Pflügers Archiv - European Journal of Physiology

, Volume 458, Issue 6, pp 1103–1114

Microarray analysis identifies matrix metalloproteinases (MMPs) as key genes whose expression is up-regulated in human adipocytes by macrophage-conditioned medium


  • Adrian O’Hara
    • Obesity Biology Research Unit, School of Clinical SciencesUniversity of Liverpool
  • Fei-Ling Lim
    • Unilever R&D Discover
  • Dawn J. Mazzatti
    • Unilever R&D Discover
    • Obesity Biology Research Unit, School of Clinical SciencesUniversity of Liverpool
Molecular and Genomic Physiology

DOI: 10.1007/s00424-009-0693-8

Cite this article as:
O’Hara, A., Lim, F., Mazzatti, D.J. et al. Pflugers Arch - Eur J Physiol (2009) 458: 1103. doi:10.1007/s00424-009-0693-8


White adipose tissue exhibits inflammation as tissue mass expands in obesity, involving macrophage infiltration and a direct inflammatory response by adipocytes. DNA microarrays and conditioned medium have been used to examine the effects of macrophages on global gene expression in human adipocytes. SGBS adipocytes, differentiated in culture, were treated with macrophage-conditioned medium (U937 cells) for 4 or 24 h; control cells received unconditioned medium. Agilent arrays comprising 44,000 probes were used to analyse gene expression. Microarray analysis identified 1,088 genes differentially expressed in response to the conditioned medium at both 4 and 24 h (754 up-regulated, 334 down-regulated at 24 h); these included genes associated with inflammation and macrophage infiltration. A cluster of matrix metalloproteinase genes were highly up-regulated at both time-points, including MMP1, MMP3, MMP9, MMP10, MMP12 and MMP19. At 4 and 24 h, MMP1 was the most highly up-regulated gene (>2,400-fold increase in mRNA at 24 h). ELISA measurements indicated that substantial quantities of MMP1 and MMP3 were released from adipocytes incubated with conditioned medium, with little release by control adipocytes. Treatment with TNFα induced substantial increases in MMP1 (>100-fold) and MMP3 (27-fold) mRNA level and MMP1 and MMP3 release in adipocytes, suggesting that this cytokine could contribute to the stimulation of MMP expression by macrophages. In conclusion, macrophage-secreted factors induce a major inflammatory response in human adipocytes, with expression of MMP family members being strongly up-regulated. The induction of MMP1 and other MMPs suggests that macrophages stimulate tissue remodelling during adipose tissue expansion in obesity.


Adipose tissueInflammationMacrophagesObesityTNFα





macrophage conditioned


matrix metalloproteinase


tumour necrosis factor-α




There has been a substantial increase in the incidence of obesity and its associated diseases in most western countries over the past two decades. In the UK, for example, approximately 25% of adults are now obese with a BMI ≥ 30 [31]. Obesity is characterised by the expansion of white adipose tissue (WAT) mass, but until recently the functional importance of the tissue was unrecognised. However, over the past 15 years, WAT has been shown to be a major endocrine organ which is involved in the secretion of a number of factors, including lipid moieties such as free fatty acids and steroid hormones, and protein signals termed adipokines [1, 16, 37]. Production and release of these secreted factors can be greatly altered in obesity and this may underlie the development of associated diseases such as the metabolic syndrome and type 2 diabetes [40].

In the early 1990s, adipocytes were shown to synthesise and release the pro-inflammatory cytokine TNFα [20]. The concept of WAT as an endocrine organ gained universal acceptance in the mid 1990s on the discovery of the hormone leptin [52] with its wide spectrum of biological functions. A diverse range of other adipokines has subsequently been identified, a number of which have inflammatory properties or are linked to inflammation [19, 29, 36, 39]. These inflammatory-related proteins include cytokines, chemokines and acute phase proteins such as TNFα, IL-1β, IL-6, IL-8, IL-10, monocyte chemoattractant protein-1 (MCP-1), macrophage migration inhibitory factor, haptoglobin, serum amyloid-A and plasminogen activator inhibitor 1 [30, 40, 41].

One of the major developments in obesity research over the past few years has been the discovery that the condition is characterised by chronic mild inflammation [19, 33]. This is based on the observation that the circulating levels of inflammatory markers, such as C-reactive protein, IL-6, IL-8 and haptoglobin, are increased in the obese [4, 14, 15, 51]. Several studies have shown that in obesity, macrophages infiltrate adipose tissue with this being an important contributor to the inflammatory state within the expanding fat mass [5, 8, 11, 48, 49]. Macrophage-secreted factors, through increases in the key inflammatory transcription factor NF-κB, are suggested to stimulate adipocyte inflammation and induce insulin resistance [29]. Macrophage factors have also been shown to block insulin action in adipocytes via down-regulation of GLUT4 and IRS-1 [25], while co-culture of macrophages with 3T3-L1 adipocytes has been reported to induce leptin expression and dephosphorylation of insulin receptor substrate 1, leading to impaired glucose uptake [18]. Studies have also shown that in the presence of macrophages, the inflammatory state can impair adipogenesis, and also lead to increased production of IL-6, IL-1β, MCP-1 and other signals [9, 24]. A microarray study with 3T3-L1 adipocytes has suggested that co-culture with RAW264.7, a murine macrophage cell line, in the presence of LPS, leads to the up-regulation of genes linked to inflammation and angiogenesis, such as IL-6, MCP-1 and RANTES [50].

While cross-talk between adipocytes, macrophages and other cell types appears central to the inflammatory response during adipose tissue expansion in obesity, the extent of the interaction is unclear. In the present study, we have used whole-genome microarrays to identify the global effects of macrophage-secreted factors on gene expression in human adipocytes. Microarray analysis demonstrated extensive up-regulation of a range of inflammation-related genes in human fat cells by macrophage-conditioned medium; in particular, members of the matrix metalloproteinase (MMP) gene family were found to be powerfully up-regulated, especially MMP1, MMP3 and MMP10.

Materials and methods

Cell culture

Simpson-Golabi-Behmel syndrome (SGBS) human preadipocytes were maintained and differentiated into adipocytes as described previously [43]. Briefly, cells were cultured in 12-well plates (20,000 cells/well) and maintained in DMEM/Ham’s F12 (1:1 v/v) medium (Invitrogen, Paisley, UK) containing 10% (v/v) foetal calf serum (FCS; Sigma, Poole, UK), 33 µM biotin, 17 µM pantothenate, 100 U/ml penicillin and 100 μg/ml streptomycin at 37°C in a humidified atmosphere of 5% CO2/95% air. Once confluent, differentiation was induced by incubating the cells in FCS-free induction media containing 0.25 μM dexamethasone, 0.5 mM isobutyl IBMX, 10 nM insulin, 200 pM triiodothyronine (T3), 0.1 μM cortisol (Sigma) and 2 μM rosiglitazone (GlaxoSmithKline, UK) for 4 days. The cells were then fed every 3 days in feeding medium (induction medium without dexamethasone, IBMX and rosiglitazone).

U937 monocytes [36], kindly provided through Dr. Stuart Wood, were plated out at 2 × 105 cells/ml in RPMI–1640 medium supplemented with 10% FCS, 100 U/ml penicillin and 100 μg/ml streptomycin, at 37°C in a humidified atmosphere of 5% CO2/95% air and grown until 25–30% confluent. Once at the required confluence, differentiation was induced by incubating the cells in a medium containing 30 nM phorbol myristic acid (PMA). The medium was changed every 2–3 days, and collected on day 7 to be used as macrophage-conditioned medium (MC medium).

Differentiated SGBS cells, 10 days post-induction, were treated with 150 µl/ml MC medium or 150 µl/ml unconditioned medium (UC medium) for 4 and 24 h. UC medium is the medium in which macrophages are normally cultured (see above) but which has not been exposed to the immune cells. Control cells were incubated in feeding medium. In experiments on the effects of TNFα, SGBS cells in feeding medium were incubated with human recombinant TNFα (Sigma) for 24 h at doses of 5 and 100 ng/ml. The cells were harvested in 500 µl of Trizol (Invitrogen) and the medium collected; both were frozen at −80°C until required for analysis.

Microarray analysis

Total RNA was isolated using Trizol, followed by the RNeasy kit (Qiagen, Crawley, UK). The integrity of the RNA was confirmed by analysis with the Agilent 2100 bioanalyser (Palo Alto, CA, USA) using the RNA 600 LabChipTM kit. Two hundred nanograms of RNA and 2.0 µl (1:5,000 dilution) Agilent One-Colour RNA Spike-In RNA were then labelled with the Agilent Low RNA Input Linear Amplification Kit PLUS, One-Colour (Agilent Technologies UK, Wokingham, UK) according to the manufacturer’s instructions, as previously described [27]. Briefly, cDNA was synthesised for 2 h at 40°C followed by denaturation for 10 min at 65°C. The synthesis of the fluorescent-labelled cRNA was performed during the second incubation step (2 h at 40°C). The labelled cRNA was purified with the RNeasy Mini Kit (Qiagen, UK) according to the manufacturer’s protocol.

The Agilent Hybridisation Kit (catalogue number 5188-5242) was used in conjunction with Agilent Human Oligo Arrays (catalogue number G4112F, Agilent Technologies UK, Wokingham, UK). Hybridisation was performed as previously described [27], according to the manufacturer’s protocol (using 2 μg of the labelled sample cRNA as input). The slides were scanned with the Agilent G2565BA Microarray Scanner System. Six sets of cells from each group were used individually and meaned in the microarray analysis.

Data extraction and deposition into GEO

The Agilent G2567AA Feature Extraction Software v.9.1 (Agilent Technologies UK) was used for data extraction and quality control. In compliance with MIAME standards, data files were deposited into the NCBI Gene Expression Omnibus (GEO). The following link was created to allow review of these data:

The GEO accession number is GSE14312.

Bioinformatic analysis

Extracted data were analysed using GeneSpring GX 7.3.1 (Silicon Genetics, CA, USA). Agilent standard scenario normalisations for FE1-colour arrays were applied to all data sets. A subset of genes for data interrogation was generated that excluded spots of poor quality, and gene probes that were present in at least four out of six of the samples in each group. Relative expression of each probe in adipocytes cultured in control, UC and MC medium was determined; probes differentially expressed by greater than 2.0-fold were selected. One-way, parametric, ANOVA tests were performed followed by Benjamini and Hochberg multiple test correction with a false discovery rate of 0.01. Microsoft Excel templates were prepared containing genes that were over- and under-expressed following treatment with UC medium and MC medium and Ingenuity™ Pathway Analysis 3.0 (Ingenuity™ Systems, CA, USA) was utilised to assemble functional networks associated with the differentially expressed probes.

Real-time PCR

For reverse transcription polymerase chain reaction (RT-PCR) and real-time PCR studies, total RNA was isolated using Trizol and the quantity of RNA was determined from the absorbance at 260 nm. RNA (0.5 µg) was reverse transcribed using a Bio-Rad iScript cDNA synthesis kit in a total volume of 10 µl on a PCR Express thermal cycler (Hybaid, UK).

Real-time quantitative PCR (qPCR) was carried out at a final volume of 12.5 µl containing 12.5–50 ng cDNA using the Taqman Gene Expression Master Mix (Applied Biosciences, Applera). Taqman (Applied Biosciences) probes were used for qPCR; the probes used to validate the microarrays are shown in Table 1. Detection was carried out using the Bio-Rad I-Cycler with PCR conditions of 50°C for 2 min at 90°C, followed by 45 cycles of 95°C for 15 s and 60°C for 60 s. For analysis of MMP1 and MMP3 in TNFα-treated samples [44], Taqman probes for MMP1 (Table 1) and MMP3 mRNA (Applied Biosystems: Hs00968308_m1) were used, and detection was carried out on a Stratagene Mx3005P detector.
Table 1

Taqman probes used in the validation of the microarrays


Applied Biosciences assay ID

Actin, beta


CCAAT/enhancer binding protein (C/EBP), alpha


Chemokine (C-C motif) ligand 3


Chemokine (C-X-C motif) ligand 3


Forkhead box O1


Homeobox D3


Insulin-like growth factor 1 receptor


Interleukin 1, beta


Matrix metallopeptidase 1 (interstitial collagenase)


Matrix metallopeptidase 12 (macrophage elastase)


Mitogen-activated protein kinase 13


Somatostatin receptor 1


Vascular endothelial growth factor C


Two different reference genes were used, β-actin and GAPDH, and similar expression of these was obtained across groups. β-actin was subsequently used as the main reference gene. Relative quantitation values were expressed using the \( 2^{{ - \Delta \Delta C_{\text{t}} }} \) method as fold changes in the target gene normalised to the reference gene and related to the expression of the controls.


Primers for MMP1, MMP3 and MMP10 were synthesised commercially (Eurogentec, UK): MMP1 5′-CTCTGAGGGTCAAGCAGACAT-3′ (forward), 5′-CAACTTCCGGGTAGAAGGGAT-3′ (reverse) product size 482 bp, MMP3 5′-ACCCACTCTATCACTCACTCACA-3′ (forward) 5′-GGTATCCAGCTCGTACCTCAT-3′ (reverse) product size 403 bp and MMP10 5′-CTCTTTCACTCAGCCAACACT-3′ (forward) 5′-GGATAACCTGCTTGTACCTCA-3′ (reverse) product size 443 bp. RT-PCR was performed using 50 ng of cDNA in a final volume of 12.5 µl containing specific primers and 1.1× Ready Mix™ PCR Master Mix (Thermo Scientific). PCR conditions were as follows: MMP1; 94°C for 4 min, followed by 35 cycles of 94°C for 20 s, 62.6°C for 30 s and 72°C for 40 s, followed by 72°C for 10 min and 30°C for 10 s. Conditions for MMP3, MMP10 and β-actin were the same as for MMP1, apart from the annealing temperatures, which were 51.3, 55.3 and 57°C, respectively. The PCR products were sequenced (Eurofins, Ebersberg, Germany) to confirm their identity.


MMP1 and MMP3 protein were measured in cell culture medium using commercial enzyme-linked immunosorbent assay (ELISA) kits (Calbiochem, La Jolla, California: MMP1 Cat No QIA55, sensitivity 23 pg/ml; MMP3 Cat No QIA73, sensitivity 100 pg/ml). Each sample was assayed in duplicate on a 96-well microplate according to the manufacturer’s instructions, and detection carried out with a Benchmark Plus microplate spectrophotometer (Bio-Rad, UK).

Statistical analysis

The results are expressed as mean values ± SE. Differences between groups were analysed using one-way ANOVA or Student’s t test (real-time PCR data).


Human SGBS adipocytes [43] were incubated for either 4 or 24 h with conditioned media from U937 monocytes that had been induced to differentiate into macrophages. Pilot studies using selected gene expression markers (IL-6 and MCP-1) established that the peak response in the adipocytes was obtained with medium from macrophages at 7 days post-induction and that the addition of 150 μl of MC medium per millilitre of adipocyte medium was optimal. Microarrays were performed on control adipocytes (cultured only with adipocyte medium), adipocytes incubated with UC medium and adipocytes incubated with MC medium, and the differences were assessed by one-way ANOVA. Comparison of control adipocytes with adipocytes incubated with UC medium showed that the macrophage medium itself had a minimal effect on gene expression in the fat cells (results not shown). Thus, changes in the MC medium-treated group are due to the conditioning rather than to the medium itself; detailed analysis was therefore undertaken by comparing the UC and MC medium groups.

From initial analysis, a total of over 5,000 mRNAs were shown to be differentially regulated between the UC and MC medium adipocytes at either the 4 or 24 h periods. Further analysis of the data showed that from this large group of transcripts, there were 1,088 that were common at both 4 and 24 h (Fig. 1).
Fig. 1

Venn diagram of the results obtained from the microarray analysis. There were 1,648 genes differentially regulated at 4 h only, 3,673 genes differentially regulated at 24 h only and 1,088 genes common to both time-points. The analysis was based on the parameters of a false discovery rate of 0.01, probes that were present in at least four out of six of the samples in each group, and probes differentially expressed by greater than 2.0-fold with P < 0.01

Validation of microarrays by real-time PCR

To validate the microarrays, a total of 12 genes were also examined by real-time PCR; these genes ranged from those that the microarrays indicated were strongly up-regulated in response to the MC medium through to those that were markedly down-regulated. The validation was done at both 4 and 24 h. Table 2 indicates that there was good correlation between the changes in mRNA levels indicated by real-time PCR and those obtained by the microarrays. In no case were changes in the opposite direction obtained with the two procedures, but somatostatin receptor 1 could not be assessed since the mRNA level was too low to obtain meaningful results by real-time PCR.
Table 2

Validation of micorarrays by real-time PCR

Gene name

Gene identifier

Fold change 4 h microarray

Fold change 4 h real-time PCR

Fold change 24 h microarray

Fold change 24 h real-time PCR

Mitogen-activated protein kinase 13 (MAPK13)






Insulin-like growth factor 1 receptor (IGF1R)






Matrix metallopeptidase 1 (interstitial collagenase) (MMP1)






Chemokine (C-X-C motif) ligand 3 (CXCL3)






Chemokine (C-C motif) ligand 3 (CCL3)






Matrix metallopeptidase 12 (macrophage elastase) (MMP12)






Interleukin 1, beta (IL1B)






Vascular endothelial growth factor C (VEFGC)






Forkhead box O1 (FOXO1)






CCAAT/enhancer binding protein (C/EBP), alpha (CEBPA)






Somatostatin receptor 1 (SSTR1)



Expression too low


Expression too low

Homeobox D3 (HOXD3)






A total of 12 genes were used to validate the microarrays by real-time PCR; mRNA levels were compared between adipocytes treated with MC medium compared with UC medium. In the case of somatostatin receptor 1, the expression was too low to obtain a meaningful value with real-time PCR. Results are mean values (n = 6 for microarrays; n = 3 for real-time)

*P < 0.05; **P < 0.01; ***P < 0.001 compared with the UC medium (ANOVA for microarrays; Student’s t test for real-time)

Microarray analysis

Of the 1,088 transcripts that were differentially regulated in response to MC medium at both time-points employed, 754 were up-regulated and 334 down-regulated at 24 h; the expression of 189 genes was increased >10-fold, 20 >100-fold and three >1,000-fold, while 33 genes showed a >10 reduction in mRNA level and two a >100-fold decrease. At 4 h, 737 genes were up-regulated and 351 were down-regulated in response to the MC medium. Over the two time-points of the microarray analysis, 91% of the genes had the same expression pattern (up- or down-regulated at both 4 and 24 h), and only modest changes were observed in the 9% of genes where the expression pattern differed.

A heat map generated from the 1,088 probes shows all the transcripts that were altered in response to the conditioned medium at 4 and 24 h (Fig. 2). The heat map indicates clusters of mRNAs that were common to both time-points, and also those differentially expressed between 4 and 24 h. The more chronic response of adipocytes to MC medium was the primary focus of the study, as best reflecting the impact of macrophage infiltration and recruitment into adipose tissue, and examples of the genes whose expression was changed at 24 h are shown in Table 3. Major increases in the mRNA level of genes encoding several metalloproteinases (particularly MMP1, MMP3 and MMP10) and chemokines (CXCL5, CCL7 and CCL6) are evident, together with substantial increases in interleukin (IL8, IL11 and IL6) and metallothionein (MT1X, MT1H) mRNAs (Table 3). On the other hand, there were reductions in the expression of genes encoding proteins such as aquaporin 1 (AQP1), G protein-coupled receptor 109B (GPR109B) and in C/EBP alpha (CEBPA) in particular.
Fig. 2

Heat map based on the 1,088 genes that were differentially expressed at both 4 and 24 h in human adipocytes incubated in the presence of either unconditioned medium (UC) or macrophage-conditioned medium (MC). Mean expression of each probe (n = 6 in each group) is shown ordered by the degree of differential regulation at 4 h from highest to lowest fold-change difference in expression following treatment with MC medium. Red and blue colouring indicate high expression and low expression, respectively, compared to the average expression of each probe across all samples. Yellow colouring indicates no change from average expression

Table 3

Examples of genes whose expression was up- or down-regulated in human adipocytes at 24 h by macrophage-conditioned medium


Genbank ID

Fold change at 24 h

Matrix metallopeptidase 1 (interstitial collagenase) (MMP1)



Matrix metallopeptidase 3 (stromelysin 1, progelatinase) (MMP3)



Matrix metallopeptidase 10 (stromelysin 2) (MMP10)



Colony stimulating factor 3 (granulocyte) (CSF3), transcript variant 1



Chemokine (C-X-C motif) ligand 5 (CXCL5)



Matrix metallopeptidase 9 (gelatinase B, 92 kDa gelatinase, 92 kDa type IV collagenase) (MMP9)



Chemokine (C-C motif) ligand 7 (CCL7)



Matrix metallopeptidase 12 (macrophage elastase) (MMP12)



Chemokine (C-X-C motif) ligand 6 (granulocyte chemotactic protein 2) (CXCL6)



Interleukin 8 (IL8)



Interleukin 11 (IL11)



Leukaemia inhibitory factor (cholinergic differentiation factor) (LIF)



Chemokine (C-X-C motif) ligand 2 (CXCL2)



Interleukin 6 (interferon, beta 2) (IL6)



Colony stimulating factor 2 (granulocyte-macrophage) (CSF2)



Metallothionein 1X (MT1X)



Metallothionein 1H (MT1H)



Chemokine (C-C motif) ligand 2 (CCL2)



Chemokine (C-C motif) ligand 3 (CCL3)



Interleukin 1, beta (IL1B)



Suppressor of cytokine signalling 3 (SOCS3)



Matrix metallopeptidase 19 (MMP19), transcript variant 1



Metallothionein 2A (MT2A)



Interleukin 7 (IL7)



Vascular endothelial growth factor C (VEGFC)



Aquaporin 1 (Colton blood group) (AQP1)



Hypoxia-inducible protein 2 (HIG2)



G protein-coupled receptor 109B (GPR109B)



CCAAT/enhancer binding protein (C/EBP), alpha (CEBPA)



Results are means of six sets of cells

*P < 0.001 (ANOVA) compared to UC medium. A minus sign indicates that expression of the gene was down-regulated

The full list of the 1,088 genes that were differentially expressed at 24 h is given in the Supplementary Table.

Ingenuity Pathway Analysis

To further analyse the data, the list of common transcripts was imported into the Ingenuity Pathway Analysis (IPA) programme to allow any possible pathways to be identified. The Canonical Pathways function indicated a wide range of functional pathways in which the differentially expressed genes could play a role. These canonical pathways included acute phase response signalling, IL-10 signalling, IL-6 signalling and NF-κB signalling. When using the Pathways tool in IPA, the involvement of seven pathways was shown to be statistically significant, including those associated with inflammation, macrophage infiltration, glucose uptake and lipid accumulation, formation and lipolysis (Table 4). The most significant pathway was inflammation, in which a total of 37 genes was shown to be either up- or down-regulated following treatment with MC medium (Table 4).
Table 4

Pathways regulated in human adipocytes by macrophage-conditioned medium


P value





Lipid formation



Macrophage infiltration



Glucose uptake



Lipid accumulation









The pathways shown were identified by Ingenuity Pathway Analysis as being statistically significant in human adipocytes (SGBS) incubated in macrophage-conditioned medium relative to the cells incubated in unconditioned medium, based on the number of up- or down-regulated genes within the pathway from the 1,088 common, differentially regulated genes

These 37 inflammation-related genes were then visualised by IPA to view possible pathway connections in the inflammatory response following 4 or 24 h treatment with MC medium. The first pathway created highlights the central role of NF-κB. This pathway also demonstrates that the degree of gene regulation differs in the context of acute (4 h) or chronic (24 h) exposure to the conditioned medium (Fig. 3a, b). For example, at 4 h expression of the MMP9 gene was only increased by 3.9-fold, whereas after 24 h, it was up-regulated by nearly 390-fold.
Fig. 3

Ingenuity Pathway Analysis was performed using the 37 inflammation-related genes (from the original 1,088 commonly regulated probes) differentially expressed at 4 h (left) and 24 h (right) of treatment with macrophage-conditioned medium. NF-κB is the central node in this pathway, indicating its importance in the network. Differentially regulated genes are shown in red and green colouring, depicting up- and down-regulation following treatment with conditioned medium. Bold colour indicates a high degree of regulation. Non-coloured genes directly or indirectly associated with the differentially expressed genes were not found to be differentially expressed following treatment. The significance of the node shapes is shown in the legend. Positive regulatory interactions are marked by solid arrows (direct interactions) or hashed arrows (indirect interactions). Negative interactions are shown by inhibitory arrows

Matrix metalloproteinases

From the original analysis of the common gene data, a cluster of genes were identified as being highly up-regulated at both 4 and 24 h and these were members of the family of matrix metalloproteinases (Table 5). MMP1 was the most highly up-regulated gene in the adipocytes incubated with MC medium, the increase in mRNA level being as much as 2,400-fold at 24 h (Table 5). MMP3 and MMP10 both also showed >1,000-fold increases in mRNA level in the MC medium-treated adipocytes, while MMP9 and MMP12 mRNA levels rose by 390- and 216-fold, respectively. Several MMPs were up-regulated at 24 h only; these included MMP11, MMP14 and MMP2 (mRNA level increased by 12-, nine- and threefold, respectively).
Table 5

Relative expression of matrix metalloproteinases (MMPs) in human adipocytes following treatment with macrophage-conditioned medium



Fold change at 4 h

Fold change at 24 h

MMP1: matrix metallopeptidase 1

Interstitial collagenase



MMP3: matrix metallopeptidase 3




MMP10: matrix metallopeptidase 10




MMP9: matrix metallopeptidase 9

Gelatinase B



MMP12: matrix metallopeptidase 12

Macrophage metalloelastase



MMP19: matrix metallopeptidase 19




The figures refer to the mean increase (for six sets of cells) in the level of specific MMPs from human adipocytes (SGBS) incubated in macrophage-conditioned medium relative to the cells incubated in unconditioned medium. All increases are statistically significant (P < 0.01 or greater)

To substantiate and extend the results observed in the microarray analysis, RT-PCR was carried out using specific primers for the three most highly up-regulated MMPs—MMP1, MMP3 and MMP10—in response to treatment with MC medium for 24 h. In the adipocytes incubated in UC medium, only a very faint band could be seen for MMP1 and MMP10 mRNA, while no bands were evident for MMP3 (Fig. 4). However, when the MC medium-treated samples were analysed, MMP1, MMP3 and MMP10 mRNAs were present as strong bands (Fig. 4). Sequence analysis confirmed that the bands corresponded to human MMP1, MMP3 and MMP10 mRNA, respectively. Control adipocytes showed no signal for any of these three MMPs—despite a high number of cycles (35 cycles) being employed.
Fig. 4

MMP1, MMP3 and MMP10 gene expression in human adipocytes following treatment for 24 h with either unconditioned medium (UC) or macrophage-conditioned medium (MC). Control adipocytes (Con) were incubated without any addition to the normal adipocyte medium. RT-PCR was performed as described in “Materials and methods”, using 35 cycles, with two separate sets of cells in each case

In order to determine whether the induction of MMP gene expression in human adipocytes by conditioned medium corresponds with the synthesis and release of the encoded protein, the secretion of MMP1 and MMP3 into the culture medium was measured by ELISA. No detectable MMP1 could be measured in the medium of control adipocytes, nor in adipocytes treated with UC medium (Fig. 5a). However, substantial amounts of MMP1 (292 ng/ml over 24 h; P < 0.001) were evident in the medium of adipocytes incubated with MC medium (Fig. 5a). MMP3 was also undetectable with the control cells, while some was detected with UC medium (Fig. 5b); MC medium resulted, as with MMP1, in a substantial stimulation of MMP3 release (211 ng/ml over 24 h, P < 0.001; Fig. 5b).
Fig. 5

Secretion of MMP1 and MMP3 by human adipocytes during treatment with either unconditioned medium (UC) or macrophage-conditioned medium (MC) for 24 h. Control adipocytes (Con) were incubated without any addition to the normal adipocyte medium. MMP1 (a) and MMP3 (b) were measured in culture media by ELISA. Results are mean values ± SE (bars) for six groups of cells. ***P < 0.001, compared with control and UC cells; †††P < 0.001 compared with control (ANOVA)

Regulation of MMP1 and MMP3 expression and release by TNFα

To investigate whether inflammatory cytokines contribute to the up-regulation of MMPs in human adipocytes by MC medium, the effect of the key pro-inflammatory cytokine, TNFα, on MMP1 and MMP3 gene expression was examined. SGBS cells were treated with either 5 or 100 ng/ml of TNFα for 24 h and the relative levels of MMP1 and MMP3 mRNA quantified by real-time PCR. TNFα produced a substantial, dose-dependent increase in MMP gene expression; an 118-fold and a 27-fold increase in MMP1 and MMP3 mRNA level, respectively, was observed with the higher dose of TNFα, and a 56-fold and sixfold increase with the low dose (Fig. 6a and b). Correspondingly, MMP1 was secreted into the medium following treatment with TNFα (low dose) and there was a substantial increase in MMP3 secretion (Fig. 6a, b).
Fig. 6

Effect of TNFα on (a) MMP1 and (b) MMP3 gene expression and secretion by human adipocytes. SGBS adipocytes were incubated in the presence of 5 ng/ml (LD) or 100 ng/ml (HD) of recombinant human TNFα for 24 h. Control adipocytes (Con) received no addition. MMP1 and MMP3 mRNA were measured by real-time PCR; relative expression was normalised to β-actin and the results expressed compared to the control adipocytes (≡1). MMP1 and MMP3 protein was measured in the medium of the cells receiving 5 ng/ml of TNFα. Results are mean values ± SE (bars) for five to six groups of cells. *P < 0.05, **P < 0.01, ***P < 0.001, compared with control cells; †P < 0.05, compared with UC cells (ANOVA)


In this study, macrophage-conditioned medium was used to investigate the effects of macrophages on global gene expression specifically in human adipocytes. The MC medium was shown to lead to significant up- or down-regulation of over 5,000 transcripts in human adipocytes at either the 4 or 24 h periods studied. From this large number of transcripts, identified through the 44,000 probes on the chips, 1,088 were found to be differentially regulated at both 4 and 24 h, and were the focus of further analysis. Genes that are differentially regulated only at 4 h are likely to reflect transitory responses, while those altered only at 24 h may reflect secondary rather than primary effects of the conditioned medium. The total number of genes found to be differentially regulated by MC medium in the present study is substantial, despite a stringent set of criteria being used in the analysis.

From the pathway analysis, it is evident that MC medium particularly affects adipocyte genes linked to inflammation. It also shows that NF-κB is central to the macrophage-stimulated inflammation pathway, in agreement with previous results [29]. The infiltration of macrophages into adipose tissue in obesity is considered an important contributor to the inflammatory state that occurs in the tissue as fat mass expands, with macrophages secreting key inflammatory factors [5, 8, 11, 48, 49]. A microarray study with 3T3-L1 adipocytes (mouse) co-cultured in the presence of LPS with RAW264.7, a murine macrophage cell line, has shown the up-regulation of genes linked to inflammation and angiogenesis, such as IL-6, MCP-1 and RANTES [50].

In the present study on human adipocytes, members of the family of MMPs were prominent as a group of highly up-regulated genes amongst the 1,088 genes differentially regulated at both 4 and 24 h. At 24 h, the three most highly up-regulated genes in adipocytes exposed to MC medium were MMP1, MMP3 and MMP10. Expression of these three genes was also strongly up-regulated at 4 h, indicating that they respond rapidly, as well as substantially, to the stimulatory effect of the MC medium. Other highly up-regulated MMPs were MMP9 and MMP12. Interestingly, MMP9 showed only a small increase in mRNA level at 4 h, while at 24 h the increase was 100-times greater, suggesting that there is a delayed response in the expression of this particular MMP.

The MMPs are endopeptidases which are important for degrading extracellular matrix proteins, including collagens; most are secreted enzymes. MMP1 is an interstitial collagenase, while both MMP3 and MMP10 are stromelysins (stromelysin 1 and 2, respectively) and MMP9 is a gelatinase. Several MMPs have been previously shown to be expressed in adipose tissue in obesity [6, 22, 26], including particularly MMP2 and MMP9 [3, 10, 12]. The majority of studies that have identified MMPs in adipose tissue have been undertaken using adipocytes from mice, but the presence of gene transcripts of several of the same MMPs was observed here in human adipocytes. Interestingly, in a comprehensive study of the expression of 16 different MMPs in mouse adipose tissue and adipocytes, MMP1 was not described [26]. However, in a proteomic study MMP1 was reported to be one of a number of proteins secreted by neonatal pig adipose tissue and stromal-vascular cells, though adipocytes as such were not investigated [17]. Again in a proteomic study, MMP1 has also been shown to be secreted, along with MMP2 and MMP9, by human adipose tissue, though not specifically adipocytes [2]. Expression of the MMP1 gene has, however, recently been described in human adipocytes and in human adipocyte cell lines [42].

In the present study, MMP1 was not found to be expressed in control human adipocytes, nor in adipocytes incubated with UC medium. MMP3 and MMP10 also showed little or no expression in control adipocytes. However, as noted above, all three of these MMPs were highly induced by the MC medium. Analysis of the medium itself confirmed that neither immunoreactive MMP1 nor MMP3 were present in the medium from control cells. MMP1 was also not released from cells incubated with UC medium, though some MMP3 was evident. However, substantial quantities of both proteins were released from SGBS adipocytes incubated with the MC medium, consistent with the results from the gene expression studies. This is the first report of MMP1 release from adipocytes and shows that this particular collagenase is up-regulated, and the protein secreted, in response to factors secreted from macrophages. It indicates that MMP1, which can be termed an adipokine, and MMP3 are released in association with an inflammatory state in adipocytes, which is a characteristic of obesity.

A well-recognised inflammatory protein secreted by macrophages is the pleiotropic, pro-inflammatory cytokine TNFα, and this is a strong candidate in the stimulation of MMP expression and secretion from adipocytes. Indeed, TNFα is an important factor in the activation of the expression of a number of inflammation-related genes in adipocytes [44, 45]. TNFα has been directly implicated in a paracrine loop, also involving fatty acids, between adipocytes and macrophages in the stimulation of inflammation in adipose tissue [35]. Furthermore, incubation with an anti-TNFα neutralising antibody has been shown to inhibit the inflammatory state of preadipocytes, implying that the macrophage-derived cytokine is an important mediator of inflammation in these adipocyte precursor cells [24].

TNFα was shown to strongly up-regulate the expression of MMP1 in human adipocytes, indicating that it could contribute to the stimulation of the production of this adipokine by MC medium. This stimulatory effect of TNFα on MMP1 gene expression in adipocytes is similar to that which has been described in other cells [7, 32, 47]. However, although TNFα induced a >100-fold increase in MMP1 mRNA level in adipocytes, this was only one-tenth of the increase in the level observed with the MC medium. This suggests that additional factors are likely to be required for the full stimulation. Macrophages secrete a multiplicity of factors which may be implicated in the induction of MMP production, and in the synthesis of other inflammation-related proteins in adipocytes. Identification of the most important factors is complex and requires further investigation. Candidates would include the cytokines IL-1β and IL-6; indeed, IL-1β stimulates MMP1 expression in other cell types [21, 23]. A similar situation is evident for MMP3, the expression and release of which were also markedly stimulated by TNFα.

Apart from the MMPs, a large number of other genes were also highly up-regulated in human adipocytes by the MC medium. These genes included those encoding IL-6, IL-8, MIP2b (CXCL3) and granulocyte colony-stimulating factor 3. Other important inflammation-related genes that were up-regulated include MCP-1 and IL-1β. The stimulation of their expression by MC medium indicates the pervasive extent to which macrophages activate the inflammatory response in adipocytes. Some are also those associated with macrophage chemoattraction, implying that macrophages stimulate the production and release of factors which lead to their own recruitment and maintenance within adipose tissue. IL-1β, Il-6 and MCP-1, in particular, are implicated in the development of insulin resistance [45], which suggests that macrophages play a role in the loss of insulin sensitivity in adipocytes in obesity.

The metallothioneins are another family of genes which were widely seen to be up-regulated within the list of common genes whose expression changed at both 4 and 24 h. The metallothioneins are a family of low molecular weight (approximately 6,000 Da), cysteine-rich metal-binding proteins. Metallothionein 2A (MT2A) has previously been shown to be expressed in human adipocytes [13], and more recently the expression of MT3 has been found to be strongly and rapidly induced in human fat cells in response to hypoxia, there being essentially no expression under normal conditions [46]. In the current study, MT-2A mRNA level was shown to be up-regulated sixfold by the MC medium, whereas MT3 expression was not induced. This indicates that the expression of MT3 is not dependent on inflammatory factors, the induction being a specific response to hypoxic conditions [46].

The expression (and secretion) of MT-1 has been previously described in rodent adipose tissue [38, 39]. However, in the present study several isoforms of the MT-1 gene were shown to be up-regulated in human adipocytes by the MC medium; these were MT-1B, 1E, 1G, 1H and 1X (each up-regulated between 14- and 23-fold at 24 h). The function of the metallothioneins remains unclear, but the probable roles include metal binding, angiogenesis and anti-oxidant defence. Extensive up-regulation of MT expression in adipocytes by MC medium could relate to angiogenesis (VEGF expression was increased by 4.7-fold at 24 h), or to protect against a putative macrophage-induced increase in reactive oxygen species.

The studies reported here used two cell systems—SGBS adipocytes and U937 monocytes. SGBS preadipocytes have a high capacity for differentiation to mature cells that are functionally similar to human adipocytes [43]. U937 monocytes [36] are established cell lines which, with the addition of PMA to the medium, differentiate into macrophages. However, PMA can have an effect on TNFα production in macrophages, and can also affect the expression of MMPs. This is unlikely to be a significant problem in the present work, but it may account for the presence of very faint bands on the PCR gels in the case of MMP1 and MMP10 with UC medium. Such an effect has been shown in other cell lines [34]. With respect to methodology, it should also be noted that in a study on preadipocytes, conditioned media produced by macrophages isolated from human adipose tissue exerted comparable effects to activated macrophages [24].

In summary, the present study employing DNA microarrays has indicated that macrophages have extensive effects on gene expression in human adipocytes, with the MMPs in particular being powerfully up-regulated. The substantial stimulatory effect on the expression of MMPs suggests that one of the key effects of macrophage recruitment in adipose tissue in obesity is to stimulate major remodelling within the tissue—and this may include angiogenesis and neovascularisation [28].


We are grateful to Dr. Stuart Wood and to Mr. Leif Hunter for their help and advice. PT is a member of COST BM0602. AOH is in receipt of an Industrial CASE Studentship from the Biotechnology and Biological Sciences Research Council (UK), which is also thanked for grant support to PT.


The authors have no conflicts of interests to declare; this study has no commercial implications.

Supplementary material

Copyright information

© Springer-Verlag 2009