Functional role of ADAMTS5 in adiposity and metabolic health

Previous studies with gene-deficient mice (ADAMTS5-P) revealed that ADAMTS5 (A Disintegrin And Metalloproteinase with Thrombospondin type 1 motifs, member 5) plays a functional role in adiposity and metabolic health. To confirm these observations, we have performed similar studies with an independently generated strain of ADAMTS5 deficient mice (ADAMTS5-J). Upon cold exposure as well as after high-fat diet feeding (diet-induced obesity or DIO model), these knockout (KO) mice developed less subcutaneous and gonadal white adipose tissue (WAT) as compared to their wild-type (WT) littermates (reduction was more pronounced in ADAMTS5-P mice). Enhanced browning of WAT, as monitored by expression of UCP-1 was seen in the ADAMTS5-J KO mice upon cold exposure but not in the DIO model (seen in both conditions with the ADAMTS5-P mice). Brown adipose tissue (BAT) mass was not different between KO and WT ADAMTS5-J mice, either upon cold exposure or in the DIO model (in contrast to the enhanced BAT mass with the ADAMTS5-P mice). Energy expenditure and thermogenesis were not significantly different between KO and WT ADAMTS5-J mice (in contrast to somewhat enhanced levels in ADAMTS5-P mice). Insulin sensitivity was improved in the ADAMTS5-J KO mice, and they were protected against non-alcoholic steatohepatitis in the DIO model (as the ADAMTS5-P mice). These data are thus similar for both strains of KO mice, confirming specificity of the phenotype, but some quantitative and qualitative differences are also observed.


Introduction
Increasing energy expenditure by stimulating thermogenesis through activation of brown adipose tissue (BAT) and/or induction of browning of white adipose tissue (WAT) is considered a promising strategy to treat/prevent obesity and related metabolic diseases [1][2][3]. Whereas WAT is adapted to store energy as triglycerides, BAT produces heat (non-shivering thermogenesis). In brown adipocytes, the uncoupling protein-1 (UCP-1) regulates conversion of energy into heat by uncoupling ATP production from mitochondrial respiration [4]. Also in WAT adaptive UCP-1 positive adipocytes (brown in white: brite or beige) can arise, predominantly in subcutaneous (s) WAT [5]. This browning of WAT is enhanced by exposure to cold and in response to β 3 -adrenergic receptor (β 3 -AR) agonists [6]. It was reported that increasing the mass of brown and beige adipocytes in man indeed resulted in reduction of body weight and fat mass, associated with improved glucose metabolism and insulin sensitivity [7].
Several mechanisms have been identified that contribute to browning of WAT (reviewed in [8,9]), but a specific causative pathway constituting a clear therapeutic target has not yet been proposed. We have previously identified ADAMTS5, the main aggrecanase [10], as a protein that appears to play a functional role in development of WAT and BAT and in browning of WAT [11,12]. ADAMTS5 deficient or knockout (Adamts5 -/or KO) mice [13] indeed showed lower WAT mass, higher mass of interscapular BAT and enhanced browning of WAT, as indicated by elevated UCP-1 levels. KO mice displayed a small but significant improvement in metabolism and insulin sensitivity as well as enhanced thermogenesis [12,14]. These findings suggested that ADAMTS5 antigen and/or activity would promote WAT development and impair BAT development or browning of WAT. In addition, Adamts5 -/mice appeared to be protected from non-alcoholic steatohepatitis (NASH) when kept on a Western type diet [14]. Thus, neutralization of ADAMTS5 could be a potentially interesting strategy to combat obesity and to improve metabolic health.
To provide confirmation of these observations, we have obtained a second strain of ADAMTS5 deficient mice, generated independently [15]. In the present study, we report that the phenotype of these KO mice shows some similarities with the originally studied strain, but is much less pronounced.

Animals and experimental models
Heterozygous ADAMTS5-P mice [13] were a kind gift from Prof J. Sandy (Rush University Medical Center, Chicago, USA), and heterozygous ADAMTS5-J mice [15] from Prof A. Fosang (Murdoch Childrens Research Institute, University of Melbourne, Australia). The latter are also available from Jacksons Lab (#005771). The ADAMTS5-P mice were originally generated by Pfizer, in collaboration with Lexicon Genetics (The Woodlands, TX, USA) [13,16]. Briefly, exon 2 was flanked by loxP sites, in combination with Cre-recombinase under control of the EIIa-promoter. Mice were backcrossed with C57Bl/6J mice for several generations to obtain a 99.6% C57Bl/6J genetic background. Alternatively, the ADAMTS5-J mice were generated by Deltagen Inc. [15] and are now commercially available through The Jackson Laboratory (http://www.informatics.jax.org/external/ko/deltagen/1232_MolBio.html, Bar Harbor, USA). The targeting construct, containing an IRES-lacZ-neomycin cassette, substituted 134 nucleotides of exon 2.
For the diet-induced obesity (DIO) model, eight-week-old male mice had ad libitum access to drinking water and were kept on standard chow (SFD, 10.9 kJ/g) or on western high-fat diet (HFD; 22 kJ/g; kcal from 42% fat, 43% from carbohydrates and 15% from protein; E15721-34, Ssniff, Soest, Germany) for 15 weeks (http://dx.doi.org/10.17504/protocols.io.kbacsie). The animals were housed in a temperature-controlled room with a 12-hour light/dark cycle. Body weight and food intake were measured at regular intervals.
At the end of the experiments, mice were fasted for 6 h and sacrificed by i.p. injection of 60 mg/kg Nembutal (Abbott Laboratories, North Chicago, IL, USA). Blood was collected via the retro-orbital sinus on trisodium citrate (final concentration 0.01 M), and plasma was stored at -80˚C. Intra-abdominal gonadal (GON) and sWAT, interscapular BAT and liver were removed and weighed. Portions were snap-frozen in liquid nitrogen for RNA or protein extraction or processed for histology.

Indirect calorimetry
After 10 weeks of HFD feeding, male WT and KO ADAMTS5-J mice were individually housed in automated Calocages for indirect calorimetry (PhenoMaster CaloCages; TSE systems, Bad Homburg, Germany) during 72 h on a 12h-dark/light cycle at 22˚C with ad libitum access to HFD and water. Prior to actual measurements, mice were adapted in regular filter-top cages for 7 days to single housing and specific drinking nipples, followed by a 48 h adaptation period in the Calocages. Food intake, oxygen consumption (VO 2 ), carbon dioxide production (VCO 2 ), heat generation and ambulatory activity were continuously recorded over a 24 h period. Respiratory exchange ratio (RER = VCO 2 / VO 2 ) and energy expenditure (EE = 1.44 x RER (3.815 + 1.232 x VO 2 )) were calculated [17]. Spontaneous locomotive activity was defined as total counts of infrared light beam breaks along the X-Y axes.

Gene expression analysis
DNA-free total RNA was extracted using the RNAeasy kit (Qiagen, Basel, Switzerland) according to the manufacturer's instructions. RNA concentrations were measured spectrophotometrically and total RNA samples were stored at -80˚C. Complementary DNA was prepared from total RNA using the TaqMan 1 Reverse Transcription Reagents (Applied Biosystems, Foster City, CA, USA). PCR reactions were performed from 10 ng/μl total RNA at 25˚C for 10 min, followed by amplification at 48˚C for 1 h and finally 5 min at 95˚C. Quantitative real time PCR was performed in the ABI 7500 Fast Sequence detector using the TaqMan 1 Fast Universal PCR Master Mix and TaqMan 1 Gene Expression Assays (Applied Biosystems), reported elsewhere [12,14]. Fold differences in gene expression were calculated with the ΔΔCt method, using β-actin as housekeeping gene (http://dx.doi.org/10.17504/protocols.io.kbgcsjw).

Statistical analysis
Data are expressed as means ± SEM. Differences between two groups were analyzed with the non-parametric Mann-Whitney U test, compatible with small sample sizes. Analysis was done with Prism 5 (GraphPad Software Inc., San Diego, CA, USA). Two-sided values of p < 0.05 are considered statistically significant.

Characterization of ADAMTS5-P and ADAMTS5-J mice
Heterozygous breeding pairs of ADAMTS5-J mice (26 couples monitored) produced on average 5.5 ± 0.5 pups per litter with a mean time of 33 ± 1.7 days between litters; corresponding values for ADAMTS5-P mice (14 couples) were 6.4 ± 0.6 pups and 31 ± 2.0 days. Gender distribution was equal, with 51% male and 49% female for ADAMTS5-P versus 52% and 48% for ADAMTST5-J mice. For 295 ADAMTS5-P and 306 ADAMTS5-J mice that were genotyped, the genotype distribution followed the expected inheritance pattern for both males and females ( Table 1). The somewhat skewed distribution for the ADAMTS5-J mice may be due to a higher mortality of the KO offspring.
At 8 weeks of age, body weights of male ADAMTS5-J WT and KO mice ( Table 2) were indistinguishable from their ADAMTS5-P counterparts [11]. Both strains were genotyped as described [13,15] yielding the expected fragments at 650 bp or 380 bp for WT or KO ADAMTS5-P mice, with corresponding bands at 271 bp or 424 bp for WT or KO ADAMTS5-J mice (Fig 1).

Effect of ADAMTS5 deficiency on adipose tissue
ADAMTS5-J homozygous deficient and WT mice showed comparable body weights before and after (p = 0.08) 15 weeks of HFD feeding. Food intake was comparable for both genotypes ( Table 2). After 15 weeks, no significant differences were observed in sWAT or GON WAT, nor BAT mass ( Table 2). Gene expression analysis in sWAT confirmed the ADAMTS5 KO genotype (Fig 2A). Expression of Cidea or Ucp1 was, however, not different between WT and Adamts5 -/--J strains (Fig 2B and 2C). UCP-1 protein was around the detection limit of Western blotting for WT ADAMTS5-J samples (as also observed for WT ADAMTS5-P mice), but also for Adamts5 -/--J samples (in contrast to Adamts5 -/--P mice) (Fig 2D).
Analysis of plasma metabolic parameters did not reveal significant differences in glucose, cholesterol or triglyceride levels between both strains. Insulin levels were lower, resulting in a reduction of the HOMA-IR by 50% (Table 3).
Indirect calorimetry with ADAMTS5-J mice after 10 weeks of HFD feeding revealed weight loss for WT as well as KO mice during the 72 h experimental period ( Fig 3A). Food ( Fig 3B) and water (Fig 3C) intake, as well as VO 2 , VCO 2 and calculated RER were not different between genotypes (Fig 3D-3F). Locomotive activity was higher during phases of darkness, but was not different between genotypes (Fig 3G). Energy expenditure ( Fig 3H) and heat production ( Fig 3I) were also not significantly different.
In our previous study with ADAMTS5-P mice exposed to cold for 2 weeks, we observed for the KO as compared to the WT mice, (i) a significantly higher BAT mass, (ii) loss of body weight, and (iii) markedly higher expression of UCP-1 in sWAT. Therefore, we have also exposed WT and Adamts5 -/--J mice to 4˚C for 2 weeks. Whereas sWAT and GON AT mass at the end of the experiments were significantly lower for KO mice, no effect was observed on total body weight or BAT mass (Fig 5A and 5B). Expression of the browning markers Ucp1, Cidea and Pgc1α, but not PR domain-containing 16 (Prdm16), was slightly enhanced for KO mice (Fig 5C). Western blot analysis confirmed enhanced UCP-1 protein levels in sWAT of Adamts5 -/--J mice, as compared to WT mice ( Fig 5D). Fig 5E illustrates the enhanced sWAT browning of Adamts5 -/--J versus WT mice in sections stained with H&E or UCP-1 antibodies.

Effect of ADAMTS5 deficiency on liver function
After 15 weeks of HFD feeding, liver weight ( Table 2) and hepatic triglyceride content ( Fig 6A) were significantly lower for KO mice. Plasma triglyceride levels were, however, comparable for WT and KO mice (Table 3). Liver enzymes including alkaline phosphatases, aspartate aminotransferase and alanine aminotransferase were non-significantly lower for the Adamts5 -/--J mice (Table 3).

Wild-type
Adamts5 -/--J in gene expression between both groups (S1 Fig). Quantitative analysis of H&E and Sirius Red stainings of liver sections (Fig 7A) confirmed reduced steatosis, hepatocyte ballooning, inflammation and fibrosis, resulting in a lower NAS score for KO mice (Fig 7B). This is in accordance with the disease severity distribution (Fig 7C). These differences between WT and Adamts5 -/--J mice are comparable, although somewhat less pronounced, to those observed in our previous study in Adamts5 -/--P mice [14], thus confirming specificity of the phenotype.
Comparison of the phenotype of the Adamts5 -/--J mice of the present study with that of the Adamts5 -/--P mice reveals some similarities supporting specificity of the phenotype, but also some quantitative and qualitative differences: (i) with respect to WAT mass: in Adamts5 -/--J mice lower sWAT and GON fat mass was found, which was significant after 2 weeks of cold exposure, but only showed a trend in the DIO model; (ii) with respect to BAT mass: in Adamts5 -/--J mice no significant difference with WT mice was observed after 2 weeks of cold exposure nor in the DIO model; (iii) with respect to browning of WAT: in Adamts5 -/--J mice higher expression of UCP-1 mRNA and protein was found after 2 weeks of cold exposure (mRNA: 1.5-to 2-fold as compared to 5-fold for Adamts5 -/--P mice). Furthermore, cold exposure for 72h led to lower UCP-1 protein levels (relative to ADAMTS5-P WT) in Adamts5 -/--J versus Adamts5 -/--P sWAT extracts. However, in the DIO model UCP-1 or Cidea expression levels in sWAT were not different between WT and Adamts5 -/--J mice (as compared to 30-fold enhancement for Ucp1 or 5-fold for Cidea for Adamts5 -/--P versus WT) [12]. Indirect calorimetry did not reveal differences in VO 2 , VCO 2 or RER (same as for Adamts5 -/--P mice), and no effect of genotype was observed on energy expenditure or heat production, in contrast to Adamts5 -/--P mice [12]. These observations are in agreement with the finding that BAT mass is only decreased in ADAMTS5 -/--P and not in ADAMTS5 -/--J mice as compared to their respective control littermates. In addition, (iv) with respect to insulin sensitivity: for Adamts5 -/--J mice a 50% lower HOMA-IR was found in the DIO model as compared to WT mice, as also seen in the Adamts5 -/--P mice (which was confirmed by insulin tolerance tests) [14], and (v) with respect to NASH: Adamts5 -/--J mice had lower liver weight and hepatic triglyceride content and developed less steatosis or fibrosis, as shown by histological stainings and expression of marker genes (as observed for the Adamts5 -/--P mice). Furthermore, with respect to the hepatic inflammatory profile: both KO strains showed reduced mRNA levels of the pan-macrophage markers F4/80 and Csf1r.
Thus, ADAMTS5 deficiency is associated with reduced WAT mass in both KO strains, and with enhanced BAT mass only in Adamts5 -/--P mice. Browning of WAT is observed in both strains upon cold exposure, but only for the ADAMTS5 -/--P strain in the DIO model (less stringent condition). The activated thermogenic profile in Adamts5 -/--P mice was due to an increased activation of the β 3 -AR signaling pathway via protein kinase A and CREB [12]. Therefore, one possible mechanism for the lack of diet-induced sWAT browning in ADAMTS5 -/--J mice may be a lower activation of the β 3 -AR pathway. Insulin sensitivity is improved in both KO strains, and both strains are protected against NASH.
Specificity of the ADAMTS5-P phenotype is further supported by our finding that in vitro differentiation into mature adipocytes of 3T3-F442A preadipocytes with stable Adamts5 gene silencing, or of embryonic fibroblasts derived from Adamts5 -/--P mice, was significantly impaired as compared to control cells [11]. De novo fat pad formation following injection of 3T3-F442A cells with Adamts5 knockdown in NUDE mice was also significantly reduced as compared to controls [11]. Furthermore, the DIO studies in Adamts5 -/--P mice were performed in total in 60 WT and 54 KO mice on HFD, in several independent studies conducted over a 3 year period. In addition, similar data were obtained using a different diet composition [14].
There may be several explanations for the quantitative differences observed between both ADAMTS5 KO strains. For both sets of studies true littermates were used, all males of comparable age and body weight and all correctly genotyped. Housing conditions and diets were also comparable. Although we have thus done every effort to perform the studies with both strains under comparable external and experimental conditions, we can not exclude differences in microbiota. It is indeed known that the bacterial composition of the gut can significantly affect the metabolic state [19,20]. The small difference in genetic background of both strains (99.6% C57Bl/6J plus 129S5 plus FVB/N for Adamts5 -/--P versus 100% C57Bl/6J for Adamts5 -/--J [13,15,21]) may also contribute, as it is known that this strongly affects DIO and metabolism in mice [22,23]. Also, modifier genes (i.e. inherited genetic variations leading to qualitative and quantitative differences in disease phenotype) may be variably expressed in different inbred mouse strains [24,25]. Furthermore, differences between KO and WT mice may become more apparent using stronger stimuli, such as prolonged cold exposure, different diet compositions, prolonged monitoring in Calocages,. . . It is also possible that during the KO procedure another gene locus has been affected, which may result in activation/ inactivation of a confounding gene. To generate the Adamts5 -/--P mice, exon 2 (encoding part of the proteolytic domain) was flanked by loxP sites, in combination with Cre-recombinase under control of the Ella promoter [13]. For the Adamts5 -/--J mice, the targeting construct containing an IRES-lacZ-neomycin cassette, substituted 134 nucleotides of exon 2, including the catalytic site [15]. Although the KO in both strains thus results in inactivation of the enzymatic activity, it remains unclear whether a truncated form of the protein may be expressed. This may affect signaling pathways (e.g. CREB mediated [12]) that are involved in UCP-1 expression. Alternatively, Gorski et al. reported the presence of a truncated ADAMTS5 species (34-40 kDa) in muscle extracts of WT mice that was undetectable in equivalent KO samples (both Adamts5 -/--P and Adamts5 -/--J) [16]. Some functions of ADAMTS5 may indeed be independent of its proteolytic activity. Thus, an anti-angiogenic and anti-tumorigenic action has been reported mediated via the first thrombospondin type 1 repeat, independent of the proteoglycanase activity [26,27]. A non-proteolytic role of ADAMTS5 has also been proposed to explain its anti-chondrogenic and pro-fibrotic effects in murine models of wound repair [16].
In any case, our observations strongly suggest that data obtained in a single strain of KO mice should be interpreted with great care. Ideally, a phenotype rescue by (tissue)specific (over)expression of the missing component should confirm specificity. In addition, a tissuespecific KO or in vivo gene silencing of the target may help to confirm the phenotype.