LOXL4 Is Induced by Transforming Growth Factor β1 through Smad and JunB/Fra2 and Contributes to Vascular Matrix Remodeling

Cell culture.

Primary bovine aortic endothelial cells (BAEC) were isolated from thoracic aortas and maintained in culture using previously described methods (12). Mouse aortic endothelial cells (MAEC) were purchased from Angio-Proteomie (Boston, MA) and cultured according to the manufacturer's instructions. Smad3 knockout and wild-type embryonic fibroblasts (MEF) were isolated from newborn mice and kept in culture using standard procedures (13). Tetracycline (Tet)-inducible human embryonic kidney 293 (HEK 293) cells stably expressing LOXL4-green fluorescent protein (GFP) constructs were generated and maintained in culture according to protocols previously published (14).

RNA expression by quantitative PCR and microarray analysis.

For RNA experiments, BAEC or MAEC were cultured in 100-mm cell culture dishes and treated with 5 ng/ml TGF-β1 (R&D Systems, Minneapolis, MN). Total RNA was isolated by guanidinium thiocyanate/phenol-chloroform extraction (15). cDNA was synthesized from 1 μg of RNA using an iScript cDNA synthesis kit, and quantitative PCR for the detection of LOX isoforms or Fra2 mRNA was performed using iQ SYBR green Supermix and specific primers (see Table S1 in the supplemental material) in a CFX96 thermocycler (Bio-Rad, Hercules, CA). Relative mRNA expression was determined using the glyceraldehyde 3-phosphate dehydrogenase gene as a housekeeping gene according to the ΔΔC_T method (16).

cDNA microarray analysis was carried out using a GeneChip bovine array (Affymetrix, Santa Clara, CA) containing 24,027 probe sets representing over 23,000 transcripts (including approximately 19,000 UniGene clusters). The purity and integrity of total RNA were checked with an RNA Bioanalyzer (Agilent 2100). RNA labeling, hybridization, scanning, and analysis were performed in the Genomics Unit of the Centro Nacional de Biotecnología (Madrid, Spain) according to the manufacturer's instructions. Genes showing statistically significant expression changes in response to TGF-β1 exposure were identified by using a cutoff value of 2.

Western blot analyses.

For protein studies, bovine aortic endothelial or HEK 293 cells were cultured in 100-mm cell culture dishes and processed essentially as previously described (12). Briefly, cells were washed with phosphate-buffered saline (PBS) and lysed with 300 μl Tris-SDS buffer (60 mM Tris-HCl [pH 6.8], 2% sodium dodecyl sulfate [SDS]) to obtain total cell lysate. Protein concentrations were determined with the bicinchoninic acid (BCA) protein assay kit (Thermo Scientific, Rockford, IL). Equal amounts of protein (50 μg) were separated on 7.5% SDS-polyacrylamide gels, and fractioned proteins were transferred onto nitrocellulose membranes at 12 V for 50 min in a semidry Trans-Blot Turbo system (Bio-Rad). Membranes were blocked by incubation for 1 h with 3% bovine serum albumin (BSA) in PBS containing 0.5% Tween 20, and antigens were detected using specific primary antibodies (anti-LOXL4 [1:2,000, polyclonal, generated in our own laboratory by rabbit immunization with the C-terminal peptide ELSQEQEQRLRN, not conserved in the remaining LOX isoforms but highly homologous among bovine, mouse, and human LOXL4], anti-β-actin [1:20,000, monoclonal; Sigma], anti-Fra2 [1:1,000, rat antibody generated in our own laboratory using a histidine-tagged mouse full-length Fra2 protein], anti-Flag [1:2,000, mouse monoclonal from Sigma], anti-JunD [1:1,000, rabbit polyclonal; Santa Cruz], anti-JunB [1:1,000, rabbit polyclonal; Santa Cruz], anti-c-Jun [1:1,000, rabbit polyclonal; Santa Cruz], anti-extracellular signal-regulated kinase [anti-ERK] total [p44/42 MAPK] [1:2,000, rabbit polyclonal; Cell Signaling], anti-phospho-ERK [phospho-p44/42 MAPK, T202/T204] [1:1,000, rabbit polyclonal; Cell Signaling], and anti-GFP [1:2,000, mouse monoclonal; Roche]). Blots were then developed and quantified using corresponding IRDye 680- and IRDye 800-labeled secondary antibodies with the Odyssey infrared imaging system (Li-Cor).

5′ RACE.

To define the transcription initiation site of the LOXL4 gene, 5′ rapid amplification of cDNA ends (5′ RACE) analysis was performed using the SMART RACE cDNA amplification kit (Clontech, Mountain View, CA). Briefly, total RNA from BAEC incubated with TGF-β1 was reverse transcribed in order to generate a first-strand cDNA suitable for 5′ RACE using SMART technology. A universal primer and a gene-specific primer (see Table S1 in the supplemental material) were then used for the generation of the 5′-end product. PCR fragments were analyzed by agarose electrophoresis and visualized with ethidium bromide staining. The resulting bands were excised from the gel, purified using the Geneclean Turbo kit (MP Biomedicals, Santa Ana, CA), and TA cloned into pCR2.1 (Life Technologies, Carlsbad, CA). A significant number of Escherichia coli colonies were picked, grown up, and analyzed by sequencing using M13 reverse and forward primers.

Construction of LOXL4 gene promoter reporter constructs and cell transfection.

Analysis of LOXL4 gene transcriptional activity was performed by luciferase reporter gene studies using genomic fragments of the LOXL4 gene. Human LOXL4 promoter fragments were obtained by restriction cutting from clone RP11-34A14 on chromosome 10 containing complete LOXL4 gene (BACPAC resources; Children's Hospital Oakland Research Institute [CHORI]). Preliminary attempts to generate by PCR fragments of the LOXL4 promoter covering about 4 kb of 5′ flanking sequence and the complete exon 1 using bovine and mouse genomic DNAs were unsuccessful, likely due to the presence of a high GC sequence in the proximal promoter. Therefore, PCR assays were designed to generate the corresponding bovine and mouse LOXL4 promoter fragments upstream from the high-GC segment, and proximal promoter sequences were obtained by chemical synthesis (inserts in pBlueScript vector [Epoch Life Sciences, Missouri City, TX]) (see Table S2 in the supplemental material). Luciferase cassettes under the control of the LOXL4 promoter were generated by assembly of upstream PCR fragments and synthetic proximal sequences into pGL3 vector (Promega, Madison, WI) by standard cloning procedures. 5′ deletion constructs of the bovine LOXL4 promoter were generated by PCR and assembled into the luciferase reporter proximal promoter plasmid as described above. A distal fragment of the bovine LOXL4 promoter containing AP-1 and Smad binding elements (113 bp) was cloned in front of a cassette composed of a constitutive simian virus 40 (SV40) promoter fused to a luciferase gene (pGL3-Promoter; Promega) or assembled to the proximal promoter luciferase vector in order to analyze its potential enhancer capacity. Reporter luciferase constructs of the bp −3870 LOXL4 promoter with specific mutations in AP-1 and in Smad binding sites were generated by PCR-based site-directed mutagenesis as previously described (12). The AP-1 site TGAGTCA was replaced by GTAGTCA. Smad binding elements were mutated as follows: the SmadA CAGA was replaced by TAAT, and the SmadB GTCTG was changed to ATGTA. LOXL4 promoter constructs with a specific mutation in the Sp1 site (GGGGGCGGGG changed to GTTTTCGGGG) were obtained by chemical synthesis (Epoch Life Sciences). All constructs were verified by sequencing.

The vectors expressing the Smad3 and Smad7 transcription factors under the control of strong constitutive cytomegalovirus (CMV) promoters were kindly provided by Liliana Attisano (Toronto, Canada) and Aristidis Moustakas (Uppsala, Sweden), respectively. Mammalian expression vectors for Sp1 and Sp3 were kindly provided by Guntram Suske (Marburg, Germany). Mouse Fra2 cDNA was excised from pBabe-Fra2 plasmid and cloned into pcDNA3 (CMV promoter; Life Technologies) (17). A site-directed mutagenesis approach was used to generate a Fra2 expression vector with a C-terminal Flag tag (pcDNA3-Fra2-Flag) and, from this construct, Fra2-Flag phosphorylation mutants with inactivating mutations threonine to nonphosphorylatable alanine (T→A) and phosphomimetic mutations replacing threonine by glutamic acid (T→E) in residues T263 and T274 (see Table S3 in the supplemental material). Constructs for the overexpression of tethered JunB/Fra2, JunD/Fra2, and c-Jun/Fra2 dimers have been previously described (18).

DNA constructs were transiently transfected into BAEC, MAEC, and wild-type or Smad3 knockout mouse embryonic fibroblasts on 24-well plates (60 to 70% confluence), and promoter activity was estimated by luminometry as described previously, using the pRL-CMV plasmid (containing a Renilla luciferase gene under the control of the CMV promoter) for normalization purposes (12)

Comparative genomic DNA analysis was performed by using the Mulan multiple-alignment server (http://mulan.dcode.org) available at the National Center for Biotechnology Information DCODE server (ECR browser; http://www.dcode.org/) (19). LOXL4-orthologous sequences from Homo sapiens, Mus musculus, Rattus norvegicus, Canis familiaris, Felis catus, Bos taurus, Pongo pygmaeus, Pan troglodytes, and Equus caballus were obtained from the ENSEMBL genome database (http://www.ensemble.org/). The CPGPlot program was used to detect CpG-rich regions that may correspond to potential CpG islands. The ClustalW2 program was used to align protein sequences from Fra2 and c-Fos orthologs in order to screen for evolutionary conserved MAPK phosphorylation sites (http://www.ebi.ac.uk).

ChIP.

Chromatin immunoprecipitation (ChIP) experiments were performed using a commercially available kit (Pierce agarose ChIP kit; Thermo Fisher Scientific) according to the manufacturer's instructions. Briefly, cells were fixed with 1% formaldehyde (Calbiochem, La Jolla, CA) for 10 min at room temperature with swirling. Glycine was added to a final concentration of 0.125 M, and the incubation was continued for an additional 5 min. Cells were washed twice with ice-cold PBS, harvested by scraping, pelleted, and resuspended in 100 μl of lysis buffer containing protease inhibitors. Samples were centrifuged and the nuclear pellet resuspended in 100 μl of micrococcal nuclease (MNase) solution. Digestion proceeded for 15 min at 37°C with occasional mixing. Afterwards, the reaction was stopped, samples were centrifuged, and supernatants were saved as digested chromatin for immunoprecipitation. After removal of a control aliquot (input), samples were incubated at 4°C overnight with anti-Smad2/3 (mouse monoclonal; BD Transduction Laboratories, Franklin Lakes, NJ), anti-Fra2 (in-house, rat), or negative-control IgG. Protein A/G Plus agarose beads were then added, incubated for 1 h, and then extensively washed before chromatin was released by proteinase K digestion and DNA was further recovered using spin columns. Purified DNA was used for amplification of a region of 104 bp located at the upstream enhancer by real-time quantitative PCR (see Table S4 in the supplemental material). Quantities of target DNA in the ChIP and input samples were determined based on a standard curve and expressed as percentage of input after IgG background subtraction.

EMSA and oligonucleotide pulldown experiments.

After different treatments, nuclear extracts were prepared according to a procedure described previously (20). Briefly, cell plates were washed once with ice-cold PBS and scraped in the same medium. Upon centrifugation, they were resuspended in 10 mM HEPES (pH 7.9), 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol (DTT), 0.5 mM phenylmethylsulfonyl fluoride, and 1 μg/ml each of leupeptin, aprotinin, and pepstatin A (Sigma). After 15 min of incubation on ice, Nonidet P-40 was added to a final concentration of 0.5% and cells were vortexed for 10 s. Then nuclei were sedimented by centrifugation for 5 min at 14,000 × g. The nuclear pellets were extracted by shaking on ice for 30 min in 20 mM HEPES (pH 7.9), 400 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, and protease inhibitors as described above. After centrifugation, nuclear debris was discarded by centrifugation at 14,000 × g for 5 min, and nuclear extracts were stored at −80°C. Protein content was determined using the Bradford method (Bio-Rad). Constructs for bacterial recombinant expression of the glutathione S-transferase (GST) fusion proteins GST-Smad2, GST-Smad3, and GST-Smad4 were kindly provided by Liliana Attisano (Toronto, Canada). Fusion proteins were expressed in E. coli and purified as described previously (12). JunB/Fra2 forced dimers, wild type or with mutations in residues T263 and T274, were prepared in vitro using the TNT rabbit reticulocyte coupled transcription-translation system (Promega). JunB/Fra2 protein expression was checked by Western blot analysis using anti-JunB antibody.

Electrophoretic mobility shift assays (EMSA) with nuclear extracts (4 μg), purified GST fusion forms (0.1 to 1 μg), or equivalent amounts of in vitro-synthesized proteins were performed as described previously using infrared-labeled oligonucleotides for the AP-1 or the Smad binding sites (see Table S5 in the supplemental material) (12). Free oligonucleotides and DNA-protein complexes were visualized with the Odyssey infrared imaging system (Li-Cor). For competition experiments, unlabeled oligonucleotides corresponding to wild-type Smad and AP-1 binding sites, or including specific mutations, were added to the incubation mixture in a 1- to 100-fold excess.

Oligonucleotide pulldown experiments were designed in order to identify and characterize proteins interacting with the DNA sequence containing the AP-1 and Smad binding sites. 5′-biotinylated wild-type and mutated AP-1/SmadB DNA fragments corresponding to positions −3866/−3754 were generated by PCR using the corresponding 5′-biotinylated forward primer and wild-type and mutated reverse primers (see Table S4 in the supplemental material), purified using Geneclean (Turbo kit MP Biomedicals), and verified by electrophoresis and ethidium bromide staining. For the reaction, 100 μg of nuclear extracts from bovine aortic endothelial cells incubated with or without TGF-β1 was incubated for 20 min at 4°C in a volume of 100 μl of binding buffer [10 mM Tris-HCl (pH 7.4), 50 mM NaCl, 1 mM MgCl₂, 5 mM DTT, and 5% glycerol and containing 0.02 mg/ml of poly(dI-dC) (General Electric, Little Chalfont, United Kingdom)] with 5′-biotinylated wild-type and mutated oligonucleotides. Afterwards, 50 μl of SoftLink soft release avidin resin (Promega; 1:1 slurry, equilibrated according to the manufacturer's instructions) was added and the mixture further incubated for 1 h at 4°C with occasional shaking. Beads were then washed five times with 500 μl of binding buffer without poly(dI-dC). Retained proteins were finally eluted from the beads by addition of 50 μl Laemmli buffer and centrifugation at 14,000 × g for 15 min, fractioned by SDS-PAGE, and visualized by Coomassie blue staining.

Bands corresponding to candidate proteins were dissected and gel pieces digested in situ with sequencing-grade trypsin (Promega) as previously described (21). Mixtures of proteolytically generated peptides were analyzed by reverse-phase liquid chromatography–tandem mass spectrometry (RP-LC-MS/MS) in an Agilent 1100 system coupled to a linear ion trap LTQ-Velos mass spectrometer (Thermo Fisher Scientific). The peptides were separated using a 0.18- by 150-mm Bio-Basic C₁₈ RP column (Thermo Fisher Scientific) operating at 1.8 μl/min. Peptides were eluted using a 35-min gradient from 5 to 40% solvent (solvent A, 0.1% formic acid in water; solvent B, 0.1% formic acid and 80% acetonitrile in water). Electrospray ionization (ESI) was done using a microspray “metal needle kit” interface (Thermo Fisher Scientific). Peptides were detected in survey scans from 400 to 1,600 amu (1 microscan), followed by five data-dependent MS/MS scans (Top 5), using an isolation width of 2 U (in mass-to-charge ratio units), a normalized collision energy of 35%, and dynamic exclusion applied during 30-s periods. Peptide identification from raw data was carried out with the SEQUEST algorithm (Proteome Discoverer 1.2; Thermo Fisher Scientific) using the Bos taurus UniProt database. The following constraints were used for the searches: tryptic cleavage after Arg and Lys, up to two missed cleavage sites, and tolerances of 1 Da for precursor ions and 0.8 Da for MS/MS fragment ions. The searches were performed allowing optional Met oxidation and Cys carbamidomethylation. Searching against a decoy database (interpreted decoy approach) was performed using a false discovery rate of <0.01. The complete proteomic analysis was carried out in the Centro de Biología Molecular Severo Ochoa Protein Chemistry Unit, a member of ProteoRed network (www.proteored.org).

GST-Fra2 pulldown assays.

Mouse Fra2 cDNA sequence was amplified from plasmid pBabe-Fra2 with primers containing adequate restriction sites for cloning into the pGEX-2T vector (GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom) (see Table S6 in the supplemental material). Specific mutations in residues T263 and T274 were introduced by site-directed mutagenesis using primers shown in Table S3 in the supplemental material. Fusion proteins were expressed in E. coli and purified as described previously (12). For pulldown experiments, equivalent amounts of GST-Fra2 beads were incubated with total extracts from bovine aortic endothelial cells for 2 h at room temperature with rocking. The beads were then washed four times with PBS and then resuspended in Laemmli buffer for SDS-PAGE and Western blot analysis of endogenous JunB and recombinant Fra2.

Adenoviral infection and siRNA-mediated inhibition of gene expression.

Recombinant-deficient human adenovirus type 5 (dE1/E3) overexpressing a dominant negative form of AP-1 (Adv-TAM67) and empty vector adenovirus (Adv-Null) were purchased from Vector Biolabs (Philadelphia, PA). Mouse Fra2 and bovine LOXL4 cDNA (the former excised from pcDNA3-Fra2 and the latter generated by reverse transcription coupled to PCR using total RNA from TGF-β1-incubated bovine aortic endothelial cells) were cloned into a suitable shuttle vector to generate adenoviral constructs for mammalian overexpression (Vector Biolabs). Adenoviruses were amplified using embryonic kidney cell line HEK 293. Briefly, HEK 293 cells were cultured until the monolayer was 90% confluent and then infected with cytopathic levels of adenoviruses for 48 h. For purification of adenoviruses, infected cells and supernatants were taken and centrifuged at 1,000 × g for 5 min. Supernatants were discarded, and cell pellets were lysed with three freezing-thawing cycles to release virus particles. The viruses were then purified using the Adeno-X Maxi purification kit from Clontech (Mountain View, CA) according to the manufacturer's instructions. Virion concentrations were determined using the Adeno-X-Rapid titer kit from Clontech, and viral stocks were stored in aliquots at −80°C. For adenoviral infection of primary bovine aortic endothelial cells, these were seeded in 100-mm culture dishes and exposed to adenoviral constructs for 24 h. Cells were then washed and incubated in serum-free medium for an additional 24-h period. In some experiments, infected cells were transfected with DNA plasmid during this 24-h period. Cells were then finally stimulated with TGF-β1 or left untreated as indicated.

Specific small interfering RNAs (siRNAs) for the inhibition of the expression of bovine Fra2, c-Jun, JunB, and JunD and a nontargeting control were purchased from Ambion (Silencer Select siRNA; Life Technologies) (see Table S7 in the supplemental material). Cell transfections were performed in 6-well plates as described above.

Immunofluorescence microscopy.

Fluorescence microscopy was performed as previously described (22). Briefly, primary bovine aortic endothelial cells were seeded onto 24-well plates containing 10-mm-diameter glass coverslips. After the corresponding treatment, cells were fixed with 4% paraformaldehyde in PBS for 10 min and then permeabilized with 0.2% Triton X-100 in PBS for 5 min at room temperature. Cells were washed with PBS, blocked with 1% BSA in PBS for 1 h, and then incubated overnight at 4°C with primary antibodies (anti-Fra2 [1:200, rat antibody described above], anti-GFP [1:200, mouse monoclonal from Roche], and anticalnexin [endoplasmic reticulum [ER] marker, 1:200, rabbit polyclonal; Stressgen, San Diego, CA), followed by the corresponding fluorescent secondary antibodies. For collagen IV staining (anti-collagen IV [1:200, rabbit polyclonal; Novotec, Lyon, France), cells were fixed with cold methanol for 5 min and blocked with 10% goat serum for 20 min after three washes with PBS before the incubation with the antibody. Nuclear staining was performed with DAPI (4′,6′-diamidino-2-phenylindole) (Sigma). Cell fluorescence was visualized by microscopy with a Nikon Eclipse T2000U (Nikon, Amstelveen, The Netherlands) or an LSM710 laser confocal microscope (Carl Zeiss, Overkochen, Germany).

Mammalian expression of recombinant LOXL4-GFP isoforms.

To study subcellular localization and potential extracellular secretion of LOXL4, we generated constructs for the overexpression of a fusion protein of bovine LOXL4 and green fluorescent protein (LOXL4-GFP). For that purpose, LOXL4 cDNA was cloned into pEGFP-N1 vector (Life Technologies) to yield full-length LOXL4-GFP. Deletion mutants from this construct lacking the scavenger receptor cysteine-rich (SRCR) domains and the C-terminal catalytic region were generated by site-directed mutagenesis (see oligonucleotide sequences in Table S8 in the supplemental material). These constructs were used for transient transfection in primary aortic endothelial cells as described above. To generate stable transfectants of HEK 293 cells overexpressing LOXL4-GFP isoforms, LOXL4-GFP fragments were cloned into the pcDNA5/FRT/TO vector (Life Technologies) to give pcDNA5/FRT/TO/LOXL4-GFP constructs. These constructs were then cotransfected with the Flp recombinase expression plasmid pOG44 into the Flp-In T-Rex 293 cell line (Life Technologies). Cells of this line stably express the tetracycline (Tet) repressor and contain a single integrated Flp recombination target (FRT) site. Flp recombinase expression from the pOG44 vector mediates insertion of pCDNA5/FRT/TO-based constructs into the genome at the integrated FRT site through site-specific DNA recombination. After 48 h, cells were selected with hygromycin B, and clones appeared after 10 to 15 days. Isogenic pooled clones were expanded and used for experiments involving induction by the Tet analog doxycycline (Dox). Expression of LOXL4-GFP chimeras upon Dox incubation was checked by Western blotting using anti-GFP antibodies as described above.

Detection of LOXL4 in the extracellular medium.

The release of LOXL4 to the extracellular medium was investigated in HEK 293 cells overexpressing LOXL4-GFP isoforms. Cells were seeded on 100-mm culture dishes. When 90% confluence was reached, they were incubated with 5 ml of serum-free medium containing doxycycline (1 μg/ml) for 48 h. Cell supernatants were collected, centrifuged at 100,000 × g for 1 h at 4°C, and then concentrated down to 100 μl using Amicon ultracentrifugal filters (Ultracel-10K; Millipore, Billerica, MA) and further immunoprecipitated by incubation overnight at 4°C with anti-GFP antibody. After addition of 50 μl of SoftLink soft release avidin resin (Promega), samples were incubated for 1 h at 4°C. Beads were then washed five times with buffer and retained proteins eluted out from the beads by adding Laemmli buffer. Supernatants were subjected to SDS-PAGE, fractioned proteins transferred to nitrocellulose membranes, and GFP-immunoreactive bands detected by Western blotting as described above.

In vitro collagen remodeling and lysyl oxidase activity assays.

Conditioned medium from primary bovine aortic endothelial cells incubated with or without TGF-β1 or infected with Adv-LOXL4 or control Adv-Null was concentrated by means of Amicon ultracentrifugal devices as described above. Type I collagen solution (2 mg/ml) was mixed with concentrated medium and incubated at 37°C for 30 min. Additional medium was added on top of the gelled collagen for further incubation at 37°C for 16 h. The gel was analyzed under confocal reflection microscopy in an LSM710 laser confocal microscope (Carl Zeiss) as previously described (23).

Cell supernatants were also assayed for lysyl oxidase activity using diaminopentane and Amplex red as previously described (24). In a typical assay, samples of 100 μl of concentrated medium were mixed with the reaction mixture containing 10 mM diaminopentane, 10 μM Amplex red, and 1 U horseradish peroxidase in 1.2 M urea–0.05 M sodium borate (pH 8.2) in the presence or absence of 0.5 mM β-aminopropionitrile (BAPN), an active-site inhibitor of lysyl oxidases. H₂O₂ release was measured every 15 min for a total period of 24 h at excitation and emission wavelengths of 525 and 580 to 640 nm, respectively, on a Glomax multidetection system (Promega). Enzyme activities were expressed as fluorescence values, corrected for background levels of H₂O₂ release determined in the reaction mixture supplemented with BAPN. Recombinant human LOXL2, used as a positive control for lysyl oxidase activity, was purchased from R&D Systems.

Statistical analysis.

Experimental data were analyzed using the unpaired Student t test in the case of normal distribution of data or using nonparametric tests as appropriate. The P values obtained are indicated in the figure legends when statistically significant (P < 0.05).

Microarray data accession number.

The microarray data are publicly available at the National Center for Biotechnology Information Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE42932.

TGF-β1 strongly induces LOXL4 expression in aortic endothelial cells through Smad and AP-1.

In order to gain insight about the potential ECM-related TGF-β1 target genes in the cardiovascular system, we profiled mRNAs from bovine aortic endothelial cells under basal conditions and incubated with TGF-β1 for 24 h. Comparing with control cells, it was found that TGF-β1 exerted a profound effect on global expression, with more than 500 genes upregulated under our experimental conditions (GEO accession number GSE42932). A significant number of genes positioned at the top of the list can be classified as ECM-related targets, including several collagen isoforms as well as collagen-modifying enzymes (Fig. 1A). A candidate TGF-β1 target gene, that for lysyl oxidase-like isoform 4 (LOXL4), attracted our attention as it showed the strongest induction, in addition to its relevance for ECM synthesis and remodeling. Analysis of the ability of TGF-β1 to modulate the expression of lysyl oxidase family members revealed that only LOXL4 and, to a lesser extent, the canonical LOX were upregulated in this cell model (Fig. 1B). Careful examination of the time course response of the LOXL4 gene to TGF-β1 showed that mRNA and protein begin to accumulate at as early as 2 h of incubation with the cytokine, indicating that LOXL4 might be a direct target for the action of TGF-β1 in the endothelium (Fig. 1C and D).

To elucidate the transcriptional mechanism for the control of LOXL4 expression by TGF-β1, we isolated a genomic fragment of the promoter of the bovine gene. Conflicting reports exist on the position of the transcriptional start site and therefore on the precise location of the promoter in the human gene (25, 26). Additionally, experimental evidence about the site(s) for the initiation of transcription in the bovine model has not been reported. Therefore, we first endeavored to identify the transcription initiation sites in the bovine LOXL4 gene by 5′ rapid amplification of cDNA ends (5′ RACE). A major 5′ RACE product was obtained and sequenced, yielding several start sites tightly clustering around the annotated site for the initiation of transcription located 121 nucleotides upstream of the translation initiation codon ATG (see Fig. S1 in the supplemental material). Based on this finding, a genomic fragment of approximately 4 kb (−3870/+87) including the 5′ flanking sequence and complete exon 1 was generated and cloned in front of a luciferase reporter gene to study the effect of TGF-β1. In a set of preliminary experiments, we tested this fragment and also a pair of truncated constructs of 2.2 and 0.3 kb. As shown in Fig. S2 in the supplemental material, while the activity of these truncated promoter fragments was not significantly modified by TGF-β1 incubation, the cytokine induced a potent increase in luciferase activity (5- to 6-fold at 24 h) in cells transfected with the longest, 4-kb construct. These results indicate that a promoter region between bp −3870 and −2266 contains the critical elements to mediate the transcriptional activation of the gene.

We screened the bp −3870/−2266 promoter region for putative binding sites for Smad and AP-1 transcription factors, both of which have been described to be important for the effect of TGF-β1 on gene expression (12, 27, 28). An upstream AP-1 consensus site (TGAGTCA) and several elements identical or similar to the Smad consensus sequence (CAGAC), either alone or in a palindromic form, were identified by in silico analysis (see Fig. S3 in the supplemental material). Based on the location of these putative regulatory elements, we generated a set of 5′ deletion luciferase constructs and tested their inducibility by TGF-β1. As shown in Fig. 2A, the TGF-β1 response observed with the −3870/+87 fragment was eliminated as soon as a short segment of 113 bp containing putative AP-1 and Smad elements was removed. In fact, this fragment can be considered an enhancer, as it is able to confer TGF-β1 responsiveness to a core promoter in a manner that is independent of its orientation and distance relative to the RNA start site, as shown in experiments with constructs generated in the context of its cognate proximal promoter and of the viral SV40 promoter (Fig. 2B and C). In order to assess the relevance of the AP-1 and Smad elements identified in this enhancer region, we generated full promoter constructs (bp −3870/+87) with specific mutations in these sites (AP-1, SmadA, and SmadB) and tested their activity upon TGF-β1 stimulation. As shown also in Fig. 2D, the disruption of the AP-1 element and the downstream Smad site, the consensus SmadB, either in individually or combined in the same construct, caused a significant reduction in their ability to respond to TGF-β1. In contrast, the specific mutation of the upstream Smad-like site (SmadA) did not show any significant effect.

Multiple-sequence alignment analysis is a powerful approach for identifying functional regulatory elements, such as exons or enhancers. We used the Mulan multiple-alignment server (http://mulan.dcode.org) to analyze the promoter regions of several mammalian orthologs of LOXL4 in order to evaluate the relevance of the DNA elements identified in the present study and also to screen for an additional control region(s) (19). As shown in Fig. 3A, at least three well-conserved regions were identified across these genomes, namely, a short stretch at the proximal promoter, a broader region in the central portion of the promoter fragment used for comparison, and finally, an upstream region partially overlapping the enhancer involved in the TGF-β1 induction. Surprisingly, rodent genomes differed from the rest of mammalian genomes compared in this study, as they did not show significant homology in the upstream enhancer. Careful examination of the sequence of this DNA segment revealed that the consensus AP-1 site is not present in the equivalent regions from rat and mouse (Fig. 3B), an observation that predicts that the LOXL4 gene may not respond to TGF-β1 in these mammalian species. To test this hypothesis, the corresponding fragments of human and mouse LOXL4 promoters were isolated and fused to a luciferase reporter for analysis of their TGF-β-dependent transcriptional activity in BAEC and MAEC. As shown also in Fig. 3C, human and bovine LOXL4 promoters displayed significant TGF-β1 inducibility in both bovine and mouse cells. In contrast, the construct containing the mouse LOXL4 promoter was unresponsive to TGF-β1 in both bovine and mouse endothelial cells. Accordingly, LOXL4 gene expression was not altered by TGF-β1 in MAEC, whereas was highly upregulated in BAEC (Fig. 3D). These results are consistent with the analysis of conserved regions, and altogether the results underscore the contribution of the AP-1 and Smad elements within this enhancer region to the induction of LOXL4 expression by TGF-β1.

The well-conserved proximal promoter of the LOXL4 gene lacks a TATA box, but it contains a GC-rich region that may correspond to a potential CpG island. Using the CPGPlot program (http://www.ebi.ac.uk), analysis of the proximal LOXL4 promoter from bp −307 to +87 revealed a CpG island that lies between bp −260 and +29 (see Fig. S4A in the supplemental material). Sp1 is a ubiquitous transcription factor that binds GC-rich sequences that are present in many CpG islands, and it is reported to prevent their methylation (29). A consensus Sp1 site (GGGGGCGGGG) has been also identified in the CpG island of the LOXL4 proximal promoter (see Fig. S4A in the supplemental material). Site-directed mutagenesis of this putative Sp1 binding site strongly reduced the basal activity of LOXL4 promoter without significantly affecting the induction by TGF-β1 (see Fig. S4B in the supplemental material). Accordingly, the overexpression of Sp1 family members Sp1 and Sp3 increased both basal and TGF-β1-stimulated promoter activity (see Fig. S4C in the supplemental material), indicating that the proximal promoter importantly contributes to the basal transcription machinery but seems to be dispensable in the induction by TGF-β.

Smad and JunB/Fra2 transcription factors contribute to TGF-β1 induction of the LOXL4 gene.

TGF-β regulates gene expression through Smad-dependent and -independent pathways (30). To investigate the role of Smad family members in the mechanism of LOXL4 upregulation by TGF-β1, the activating Smad3 and the inhibitory Smad7 were overexpressed in BAEC and their effect on LOXL4 promoter activity analyzed in luciferase assays. Figure 4A shows that Smad3 potentiated, whereas Smad7 significantly reduced, both basal and TGF-β1-induced promoter activity. The implication of Smad3 was further confirmed by experiments with mouse embryonic fibroblasts (MEF) isolated from wild-type and Smad3 knockout embryos. As shown also in Fig. 4B, TGF-β1 activated the LOXL4 promoter in transfected wild-type MEF, whereas it showed no effect on Smad3 knockout cells. TGF-β1-mediated induction of the LOXL4 promoter was recovered in these cells upon reintroduction of Smad3.

On the other hand, the implication of AP-1 members was investigated by adenovirus-mediated overexpression of a dominant negative form of AP-1 (TAM67) that binds to DNA but is not able to transactivate (31). TAM67 dose dependently inhibited the induction of LOXL4 expression by TGF-β1 at the levels of both mRNA and promoter activity (Fig. 4C and D).

EMSA and ChIP experiments were performed to detect and characterize protein-DNA complexes at the LOXL4 promoter. EMSA with nuclear extracts from cells under basal conditions showed an AP-1 binding activity that was significantly increased upon incubation with TGF-β1 (see Fig. S5A and B in the supplemental material). The binding to this complex was competed out by an excess of unlabeled AP-1 oligonucleotide but not by a mutated one. Smad proteins bind to Smad sites with low affinity, and this feature makes their detection by EMSA difficult (32). ChIP using a specific Smad2/3 antibody successfully revealed the recruitment of Smad factors to the enhancer region within the LOXL4 promoter in a TGF-β1-dependent manner (see Fig. S5C in the supplemental material). Smad binding was further confirmed by EMSA with the use of recombinant proteins. Recombinant Smad4 and, to a lesser extent, Smad3 displayed strong binding activity to a Smad probe (see Fig. S5D in the supplemental material). In agreement with site-specific mutagenesis experiments, a wild-type Smad probe or a probe with a specific mutation in the Smad-like site, SmadA, efficiently competed out the binding observed to recombinant Smad4 compared with a probe mutated in the consensus SmadB (see Fig. S5E and F in the supplemental material).

In an effort to identify the proteins involved in the regulation of LOXL4 expression by TGF-β1, a DNA pulldown assay was employed to isolate proteins that specifically and differentially bind to a DNA sequence derived from the upstream enhancer. 5′-biotinylated wild-type and mutated AP-1/SmadB DNA fragments corresponding to positions −3870/−3756 were exposed to nuclear extracts from BAEC incubated with or without TGF-β1, and DNA-bound proteins were separated by SDS-PAGE and visualized by Coomassie blue staining (see Fig. S6A in the supplemental material). A protein band of approximately 35 kDa was significantly enriched in samples obtained from TGF-β1-treated cells and incubated with the wild-type oligonucleotide compared with those incubated with the mutant DNA probe or isolated from control cells. SEQUEST searches of LC-MS analyses of tryptic peptides from the excised band showed that one of the proteins identified corresponded to a member of the AP-1 family, FOS-like antigen 2 (FOSL2) or Fra2 (see Fig. S6B and C in the supplemental material).

Fra2 has recently evolved as an extremely interesting factor in the context of vascular ECM remodeling and fibrosis (33–36). In order to confirm the implication of Fra2 in the upregulation of LOXL4 expression, we performed gain- and loss-of-function experiments. Adenovirus-mediated overexpression of Fra2 potentiated TGF-β1-induced LOXL4 expression at the level of promoter activity and protein (Fig. 5A). Fra2 overexpression also associated with its binding to the enhancer region within the LOXL4 promoter as assessed by ChIP (validation of overexpression by adenoviral Fra2 and ChIP experiments is shown in Fig. S7A and B in the supplemental material). Conversely, specific silencing of endogenous Fra2 by siRNA technology significantly diminished the level of induction of LOXL4 expression by TGF-β1 (Fig. 5B) (validation of silencing by siRNA Fra2 is shown in Fig. S7C in the supplemental material). Proteins belonging to the AP-1 family bind DNA as obligate homo- or heterodimers. While Jun members can dimerize with each other and with Fos proteins, the members of the Fos subfamily, c-Fos, FosB, and Fra1 and Fra2, cannot form homodimers (37). To identify the Jun partner that dimerizes with Fra2 in TGF-β1-mediated activation of LOXL4 expression, we took advantage of the existence of tethered AP-1 dimers, including JunB/Fra2, JunD/Fra2, and c-Jun/Fra2, with AP-1 molecules joined via a flexible polypeptide tether to force specific pairing (18). As shown in Fig. 5C, overexpression of forced JunB/Fra2 dimer resulted in a potentiation of TGF-β1-induced promoter activity, while Fra2/JunD and c-Jun/Fra2 did not show any significant effect, compared with empty vector (validation of the overexpression of AP-1 dimers by anti-Flag Western blotting is shown in Fig. S7D in the supplemental material). The involvement of JunB was further confirmed by siRNA experiments. Specific downregulation of JunB expression resulted in a significant reduction of the effect of TGF-β1 on LOXL4 promoter activity, whereas the silencing of JunD and c-Jun did not have a significant effect (Fig. 5D) (validation of silencing of Jun proteins by siRNA is shown in Fig. S7E in the supplemental material).

Experiments involving Fra2 were done with an anti-Fra2 antibody raised against mouse Fra2 protein, which unfortunately did not react with the endogenous Fra2 from bovine aortic endothelial cells but very efficiently recognized the overexpressed mouse isoform. Using this antibody in overexpression experiments, we noticed that the reactivity of Fra2 bands increased significantly upon TGF-β1 stimulation, an observation that was accompanied by the detection of higher-molecular-weight (higher-MW) immunoreactive bands (Fig. 6A; see Fig. S7A in the supplemental material). Immunofluorescence experiments performed with the same protocol showed specific nuclear Fra2 staining that strongly increased after TGF-β1 incubation (see Fig. S7B in the supplemental material). Protein phosphorylation may be compatible with the observed changes in immunoreactivity and MW, and as such it is a posttranslational modification previously reported for Fra2 (38, 39). To demonstrate the potential use of this antibody to reveal phosphorylated Fra2 and therefore the ability of TGF-β1 to induce the phosphorylation of this factor, cell extracts were incubated with calf intestine alkaline phosphatase (CIAP) and λ-phosphatase (λ-PPase), conditions shown to dephosphorylate serine, threonine, and tyrosine residues in proteins (40). As shown also in Fig. 6A, dephosphorylation of cell extracts diminished significantly the increase in immunoreactivity, an effect particularly evident with λ-phosphatase, and, more interestingly, eliminated the higher-MW Fra2 bands, which presumably correspond to heavily phosphorylated forms of the protein. The noncanonical TGF-β signaling pathway has been associated with the activation of members of the mitogen-activated protein kinase (MAPK) family (30). Screening of the sequences of the Fra2 protein among different species, including mouse, rat, human, bovine, chicken, and zebrafish, revealed the presence of two MAPK phosphorylation consensus sites (P-X-X-S/T-P or P-X-S/T-P), in particular, two threonine residues in the C-terminal domain (T263 and T274 in the mouse Fra2), highly conserved in most of Fra2 orthologs and in human and mouse c-Fos proteins (Fig. 6B). Flag-tagged Fra2 constructs with inactivating mutations of threonine to nonphosphorylatable alanine were generated, and the pattern of Flag-tagged Fra2 expression upon TGF-β1 stimulation was analyzed by SDS-PAGE and anti-Flag Western blotting. As shown in Fig. 6C, the T274A substitution resulted in a clear decrease in the detection of high-MW, hyperphosphorylated bands that otherwise were observed with the wild type and the T263A mutant. The construct including both mutations behaved essentially like the T274A mutant. As an alternative approach, glutamic acid substitution mutagenesis, a modification mimicking the structure of a phosphorylated threonine residue, was also introduced. Figure 6C shows that the T274E and, to a lesser extent, T263E substitutions shifted the detected immunoreactive bands into high-MW forms, an effect more clearly evident in the double mutant. These results indicate that phosphorylation of Fra2 in these residues, particularly in the highly conserved T274, may play an important role in the action of TGF-β1. In order to test the ability of the Fra2 mutants to induce TGF-β1-induced LOXL4 expression, we cotransfected wild-type, T263/274A (inactivating), and T263/274E (phosphomimetic) Fra2 plasmids together with the LOXL4 promoter luciferase construct. As shown also in Fig. 6D, the overexpression of wild-type or T263/274E Fra2 strongly potentiated TGF-β1-induced LOXL4 promoter activity, whereas the mutant T263/274A was without any effect. We have also investigated the mechanism(s) by which Fra2 phosphorylation gives rise to a potentiation of LOXL4 expression. First, we analyzed whether the interaction of JunB/Fra2 complex with DNA is affected by Fra2 phosphorylation. In vitro-synthesized JunB/Fra2 forced dimers, either wild type, T263/274A, or T263/274E, showed very similar DNA binding activity as assessed by EMSA (see Fig. S8A and B in the supplemental material). We next investigated whether Fra2 phosphorylation facilitates JunB/Fra2 dimerization using GST-Fra2 forms in pulldown experiments. As shown in Fig. S8C in the supplemental material (input), JunB expression was induced upon TGF-β1 stimulation in BAEC, a result that was also observed by microarray analysis (GEO accession number GSE42932). GST-Fra2 beads were incubated with total extracts from BAEC incubated under basal conditions or with TGF-β1 for 24 h, extensively washed, and loaded onto SDS-polyacrylamide gels for Western blot analysis of JunB. GST-Fra2, either wild type, T263/274A, or T263/274E, pulled down endogenous JunB protein with similar capacities. These results indicate that Fra2 phosphorylation does not significantly modify DNA binding activity or dimer formation of AP-1 complexes.

TGF-β1-dependent Fra2 phosphorylation likely involves the activation of a MAPK. We next analyzed the effects of specific pharmacological inhibitors of p38, ERK, and JNK MAPKs on the upregulation of LOXL4 promoter activity by TGF-β1. As shown in Fig. 7A, the blockade of the ERK pathway strongly attenuated the action of TGF-β1, whereas the inhibitors of p38 and JNK did not show any significant effect. Dose-dependent analysis of the effect of the ERK inhibitor, CI-1040, showed that this compound diminished TGF-β1-induced LOXL4 mRNA and promoter activity at concentrations reported to specifically inhibit ERK phosphorylation (100 to 500 nM) (Fig. 7B and C) (41). We also studied the time course of the effect of TGF-β1 on ERK and Fra2 phosphorylation by Western blotting. TGF-β1 very rapidly induced an increase in the phosphorylation status of ERK, which was followed by the detection of hyperphosphorylated forms of Fra2 (Fig. 7D). Both actions of TGF-β1 were strongly abolished by preincubation of the ERK inhibitor. These results indicate that TGF-β1 induction of LOXL4 expression requires the activation of the ERK pathway and subsequent ERK-dependent phosphorylation of Fra2.

LOXL4 contributes to the matrix remodeling action of TGF-β1 in aortic endothelial cells.

LOXL4 is the most recently identified member of the lysyl oxidase family. Therefore, little information is available on LOXL4 subcellular localization and function, particularly in the cardiovascular system (42–44). The C-terminal region of LOXL4 protein contains a catalytic domain highly homologous to other lysyl oxidases. On the other hand, the N-terminal region has four scavenger receptor cysteine-rich (SRCR) domains that are conserved only in the LOXL2 and LOXL3 isoforms. To gain insight into the subcellular localization of LOXL4 protein and to individually analyze the contributions of its different domains, we generated a LOXL4-GFP fusion protein and deletion mutants thereof lacking the SRCR domains and the C-terminal catalytic region (Fig. 8A). Cell transfection with the full-length LOXL4-GFP construct showed that the fluorescent signal was perinuclear, spatially overlapping with the immunocytochemical detection of calnexin, a typical endoplasmic reticulum (ER) marker (Fig. 8B). This subcellular distribution is roughly maintained upon removal of the SRCR domains, but it is heavily affected by deletion of the C-terminal catalytic region, whose fluorescence was homogeneously diffuse throughout the cell, much in the same way as GFP only (see Fig. S9 in the supplemental material). The localization of the full-length protein in the ER compartment suggests that LOXL4 may transit through the ER as a part of the secretory pathway, yielding its release into the extracellular medium. To evaluate this hypothesis, we generated stable transfectants of human embryonic kidney 293 (HEK 293) cells expressing the aforementioned set of LOXL4-GFP constructs. Analysis by anti-GFP Western blotting showed that these isogenic clones of HEK 293 cells expressed the corresponding LOXL4-GFP proteins (Fig. 8C, Cells). The extracellular medium accumulated for 48 h was concentrated and analyzed by Western blotting. As shown also in Fig. 8C (Media), full-length LOXL4-GFP was clearly detectable in the extracellular medium, indicating that this protein is indeed extracellularly secreted. The presence of multiples copies of the SCRC domains seems to be essential for the extracellular release, as elimination of more than one of these repeats impaired the accumulation of the protein in the extracellular medium. LOXL4-GFP ΔC-t, a deletion construct with intact SRCR domains, was again recovered from the extracellular medium. Similar experiments performed with BAEC incubated with TGF-β1 or infected with a LOXL4-overexpressing adenovirus also showed the presence of extracellularly secreted LOXL4 protein (Fig. 8D).

To explore the contribution of LOXL4 to collagen cross-linking, we incubated soluble type I collagen with conditioned medium generated by aortic endothelial cells incubated for 48 h in the presence or absence of TGF-β1 and with empty vector or LOXL4-overexpressing adenoviruses, and fibrillar collagen was imaged by confocal reflection microscopy. As shown in Fig. 9A, based on fiber size, fibrillar collagen formation was strongly enhanced in samples incubated with medium from cells exposed to TGF-β1 or infected with LOXL4 adenovirus. This matrix-remodeling capacity was consistent with the classical determination of lysyl oxidase activity by detection of H₂O₂ using Amplex red as assessed in both endothelial cells and HEK 293 cells overexpressing LOXL4 (Fig. 9B). The ability of LOXL4 to regulate the deposition and assembly of collagen IV was analyzed by immunofluorescence microscopy. As shown in Fig. 9C, cells overexpressing LOXL4 showed increased collagen IV network assembly compared to control cells. These data suggest that LOXL4 may play a role in collagen IV assembly in the subendothelial basement membrane, therefore contributing significantly to vascular extracellular microenvironment.