Accumulation and Changes in Composition of Collagens in Subcutaneous Adipose Tissue After Bariatric Surgery

Abstract

Context:

Extracellular matrix (ECM) in sc adipose tissue (scAT) undergoes pathological remodeling during obesity. However, its evolution during weight loss remains poorly explored.

Objective:

The objective of the investigation was to study the histological, transcriptomic, and physical characteristics of scAT ECM remodeling during the first year of bariatric surgery (BS)-induced weight loss and their relationships with metabolic and bioclinical improvements.

Design, Setting, Patients, and Interventions:

A total of 118 morbidly obese candidates for BS were recruited and followed up during 1 year after BS.

Main Outcome Measures:

scAT surgical biopsy and needle aspiration as well as scAT stiffness measurement were performed in three subgroups before and after BS. Fourteen nonobese, nondiabetic subjects served as controls.

Results:

Significantly increased picrosirius-red-stained collagen accumulation in scAT after BS was observed along with fat mass loss, despite metabolic and inflammatory improvements and undetectable changes of scAT stiffness. Collagen accumulation positively associated with M2-macrophages (CD163⁺ cells) before BS but negatively afterward. Expression levels of genes encoding ECM components (eg, COL3A1, COL6A1, COL6A2, ELN), cross-linking enzymes (eg, lysyl oxidase [LOX], LOXL4, transglutaminase), metalloproteinases, and their inhibitors were modified 1 year after BS. LOX expression and protein were significantly decreased and associated with decreased fat mass as well as other cross-linking enzymes. Although total collagen I and VI staining decreased 1 year after BS, we found increased degraded collagen I and III in scAT, suggesting increased degradation.

Conclusions:

After BS-induced weight loss and related metabolic improvements, scAT displays major collagen remodeling with an increased picrosirius-red staining that relates to increased collagen degradation and importantly decreased cross-linking. These features are in agreement with adequate ECM adaptation during fat mass loss.

The extracellular matrix (ECM) in sc adipose tissue (scAT) undergoes substantial pathological remodeling during obesity. ECM accumulation, usually called fibrosis, is defined as an excessive deposition of ECM components (mainly cross-linked collagens) and impaired degradation (1). ECM accumulation is important in the regenerative step in which it replaces damaged cells. However, if the damage persists, excessive ECM deposition harms tissue homeostasis and function (2). In obesity, scAT ECM accumulation reduces tissue plasticity and results in adipocyte dysfunction, ectopic fat storage, and metabolic disorders (1). Studies have shown the detrimental consequences of ECM accumulation in obesity and their associations with comorbidities. In mice, genetic ablation of membrane type 1-matrix metalloproteinase (MT1-MMP) a membrane anchored metalloproteinase degrading collagen I, leads to increased periadipocyte fibrosis and severe metabolic complications such as hepatic steatosis (3). Likewise, collagen VI accumulation in obesity is associated with insulin resistance (4, 5). By contrast, the absence of collagen VI in high-fat diet or ob/ob mice results in uninhibited adipocyte expansion and associates with metabolic and inflammatory improvements (6). In obese subjects, scAT fibrosis is increased (7, 8). Moreover, higher scAT fibrosis at baseline is associated with lower weight loss 1 year after bariatric surgery (BS) (7, 9). In addition, scAT pericellular fibrosis is associated with liver fibrosis, suggesting that obesity is a profibrotic condition (9). Finally, the pericardial fat secretome was also found to promote myocardial fibrosis (10). Overall, these studies underline the potential deleterious effects of obesity-induced scAT ECM accumulation.

Mechanistically, scAT fibrosis leads to adipocyte dysfunction and fibro-inflammation through a mechanotransduction pathway (11). Lysyl oxidase (LOX), an important matrix fibers' cross-linking enzyme, contributes to tissue mechanical properties (12). In AT, LOX expression is up-regulated in high-fat diet or ob/ob mice. By contrast, inhibition of LOX activity leads to improved metabolism and inflammation (13). In obese subjects, scAT LOX expression is also increased (11). scAT stiffness, measured noninvasively by transient elastography, associates with picrosirius-red-stained scAT fibrosis and altered glucose homeostasis (9).

ECM turnover, a crucial process during excess ECM accumulation, is predominantly regulated by the balance between MMPs and their endogenous tissue inhibitor of metalloproteinases (TIMPs). In obesity, a new relationship between MMPs and TIMPs is established and enables tissue remodeling. Enzymes (eg, MMP-3, -9, -11, -12, -13, -16, and -24) are expressed at low level in scAT but are rapidly up-regulated during obesity, which eventually favors scAT expansion (1). Weight loss represents another condition that induces scAT remodeling, exhibiting by changes in expression of many ECM genes soon after BS (8). Some studies have shown increased ECM deposition, eg, up-regulated collagens, particularly COL6A3, after major weight loss in the long term (14). However, most of these studies focused on selected collagens at expression levels and did not explore the overall ECM characteristics. Furthermore, no study has yet evaluated the links between scAT ECM remodeling, stiffness, and modifications in cross-linking enzymes and improved metabolic parameters after weight loss.

Herein we examined fibrillar collagen accumulation, synthesis, and degradation as well as cross-linking enzymes, macrophage infiltration, and scAT stiffness during the first year after BS. We also analyzed relationships between ECM properties and metabolic and inflammatory parameters improvements observed after BS.

Materials and Methods

Study population

A total of 118 morbidly obese candidates for BS who met the recruitment criteria as previously described (7) were enrolled at the Institute of Cardiometabolism and Nutrition, Department of Nutrition, and operated in the Department of Surgery, Pitié-Salpêtrière Hospital (Paris, France). Due to the difficulties to obtain large amount of scAT surgical biopsy sample in every subject during the follow-up and the number of experiments needed to perform on these samples, we divided our overall cohort into three groups according to different scAT measurements that were realized (study flow chart; see Figure 1). However, subjects were part of the same prospective cohort and baseline (T0) characteristics were not significantly different (Table 1).

Figure 1.

Study flow chart of 118 obese subjects.

Three groups of obese individuals were recruited for scAT exploration at baseline (T0) and during the follow-up after bariatric surgery at T3 and T12. There are 11 common subjects (female, n = 6) in group 1 and group 2 in whom we performed both histological analysis (P-R, IHC) and scAT stiffness measurement (stiffness; upper gray part). A subgroup of 14 subjects in group 2 underwent scAT needle aspiration for RT-PCR analysis (bottom gray part). P-R, Picrosirius-red staining.

Table 1.

Clinical and Biological Parameters of the Study Population

Variable	Group 1 (n = 52)			Group 2 (n = 35)			Group 3 (n = 42)
	T0	T3	T12	T0	T3	T12	T0	T12
BMI, kg/m2	45.8 ± 6.8	38.6 ± 6.0a	33.1 ± 6.0a,b	46.9 ± 7.7	39.2 ± 6.1a	33.3 ± 5.4a,b	47.1 ± 6.0	33.0 ± 5.0a
Type 2 diabetes, n, %	15 (29)	7 (13)	8 (15)	12 (34)	2 (6)a	3 (10)a	14 (33)	4 (10)a
Hypertension, n, %	18 (35)	16 (31)	12 (23)	13 (37)	10 (29)	10 (29)	12 (29)	8 (19)
Sleep apnea, n, %	31 (60)	24 (46)	15 (29)a	19 (54)	17 (49)	12 (34)	24 (57)	7 (17)a
Body composition
Fat mass, kg	57.96 ± 12.98	48.73 ± 11.93a	36.40 ± 11.03a,b	61.75 ± 12.25	50.37 ± 11.00a	37.73 ± 9.23a,b	61.43 ± 11.67	36.01 ± 10.91a
Fat-free mass, kg	62.61 ± 10.96	53.78 ± 9.56a	51.71 ± 9.31a,b	62.09 ± 9.50	55.63 ± 9.93a	53.52 ± 9.73a,b	58.64 ± 7.33	50.42 ± 8.45a
Fat mass, %	46.74 ± 6.27	45.91 ± 6.25a	39.58 ± 7.16a,b	48.55 ± 5.51	46.10 ± 5.78a	39.95 ± 6.91a,b	49.72 ± 3.55	39.79 ± 5.14a
Fat-free mass, %	50.85 ± 6.03	51.12 ± 6.06a	57.34 ± 6.80a,b	49.09 ± 5.30	51.15 ± 5.55a	57.03 ± 6.67a,b	47.92 ± 3.44	57.22 ± 4.74a
Glycemic parameters
Fasting glycemia, mM	5.99 ± 2.34	4.95 ± 0.76a	4.81 ± 0.77a	6.11 ± 2.25	4.99 ± 0.97a	4.87 ± 0.65a	6.29 ± 2.21	4.73 ± 0.64a
Fasting insulin, μU/mL	20.74 ± 11.68	12.52 ± 7.88a	11.06 ± 11.04a	22.94 ± 11.57	12.74 ± 5.27a	10.21 ± 5.75a	20.84 ± 10.09	9.07 ± 4.16a
HbA1c, %	6.28 ± 1.39	5.57 ± 0.49a	5.55 ± 0.42a	6.37 ± 1.55	5.63 ± 0.83a	5.44 ± 0.44a	6.36 ± 1.13	5.54 ± 0.52a
HOMA-IR	2.74 ± 1.62	1.62 ± 0.96a	1.41 ± 1.30a	2.94 ± 1.46	1.61 ± 0.65a	1.30 ± 0.73a	2.67 ± 1.28	1.15 ± 0.52a
HOMA-B%	152.48 ± 62.89	85.13 ± 53.04	129.61 ± 81.99	175.09 ± 72.51	145.40 ± 52.62a	123.83 ± 54.05a,b	163.60 ± 66.38	121.90 ± 41.12a
HOMA-S%	49.08 ± 30.88	132.95 ± 66.10a	105.75 ± 55.32a	43.36 ± 26.92	75.31 ± 37.80a	104.32 ± 60.00a,b	47.79 ± 30.99	113.60 ± 86.29a
Lipid parameters
Total cholesterol, mM	4.86 ± 0.93	4.60 ± 0.83a	4.82 ± 0.92	4.98 ± 0.87	4.59 ± 0.87	4.62 ± 0.82	4.82 ± 0.91	4.11 ± 0.65a
Triglycerides, mM	1.56 ± 1.02	1.12 ± 0.46a	0.91 ± 0.40a,b	1.50 ± 0.72	1.17 ± 0.40	0.91 ± 0.35a,b	1.57 ± 0.87	0.89 ± 0.34a
HDL cholesterol, mM	1.13 ± 0.34	1.22 ± 0.35	1.50 ± 0.36a,b	1.14 ± 0.32	1.10 ± 0.33	1.45 ± 0.37a,b	1.09 ± 0.29	1.40 ± 0.36a
Hepatic factors
AST, IU/L	29.3 ± 9.8	27.5 ± 15.2	24.6 ± 8.2a	30.6 ± 8.9	31.9 ± 17.9	24.1 ± 7.6a,b	28.3 ± 11.7	24.7 ± 5.2
ALT, IU/L	34.7 ± 20.9	28.2 ± 13.6	22.4 ± 11.6a,b	38.9 ± 21.4	41.2 ± 34.1	20.4 ± 10.7a,b	33.0 ± 22.5	24.7 ± 8.2a
γGT, mg/dL	40.8 ± 24.2	22.8 ± 12.0a	19.2 ± 8.5a,b	42.6 ± 25.4	26.0 ± 15.0a	21.4 ± 11.5a	40.6 ± 25.5	21.6 ± 28.3a
Adipokines
Leptin, ng/mL	59.2 ± 31.8	28.8 ± 17.1a	27.4 ± 21.8a	71.5 ± 35.2	28.2 ± 16.3a	23.9 ± 16.9a	57.4 ± 27.3	24.8 ± 15.4a
Adiponectin, μg/mL	4.88 ± 2.81	5.71 ± 2.49	6.56 ± 3.05a,b	3.84 ± 1.56	5.68 ± 2.66a	6.98 ± 2.73a	3.56 ± 1.53	6.59 ± 2.70a
Inflammatory factors
IL-6, pg/mL	4.14 ± 2.56	3.82 ± 1.92	2.95 ± 1.78	4.35 ± 2.32	4.27 ± 2.45a	3.05 ± 2.55a	3.90 ± 1.27	2.74 ± 1.43a
hsCRP, mg/L	7.86 ± 5.67	4.38 ± 3.94a	2.30 ± 2.03a,b	9.25 ± 7.08	3.41 ± 2.73a	3.49 ± 5.38b	9.80 ± 7.18	1.48 ± 1.72a
Orosomucoid, g/L	0.92 ± 0.17	0.83 ± 0.21a	0.65 ± 0.17a,b	0.95 ± 0.19	0.82 ± 0.20	0.68 ± 0.21a,b	0.98 ± 0.17	0.69 ± 0.17a

Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; rGT, gamma-glutamyltransferase; HbA1c, glycated hemoglobin; HOMA-B, homeostatic model assessment of β-cell function; HOMA-IR, homeostatic model assessment of insulin resistance; HOMA-S, homeostatic model assessment of insulin sensitivity; hsCRP, high-sensitivity C-reactive protein. Values are expressed as mean ± SD unless otherwise stated. There are 11 common subjects in group 1 and group 2. For continuous data, repeated ANOVA was used to compare three time points in group 1 and group 2, when P value of ANOVA was significant (ie, P < .05); Holm-Sidak's parametric multiple comparison test was used to compare each of the two time points; and Student's t test was used in group 3. For categorical data, Fisher's exact test was used.

^a

P < .05 when compared with T0 in each group.

^b

P < .05 when compared with T3 in group 1 and group 2. Subjects' baseline characteristics (T0) in three groups were compared by ANOVA for quantitative data and by Fisher's exact test for qualitative data. No significant differences were found except the BS type.

Group 1 subjects (n = 52; aged 40.1 ± 10.2 y; female, n = 37 [71%]; BS procedures: gastric banding, n = 8 [15%]; sleeve gastrectomy, n = 16 [31%]; Roux-en-Y gastric bypass [RYGB], n = 28 [54%]) accepted surgical scAT biopsy before (T0), 3 months (T3), and 12 months (T12) after BS. Surgical biopsy was performed under local anesthesia in the periumbilical area as described (15, 16). The collected scAT samples were used for explant experiments and histological analysis.

Group 2 (n = 35, aged 38.0 ± 10.0 y; female, n = 24 [69%]; BS procedures: gastric banding, n = 3 [8%], sleeve gastrectomy, n = 16 [46%], RYGB, n = 16 [46%] underwent at T0, T3, and T12 scAT stiffness measurement (see below). A subgroup of 14 nondiabetic women from group 2 underwent scAT needle aspiration for RT-PCR analysis. Notably, 11 individuals with stiffness measurement were also part of group 1.

Group 3 (n = 42; aged 42.9 ± 10.5 y; female, n = 42 [100%]; BS procedures: RYGB, n = 42 [100%]) underwent scAT needle aspiration at T0 and T12 for microarray analysis.

Fourteen nonobese, nondiabetic subjects (aged 41.6 ± 14.1 y; female, n = 4 [29%]; body mass index [BMI] 23.2 ± 3.3 kg/m²), who had elective abdominal programmed surgery without inflammatory diseases as described (7), were recruited as a control group. Perioperative scAT biopsy samples were collected in the same location as in obese subjects. Ethical approval was obtained from the Research Ethics Committee of Pitié-Salpêtrière Hospital (Comité de Protection des Personnes, Ile de France). Informed written consent was obtained from all subjects.

Clinical, anthropological, and biological parameters

Body composition was evaluated by whole-body, fan-beam, dual-energy X-ray absorptiometry scan (Hologic Discovery W) as described (9). Blood samples were collected after a 12-hour overnight fast at T0, T3, and T12. Clinical variables were measured as described (7). Pancreatic β-cell function (insulin secretion), insulin sensitivity, and resistance were estimated using homeostatic model assessment of continuous infusion glucose model assessment (17).

Measurement of scAT shear wave speed by transient elastography

A new noninvasive medical device based on transient elastography (18), AdipoScan (Echosens), was developed to measure scAT shear wave speed (SWS) associated with scAT stiffness (9, 19). scAT stiffness was measured by the same operator in obese subjects (group 2) in the right periumbilical region at T0, T3, and T12.

Transcriptomic experiments

scAT samples obtained by needle aspiration at T0 and T12 (group 3) were stored at −80°C for microarray analysis. Total RNA extraction, amplification, hybridization, and raw data analysis were performed as described (20). The complete data set is published in the National Center for Biotechnology Information Omnibus (http://www.ncbi.nlm.nih.gov/geo/) through the series accession number GSE72158.

RT-PCR for selected genes was performed as described (20), using total RNA extracted from scAT needle aspiration in 14 nondiabetic obese women (group 2) at T0, T3, and T12.

Tissue preparation and histological analysis of scAT

A piece of surgical biopsy sample was fixed and embedded in paraffin and sliced into 5-μm-thick sections. Collagen was stained with picrosirius-red (mainly collagen I and III) and analyzed using Calopix software (Tribvn) in 36 subjects (group 1) at T0, T3, and T12 as described (9). Total collagen accumulation represents the ratio of the stained fibrous area to the total tissue surface. Pericellular collagen accumulation (ie, collagen surrounding adipocytes) represents the ratio of the stained area in 10 random fields, avoiding fibrosis bundles. Adipocyte diameters were evaluated in the same 10 fields. Pericellular collagen accumulation was adjusted by adipocyte size to eliminate the effects of different adipocyte sizes in measure fields. Collagen I and VI, degraded collagen I, LOX, and macrophages were detected by immunohistochemistry (IHC) using specific antibodies (Supplemental Table 1). Total macrophages were defined as CD68⁺ cells and M2-macrophages as CD163⁺ cells. Their results are expressed as the number of CD68⁺ or CD163⁺ cells related to 100 adipocytes (21). Collagen and elastin structures were analyzed using confocal microscopy and second-harmonic generation (SHG) microscopy on another piece of fixed scAT sample in three random obese subjects (group 1) as described (11).

ScAT explant in vitro

A piece of surgical biopsy samples (group 1) was placed in a culture medium enriched in endothelial cell basal medium (Promocell), supplemented with 1% albumin free fatty acids (PAA Laboratories) and antibiotics. After 24-hour incubation at 37°C, scAT explant secretion media were collected and frozen at −80°C for ELISA and zymography. The explant secretion was normalized to adipose tissue weight according to the ratio of 1 mL of culture medium for 0.1 g scAT.

Protein determination in scAT explant

The concentrations of collagen III formation marker, N-proteases cleaved N-terminal propeptides of collagen III (pro-C3), and degradation marker, MMP-9-generated neoepitope fragments of collagen III (C3M) in scAT explants were evaluated using two competitive ELISA kits developed by Nordic Bioscience A/S (22). The protein profiles of proMMP-2 and proMMP-9 were analyzed by zymography as described (23).

Statistical analyses

Data are expressed as mean ± SD, categorical variables as numbers and percentages, and values in graphs as mean ± SEM. Categorical data were analyzed using a Fisher's exact test. For continuous data, a repeated one-way ANOVA was used to compare more than two groups and Holm-Sidak's parametric multiple comparison for post hoc analysis; a Student's t test was used to compare two groups. For small sample size (ie, n < 30), data were first transformed by natural logarithm if they did not follow a Gaussian distribution. Two-tailed P values were considered significant at P < .05. All analyses were conducted using R software version 3.0.3 (http://www.r-project.org) and GraphPad Prism 6.0.

Results

Increased collagen deposition in scAT during BS-induced weight loss

Using picrosirius-red staining, scAT collagen was quantified in 36-paired obese subjects (group 1) at baseline (T0) and during the follow-up (T3 and T12). No significant difference in collagen accumulation was found among the different BS procedures at baseline (Supplemental Table 2). More abundant and thicker bundles of collagen fibers traversing the scAT were observed at T3 and T12 (Figure 2, A–C). Several parenchymal areas were filled with less well-organized collagen in postoperative tissues (Figure 2C, enlarged image). A significant increase in the average of total and pericellular collagen was observed at T3 and T12 (Figure 2D). As expected, adipocyte size significantly decreased after BS (Figure 2E), but this reduction was not correlated with collagen accumulation. Moreover, the increase in pericellular collagen remained significant after adjustment for adipocyte size reduction. Importantly, the fat mass reduction was negatively correlated with pericellular collagen accumulation (r = −0.40, P < .05). No other associations were observed between collagen accumulation and metabolic or inflammatory variables except for systemic high-density lipoprotein (HDL)-cholesterol (Supplemental Table 3). Notably, the results held true in subgroup analysis of each BS procedure (data not shown).

Figure 2.

scAT evaluation.

Collagen accumulation in scAT stained by picrosirius-red in one representative obese subject at baseline (T0) (A), T3 (B), and T12 (C) after BS. Total and pericellular collagen accumulation (D) and adipocyte size (E) at T0, T3, and T12 in 36 obese subjects from group 1 are shown. Repeated ANOVA test and Holm-Sidak's parametric multiple comparison test were used, *, P < .01. F, scAT stiffness, SWS, was evaluated at T0, T3, and T12 after BS measured by transient elastography in 35 subjects (group 2). Repeated ANOVA test was used (P > .05).

Undetectable changes in tissue stiffness despite increased scAT collagen accumulation in scAT

Because we previously showed that collagen accumulation was associated with scAT rigidity and metabolic alterations in obesity (9), we next aimed to investigate scAT stiffness changes after BS using AdipoScan (Echosens) at T0, T3, and T12 (group 2). To our surprise, despite increased collagen accumulation, no significant change in average SWS was detected at T3 and T12 compared with T0 (T0: 0.90 ± 0.29 m/sec, T3: 0.88 ± 0.28 m/sec, T12: 0.93 ± 0.43 m/sec, P = .58, Figure 2F). Even though we observed two major clusters of scAT stiffness individuals trajectories using K-means for longitudinal data cluster (Supplemental Figure 1), we did not observe significant bioclinical differences at any time points that could possibly explain these trajectories (Supplemental Table 4). Overall, increased picrosirius-red-stained collagen along with undetectable change in average scAT SWS could be considered as an adaptive phenomenon of ECM remodeling during weight loss, which requires further investigation.

M2-macrophages associate with collagen accumulation in scAT

M2 cells, alternatively activated macrophages, are implicated in the resolution phase of inflammation and tissue remodeling (24). Using IHC, M2 cells (ie, CD163⁺ cells) and total macrophages (ie, CD68⁺ cells) in scAT were quantified in 15 obese subjects from group 1 at T0 and T12. The CD163⁺ to CD68⁺ ratio increased between T0 and T12 (0.38 ± 0.20 vs 0.78 ± 0.58, P < .01, Figure 3A), in agreement with a switch toward M2 macrophages during weight loss and their role in tissue remodeling. At T0, a strong positive association between CD163⁺ cells and pericellular collagen accumulation was observed (r = 0.76, P < .01, Figure 3D, left panel). Although the number of CD163⁺ cells moderately increased at T12 (6% ± 3% vs 9% ± 4%, P = .04, Figure 3B), a negative association between CD163⁺ cells and pericellular collagen accumulation was found (r = −0.65, P = .02, Figure 3D, right panel). By contrast, the number of CD68⁺ cells decreased between T0 and T12 (17% ± 8% vs 14% ± 7%, P = .04, Figure 3C) but was not associated with collagen deposition at T0 or T12.

Figure 3.

Macrophage infiltration in scAT.

Evolution of CD163⁺ to CD68⁺ ratio (A) and CD163⁺ cells (B) and CD68⁺ cells (C) in scAT was evaluated by IHC at baseline (T0) and T12 after BS in 15 obese subjects from group 1 (IHC, T0–T3-T12 subgroup). Solid lines represent nondiabetic subjects (n = 8), and dotted lines represent type 2 diabetic subjects (n = 7). Pearson's correlation between CD163⁺ cells and pericellular collagen accumulation at baseline (T0) (D, left panel) and 12 months after BS (D, right panel) is shown, and hollow points represent type 2 diabetic subjects.

Major ECM remodeling at transcriptomic level after weight loss

Because BS-induced weight loss is accompanied by increased collagen deposition without detectable 1-year change in SWS, we next characterized transcriptomic signatures of scAT at T0 and T12 in 42 women (group 3). Using a 5% false discovery rate, we detected 4236 up- and 2989 down-regulated genes (for functional annotations, see Supplemental Figure 2). We focused our analysis on genes encoding proteins involved in ECM structural components, profibrotic proteins, remodeling, and mechanotransduction. We found differential patterns of gene changes (Figure 4A). In particular, genes encoding collagen III (COL3A1), collagen VI (COL6A1, COL6A2), and elastin (ELN) were significantly down-regulated, whereas collagen I (COL1A1) was unchanged. Collagen VI α3-chain (COL6A3) was modestly up-regulated. Connective tissue growth factor (CTGF) and secreted phosphoprotein 1 (osteopontin, SPP1) were significantly down-regulated. MMP-9, TIMP1, TIMP2, and TIMP4 were also significantly modified. Importantly, some genes previously shown to be stimulated by mechanical stress (11), such as Yes-Associated Protein, TEA Domain Family Member (TEAD)2, TEAD3, and TEAD4, were not modified after weight loss (P > .05).

Figure 4.

Transcriptomic signature of scAT ECM genes in obese subjects 1 year after BS.

Gene expression levels from microarray data in scAT at baseline represented as dotted line (T0) and T12 months after BS are represented as bars in 42 women from group 3. A, Genes involved in ECM remodeling (matrix fibers, cross-linking, profibrotic protein, degradation proteins, and adhesion protein) show important changes. B, Most genes involved in posttranscriptional modifications of collagen are down-regulated 1 year after BS. They include the following: 1) enzymes involved in the hydroxylation of proline: prolyl 4-hydroxylase; prolyl-3 hydrolase; 2) enzyme involved in glycosylation of hydroxylysine: GLT25D1; 3) molecular chaperones HSP47, GRP94, calexin (CANX), and disulphideisomerase (PDI) (HSPA5, DNAJC10, ERP29, PDIA4, PDIA6); and 4) enzymes involved in N- and C-propeptides of procollagens: ADAMTS1, ADAMTS2, ADAMTS5, and ADAMTSL4. By contrast, prolyl-3 hydrolase (P3H2, P3H3) and ADAMTS9 genes were up-regulated. Data are presented as changes from baseline. *, P < .05.

During collagen biosynthesis, major posttranslational modifications take place and are mediated by important enzymes and chaperones. We found that the expression levels of most of these molecules were decreased at T12 (Figure 4B). Finally, we observed a significant down-regulation of genes encoding cross-linking enzymes such as LOX, lysyl oxidase-like (LOXL)-4, transglutaminase1, procollagen-lysine, and 2-oxoglutarate 5-dioxygenase-2 and -3, suggesting that matrix fibers' cross-linking was decreased after BS (Figure 4A). These transcriptomic analyses confirm the strong remodeling of scAT after BS and show major transcriptomic modifications of enzymes involved in collagen biosynthesis, cross-linking, and degradation.

Decreased cross-linking of matrix fibers during weight loss associates with improved metabolic phenotype

We next explored cross-linking enzymes and their associations with metabolic phenotypes. We confirmed microarray data by RT-PCR and observed that LOX gene expression was significantly down-regulated at T3 and T12 (group 2) (Figure 5A). This was substantiated by decreased LOX protein staining surrounding adipocytes at T3 and T12 (Figure 5B) using IHC. By confocal microscopy and SHG in fixed scAT samples in three random obese subjects (group 1), we found a trend toward reduced collagen and elastin intensity at T3 (Supplemental Figure 3). Elastin protein at T3 had more twisted structures (Figure 5C), suggesting that scAT might become less rigid after weight loss.

Figure 5.

Cross-linking of matrix fibers and collagen degradation and synthesis in scAT.

A, LOX gene expression levels at baseline (T0), T3, and T12 after BS in 14 obese nondiabetic women from group 2. B, LOX stained by IHC in obese and nonobese subjects (magnification, ×20). C, scAT elastin structure (magenta) was observed by second harmonic generation at T0 and T3 in one representative obese subject among the three (magnification, ×20). D, Correlation heat map between changes of bioclinical parameters and changes of genes regulating cross-linking from T0 to T12 in 42 women from group 3. Correlations between gene expression and changes of glycated hemoglobin and glycemia were analyzed separately in nondiabetic (nonDM, n = 28) and type 2 diabetic (DM, n = 14) subjects. Homeostatic model assessment of insulin resistance (HOMA-IR), homeostatic model assessment of β-cell function, and homeostatic model assessment of insulin sensitivity were analyzed only in nondiabetic subjects. Pearson's coefficients of each correlation are represented as blue (negative correlation) or red (positive correlation). *, P < .05. E, Degraded collagen I in scAT stained by IHC in one representative obese subject at T0, T3, and T12 and one representative nonobese subject. F, Degraded collagen III (C3M, left panel) and newly synthesized collagen III (pro-C3, right panel) in scAT explant measured by ELISA in five nonobese (Non Ob) and 10 obese subjects (group 1). Diabetic subjects are shown in red. *, P < .05. G, Analysis of (pro)MMP-2 and (pro)MMP-9 presence in scAT by gelatin zymography in three obese nondiabetic (Ob), three obese diabetic (Ob Diab), and two nonobese subjects. Ob Diab 1 was under metformin at T0, treatment that was stopped after BS and therefore absent at T3 and T12; Ob Diab 2 was taking sitagliptin, glimepiride, and metformin at T0, metformin alone at T3, and no more treatments at T12; Ob Diab 3 was taking basal insulin, liraglutide, glimepiride, and metformin at T0, basal insulin and metformin at T3, and glimepiride and metformin at T12. U937 cells (ATCC CRL-1593.2) stimulated with 100 U/mL recombinant TNF for 48 hours were used as positive control. ProMMP-2 (72 kDa) and proMMP-9 (92 kDa) were detected as transparent bands on the background of Eza-blue-stained gelatin.

We next examined the relationships between 1-year changes in cross-linking enzyme expression and that of clinical variables (ie, T12-T0 variation) in group 3 (Figure 5D). The reduction of LOX gene expression was positively associated with the reduction of BMI, fat mass (kilograms), adipocyte volume, serum leptin, and orosomucoid. Variation of LOXL1 was also associated with BMI, fat mass (kilograms), leptin, and total- and HDL-cholesterol. Our gene expression results suggest that decreased post-BS cross-linked scAT matrix fibers link with improved weight loss.

Increased collagen degradation during BS-induced weight loss

Our team (7) and others (4) have shown that collagen I and III are more frequently observed in fibrous bundles, whereas collagen VI is surrounding adipocytes. Despite increased scAT collagen accumulation after BS, we found decreased collagen I and VI staining at T12 (Supplemental Figure 4), suggesting that increased picrosirius-red staining may also indicate (at least partially) degraded collagen fragments. We tested this hypothesis by measuring collagen fragments with immunostaining, ELISA, and zymography from scAT explants. We observed increased stained degraded collagen I surrounding adipocytes in scAT at T3 and T12 (Figure 5E). Accordingly, the concentration of degraded collagen III in scAT at T12 was significantly increased compared with T0 (Figure 5F, left panel). ProMMP-9 and proMMP-2 entities at 92 and 72 kDa were respectively observed (Figure 5G). Despite individual variability, an increased trend of proMMP-2 at T3 in one nondiabetic obese subject and an increase of proMMP-9 at T3 followed by stabilization at T12 in two others were observed. These changes in proMMPs were not detected in samples from type 2 diabetic obese subjects (Figure 5G). In parallel, newly synthesized collagen III (propeptides of collagen III) concentration was significantly decreased in obese compared to nonobese subjects and showed a nonsignificant increase at T3, T12 (Figure 5F, right panel).

Discussion

Collagen accumulation in white adipose tissue is considered as an important pathological alteration associated with several comorbidities of obesity (1, 7, 9). Our results provide new insights into weight-loss-induced adipose tissue remodeling in paired humans individuals before and 1 year after BS. Our results suggest that picrosirius-red-stained collagen in scAT does not always refer to pathological collagens but could be a signature of extensive tissue remodeling and collagen degradation after adipocyte shrinkage during weight loss along with improved clinical, metabolic, and inflammatory outcomes.

During physiological tissue repair, ECM accumulation is a key regenerative step replacing tissue debris and dead cells (2). In pathological conditions, increased collagen deposition is not always synonymous with deleterious fibrosis. In myocardial injury, different types of fibrosis have been reported according to the progression and history of cardiomyopathies: a reactive interstitial fibrosis with ECM deposition in response to deleterious stimuli is considered pathological. Conversely, a replacement fibrosis that replaces myocytes after cell damage or necrosis may preserve the structural integrity of the myocardium (25, 26). Because we did not find any association between adipocyte diameter reduction and pericellular collagen increase, we attribute this increased pericellular collagen to replacement collagen that occurs at adipocyte shrinkage sites. This is further supported by our observation of some large parenchyma areas filled with less well-organized collagen. This replacement collagen, as part of the remodeling process, might be an adaptive and physiological phenomenon during weight loss.

Cross-linking is necessary for matrix fibers maturation and stabilization (27) and contributes to increased tissue stiffness. LOX is a major enzyme mediating collagen and elastin cross-linking. A relationship between LOX enzymatic activity and tissue stiffness was established in colorectal cancer and indicated a pivotal role of LOX-associated stiffness in driving colorectal cancer progression (12, 28). In obese subjects, scAT LOX gene expression is increased (11). Increased perioperative scAT pericellular collagen is associated with increased tissue stiffness measured by AdipoScan (Echosens) (9). Moreover, pericellular collagen leads to adipocyte constraints and stimulates genes encoding mechanosensitive, inflammatory, and profibrotic proteins such as CTGF in a three-dimensional model (11). Herein decreased LOX gene expression and protein and increased elastin twist structure evaluated by SHG after weight loss clearly suggest decreased cross-linking and relaxed fibers.

scAT stiffness measured by AdipoScan (Echosens) relates to adipose tissue rigidity in severe obesity before weight loss and is associated with picrosirius-red-stained collagens and metabolic alterations (9). Surprisingly, we found increased post-BS collagen accumulation without significant change in average scAT stiffness measured by AdipoScan despite large interindividual variability. These results, associated with improved metabolic alterations after BS, suggest that the major ECM remodeling observed after weight loss might be adaptive. Profiles of stiffness changes were observed but without significant link with clinical parameters at this stage. In addition, our results also suggest that transient elastography AdipoScan might be more sensitive to severe cross-linked and dense fibrosis (ie, pathological fibrosis) as shown in liver stiffness measurements (29, 30), thus explaining why AdipoScan fails to detect small decreases in post-BS stiffness or alternatively to quantify adaptive ECM remodeling (ie, less cross-linked and more degraded collagens) not linked to pathological conditions. Therefore, AdipoScan might be more appropriate to better stratify obese individuals before any drastic weight intervention or to noninvasively predict weight loss outcomes (9), a feature that needs further study in extended cohorts. Furthermore, other scAT changes occurring after weight loss might also influence tissue stiffness, such as the amount and types of lipids in adipocyte or scAT vascularization. In addition, some genes involved in mechanotransduction pathway YAP/TEAD were unchanged, whereas the downstream profibrotic gene CTGF was down-regulated, suggesting again that weight loss induced increased collagen deposition was not associated with pathological constraint.

The transcriptomic study performed before and 1 year after BS confirmed intense tissue remodeling. These results align with other observations of decreased major ECM gene and profibrotic proteins both after short-term BS-induced weight loss (8) or dietary intervention (31). We herein suggest that increased picrosirius-red staining is, at least partially, due to increased degraded collagens (collagen I, III) and eventually less newly synthesized collagens (collagen III) as shown by IHC and ELISA. Indeed, we found decreased staining of specific collagens such as collagen I and VI. Importantly, we went beyond the transcriptomic results obtained by McChulloch et al (14), who observed only increased post-BS COL6A3 expression but not other collagen VIα chains or their protein content. Our microarray analysis displayed different expression changes of collagen VI α chain: COL6A1 and COL6A2 decreased, whereas COL6A3 increased. It is well known that transcriptomic changes of subtype chains do not always relate to the same changes at the protein level. According to our immunostaining results, we found less collagen VI surrounding adipocytes after BS.

Our zymography analysis in scAT revealed the presence of proMMP-2 and/or proMMP-9 proteins in obese nondiabetic subjects. proMMPs are the inactive zymogen forms. There are growing evidences of the ability of proMMP-2 and proMMP-9 to directly activate classical signaling pathways involved in cell growth, survival, migration, and angiogenesis (32). In metabolically healthy obese individuals, scAT proMMP-9 zymographic activity is increased, suggesting that proMMP-9 might be linked with better metabolic profile (33). The fact that we did not observe the difference in obese diabetic individuals seems to be in accordance with this last point or could also be due to the effect of antidiabetic drugs. Exploring the coexpression of proMMPs and TIMPs in the context of scAT remodeling and improved metabolism deserves further consideration.

The mechanisms leading to fibrosis synthesis and degradation at the cellular level need to be better delineated in AT. Adipose tissue macrophages are triggers of fibrosis (34). We previously showed that both diet and BS-induced weight loss improve inflammatory profiles despite nonnegligible interindividual variations (24, 35). Here we observed increased CD163⁺ to CD68⁺ ratio due to increased CD163⁺ cells and decreased CD68⁺cells during weight loss, a profile of activated state of adipose tissue macrophages shifted toward M2 relative to M1, as previously shown after 3 months after BS (24). In addition, CD163⁺ cells before BS associated with pericellular collagen accumulation, indicating a role in the generation of fibrosis in obese scAT. M2 cells have a complex role in tissue repair and fibrosis: in addition to direct effects of M2 cells on promoting and suppressing collagen synthesis and fibrosis development, M2 cells are inducers of T-regulatory cells, which are implicated in fibrosis suppression and can directly produce MMPs and TIMPs, thus controlling ECM turnover (36). The reason we found a significant negative association between CD163⁺ cells and collagen accumulation at T12 is unknown but may suggest a balanced involvement of several cell types during this remodeling process and warrants further exploration. Our previous studies that described changes of scAT immune cells before and after weight loss have used the IHC method for cell quantification (24, 37). However, due to the clinical difficulties in acquiring sufficient and repeated post-BS scAT surgical biopsy samples in obese subjects during the follow-up, it was hard to compare our IHC observation with other methods such as fluorescence-activated cell sorting for quantifying immune cells infiltration.

Some questions remain unanswered. Our clinical study aimed at evaluating the changes in scAT ECM until 1 year, the nadir point of post-BS weight loss in many individuals (38). The kinetic changes (amount, type, cross-linking) of collagen fibers with longer duration of post-BS weight loss, stabilization, or weight regain remain to be evaluated. Some studies showed interesting results. For example, 2 years after BS weight stabilization, ex-obese subjects still presented the same amount of picrosirius-red-stained scAT fibrosis as morbidly obese subjects, despite improvements in adipocyte hypertrophy and inflammation infiltration (39). However, this ex-obese group was compared with an independent group of pre-BS obese individuals, which might induce bias in the results due to important intervariability in adipose tissue fibrosis. Therefore, these findings should be confirmed in samples from same individuals obtained before and after BS, as we herein assessed. Furthermore, the type of collagens and cross-linking enzymes were not investigated. In addition, obese subjects experience periods of weight fluctuations even after BS (38) that could possibly subsequently modify their adipose tissue ECM characteristics. We previously showed that 59 subjects who underwent RYGB after an initial failure of gastric banding displayed significantly higher total collagen accumulation than primarily operated subjects (9), suggesting again that weight fluctuations impact on ECM remodeling. Therefore, it is of interest to pursue the follow-up of our obese subjects, who were already investigated at baseline and follow-up until 1 year, to evaluate longer-term scAT remodeling and potential relationships with BS outcomes. In addition, there are very few data concerning the change of visceral adipose tissue characteristics. In one human study, obese subjects displayed decreased fat diameter in visceral adipose tissue as measured by ultrasound (40). In rodent, mice that underwent BS demonstrated decreased infiltration of T lymphocytes and macrophages in visceral adipose tissue (41). Further study in these post-BS features in humans would be of major interest. However, there are clinical and ethical limitations to such explorations, and the development of noninvasive measures (eg, imaging) is indispensable.

In conclusion, this study provides new insights into scAT adaptation during drastic weight-loss and shows that increased picrosirius-red staining is a signature of tissue remodeling with increased collagen degradation and fewer cross-linked fibers. It will be critical to follow up patients during long-term weight loss and to determine the impact of scAT remodeling on metabolic improvements.

Acknowledgments

We are grateful to the patients who contributed to this work and especially those who accepted repeated surgical biopsies during the follow-up. We thank Valentine Lemoine for the patients' follow-up, Florence Marchelli for the data management, and Rohia Alili for her contribution in biobanking. We also thank Frédéric Charlotte, Annette Lescot, and Anne Gloaguen for the scAT tissue preparation and picrosirius-red staining. We thank Victoria Dubar for helping in the immunohistostaining. We also thank Claire Lovo, Aurélien Dauphin, and Christophe Klein for performing the SHG acquisition and for their help in the analysis (Plate-forme d'Imagerie Cellulaire Pitié Salpêtrière). We thank Nataliya Sokolovska for her help in the K-means for longitudinal data analysis. We also thank Brandon Kayser (Institute of Cardiometabolism and Nutrition) for editorial/writing support.

Clinical trial registration number for this study (clinicaltrials.gov) was NCT01655017.

This work was supported by several clinical research contracts (Assistance Publique-Hôpitaux de Paris Clinical Research Center Grant FIBROTA [to J.A.-W. and K.C.] and Programme Hospitalier de Recherche Clinique Grant 0702 to [K.C.]) and funding from the Fondation pour la Recherche Médicale (Grant DEQ20120323701), the National Agency of Research (Adipofib), the national program “Investissements d'Avenir” with the reference Grant ANR-10-IAHU-05 and Conventions Industrielles de Formation par la Recherche Grant 2012/1180.

Disclosure Summary: Y.L. received support from Echosens for her PhD program, M.S. and V.M. are employees of Echosens. The other authors have nothing to disclose.

Abbreviations

 
• BMI

  body mass index

 
• BS

  bariatric surgery

 
• CTGF

  connective tissue growth factor'

 
• ECM

  extracellular matrix

 
• HDL

  high-density lipoprotein

 
• IHC

  immunohistochemistry

 
• LOX

  lysyl oxidase

 
• LOXL

  LOX-like

 
• MMP

  matrix metalloproteinase

 
• RYGB

  Roux-en-Y gastric bypass

 
• scAT

  sc adipose tissue

 
• SHG

  second-harmonic generation

 
• SWS

  shear wave speed

 
• T0

  baseline

 
• T3

  3 months

 
• T12

  12 months

 
• TIMP

  tissue inhibitor of metalloproteinase.