Abstract
BACKGROUND:
Obesity-induced chronic inflammation and fibrosis in adipose tissue contributes to the progression of type 2 diabetes mellitus (DM). While fibrosis is known to induce mechanical stiffening of numerous tissue types, it is unknown whether DM is associated with alterations in adipose tissue mechanical properties.
OBJECTIVE:
The purpose of this study was to investigate whether DM is associated with differences in bulk viscoelastic properties of adipose tissue from diabetic (DM) and non-diabetic (NDM) obese subjects.
METHODS:
Bulk shear rheology was performed on visceral (VAT) and subcutaneous (SAT) adipose tissue, collected from obese subjects undergoing elective bariatric surgery. Rheology was also performed on the remaining extracellular matrix (ECM) from decellularized VAT (VAT ECM). Linear mixed models were used to assess whether correlations existed between adipose tissue mechanical properties and DM status, sex, age, and body mass index (BMI).
RESULTS:
DM was not associated with significant differences in adipose tissue viscoelastic properties for any of the tissue types investigated. Tissue type dependent differences were however detected, with VAT having significantly lower shear storage and loss moduli than SAT and VAT ECM independent of DM status.
CONCLUSION:
Although DM is typically associated with adipose tissue fibrosis, it is not associated with differences in macroscopic adipose tissue mechanical properties.
Keywords
Introduction
Adipose tissue metabolic dysfunction plays a central role in the pathogenesis of obesity-associated metabolic disease including type 2 diabetes mellitus (DM). Chronic inflammation and fibrosis are central features of dysfunctional adipose tissue. These observations suggest an important role for the adipose tissue extracellular matrix (ECM) in regulating tissue and systemic metabolism. Adipose tissue ECM deposition is increased in murine and human obesity, with upregulation of collagen and down-regulation of matrix metalloprotease gene expression [1–5], while transgenic manipulations in mice that reduce ECM deposition specifically in adipose tissue attenuate adipocyte and systemic metabolic dysfunction, and the opposite is observed in models in which ECM deposition is increased [6,7]. Other data demonstrate disease-specific regulation of adipocyte cellular metabolism by the ECM in humans: for example, our group has demonstrated that ECM from healthy non-diabetic (NDM) subjects can rescue cellular insulin resistance in tissues from unhealthy DM subjects [8].
While the ECM mediates its effects on adipocyte function in part via specific ligand-receptor interactions between ECM proteins and adipocyte scavenger receptors [5,9], other data support a role for mechanical effects of ECM on cellular function in multiple tissues [10–12], including adipose. Adipocytes cultured in stiffer substrates manifest decreased adipogenesis and cellular metabolic dysfunction [10,11], while adipocyte hypertrophy is associated with cellular stiffening [12]. Furthermore, in vivo transcutaneous shear wave elastography demonstrated increased SAT stiffness in human DM relative to NDM subjects [1]. Despite these observations, few studies focus on mechanical properties of adipose tissue in the context of obesity and metabolic disease. Indeed, most published literature uses histologic and molecular measures of ECM, including collagen-avid dye staining and alterations in expression of ECM-related genes. The role of alterations in the mechanical properties of human adipose tissue in metabolic disease are sparse.
In this manuscript, we used shear-rheology to characterize the viscoelastic mechanical properties of visceral adipose tissue (VAT), decellularized VAT ECM and subcutaneous adipose tissue (SAT) from DM and NDM humans. Given the preponderance of data demonstrating increased ECM deposition in adipose tissue in obesity and metabolic disease [1–7] as well as data demonstrating adverse effects of culture matrix stiffness on adipocyte metabolism [10,11], we hypothesized that adipose tissues from obese DM subjects would exhibit increased stiffness relative to tissues from NDM subjects. We further hypothesized that VAT would exhibit increased stiffness relative to SAT, given multiple studies suggesting VAT is more strongly associated with metabolic disease [13,14]. Considering the central role that the ECM plays in fibrosis and the role of the ECM in regulating adipose tissue metabolism [8], we also characterized mechanical properties of VAT ECM. We then assessed whether patient characteristics such as age, sex, and body mass index (BMI) correlated with intact adipose tissue sample mechanics. This is one of only a few published studies describing detailed rheological properties of human adipose tissue in the context of obesity and diabetic status.
Materials and methods
Human subjects, adipose tissue
Human subjects were consented and enrolled with Institutional Review Board approval at University of Michigan and Ann Arbor Veterans Affairs Healthcare System conforming with The Code of Ethics of the World Medical Association (Declaration of Helsinki). DM subjects were defined by clinical diagnosis requiring medication and hemoglobinA1c (HbA1c) ≥ 6.5%. NDM subjects were defined by no clinical history of diabetes and HbA1c < 5.7% per American Diabetes Association criteria [15]. VAT from the greater omentum, and SAT from the abdominal wall were collected from obese subjects during bariatric surgery at the beginning of operation and stored at −80 °C until rheological analysis. Adipose tissue samples were prepared for rheologic analysis by snap-freezing 3–5 gm pieces of tissue in liquid nitrogen, followed by cutting 200 mg tissue fragments from these frozen samples with a scalpel. Samples were kept on ice at all time for cutting.
Subject details for entire study
Subject details for entire study
Decellularized VAT ECM was prepared from freshly collected VAT as described in detail [8,16]. Briefly, VAT explants were freeze-thawed 3X from −80 °C to 37 °C in 10 mM Tris, 5 mM EDTA, 1% phenylmethanesulphonylfluoride (PMSF, pH 8.0), incubated for 24 hrs in 0.25% Trypsin/0.1%EDTA at 37 °C, then washed 3X in rinsing buffer (8 g/L NaCl, 200 mg/L KCl, 1g/L Na2HPO4, 200 mg/L KH2PO4, 1% PMSF) at 37 °C for 20 min. Nucleic acids and lipids were removed by incubation at 37 °C for 24 hrs in 55 mM Na2HPO4, 17 mM KH2PO4, 4.9 mM MgSO
Rheological methods were based on previous procedures used to determine the viscoelastic properties of porcine SAT [17]. The bulk viscoelastic properties of samples were measured by parallel plate shear rheology using an AR-G2 rheometer (TA Instruments, New Castle, DE, USA) equipped with an 8 mm diameter measurement head and a Peltier stage. The rheometer stage was maintained at 37 °C and active normal force was applied to maintain a constant axial force of 0.07 N for the duration of testing. The linear viscoelastic regime was qualitatively determined using strain sweeps from 0.05% to 10% with 4 points per decade at frequencies of 1, 10, and 100 rad/sec. Subsequent frequency sweeps from 1 to 100 rad/sec with 4 points per decade were performed at 0.1% strain to assess differences in tissue mechanics depending on DM status. Characteristic metrics of storage modulus (G ' ), loss modulus (G " ), gap height, and normal force were extracted from this frequency sweep at 1 rad/sec to assess intra-subject variability for subjects from which 3 samples were measured. Similar amounts of tissue samples were used for rheological testing. For whole VAT; 2–3 samples per subject were measured to characterize intra-subject variability. Only one sample per subject was analyzed for SAT and decellularized VAT ECM. In preparation for rheological testing, samples were retrieved from −80 °C storage and then thawed on ice. Immediately prior to measurement of a particular sample it was transferred to a 48-well plate (Corning Inc., Corning, NY, USA). Measurements were performed directly in this 48-well plate with both the bottom of the well and the measurement head coated with sand-paper to prevent sample slippage.
Linear mixed models showing the effect of DM status and tissue type on storage modulus (G
′
) for SAT, VAT and VAT ECM
Linear mixed models showing the effect of DM status and tissue type on storage modulus (G ′ ) for SAT, VAT and VAT ECM
Standardized beta coefficients; Standard errors in parentheses. All models corrected for the covariates of sex, age, preoperative BMI and sample weight as confounding variables. Model intercepts not shown. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
Linear mixed models showing the effect of DM status and tissue type on loss modulus (G ′′ ) for SAT, VAT and VAT ECM
Standardized beta coefficients; Standard errors in parentheses. All models corrected for the covariates of sex, age, preoperative BMI and sample weight as confounding variables. Model intercepts not shown. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
All data was graphed using Prism 8 (GraphPad Software Inc., San Diego, CA, USA) and presented as mean ± standard deviation while STATA version 15 (StataCorp Inc., College Station, TX, USA) was used for inferential statistics. The alpha level was set a-priori at p < 0.05. The storage (G
′
) and loss (G
′′
) moduli were log-transformed to approximate a normal distribution. Occasional negative measures of G
′
and G
′′
which can occur at high frequency measurements due to erroneous instrument inertial effects were discarded. A linear mixed model was used to estimate the effect of diabetes status and the tissue type on the log transformed storage and loss moduli from the frequency sweeps. Linear mixed models are a class of analytical models which are used to analyze non-independent or clustered data as is the case with repeated measures. These models allow both fixed and random effect estimations, wherein the random effects are used to control for correlations in the repeated measurement and thus generate unbiased model parameters and standard errors of parameters [18]. A polynomial quadratic term for log transformed frequency was used to control for the curvilinear relationship between frequency and the moduli. Additional covariates were included in the model to control for sample weight, sex, age, and preoperative BMI along with a random intercept to control for correlations within repeated experiments on samples obtained from the same individual. A dummy variable was used to estimate the main effect of diabetes and main effect of tissue type followed with separate models to estimate the effect of diabetes for each tissue type. The functional form for the regression equation (1) for G
′
is presented below (the subscripts i and j denote the ith sample of jth patient):
Average predicted values from linear mixed model and standard errors of G ′ and G ′′ by tissue type at characteristic frequencies, controlling for sample weight, gender, age and BMI
Linear Mixed Models showing the effect of sample weight, sex, age, and preoperative BMI on storage modulus (G ′ ) for SAT and VAT
Standardized beta coefficients; Standard errors in parentheses. Model intercepts not shown. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
Adipose tissue samples were taken from 41 distinct subjects, with some subjects contributing multiple tissue types for characterization. Overall, measurements were made on 67 VAT samples from 29 different subjects (14 NDM subjects, 15 DM subjects), 16 SAT samples from 16 different subjects (8 NDM subjects, 8 DM subjects), and 17 VAT ECM samples from 17 subjects (8 NDM subjects, 9 DM subjects). Insufficient SAT was available to produce decellularized SAT ECM samples. All subjects included in this study had a body mass index greater than 30 kg/m2. Additional details regarding subject sex, age distributions, the mean mass of samples measured, and the number (n) of individual subjects from which VAT, SAT, and VAT ECM samples were taken are provided in Table 1. For each of these 12 distinct categories, the mean age, BMI and sample mass are shown with the corresponding standard deviation (stdv). A schematic of the instrument set-up is shown in Fig. 1A and representative images for each type of sample is presented in Fig. 1B. Qualitatively, SAT samples tended to contain more blood than VAT samples.
(A) Cartoon schematic for rheology testing procedure in a 48-well tissue culture plate. (B) Representative morphological appearance of whole visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) and decellularized VAT ECM in a 48-well plate prior to testing. Best when viewed in color.
Strain sweeps from 0.05–10% strain for intact VAT and SAT and decellularized VAT ECM at frequencies of 1, 10 and 100 rad/sec. Storage (G ′ , solid lines) and loss moduli (G ′′ , dotted lines) are plotted for both NDM (red) and DM (blue) patient samples. Interpatient means and standard deviations are shown for each data point. Error bars are only shown in the positive direction for clarity.
Frequency sweeps from 1–100 rad/sec for intact VAT and SAT and decellularized VAT ECM at 0.1% strain. Storage (G ′ , solid lines) and loss moduli (G ′′ , dotted lines) are plotted for both NDM (red) and DM (blue) subject samples (A). The dynamic modulus (B) and phase angle (C) depending on frequency are also presented for ease of interpretation. Inter-subject means and standard deviations are shown for each data point. Error bars are only shown in the positive direction for clarity (A, B).
The results from strain sweep tests for each type of sample from 0.05 to 10% strain for 1, 10, and 100 rad/sec for both the storage modulus (G ′ ) and loss modulus (G ′′ ) are shown in Fig. 2. For each interrogated frequency, G ′ and G ′′ for each sample type were nearly independent of strain up to at least 0.1% indicating all samples acted as linearly visco-elastic (LVE) solids for small deformations. Within the LVE range, G ′′ was significantly lower than G ′ , however all three sample types exhibited strain softening for progressively increasing strains (Fig. 2). Subsequent frequency sweeps were performed at 0.1% strain from 1 to 100 rad/sec, with the results for both G ′ and G ′′ shown in Fig. 3a. As an alternate representation, the magnitude of the shear modulus (dynamic modulus) is presented in Fig. 3b, and the phase angle is presented in Fig. 3c. All three sample types demonstrated strain rate dependent stiffening. This strain rate dependent stiffening was accompanied by a general trend towards decreasing phase angle with increasing frequency for SAT and VAT samples, but not for VAT ECM samples (Fig. 3c). The intra-class correlation coefficient for all VAT samples was 0.74 and in a small subset of subjects, 3 VAT samples per subject were collected (NDM = 4, DM = 5) to provide an additional estimate of intra-subject variability. The intra-subject mean standard deviation for G ′ from the frequency sweeps at 1 rad/sec for these 9 subjects was 3.16 kPa (mean characteristic storage modulus of 8.41 kPa).
Linear mixed models estimating the effect of diabetes in samples obtained from SAT, VAT, and VAT ECM showed that overall, the storage modulus (G ′ ) and loss modulus (G ′′ ) did not significantly differ between DM and NDM samples (Table 2, 3). However, samples obtained from SAT (𝛽 = 0.458 ± 0.041; p < 0.001) and VAT ECM (𝛽 = 0.633 ± 0.076; p < 0.001) had significantly higher storage moduli (G ′ ) than samples obtained from VAT after controlling for frequency, sample weight, sex, age and BMI (Table 2). Similar to G ′ , samples obtained from SAT (𝛽 = 0.440 ± 0.041; p < 0.001) and VAT ECM (𝛽 = 0.922 ± 0.078; p < 0.001) had significantly higher G ′′ than samples obtained from VAT (Table 3). The significant coefficients for the quadratic log frequency shows the curvilinear relationship between log(frequency) and log(moduli) as depicted in Fig. 3 (Tables 2, 3). Predicted average values and standard errors of G ′ and G ′′ by tissue type at characteristic frequencies are presented in Table 4. Additional linear mixed models showed that there was no signficant effect of sample weights, sex, age, and preoperative BMI on either the storage or loss moduli of SAT and VAT samples after controlling for frequency (Tables 5, 6).
Linear Mixed Models showing the effect of sample weight, sex, age, and preoperative BMI on loss modulus (G ′′ ) for SAT and VAT
Standardized beta coefficients; Standard errors in parentheses. Model intercepts not shown. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
We demonstrated no association between DM status, age, sex, or BMI and tissue mechanics as measured by macro-rheology in excised human VAT or SAT. Together, these observations suggest that published literature demonstrating increased fibrosis in DM adipose tissue [1–4] is not necessarily paralleled by increased tissue stiffness in metabolic disease, suggesting complex relationships between histologic measures of tissue fibrosis and mechanical properties. Nonetheless, we did observe decreased stiffness of VAT relative to SAT, suggesting depot-specific differences in mechanical properties may play a role in adipose tissue biology.
Our observation of VAT stiffness being lower than SAT stiffness is in contradiction to a previous report in which tensile testing demonstrated SAT had a lower modulus than VAT [19]. However, the mean BMI of subjects from which VAT and SAT were taken (30.2 ± 6.4 and 28.9 ± 6) in that study was much lower than in this study (44.9 ± 5.7 and 45.9 ± 5.8). Human SAT adipocytes demonstrate increased adipocyte hypertrophy relative to VAT adipocytes [20,21], and adipocyte hypertrophy results in cellular stiffening [12], which may explain our results. Our study also failed to corroborate a previous study in which vibration controlled transient elastography (VCTE) was used to demonstrate that subjects with diabetes and high fibrosis scores in SAT biopsies tended to have higher shear wave velocity measurements in SAT [1]. VCTE is a non-invasive technique in which the velocity of shear waves propagating through tissue is measured to assess differences in tissue stiffness based on shear waves propagating faster in stiffer tissues. This discrepancy may in part be due to differences between the in vivo and ex vivo environment, and as such we acknowledge that caution must be exercised in extrapolating our ex vivo results to in vivo characteristics. Nonetheless, all samples were processed and measured identically, supporting validity of the comparisons between groups studied. However, limitations in sample number did not permit controlling for multiple clinical characteristics of the subjects from whom samples were collected, including medications, sex, age, other metabolic diseases, and other clinical variables. Furthermore, tissues were derived from obese patients undergoing bariatric surgery, a select subgroup of all patients with metabolic disease. Larger studies will be required to address these and other subgroup analyses and permit extrapolation of our results to more general patient populations. Limitations in tissue amounts precluded analysis of protein and lipid contents in the samples studied in this manuscript, but prior data from our group demonstrate increased adipocyte size, and thus possibly increased lipid content, in DM relative to NDM adipose tissues [22]. This may be a contributing factor to differences in rheologic properties, although other factors such as differences in ECM composition and quantity likely also contribute. Finally, fibrosis may result in more localized changes to tissue mechanics, rather than macroscopic tissue stiffening. As such, more refined techniques such as atomic force microscopy may be more appropriate for assessing microscale spatial differences in tissue stiffness for ex vivo tissue samples on a scale comparable to a cell.
Our measurements of human SAT is in good agreement with previous measurements of human and porcine SAT [17,23]. Considering the role adipose tissue plays in providing mechanical protection for organs and various anatomical structures, these data may inform the development of a constitutive model to predict mechanical load transfer through adipose tissue [17,24]. These results may also be relevant in the context of fat grafting for surgical reconstruction of soft-tissue deficits. The cannula geometry used for fat graft injection has previously been shown to affect the modulus of the graft, with higher moduli grafts associated with more intact tissue and higher associated graft retention [25]. Surgeons may need to consider both the intra- and inter-subject variability of adipose samples when optimizing the efficacy of fat grafts. Our results also demonstrate that decellularization results in a stiffer ECM substrate. This may be the result of protein crosslinking, denaturation, and other biochemical changes resulting from the decellularization process, which involves multiple freeze-thaw treatments. These changes to tissue mechanical properties may be important to consider when employing decellularized adipose tissue as scaffolds for tissue engineering applications [23,26]. In conclusion, these data are among the few detailed characterizations of human adipose tissue mechanical properties and help provide initial insights into the effects of anatomic depot and disease on adipose tissue mechanics with implications for both metabolic disease and tissue engineering.
Footnotes
Acknowledgements
We thank Danielle Berger NP, Terra Babas MA, Christine Bridge MA, Simone Correa MA, Retha Geiss MA, Jennifer Lavey, Andrew Schlaud, MA, and Marilyn Woodruff NP for assistance with study coordination. This project was supported by NIH grants R01DK115190 (RWO, CNL), R01DK090262 (CNL), R01-HL085339 (AJP), and by the Veterans Affairs Merit Review Award I01CX001811 from the U.S. Department of Veterans Affairs Clinical Sciences Research and Development Service (RWO). BAJ was partially supported by the Tissue Engineering and Regeneration Training Program at the University of Michigan (T32-DE007057) and by the Training Program in Translational Cardiovascular Research and Entrepreneurship (T32-HL125242).