Multiple actions of lysophosphatidic acid on fibroblasts revealed by transcriptional profiling
© Stortelers et al. 2008
Received: 25 April 2008
Accepted: 14 August 2008
Published: 14 August 2008
Skip to main content
© Stortelers et al. 2008
Received: 25 April 2008
Accepted: 14 August 2008
Published: 14 August 2008
Lysophosphatidic acid (LPA) is a lipid mediator that acts through specific G protein-coupled receptors to stimulate the proliferation, migration and survival of many cell types. LPA signaling has been implicated in development, wound healing and cancer. While LPA signaling pathways have been studied extensively, it remains unknown how LPA affects global gene expression in its target cells.
We have examined the temporal program of global gene expression in quiescent mouse embryonic fibroblasts stimulated with LPA using 32 k oligonucleotide microarrays. In addition to genes involved in growth stimulation and cytoskeletal reorganization, LPA induced many genes that encode secreted factors, including chemokines, growth factors, cytokines, pro-angiogenic and pro-fibrotic factors, components of the plasminogen activator system and metalloproteases. Strikingly, epidermal growth factor induced a broadly overlapping expression pattern, but some 7% of the genes (105 out of 1508 transcripts) showed differential regulation by LPA. The subset of LPA-specific genes was enriched for those associated with cytoskeletal remodeling, in keeping with LPA's ability to regulate cell shape and motility.
This study highlights the importance of LPA in programming fibroblasts not only to proliferate and migrate but also to produce many paracrine mediators of tissue remodeling, angiogenesis, inflammation and tumor progression. Furthermore, our results show that G protein-coupled receptors and receptor tyrosine kinases can signal independently to regulate broadly overlapping sets of genes in the same cell type.
Lysophosphatidic acid (LPA; monoacyl-glycerol-3-phoshate) is a lipid mediator that stimulates the proliferation, migration and survival of many cell types . LPA acts through at least five distinct G protein-coupled receptors (GPCRs), termed LPA1–5, which show overlapping signaling properties and tissue distribution [2, 3]. LPA signaling has been implicated in a great variety of biological processes, ranging from embryonic development to wound healing and tumor progression [1–5]. This multitude of activities is consistent with the broad distribution of LPA receptors and their coupling to multiple G proteins. LPA is produced from more complex lysophospholipids by a secreted lysophospholipase D known as autotaxin (ATX), originally identified as an 'autocrine motility factor' for tumor cells (for review see ref. ). ATX is essential for vascular development [6, 7] and in addition promotes tumor aggressiveness and angiogenesis . This strongly suggests that LPA is a key pro-angiogenic factor during development and a signifcant effector of cancer progression in the stroma-tumor microenviroment.
Although LPA-induced signaling pathways and cellular responses have been extensively analyzed over the years , it remains unknown how LPA affects global gene expression in its target cells. Gene expression analysis may uncover previously unknown activities of LPA, lead to a better understanding of GPCR signaling in general, and help to predict the behavior of cells in an LPA-enriched environment.
In this report, we analyze the global transcriptional response to LPA in mouse embryo fibroblasts (MEFs). Fibroblasts are abundant mesenchymal cells in the stroma of many tissues and organs where they regulate epithelial-mesenchymal interactions during development, tissue regeneration and tumor progression [9, 10]. LPA has long been known to stimulate the proliferation and migration of fibroblasts [1, 11], while excessive LPA signaling in these cells can lead to fibrosis . The present study identifies many novel LPA-regulated genes and shows especially that LPA commits fibroblasts, at the transcriptional level, to create a microenvironment that supports tissue remodeling, leukocyte recruitment, angiogenesis and tumor progression. Since fibroblasts are also responsive to epidermal growth factor (EGF), acting on its cognate receptor tyrosine kinase, we examined in addition to what extent the LPA- and EGF-induced expression patterns overlap.
The immediate transcriptional response to LPA was characterized by >100 upregulated immediate-early genes (mRNA levels peaking at 0.5–1 hr) (Figure 2; see additional files 2 and 2/Table 1). This gene set contained various transcription factors associated with growth stimulation and cell cycle progression, notably genes encoding the AP-1 complex (Fos, Fra1/FosL1, Jun, Atf3) and other growth regulatory genes (Egr1, Egr2, Klf6, Myc). Strong upregulation was also observed for Ptgs2, a highly inducible gene that encodes cyclooxygenase-2 (Cox-2) and has important roles in normal tissue homeostasis and inflammation. Upregulation of some of these immediate-early genes, including Egr1, Jun, Myc and Ptgs2, has earlier been found in LPA-treated MEFs and ovarian carcinoma cells [14, 15]. LPA also induced the expression of genes encoding growth-regulatory protein kinases, including Sgk1, Dyrk3, Nuak2 and Map2k3 (Mek3). In addition to these 'forward-driving' genes, the early gene clusters contained various 'feedback regulators' known to turn off gene expression and/or attenuate prolonged signaling (see Additional file 4: Table 1). Co-expression of both negative and positive regulators may be critical for the precise control of cell cycle progression. Among the feedback regulators induced by LPA were transcriptional repressors (Ztbtb16, Lfrd1/Tis7, Pawr), genes that promote the degradation of inducible mRNAs (Zfp36, Nocturnin/Ccrnl2) and a number of dual-specificity phosphatase (DUSP) genes, notably Dusp1, Dusp5, Dusp6, and Dusp10, whose products attenuate the activity of MAP kinases . It is further of note that LPA induced a very robust upregulation of Mig-6 also known as Errfi1 (ErbB receptor feedback inhibitor 1; ~60-fold induction after 2 h). MIG-6 is a scaffold protein that interacts with the EGF receptor to inhibit its catalytic activity and all downstream signaling events [17, 18]. By upregulating Mig-6, LPA may keep fibroblasts poised to prevent uncontrolled EGF receptor activation.
In addition to the above growth-regulatory genes, prominent upregulation was observed for genes whose products regulate the cytoskeleton, including various actin isoforms, vinculin and integrin alpha-subunits (Itga5 and Itga6), consistent with LPA's function as a regulator of cell shape and motility (see Additional file 4: Table 1).
Strongest downregulation of gene expression was mostly observed at ~4–6 hrs after LPA stimulation, i.e. coincident with the mid-G1 phase of the cell cycle (Figure 2; see Additional file 3). The most strongly down-regulated genes are listed in Additional file 4 (Table 2); it comprises the transcription factor Sox4, recently identified as a mediator of metastasis , the transcriptional repressor Slug (Snai2), an inducer of the mesenchymal phenotype and a marker of malignancy , as well as genes that promote growth arrest. The latter set included the cell cycle inhibitors Ccng2 (encoding cyclin G2) and Cdkn1b (p27Kip), and the 'growth arrest-specific' genes Gas1-3 and Ccn5. The latter gene encodes a growth inhibitory matrix protein (CCN5) whose reduced expression promotes mesenchymal cell motility . Other strongly downregulated genes were Nedd9/Hef1 and Cdh2. The Nedd9 scaffold is a downstream effector of focal adhesion kinase that may transduce integrin "inside-out" signaling to regulate cell-matrix adhesion and invasion [22, 23], while Cdh2 encodes N-cadherin, a key regulator of intercellular adhesion strength in fibroblasts . Through the coincident loss of N-cadherin, Nedd9 and CCN5, together with the upregulation of secreted metalloproteases (see below), LPA-stimulated fibroblasts may reduce their adhesive contacts and acquire a more motile and invasive phenotype.
Components of the urokinase-type plasminogen activator (uPA) system, notably PlauR (uPAR; peaking at 2 hrs) and the two major inhibitors of this system, plasminogen activator inhibitor-1 (PAI-1; peaking at 2–4 hrs) and PAI-2 (peaking at 4–6 hrs) were also strongly upregulated by LPA. Through its action on extracellular matrix and cell-surface proteins, the uPA system modulates cell migration and cell-matrix interactions and thereby plays a key role in wound healing, angiogenesis and tumor progression.
Genes encoding additional extracellular mediators and surface-exposed proteins were induced in a second wave of transcription with expression peaking at 4–6 hr after LPA stimulation (Figure 2; see additional files 3 and 4/Table 1). These included the chemokines CCL-2, CCL-7 and Cx3Cl1, the matrix metalloproteinases Adam19 and MMP3, and the transmembrane glycoprotein CD44, which plays a key role in cell-cell interactions. A very strongly upregulated gene was ILl1rl1, which encodes an IL-1 receptor family member (murine ST2; ~40-fold induction at 4 hrs). The secreted form of ST2 inhibits the production of cytokines in inflammatory cells . Thus, ST2 upregulation in LPA-stimulated fibroblasts may serve to temper inflammatory responses. Consistent with our data, ST2 was recently identified as major LPA target gene in osteoblastic cells, where it is thought to play an anti-inflammatory role during bone healing . Strong upregulation was also observed for tissue factor (coagulation factor III, encoded by F3), a cell-surface glycoprotein that initiates the clotting cascade and has additional roles in cell migration and angiogenesis . Late upregulation (peak expression at ~6 hrs: see Additional file 3) was observed for Lgals3 and Timp1. Lgals3 encodes galactin-3, a mediator of inflammation, while secreted Timp-1 has a role in wound healing and the creation of a prometastatic niche.
In conclusion, LPA-stimulated fibroblasts are transcriptionally committed to produce numerous factors known to act on nearby epithelial cells, leukocytes and endothelial cells. LPA stimulation thus enables fibroblasts to promote tissue remodeling, inflammation, angiogenesis, wound healing and, in a tumor context, cancer progression. A schematic representation of the LPA-induced gene expression program over time is shown in Additional file 7.
Despite this large overlap of the LPA and EGF expression profiles, approx. 7% of the genes (105 out of 1508 transcripts) was differentially regulated by LPA (Figure 8; see additional files 8 and 9). Immediate-early genes that were upregulated by LPA, but not EGF, include Edn1 (endothelin-1), Fgf16, Nfkbia (NF-kappaB inhibitor alpha) and several protein kinase genes (Bmp2k, Plk2, Tesk2, Pim1), as shown in Additional file 8 (cluster 3). Interestingly, LPA also induced the expression of a newly identified GPCR for LPA, termed P2Y5 (encoded by P2ry5 ), which adds an element of feedback to the fibroblast response to LPA (see Additional file 8: cluster 3). At 2–4 hrs, more LPA-specific transcripts could be identified (Figure 8B; see additional files 8 and 9). Gene ontology analysis revealed that the LPA-specific gene set was enriched for genes associated with cytoskeletal organization and integrin signaling, notably those encoding various actin isoforms (Actb, Acta1, Actg2), palladin (Palld), vinculin (Vcl), an Arp2/3 subunit (Arpc5), calponins (Cnn1, Cnn3), a Rho GTPase (Rhoj), Rho-kinase (Rock2), myosin X (Myo10) and an integrin subunit (Itga5) (see Additional file 9). Specific upregulation of cytoskeleton-associated genes is in line with LPA's role as an efficacious regulator of cell shape and motility.
Dissecting the transcriptional response to growth factors in selected cell systems may help to better understand various aspects of embryonic development, adult tissue homeostasis and cancer. The present study characterizes the global transcriptional program of MEFs to LPA and thereby provides new insights into the normal physiological response of quiescent fibroblasts to this multifunctional lipid mediator. In addition to genes associated with cell proliferation, adhesion and migration, LPA induces a host of genes that encode secreted factors known to promote tissue remodeling, wound healing, inflammation, angiogenesis and tumor progression, depending on cellular context. This highlights the importance of LPA signaling in profoundly modifying the fibroblast microenvironment.
Previous transcriptional profiling of serum-stimulated human skin fibroblasts has identified a 'core serum response' (CSR) that is characterized by cell-cycle-independent genes and reflects various aspects of wound healing, notably the induction of genes involved in matrix remodeling and re-epithelialization . This fibroblast CSR or "wound-response signature" is recapitulated in human carcinomas and may help predict tumor progression . Comparative analysis revealed, however, that the expression profile of LPA-stimulated MEFs shows only limited overlap with the canonical CSR of serum-stimulated human fibroblasts. At first sight, this result is somewhat unexpected since LPA is a major serum constituent [1, 13]. On the other hand, serum is an ill-defined mixture of numerous bioactive factors and it is likely that the combined action of many different factors obscures the comparison between serum and LPA. Moreover, MEFs differ from human skin fibroblasts not only in their biological and anatomic origin, but also in their LPA receptor expression pattern (data not shown). Yet, it should be emphasized that LPA-stimulated MEFs and serum-stimulated human skin fibroblasts both show a gene expression profile that is strongly associated with tissue remodeling as well as tumor progression.
Of final note is our finding that the transcriptional response of MEFs to LPA versus EGF shows an overlap of >90%, at least qualitatively, despite the fact that LPA and EGF signal via completely different mechanisms. LPA-induced gene expression was largely independent of EGF receptor activity, which argues against the notion that LPA exploits the EGF receptor as a signaling intermediate. The broad overlap between LPA- and EGF-induced gene expression shows that GPCRs and RTKs have more in common than previously appreciated. One should not conclude, however, that the transcriptional response to receptor stimulation is less dependent on the nature of the receptor than on the cell type. For example, the common expression pattern of LPA and EGF shows hardly any overlap with that induced by Wnt signaling in fibroblasts . The great diversity of fibroblast responses to LPA as reported here is an important area for further study.
Mouse embryonic fibroblasts (MEFs) were immortalized at passage 2 by retroviral introduction of the T-box member Tbx2 (LZRS-TBX2-ires-EGFP) to represses p53 function . MEFs were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 8% fetal calf serum (FCS), penicillin and streptavidin, and seeded every three days at a density of 1.6 × 104/cm2 according to the 3T3 protocol.
Detailed protocols for RNA isolation, amplification, labeling and microarray hybridization can be found at http://microarray.nki.nl/download/protocols.html. MEFs were seeded at density of 2.5 × 104/cm2, and 24 h later were starved for 40 h in serum-free medium supplemented with 0.2% bovine serum albumin (BSA; Sigma). Cells were treated with oleoyl LPA (0.3–5 uM) coupled to fatty acid free BSA (Sigma) at a 3:1 molar ratio, or EGF (20 ng/ml) for the indicated time period. Before harvesting, cells were washed twice with ice-cold phosphate buffered saline. Isolation of total RNA was performed with RNAzol Bee (Campro Scientific, Amersfoort, the Netherlands). Isolated total RNA was subsequently Dnase1-treated by using the Qiagen RNase-free DNase kit (Cat. no. 74106) and RNeasy spin columns (Qiagen, West Sussex, UK, Cat. no. 79254) and dissolved in RNase-free H2O. RNA concentration and purity was measured on a NanoDrop ND-1000 spectrophotometer (Isogen Life-Science), while RNA integrity was determined by agarose gel electrophoresis. Four micrograms of total RNA was used to synthesise cDNA with a T7-(dT)24 primer and RT superscript III (Invitrogen Life Technologies; Cat. no. L1016-01). After second-strand synthesis and removal of contaminating RNA, cRNA was generated by in vitro transcription using T7 RNA polymerase. Amplification yields were 1,000- to 2,000-fold.
The cRNA of each sample was labeled with Cy5 or Cy3 (Universal Linkage System, Kreatech Biotechnology, Cat. no. EA-006) according the manufacturer's protocol, with minor adjustments. For each labeling, 0.3 μl of the Cy5-ULS or 1 μl Cy3-ULS was mixed with 1 μg of cRNA and 2 μl of labeling solution. The volume of this mix was adjusted to 20 μl and incubated for 30 minutes at 85°C, followed by purification with a KREA pure column. Dye incorporation was determined after measuring the labeled cRNA on a NanoDrop ND-1000 spectrophotometer and mixed with the same amount of reverse color Cy-labeled cRNA from the untreated control. Labeled cRNAs were fragmented to pieces of 60–200 nt (Ambion, Cat. no. AM8740). Before hybridization, 20 μg of COT-1 DNA (Invitrogen-Life Technologies, Cat. no. 15279-011), 8 μg of yeast tRNA (Roche Diagnostics B.V., Cat. no. 10109495001) and 20 μg of poly-d(A) (GE Healthcare Biosciences Europe GmbH, Cat. no. 27-7836-03) was added and the volume was adjusted to 60 μl. To this volume 60 μl of a formamide buffer containing 50% Formamide, 50% 20 × SSC (NaCl/Na-citrate) and 0.1% SDS was added. All hybridizations were performed in a hybridization station (Tecan, Cat. no. Hs4800). Before hybridization with the labeled RNA samples, the microarray was pre-hybridized with a bovine serum albumin solution (1% BSA, 5 × SSC and 0.1% SDS) for one hour at 42°C and washed with water and a 5 × SSC, 0,1% SDS solution. The labeled material was denatured at 95°C for 3 minutes and cooled to 42°C before injection in the hybridization chamber. After hybridization for 16 hrs, slides were washed in the hybridization station with a 5 × SSC, 0,1% SDS solution, a 2 × SSC, 0,1% SDS solution and a 1 × SSC solution at 42°C and a 0,2 × SSC solution at 23°C. Finally, the slides were dried with medical grade nitrogen for 3 min. at 30°C. A DNA Microarray scanner (Agilent Technologies, Cat. no. G2505B) was used to scan the slides. To monitor the consistency of the array experiments, "self-self" experiments were performed using the same sample as reference. Fluorescent intensities of the images were quantified by using ImaGene v6.0 software (Biodiscovery Inc.). This software has an output of two text files which were uploaded to the CMF database (CMFdb, http://cmfdb.nki.nl) for further analysis. The background-corrected intensities from the Cy5 and Cy3 channel were used to calculate log2transformed ratios. These ratios were normalized using a lowest fit per subarray . Experiments done in dye-swap fashion were combined to create one dataset on which an outlier analysis was performed. A weighted average ratio and confidence level (P-value) was calculated per gene by a NKI platform adjusted error model , which was fine-tuned by self-self hybridizations. Differentially expressed genes between sample and reference were selected based on their P-value (a gene with a P-value < 0.01 is considered an outlier). Genes were selected for further analysis if they had p-values < 0.01 and 2log ratios greater than 0.67 or smaller than -0.67 corresponding to a 1.7-fold change, unless indicated otherwise. To identify structural patterns of gene expression, selected outliers were used for hierarchical clustering using the complete linkage algorithm and for K-means clustering analysis (Euclidian distance) using the GENESIS program . To identify genes that differ between LPA- and EGF induced treatment Anova analysis was applied (two groups) in which the expression values of two consecutive time points were used. "Ingenuity" pathway analysis was used for functional analysis on the set of significantly modulated genes to identify affected biological pathways and functional processes. In addition, the web-based platform Gene Ontology Tree Machine (GOTM) was used to identify GO terms with relatively enriched gene numbers .
Mouse 32 k Operon v3.0 oligo arrays from the Central Microarray Facility (CMF) at the Netherlands Cancer Institute were used for hybridization. A complete list of genes and controls present on the slides is available on the CMF web site http://microarrays.nki.nl/download/geneid.html. The 70-mer oligo's (Operon, AROS v3.0) were printed on UltraGaps slides (Corning) with a BioRobotics MicroGrid II (Genomic Solutions) print robot. The description of our microarray study follows the MIAME guidelines and the entire microarray data set has been deposited in the EBI/ArrayExpress database and is accessible through accession numbers E-NCMF-16, E-NCMF-17, E-NCMF-18 and E-NCMF-19).
Total RNA was extracted using the RNeasy minikit combined with on-column Dnase-I treatment (Qiagen, West Sussex, UK) and resolved in diethyl pyrocarbonate-treated H2O (DEPC; Sigma). First-strand cDNA synthesis was done using 2 ug of total RNA in presence of 0.5 ug oligo-dT primers (Invitrogen), 40 units RNase inhibitor Rnasin (Promega), 500 nM deoxynucelotide triphosphates (Roche), 10 uM DTT and 10 units Superscript II RT in 1× reverse transcriptase buffer (Invitrogen). Sequences of real-time quantitative PCR primers were designed using Primer express software (PE Biosystems, Foster City, CA). Primer sequences are available upon request. Detection and quantification of each gene was accomplished by SYBRgreen incorporation using the ABI PRISM 7700 sequence detection system (Applied Biosystems). Quantitative RT-PCR was carried out using 40 ng cDNA, 300 nM of each oligo in presence of 1× SYBRgreen mix in 20 ul reactions (Applied Biosystems). Cycling parameters were: 2 min incubation at 50°C, 10 min. incubation at 95°C, followed by 50 PCR cycles consisting of 15 sec at 95°C and 1 min. at 60°C. Product sizes were routinely verified by collecting a melting curve from 55°C to 95°C after final amplification. The relative product levels were quantified using the 2-ΔΔCT method. Data are presented as relative induction of each target gene, normalized to the expression of HPRT, and are representative of two independent experiments.
We thank Marloes Bagijn for PCR and biochemical experiments, Mike Heimerikx and Marja Nieuwland for assistance with the microarray studies, Arno Velds and Daoud Sie for bioinformatics assistance and Lodewijk Wessels for help with statistical analysis. We thank Panthea Taghavi for MEFs and the TBX2 vector. This study was supported by the Dutch Cancer Society and the Centre for Biomedical Genetics. The authors declare no conflict of interest.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.