The sulphur amino acids (SAA) are fundamental to protein synthesis. Both methionine and cysteine are proteinogenic amino acids; methionine is required for translation of all eukaryotic proteins, and cysteine contributes to protein structure by forming disulphide bonds(Reference Brosnan and Brosnan1). Although it has not been established that SAA can contribute to the development of obesity, methionine and cysteine intakes can promote growth and weight gain in mammals(Reference Byington, Howe and Clark2, Reference Sarwar Gilani, Nimal Ratnayake and Peace3). In humans, plasma total cysteine (tCys) levels are positively associated with BMI and obesity(Reference El-Khairy, Ueland and Nygard4–Reference Elshorbagy, Nurk and Gjesdal6), mediated through fat mass(Reference Elshorbagy, Nurk and Gjesdal6). Among men and women in a population-based study, an increase from low to high tCys quintile was associated with an increase in fat mass of 6–9 kg, after adjustment for age, diet, lifestyle factors and serum lipids. Longitudinal studies have revealed that an increase or decrease in tCys over 6 years was linked with corresponding changes in BMI and fat mass(Reference Elshorbagy, Nurk and Gjesdal6, Reference El-Khairy, Vollset and Refsum7).
Obesity is also associated with changes in compounds metabolically related to cysteine. Plasma glutathione (GSH) correlates negatively with BMI(Reference Kennedy, Rao and Botiglieri8), while the amino acid glutamate is elevated in the blood(Reference Proenza9) and plasma(Reference Newgard, An and Bain10) of obese individuals. Glutamate is linked to cysteine via GSH metabolism: cysteine and glutamate combine to form γ-glutamylcysteine in the rate-limiting step of GSH synthesis(Reference Lu11). The membrane enzyme γ-glutamyltransferase (GGT) hydrolyses GSH, with the subsequent release of cysteine and glutamate(Reference Whitfield12). Serum GGT activity is elevated in obese persons(Reference Elshorbagy, Refsum and Smith5, Reference Lee, Silventoinen and Jacobs13), and decreases with weight loss(Reference Dixon, Bhathal and O'Brien14). Conversely, plasma concentrations of the cysteine precursor, homocysteine, often increase after weight loss(Reference Dixon, Dixon and O'Brien15).
Despite the diverse evidence pointing to a link between cysteine metabolism and body weight, the effect of weight loss on cysteine metabolism is largely unexplored. The causal direction for the relationship between tCys and fat mass remains uncertain: does tCys promote body fat accumulation; or does fat mass influence cysteine production, catabolism, uptake or release? In favour of tCys promoting fat accumulation, intake of cysteine-rich diets has been associated with weight gain both in animal models(Reference Elshorbagy, Valdivia-Garcia and Refsum16, Reference Pollack, Rivera and Rassin17) and in patients with cancer(Reference Tozer, Tai and Falconer18). However, recent studies on the role of adipose tissue in regulating cysteine conversion to taurine(Reference Stipanuk, Ueki and Dominy19) have suggested that adipose tissue mass could be an important determinant of cysteine and taurine levels. The hypothesis that fat mass determines plasma tCys has not been addressed in humans. Severely obese patients undergoing bariatric surgery may be an ideal population for testing the hypothesis: if fat mass determines tCys, then tCys concentrations would be expected to be markedly elevated before and drop significantly after bariatric surgery due to major weight loss. We, therefore, studied the effects of bariatric surgery in obese patients on serum concentrations of tCys and metabolically related compounds, such as methionine, homocysteine, cystathionine, taurine, GSH and glutamate, and serum GGT activity.
Experimental methods
The patients have been described in detail previously(Reference Aasheim, Bjorkman and Sovik20). Briefly, sixty obese patients aged 20–50 years, from two centres (Oslo University Hospital Aker, n 30; Sahlgrenska University Hospital, n 30), were included in a randomised trial of gastric bypass and duodenal switch for weight loss (February 2006–August 2007; ClinicalTrials.gov identifier: NCT00327912). All patients had a BMI between 50 and 60 kg/m2. In addition to the baseline visit, the patients were examined 6 weeks, 6 months and 1 year after surgery. Patients followed a 4184 kJ (1000 kcal) diet for 3 weeks before surgery. During surgery, the alimentary limb, biliopancreatic limb and common channel were created as follows: gastric bypass – 150, 50 cm and variable; duodenal switch – 200 cm, variable and 100 cm(Reference Aasheim, Bjorkman and Sovik20). After surgery, all patients were daily prescribed Fe, Ca, vitamin D and multivitamin/mineral supplements (selected contents: 2 mg vitamin B6, 200 μg folic acid and 1 mg vitamin B12 (Nycoplus multi; Nycomed, Asker, Norway)). Gastric bypass patients also received cyanocobalamin supplementation (intramuscular injections, 1 mg every 3 months (Betolvex; Actavis, Oslo, Norway) or oral supplement, 1 mg daily (Behepan; Pfizer, Sollentuna, Sweden)). Ursodeoxycholic acid was given until 6 months after surgery. Duodenal switch patients were more often taking additional fat-soluble vitamin supplements, but otherwise, supplement use was similar in the two groups(Reference Aasheim, Bjorkman and Sovik20).
The controls were fifty-eight healthy men and women with BMI between 17 and 31 kg/m2 and aged 19–59 years, recruited January–May 2007 at Oslo University Hospital Aker(Reference Aasheim, Hofso and Hjelmesaeth21). The controls were recruited as a convenience sample, with no attempt made to match controls and patients. Exclusion criteria were chronic disease and regular medication or multivitamin use. Persons using contraceptive medication (n 10) or thyroxine substitution (n 4) were included. As a population-based control group for plasma tCys and BMI comparisons (see Fig. S1 of the supplementary material, available online at http://www.journals.cambridge.org/bjn), we used the Hordaland Homocysteine Study participants (n 3721, men and women aged 47–49 years)(Reference Elshorbagy, Nurk and Gjesdal6).
The study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the appropriate regional ethics committees for medical research. Written informed consent was obtained from all subjects.
Biochemical analysis
Overnight fasting blood samples were clotted for 30 min at room temperature, and serum was separated by centrifugation. Samples collected at Oslo University Hospital Aker were frozen and stored at − 80°C until assayed. Samples collected at Sahlgrenska University Hospital were thawed once for the analysis of 25-hydroxyvitamin D(Reference Aasheim, Bjorkman and Sovik20) and were subsequently stored at − 20°C until assayed. Assays were performed within 2 years of blood sampling.
Serum concentrations of SAA and glutamate were analysed by liquid chromatography–tandem MS using modifications of described methods(Reference Elshorbagy, Valdivia-Garcia and Refsum16, Reference Antoniades, Shirodaria and Leeson22). Methionine, total homocysteine (tHcy), cystathionine, tCys and total glutathione (tGSH) were analysed in a single run. A separate assay was used to simultaneously measure taurine and glutamate. Inter-assay CV was < 4 % for tCys, tHcy and taurine; < 8 % for methionine, cystathionine and tGSH; and 10·5 % for glutamate. Serum cobalamin was determined using Lactobacillus leichmannii microbiological assays(Reference Refsum, Johnston and Guttormsen23).
Serum creatinine, alanine aminotransferase, GGT and folate were measured during the follow-up with Hitachi (717 or 800) Modular multianalysers (Boehringer Mannheim, Mannheim, Germany)(Reference Aasheim, Bjorkman and Sovik20). Serum pyridoxal-5′-phosphate (vitamin B6) was measured by HPLC (Chromsystems, Munich, Germany). Plasma tHcy concentrations measured during the follow-up in some patients (using Hitachi 717 multianalyser) correlated well with the reported serum concentrations obtained by liquid chromatography–tandem MS (Pearson's r 0·95, P < 0·001; n 18).
In total, frozen serum samples were available from 223 out of 240 study visits (93 %). As outlined above, serum handling and storage varied between the study sites (Oslo and Sahlgrenska). In agreement with our laboratory's experience that sample processing and freeze–thaw cycles may influence assay values, the serum concentrations for methionine, taurine, tGSH and glutamate differed markedly between the centres. We, therefore, report data on those metabolites only for patients with specimens kept at − 80°C (n 30; followed in Oslo), and data on tHcy, cystathionine and tCys from all patients (n 60).
Statistical analysis
For cross-sectional analyses, we compared patients and controls separately for men and women, using Fisher's exact test for proportions and unpaired Student's t tests for continuous data. These tests, correlation analyses and multiple linear regressions were calculated using SPSS 14.0 (SPSS, Inc., Chicago, IL, USA).
For longitudinal analyses, we used linear mixed-effects models (with a random intercept and assuming compound symmetric correlation structure) for continuous dependent variables to assess a trend (time effect) across the four time points in each patient. Estimates of fixed-effects coefficients and their standard errors were extracted after fitting the model. Standardised values of the outcome variables were computed as predicted values from the model. These predictions were done for each time point in a new dataset, where potential confounders (age, sex and type of surgery) were replaced by their overall average values in the full dataset. Standard errors of the predicted values were estimated by assuming the predictors as non-random and using the estimated variance–covariance matrix for the fixed-effects estimates. Non-normally distributed variables (including tHcy, cystathionine, cobalamin, creatinine, alanine aminotransferase and GGT) were log-transformed before analysis. Data are expressed as standardised means or geometric means with 95 % CI. Linear mixed-effects models (lme function in the R-library nmle) were computed in R 2.8.1 for Windows(24). The P values are two-sided, and the significance level was 0·05.
Results
Participant characteristics
Most participants (97 %) were of North European descent. Mean age was not significantly different between patients and controls, who had a mean BMI of 55 and 24 kg/m2. Patients had lower serum pyridoxal-5′-phosphate and folate concentrations than controls (Table 1).
ALT, alanine aminotransferase; GGT, γ-glutamyltransferase; PLP, pyridoxal-5′-phosphate; tHcy, total homocysteine; tCys, total cysteine; tGSH, total glutathione.
* Comparisons by the t test or Fisher's exact test. ALT, GGT, PLP, folic acid, vitamin B12, homocysteine and cystathionine were compared using log-transformed values.
† Data available only from the Norwegian patients in the study (twenty-two women and eight men).
‡ One male control (outlier) was excluded from comparisons of GGT and glutamate.
Total cysteine and related compounds in obese patients and healthy controls
Before surgery, the severely obese patients had significantly but modestly (approximately 17 %) higher serum tCys concentrations than controls. Compared with controls, the obese patients also had significantly higher serum concentrations of amino acids both upstream (tHcy) and downstream (taurine) of cysteine, as well as of glutamate (Table 1).
Effect of bariatric surgery on total cysteine and related compounds
In the same obese patients, serum tCys and related compounds were also measured 6 weeks, 6 months and 1 year after a bariatric surgical procedure, which was either gastric bypass or duodenal switch. Unless stated otherwise, patients in the two surgical groups had similar trends in biomarker values during the follow-up (i.e. no significant time × procedure interaction was found).
SAA upstream of cysteine changed considerably after bariatric surgery (Fig. 1). Methionine concentrations declined (P < 0·001), with the lowest concentrations (79 % of pre-surgery levels) measured 6 months postoperatively. Trends for serum tHcy concentrations differed in the two surgical groups (P < 0·001 for time × procedure interaction): concentrations increased during weight loss after duodenal switch (P < 0·001) but did not change significantly after gastric bypass (P = 0·63). Cystathionine dropped steeply from baseline to 6 weeks after surgery (by approximately 36 %) and then remained stable (P < 0·001). Trends for serum tCys differed according to the surgical procedure (P = 0·02 for time × procedure interaction): duodenal switch patients showed a gradual decline in tCys concentrations, reaching 12 % decrease 1 year post-surgery (P < 0·001), while gastric bypass patients showed no significant change (P = 0·22). Duodenal switch patients did not have significantly higher tCys than controls 1 year after surgery (P = 0·11).
Biomarkers downstream of cysteine are shown in Fig. 2. No significant changes were found for concentrations of taurine (P = 0·51) or tGSH (P = 0·98) during the follow-up. However, glutamate concentrations dropped briskly after surgery (P < 0·001). Serum GGT activity also decreased (P < 0·01 in both surgical groups); this trend was more pronounced after gastric bypass than duodenal switch (P = 0·03 for time × procedure interaction).
Effect of bariatric surgery on other variables
For the same patients, we have previously reported 1-year outcomes for the following variables: mean weight loss (30 % after gastric bypass and 41 % after duodenal switch); serum folate and pyridoxal-5′-phosphate (stable and increased concentrations after both procedures); Hb, total cholesterol and plasma albumin (greater decreases after duodenal switch)(Reference Aasheim, Bjorkman and Sovik20).
Further variables relevant to SAA metabolism are shown in Fig. 3. Serum cobalamin concentrations increased in gastric bypass patients, who received additional vitamin B12 supplementation. Blood FAD concentrations were mostly stable after surgery (P = 0·054). Serum creatinine declined (P < 0·001). Alanine aminotransferase activity increased transiently 6 weeks after surgery, before declining (P = 0·005).
Associations of serum total cysteine and related compounds with other variables
The relationships (Spearman's correlations) of tCys and related compounds with other biomarkers are shown in Table S1 (supplementary material for this article can be found at http://www.journals.cambridge.org/bjn). No consistent associations were found with BMI. However, at all patient visits, serum GGT activity correlated positively with glutamate (ρ = 0·79 to 0·84) and inversely with tGSH (ρ = − 0·41 to − 0·76) concentrations (Fig. 4). Significant correlations were also noted for creatinine (with tHcy and cystathionine), alanine aminotransferase (with glutamate), folate (with tHcy) and albumin (with tCys).
Mixed-effects models identified the following associations with tCys and related compounds in the patients: age was associated positively with tCys (P = 0·001) and inversely with tGSH (P = 0·03), while male sex was associated with higher methionine (P = 0·003), tHcy (P = 0·002), cystathionine (P = 0·02) and tCys (P = 0·03) concentrations.
Relationship between total cysteine and weight loss
The relationship between tCys and weight loss was explored in multiple linear regression models, using percentage of weight loss (from baseline to 1 year after surgery) as a dependent variable and surgical procedure (gastric bypass or duodenal switch) as well as baseline BMI, age, sex and tCys as independent variables. In this model (r 2 0·59), the only significant predictors of weight loss were surgical procedure and baseline BMI (P < 0·001 for both) (baseline tCys: P = 0·43). Replacing baseline tCys with change in tCys (from baseline to 1 year after surgery) did not change the findings (r 2 0·59; change in tCys: P = 0·64).
Total cysteine in bariatric surgery patients and the general population
Finally, we compared the obese patients with a large population-based cohort. Before surgery, the obese patients had a BMI much higher than the 95th percentile (32 kg/m2) of the Hordaland Homocysteine Study population, while serum tCys was below the 95th percentile. After surgery, duodenal switch patients had a modest but significant decline in tCys, whereas the gastric bypass patients had no significant change in tCys despite a BMI loss of 16·3 kg/m2 (see Fig. S1 of the supplementary material, available online at http://www.journals.cambridge.org/bjn).
Discussion
In the present study, we showed changes in serum tCys and metabolically related amino acids in severely obese patients undergoing surgery-induced weight loss. The patients had modestly elevated tCys concentrations before surgery. At 1 year after surgery, gastric bypass patients showed no significant change in tCys, despite 30 % weight loss. Given the strong association between tCys and fat mass in the general population(Reference Elshorbagy, Nurk and Gjesdal6), the present data suggest that fat mass does not cause elevation of tCys, since these very obese patients neither had very high tCys at baseline nor showed a marked drop in tCys after 30 % weight loss. This contrasts with glutamate concentrations, which were markedly elevated before surgery and decreased with both procedures; reflecting the changes in GGT.
It has been known for decades that obese persons may have elevated plasma amino acid levels(Reference Felig, Marliss and Cahill25). Given that type of dietary protein influences the risks of having obesity and insulin resistance(Reference Rosell, Appleby and Spencer26, Reference Tremblay, Lavigne and Jacques27), it is conceivable that specific amino acids could contribute to causing these conditions. Recent interest in this field has been fuelled by the finding that certain plasma amino acid profiles, in particular elevated branched-chain amino acid (BCAA) and glutamate levels, may promote insulin resistance in obesity(Reference Newgard, An and Bain10, Reference Tremblay, Lavigne and Jacques27). Then again, obese rodents with elevated BCAA levels had a reduction in BCAA-metabolising enzymes; and in obese humans, after gastric bypass, a decrease in plasma BCAA was coupled with an increase in BCAA-metabolising enzymes(Reference She, Van Horn and Reid28). Hence, the elevated plasma BCAA in obesity could be due to disturbances in metabolising enzymes.
The present study targets a different group of amino acids, the SAA. SAA are relevant to clinical obesity not only because of the association of tCys with BMI and fat mass in large cohort studies(Reference El-Khairy, Ueland and Nygard4–Reference Elshorbagy, Nurk and Gjesdal6), but also because rodent experiments show that restricted intake of methionine, the precursor of all SAA, protects against obesity and insulin resistance(Reference Rizki, Arnaboldi and Gabrielli29, Reference Malloy, Krajcik and Bailey30). Furthermore, adding cysteine to a methionine-restricted diet reverses the effect of the diet on rat adiposity and the serum fatty acid profile(Reference Elshorbagy, Valdivia-Garcia and Mattocks31), raising the possibility that cysteine may influence lipid metabolism and weight in humans.
To our knowledge, these are the first data on tCys levels in patients undergoing current bariatric procedures. Early studies found high tCys and glutamic acid levels in obese patients, with varying changes after jejunoileal bypass surgery(Reference Backman, Hallberg and Kallner32, Reference Kihlberg, Bark and Hallberg33). Jejunoileal bypass more often led to metabolic complications (e.g. liver cirrhosis)(Reference Hocking, Duerson and O'Leary34), which may hamper comparison with modern operations. More recent studies have found both increases(Reference Dixon, Dixon and O'Brien15, Reference Borson-Chazot, Harthe and Teboul35) and decreases(Reference Sledzinski, Goyke and Smolenski36, Reference Williams, Hagedorn and Lawson37) in tHcy in obese patients following bariatric surgery; perhaps related to differences in vitamin supplements, or, as in the present study, the procedure used.
Possible mechanisms for the changes in total cysteine and related compounds
Methionine and cysteine are normally efficiently digested and absorbed in the intestines, and are removed from the portal blood by the liver for use in the synthesis of proteins and GSH, or for catabolism to taurine(Reference Stipanuk, Ueki and Dominy19) (ses Fig. S2 of the supplementary material, available online at http://www.journals.cambridge.org/bjn). While gastric bypass patients presumably lose weight mainly by lowering energy intake, biliopancreatic diversion with duodenal switch causes more protein and fat malabsorption(Reference Puzziferri, Blankenship and Wolfe38, Reference Scopinaro, Marinari and Camerini39) and more often leads to nutrient deficiencies(Reference Aasheim, Bjorkman and Sovik20). Protein depletion and methionine restriction lead to impaired cystathionine β-synthase enzyme activity: this limits trans-sulphuration of homocysteine to cystathionine to preserve methionine(Reference Ingenbleek, Hardillier and Jung40, Reference Tang, Mustafa and Gupta41). Sparse protein availability in duodenal switch patients could thus help explain their higher tHcy concentrations and, secondary to diminished flux through the trans-sulphuration pathway, their lower tCys concentrations. Elevated tHcy could potentially in itself also contribute to decreasing serum tCys by displacing protein-bound cysteine from albumin(Reference Mansoor, Svardal and Schneede42). Finally, given that the gastrointestinal tract is an active site for SAA metabolism(Reference Bauchart-Thevret, Stoll and Burrin43), it is possible that differences in tCys after gastric bypass and duodenal switch could in part relate to varying intestinal SAA metabolism following these procedures.
Changes in dietary intakes and food preferences(Reference Olbers, Bjorkman and Lindroos44) and use of vitamin supplements could have contributed to changes in SAA concentrations after surgery. It has been suggested that higher serum cobalamin and folate levels are needed to maintain tHcy during weight loss(Reference Dixon, Dixon and O'Brien15). Cobalamin is a cofactor in homocysteine remethylation to methionine(Reference Brosnan and Brosnan1). Duodenal switch patients did not receive parenteral cobalamin supplementation, so a relative cobalamin deficiency may have contributed to the elevation of their tHcy concentrations. Both surgical groups showed increases in serum vitamin B6 postoperatively, which probably explains the abrupt 36 % reduction in cystathionine concentrations after surgery(Reference Bleie, Refsum and Ueland45): vitamin B6 supplementation increases the conversion of cystathionine to cysteine by the cystathionine γ-lyase enzyme, which is very sensitive to pyridoxal-5′-phosphate depletion(Reference Martinez, Cuskelly and Williamson46).
Taurine can be obtained from the diet, or via conversion of cysteine by cysteine dioxygenase(Reference Ueki and Stipanuk47). Since cysteine dioxygenase is up-regulated in the liver and adipose tissue in response to high cysteine availability(Reference Stipanuk, Ueki and Dominy19), the higher taurine concentrations in obese patients compared with controls could be secondary to increased conversion of cysteine to taurine. Glutamate was much higher in obese patients than controls, as shown by others(Reference Proenza9, Reference Newgard, An and Bain10), and decreased after surgery. Elevated GGT levels in obesity, and their decrease during weight loss, also correspond with previous research(Reference Dixon, Bhathal and O'Brien14). Perhaps surprisingly, despite losing more weight, duodenal switch patients had higher GGT levels than gastric bypass patients during the follow-up. This could reflect either higher alcohol intake, or less regression of non-alcoholic fatty liver disease(Reference Mathurin, Hollebecque and Arnalsteen48) in duodenal switch patients relative to gastric bypass patients post-surgery. Strong correlations were found for GGT with tGSH (inverse) and glutamate (positive) before and after surgery. Despite a decline in GGT activity, however, patients had stable tGSH concentrations. Other researchers reported increased erythrocyte GSH after gastric banding(Reference Uzun, Konukoglu and Gelisgen49). The varying results could relate to our assessment of tGSH in the serum; and not erythrocytes, where GSH is more abundant. Our finding of a strong correlation (r 0·8) between serum GGT activity and glutamate concentrations may reflect the role of GGT in hydrolysing GSH to ultimately yield glutamate (and cysteine), and could suggest that GGT influences plasma glutamate. Increased glutamate levels in obesity may thus be secondary to disturbances in metabolic enzymes, similar to BCAA(Reference She, Van Horn and Reid28).
A longitudinal study of predictors of change in tCys in a population-based cohort showed that a decrease in BMI of 1 kg/m2 or more was associated with a corresponding decrease in tCys(Reference El-Khairy, Vollset and Refsum7). In the present trial, despite 30 % weight loss, there was no significant change in tCys after gastric bypass surgery. This suggests that the epidemiological association between changes in tCys and changes in BMI(Reference El-Khairy, Vollset and Refsum7) is not explained by an effect of body weight on plasma tCys levels. The 12 % decrease in tCys seen in the present study after duodenal switch surgery may relate to additional malabsorption with this procedure(Reference Puzziferri, Blankenship and Wolfe38, Reference Scopinaro, Marinari and Camerini39). Consistent with this, duodenal switch patients had lower serum albumin than gastric bypass patients(Reference Aasheim, Bjorkman and Sovik20), and serum albumin correlated positively with tCys in the combined patient group.
The strengths of the present study include a broad biochemical characterisation of patients throughout major weight loss. Unlike previous reports on other amino acids(Reference She, Van Horn and Reid28), we did not investigate changes in metabolic enzymes, apart from plasma GGT. However, the serial examinations enabled us to distinguish between amino acid changes that occurred rapidly (e.g. decrease in cystathionine) or more gradually (e.g. decrease in glutamate) during weight loss. The randomised trial design allowed us to compare the effects of two different surgical techniques. We did not assess dietary SAA content, but monitored the intake of dietary supplements(Reference Aasheim, Bjorkman and Sovik20), and use of fasting blood samples minimised bias from acute variations in SAA intake(Reference Guttormsen, Solheim and Refsum50).
In summary, the stability of tCys during weight loss contrasts with the marked drops in serum methionine, cystathionine and glutamate. This may illustrate that elevated plasma amino acid levels associated with obesity fall into at least two categories: those that are an adjunct to obesity and those that may contribute towards the development of obesity. This can potentially have implications for the modulation of dietary protein content in future strategies to prevent obesity.
Acknowledgements
The authors’ responsibilities were as follows: E. T. A. (guarantor), A. K. E. and H. R. organised the amino acid studies; E. T. A., T. T. S., T. M. and T. O. organised the clinical trial; E. T. A. and K. I. B. organised examination of the healthy controls; M. V.-G. performed the amino acid assays; E. T. A., A. K. E., L. M. D. and H. R. cross-checked and analysed the data; E. T. A. and L. M. D. performed the statistical analyses; E. T. A. and A. K. E. wrote the manuscript; L. M. D., T. T. S., T. M., M. V.-G., T. O., T. B., K. I. B. and H. R. revised the manuscript; E. T. A., T. B. and H. R. obtained funding; H. R. supervised the amino acid studies. The authors acknowledge Carl Fredrik Schou, Jon Kristinsson and Hans Lönroth (bariatric surgery); and My Engström and Kari Julien (biobanking). The present study was supported by a research fellowship grant from the South Eastern Norway Regional Health Authority and research grants from the South Eastern Norway Regional Health Authority (AUS-20-2006) and Aker University Hospital (08/835 and 09/40) (all to E. T. A.), and the Norwegian Research Council (197195/HR). The sponsors had no role in designing the study or in collecting, analysing or interpreting the data. The authors declare that they have no conflict of interest.