Changes in abdominal subcutaneous adipose tissue phenotype following menopause is associated with increased visceral fat mass

Menopause is associated with a redistribution of adipose tissue towards central adiposity, known to cause insulin resistance. In this cross-sectional study of 33 women between 45 and 60 years, we assessed adipose tissue inflammation and morphology in subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) across menopause and related this to menopausal differences in adipose tissue distribution and insulin resistance. We collected paired SAT and VAT biopsies from all women and combined this with anthropometric measurements and estimated whole-body insulin sensitivity. We found that menopause was associated with changes in adipose tissue phenotype related to metabolic dysfunction. In SAT, postmenopausal women showed adipocyte hypertrophy, increased inflammation, hypoxia and fibrosis. The postmenopausal changes in SAT was associated with increased visceral fat accumulation. In VAT, menopause was associated with adipocyte hypertrophy, immune cell infiltration and fibrosis. The postmenopausal changes in VAT phenotype was associated with decreased insulin sensitivity. Based on these findings we suggest, that menopause is associated with changes in adipose tissue phenotype related to metabolic dysfunction in both SAT and VAT. Whereas increased SAT inflammation in the context of menopause is associated with VAT accumulation, VAT morphology is related to insulin resistance.

Adipocyte size and number. Adipocyte size distribution differed significantly between pre-and postmenopausal women in both SAT (p < 0.01) and VAT (p < 0.01) with a higher frequency of smaller adipocytes in premenopausal women and of bigger adipocytes in postmenopausal women. Adipocyte size distribution in perimenopausal women was not significantly different from either pre-or postmenopausal women ( Fig. 1A-D). Menopausal status did not affect total number of adipocytes in SAT (p = 0.31) or VAT (p = 0.86) (Fig. 1E). Adipocyte number per kg SAT was only borderline associated with menopausal status with a tendency to fewer adipocytes after menopause (p = 0.07). Postmenopausal women had 15% (95% CI 5-24%) fewer adipocytes per kg VAT compared to premenopausal women (Fig. 1F).
Aging. SAT inflammation and -morphology was not correlated to age. Increasing age was significantly asso-

Discussion
The key findings of this study were that postmenopause is associated with changes in AT phenotype related to metabolic dysfunction in both SAT and VAT. Whereas increased SAT inflammation in the context of menopause is associated with VAT accumulation, changes in VAT morphology is associated with decreased estimated IS.
Adipocyte hypertrophy was evident in both SAT and VAT of postmenopausal women and the number of adipocytes per kg fat was reduced with menopause. Even the calculated total number of adipocytes seemed to be 5-15% lower in the postmenopausal women, although this difference did not reach statistical significance. The fact that loss of endogenous sex hormone production is associated with adipocyte hypertrophy is in accordance with studies in rodents, where oophorectomy leads to an increase in adipocyte size in both inguinal and gonadal adipose depots [18][19][20] . The lower number of adipocytes in postmenopausal women indicate that a number of adipocytes might perish in relation to menopause. Estrogen has been shown to regulate cell death in several cell types 22,23 . However, whether the lower adipocyte number after menopause in this study reflects adipocyte apoptosis induced by estrogen withdrawal remains speculative. What drives the adipocyte hypertrophy with menopause is yet to be understood. In this study, we investigated the mRNA expression of a subset of adipogenic and lipogenic markers. We found the key rate-limiting enzyme in de novo lipogenesis, FAS, to be decreased in SAT of postmenopausal women. FAS expression has previously been shown to be lower with obesity 24 , and we speculate that the downregulation of FAS in this study could reflect a compensatory attempt to avoid further lipid deposition in the adipocyte. Inability to store excess lipids in SAT is known to cause lipid overflow and ectopic lipid deposition. In accordance with this, we found the low FAS expression in SAT to be strongly associated with increased VAT deposition. VAT PPAR-γ expression was increased in postmenopausal women which could reflect an attempt to overcome the increased need for lipid storage in this depot. However, this alone is unlikely to explain the substantial differences in VAT mass across menopause. Studies in rodents generally show increased mRNA expression of lipogenic and adipogenic markers in both inguinal and gonadal depots following oophorectomy 17,25,26 but functional and mechanistic studies are highly warranted to understand how adipogenesis and lipogenesis are regulated with loss of ovarian function.
Adipocyte hypertrophy is believed to cause hypoxia contributing to the induction of AT inflammation 27,28 . In relation to this, we found that increased adipocyte size was associated with increased HIF-1α expression (the major mediator of the hypoxic response 28 ) in both SAT and VAT. Furthermore, the HIF-1α promotor region is known to contain estrogen response elements and estrogen has been shown to ameliorate HIF-1α signaling to prevent AT hypoxia and fibrosis 29 . Consequently, both the increased adipocyte size and estrogen withdrawal could contribute to the vastly increased expression of HIF-1α in SAT after menopause. Postmenopausal women displayed increased PcF in both SAT and VAT. A previous study showed that increasing AT PcF was associated with a decreased ability to loose fat 15 . Thus, we speculate, that the pathologic AT remodeling could compromise the ability to lose weight, e.g. through exercise, following menopause. Estrogen has been suggested to prevent the production of reactive oxygen species (inducers of HIF-1α expression), deteriorate mitochondrial function, and counteract healthy adipose tissue expansion [30][31][32][33] . Taking this into consideration, future studies should investigate the possible role of reactive oxygen species in the relationship between loss of ovarian function and changes in adipose tissue phenotype.
Estrogen stimulates the expression of VEGFA (inducing neovascularization in the tissue to avoid hypoxia) 34 . Paradoxically, we showed that VEGFA expression in VAT was increased with menopause and propose that this could be a compensatory reaction to a hypoxic environment in the AT 35 . Adiponectin improves oxidative conditions and enhance insulin sensitivity 36,37 . Thus, we speculated that an increased expression of adiponectin in premenopausal adipose tissue could contribute to the beneficial metabolic phenotype in this group of women. However, we saw no differential expression of adiponectin in either SAT or VAT across menopause. This is in accordance with a previous study investigating circulating adiponectin levels in both humans and mice 38 .
The mechanisms underlying the redistribution of fat towards increased VAT deposition with menopause are unknown. In this study, increased expression of inflammatory markers, immune cell infiltration, hypoxia, and decreased expression of FAS in SAT were all associated with increased VAT deposition, even after correcting for www.nature.com/scientificreports/ total fat mass. This finding could suggest that subcutaneous AT dysfunction might contribute to VAT deposition following menopause. It is widely debated whether subcutaneous AT serves a protective role or contributes to metabolic deterioration and this is likely to depend on the anatomic location-as well as the morphologic phenotype of the subcutaneous adipose depot 8,39 . In relation to this, immune cell infiltration in the subcutaneous fat compartment have previously been associated with increased VAT mass in both humans and rodents and genetically modified rodent models where immune cell infiltration in AT are depleted are protected against IR [39][40][41][42] . Low ESR1 expression in adipose tissue has been related to adiposity in humans 43 and studies in rodents show that AT depletion of ESR1 leads to mitochondrial dysfunction, adipocyte hypertrophy and IR 29,43,44 . In this study, we found no differences in expression of ESR1 across menopausal status in either SAT or VAT indicating that it is not a differential expression of ESR1 per se that drives the menopausal differences in AT phenotype. As expected, serum follicle stimulating hormone (FSH) increased across menopause. A recent study in rodents suggest an independent effect of FSH on adiposity as blocking FSH reduced body fat 45 . This mechanism has not been demonstrated in human but could indicate that the metabolic deterioration with menopause might be driven through both low estrogen and high FSH.
We did not find any differences in circulating lipids across menopause, despite considerable differences in both SAT and VAT adipose morphology. However, in this study, we only assessed fasting plasma lipids. Thus, it is possible that a metabolic challenge, such as energy intake or insulin clamping, is necessary to detect menopausal differences in lipid flow in this, metabolically, relatively healthy group of middle-aged women.
Apart from increased VAT MCP-1 expression, VAT inflammation was not related to estimated IS after correcting for VAT mass, suggesting that it is unlikely to be VAT inflammation per se that drives the IR following menopause. Opposite to this, markers of VAT morphology (increased adipocyte size, -PcF, and -HIF-1α) tended to be associated with decreased estimated IS and the association between all three and estimated IS were stronger than the association between estimated IS and VAT mass itself. Neither SAT inflammation nor morphology was related to estimated IS. These findings are in accordance with a previous study in males, that found increased VAT adipocyte size, but not inflammation, to predict IR and further showed that SAT phenotype was unrelated to IR 46 . The fact that VAT, but not SAT, phenotype was associated with estimated IS might be explained by the anatomic differences in location of adipose depots, with the close proximity between the VAT depot and the liver, where lipid overflow is believed to cause hepatic lipotoxicity and thereby IR 47 . IR itself has also been shown to cause adipose tissue inflammation 48 . Thus, we cannot exclude that the IR could precede and promote the observed adipose tissue inflammation in postmenopausal women and the causal line of these actions should be investigated in future longitudinal studies.
Menopause was associated with increased CD163 + cell infiltration in VAT. CD163 + cells, originally referred to as M2 macrophages, are believed to play a role in tissue repair and matrix remodeling 49 . Total number of CD163 + cells in AT is known to increase with obesity and has been associated with IR 50 . Whether the increased CD163 + cell number is compensatory or contribute to AT inflammation remains debated 51,52 . In this study, we found no association between number of CD163 + cells, in either SAT or VAT, and estimated IS suggesting that the link between the two might be subsidiary. CD20 + cell number was also increased in VAT of postmenopausal women, reflecting an increased B-cell infiltration in the adipose depot. Studies in rodents suggest an important role for AT B-cells in the early stages of high-fat diet induced IR 42 .
Our study has some limitations. The study was based on cross-sectional data and consequently, causal conclusions could not be drawn. Furthermore, we cannot exclude, that differential indications for the gynecological surgeries could be related to metabolic phenotype and thereby confound the results as both uterine fibroids and heavy menstrual bleedings are more frequent with obesity 53,54 . However, types of surgeries were evenly distributed between pre-and postmenopausal women suggesting that the confounding effect was limited.
Five women were defined as perimenopausal and due to the limited size of this study population, results from the perimenopausal group should be interpreted with precaution. The inclusion of the small but complex group of perimenopausal women in the study, could impact the statistical comparisons between pre-and postmenopausal women. Thus, statistical analyses were explored without the perimenopausal group, with no major impact on the overall conclusions.
With the cross-sectional nature of the study, we cannot rule out, that especially women in the perimenopausal group had had their last menstrual period and would therefore be allocated to the postmenopausal group if sampled shortly after.
Postmenopausal women were approximately 5 years older than premenopausal women. Studies in rodents show, that aging is associated with increased inflammation in gonadal AT [55][56][57] . In accordance with this, we showed that high age to some extent was associated with increased VAT inflammation. Due to the multicollinear nature of menopause and age, analyses were not adjusted for age. However, the age span in this study was relatively  www.nature.com/scientificreports/  Table 3. Visceral adipose tissue inflammation and morphology associated with estimated insulin sensitivity.  www.nature.com/scientificreports/ narrow and the effects of age on adipose phenotype, especially in SAT, were sparse. Thus, the contribution of age to the overall findings are believed to be modest. We conclude that menopause is associated with changes in AT phenotype related to metabolic dysfunction in both SAT and VAT. Whereas increased SAT inflammation in the context of menopause was associated with VAT accumulation, changes in VAT morphology was associated with decreased estimated IS. Based on these findings, we speculate that changes in SAT and VAT phenotype in relation to menopause could represent important mechanistic links between menopause, adipose redistribution and IR. Furthermore, the findings suggest a metabolically more detrimental effect of expanding the SAT and VAT depots after menopause. Thus, future studies should focus on efficient strategies such as diet or exercise to prevent postmenopausal metabolic deterioration and adipose tissue dysfunction.

Methods
Participants. Women between 45 and 60 years were recruited through the Department of Gynecology, Rigshospitalet, Denmark, and were all undergoing benign gynecological surgeries. Types of surgeries included hysterectomy, and uni-and bilateral oophorectomies. Included indications for hysterectomy were uterine fibroids and menstrual bleeding disorders. Indications for unilateral oophorectomies were ovarian cysts and bilateral oophorectomies were performed due to genetic predisposition to cancer.
The women were divided into premenopausal, perimenopausal and postmenopausal based on menstrual status and FSH. In accordance with international guidelines, a period of 12 months of amenorrhea is a prerequisite for the label of postmenopause 58 . A cut-off of FSH = 20 IU/l were chosen to distinguish a regularly ovulating ovary from an ovary with a gradually declining function, in accordance with larger population studies 59 .
Premenopausal women were defined as having regular menstrual bleedings (10 menstrual bleeding within the last 12 months) and serum FSH < 20 IU/l. Postmenopausal as having amenorrhea for > 12 months and FSH > 20 IU/l. Women who did not meet the subdivision criteria because of menstrual bleedings within the last 12 months despite serum FSH > 20 IU/l were defined as perimenopausal.
Premenopausal women with regular menstrual bleedings went through all tests in the follicular phase of their menstrual cycle (day [1][2][3][4][5][6][7][8]. Women with irregular bleedings were tested on a random cycle day, where serum FSH were obtained simultaneously. Postmenopausal women were also tested on a random day. A subset of the women (n = 18) also participated in a sub-study investigating ectopic lipid deposition with menopause and systemic inflammation, published previously 4,21 .
Informed consent was obtained in writing and verbally from all participants. The study was approved by the ethical committee of the Capital Region, Denmark (H-3-2014-096) and performed according to the declaration of Helsinki.

Study design.
A maximum of 1 month prior to surgery, women reported to the laboratory in the morning, on two separate days, following an overnight fast. All women were instructed not to perform any vigorous exercise 48 h prior to the experiment and abstain from coffee, tea, or alcohol 24 h before. The first experimental day included (1) health exam, and (2) oral glucose tolerance test (OGTT). The second day included (1) whole-body dual-energy X-ray absorptiometry (DXA) scan, (2) biopsies from the SAT and skeletal muscle, and (3) fitness test (VO 2 max test).
On the day of surgery, a biopsy was collected from the omentum majus when surgery was initiated. All biopsies (including omental) were taken before noon, following an overnight fast.
Body composition. Fat-and lean body masses were assessed through a DXA scan (Lunar Prodigy Advance; GE Medical Systems Lunar, Milwaukee, WI). Software (Prodigy, enCORE 2004, version 8.8; GE Lunar Corp, Madison, WI) was used to estimate regional-and total fat, and fat-free tissue masses.
Magnetic resonance imaging (MRI) of the abdomen was performed using a Siemens Magnetom Prisma 3 Tesla matrix magnetic resonance scanner (Erlangen, Germany) at 3-mm intervals as described previously 4 . All adipose tissue located from the diaphragm to pelvic floor inside the peritoneum was traced manually as the visceral fat region of interest. Multi-image analysis software 60 (Mango; Research Imaging Institute, Houston, TX) was used to calculate the total volume of visceral fat from the T1-weighted MRI sequence. A single reader, who was blinded to the menopausal status of the subjects, performed all image analyses.
Estimated insulin sensitivity. Whole-body estimated IS was calculated from an OGTT using the Composite Matsuda IS Index 61 . Blood samples were drawn at time points − 10, 0, 15, 30, 60, 90, and 120 min after drinking 83 g of glucose monohydrate (corresponding to 75 g anhydrous glucose) dissolved in 300 mL of water 62 . Baseline values were calculated as a mean of the − 10-and 0-min samples.
Adipose tissue biopsies. Biopsies from abdominal SAT were obtained with a modified Bergström needle.
VAT biopsies were collected as soon as surgery was initiated. All AT biopsies were divided in three; (1) frozen in liquid nitrogen and stored at − 80 °C, (2) fixed in 2% depolymerized paraformaldehyde plus 0.15% picric acid and embedded in paraffin, (3) isolation of adipocyte precursors. Stained cells were counted manually by a single person blinded to the menopausal status of the subjects. In order to get a representative sampling at least 100 adipocytes were counted for each ID 63 . Cells of interest were counted in at least two different regions of the embedded tissue. Quantification of cells was performed using ImageJ software (NIH, Bethesda, MD, USA).
Median volume of adipocytes was estimated using the formula V = 4/3·π·r 2 , and total number was calculated dividing the volume of the fat depot with median volume of adipocytes.
RNA isolation and quantitative real-time PCR. Total RNA was extracted from AT using RNeasy lipid tissue mini kit (Qiagen, Hilden, Germany). RNA was reverse transcribed according to the manufacturer's protocol (Applied Biosystems, Foster City, CA, USA). The RNA levels were determined by quantitative real time PCR using a ViiA 7 sequence detector (Applied Biosystems, Foster City, CA, USA). A mean value of 18S rRNA and GAPDH mRNA levels were used as endogenous control. Primer sequences are shown in Supplementary Information 4.
Due to limited tissue, five pre-and five postmenopausal women were not included in SAT PCR analyses. This was the case for two pre-and two postmenopausal women in VAT PCR analyses. VO 2 max test. VO 2 max was determined using a progressive exercise test on a bicycle ergometer (Monark 839E, Monark, Varberg, Sweden) with a 5-min warm up at 70 watts followed by a 20-W increase in workload each minute until exhaustion. VO 2 was continuously measured by indirect calorimetry (Quark b2, Cosmed, Rome, Italy).
Questionnaires. Participants completed a Minnesota Leisure Time Physical Activity Questionnaire (MLT-PAQ) to determine their physical activity levels over the last 4 months, estimated as an activity metabolic index 64 . The daily energy intake was estimated from a comprehensive 3-day food diary before the first visit. Based on this, mean caloric intake was calculated.
Statistics. All variables were log-transformed if not normally distributed. Data are presented as median (IQR) unless otherwise stated. Variables of interest were compared using a one-way ANOVA with a Fisher's Least Significance Difference post-hoc test. Regression models were checked for assumptions of the linear model, including normal distribution of the residuals, homogeneity of variance, linearity, and independent observations. Standardized β (β S ) was used to estimate the contribution of each predictor to the dependent variable, within a multiple regression analysis. Distribution of adipocyte size was assessed using a mixed model. A random effect accounting for an individual specific level was included.
Simple correlations between variables were investigated using Pearson correlation analyses. Mixed model analyses were performed in SAS Enterprise Guide version 7.1. Other statistical analyses were performed using IBM SPSS statistics version 25. P < 0.05 was considered statistically significant.

Data availability
The datasets generated during and/or analyzed during the current study are not publicly available due to Danish legislation on data protection but are available from the corresponding author on reasonable request and with permission from the Ethical Committee of the Capital Region, Denmark.