0.112	0.331	0.966
Cholesterols	−0.018	0.920	0.920	−0.029	0.926	0.966
Estradiol	0.245	0.031*	0.574	0.08	0.487	0.966
Progesterone	0.115	0.651	0.834	0.117	0.499	0.966
Oleic Acid	0.076	0.790	0.860	−0.096	0.864	0.966
Glycerol	0.215	0.468	0.834	0.031	0.964	0.966
Stearic Acid	0.118	0.685	0.834	−0.31	0.646	0.966
Palmitic Acid	0.113	0.696	0.834	−0.261	0.672	0.966
Linoleic Acid	0.101	0.721	0.834	−0.171	0.770	0.966
Triglycerides	0.034	0.905	0.920	0.833	0.331	0.966
Diglycerides	0.167	0.610	0.834	−0.374	0.788	0.966
Ceramides	0.148	0.634	0.834	−0.075	0.938	0.966

Median normalization and log 2 transformation method are used for all these metabolites (from L-Alanine to Ceramides) before using the linear model.

Results

The patient characteristics of the prospective LAGB cohort from which samples were drawn to conduct this post-hoc analysis are described in Table 1. Briefly, subjects had a mean age of 42.5 ± 9.6 years old. 72% of subjects recruited were female and 84% (21/25) reported their ethnicity as Caucasian. Baseline weight was 129.9±20.6𝑘𝑔, with a baseline BMI of 46.3 ± 6.8 kg/m2 and a baseline hemoglobin A1C of 6.1 ± 1.2%. The baseline study population studied had relatively few complications from excess weight, with hypertension (32%), diabetes (24%), and hyperlipidemia (24%) being the most common comorbidities. As previously reported, the mean weight loss was 12.3 kg ± 4.9, the mean percent excess BMI lost (EBMIL) was 22.6 ± 9.9% and the mean percentage of total weight loss was 9.7 ± 3.5%¹⁴. The variance in weight loss in the group was 53.1 ± 4.9. Table 2 shows a summary of the previously published data comparing the levels of various relevant hormones pre- and post-operatively. Of the adipokines and gut hormones measured, only leptin decreased significantly after surgery. Neither ghrelin, GIP, GLP-1, Insulin, PP, PYY nor amylin changed significantly after LAGB.

Table 2 –

Baseline and post-operative levels of measured gastrointestinal hormones pre and post LAGB. Data is shown as a median value with interquartile range.

Hormones	Pre-operative fasting serum hormone concentration (pg/mL) (n=17)	Post-operative fasting serum serum hormone concentration (pg/mL) (n=17)	P value
Ghrelin	11.2(0–32.7)	0.0(0–11.51)	0.0756
GIP	58.20 (41.30–79.90)	66.9 (35.8–89.0)	0.4874
GLP-1	0.0(0–3.29)	0.0(0–5.68)	0.9326
Insulin	503.7 (417.6–1025.5)	520.4(361.2–941.5)	0.6441
Leptin	33884.6(27847.8–40751)	28253.5(14636.1–34813.1)	4.58E-05*
PP	16(9.7–34.3)	20.6(7.6–47)	0.7368
PYY	29.0(0–63.1)	25.1(0–69.75)	1
Amylin	28.8(0–80.21)	24.1(0–66.67)	0.2553

The plasma metabolite profiles for each of these subjects was characterized via reverse-phase liquid chromatography time-of-flight mass spectrometry (LC-TOF-MS), and supervised heatmap analysis of these profiles is displayed in Figure 1A. We supervised the clinical trial sample categories by operative status (baseline v post-operative) as well as the nature of the plasma sample in relation to fasting (fasting v postprandial). We performed unsupervised clustering of metabolites as shown in the upper right dendrogram to allow for self-segregation of metabolites with similar profiles. To this end, two major clusters of metabolites emerged. We observe a first cluster of high fasting levels of lipid families including prostaglandins (PG), phosphatidylethanolamines (PE), phosphatidic acid (PA), phosphatidylserines (PS) and docosahexaenoic acid (DHA)-derived lipids that decrease postprandially in all plasma samples, regardless of surgical status. A second cluster includes low fasting levels of FFA, decorated lipids and lipokines such as 12,13-diHOME, that in turn increase postprandially. Venn Diagram analysis of the metabolite profiles (Figure 1B) shows 73 metabolites (5 unique) in the baseline fasting state, 59 metabolites (5 unique) in the baseline postprandial state, 74 metabolites (3 unique) in the postoperative fasting state and 56 metabolites (7 unique) in the postoperative prandial state (Figure 1B). A principal component analysis of the metabolite profiles, demonstrated in Figure 1C, yielded three distinct clusters that segregate based on nutritional status. The first principal component accounts for 17.5% of the variance, while the second and third principal components account for 8.5% and 5.9% of the variance, respectively. Metabolites identified from postoperative postprandial samples tended to cluster together, as did metabolites identified from the fasted plasma samples cluster, with significant overlap between the baseline and post-operative fasting samples.

Figure 1 -. Plasma metabolomic survey of LAGB-induced weight loss.

A. Targeted plasma metabolomic heatmap by prandial and operative condition.

B. Venn Diagram highlighting overlapping and differing plasma metabolites by operative and prandial status.

C. Principal Component Analysis of plasma metabolomic signatures yields three distinct clusters that segregate by nutritional status.

To further investigate these clusters and additional known metabolic changes with LAGB, we focused on known lipid, AA and glucose-derived metabolites while further exploring the lipidomic signatures borne out of the initial unsupervised clustering. Baseline serum levels of specific FFA species were compared to post-operative specimens obtained under fasting conditions and after a 1-hour mixed meal test (Figure 2). Compared to baseline, surgical intervention resulted in significant decreases in serum palmitic acid, stearic acid, and oleic acid levels both fasting and postprandially (Figure 2A–C). Although on average linoleic acid levels were also lower postoperatively, these changes were not significant (p = 0.053 for fasting baseline vs fasting post-op and p = 0.07 for postprandial baseline vs postprandial post-op). Similar trends were seen towards lower post-operative levels of serum free triglycerides and cholesterol, though these differences did not reach significance (Figure 2E–F). While there were no significant differences between fasting and postprandial FFA levels at baseline, postoperative samples had significantly decreased levels of free palmitic acid, oleic acid, linoleic acid, and triglycerides after the mixed meal test compared to fasted samples (Figure 2A, 2C, 2D, 2E).

Figure 2 – Detected free lipids are primarily modulated by surgical status.

Plasma free palmitic acid (A), stearic acid (B), oleic acid (C), linoleic acid (D), triglycerides (E), cholesterol (F), glycerol (G), and ceramide (H) levels after a 10 hour fast and 1 hour postprandially in obese subjects before and after LAGB. Values are means ± SEM. * denotes statistically significant results (p ≤ 0.05), ** denotes statistically significant result (p ≤ 0.01), *** denotes statistically significant result (p ≤ 0.001).

Glycerol levels showed an opposite trend, with higher postprandial levels both at baseline and postoperatively when compared to fasting levels of the same operative status (Figure 2G). However, these changes were not significant. No significant change in ceramide levels was seen in any condition tested (Figure 2H).

Effects of LAGB and feeding on serum free AA levels were highly variable, with certain AA being affected more by operative status and others by prandial status (Figure 3). Figure 3A–E highlights those AAs whose serum levels were more heavily influenced by operative status rather than prandial status. Both alanine and glycine exhibited a trend towards decreased levels after surgery (Figure 3A and 3B, respectively), with a decrease in postprandial postoperative levels when compared to postprandial baseline levels that approached significance (P=0.059 and P=0.067, respectively). Cystine levels fluctuated in a similar pattern (Figure 3E), whereas glutamate and arginine showed a reverse pattern, significantly increasing postoperatively (Figure 3C, D). Serum free valine, histidine, methionine, glutamine, asparagine, cysteine, tyrosine, proline, and phenylalanine levels oscillated with prandial status, typically increasing postprandially (Figure 3F–O). Leucine, which is a branched chain amino acid, followed this pattern as well, but was also affected by operative status, significantly decreasing post-operatively (Figure 3G). Tryptophan, serine, lysine, and aspartic acid levels were unaffected by either operative or prandial status (Figure 3P–S).

Figure 3 – Detected free amino acids show prandial and body status variation.

Plasma free L-alanine (A), L-glycine (B), L-glutamate (C), L-arginine (D), L-cystine (E), L-valine (F), L-leucine (G), L-histidine (H), L-methionine (I), L-glutamine (J), L-asparagine (K), L-cysteine (L), L-tyrosine (M), L-proline (N), L-phenylalanine (O), L-tryptophan (P), L-serine (Q), L-lysine (R), and L-aspartic acid (S) levels after a 10 hour fast and 1 hour postprandially in obese subjects before and after LAGB. Values are means ± SEM. * denotes statistically significant results (p ≤ 0.05), ** denotes statistically significant result (p ≤ 0.01), *** denotes statistically significant result (p ≤ 0.001). # denotes p=0.059 for alanine and P=0.067 for glycine.

In order to better understand the dynamics of glucose handling under these conditions, levels of various ketone bodies and glucose-derived metabolites were measured (Figure 4). α-ketobutyric acid was unaffected by operative status or prandial condition (Figure 4A). In contrast, β-hydroxy-β-methylbutyric acid (HMB), which is a byproduct of leucine metabolism, was significantly decreased post-operatively, both in fasting and postprandial conditions (Figure 4B). β-hydroxybutyrate (BHB), which is synthesized in the liver from fatty acids, HMB, or ketogenic amino acids under conditions of glucose deprivation, was unexpectedly increased postprandially (Figure 4C). Though not significant, there was a decreased trend in BHB levels in fasting serum samples after LAGB (Figure 4C). Methylglyoxal is an advanced glycation endproduct (AGE) precursor that is formed primarily as a byproduct of glycolysis but can also be synthesized via the degradation of acetone and threonine. As expected, methylglyoxal levels increased postprandially but this increase was significantly blunted after LAGB (Figure 4D).

Figure 4 – Detected ketone body and preAGE levels are modulated by surgical status.

Plasma free alpha-ketobutyric acid (A), beta-hydroxybetamethylbutyric acid (B), beta-hydroxybutyrate (C), and methylglyoxal (D) levels after a 10 hour fast and 1 hour postprandially in obese subjects before and after LAGB. Values are means ± SEM. * denotes statistically significant results (p ≤ 0.05), ** denotes statistically significant result (p ≤ 0.01). # denotes p=0.09 for beta-hydroxybutyrate.

A linear-mixed effects model was employed to detect significant differences in variables of interest between pre- and post-surgical samples (Table 3). Under fasting conditions, leptin, ghrelin, alanine, β-hydroxybutyrate, β-hydroxy-β-methylbutyric acid, and 3-deoxyglucosone levels were significantly decreased after LAGB when compared to their pre-surgical baseline. Significant decreases in leptin, ghrelin, glycine, alanine, cystine, β-hydroxy-β-methylbutyric acid, 3-deoxyglucosone, and oleic acid were also identified in postprandial samples. In order to assess for relationships between percent weight loss, hormone, and plasma metabolite levels and potentially identify predictors of surgical response we used a linear regression model to compare baseline sample levels with surgical response, adjusting for age, race, and gender (Table 4). We found that at baseline, postprandial beta-hydroxybutyrate levels were directly associated with post-operative weight loss whereas fasting phenylalanine, fasting estradiol, and postprandial lysine levels were inversely correlative. A comparable analysis was done to identify associations between absolute weight loss and variables of interest at baseline, with similar results (Supplemental Table 1).

Discussion

Further understanding of the metabolic changes that contribute to weight loss and improved metabolic health after bariatric surgery can help to create a framework for the development of novel leptogenic pharmaceuticals as well as allow us to predict weight loss response in obese patients based on pre- and post-surgical biochemical characteristics. This in turn can empower a more personalized approach to bariatric medicine where specific procedures, surgical interventions, and/or medications can be recommended according to an individual’s metabolite or hormonal profile. LAGB has been shown to have comparable results to vertical sleeve gastrectomy (VSG) in terms of weight loss on 10-year followup¹⁵. While many studies have been done to identify changes in serum metabolites after RYGB and VSG procedures^6,16–20, this is the first comprehensive metabolite profiling that has been in done in patients undergoing LAGB. LAGB differs significantly from the aforementioned interventions in that it is a reversible procedure in which normal gut anatomy is maintained. Given this reversibility, LAGB has been suggested as an ideal procedure for teenagers or pregnant women with obesity in whom transient treatment of excess weight is needed^21,22. However, the metabolic and hormonal changes that have been shown with other bariatric surgeries cannot necessarily be extrapolated to LAGB patients and more specific studies are necessary to fully understand the effects of LAGB-induced weight loss in these individuals.

To this end, we have taken a targeted metabolomic approach to elucidate metabolite changes that occur after LAGB. Serum metabolite levels are constantly in flux within the body and fluctuate in response to multiple external influences including circadian rhythms, prandial status, recent exercise, and environmental stress. As such, static measurements are likely insufficient to portray the full range of metabolic transformation that occurs after a weight loss intervention. In order to better incorporate some of these dynamic changes and to more fully understand changes in glucose tolerance and nutrient handling in our study participants, we collected serum from our participants pre- and post-surgery under both fasting and postprandial conditions. These serum samples were then subjected to reverse phase LC-TOF-MS, allowing us to both qualitatively and quantitatively detect dynamic changes in over 4,000 metabolites and lipid species under these conditions.

In focusing our analysis on target lipid species, amino acids, ketone bodies, and free fatty acids, we found a varied metabolite response, with some classes of metabolites being more affected by surgical status while others were more affected by prandial condition. There was a decrease in postoperative levels of free fatty acid species, including palmitic acid, stearic acid, oleic acid, linoleic acid, and triglycerides that was most pronounced in the postprandial samples. From other studies, it seems that the effect of bariatric procedures on lipid species can be quite variable and appears to be dependent on time from initial weight loss intervention. Our post-operative samples were obtained in the early weight loss phase, at 2 months post-op. Similar effects on lipid metabolism have been shown two weeks after RYGB and, to a lesser extent, after calorie restriction-induced weight loss and are thought to be reflective of decreased body weight and subsequent improvements in insulin sensitivity resulting in a higher rate and efficiency of fatty acid oxidation¹⁷. In contrast, studies that have analyzed serum lipid levels after longer post-operative periods following RYGB have reported increases in the levels of certain fatty acid species at 6 months²⁰ and 1 year¹⁸ when compared to pre-surgical baseline. These contradictory findings may be the result of continued metabolic adaptation with increased time from initial weight loss intervention, perhaps as patients begin to approach the weight maintenance phase, or due to fundamental differences in the metabolic response to LAGB compared to RYGB. Further studies examining intra- and inter-individual fluctuations in metabolite levels at multiple post-surgical timepoints and between different surgical interventions are necessary to clarify these discrepancies.

Amino acid levels fluctuated in response to the mixed meal test, with minimal changes being seen after LAGB when compared to baseline for the majority of amino acids tested (Figure 3F–S). Notable outliers, as demonstrated in Figure 3A–E were glycine, alanine, glutamate, arginine, and cystine. While our study reports significant changes in only a small subset of amino acids after LAGB, more global effects on plasma amino acid levels and their metabolism have been reported after RYGB ^17,23. This may be reflective of the anatomic differences between the surgical techniques or due to differences in study timing, study design or data acquisition.

Alanine and glycine levels were significantly lower at the end of the meal challenge after LAGB than pre-surgically, suggesting improved peripheral utilization of these metabolites. Multiple recent studies have found that decreased fasting serum glycine levels are associated with increased branched chain amino acid (BCAA; leucine, isoleucine, and valine) levels and, consequently, with increased insulin resistance ^24–27. Improvement in biomarkers of insulin resistance is expected after clinically significant weight loss and so it is unclear how to relate this to our findings given that we saw no significant change in fasting serum glycine levels despite decreases in fasting leucine levels (Figure 3G). Surprisingly, valine levels were unchanged postoperatively (Figure 3F). It is possible that not enough time had elapsed after surgery for significant changes in the pathways that regulate fasting glycine and BCAA metabolism to occur. Still another possible explanation is that the metabolic pathways regulating these effects are significantly altered after LAGB in such a way as to uncouple the effects of BCAA on glycine metabolism, thereby lessening its association with insulin resistance. Postoperative decreases in postprandial alanine levels have also been reported after RYGB, similar to the present study, but not with calorie restriction-induced weight loss implying that the metabolic changes after LAGB may not solely be attributable to restricted food intake ¹⁷. Decreases in alanine levels have been associated with diabetes remission after RYGB ¹⁹. Our study cohort consisted of only two patients with diabetes so data correlating alanine levels with diabetes remission after LAGB are not available for the present study.

Glutamate and arginine levels were significantly higher at the end of the meal challenge after LAGB than pre-surgically. A postsurgical increase in glutamate levels has also been reported 2 weeks ¹⁷, 1 month²⁸ and one year ^18,19 after RYGB and VSG but not after calorie restriction-induced weight loss ¹⁷. In cell culture models, glutamate uptake into insulin granules has been shown to amplify incretin-induced insulin secretion in pancreatic beta cells via a mechanism that links glucose metabolism to cAMP action ²⁹. Additionally, the postprandial kinetics of GLP-1 and glutamate appearance and clearance after a mixed meal in patients who underwent RYGB suggests that a postsurgical increase in serum glutamate may augment GLP-1 response¹⁷. Thus, while absolute levels of GLP-1 were unchanged in our study subjects after LAGB¹⁴, our data suggest the potential for metabolite-induced changes in incretin activity that confer additional metabolic benefits beyond changes in the hormone levels themselves. Arginine variation after bariatric procedures is largely unreported in the majority of metabolomic studies but the data that we do have is mixed ^16,17. Long term dietary supplementation with L-arginine has been shown to improve hepatic and peripheral insulin resistance in type 2 diabetic patients via modulation of nitric oxide homeostasis³⁰. How this relates to systemic plasma arginine levels in the absence of increased dietary intake is unclear, however it provides a potential mechanism for increased serum arginine levels in contributing to an improved metabolic phenotype after bariatric surgery.

The levels of various ketone bodies and glucose-derived metabolites were measured in order to understand how glucose handling and glucose metabolism changes post-operatively in both the fasted and fed states. α-ketobutyric acid levels were unchanged by either surgical or prandial status. This is surprising given that α-ketobutyric acid levels have been associated with increased insulin resistance, likely due to increased production of α-hydroxybutyrate, which is formed as a byproduct of α-ketobutyrate metabolism ²⁶. α-hydroxybutyrate is known to be an early biomarker of insulin resistance and its levels track proportionally with other measures of dysglycemia ^26,31. A study examining metabolic changes during the two week pre-operative liquid or restricted diet period and one month after bariatric surgery showed that patients experience a transient increase in α-ketobutyric acid levels during the two week pre-operative liquid or restricted diet, likely due to increased metabolic stress during this time, and return to pre-surgical baseline levels at one month post-op ²⁸. α-ketobutyric acid levels have not been reported at longer post-surgical time periods so we cannot directly compare our findings to this study. HMB, a leucine metabolite that is a component of the signaling cascade leading to mTOR phosphorylation and activation^32,33, was significantly decreased postoperatively (Figure 4B). Coupled with decreases in postoperative serum leucine levels (Figure 3G), this may reflect decreased flux through this pathway after LAGB. Decreases in postoperative BHB and methylglyoxal levels are consistent with improved insulin sensitivity and glucose disposal after LAGB (Figure 4C–D). Population studies have found that BHB levels correlate significantly with fasting plasma glucose and are inversely associated with insulin sensitivity³⁴. Methylgloxal is a byproduct of glycolysis that is directly involved in the formation of advanced glycation end-products (AGEs) and has been implicated in promoting atherogenesis and neuropathy in patients with diabetes^35,36. Thus, decreases in the levels of both of these metabolites are expected with an improved metabolic phenotype.

Linear modeling was used to assess for relationships between both percent and absolute weight loss and metabolites of interest at baseline (Table 4, Supplemental Table 1). Clinically meaningful weight loss is defined in percent of total body weight lost³⁷ and so we have chosen to focus our analysis on the associations identified for percent rather than absolute weight loss in order to better relate our findings to these standard definitions. Postprandial beta-hydroxybutyrate levels were directly associated with percent weight loss, whereas there was an inverse relationship between weight loss success and fasting phenylalanine, fasting estradiol, and postprandial lysine levels. These metabolites were largely unchanged postoperatively indicating them as markers for susceptibility to weight loss success after LAGB rather than drivers of disease pathology. It will be interesting to see if this metabolite signature bears out as being predictive of surgical response when examined in larger cohorts and for other bariatric interventions. Additionally, it would be helpful to determine whether these or other metabolites can also predict improvement in measures of obesity-related comorbidities, including dysglycemia, hypertension, hyperlipidemia, and coronary artery disease after LAGB.

This study has several limitations. Our study population was small, with only 18 individuals completing the study, and so it is likely that we were unable to detect smaller changes, especially given the degree of inter-individual variability in metabolite levels between the subjects. A larger cohort might have revealed more significant differences between the various conditions tested. Additionally, our sample group was limited to LAGB patients and there was no non-surgical control group, thus we are unable to differentiate which, if any, metabolite changes are direct sequelae of LAGB and would not be present after weight loss achieved via other mechanisms.

Future studies comparing metabolite profiles after LAGB to those after calorie restriction-induced weight loss or to weight loss induced by other bariatric procedures would be helpful in making this distinction. Though we can use these data to generate hypotheses regarding metabolic changes after LAGB-induced weight loss, these are static measurements and cannot be used to garner information about metabolic flux or make definitive conclusions about interactions between hormonal and metabolic pathways.

Despite these limitations, our study provides insight into the metabolic changes that occur after LAGB. Due to the lack of anatomic alteration, LAGB-induced weight loss has been likened more to calorie restriction-induced weight loss than the weight loss that occurs after more invasive bariatric procedures. However, we identified multiple metabolite changes that mirror those that occur after RYGB but not calorie restriction-induced weight loss, suggesting that the effects of LAGB likely fall somewhere between the two. Given the degree of weight loss achieved by our study subjects, it is not surprising that many of the metabolite changes identified are consistent with a post-surgical improvement in insulin sensitivity. This cannot be explained by post-operative changes in the levels of insulin or insulin-sensitizing hormones, such as GIP and GLP-1,¹⁴ and so further investigation under controlled conditions such as euglycemic-hyperinsulinemic clamp will likely be necessary to understand how changes in these metabolic pathways evoke changes in hormone function and activity as well as how they relate to acute and sustained weight loss after surgery.

Conclusions

In conclusion, our study is the first comprehensive metabolite profiling that has been in done in patients undergoing LAGB under dynamic conditions. In our focused analysis, we identified significant changes in the levels of fatty acids, ketone species, and advanced glycation endproducts postoperatively that are consistent with improvements in insulin sensitivity. Amino acid levels, in contrast, fluctuated with prandial status over changes in body weight or surgical status. Future studies are warranted to understand how these post-surgical adaptations relate to longitudinal health outcomes in this patient population.

Supplementary Material

MMC1

Supplemental Table 1 – P-values based on linear model of weight loss (lb) regressed by the variables of interest at baseline, adjusted for age, race and gender.

Highlights:

• We profiled metabolite changes following gastric banding with mixed meal tests.

• Weight loss by gastric band changes lipid metabolites in proportion to body weight.

• Gastric banding changes amino acids and glucose metabolites in relation to feeding.

• Beta-hydroxybutyrate and lysine associate with weight loss by gastric banding.