Introduction
According to the World Health Organization, obesity prevalence has almost tripled since 1975, with over 650 million adults worldwide suffering from this condition in 2016 [
1]. Obesity is related to many comorbid conditions including cardiovascular disease, type 2 diabetes, non-alcoholic fatty liver disease, obstructive sleep apnea, reproductive dysfunction, musculoskeletal disorders, certain types of cancer, as well as psychosocial consequences [
2,
3]. Unfortunately, non-surgical treatment of severe obesity using modalities such as diet, exercise, or even pharmacological treatment have low to moderate success rates, especially when considering medium- to long-term weight maintenance [
4]. Bariatric surgery, also known as metabolic surgery, emerged as an effective therapy for individuals with severe obesity to achieve significant sustainable weight loss, as well as to reduce the associated comorbidities [
3,
5,
6]. The term metabolic surgery was proposed to acknowledge the physiological changes induced by these approaches, which contribute to a more favorable metabolic profile following the surgery, beyond the traditional view that these surgeries induce beneficial effects through weight loss alone [
6‐
8]. Yet, these physiological changes are still not fully elucidated.
Among the different types of bariatric surgery, sleeve gastrectomy (SG) is one the most common and simple from a surgical standpoint, consisting of a surgical removal of about 80 percent of the stomach along the greater curvature, physically restricting food intake. In addition to reducing gastric volume, Roux-en-Y gastric bypass (RYGB) also decreases the efficiency of nutrient absorption in the small intestine by creating a little gastric pouch directly connected to the jejunum, bypassing the duodenum. Biliopancreatic diversion with duodenal switch (BPD-DS) is the most complex of these bariatric procedures, combining gastric restriction induced by SG and a more marked malabsorption than that observed after RYGB. Briefly, a pylorus-preserving SG procedure is combined with a transection of the duodenum near the pylorus creating an alimentary limb that connects the biliary limb near the ileo-cecal valve to create a short common channel where nutrients are absorbed [
9].
Epigenetics may help in our understanding of how an individual will respond to bariatric surgery as the latter may be viewed as an environmental factor modifying the epigenome, although certain epigenetic marks may be inheritable [
10,
11]. DNA methylation is the most widely investigated epigenetic mechanism, and some studies have predicted weight loss or weight regain after bariatric surgery according to baseline gene methylation [
12,
13]. Bariatric surgery also promotes modifications in methylation profiles of individuals with obesity, which are more akin to those who are lean [
14,
15]
. Some authors have also observed lower overall methylation levels after RYGB in subcutaneous adipose tissue (SAT) [
16,
17]. Conversely, more recent studies have observed higher methylation levels at cytosine-phosphate-guanine dinucleotides, or CpG sites, after RYGB and SG procedures in peripheral blood [
15], as well as higher global methylation levels in skeletal muscle after RYGB [
18]. These discrepancies may partly be explained by tissue-specific DNA methylation [
19]. As compared to RYGB and SG, BPD-DS is a surgery that creates a greater nutrient malabsorption, due to a reduced length of the intestinal segment allowed for absorption, especially impacting dietary lipids [
20,
21], and it has been proven to be particularly effective among individuals with severe obesity [
22,
23]. By contrast, the impact of BPD-DS on the epigenetic profile of SAT is almost completely unknown.
The SAT is much more than a site of storage for excess energy intake, and it is recognized to play a crucial role in energy homeostasis control and inflammation [
24]
. Although significant modifications occur in the SAT after surgically induced weight loss [
25,
26], only few studies have investigated gene expression changes in this tissue [
27‐
31]. Genes previously identified as differentially expressed in SAT following bariatric surgery are involved in lipid and energy metabolism, inflammatory and immunological responses, insulin signaling, cell differentiation, oxidative stress regulation and gene transcription [
27,
30]. A recent study [
32] has also observed long-term effects of RYGB on gene expression in abdominal SAT, with enriched pathways related to lipid metabolism, fat cell differentiation and immune response. Again, most of the studies examining gene expression changes in SAT have been conducted among individuals with obesity who had undergone RYGB or SG, but none had investigated the impact of BPD-DS on the transcriptomic SAT profile [
27‐
31].
Examining molecular changes taking place following a weight-loss procedure may then help in gaining further insight into its distinct metabolic impact and may eventually aid in targeting the most beneficial intervention. Thus, our objective was to compare methylation and gene expression changes in SAT following three different types of bariatric surgeries, namely BPD-DS, RYGB and SG. We hypothesized that the more metabolically effective BPD-DS leads to greater modifications at the methylation and gene expression level than the extensively studied RYGB and SG bariatric surgeries.
Discussion
Following bariatric surgery in any of the procedures analyzed, most of the participants lost 20% or more of their initial weight, which has been suggested as a threshold for intervention success [
34,
35]. Also, global transcriptomic and methylomic findings highlighted that, independently of the type of procedure, bariatric surgery induced a profound molecular remodeling in the SAT of patients with severe obesity, mainly through gene down-regulation and hypermethylation. However, weight loss at 12 months postoperative was far more important in participants who underwent BPD-DS, as compared to those undergoing RYGB + SG, which may partly explain the more extensive transcriptomic and methylomic modifications observed in the SAT of BPD-DS participants. It should be noted that less than 35% of all DEGs and only 17% of DMGs were common to both BPD-DS and RYGB + SG surgery groups. Moreover, the extent of gene down-regulation and hypermethylation among common DEGs and DMGs was 50% greater following BPD-DS than after RYGB + SG. Even more striking was that more than 50% of DEGs and almost 80% of DMGs were exclusive to BPD-DS, suggesting more extensive transcriptomic and methylomic modifications in SAT following this surgery, which also represents 70% of genes being simultaneously identified as DEGs and DMGs. Interestingly, an important proportion of differentially methylated CpG sites was located within promoter regions, through which gene expression can be regulated. Part of the methylation profile may be acquired during embryogenesis and is thought to remain stable across the lifespan [
36]. However, some methylation marks may also be modified from the exposition to environmental factors such as diet, exercise, obesity, ageing or bariatric surgery [
10,
11,
36].
Greater remission rates in type 2 diabetes following BPD-DS in comparison to RYGB and SG procedures have been previously reported [
23,
37,
38]. These beneficial effects on type 2 diabetes are maintained over time with slightly more than 90% of the patients who discontinued diabetic therapy 10 years after undergoing BPD-DS surgery [
21]. Besides the impressively favorable impact on plasma glucose and insulin levels, other beneficial metabolic shifts have been reported to be more persistent in long-term follow-ups with BPD-DS compared to RYGB, such as improved lipid profile and blood pressure lowering [
21,
39,
40]. Whether these greater modifications are due to a more important and sustained weight loss or to more profound metabolic modifications after BPD-DS is still unknown. In this regard, it is worth highlighting that other factors than surgery type itself, such as sex and initial BMI, have been revealed to have a critical impact on weight loss and health outcomes following a bariatric surgery, with heavier men usually having the worst prognosis [
41‐
43]. We also previously observed that men are overrepresented among subjects with a reduced weight loss response after surgery and, more importantly, that initial BMI is the best predictor of weight loss [
44]. Concretely, we reported that the probability to show reduced weight loss following bariatric surgery significantly increases with initial BMI and mostly in men. In the present study, we tried to minimize a potential sex bias by keeping similar proportions of men and women among the different procedures, as well as by excluding for the analysis all the transcripts and CpG sites located on sexual chromosomes. Similarly, we used %TWL instead of %EWL as a measure of body weight loss, in an effort to reduce the impact of initial BMI, as recently reported [
45]. Also, in order to take into account baseline differences between participants, differential gene expression analysis was performed by using a paired design, and linear as well as logistic models were adjusted for age, sex and initial BMI. Herein, having these considerations into account, we have established a potential link between a greater weight loss reduction with a more extensive transcriptomic and methylomic remodeling in SAT, which ultimately may contribute to these metabolic improvements. Similarly, gene expression changes in SAT were strongly associated with the reduction of neck circumference following bariatric surgery, which was previously suggested to be a reliable predictor of remission rates [
46].
In animal models, metabolic changes seem to be mostly attributable to the malabsorptive effect of BPD-DS [
47]. Vink et al. [
48] have observed that a very-low caloric diet compared to a low caloric diet with similar weight loss induced greater gene expression modifications in pathways related to mitochondrial function, adipogenesis, immunity and inflammation. Thus, it is possible that the more pronounced effects ascertained in this study in BPD-DS compared to RYGB + SG were also partly attributable to a decrease in lipid absorption [
49]. Moreover, the jejunum maybe crucial in regulating insulin sensitivity and is completely bypassed in BPD-DS [
49]. Among mouse models, a high-fat compared to a low-fat isocaloric diet led to greater modifications in gene expression and methylation [
50]. However, mice in the high-fat diet group were also significantly heavier following the diet than mice in the low-fat group [
50]. It is not known whether diet or weight gain was responsible to induce these greater changes. In this regard, different bariatric procedures may also lead to distinct changes in diet, but actual data do not allow to test whether these changes are also impacting the present results.
In our study, many significantly enriched biologic processes related to protein translation and ribosomal activity were observed for up-regulated DEGs exclusive to BPD-DS and common to both surgery groups. Enrichments in similar pathways have been previously reported after RYGB [
32] and diet-induced weight loss [
48]. Genes involved in protein translation are also differently expressed among metabolically unhealthy individuals with obesity [
51]. Protein translation may be regulated through an enhanced insulin sensitivity following bariatric surgery or dietary-induced weight loss [
52]. For instance, insulin activates eukaryotic initiation and elongation factors, and increases the cellular content of ribosomal proteins [
53]
. Herein, many genes encoding ribosomal proteins (
RPS and
RPL genes) were up-regulated. In addition, genes encoding eukaryotic translation initiation factor 4E binding proteins (
EIF4EBP1,
EIF4EBP2 and
EIF4EBP3 genes) and eukaryotic translation initiation factors (
EIF genes) were found among the most up-regulated DEGs
. Although these genes all encode proteins involved in translation, they may also be linked to adiposity, adipogenesis or glucose homeostasis. In fact, some
EIF genes have been reported to correlate with genes encoding insulin receptor (
INSR) and insulin receptor substrate-1 (
IRS-1) [
54], which were both also significantly up-regulated in the present study.
More than a simple reservoir of energy surplus, SAT has important endocrine and paracrine functions which regulate many biological processes [
55]. Overweight and obesity may lead to SAT dysfunction including several perturbations, as an impaired expandability and adipocyte hypertrophy, altered innate and adaptative immune functions, changes in the secretion of pro- and anti-inflammatory peptides and eventually fibrosis [
56,
57]. Transcriptomic modifications in pathways related to immunity and/or inflammation following weight loss, induced either by bariatric surgery or by diet, have been previously observed [
27,
28]. Here, almost all the biological processes found to be significantly enriched with down-regulated DEGs were related to immune-related functions. Interestingly, most of such pathways were also significantly enriched with hypermethylated DMGs, pointing to a marked rearrangement of inflammation and immune molecular processes in the SAT of study participants. Although this effect was observed independently of the bariatric procedure analyzed, it was particularly pronounced following BPD-DS. By contrast, the fact that none of these pathways was significantly enriched with DEGs or DMGs exclusive to RYGB + SG emphasizes the procedure-specific nature of many of these molecular changes. Among significantly enriched immune-related pathways, biological processes such as T cell activation, leukocyte cell–cell adhesion, neutrophil activation and ERK1/2 cascades were significantly enriched. It has been observed that individuals with obesity have an increased quantity of T lymphocytes in their SAT [
58]. Moreover, T regulatory lymphocytes have been shown to be reduced following bariatric surgery [
59]. Neutrophils are an essential component of the innate immune response. Following RYGB, Poitou et al. [
60] identified several DEGs which were related to neutrophil-mediated inflammation. In their study [
60], DEGs related to neutrophils function or activity included
S100A8,
S100A9,
S100A12,
CD300E,
VNN2,
FPR2 and
APOBEC3A, which were all also differentially expressed in both surgery groups of the present study, but around twice as much in BPD-DS than in RYGB + SG. Interestingly, Kerr et al. [
32] have observed that there were continuously down-regulated genes 5 years following RYGB, despite significant weight regain occurring from 2 to 5 years postoperative. The authors observed that these continuously down-regulated genes were involved in biological processes such as cytokine production, cell chemotaxis and neutrophil activation [
32], suggesting that these gene modifications might not be linked exclusively to weight variations. Moreover, these long-term changes in gene expression may be sustained through epigenetic mechanisms.
Two genes encoding for cytokines were among DEGs with the greatest change extent for both surgery groups,
CSF3, which encodes for colony stimulating factor 3, a cytokine that has been reported to be elevated among individuals with obesity [
61], and
IL6, which is a well-known cytokine involved in inflammation. Following weight loss induced either by diet or bariatric surgery, a down-regulation in gene expression was previously noticed for genes such as
CCL2 [
28],
NFKB1 [
32],
NLRP3 [
32],
HIF1A [
27],
CLEC7A [
27] and
IL4R [
27]. These genes have also been found to be significantly down-regulated herein. Among them,
HIF1A may indirectly activate
NLRP3, which encodes for NLRP3 inflammasomes, contributing to the inflammatory responses via IL-1β activation, which is down-regulated to a greater extent in BPD-DS in this study, and could be of key importance in the development of type 2 diabetes [
62].
Biological processes significantly enriched with hypomethylated DMGs were mostly related to extracellular structure and matrix organization, actin filament organization, cell-substrate and matrix adhesion, as well as cell-substrate junction and assembly, among others. These changes were again more pronounced following BPD-DS. Kelehmainen et al.[
63] reported a down-regulation of DEGs involved in extracellular matrix following weight loss. The excessive accumulation of extracellular matrix components associated with obesity can lead to adipose tissue fibrosis which contributes to the dysfunction of adipocytes [
64]. Moreover, higher SAT fibrosis may lessen the weight loss response following RYGB [
64]. In the present study, these pathways were hypomethylated but not up-regulated. Thus, the functional impact of this hypomethylation remains unknown. It is possible that changes in gene expression were transient and no longer present at 12 months postoperative, since they potentially occurred earlier following the bariatric surgery, as previously shown in skeletal muscle [
31].
Among significantly down-regulated DEGs common to both surgery groups, a strong inverse association with %TWL was observed for
CD248, a gene which encodes for tumor endothelial marker 1/endosialin, a transmembrane glycoprotein known to be expressed in proliferating tissues, especially during embryogenesis, tumor growth and inflammatory lesions [
65]. More recently, Petrus et al. [
66] have demonstrated that
CD248 is up-regulated in the SAT of individuals with obesity and insulin resistance and is potentially involved in the response to hypoxia. They reported that both
CCL2 and
IL6, respectively involved in extracellular matrix remodeling and inflammation, were correlated positively with
CD248 gene expression [
66]. Among up-regulated DEGs common to both surgery groups,
GJC3 was associated with %TWL and it has been reported to be down-regulated in obesity [
67]. On the other hand,
GPD1L was the up-regulated DEG exclusive to BPD-DS most strongly associated with %TWL. In a long-term follow-up study,
GPD1L was reported to be regulated by weight loss and regain after RYGB [
32]. Furthermore,
GPD1L was recently identified as potentially playing a causal role in obesity and insulin resistance [
68]. During weight loss and weight maintenance induced by a low caloric diet,
GPD1L was found to be up-regulated, while being down-regulated during weight gain induced by a high-fat diet [
68]. It is worth noting that in the present study, most of DEGs exclusive to BPD-DS and associated with %TWL were also identified as DMGs, while none of the DEGs exclusive to RYGB + SG group were significantly associated to %TWL. Moreover, a total of four down-regulated DEGs exclusive to BPD-DS also showed an association with adipocyte size change. Among them,
CORO1C, a gene recently identified to be up-regulated in the SAT of individuals with obesity [
69], was found to be closely linked to %TWL and was also identified as hypomethylated. From a broader perspective, methylomic changes observed in this study were mostly exclusive to BPD-DS, which points to an epigenetic-mediated mechanism by which gene expression changes in SAT may occur in a greater extent in patients undergoing this type of surgery.
Present findings thus provide evidence that BPD-DS induce larger methylomic and transcriptomic modifications than RYGB + SG, which may be partly explained by greater weight loss and malabsorption created by this surgical approach. However, it is also possible that BPD-DS participants, who had higher BMI and waist circumference before surgery, started with a more deteriorated metabolic profile than participants who underwent RYGB or SG, which ultimately led to the more extensive transcriptomic and methylomic modifications observed. Results shown herein were obtained at 12 months postoperative, and it is possible that participants may still be losing weight or not being weight stable, which could affect transcriptome and methylome profiles. However, it has been reported that weight loss is as its nadir around 12 to 18 months following either RYGB [
70] or BPD-DS [
71].
Acknowledgements
We would like to acknowledge the contribution of surgeons, nurses, and the medical team of the bariatric surgery program at IUCPQ. The coinvestigators and collaborators of the REMISSION study are (alphabetical order): Bégin C, Biertho L, Bouvier M, Biron S, Cani P, Carpentier A, Dagher A, Dubé F, Fergusson A, Fulton S, Hould FS, Julien F, Kieffer T, Laferrère B, Lafortune A, Lebel S, Lescelleur O, Levy E, Marette A, Marceau S, Michaud A, Picard F, Poirier P, Richard D, Schertzer J, Tchernof A, Vohl MC.