Obesity is regarded as a pro-inflammatory state. It is associated with low circulating levels of the adipokine, adiponectin, which is considered to be an anti-inflammatory. However, adiponectin knockout mice do not consistently demonstrate pro-inflammatory phenotypes, suggesting more complexity in the in vivo immunomodulatory effects of adiponectin than originally anticipated. Moreover, adiponectin exerts pro-inflammatory effects in some experimental systems. This contradiction has been resolved by hypothesizing that adiponectin induces tolerance to inflammatory stimuli, notably Toll-like receptor (TLR) ligands. We noticed that this effect resembled lipopolysaccharide (LPS) tolerance and therefore tested adiponectin from a variety of sources for LPS contamination. All adiponectin tested carried low levels of LPS in the range of 1–30 pg/μg of adiponectin, sufficient to produce final LPS concentrations in the pg/ml range under experimental conditions. We found that induction of tolerance to TLR ligands by adiponectin in human monocyte-derived macrophages could be reproduced by such LPS concentrations. Moreover, the LPS antagonist, polymixin B, substantially inhibited induction of tolerance by adiponectin. Furthermore, polymixin B and a naturally occurring antagonist LPS were able to partially attenuate induction of tumour necrosis factor-α and interleukin-6 in human monocyte-derived macrophages by adiponectin. Polymixin B also inhibited nuclear factor-κB and mitogen-activated protein kinase signalling elicited by adiponectin. We therefore propose that some of adiponectin’s immunomodulatory effects, in particular, its TLR-tolerising actions in human monocyte-derived macrophages, may be confounded by induction of tolerance by contaminating LPS.
Low circulating levels of adiponectin have been implicated in the pathogenesis of insulin resistance [1–3], dyslipidemia [1, 4], atherosclerosis [2, 3, 5] and inflammation [2, 6, 7] seen in association with obesity. In vitro models have suggested that the immunomodulatory actions of adiponectin are anti-inflammatory [8–14], as exemplified by its in vitro ability to attenuate lipopolysaccharide (LPS) induction of tumour necrosis factor (TNF-α) [8, 14, 15].
If this hypothesis is correct, adiponectin-deficient animal models would be expected to display inflammation-prone phenotypes, and while this has been found to be true in some instances [16–18], it has not been a consistent finding [18–20], highlighting a significant degree of complexity in immunomodulatory actions of adiponectin. Moreover, it has also been reported that adiponectin induces the inflammatory cytokines, interleukin (IL)-6 and TNF-α in some systems [15, 21, 22]. One hypothesis proposed to reconcile this apparent paradox postulates stable high circulating levels of adiponectin, typically seen in lean healthy individuals , and may induce tolerance to the pro-inflammatory effects of adiponectin and to other inflammatory stimuli including Toll-like receptor (TLR) ligands such as LPS [15, 24]. However, adiponectin itself avidly binds LPS  with the result that producing totally LPS-free recombinant adiponectin is technically very challenging. Combined with the necessity to use adiponectin at μg/ml concentrations to recreate in vivo conditions, cells may be exposed to potentially significant levels of LPS under experimental conditions. We were concerned that contamination may confound the interpretation of experimental findings and such a scenario is well described in the literature for several other immunologically active proteins (e.g. heat shock protein 60 ). Furthermore, there are several similarities between some immunomodulatory effects of adiponectin and the phenomena of LPS tolerance , specifically the ability of adiponectin to induce tolerance to LPS  and to other TLR ligands  as well as the necessity to pre-incubate cells with adiponectin for several hours prior to stimulation to observe attenuation of TNF induction [15, 24].
We therefore re-investigated the in vitro anti-inflammatory effects of adiponectin and found that low levels of LPS are present contaminating all preparations of adiponectin that we examined. We also found that these are potentially sufficient to contribute to some immunomodulatory effects of adiponectin.
Materials and methods
Reagents. Cell culture reagents used were Penicillin–Streptomycin and RPMI-1640 obtained from Cambrex (Verviers, Belgium) and fetal bovine serum from PAA (Pasching, Austria). Human recombinant macrophage-colony stimulating factor was from Peprotech (London, UK). Chloroform-extracted Escherchia coli LPS was from Alexis (Nottingham, UK) and a single batch was used throughout. Pam3Cys-Ser-Lys4 (Pam3Cys) was from Alexis, macrophage-activating lipopeptide (Malp-2), peptidoglycan and Flagellin (purified) were from Invivogen (SanDiego, CA, USA). Human recombinant adiponectin was purchased from R&D Systems (Minneapolis, MN, USA); Alexis (Lausen, Switzerland); Biovendor (Heidelberg, Germany) and Biovision (Mountain View, CA, USA). Human full-length recombinant adiponectin from R&D was used throughout for all cell culture work. Antagonist LPS, E. coli K12 msbB LPS was from Invivogen. Antibody for p38 mitogen-activated protein kinase (MAPK) (SAK7) was kindly provided by Prof. J. Saklatvala (Kennedy Institute, London, UK) and phospho-specific Abs against IκBα and p38MAPK were from New England Biolabs (Hitchin, Herts, UK). The antibody to IκBα was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and anti-mouse horseradish peroxidase from DAKO (Glostrup, Denmark).
Production of recombinant adiponectin in HEK293 cells. Human full-length adiponectin cDNA (Accession number: NM004797) was cloned into the pCEP4 expression vector by PCR (Forward primer: 5′-CAGAAGCTTATTCCATACCAGAGGGGC-3′, reverse primer: 5′-TATGCGGCCGCTCACTTGTCATCGTCGTCCTTGTAGTCGTTGGTGTCATGGTAGAG-3′) between the Hind III and NotI restriction sites, with a C-terminus FLAG-tag. After lipofectamine transfection into human embryonic kidney (HEK293) cells, and selection with hygromycin, cells were left to express in serum-free medium for 48 h and adiponectin was purified by FLAG M2 affinity column according to standard protocols .
Limulus amoebocyte lysate assay. All reagents were tested for LPS using the limulus amebocyte lysate (LAL) assay  from Cambrex (USA), according to the manufacturer’s instructions. All reagents were found to have undetectable levels apart from human recombinant adiponectin.
Cell culture. Primary human monocytes were isolated and differentiated to macrophages as previously described  by culturing in RPMI-1640 medium supplemented with 5% fetal calf serum, 1% penicillin–streptomycin and M-CSF (100 ng/ml) for 72 h. Macrophages 105/well were plated in 96-well plates and rested overnight. A final concentration of 1 or 3 μg/ml of adiponectin or LPS at final concentrations corresponding to the LPS contamination of the adiponectin i.e. 3 or 10 pg/ml LPS, respectively, was added and after 16 h, either supernatants were harvested and assayed for cytokine induction or cells were washed and fresh medium was added to stimulating concentrations of LPS (0.1, 1 or 10 ng/ml) or other TLR ligands (Malp-2 at 30 ng/ml, Flagellin at 20 ng/ml, R848 at 1 μg/ml and Pam3 at 100 ng/ml) for a further 6 h prior to harvesting and assaying supernatants. Cell viability was routinely assessed after each experiment by assaying uptake and metabolism of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) as previously described . Cells from one donor were used for each individual experiment.
ELISA. The concentration of TNF-α and IL-6 in cell culture supernatants was determined by ELISA (BD Pharmingen, San Diego, CA, USA), following the manufacturers’ instructions. Absorbance was read and analysed at 450 nm on a spectrophotometric ELISA plate reader (Labsystems Multiskan Biochromic, Vienna, VA, USA) using the Ascent version 2.4.2 software. Results are expressed as the mean concentration of triplicates ± SEM.
Western blot analysis. Western blotting for p38 MAPK and IκBα was performed as previously described . Briefly, 106 macrophages were plated per well in 6-well plates, additions (Adiponectin, final 3 μg/ml; LPS, final 10 pg/ml and polymixin B, final 100 μg/ml) were made in minimum volumes of media at the indicated time points and at the end of the time course, cells were washed with ice-cold phosphate-buffered saline, lysed in 1% Triton lysis buffer with protease inhibitor cocktail [1 mm, 4-(2-aminoethyl) benzensulfonyl fluoride hydrochloride, 800 nm Aprotinin, 20 μm Leupeptin, 36 μm Bestatin, 15 μm Pepstatin A and 14 μm trans-Epoxysuccinyl-l-leucylamido (4-guanidino) butane] and phosphatase inhibitors (1 mm dithiothretol, 1 mm sodium vanadate and 5 mm sodium fluoride) were added and electrophoresis was performed under reducing and denaturing conditions on 10% polyacrylamide gels.
Adiponectin from a range of sources was LPS contaminated
Using the LAL assay, we tested adiponectin for LPS contamination. All sources of adiponectin examined, including five commercially available and that produced in our laboratory, contained LPS at concentrations up to 30 pg/μg of protein (Table 1). For the commercially available adiponectin, this was consistent with the manufacturers’ estimates of contamination .
Table 1. Endotoxin contamination of a range of recombinant adiponectin preparations.
HEK293, human embryonic kidney cells; LAL, limulus amebocyte lysate; LPS, lipopolysaccharide; EU, endotoxin unit.
aEndotoxin content assayed by the LAL assay  and expressed as pg endotoxin/μg protein where 100 pg LPS is equivalent to 1 EU; bHis; 6x Histidine-tagged; crecombinant full length human adiponectin produced in our laboratory.
Murine myeloma line, NS0
1 to 3
LPS contamination is a sufficient and necessary condition for induction of tolerance by adiponectin
Consistent with previous reports [8, 13, 15], pre-incubation of human monocyte-derived macrophages with 1 or 3 μg/ml adiponectin for 16 h almost completely suppressed subsequent induction of TNF-α by 0.1, 1 and 10 ng/ml LPS (Fig. 1A). To assess the contribution of contaminating LPS to this effect, macrophages were pretreated for 16 h with 3 or 10 pg/ml of LPS, corresponding to the contamination seen in 1 and 3 μg/ml of adiponectin and then underwent 6 h stimulation with LPS in the ng/ml range (Fig. 1B). The attenuation of TNF induction closely mirrored that due to adiponectin (Fig. 1A). The attenuation of TNF induced by pretreating the cells with adiponectin or LPS was highly statistically significant, P < 0.0001 in all cases (Fig. 1A, B). These experiments were then repeated with Polymixin B (100 μg/ml) present during the pre-incubation with adiponectin (1 or 3 μg/ml) in order to block signalling from contaminating LPS (Fig. 1C) or during pre-incubation with LPS (3 or 10 pg/ml) as positive control (Fig. 1D). Polymixin B completely abolished the attenuation of TNF induction by adiponectin (P = not significant, Fig. 1C) and by pre-incubation with LPS. Indeed, when the cells were pretreated with LPS (3 or 10 pg/ml) in the presence of Polymixin, subsequent stimulation with 1 ng/ml LPS induced a statistically significant (P < 0.0001 and P < 0.05 respectively) enhancement in TNF induction relative to the non-pretreated cells (Fig. 1D). Furthermore, with Polymixin B present during the pre-incubation of macrophages with adiponectin, a direct dose–response relationship between LPS concentration and TNF induction is restored (Fig. 1C, D). This effect was seen in cells from all donors that we studied. No adverse effects of Polymixin B on cell viability as assessed by MTT assay were detected (Fig. 1E).
Heterologous TLR tolerance induced by adiponectin is also reproduced by contaminating levels of LPS
We also observed that adiponectin induced tolerance to a wide range of other TLR ligands, specifically, Malp-2 (agonist at TLR2/6), Pam3 (TLR1/2), R-848 (TLR7/8) and flagellin (TLR5). Pre-incubation of macrophages with 1 or 3 μg/ml adiponectin or with 3 or 10 pg/ml LPS significantly attenuated TNF-α induction by all TLR ligands investigated (Table 2). Moreover, the magnitude of TNF attenuation by each concentration of adiponectin and the contamination-equivalent LPS concentration was very similar. Specifically, the attenuation induced by 3 μg/ml of adiponectin ranged from 83.4 ± 5.25% to 93.8 ± 4.4% (mean ± SEM) and that induced by 10 pg/ml LPS ranged from 82.8 ± 5% to 93.1 ± 4.5% according to the TLR ligand used. In each case the difference in attenuation induced by 3 μg/ml adiponectin and attenuation induced by corresponding LPS concentration was not statistically significant. When 1 μg/ml adiponectin and corresponding LPS concentration was examined the same was observed except for R-848 where 1 μg/ml adiponectin induced significantly greater attenuation than 3 pg/ml LPS (72.0 ± 6.9% versus 93.9 ± 3.1%, P = 0.04) (Table 2).
Table 2. Heterologous desensitization of TLR signalling by adiponectin and LPS.
Percentage reduction (± SEM) of TNF-α induction by a range of TLR ligands in human monocyte-derived macrophages following pre-incubation with adiponectin or LPS.
Macrophages were maintained for 16 h in media alone, 1 or 3 μg/ml adiponectin or in 3 or 10 pg/ml of LPS. Cells were then stimulated with flagellin, Malp-2, R-848, Pam3Cys or LPS at the indicated concentrations for 6 h.
Data are expressed as mean (± SEM) per cent reduction of TNF-α induction compared with media-alone pretreated cultures. Incubations were carried out in triplicate cultures and results are pooled from three independent experiments from unrelated donors.
aDenotes P < 0.05 when compared to 1 μg/ml adiponectin, ns denotes result not statistically significantly different from corresponding adiponectin concentration i.e. 3 pg/ml LPS compared to 1 μg/ml adiponectin and 10 pg/ml LPS compared to 3 μg/ml adiponectin.
Flagellin (20 ng/ml)
63.6 ± 12.7
83.4 ± 5.25
78.7 ± 5.8, ns
82.8 ± 5.0, ns
Malp-2 (30 ng/ml)
69.3 ± 16.2
91.1 ± 3.2
83.7 ± 4.4, ns
88.2 ± 6.9, ns
R-848 (1 μg/ml)
72.0 ± 6.9
91.5 ± 4.0
93.9 ± 3.1a
93.1 ± 4.5, ns
Pam3Cys (100 ng/ml)
72.2 ± 15.1
93.8 ± 4.4
84.8 ± 3.7, ns
91.2 ± 3.9, ns
LPS (10 ng/ml)
68.6 ± 12.7
92.6 ± 3.44
81.5 ± 8.8, ns
93.0 ± 1.2, ns
LPS antagonists suppress direct cytokine induction by adiponectin
Further studies were performed to investigate whether other effects of adiponectin were also potentially contributed to by contaminating LPS. Incubation of monocyte-derived macrophages for 16 h with adiponectin (1, 3 and 10 μg/ml) or LPS at concentrations equivalent (3, 10 and 30 pg/ml) to the contamination seen in the adiponectin was performed. Adiponectin induced TNF-α and IL-6 (Fig. 2A, B). LPS at these concentrations was able to induce more than 50% of TNF-α induction by adiponectin (Fig. 2A) and more than 30% of IL-6 induction by adiponectin (Fig. 2B). However, when polymixin B (100 μg/ml) was included in the incubations, complete suppression of cytokine induction was observed for both LPS and adiponectin (Fig. 2A, B). A naturally occurring TLR4 antagonist LPS (E. coli K12 msbB LPS) , used at 1 μg/ml, also fully suppressed TNF-α induction by both LPS (30 pg/ml) and adiponectin (10 μg/ml) (Fig. 2D). MTT assay showed that inhibition of cytokine induction by polymixin B or antagonist LPS was not due to loss of cell viability (Fig. 2C, E).
NF-κB and p38 MAPK signalling induced by adiponectin is recapitulated by contaminating levels of LPS and is blocked by Polymixin B
Investigation of signal transduction induced by adiponectin in human monocyte-derived macrophages revealed that adiponectin (3 μg/ml) activated IκBα phosphorylation, IκBα degradation and p38 MAPK phosphorylation (Fig. 3). This was also observed when 10 pg/ml LPS was used in place of adiponectin. Polymixin B (100 μg/ml) was able to substantially inhibit signalling elicited by both stimuli (Fig. 3).
Adiponectin is regarded as an anti-inflammatory adipokine [8, 14, 15]. However, gene knockout models of adiponectin deficiency have not consistently demonstrated anticipated inflammation-prone phenotypes and in some cases are protected from inflammatory stimuli . Of three adiponectin knockout models described [1, 2, 4], that of Maeda et al.  has been shown to have an inflammation-prone phenotype in the LPS-galactosamine model of hepatitis  and in the dextran sodium sulphate (DSS) colitis model . Conversely, Pini et al.  report that adiponectin null mice do not have exaggerated inflammatory responses to systemic LPS challenge, and Fayad et al.  report that adiponectin null mice are protected from inflammation in the DSS-colitis model. We have also found that adiponectin null mice do not exhibit increased susceptibility to systemic LPS challenge as assessed by circulating cytokine levels (unpublished observations). These observations suggest that additional complexity in the immunomodulatory effects of adiponectin may exist which is yet to be fully elucidated.
It has recently been reported that avidly binding LPS may be one of adiponectin’s physiological functions . Analysis of adiponectin from a range of sources revealed consistent contamination with low levels of LPS (Table 1). Production of our own recombinant adiponectin in HEK293 was therefore undertaken and it was found that this also consistently had low-level LPS contamination. When adiponectin is used at μg/ml concentrations under experimental conditions, the resultant final LPS concentrations are in the pg/ml range, capable of signalling in human monocyte-derived macrophages. We were therefore concerned that some immunomodulatory effects of adiponectin may be contributed by LPS contamination. We found that LPS in the concentration range of 3–30 pg/ml, recapitulated some effects of adiponectin including attenuation of TNF induction by LPS after adiponectin pre-incubation (Fig. 1), heterologous desensitization of TLR signalling , as has been previously described for adiponectin by other authors  (Table 2), as well as direct induction of TNF-α and IL-6 and activation of p38 MAPK and nuclear factor-κB signalling (Figs. 2 and 3). These observations suggest that interpreting certain in vitro immunomodulatory effects of adiponectin may be complicated by LPS contamination. This is supported by the ability of LPS antagonists including Polymixin B and E. coli K12 msbB LPS to inhibit effects of adiponectin on monocyte-derived macrophages. Although neither inhibitor is likely to be absolutely specific for LPS and off-target effects of polymixin have been reported , the observation that both agents attenuate certain adiponectin effects further supports a role for LPS contamination. We also attempted to control the contamination by heat inactivation but as previously reported  found that heating for 60 min at 70 °C was also sufficient to inactivate low levels of LPS.
While clear evidence derived from knockdown of adiponectin’s specific receptors (Adipo R1 and Adipo R2) shows that immunomodulatory actions of adiponectin are mediated through these receptors [38, 39], we believe that LPS contamination may contribute under certain circumstances and should be considered. Several other factors are now recognized as contributing to the emerging complexity of adiponectin’s immunomodulatory signalling. These include differences due to use of full-length or globular adiponectin , whether high molecular weight or monomeric adiponectin is investigated , and the cell-type under investigation. We propose that some role of LPS contamination may also contribute to this complexity.
In summary, all sources of adiponectin we examined contained sufficient contaminating LPS to contribute to adiponectin’s capacity to attenuate TNF induction, which we propose may at least in part, be due to induction of LPS tolerance. Moreover, the potential contribution of contaminating LPS that we have reported may reconcile the apparent contradiction in previous reports that adiponectin has both pro- and anti-inflammatory effects [8, 15, 21, 22] and may also underlie the capacity of adiponectin to induce tolerance to a range of TLR ligands [15, 24]. Given the avidity of adiponectin to bind ambient LPS , it is highly likely that most adiponectin preparations carry some degree of LPS contamination and this effect should therefore be considered when interpreting the tolerising effects of adiponectin.
This research is funded by the Arthritis Research Campaign UK, programme grants from the MRC (G7811974) and Wellcome Trust (072643/Z/03/Z) and by an EU FP6 Integrated Project Grant LSHM-CT-2003-503041. We are also grateful for support from the NIHR Biomedical Research Centre funding scheme. JJOT holds a BHF intermediate fellowship (FS/06/001).
Contributions of the authors
JT, SS and BF conceived and designed the study. JT, MB and SS performed the experiments and data analysis. JT, SS and BF wrote and edited the manuscript.