Identification of Genes Selectively Regulated by IFNs in Endothelial Cells

Cell lines and in vitro culture

HUVEC were isolated from freshly obtained umbilical cord samples by collagenase digestion of the umbilical vein (19). Primary cultures were serially split in flasks coated with collagen type I (10 μg/cm²; Sigma-Aldrich) and maintained in TC 199 medium supplemented with 20% FCS (Invitrogen Life Technologies), human rβ-EC growth factor (0.1 ng/ml; Sigma-Aldrich), human recombinant basic fibroblast growth factor (10 ng/ml; Peprotech), and heparin (50 μg/ml). Human primary fibroblasts (HF) were obtained by mechanical disruption of the connective tissue of the umbilical cord and cultured in DMEM (Sigma-Aldrich) supplemented with 10% FCS and 2 mM l-glutamine. Primary cultures of human dermal microvascular endothelial cells (HMVEC) were purchased from Cascade Biologics and maintained in M131 medium supplemented with microvascular growth supplement (Cascade Biologics). Primary cells were used between the fourth and the seventh in vitro passage. Cultures of human macrovascular EC obtained from the renal artery (HMVAR) were donated by Dr. A. Caruso (University of Brescia, Brescia, Italy) and were maintained in M131 medium. The murine 1G11 lung endothelial cell line and dermal microvascular EC of the SIEC cell line were obtained from Dr. A. Vecchi (Mario Negri Institute, Milan, Italy) and were cultured in DMEM supplemented with 20% FCS, 1 mM l-glutamine, 1% nonessential amino acids, 1 mM sodium pyruvate, EC growth supplement (Sigma-Aldrich), and heparin (100 μg/ml). The human breast cancer cell line MCF7, obtained from American Type Culture Collection, was grown in DMEM supplemented with 10% FCS and 2 mM l-glutamine; IFN-α-producing MCF7 cells were obtained by the transduction of parental MCF7 cells with a murine IFN-α-encoding retroviral vector as reported elsewhere (4).

Before processing as detailed below, human EC and HF were incubated at 37°C for 5 h in a humidified atmosphere of 5% CO₂ in air alone or in the presence of rIFN-α, rIFN-β (both purchased from Schering-Plough), or rIFN-γ (purchased from Peprotech) at 1000 IU/ml; SIEC and 1G11 cells were incubated with 1000 IU/ml murine rIFN-α (PBL Laboratories).

Preparation of RNA and cRNA

Total RNA was isolated from EC and HF cultured as above using the RNeasy Mini kit (Qiagen) according to the manufacturer’s instructions. cDNA synthesis was performed using T7-(dT)₂₄ oligonucleotide primers and the Custom SuperScript double-stranded cDNA synthesis kit (Invitrogen Life Technologies). Double-stranded cDNA was extracted with phenol-chloroform-isoamyl alcohol (25:24:1), precipitated with ethanol, and used to prepare cRNA using the BioArray high-yield RNA transcription kit (Affymetrix) according to manufacturer’s instructions. cRNA was purified using the RNeasy Mini kit as described above, controlled by agarose gel electrophoresis and RNA 6000 Pico assay (Agilent), and subjected to fragmentation for 35 min at 94°C in fragmentation buffer (40 mM Tris-acetate (pH 8.1), 100 mM CH₃COOH, and 30 mM Mg(CH₃COO)₂·4H₂O).

GeneChip microarray analysis and data normalization

Labeled cRNA was used for screening of GeneChip human genome U133A arrays (Affymetrix). The experiment consisted of three biological replicates for control and IFN-treated HUVEC and two biological replicates for all the other conditions; in some cases, technical replicates were also performed. Each biological replicate consisted of four independent experiments with different donors that were pooled before hybridization. Data concerning the effects of IFN-β are based on a single biological replicate with two technical replicates. Microarray analysis of HMVEC was performed on cDNA obtained from IFN-treated or control HMVEC cultures from two different donors. Hybridization and scanning was conducted on the Affymetrix platform. Data were normalized following the GeneChip robust multiarray average (GCRMA) procedure (20) of Bioconductor 1.4 (21) (www.bioconductor.org). Statistically significant expression changes were determined using permutation tests (Significance Analysis of Microarrays (SAM); www-stat.stanford.edu/∼tibs/SAM/) (22). Genes regulated at least 2-fold in comparison to untreated controls were considered. The delta value was set to return a median false significant number of <1. The q values were between 0.012 and 0.0031 for the different gene lists created. Annotations were obtained through the Database for Annotation, Visualization, and Integrated Discovery (DAVID; http://david.niaid.nih.gov/david/beta/index.htm) (23). The microarray data discussed in this publication have been deposited in the Gene Expression Omnibus (GEO; www.ncbi.nlm.nih.gov/geo/info/linking.html) of the National Center for Biotechnology Information and are accessible through GEO series accession number GSE3920.

RT-PCR end-point analysis

The expressions of murine CXCL10, CXCL11, TRAIL, and β-actin were analyzed by qualitative PCR. Briefly, total RNA was extracted from in vitro cultured cells and homogenized tumors as described above, and RNA was reverse transcribed using 12.5 U of avian myeloblastosis virus reverse transcriptase and 0.5 μg of (oligo)dT as primer for 45 min at 42°C and 5 min at 99°C. The following set of primers (Sigma-Genosys) was used: CXCL10 forward, 5′-ACCATGAACCCAAGTGCTGCCGTC-3′; CXCL10 reverse, 5′-GCTTCACTCCAGTTAAGGAGCCCT-3′; CXCL11 forward, 5′-AGGAAGGTCACACCATAGC-3′; CXCL11 reverse, 5′-CAGGTTCCTGGCACAGAGTT-3′; TRAIL forward, 5′-TCACCAACGAGATGAAGCAG-3′; TRAIL reverse, 5′-GCCTAAGGTCTTTCCATCC-3′; murine β-actin forward, 5′-CCTTCCTGGGCATGGAGTCCTG-3′; and murine β-actin reverse 5′-GGAGCAATGATCTTGATCTTC-3′.

In each amplification, cDNA from a cell line known to express a given factor was included as the positive control. Water was used as a negative control. PCR was performed for 30 cycles at 95°C for 1 min, 58°C for 1 min, and 72°C for 1 min in a total volume of 50 μl using GoldTaq and standard PCR buffers (all from Applied Biosystems). The amplified products were resolved by 1.5% agarose gel electrophoresis and stained with ethidium bromide. No bands were obtained by PCR analysis of RNA samples amplified without RT.

Quantitative RT-PCR

Expression data validation was performed by quantitative real-time RT-PCR using the RNA extracted from IFN- or mock-treated cells and reverse transcribed as described above with oligo(dT) primers in a 20-μl final volume. All primers for the genes tested were designed using Primer3 software (24) with a T_m optimum of ∼60°C and a product length of 100–150 nt (Table I⇓). Real-time PCR was performed on an I-Cycler (Bio-Rad) using iQ Supermix (Bio-Rad) supplemented with 10 nM fluorescein (Bio-Rad), 0.1× SYBR Green I (Sigma-Aldrich), 2.5 μl of cDNA (5× diluted), and 0.3 μM sense and antisense primers in a final reaction volume of 25 μl. After an initial denaturation step of 3 min during which the well factor was measured, 50 cycles of 15 s at 95°C followed by 30 s at 60°C were performed. Fluorescence was measured during the annealing step in each cycle. After amplification, melt curves with 80 steps of 15 s and 0.5°C increase were performed to monitor amplicon identity. Amplification efficiency was assessed for all primer sets used in separate reactions, and primers with efficiencies >94% were used. Expression data were normalized on GAPDH and RNA polymerase II gene expression data obtained in parallel. Relative expression values with SE values and statistical comparisons (unpaired two-tailed t test) were obtained using QGene software (25). Expression changes were calculated from the mean value of normalizations obtained using the above genes as references.

Quantification of CXCL10, CXCL11, and TRAIL protein levels

ELISA (R&D Systems) were used to quantify the levels of CXCL10 and CXCL11 in the supernatants of EC and HF incubated for 48 h in the presence of IFN-α or IFN-γ as described above. In the kinetics experiments, HUVEC were treated with either IFN-α or IFN-γ for 5 h and washed and cultivated for various intervals (range, 18–72 h) in complete medium without IFNs. The range of sensitivity of the assays was 7.8–500 pg/ml and 62.5–4000 pg/ml for CXCL10 and CXCL11, respectively. Concentrations of TRAIL in cell-free supernatants and cell lysates were measured by ELISA (Diaclone Research), according to manufacturer’s instructions.

In vivo tumor growth

Parental or IFN-α-producing MCF7 cells were injected s.c. into the flanks of female SCID mice (5 × 10⁵ cells/mouse) obtained from Charles River Breeding Laboratories. Tumor size was measured daily with calipers; when 8control tumors were ∼300 mm³, the animals were sacrificed and the tumors collected. The tumors were either snap frozen or fixed in formalin, paraffin-embedded, sectioned, and stained with H&E for histological analysis. Procedures involving animals and their care conformed with institutional guidelines that comply with national and international laws and policies (European Economic Community Council Directive 86/609, OJ L 358, 1, Dec. 12, 1987; National Institutes of Health Guide for the Care and Use of Laboratory Animals, National Institutes of Health Publication 85-23, 1985).

Histology and immunohistochemistry

Four-micrometer thick sections of MCF7 tumors were rehydrated and either stained with H&E or processed for immunohistochemistry by standard procedures. The sections were labeled with a goat anti-CXCL11 Ab (Santa Cruz Biotechnology), and immunostaining was performed using the avidin-biotin-peroxidase complex technique and 3,3′-diaminobenzidine as a chromogen (Vector Laboratories). Sections were finally lightly counterstained with Mayer’s hematoxylin. Parallel negative controls obtained by replacing primary Abs with PBS were run in all cases.

Western blot analysis of guanylate-binding protein (GBP)-1 expression

GBP-1 expression in tumor and cell lysates was determined by Western blotting according to standard protocols. Briefly, tumor samples were run on 10% polyacrylamide gels; separated proteins were then blotted for 2 h at 400 mA onto a nitrocellulose membrane. The membrane was then saturated with PBS plus 1% nonfat dry milk (Sigma-Aldrich) as blocking buffer for 3 h at room temperature and then stored at −20°C until use. Immunoprobing was performed with rabbit polyclonal Abs against human GBP-1 used at 0.1 μg/ml (Santa Cruz Biotechnology), followed by hybridization with a 1/5000 diluted anti-rabbit HRP-conjugated Ab (Amersham Biosciences). Ags were identified by luminescent visualization using the SuperSignal kit (Pierce). Tubulin was used to normalize differences among the different lanes.

Migration assay

To test the inhibitory activity of IFNs, HMVEC were cultured for 48 h with complete EBM-2 medium supplemented with either IFN-α or IFN-γ (1000 U/ml) before the migration assay. In some wells, CXCL10 (50 nM; Peprotech) was added to the HMVEC as a positive control. To inhibit CXCR-3-mediated effects on EC migration, a neutralizing anti-human CXCR3 mAb was used (IgG1, clone 49801.111, final concentration 10 μg/ml; R&D Systems). Isotype-matched (IgG1) controls were also included in each experiment to check for the specificity of the effects of the anti-CXCR3 mAb.

Chemotaxis assays were performed in Boyden chambers using polyvinylpyrrolidone-free polycarbonate filters with 12-μm pores for HMVEC cells coated with 5 μg of collagen IV. HMVEC (1.2 × 10⁵) pretreated with IFNs, CXCL10, or appropriate Abs were placed in the upper compartment in a serum-free medium containing 0.1% BSA; the lower compartment of the chamber was filled with either conditioned medium from the Kaposi’s sarcoma-derived KS-IMM cell line (26) as a chemoattractant or SFM as a negative control. The chambers were then incubated at 37°C in 5% CO₂ for 3 h. Cells remaining on the upper surface of the filter were then mechanically removed, and 5–10 random fields of cells that had migrated to the lower surface of each filter were stained and counted. Assays were performed in triplicate and repeated at least three times.

Angiogenesis assay

The Matrigel sponge model of angiogenesis was used as previously described (26). Angiogenic growth factors, including vascular endothelial growth factor (100 ng/ml), TNF-α (2 ng/ml), and heparin (24–26 U/ml) (VTH; all from PeproTech), were added to liquid Matrigel at 4°C. In some experiments, IFN-γ (1000 U/ml), anti-mouse-CXCR3 (15 μg/ml), or control isotype Ab (15 μg/ml) (R&D Systems) were added to the Matrigel. This suspension (final volume of 600 μl) was slowly injected s.c. into the flanks of 6-wk-old male C57BL/6 mice, where the Matrigel polymerizes to form a solid gel. After 4 days, gels were recovered, weighed, minced, and diluted in water and the hemoglobin content was measured with a Drabkin reagent kit (Sigma-Aldrich). The final hemoglobin concentration was calculated from a standard calibration curve.

Genes induced by IFNs in HUVEC

Analysis of the gene expression profile in HUVEC after 5 h of activation by IFN-α indicated that numerous genes were induced and only a few were down-regulated; 356 probe sets were >2-fold up-regulated and 23 probe sets were >2-fold down-regulated. Two hundred forty-two of the up-regulated and none of the down-regulated probe sets were found to be statistically significant after testing using Significance Analysis of Microarrays (delta value set to return a median false significant number <1, q = 0.003; see also supplemental Table I).⁵ Of these probe sets, 137 corresponded to unique, characterized genes, 67 of which were present with two or more probe sets, and 38 corresponded to unknown genes.

We next analyzed gene ontology classes of the genes induced. The ratio of IFN-α-induced genes belonging to a given gene ontology class (list hits) and the total of annotated genes present in this list (list total) were compared with the ratio of expressed, annotated genes belonging to the same ontology (population hits) and the total of expressed, annotated genes present on the array (population total). Statistical significance was analyzed using the Expression Analysis Systematic Explorer (EASE) score as well as Bonferroni multiparameter posttesting. The most significant gene ontology classes corresponded to the biological processes of defense/immune/inflammatory responses and molecular functions compatible with their involvement in these processes, including cytokine and chemokine activity (Table II⇓). This is in general agreement with other reports on the classes of IFN-α-stimulated genes (12, 15, 16, 18).

Viperin, an IFN-inducible antiviral protein that is also directly induced by the CMV (27), and IFIT1, a member of the IFN-regulated IFI54/IFI56 family (28), were the highest responding genes in HUVEC, being induced 1362-fold and 722-fold, respectively, following IFN-α treatment (Table III⇓). Interestingly, transcription of two members of the ELR⁻ CXC chemokine family (29), CXCL10 and CXCL11, was also strongly induced by IFN-α in these cells; the relative probe sets showed 459/374-fold induction for CXCL11 and 355-fold induction for CXCL10 as compared with untreated controls. Table III⇓ reports the 50 most strongly IFN-α-induced genes in HUVEC (for a complete list of induced genes see supplemental Table I⁵).

IFN-β treatment induced changes in the transcriptome of HUVEC that were virtually identical with those obtained following IFN-α stimulation (data not shown); for this reason, no replicates and statistical evaluations for IFN-β were performed. In contrast, several genes were differentially up-regulated by class I and class II IFN in HUVEC. Forty-seven genes showed a >5-fold difference upon induction with the two types of IFN and 14 of them were induced >5-fold by IFN-α and <2-fold by IFN-γ (Table IV⇓). Both CXCL10 and CXCL11 were strongly up-regulated by IFN-γ also (591/550-fold and 408-fold, respectively). A few genes were up-regulated by IFN-γ but showed little response to class I IFNs; among these, we found CXCL9, which was increased by 228-fold following IFN-γ treatment. Due to high variability between independent experiments (69- to 387-fold induction), CXCL9 was not among the genes that passed statistical testing; notwithstanding, its preferential up-regulation by IFN-γ was confirmed by real-time PCR analysis for other EC types (see below).

Validation of expression data by real-time PCR

We then validated the microarray gene expression data by means of real-time PCR for a group of genes (Fig. 1⇓A). In addition to the most strongly up-regulated genes, including CXCL10 and CXCL11, we selected several other genes that were induced to a lesser extent (comprised in supplementary Table I), including some genes potentially involved in angiogenesis control such as TRAIL (also called TNFSF10) and GBP-1. Although some variation in the extent of regulation was observed, the data obtained with microarrays were substantially confirmed by real-time PCR analysis (Fig. 1⇓A). Only in the case of OASL and MX2 did we observe a dramatic discrepancy between microarray and real-time PCR data for IFN-γ-treated HUVEC (Fig. 1⇓A); however, if the ratio between the signals in IFN-α-treated and IFN-γ-treated cells was considered, the differential responsiveness was still confirmed (MX2: microarray IFN-α vs IFN-γ = 232-fold, RT-PCR 1115-fold; OASL: microarray IFN-α vs γ = 146/113-fold, RT-PCR 102-fold). All expression variations were statistically significant (unpaired t test, p < 0.05). Major quantitative differences between microarray and RT-PCR analyses occurred for genes that were not expressed in untreated HUVEC, because background adjustment by GeneChip robust multi-array average leaves unexpressed genes at an intensity above zero.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Real-time PCR validation of microarray data. A, Real-time PCR analysis was performed as described in Materials and Methods using primers listed in Table I⇑ and cDNA prepared from IFN-α- or IFN-γ-treated HUVEC. Results indicate the log of relative expression of the individual genes listed on the y-axis, as judged by microarray analysis (MA; gray columns) and real-time PCR (RT; filled columns). Real-time PCR data were normalized on GAPDH and RNA polymerase II expression levels; the mean value is reported as log 2. B, Real-time PCR analysis of CXCL10 and CXCL11 expression in HUVEC and HF incubated in the absence (gray columns) and the presence of IFN-α (open columns) or IFN-γ (filled columns).

EC-specific transcriptional changes induced by IFN-α

To explore to what extent the transcriptional changes detected in HUVEC following IFN treatment were lineage-specific, we compared the effect of IFN-α on the microarray expression profile of HUVEC and primary HF. Of note, HUVEC and HF were both derived from single umbilical cord samples and were thus representative of two different normal cell types from a same donor. IFN-α treatment produced marked transcriptional changes also in HF, with >2-fold statistically significant (q = 0.012) changes in the expression levels of 235 genes. The great majority (80%) of genes induced by IFN-α were the same in both cell types (<5-fold difference); however, some quantitative and/or qualitative differences were observed. A list of the genes differentially expressed in HUVEC and HF upon IFN-α treatment is shown in Table V⇓; in particular, CXCL11 was apparently not induced in HF by IFN-α, whereas CXCL10, although induced, showed a 18 times lower induction than in HUVEC. Real-time PCR analysis confirmed the EC-specific up-regulation of CXCL10 and CXCL11 following both IFN-α and IFN-γ treatment (Fig. 1⇑B).

Kinetics of gene expression in IFN-stimulated EC

To investigate the dynamics of the transcriptional changes induced by IFNs in EC, a time course analysis was performed. HUVEC were stimulated with 1000 IU/ml IFN-α or IFN-γ followed by the extraction of mRNA at different time points (0, 2, 5, 18, 24, 48, 72, and 96 h); in all cases, IFN-containing medium was removed after 5 h of stimulation. Expression of some of the most strongly up-regulated genes, including CXCL9–11, GBP-1, and TRAIL, was analyzed by quantitative PCR analysis, and results are illustrated in Fig. 2⇓. The most striking finding was the different kinetics of the transcriptional up-regulation induced by type I and type II IFN in HUVEC. A marked increase in the expression of all these genes was noticed already after 2 h of stimulation in both IFN-α- and IFN-γ-stimulated HUVEC and peaked in most cases at 5 h after stimulation. In contrast, in the case of IFN-α stimulation the expression of all genes analyzed substantially returned to baseline levels by 18 h, whereas it remained persistently up-regulated after stimulation with IFN-γ, up to 96 h after stimulation in the case of GBP-1 and TRAIL transcripts.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Time course analysis of gene expression in IFN-activated HUVEC. Real-time PCR analysis of CXCL9–11, GBP-1, and TRAIL expression was performed as described in Materials and Methods using primers listed in Table I⇑ and cDNA prepared from IFN-α- or IFN-γ-treated HUVEC. RNA was extracted from HUVEC at the time points indicated in the x-axis. Treatment with IFNs (1000 U/ml) lasted up to 5 h and was then followed by washing, replacement with fresh culture medium without IFN, and cultivation for the indicated times before RNA extraction (range, 18–96 h). Results indicate the log of relative (rel.) expression of the individual genes on the y-axis as judged by real-time PCR. Data were normalized on GAPDH and RNA polymerase II expression levels; the mean value is reported as log 2.

Effects of IFN-α and IFN-γ stimulation on gene expression in various EC types

We next addressed the effect of IFN-α and IFN-γ on the expression profile of CXCL10, CXCL11, and other selected genes in different types of normal human EC, including two isolates of microvascular cells (HMVEC no.1 and no.2) and HMVAR by both microarray and real-time PCR analysis. Comparison of the transcriptional responses to IFN-α treatment of HUVEC and HMVEC by microarrays revealed very similar expression profiles; Fig. 3⇓ shows a scatter plot of microarray gene expression values of IFN-α-induced genes in the two EC types. Data were confirmed by real-time PCR of selected genes (Table VI⇓); no major qualitative differences were observed for the different EC types as far as the direction of regulation and the ratio between IFN-α and IFN-γ were concerned. The only exceptions were CXCL9, which was much more strongly induced by IFN-γ in HMVEC and HMVAR than in HUVEC, and GBP-1, which was not induced in one of the two HMVEC cell lines (Table VI⇓). The preferential induction of IFIT1, MX2, and OASL by IFN-α observed for HUVEC (see Table III⇑) was also confirmed for the other EC types examined (Table VI⇓).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Scatter plot of expression values of genes induced by IFN-α or IFN-γ in HUVEC and HMVEC. Very few genes show differential expression exceeding the 2-fold interval delimited by the diagonals. Bars indicate SEM; r and p values were obtained by Pearson’s correlation.

Expression of CXCL10, CXCL11 and TRAIL in the supernatant of IFN-stimulated EC

To confirm some of the more marked transcriptional changes at the protein level, we quantified by ELISA the levels of some chemokines in the supernatants of IFN-treated cells. CXCL10 was produced at low levels (range 0.1–0.7 ng/ml) by IFN-α-treated EC and at much higher levels following IFN-γ treatment (range 1–30 ng/ml); among EC, HMVARs produced the highest CXCL10 amounts (Fig. 4⇓A). In contrast, HF released negligible levels of this cytokine in response to IFN-α, and only IFN-γ induced measurable CXCL10 production by these cells. Substantially similar findings were observed when we analyzed CXCL11 levels in the same samples. Following IFN-α treatment, CXCL11 was produced by all the different types of EC, but not by HF (Fig. 4⇓A). Among the different EC types, however, some quantitative differences were evident, as CXCL11 levels were ∼7-fold higher in HMVAR than in HMVEC. A much stronger difference emerged when IFN-α- and IFN-γ-treated cells were compared; 5–23 ng/ml CXCL11 was measured following IFN-γ treatment as compared with 0.02–0.6 ng/ml following IFN-α treatment (Fig. 4⇓A). Finally, when TRAIL levels were measured, modest TRAIL release was observed in the supernatants of IFN-α- and IFN-γ-treated cells; in contrast, as recently observed for other cell types (30), treatment of EC with both cytokines was associated with a substantial increase in cell-associated TRAIL, whereas sizable TRAIL expression by HF was only induced by IFN-γ (Fig. 4⇓B). To measure variations in protein levels over time, we treated HUVEC for 5 h with IFN-α or IFN-γ and then collected the conditioned medium at various time points and analyzed CXCL10 and CXCL11 concentrations by ELISA. Results are summarized in Fig. 4⇓C. Following treatment with IFN-γ, CXCL10 and CXCL11 levels increased steadily during the 96-h interval. In contrast, following treatment with IFN-α, CXCL10 levels increased at early time points (18–48 h) but subsequently dropped at later time points (72–96 h). A modest increase in CXCL11 levels was measured during the same interval in IFN-α-treated HUVEC. Notably, at any given time point analyzed CXCL10 and CXCL11 levels were ∼40- to 100-fold higher in supernatants of IFN-γ-treated HUVEC as compared with the corresponding IFN-α-treated cultures. A similar trend was also observed in kinetics experiments with HMVEC cells (data not shown). In general, the kinetics of CXCL10 and also partially CXCL11 protein expression were compatible with the transcriptional changes observed at the mRNA level.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Production of CXCL10, CXCL11, and TRAIL in the supernatants of IFN-stimulated cells. A, Different types of EC including microvascular EC (HMVEC 1 (HVEC#1) and HMVEC 2 (HMVEC#2)), HUVEC, and HMVAR, as well as human primary fibroblasts (HF E (HF#E) and HF 16 (HF#16)) were cultured for 48 h alone and in the presence of IFN-α or IFN-γ (1000 IU/ml). The amount of CXCL10, CXCL11, and TRAIL in the supernatants or cell lysates was measured in triplicate by ELISA. CXCL10 and CXCL11 levels in the supernatants of untreated cells were always below the sensitivity limits of the assay. Three independent experiments were performed; the error bars indicate the SD from the mean. nt, Untreated samples. Kinetics of CXCL10 and CXCL11 production by HUVEC. EC were cultured for 5 h in the presence of IFN-α or IFN-γ (1000 IU/ml); after washing, fresh medium without IFNs was added and the supernatants were collected at the indicated time points thereafter. Supernatants from four different HUVEC cultures were pooled and analyzed by ELISA to measure the amounts of CXCL10 and CXCL11. Three independent experiments were performed; the error bars indicate the SD from the mean.

Expression of CXCL10, CXCL11, and TRAIL in IFN-α-transduced xenografts

In view of the above data, which indicated that IFNs are able to up-regulate in vitro expression of some antiangiogenic chemokines by EC, we were interested in ascertaining whether a similar activity could also be recorded in vivo. Indeed, preliminary evidence (Fig. 5⇓) showed that IFN-α was able to profoundly affect the pattern of chemokine expression by murine lung and dermal microvascular EC, as CXCL10, CXCL11, and TRAIL expressions were strongly up-regulated following in vitro treatment with murine rIFN-α (Fig. 5⇓A). We thus wondered whether IFN-α could also affect in vivo the expression of some of the genes highlighted by the above analysis. To this end, we exploited a previously established xenotransplantation model based on parental or IFN-α-transduced MCF7 human breast cancer cells injected s.c. into SCID mice (4). As previously reported (4), the in vivo growth of IFN-α-producing tumors was greatly impaired compared with parental cells, and large areas of necrosis were present in IFN-α-expressing MCF7 tumors but not in control tumors (Fig. 5⇓B). RT-PCR analysis of the RNA extracted from tumor tissues disclosed that CXCL10, CXCL11, and TRAIL transcription was dramatically up-regulated in IFN-α-expressing tumors as compared with control tumors (Fig. 5⇓C). To identify which cells within the tumors were actually releasing CXCL11, we performed immunohistochemical analysis of MCF7 tumors. Although relatively few blood vessels could be analyzed because of the marked reduction in microvessel density in IFN-α-expressing tumor samples, CXCL11-producing cells were uniquely found in IFN-α-producing tumors and CXCL11 was predominantly expressed by EC within mid-size vessels (Fig. 5⇓D). Among the other genes increased by IFN-α treatment of EC in vitro, GBP-1 expression was also dramatically increased in IFN-α-expressing tumors as compared with control tumors, both at the mRNA (not shown) and the protein level (Fig. 5⇓E). Because both MCF7 and MCF7/IFN-α tumor cells only barely expressed GBP-1 (Fig. 5⇓E), it is conceivable that GBP-1 in MCF7/IFN-α tumors was mainly produced by stromal cells under the influence of IFN-α. Unfortunately, the poor performance of the anti-GBP-1 Ab in immunohistochemistry assays precluded the possibility of further investigating the cellular source of GBP-1 in these samples.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

In vitro and in vivo effects of IFN-α on the expression of CXCL10, CXCL11, and TRAIL by murine EC. A, SIEC (lung microvascular cell line) and 1G11 (dermal microvascular cell line) were incubated alone (nt) and in the presence of IFN-α; the expression of the genes listed on the left was analyzed by RT-PCR. B, Groups of SCID mice were injected s.c. with cells of the MCF7 tumor cell line, either untreated or engineered to express murine IFN-α (MCF7/IFN-α). H&E staining of tumor masses shows large areas of necrosis in IFN-α-expressing tumors. C, The RNA obtained from IFN-α-producing or control MCF7 tumor samples was analyzed by RT-PCR for expression of murine CXCL10, CXCL11, and TRAIL. β-Actin was used as an amplification control. D, Immunohistochemical analysis of CXCL11 expression in IFN-α-expressing tumors. Staining with anti-CXCL11 Ab disclosed CXCL11 expression in some blood vessels (arrow) within IFN-α-producing but not control tumors (original magnification, ×400). E, Western blot analysis of GBP-1 expression in IFN-α-expressing tumors. Hybridization with anti-GBP-1 Ab showed increased murine GBP-1 (muGBP-1) expression in IFN-α-expressing tumors (MCF7/IFN-α) than in controls (MCF7). Tubulin hybridization is shown as a loading control. Expression of GBP-1 in cell lysates from MCF7 cells treated in vitro with IFN-α or control cells is shown in the right panel.

CXCR-3 neutralization impairs but does not abrogate the antiangiogenic effects of IFNs

To analyze the functional role exerted by CXCL10 and CXCL11 in the antiangiogenic activity of IFNs, we performed an in vitro EC migration assay. As shown in Fig. 6⇓A, EC migration was partially inhibited by both IFN-α and IFN-γ, albeit more markedly by the latter. Neutralization of CXCR3 partially restored the migration of IFN-γ-treated EC, whereas it had minimal effects on IFN-α-treated cells. In contrast, CXCL10 inhibited EC migration similarly to IFN-γ, and its effects were virtually blocked by anti-CXCR3 as expected. Altogether, these findings indicated that although both IFN-α and IFN-γ impaired EC migration in vitro, CXCR3-binding chemokines contributed mainly to the negative modulation of EC chemotaxis by IFN-γ.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

The antiangiogenic effects of IFN-γ are mediated by CXCR3, the receptor for CXCL10 and −11. A, Migratory response of HMVEC to conditioned medium from NIH 3T3 cells and the effects of IFN-α, IFN-γ and anti-CXCR3 Ab. An isotype-matched Ab (IgG1) was used to control the specificity of the anti-CXCR3 mAb. Migration was assessed using a modified Boyden chamber as described in Materials and Methods. neg., Negative; pos., positive. B, Matrigel pellets containing different combinations of VTH, IFN-γ, anti-CXCR3 mAb, and its isotype-matched control (as outlined at the bottom of the panel) were injected s.c. in C57BL/6 mice. Plugs were removed 6 days later and vascularization was measured in terms of hemoglobin content normalized over plug weight (vascularization index). Anti-CXCR3 Abs reversed the antiangiogenic effect of IFN-γ, whereas isotype-matched control Ab had no effect. The Abs alone in the absence of IFN-γ did not influence the vascularization of the plug.

To investigate whether this may also occur in vivo, Matrigel pellets containing the angiogenic mixture VTH and VTH plus IFN-γ with or without a neutralizing Ab to murine CXCR3 were injected under the skin of C57BL/6 mice. Four days later, Matrigel plugs were removed and vascularization measured in terms of hemoglobin content normalized over plug weight. As shown in Fig. 6⇑B, anti-CXCR3 Ab reversed almost completely the antiangiogenic effects of IFN-γ, whereas control Abs did not. In conclusion, these results indicate that CXCR3 ligands substantially contribute to the antiangiogenic activity of IFN-γ.