Similar NF-κB Gene Signatures in TNF-α Treated Human Endothelial Cells and Breast Tumor Biopsies

Background Endothelial dysfunction has been implicated in the pathogenesis of diverse pathologies ranging from vascular and immune diseases to cancer. TNF-α is one of the mediators of endothelial dysfunction through the activation of transcription factors, including NF-κB. While HUVEC (macrovascular cells) have been largely used in the past, here, we documented an NF-κB gene signature in TNFα-stimulated microvascular endothelial cells HMEC often used in tumor angiogenesis studies. Methodology/Principal Findings We measured mRNA expression of 55 NF-κB related genes using quantitative RT-PCR in HUVEC and HMEC. Our study identified twenty genes markedly up-regulated in response to TNFα, including adhesion molecules, cytokines, chemokines, and apoptosis regulators, some of them being identified as TNF-α-inducible genes for the first time in endothelial cells (two apoptosis regulators, TNFAIP3 and TNFRSF10B/Trail R2 (DR5), the chemokines GM-CSF/CSF2 and MCF/CSF1, and CD40 and TNF-α itself, as well as NF-κB components (RELB, NFKB1or 50/p105 and NFKB2 or p52/p100). For eight genes, the fold induction was much higher in HMEC, as compared to HUVEC. Most importantly, our study described for the first time a connection between NF-κB activation and the induction of most, if not all, of these genes in HMEC as evaluated by pharmacological inhibition and RelA expression knock-down by RNA interference. Moreover, since TNF-α is highly expressed in tumors, we further applied the NF-κB gene signature documented in TNFα-stimulated endothelial cells to human breast tumors. We found a significant positive correlation between TNF and the majority (85 %) of the identified endothelial TNF-induced genes in a well-defined series of 96 (48 ERα positive and 48 ERα negative) breast tumors. Conclusion/Significance Taken together these data suggest the potential use of this NF-κB gene signature in analyzing the role of TNF-α in the endothelial dysfunction, as well as in breast tumors independently of the presence of ERα.


Introduction
Tumor necrosis factor (TNF-a) is a potent multi-functional cytokine produced by various cells, including neutrophils, activated macrophages, T and natural killer cells, and fibroblasts, in response to inflammation, infection, tissue injury and other environmental challenges. TNF-a elicits a large number of biological activities including the control of cell proliferation, differentiation and apoptosis, and is an important mediator of inflammation and immunity [1]. Considerable progress has been made in understanding the molecular mechanisms that mediate TNF-a-induced cellular responses. TNF-a binds to membrane receptors, TNFR1 (or TNF-Rp55) and TNFR2 (or TNF-Rp75), which initiate signalling pathway leading to gene induction via activation of transcription factors [2][3]. TNFR1 is ubiquitously expressed, while TNFR2 expression is restricted to endothelial and hematopoietic cells [4].
Through complex signaling cascades and networks, signal transducers lead to the activation of transcription factors (NF-kB, AP-1 and SP1) [2] and their binding to the promoters of specific genes.
The transcription factor NF-kB is a key inducible transcription factor which regulates numerous genes controlling inflammation, immune response, cell growth, tissue differentiation and apoptosis [5][6][7]. NF-kB-activating signals emanate from cell surface receptors for inflammatory cytokines (such as TNF-a, IL-1, or LPS), growth factors, lymphokines, reactive oxygen intermediates, viral infection, and B or T cell activation [5]. NF-kB factors are homo-and heterodimeric transcription factors that belong to the Rel family; this family is composed of five homologous subunits in mammals: NF-kB1 (p50 and its precursor p105), NF-kB2 (p52 and its precursor p100), RelA/p65, RelB and c-Rel. Biological activity mediated by NF-kB occurs through two main signalling pathways, the classical and the alternative pathways, the former being specifically induced in response to TNF-a, whereas the latter is induced by other specific stimuli [6][7][8]. In the classical pathway, the inactive NF-kB complexes are sequestered in the cytoplasm through interaction with specific inhibitors, the IkB proteins (IkBa, IkBb, especially). Following stimulation by TNF-a, activation of the IkB kinase (IKK) complex leads to the degradation of IkB proteins and subsequent release of mainly RelA/p50 dimers, which subsequently translocate into the nucleus where they bind to specific sites within the promoter of specific target genes. Based on studies on fibroblasts and B cell maturation, biological activity mediated by NF-kB may also occur through alternative pathway involving the activation of IKKa (IKK1) and the activation of RelB/p52 and RelB/p50 dimers [9][10][11].
Endothelial dysfunction has been implicated in the pathogenesis of diverse pathologic processes ranging from vascular diseases to cancer and metastasis. TNF-a mediated endothelial changes are relevant to acute inflammation, a highly complex biologic cascadeinvolving cytokines (such as IL6), chemokines IL-8 and MCP-1, and the induction of adhesion molecules (VCAM, ICAM and E selectin) [12][13][14][15][16][17] that recruit and activate granulocytes, monocytes/macrophages, and lymphocytes at the damaged tissue site. NF-kB often plays a major role in these pathways [17][18]. The diversity of the vascular bed is crucially determined by the type of endothelial cells which can be divided into microvascular and macrovascular endothelial cells. Although many reports have addressed the biological responses of human macrovascular endothelial cells (HUVEC) to TNF-a [14][15]19], the detailed dissection of the coordinated expression of TNF-a-inducible genes via NF-kB has not been performed in human microvascular endothelial cells (HMEC).
In this study, we have determined the expression profile of 55 selected key genes in TNF-a-stimulated HMEC, that are all known as being regulated by NF-kB in other cell types in response to various stress responses [20][21]. HMEC were chosen, in addition to HUVEC which have been largely used in the past, because they originate from the microcirculation, their varying thresholds to cytokines have been previously described and they are often used in tumor angiogenesis studies. Using real-time reverse transcription (RT)-PCR, we demonstrated that 36.4 % (20/55) genes are markedly up-regulated (.5-fold) by TNF-a in a dose-and time-dependent pathway in HMEC. Using two NF-kB inhibitors and a direct approach of RelA expression knock-down by RNAi, we demonstrated that the gene signature observed in TNF-a stimulated HMEC is NF-kB-dependent. Intracellular signalling pathways triggered by TNF-a leading to numerous tumors have been partially delineated, with NF-kB playing an important role especially in breast cancer cells [20][21][22][23][24][25]. However, the stroma of a solid tumor is vital for its survival, and key components in this respect are the blood vessels. We further demonstrated that tumors producing high amount of TNF-a could have similar dysregulated NF-kB endothelial pathways.

Cell transfections
Subconfluent HMEC were transfected in OPTI-MEM medium (Gibco) with Fugene (4 mg per P30 flask; Roche) containing a construct of firefly luciferase ligated to a NF-kB -sensitive promoter of the gene [29]. Twenty hours later, cells were incubated with Bay 11-7082 (10 mM) and stimulated with TNFa (40 ng/ml) for 4 h in the presence of the inhibitor, before they were harvested; pellets were analyzed for luciferase activity. Results are given as firefly luciferase activity as compared to TNF stimulation Transient of siRNA. TNFR1 and TNFR2 were silenced using a specific siRNA mix containing three different siRNAs provided by Santa Cruz Biotechnology, as previously described [30]. As a control, siRNA with a nonsense sequence was usedg. Briefly, 10 mM of siRNA was added to the transfection reagent (Oligofectamine, InVitrogen; 1.25 mg) and subsequently transferred to culture plates P100 flask with 4610 5 /flask HMEC. Then, 24 hours after transfection, cells were stimulated with TNF-a for 4 h and the specific experiment was performed.
Plasmid constructs. pTrip-shRNA RelA was generated by subcloning an oligonucleotide designed to target RelA under control of the H1 promoter into pTRIP-DU3-MND-GFP lentiviral vector [31]. pTRIP-shRNA control was generated following the same approach using a scrambled control oligonucleotide. Sequences are available upon request.
Lentiviral infection. Production of infectious recombinant lentiviruses was performed by transient transfection of 293T cells, as previously described [31]. For infections, 10 5 cells in 35-mm dishes were transduced with 5000 ng/ml of p24 (HIV-1 capsid protein). 48 h later, cells were washed and fresh medium was added. The culture was then continued as described above.

Electrophoretic Mobility Shift Assays (EMSA)
Nuclear extracts were prepared and analyzed for DNA binding activity using the HIV-LTR tandem kB oligonucleotide as kB probe [32]. For supershift assays, nuclear extracts were incubated with specific antibodies for 30 min on ice before incubation with the labeled probe. Anti-RelA, RelB, cRel, and p105/p50 supershifting antibodies were purchased from Santa Cruz Biotechnology.
Ratio of the mRNA content in stimulated cells for 4 hours by TNF-a to the mRNA content in the unstimulated cells was calculated. This time was chosen from preliminary kinetic analysis on gene candidates (such as E-selectin and IL-8) (see also

Patients
To investigate the in vivo inter-relationships between mRNA levels of genes of interest, we analyzed 96 (48 ERa positive and 48 ERa negative) breast tumors excised from women at our institution (Centre René Huguenin, St Cloud, France). Samples containing more than 70% of tumor cells were considered suitable for this study. Immediately following surgery, the tumor samples were placed in liquid nitrogen until RNA extraction. The patients met the following criteria: primary unilateral non metastatic breast carcinoma; complete clinical, histological and biological information available; no radiotherapy or chemotherapy before surgery; and full follow-up at our institution. Estrogen receptor alpha status was determined at the protein level by biochemical method (enzymatic immunoassay) and confirmed by ERa real-time quantitative RT-PCR assay [33]. The malignancy of infiltrating carcinomas was scored according to Scarff Bloom and Richardson's histoprognostic system [34]. The characteristics of these patients are shown in Table 1.

RNA extraction
RNA was isolated using TRIzol reagent according to the manufacturer's instructions (Invitrogen, Cergy Pontoise, France). The quality of the RNA samples was determined by electrophoresis through agarose gels and staining with ethidium bromide, and the 18S and 28S RNA bands were visualized under ultraviolet light.

Real-time quantitative-PCR
Transcript quantification of NF-kB selected genes was performed as previously described [35], using the TATA box-binding protein (TBP) as the endogenous RNA control; each sample was normalized on the basis of its TBP content. Primers for TBP and the 55 target genes ( Table 2) were chosen with the assistance of the Oligo 5.0 computer program (National Biosciences, Plymouth, MN). We searched the dbEST and nr databases to confirm the total gene specificity of the nucleotide sequences chosen as primers, and the absence of single nucleotide polymorphisms. In particular, the primer pairs were selected to be unique relative to the sequences of closely related family member genes or of the corresponding retropseudogenes. To avoid amplification of  contaminating genomic DNA, one of the two primers was placed at the junction between two exons, if possible. In general, amplicons were between 60 and 120 nucleotides long. Gel electrophoresis was used to verify the specificity of PCR amplicons. For each primer pair we performed no-template control (NTC) and no-reverse-transcriptase control (RT-negative) assays, which produced negligible signals (usually .40 in Ct values), suggesting that primer-dimer formation and genomic DNA contamination effects were negligible. The cDNA synthesis and PCR conditions have been described in detail elsewhere [35]. Results are expressed as N-fold differences in target gene expression relative to TBP gene expression.

Statistical analysis
Statistical analysis of in vitro experiments was performed by analysis of mean with the Student t test. The terms n = x is used to indicate the number of independent experiments performed. Correlations between TNF and the gene target expressions were tested using the non parametric Spearman rank correlation test (which compared continuous variables). Differences were judged significant at confidence levels greater than 95% (p,0.05).

Identification of TNFa-dependent NF-kB-associated genes in microvascular endothelial cells
To determine the expression pattern of NF-kB-associated genes in response to TNF-a in microvascular endothelial cells, we have performed real-time RT-PCR in HMEC stimulated by TNF-a for 4 h. Our preliminary kinetic analysis on gene candidates (such as E-selectin and IL-8) suggested an induction at 2-4 h time point induction. Among the 55 genes tested ( Table 2), only one gene, IL12B, presented detectable but not reliably quantifiable mRNA expression (Ct.35). Twenty (37 %) of the other 54 genes were upregulated by 5-fold or more (until 1200-fold) in HMEC stimulated with TNF-a as compared to untreated cells ( Table 3). These robust selection criteria (a cut-off of 5-fold expression difference in HMEC stimulated with TNF-a) ensure identification of genes with marked interest.
In contrast, TNF-a did not modify the expression of a few set of genes known to be regulated by TNFa in tumor cells (such as MMP9). In the same set of samples, TNF-a did not modify the expression of MKI67 gene encoding the proliferation-related Ki-67 antigen cell proliferation marker.
To confirm that the increase of mRNA expression correlates with induction of protein expression, cell lysates from TNF-a stimulated HMEC were subjected to western immunoblotting using antibodies directed against ICAM, IL-1b or several proteins of the NF-kB and IkB families. As shown in Fig. 1, treatment of HMEC with TNF-a strongly up-regulated the expression of IL-1b ICAM and RelB. TNFa also increased NF-kB1/p50 and NF-kB2/p52 expression in endothelial cells, whereas modulation of RelA protein level was not observed, in agreement with findings at the mRNA levels.

TNF-a modulates NF-kB -associated genes in a dose-and time-dependent manner
Expression of the NF-kB -associated genes was up-regulated with TNF-a in a dose-dependent manner (1-80 ng/ml) in HMEC with the same profile of response (maximum degree of response for 40 ng/ ml of TNF-a) ( Table 4). Ten of these genes were up-regulated more than 10-fold at 0.2 ng/ml, as compared to controls, with VCAM, CSF2/GMCSF and IL8 being stimulated more than 100-fold.
We also investigated the kinetics of TNF-a-induced mRNA expression. HMEC were treated with TNF-a (40 ng/ml) at different time periods (0 to 72 h). We observed the same profile of response for all genes, with a maximum induction always observed at 4 hours of TNF-a treatment ( Table 5), associated to a marked decreased at 24 and 72 hours.
Quantitative and qualitative differences in the expression of the NF-kB-associated genes between microvascular and macrovascular endothelial cells Quantitative RT-PCR was used to compare the expression of NF-kB associated genes in TNF-a stimulated HUVEC and HMEC (40 ng/ml for 4 h). As shown in Table 6, TNF-a upregulated the expression of the same 20 NF-kB associated genes on the two classes of endothelial cells, except for CD40, IL6 and TNFRSF10B that are hardly induced in HUVEC. Interestingly, for VCAM, ICAM, IL8, CSF2, CXCL1/GRO1, RELB, IL6 and TNFRSF10B expression, the fold induction was much higher in HMEC, as compared to HUVEC. However, these results were probably due to a lower basal expression level in HMEC compared to HUVEC ( Table 6). Fig. 2 shows the mRNA levels of three characteristic genes (CCL2/MCP1, IL8 and IL1A) in TNFa-stimulated HUVEC and HMEC cells. Finally, it is interesting to note that among the 35 genes whose expression was not upregulated (.5-fold) in stimulated HMEC, no gene was upregulated in HUVEC (.5-fold) (not shown). TNFR1 and, to a lesser extent, TNFR2 stimulate the NF-kB --associated genes We then analysed the mechanisms of the regulation by TNF-a observed in endothelial cells first using silencing of TNF receptors. Since TNFR1 (or TNF-Rp55) and TNFR2 (or TNF-Rp75) are both present in endothelial cells, we silenced the expression of each receptor by RNA interference in HMEC. In the presence of TNF-a, the expression of TNFR1 was reduced from 45 % in siRNA TNFR1-transfected cells, as compared to siRNA control ( Table 7). In these cells, the expression of twelve NF-kB associated genes including VCAM, ICAM, CSF2/GMCSF and CSF1, the cytokines TNF, IL8, IL1A, IL1B and IL6, TNFAIP3, BCL2A1 and NFKB1 was inhibited more than 30 %, as compared to TNF-a stimulated siRNA control-transfected cells ( Table 7). In contrast, TNFR2 expression was not reduced in siRNA TNFR1-transfected cells, suggesting that the decrease of TNFR1 expression was receptor specific. We further tested the putative involvement of TNFR2 in TNF-a response in endothelial cells. The expression of TNFR2, but not TNFR1, was reduced from 77 % in HMEC transfected with siRNA TNFR2. Only 4 genes including CSF2/GMCSF, IL1B, TNFRSF10B and TNF were inhibited more than 15 % (28 %, 26 %, 22 % and 17 %, respectively) when TNFR2 was silenced ( Table 7). Altogether, these results indicate that TNFR1 and, to a lesser extent, TNFR2 are involved in the control of the NF-kB associated gene signature in TNF-a stimulated endothelial cells.

NF-kB activation is involved in TNF-a-induced gene expression in endothelial cells
We next examined the effects of chemical inhibition of IkB phosphorylation and IkappaB kinase activity on gene expression in TNF-stimulated HMEC (Table 8). Bay 11-7082, an inhibitor IkB phosphorylation [27] was first tested on NF-kB activity in endothelial cells using a luciferase reporter gene whose activity depends on a kB element. Luciferase activity was decreased by 99 % after addition of Bay 11-7082 (10 mM) on TNF-a treated endothelial cells (data not shown). Most importantly, the addition of Bay 11-7082 resulted in decreased expression of all selected genes in TNF-a treated HMEC ( Table 8). As a control, TNFR receptor expression which was not changed by TNF-a stimulation was insensitive to Bay 11-7082. The addition of pharmacological IkB kinase (IKK2) inhibitor V [28] also inhibited the expression of VCAM, ICAM, CCL2/MCP1 and BCL2A1 and IL1Bat 1-10 mM (Table 8); surprisingly, it activated CSF2/ GMCSF, TNFAIP3, IL1A and IL8. Altogether these experiments using two different inhibitors suggest that canonical NF-kB signalling pathway is involved in TNF-k-induced gene expression in endothelial cells.
To directly demonstrate the involvement of NF-kB in the gene signature induced by TNF-a in endothelial cells, we first evaluated TNFa-induced activation of NF-kB complexes in HMEC cells by EMSA. As shown in Fig 3A, TNF-a treatment in HMEC resulted in  a characteristic biphasic induction of kB DNA binding (complex I). Supershift analysis of nuclear extracts stimulated for 4 h with TNF-a indicated that the NF-kB DNA-binding complexes consist of RelA-p50 dimers (Fig 3B).
We then used stable RNA interference knock-down to inhibit RelA activity in HMEC cells. We infected HMEC cells with either a lentivirus carrying a shRNA targeting RelA or a scrambled control. Whereas TNFa stimulation of scrambled control-infected HMEC cells induced RelA DNA binding with kinetics that parallel what is seen in non-infected cells, RelA expression knock-down almost completely abolished RelA-DNA binding activity (Fig 3C).
We then examined the mRNA expression levels of 16 genes by real time RT-PCR. The down-regulation of RelA activity clearly inhibited the expression of all of them except TNFRSF10B (Table 9), thus indicating that RelA is required for the TNF-a-induced expression of these genes in HMEC cells.
Positive correlation between the expression of TNF-a mRNA and of the 19 other identified putative TNF-inducible genes in 96 human breast tumors TNF-a is highly expressed in tumors where it has been considered initially as a potent tumor cell killer and an antivascular cytokine at high doses [36]. However, low chronic doses of the cytokine are thought to be pro-angiogenic [4]. In order to test the involvement of TNF-a activation/endothelial dysfunction of the breast tumor vasculature, we further explored if the correlations identified in our ex vivo cellular model (endothelial cells) between the expression of TNF-a mRNA and the expression of 19 putative TNF-a-inducible NF-kB target genes were also observed ex vivo in human breast tumors.
This study was performed on a series of 96 breast tumors (48 ERa-positive and 48 ERa-negative tumors) ( Table 1). We observed a positive correlation between TNF and 16 (84.2 %) of the 19 endothelial TNF-a-inducible genes in both ERa-negative and ERa-positive breast tumors ( Table 10). The three remaining genes (i.e., IL8, IL1A and IL6) were significantly linked to TNF only in ERa-negative tumors. In the same set of 96 tumors, we also examined the expression of the ESR1 gene, which encodes ERa as well as the proliferation-associated gene MKI67, which encodes the proliferation-related antigen Ki-67. We found no correlation between TNF and ESR1 and MKI67, suggesting that the 19 genes were associated to TNF independently to hormonal status and cell proliferation.

Prognostic value of TNF-a mRNA and of the 19 putative TNF-inducible gene mRNAs in human breast tumors
We estimated the prognostic value of the TNF-a mRNA and the 19 other identified putative TNF-a-inducible gene mRNAs in our series of 96 breast tumors (48 ERa-positive and 48 ERa-negative tumors). In the cohort of 48 ERa-negative breast tumors, nineteen patients relapsed, whereas in the cohort of 48 ERa-positive breast tumors, twenty-seven patients relapsed (Table 11). We used univariate analysis (log-rank test) to study the prognostic value of these 20 genes. For each gene, the 48 ERa-positive (or -negative) breast tumors were divided into two equal groups of 24 tumors  with 'low' and 'high' mRNA levels. Univariate analysis showed that only low expression level of NFKB2 in the cohort of 48 ERanegative breast tumors and high expression level of IL8 in the cohort of 48 ERa-positive breast tumors correlated with significantly shorter metastases-free survival, but at the limit of the significance (p = 0.039 and p = 0.022, respectively) ( Table 11).

Discussion
This study documents, for the first time, the NF-kB gene expression signature in two different subtypes of TNFa-stimulated endothelial cells, especially in HMEC which originate from the microcirculation and have been much less studied as compared to HUVEC. Twenty genes were up-regulated 5-fold or more (until 1200-fold) in TNFa-stimulated endothelial cells, in a concentration-and time-dependent manner. The up-regulated genes include those controlling cytokines, chemokines and adhesion molecules, relevant to inflammation and apoptosis, as well as factors involved in the immune system. The most highly up-regulated genes in TNFa-stimulated endothelial cells are those encoding adhesion molecules and cytokines/chemokines (i.e., TNF-a, IL-8, MCP-1 and GM-CSF/CSF2). Novel observations concern the upregulation of TNFAIP3, CXCL1/Gro1, GM-CSF/CSF2 and MCF/ CSF1, TNFRSF10B/Trail R2 (DR5), CD40 and TNF itself in TNFa-stimulated endothelial cells via NF-kB, as well as selective expression of NF-kB components (RELB, NFKB1 and NFKB2). Interestingly, while TNFR1 was involved in all the regulations observed, some of the regulated genes were shown to be regulated through both TNFR1 and TNFR2. Moreover, the correlations identified in our ex vivo cellular model (endothelial cells) between the expression of TNF-a mRNA and the expression of 19 putative TNF-a-inducible NF-kB target genes were also observed in human breast tumors.
The up-regulation of adhesion molecules (VCAM-1, ICAM-1 and E-selectin) and pro-inflammatory cytokines (IL1a, ILb and IL-6) in TNFa-stimulated endothelial cells through NF-kB agrees with previous findings reported in HUVEC [16][17][18].IL-6 expression was poorly modulated by NF-kB activation in these cells. Interestingly, our results show that TNF-a strongly up-regulates TNF-a itself (but not TNFR1 and TNFR2 expression), as well as the expression of DR5 (TNFRSF10B or TRAIL-R2) and CD40, two additional members of the TNF/TNF receptor superfamily involved in the immune system [3]. Such induction occurs through TNFR1 and TNFR2 and via NF-kB activation. Interestingly, DR5 has been detected in endothelial cells derived from blood brain barrier (HBEC), where it released MMP9 upon activation [37].
The chemokines IL-8/CXCL8, CXCL1/Gro1, CCL2/MCP-1, GMCSF/CSF2 and MCF/CSF1 were significantly up-regulated more than 5 fold in HMEC and HUVEC. Their expression was upregulated as early as 30 min-1 h and sustained for up to 24 h, as previously described for IL-8 and CCL2/MCP-1 in HUVEC [15,38]. The up-regulation of IL-8 through NF-kB has not been described before in TNF-a-induced endothelial cells. IL-8, CCL2/ MCP-1 and CXCL1/Gro1 facilitate the subset specific recruitment of neutrophils, monocytes, T lymphocytes or NK cells. In addition to their chemotactic activity, the three chemokines possess angiogenic activity in vitro [39] and are involved in the regulation of tumoral angiogenesis such as in breast cancer [40][41]. Our study further shows for the first time that GM-CSF/CSF2 and MCF/ CSF1 expression is induced through NF-kB in HMEC. CSF1 has been previously shown to be induced in HUVEC by endotoxin [42][43]. This chemokine promotes differentiation of hematopoietic cells to mature granulocytes and macrophages. GM-CSF/CSF2 is involved in the recruitment of macrophages in various inflammatory diseases and was recently found to mobilize bone marrow progenitors [44]. This chemokine is also involved in angiogenesis and is a key target of NF-kB in cancer and metastasis [45]. Bcl2A1 (A1) and TNFAIP3 (also known as A20) are antiapoptotic proteins expressed in endothelial cells [46][47][48], and have also been associated with angiogenesis [49]. Their expression was shown to be up-regulated by TNF-a in both HUVEC and HMEC, in contrast to other classical NF-kB-inducible antiapoptotic proteins, such as MCL1S, MCL1L, BIRC2 (c-IAP1), BIRC4 (XIAP) and TRAF2. To our knowledge, our study indicates for the first time the up-regulation of TNFAIP3/A20 gene expression by TNF-a through NF-kB.
In our study, several approaches were used to demonstrate the involvement of the NF-kB canonical activation pathway in TNFa-induced gene expression in endothelial cells. Our observations showing that genes encoding the kinases IkKa (CHUK), IkKb (IKBKB) and NEMO/IkKc (IKBKG) were not up-regulated by TNF-a in endothelial cells indicate that changes in the NF-kB target gene expression is not dependent upon changes in the expression of these kinases. The decreased expression of the NF-kB-mediated genes in TNF-a treated HMEC in the presence of Bay11-7082 (a specific inhibitor of cytokine-induced IkBa phosphorylation), is in agreement with the involvement of this pathway. It is also in agreement with preliminary experiments showing the time-dependent phosphorylation of IkBa and nuclear translocation of p50 (NF-kB1) in TNF-a endothelial cells (L Michel and M Applanat, not shown). The most direct demonstration that TNF-a-induced gene expression in HMEC depends on NF-kB activation arises from RelA knock-down experiments, RelA representing the major NF-kB activity in HMEC stimulated by TNFa [50] (see Fig. 3). Nonetheless, the expression of TNFRSF10B is not significantly down-regulated by RelA knock-down as compared to the effect observed with the chemical inhibitors. The significance of the selective increase of expression of RELB, NFKB1 and NFKB2/p52, Rel/NF-kB family members, but not RelA and c-rel, which is demonstrated for the first time in TNFastimulated endothelial cells, is unknown. These up-regulations occur at both the mRNA and protein levels, and correlate with nuclear translocation of p52 (not shown). These findings suggest that heterodimers NF-kB2 (p52)/RelA or p52/p52 could also favour TNF-a signalling in endothelial cells. Of note, TNFRSF10B/DR5 has previously been shown to be activated by p52 in some context [51]. The non canonical pathway is known to be activated by stimuli such as lymphotoxin-b, CD40L and BAFF, involves IKKa (IKK1) activation and nuclear translocation of RelB in the form of RelB/p52 and RelB/p50 dimers [9][10]11,[51][52].
Our study also contributes to analyze the role of TNFR1 and TNFR2 receptors, which are expressed at a similar level in endothelial cells. The decrease of gene expression is observed in TNFa-stimulated endothelial cells after silencing TNFR1; this receptor is thought to mediate the majority of the biological effects of the TNFa and the classical pathway of TNF-a signalling. However, silencing of TNFR2 also decreases the expression of IL1B, GM-CSF/CSF2, TNFRSF10B/Trail R2 and TNF, suggesting that both TNFR1 and TNFR2 are involved in TNF-a dependent up-regulation of these genes in HMEC.
TNF-a and its related NF-kB transcription factor are described as critical components of tumor progression [20,22]. NF-kB is deregulated in breast cancer patients and in breast cancer cells [25,[53][54][55]. Several studies have shown that tumor-associated neovascularisation is a prerequisite of rapid growth and metastasis [56]. Although great strides have been made in the elucidation of molecular mechanisms of tumor vasculature, different markers and many essential interactions have to be yet determined. Among genes described in this study, IL8, CCL2/MCP1, CXCL1/Gro1, Bcl2 and TNFAIP3 are known to be pro-angiogenic [38][39][40]49,57]. Moreover, one of the consequences for TNF-á activation of tumor endothelial cells is also to recruit numerous cells through chemokines and adhesion molecules, including monocytes, neutrophils, fibroblasts and immune cells inside the tumor.
Our study provides for the first time a significant correlation between TNF-a expression and the expression of putative TNF-ainducible NF-kB target genes (as identified in microvascular endothelial cells) in human breast tumors ( Table 10). The highest correlation between TNF-a and NF-kB related genes was observed for ERalpha negative-, as compared to ERpositivebreast tumors. TNF-a expression and IL8, IL1A and IL6 expression were only correlated in ER-negative, but not in ERpositive breast tumors. A distinct pattern of activation of NF-kB subunits (c-Rel, RelB, p50/ NF-kB1, p52/ NF-kB2, and Ik-Ba), as well as the absence (or low level) of nuclear p65/RelA has also been previously described in breast tumors, as compared to normal adjacent tissue [22,48]. Altogether, these data suggest that dysregulated/TNF-a stimulated endothelial cells through NF-kB are likely to play an important role in breast cancer.
In conclusion, we describe for the first time the profile of cytokines, chemokines, adhesion molecules and NF-kB components induced in endothelial cells stimulated by TNF-a through TNFR and NF-kB. This study provides a NF-kB gene signature associated with TNF-a stimulation in endothelial cells. Some of the newly identified regulated genes, such as those involved in the recruitment of bone marrow or endothelial cell progenitors, will be further explored in a mouse cancer model. The cytokines, chemokines, adhesion molecules and angiogenic factors, as well as factors involved in the immune system, that we have identified in TNF-a stimulated endothelial cells may also promote tumor progression in vivo.