The adipocyte exerts an important role in energy homeostasis, both as depot for energy-rich triglycerides and as a source for metabolic hormones. Adipocytes also contribute to inflammation and the innate immune response. Although it can be physiologically beneficial to combine these two functions in a single cell type under some circumstances, the proinflammatory signals emanating from adipocytes in the obese state can have local and systemic effects that promote atherosclerosis and insulin resistance. The transcriptional machinery in the adipocyte that mediates these pro-inflammatory responses has remained poorly characterized to date. In particular, no information is currently available on the NF-κB family of transcription factors. Here, we show that adipogenesis is associated with changes in amount and subunit composition of the NF-κB complexes. NF-κB subunits p65 (RelA), p68 (RelB), and IκB are upregulated during fat cell differentiation. Correspondingly, basal NF-κB nuclear gel shift and luciferase reporter assays are induced in parallel during differentiation. Surprisingly, endotoxin sensitivity of the classical NF-κB pathway is substantially delayed and attenuated despite increased overall inflammatory response in the mature adipocyte, as judged by induction of IL-6 and TNF-α. As a reflection of the constitutively elevated NF-κB activity in the mature adipocyte, adipocytes (but not preadipocytes) exert a strong inflammatory stimulus on macrophages in vitro, suggesting a cross talk between adipocytes and interstitial macrophages in adipose tissue in vivo. These effects are mediated by a secretory product of adipocytes that is unlikely to be IL-6 or TNF-α.
- nuclear factor-κB
a growing body of evidence demonstrates extensive interdependence between metabolic dysregulation, atherosclerosis, inflammation, and innate immunity. New studies show that the causative relationships between obesity, insulin resistance, and atherosclerosis are mediated not only by associated hyperlipidemias but also by coexisting inflammatory states. A central player in these processes is the adipocyte. A number of recent reports have demonstrated an intricate link between metabolic control, innate immunity, and inflammation at the level of the adipocyte, and dysregulation at the level of any one of these transcriptional programs has a profound impact on the other cellular processes. There is growing evidence implicating factors mediating innate immunity in the pathogenesis of atherosclerosis: elevated levels of acute-phase reactants such as IL-6 and C-reactive protein and decreased levels of the adipose-specific secretory protein adiponectin, in particular, are highly correlated with cardiovascular problems (24, 30, 33, 38, 52). Elevated levels of inflammatory mediators are also associated with insulin resistance and type II diabetes (41, 42). TNF-α secreted from adipocytes mediates insulin resistance in an autocrine fashion (13, 20, 21). Inhibition of the inflammatory signaling complex IKKβ, which acts through the transcription factor NF-κB, improves the insulin resistance and dyslipidemia associated with obesity (22, 28, 36, 51). A recent report correlated the improved inflammatory profile found in patients with mutations in toll-like receptor (TLR)-4 (the receptor for bacterial lipopolysaccharide) with decreased risk for atherosclerosis (27). Although not yet fully appreciated in the literature, the adipocyte provides a comprehensive cellular nexus for the processes of inflammation and metabolic dysregulation, sensing and secreting factors involved in both processes. The adipocyte displays a high level of sensitivity to bacterial lipopolysaccharide (LPS), TNF-α, IL-6, interferon-γ, and a host of other factors. Activation of NF-κB by TNF-α was shown to cause dedifferentiation of adipocytes in culture, an effect specifically antagonized by the adipogenic transcription factor peroxisome proliferator-activated receptor (PPAR)γ and mediated by its newfound ability to override the inhibitory effects of NF-κB on the expression of key adipocyte genes (44, 45). In addition, TNF-α and LPS both induce expression and activity of inducible nitric oxide synthase (iNOS), a downstream target of NF-κB transcription (25). Adipose tissue iNOS induction has been observed in the obese state, and iNOS-deficient mice are partially protected from obesity-induced insulin resistance and glucose intolerance (39). Together, these findings implicate NF-κB signaling as a molecular link between inflammation and metabolic dysregulation in the adipocyte. However, because the inflammatory component of the adipocytes is a relatively new concept, very little is known about NF-κB biology in fat.
Both TNF-α ligands as well as infectious pathogen cell wall constituents such as bacterial LPS and fungal zymosan mediate their inflammatory activities through NF-κB. These pyrogens activate NF-κB through the TLRs (34). We have previously shown that in the adipocyte, LPS-mediated activation of TLR4 causes upregulation of the fungal cell wall receptor TLR2 in an inflammatory feed-forward pathway (31). We observed that the time course and dose response of TLR2 induction by LPS were significantly different in the adipocyte compared with the preadipocyte, suggesting a change in NF-κB signaling during differentiation; however, there is no published data regarding the expression or regulation of NF-κB subunits during adipogenesis. With this intriguing new data indirectly suggesting regulation of this transcription factor during differentiation, and growing evidence implicating NF-κB as an important factor at the interface of metabolism and systemic inflammation, we felt further characterization of this important inflammatory transcription factor during adipocyte differentiation was necessary.
NF-κB activation is at the core of many proinflammatory transcriptional programs (reviewed in Refs. 16, 26). NF-κB activation entails cytoplasmic deactivation/degradation of inhibitors of NF-κB (IκB-α, -β, and -γ and the COOH terminus of p100) that then release the sequestered subunits of NF-κB transcription factors. These subunits, RelA (p65), p50, RelB (p68), and p52, then translocate to the nucleus where they bind their DNA response elements as hetero- and homodimers in various combinations. In a number of cell types, activated nuclear NF-κB induces transcription of serum amyloid A3 (SAA3), IL-6, and other secretory proteins of the innate immune response (49). A number of these acute-phase markers are expressed by the adipocyte (32). In the case of SAA3, we recently demonstrated that adipocytes (but not preadipocytes) specifically induce expression of SAA3 after inflammatory or hyperglycemic stimulation. Other inflammatory cytokines expressed by the adipocyte include TNF-α, IL-6, IL-1β, complement factors B and D, plasminogen activator inhibitor-1, α1-acid glycoprotein, lipocalin 24p3, migration inhibitory factor (MIF), IL-8, monocyte chemoattractant protein (MCP)-1, and macrophage inflammatory protein (MIP)-1α (7). Additional candidates that may exert pro- or anti-inflammatory activities are the adipokines leptin and adiponectin/Acrp30 (15, 50). In addition to the inflammatory signals secreted by adipocytes themselves, adipose tissue contains a significant population of interstitial macrophage cells and preadipocytes with macrophage-like activities (8, 9, 48). The heterotypic paracrine signaling events that may be occurring between these different cell populations have been completely unexplored to date. As a second aim of this investigation, we established an assay system that probes the effects of pre- and fully differentiated adipocytes on macrophages to begin characterizing these interactions. In this isolated system, adipocytes (but not preadipocytes) exert a strong proinflammatory effect on the macrophages. These effects are mediated by a secretory product of adipocytes that is unlikely to be IL-6 or TNF-α.
DMEM was purchased from Cellgro (Herndon, VA), and murine TNF-α and IL-6 were purchased from Pharmingen (San Diego, CA). LPS (Escherichia coli) was purchased from Sigma (St. Louis, MO). Insulin was purchased from Sigma and used at 100 nM. All other chemicals were purchased from Fisher Scientific (Pittsburgh, PA).
3T3-L1 murine fibroblasts (a generous gift of Dr. Charles Rubin, Dept. of Molecular Pharmacology, Albert Einstein College of Medicine) were propagated and differentiated according to the protocol described previously (12). In brief, the cells were propagated in FCS [DMEM containing 10% FCS (JRH Biosciences, Lenexa, KS) and penicillin-streptomycin (100 U/ml each)] and allowed to reach confluence (day −2). After 2 days (day 0), the medium was changed to IDX (containing FCS and 160 nM insulin, 250 μM dexamethasone, and 0.5 mM 3-isobutyl-1-methylxanthine). Two days later (day 2), the medium was switched to FCS containing 160 nM insulin. After another 2 days, the cells were switched back to FCS. Mouse macrophage J774 cells (a gift from Dr. Matthew Scharff, Albert Einstein College of Medicine) were cultured in FCS.
Western Blot Analysis
Ten-centimeter plates of 3T3-L1 cells after treatment were washed twice with PBS and lysed in 1 ml of SDS-PAGE sample buffer (0.75% sodium dodecyl sulfate, 0.5 M Tris·HCl, pH 6.8, and 16 mM EDTA) plus 1 mM phenylmethylsulfonyl fluoride, and lysates were boiled for 5 min followed by brief sonication. Thirty micrograms of total protein were resolved by SDS-PAGE on 12% acrylamide gels and transferred to BA83 nitrocellulose (Schleicher and Schuell, Keene, NH). Blots were probed with antibodies to NF-κB subunits from Santa Cruz Biotechnology (Santa Cruz, CA), rabbit polyclonal antiserum against p52 (a kind gift of Dr. Ulrich Siebenlist, National Institutes of Health, Bethesda, MD), and rabbit polyclonal antibodies to the guanine nucleotide dissociation inhibitor (GDI; a generous gift from Dr. Perry Bickel, Washington Univ., St. Louis, MO). Western blots to nuclear proteins were performed on 5 μg of nuclear protein from the extracts obtained for the EMSA experiments.
After SDS-PAGE, proteins were transferred to BA83 nitrocellulose (Schleicher and Schuell). Nitrocellulose membranes were blocked in PBS or Tris-buffered saline with 0.1% Tween 20 and 5% nonfat dry milk. Primary and secondary antibodies were diluted in PBS or Tris-buffered saline with 0.1% Tween 20 and 1% bovine serum albumin. Bound antibodies were detected by enhanced chemiluminescence according to the manufacturer's instructions (NEN Life Science Products, Boston, MA).
Electrophoretic Gel Mobility Shift Analysis
3T3-L1 adipocytes or preadipocytes were incubated with the indicated agents for the indicated times and then washed with PBS. Nuclear proteins were extracted as follows. The cells were scraped into 10 mM HEPES, 0.5 mM MgCl2, 10 mM KCl, and 1 mM phenylmethylsulfonyl fluoride and vortexed for 10 s, and Nonidet P-40 was added to 0.1%. After vigorous vortexing, cells were pelleted in a microfuge at 1,000 rpm for 5 min, and cytosolic supernatant was removed. The extracted nuclei were washed and pelleted twice with the same hypotonic lysis buffer without detergent. The nuclear pellet was resuspended in 20 mM HEPES, 0.5 mM MgCl2, 400 mM NaCl, 0.1 mM EDTA, 20% glycerol, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride. After centrifugation at 14,000 rpm for 10 min, the nuclear supernatant extract for each sample was quantitated for total protein by bicinchoninic acid assay (BCA; Pierce, Dallas, TX). The binding reaction was a 20-min incubation of 5 μg of nuclear protein with a 32P end-labeled, double-stranded oligonucleotide containing the NF-κB binding site on the vascular cell adhesion molecule (VCAM)-1 promoter (5′-CCTTGAAGGGATTTCCCTCC-3′) (37). Cold competition controls were performed by preincubating the nuclear proteins with unlabeled 50-fold molar excess of the NF-κB double-stranded oligonucleotide for 10 min before the addition of the 32P-labeled oligonucleotide. For SAA3 enhancer factor (SEF) binding studies, the oligonucleotide 5′-CACATTTCTGGAAATGCCTAGAT-3′ was used and the corresponding mutant sequence 5′-CACATTTATCAAAATGCCTATAT-3′ that lacks specific binding to SEF (43). Italicized nucleotides are the point mutations that differentiate the mutant from the wild-type sequence. The mixtures were resolved on native 5% polyacrylamide gels made and run with 0.5 × Tris-borate-EDTA buffer, which were dried and autoradiographed.
Quantitative Western and Gel Shift Data
Films from the Western blots of five independently differentiated sets of adipocytes were scanned, and the background-corrected signal from each band was quantitated by densitometry using an Alpha Innotech Multiimage Light Cabinet with Chemimager 4400 software. Signal for each sample lane (containing samples from day 0 to day 8) was normalized to levels found in the day 8 lane of that same blot. Similarly, to quantitate EMSA data, gel films were scanned and gel shift bands were quantitated by densitometry. Gel shift signal for every sample in each experiment was normalized to the amounts in unstimulated cells in that same experiment. The normalized relative levels for each experimental group are represented as means ± SE. For Fig. 1, these normalized averages were further converted to percent (expression levels in day 8 adipocytes = 100%), and results were graphed as percent change from preadipocyte to adipocyte.
Luciferase assays were performed essentially as described previously (14). 3T3-L1 preadipocytes were transfected with an NF-κB-responsive luciferase reporter construct (a gift from Dr. Richard Pestell, described in Ref. 19) containing three tandem consensus binding sites from the major histocompatibility complex (MHC) II promoter. Control luciferase constructs with promoters shown to be activated by PPARγ, c-Fos, c-Myc, cyclin D, cyclin A, and junB were also graciously provided by the laboratory of Dr. Richard Pestell. The cells were cotransfected with a construct containing a hygromycin-resistance gene, and clones were selected by hygromycin. Resistant clones were picked and screened for both baseline luciferase activity and NF-κB responsiveness to TNF-α induction. The clones were checked for competence to hormonal differentiation and lipid accumulation both by appearance under the microscope as well as oil red O staining for lipids (as described in Ref. 14). After experimental treatments, cells were subjected to lysis in 200 μl of extraction buffer, 50 μl of which were used to measure luciferase activity, as described previously (40).
Enzyme-Linked Immunosorbant Assays for IL-6 and TNF-α
Ten-centimeter plates of confluent preadipocytes, adipocytes, and J774 macrophage cells were given the indicated treatments, the cells were washed with PBS, and 3 ml of media were added for collection of secreted cytokines for 6 h. Fifty microliters of the resulting cell media supernatant were assayed for accumulation of IL-6 and TNF-α using ELISAs purchased from R&D Systems (Minneapolis, MN).
Heterotypic Stimulation Experiments
Confluent preadipocytes and adipocytes were incubated in FCS medium for 24 h, and this conditioned medium was removed and assayed for cytokines. The conditioned medium was then diluted 1:5 in fresh FCS medium and used to stimulate J774 cells for 12 h. This medium was then assayed for its accumulated cytokine content as described in results. For immunodepletion experiments, the conditioned medium was incubated for 6 h at 4°C on protein G-Sepharose resin with anti-IL-6 antibody bound. Control mouse monoclonal antibodies generated by our laboratory were used in control immunodepletion.
Separation of proteins by SDS-PAGE, fluorography, and immunoblotting was performed as described previously (46). All statistical analyses were performed with numerical data represented throughout the figures as means ± SE. Student's t-test (P values ≤ 0.05) was used as a cutoff for statistical significance.
NF-κB Subunits are Dynamically Expressed in Differentiating Adipocytes
To study changes in NF-κB signal pathways associated with differentiation, 3T3-L1 cells were used to investigate expression of NF-κB subunits throughout the process of differentiation from fibroblastic preadipocyte to lipid-laden fat cell. Cells are differentiated after they reach confluence over a period of 8 days, after which triglyceride droplets are visible by microscopy. Protein extracts were made from cells at various stages of differentiation, and Western blots were processed for expression of NF-κB subunits as well as for GDI as an equal total protein loading control. Over the 8-day differentiation process, the most striking results are a significant induction of the NF-κB p65 (RelA) and p68 (RelB) subunits, as well as the regulatory subunit IκBα (Fig. 1A). This could be explained by the autoregulation of their promoters by increased NF-κB activity (6, 47). Even more strikingly, the p52 subunit increases from undetectable levels in the preadipocyte to a continuously and dramatically induced expression in the day 8 adipocyte. Interestingly, levels of the p52 precursor p100 remain relatively unchanged, implying that it is not necessarily expression of p52 that is changed but rather the proteolytic processing of the precursor to its smaller transcriptionally active form. p50 and its precursor p105 (NF-κB1) as well as Bcl-3 levels remain relatively unchanged. A number of the subunits have differential regulation specifically on day 2, with p65 and p68 displaying a downregulation and p105, p50, and Bcl-3 displaying an increase. This is, however, most likely due to the effects of the dexamethasone in the differentiation cocktail (1, 10, 29). c-Rel was not detectable on these Western blots at any stage of differentiation (data not shown). Replicates of this experiment with independently differentiated cells and subsequent assays measuring the expression of the DNA binding subunits of NF-κB (p65, p68, p50, p52) were performed and quantitated by densitometry, with amounts in each Western blot normalized to the signal present in the day 8 adipocyte samples (Fig. 1B). These results are presented graphically as percent change in expression after differentiation.
Basal NF-κB Activity Increases During Adipocyte Differentiation
Basal NF-κB transcriptional activation increases through differentiation.
3T3-L1 preadipocytes were stably transfected with an NF-κB-responsive luciferase reporter construct containing three tandem binding sites from the MHC class II promoter to measure NF-κB transcriptional activity and induction (19). Clones were selected based on TNF-α-responsive luciferase activity. These clones could be differentiated into lipid-laden adipocytes efficiently and were used to look at constitutive NF-κB activity during the process of differentiation. Cells were harvested on days 0, 2, 4, and 8, and cell extracts were normalized for total protein. Extracts were then assayed for their unstimulated constitutive NF-κB-mediated luciferase activity. Figure 2A shows that, during the 8-day differentiation protocol, constitutive NF-κB activity increased significantly and remained significantly higher in the mature adipocyte compared with preadipocytes. To show that luciferase induction was specifically dependent on the NF-κB promoter and not a reflection of a general increase in transcriptional activity during differentiation, a number of other luciferase constructs driven by various other promoters were transfected into 3T3-L1 cells and similarly assayed in day 0 and day 8 3T3-L1 cells (Fig. 2B). The luciferase activity driven by these promoter constructs generally decreased or remained constant, whereas a construct that served as a positive control (a PPARγ-responsive construct) doubled its activity during differentiation. Significantly, the NF-κB reporter induced luciferase activity even more significantly than PPARγ, underlining that this increase is a specific differentiation-induced effect. In addition to fold activity, absolute luciferase activities (counts/min; cpm) are indicated within the data bars for preadipocytes.
Endotoxin-Stimulatable NF-κB Activity is Suppressed During Differentiation
NF-κB-responsive luciferase activity is stimulated by LPS in preadipocytes but suppressed in the adipocyte.
After observing the surprising induction of NF-κB luciferase activity through differentiation in the basal state, we tested what the effects on stimulatable NF-κB activity were during differentiation. 3T3-L1 cells carrying NF-κB luciferase reporter constructs were grown in six-well plates and analyzed at either differentiated or the preadipocyte level. Preadipocytes and adipocytes (days 0 and 8) were stimulated with either 1 μg/ml LPS or 10 ng/ml TNF-α for 8 h, and equal amounts of total cellular protein were assayed for luciferase activity. As shown in Fig. 3A, preadipocytes displayed a solid induction of NF-κB upon exposure to either LPS or TNF-α compared with unstimulated cells, whereas adipocytes were sensitive only to TNF-α. Likewise, during a time-course experiment of LPS stimulation (1 μg/ml), preadipocytes produced a significant induction of NF-κB transcriptional activation upon LPS treatment, whereas adipocytes had constitutively higher levels of NF-κB and were insensitive to further induction by LPS (Fig. 3B). Dose-response experiments further corroborate these results, showing that although preadipocytes are exquisitely sensitive to even the lowest doses of LPS (20 ng/ml), adipocytes are resistant to even the highest doses (Fig. 3C).
NF-κB nuclear translocation and DNA binding activity corroborate luciferase reporter experiments, showing stimulation by LPS and TNF-α and modulation during adipogenesis.
To corroborate luciferase activity assays and get further insight into the changes in NF-κB activity through differentiation, EMSAs were used to quantitate nuclear NF-κB DNA binding activity in 3T3-L1 preadipocytes and adipocytes with or without stimulation. Nuclear extracts were exposed to double-stranded oligonucleotide probes containing specific binding sequences for both p50/p50 and p50/p65 complexes (37), although p52 and p68 may also be part of the complexes. 3T3-L1 cells in various stages of differentiation were stimulated with 1 μg/ml LPS or 10 ng/ml TNF-α for 30 min, and nuclear extracts were isolated and probed for NF-κB DNA binding with the use of a double-stranded NF-κB consensus oligonucleotide. EMSAs reveal that TNF-α caused dramatic induction of NF-κB nuclear translocation in both preadipocytes and adipocytes, whereas in our first set of experiments, LPS-induced translocation could be observed in preadipocytes but was significantly decreased in adipocytes (Fig. 4A). A time course of LPS stimulation corroborated this result, showing a characteristic induction in preadipocytes followed by a reduction in DNA binding activity after 1 h. In differentiated adipocytes, LPS stimulation caused only a modest translocation at all time points (Fig. 4B). A similar experiment with slightly different stimulation times was performed with the use of TNF-α (10 μg/ml). In contrast to LPS, TNF-α stimulated both preadipocytes and adipocytes rapidly and with approximately equivalent amounts of translocation (Fig. 4C). Control gel shift experiments using excess unlabeled oligonucleotides show that the observed LPS-stimulated gel shift is effectively competed (Fig. 4D, left), and supershifts with anti-p65 antibodies cause the disappearance of the band, demonstrating the presence of p65 in this complex (Fig. 4D, right). Quantitation of the observed changes in LPS- and TNF-α-mediated NF-κB DNA binding through differentiation was obtained from three independent experiments using separate batches of 3T3-L1 cells, quantified and normalized in each experiment to the signal present at time 0. 3T3-L1 adipocytes did respond to LPS stimulation, but their response was significantly decreased and delayed compared with the preadipocytes (Fig. 4E). In agreement with the previous luciferase stimulation experiments and with the representative blot in Fig. 4B, adipocytes are equally responsive to TNF-α compared with preadipocytes (Fig. 4F). Because our attempts to determine the specific subunits responsible for gel shift induction were unsuccessful using commercial supershift antibodies, we sought an alternative approach to corroborate our gel shift experiments. Nuclear extracts were prepared from our LPS- and TNF-α-stimulated preadipocytes and adipocytes, and equal amounts of nuclear protein were resolved by SDS-PAGE and probed by Western blot analysis for the presence of NF-κB subunits. Similar to electromobility shift assays, changes in the relative levels of NF-κB subunits in the nuclear fraction reflect nuclear translocation. The Western blots on nuclear extracts of stimulated preadipocytes corroborate the gel shift data indicating nuclear translocation of p65, p68, and p50, with no detectable signal for p52 (Fig. 4G, “preadipocytes”). In contrast, LPS stimulation of adipocytes induces an approximately twofold increase in nuclear p65, while p68 and p52 are constitutively present (Fig. 4G, “adipocytes”). To gain further functional insights into this phenomenon, we used whole cell lysates or isolated nuclear extracts from time-course experiments of LPS and TNF-α stimulation. Consistent with the gel shift and luciferase data, cytosolic IκBα in preadipocytes is rapidly and completely degraded in response to either LPS or TNF-α and reappears 60 min after initiation of the stimulus. In contrast, IκBα in differentiated adipocytes is degraded only in response to TNF-α stimulation and not in response to LPS (Fig. 4H, top). Probing the same extracts for p65, we find that the pattern of IκBα degradation in preadipocytes and adipocytes is mirrored by nuclear p65 content: preadipocytes show rapid and striking increases in p65 nuclear content after either LPS or TNF-α stimulation, whereas LPS stimulation of adipocytes produces delayed and suppressed appearance of nuclear p65 (Fig. 4H, bottom). Together, these results corroborate our prior gel shift and luciferase results, indicating that the observed decrease in LPS-stimulatable NF-κB activation after differentiation is mediated in part by a decrease in IκBα degradation.
Despite Decreased LPS Activation of Classical NF-κB Activity Assays, Adipocytes Respond to LPS by Activating Other Inflammatory Paths and Increase IL-6 Secretion
Although our results show altered NF-κB sensitivity to LPS-induced signaling in the adipocyte, we had previously demonstrated a high level of sensitivity to LPS as judged by the transcriptional induction of TLR2 (31) as well as additional well-established NF-κB targets, such as IL-6 and SAA3 (32). To explain this apparent paradox, we sought possible additional adipocyte-specific intracellular inflammatory signals that could mediate the effects of LPS and/or TNF-α by modulating NF-κB activity in ways other than the classical nuclear IκB degradation/translocation paradigm. LPS and inflammatory cytokines induce the NF-κB target SAA3 only in adipocytes, not in preadipocytes. The ability of the mature adipocyte to induce SAA3 suggests the presence of a cofactor in addition to NF-κB that is present in the adipocyte but absent in preadipocytes or NIH-3T3 cells. The SAA3 promoter has a region called the distal regulatory element (DRE) that functions as a cytokine-inducible transcription enhancer (2, 3, 43). Within this DRE, there is an oligonucleotide that constitutes the binding site for a SEF and an overlapping NF-κB site. SEF is a transcription factor critically involved in SAA3 expression (43) (although the activity or presence of SEF has never been shown in adipocytes). Bing et al. (2) have demonstrated a functional synergy between SEF and NF-κB to achieve SAA expression and shown that this synergy may, in part, be due to the ability of SEF to recruit NF-κB through physical interactions that lead to enhancement or stabilization of NF-κB binding to the SAA3 promoter element. As in initial study, gel shift assays were performed on preadipocyte and adipocyte extracts using the SEF binding oligonucleotide, demonstrating that unstimulated and LPS-stimulated adipocytes had a strikingly increased amount of SEF binding activity compared with preadipocytes (Fig. 5A). LPS stimulation caused a modest increase in binding activity, so a time-course experiment was performed using both LPS and TNF-α as stimulants. To demonstrate specific binding, either wild-type oligos or oligos with a base pair mutation were coincubated with the SEF-binding radiolabeled probes. Notably, in this experiment, similar to the time course for p65/RelA translocation shown in Fig. 4, E and F, TNF-α treatment of mature adipocytes leads to a more rapid activation of SEF binding to DNA than treatment with LPS (Fig. 5B). Interestingly, although LPS-stimulated activity is delayed, it is significant and prolonged, whereas TNF-stimulated activity disappears after 5 h.
Adipocytes increase secretion of IL-6 in response to LPS stimulation.
The constitutive activation of the NF-κB pathway in the mature adipocytes and the loss of classical NF-κB activation by LPS through differentiation indicate a switch to an alternative mechanism of inflammatory signaling. Our observations of SEF translocation, as well as previously reported effects of LPS on SAA3 and TLR2 expression in the adipocytes, indicate that this switch does not reflect a decrease in inflammatory responsiveness, however. To demonstrate LPS responsiveness, we measured IL-6 in adipocyte-conditioned medium upon LPS treatment. 3T3-L1 preadipocytes and adipocytes were treated with or without LPS (1 μg/ml) for 24 h. After this time, aliquots of cell culture supernatant were assayed for the presence of IL-6. As seen in Fig. 6A, inset, both preadipocytes and adipocytes secreted measurable amounts of IL-6, and both increased their production dramatically in response to LPS. Although the amounts of IL-6 secreted by preadipocytes and adipocytes with or without LPS stimulation are easily measurable, they are significantly less than that secreted by an equal number of immortalized J774 macrophage cells in culture (Fig. 6A). Interestingly, when amounts of TNF-α are measured from these cultured cells, the amounts secreted by preadipocytes and adipocytes with or without LPS stimulation are at the lower limits of detection with this ELISA system (Fig. 6B, inset), although they are easily detected and LPS responsive in the J774 macrophages (Fig. 6B).
Adipocytes But Not Preadipocytes Secrete Factors That Activate J774 Macrophages
In light of increased constitutive NF-κB activity and acute-phase reactant secretion after adipocyte differentiation, we wanted to know whether this could be physiologically significant to inflammation in general. To simulate the paracrine interaction between adipocytes and adipose tissue-resident macrophages, cell culture medium was first conditioned by preadipocytes or differentiated adipocytes for 12 h. Subsequently, these conditioned media were diluted and used to treat J774 macrophages (1 ml of conditioned medium in 4 ml of fresh medium), and IL-6 as well as TNF-α secreted from the macrophages were assayed. When J774 cells were exposed to unstimulated preadipocyte medium, there was no increase in IL-6 or TNF-α production compared with untreated cells (Fig. 6C). In contrast, adipocyte-conditioned medium caused a significant increase in the secretion of both cytokines from macrophages. A comparison of the amounts of cytokines secreted from adipocytes with that from macrophages (Fig. 6, A and B) shows that the transfer of IL-6 and TNF-α secreted by preadipocytes and adipocytes does not contribute significantly to the levels of these cytokines found in the basal, unstimulated state of macrophages. This indicates that the macrophage-derived medium after exposure to conditioned medium containing IL-6 and TNF-α was almost completely derived from macrophages. Furthermore, this suggests that the factors stimulating the macrophages were adipocyte-specific inflammatory signals other than IL-6 and TNF-α. No detectable TNF-α could be found in the preadipocyte- or adipocyte-conditioned media. To further support the point that secreted IL-6 was not playing a role in J774 stimulation, the conditioned media from both preadipocytes and adipocytes were immunodepleted for IL-6 and used to stimulate J774 cells in a similar fashion. Immunodepletion with a resin coated in anti-IL-6 antibody compared with a control resin containing a nonrelevant antibody effectively depleted 90% of the IL-6 secreted from preadipocytes and adipocytes (Fig. 7A), yet there was no loss of J774-stimulating activity by these supernatants after removal of their IL-6 (Fig. 7B).
The data above represent the first comprehensive description of NF-κB levels and activity during adipogenesis. The adipocyte not only demonstrates responsiveness to inflammatory stimuli, but shows changes in NF-κB subunit expression and activity during differentiation that are unique for a cell type previously thought to act primarily as a metabolic cell. An overall increase in constitutive NF-κB signaling during differentiation is apparent, but also a concomitant repression of LPS-stimulatable classical NF-κB activity. TNF-α signaling, by comparison, remains relatively unchanged.
No characterization of adipocyte NF-κB expression has been reported thus far. Yet, the significant potential of the adipocyte as an inflammatory cell of systemic importance puts NF-κB regulation in adipocytes at center stage for the pathophysiology of diabetes. The patterns of expression through adipocyte differentiation, specifically of p65, p68, p52, and IκBα, suggest an increase in constitutive NF-κB activity. Luciferase reporter analysis of unstimulated preadipocytes and adipocytes corroborates these results, showing increased promoter activation by (at least) the p65/p50 NF-κB complex. Measurement of adipocyte-secreted IL-6 demonstrates that as a consequence of the increase in constitutive NF-κB activity through differentiation, the adipocyte constitutively produces inflammatory signals. The increased cytokine secretion by macrophages stimulated with adipocyte-conditioned media further amplifies this inflammatory activity. Thus we can infer that the adipocyte will contribute to systemic inflammation by secreting its own inflammatory cytokines, as well as inducing secretion of inflammatory signals via nearby resident tissue macrophages. This combined NF-κB-dependent inflammatory activity by adipose cellular constituents may be involved in the systemic inflammation and resulting metabolic dysregulation observed in obesity.
One possible limitation of the studies presented is that we use a tissue culture model to study the differentiation process. The 3T3-L1 cell line has been widely used for almost three decades (17, 18) and has proven to be a valuable tool for the elucidation of many processes relevant to primary adipocytes in vivo. We employed the classical differentiation cocktail containing insulin, dexamethasone, and 3-isobutyl-1-methylxanthine to induce adipogenesis in these cells in 25 mM glucose, conditions widely used in the literature. Compared with in vivo conditions this is comparable to hyperglycemic conditions. The adipocytes obtained by this differentiation protocol are therefore the product of these in vitro conditions, but offer nevertheless valuable insights into adipogenesis-related functional changes at the level of signal transduction and transcription.
In addition to the increase in constitutive NF-κB activity during adipocyte differentiation, there is a concomitant repression of LPS-stimulatable NF-κB activity. Gel shift and luciferase activity assays show that adipocytes are nearly completely resistant to LPS stimulation via the classical NF-κB pathway. This resistance correlates with a decrease in LPS-mediated IκB degradation. Western analyses on nuclear NF-κB show that the adipocyte has constitutively increased nuclear p65/RelA, p68/RelB, p52, and p50, and that LPS stimulation causes little additional translocation.
Although LPS-induced NF-κB activation is repressed after differentiation, adipocytes can induce NF-κB activity as judged by the ability of TNF-α to induce DNA binding and luciferase activity. Also, adipocytes are exquisitely sensitive to other modes of LPS-induced inflammatory signaling, as demonstrated by induction of IL-6, TLR2, and SAA3 (31, 32). Thus the significance of the changes in NF-κB signaling go beyond the previously observed phenomenon of endotoxin tolerance (53). The downstream signaling cascade initiated by the LPS receptor TLR4 may therefore differ and be unique in adipocytes.
We used an in vitro assay designed to mimic in vivo paracrine signaling between adipocytes and interstitial macrophages to demonstrate the constitutive inflammatory potential of the adipocyte. The results demonstrate that 3T3-L1 adipocytes, in addition to secreting significant amounts of IL-6, secrete additional factor(s) that have a significant impact on macrophages as judged by a massive induction of IL-6 and TNF-α production at the level of the target cell. This is the first time that any direct evidence for a heterotypic signaling event between the adipocyte and a macrophage has been presented. The in vitro approach of using adipocyte-conditioned medium and measuring its effects on macrophages has the advantage that it represents a very clean and well-defined system; however, it has the drawback that we cannot account for effects mediated in vivo through direct cell-to-cell contact. We have recently started to characterize the effects of adipocyte-secreted products on surrounding breast cancer cells during tumor progression (23) and found that many adipocyte-derived factors have a profound impact on tumor cells, promoting tumor growth by stimulating breast cancer cell growth through activation of NF-κB and cyclin D1. These observations are equally borne out in both cell culture systems as well as in vivo. There is therefore ample evidence that the adipocyte can exert a very significant role on the transcriptional programs of neighboring cells. In vivo, adipocytes in different fat pads may significantly differ with respect to their inflammatory potential (11). The cross talk between adipocytes and macrophages may be particularly relevant in the context of obesity when the local inflammatory level is upregulated. The purification of the unknown soluble factor released by adipocytes that exerts this dramatic proinflammatory activity on macrophages is currently the focus of our efforts. Systemic elevation of inflammatory markers, in particular IL-6, is strongly associated with increased risks for cardiovascular problems (reviewed recently in Ref. 4). Adipose tissue has been shown to be a major contributor to systemic IL-6 levels (35). Its contribution may be particularly important in the subclinical inflammatory state associated with syndrome X. The elevated production of IL-6 under these circumstances may not only be the result of an upregulation of IL-6 in adipocytes, but the local macrophages may contribute significantly to this phenomenon.
This work was supported by an American Diabetes Association Medical Scientist Training Grant (to A. H. Berg), National Institutes of Health (NIH) Grant R01-HL-073163-01 (to Y. Lin), the Core Laboratories of the Albert Einstein Diabetes Research and Training Center, NIH Grant R01-DK-55758 (to P. E. Scherer), and a research grant from the Juvenile Diabetes Association (no. 2001-780; to P. E. Scherer).
We thank the members of the Scherer laboratory for helpful comments.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by American Physiological Society