Recent Progress in Hormone Research 58:95-130 (2003)
© 2003 The Endocrine Society
Identification of a Nuclear Factor Kappa B-dependent Gene Network
Bing Tian and
Allan R. Brasier
Department of Medicine and the Sealy Center for Molecular Sciences, The University of Texas Medical Branch, Galveston, Texas 77555-1060
 |
ABSTRACT
|
|---|
Nuclear factor-kappa B (NF-
B) is a highly inducible transcription factor that plays an important role in the hepatic acute-phase response, innate/adaptive immunity, and cellular survival through the induction of genetic networks. The major transcriptional-activating species Rel A-NF-B is a cytoplasmic complex whose nuclear translocation is controlled by its association with a family of inhibitory proteins, termed IBs. Activation of NF-B results in the targeted proteolysis of IB, releasing NF-B to enter the nucleus and bind to specific sequences in target promoters. Because the genomic actions of NF-B are influenced by the stimulus applied and the promoter context/chromatin structure in which it binds, the spectrum of NF-B-regulated genes has not been elucidated. We have begun to address this question, exploiting a tightly regulated cellular system expressing a nondegradable IB
mutant that completely inhibits NF-B action. High-density oligonucleotide microarrays were used to identify genetic responses in response to complex biological stimuli (viral replication) in the presence and absence of NF-B. Using statistical and informatics tools, we identified two groups of NF-B-dependent genes with distinct expression profiles: 1) a group with high constitutive expression whose expression levels fall in response to viral exposure and constitutive mRNA expression increases from NF-B blockade, and 2) a group where constitutive expression was very low (or undetectable) and, after stimulation, expression levels strongly increased. In this group, NF-B blockade inhibited the viral induction of genes. This latter cluster includes chemokines, transcriptional regulators, intracellular proteins regulating translation and proteolysis, and secreted proteins (e.g., complement components, growth factor regulators). These data reveal complexity in the genetic response to NF-B and serve as a foundation for further informatics analysis to identify genetic features common to up- and downregulated NF-B-dependent promoters.
|
I. Introduction
|
|---|
The mechanisms controlling RNA polymerase II-dependent gene expression in eukaryotes have been investigated intensively over the past several decades. From this work, it is widely accepted that transcriptional initiation is the pri-mary mechanism controlling gene expression in response to tissue specific-, developmental-, and hormonal-induced cellular signaling (McKnight and Yamamoto, 1992). Transcriptional activation is a multistep process initiated by sequence-specific proteins binding to upstream regulatory sites of DNA flanking transcribed regions of a gene. In highly inducible gene promoters, enhanceosomes — nucleoprotein complexes assembled on inducible enhancer sequences in the regulatory regions — allow for multiple, intracellular signaling cascades to modulate target gene expression (Thanos and Maniatis, 1995). Formation of the enhanceosome is initiated by the binding of sequence-specific transcription factors and results in the cooperative binding of accessory chromatin-remodeling proteins, nuclear coactivators, kinases, and/or histone acetylases into the nucleoprotein complex. The presence of chromatin-remodeling factors induces formation of an activated preinitiation complex that, in turn, controls DNA-dependent RNA polymerase II activity and subsequent gene expression. Although the basic biochemistry of preinitiation complex formation and the identification of how signaling molecules control transcription factor have resulted from the study of promoters on an individual level, the spectrum of genes controlled by specific signaling pathways within the context of a complex biological stimulus has not been explored systematically.
The development and application of high-density microarrays is an enabling technology that allows global understanding of genome-wide changes in cellular mRNA abundance in response to specific stimuli (Tavazoie et al., 1999; Bussemaker et al., 2001; Berman et al., 2002). However, as it is commonly used, mRNA profiling is a correlational tool that records patterns of gene expression in response to specific cellular perturbations. Although these studies provide new insight into cellular responses, allowing the identification of genes controlled by a stimulus not previously considered, little mechanistic information arises on the interaction of downstream signaling pathways in controlling specific subsets of genes. Herein, we describe our approach to experimentally identify genetic networks under inducible transcription factor control using mRNA profiling. We have chosen the highly inducible transcription factor, NF-B, and sought to identify its role in response to a complex biological stimulus (viral replication).
A. NF-B AS AN INDUCIBLE TRANSCRIPTIONAL REGULATOR
NF-B is a ubiquitously expressed, highly inducible transcription factor that plays an important role in the hepatic acute-phase response, innate and adaptive immunity, and cellular survival through the induction of genetic networks (Barnes and Karin, 1997; Karin, 1999). Composed of five distinct gene products related by a common NH2-terminal Rel homology domain, the NF-B family is divided into two functionally distinct groups that freely heterodimerize. The first group consists of members translated as mature proteins that bind DNA weakly and contain potent COOH-terminal transcriptional activation domains (Rel A and c-Rel). The second group has members encoded by large precursor proteins that are proteolytically processed into small (
50 kDa) DNA-binding subunits with strong DNA-binding activity and weak transcriptional activation potential (p105/NF-B1 and p110/NF-B2) (reviewed in Siebenlist et al., 1994; Baldwin, 1996; Barnes and Karin, 1997). The subunit composition of the NF-B complex influences its subcellular localization, transactivation potential, and mode of regulation. For example, homodimers of NF-B1 are primarily nuclear proteins with weak transactivation potential. They are activated by phorbol myristate acid (PMA) and peptide hormones to weakly activate gene transcription (Jamaluddin et al., 2000). Conversely, the prototypical NF-B complex, composed of 50 kDa NF-B1-65 kDa Rel A heterodimers, is primarily cytoplasmic. The complex is regulated by its association with a family of inhibitors, IBs, whose members bind and specifically inactivate Rel A by masking its nuclear-localization sequence and preventing nuclear entry (Henkel et al., 1993; reviewed in Beg and Baldwin, 1993; Han et al., 1997). In response to cellular stimulation by cytokines and pathogens, Rel A-NF-B1 enters the nucleus, where it becomes a potent transactivator.
The mechanism controlling nuclear translocation of Rel A-NF-B1 has been intensively investigated. Rel A-NF-B1 activation requires inducible proteolysis of the IB inhibitors, a process initiated by IB phosphorylation (Brown et al., 1995). Intracellular NF-B-activating signals converge on the multiprotein cytoplasmic IB kinase complex (IKK), a complex that phosphorylates IB on two serine residues (Ser32 and Ser36) in its NH2-regulatory domain (reviewed in Karin, 1999). The IKK is a multiprotein, cytosolic kinase of 700 kDa, composed of catalytic basic helix-loop-helix-containing kinases, IKKß and IKK, and a regulatory subunit, IKK
, required for coupling IKK to upstream activating kinases (Karin,1999; Poyet et al., 2000; Zhang et al., 2000). The pathways for how the cytokines, interleukin-1 (IL-1) and tumor necrosis factor-alpha (TNF), activated NF-B through the IKK have been extensively investigated. IL-1 and TNF cytokine receptors are single-pass, transmembrane receptors lacking intrinsic kinase activity. Upon binding ligand, the receptors form trimeric structures, resulting in intracellular aggregation of the cytoplasmic signaling domains. This results in the recruitment of a submembranous signaling complex activating IKK; in the case of the IL-1 signaling pathway, the signal adapters include MyD88, IL-1 receptor-associate kinase (IRAK), and the TNF receptor-associated factor 6 (TRAF6) (Cao et al., 1996). TRAF6 is required for coupling the activated IL-1 receptor to the IKK; recently, it was shown that TRAF6 is ligated by unique lysine 63-linked polyubiquitin chains, a modification that is required to activate the downstream MAPKK kinase, TGFß-associated kinase (TAK1) (Ninomiya-Tsuhi et al., 1999; Wang et al., 2001). TAK1, in turn, phosphorylates the first committed step in NF-B activation, the NF-B-inducing kinase (NIK), resulting in IKKß phosphorylation and IKK activation (Malanin et al., 1997; Wang et al., 2001). In TNF signaling, TNF receptor-associated death domain (TRADD), TRAF2, and receptor-interacting protein (RIP) constitute the submembranous signaling complex (Hsu et al., 1996; Zhang et al., 2000). Here, the RIP kinase plays an indispensable role in IKK activation, as targeted disruption of RIP abolishes NF-B activation (Kelliher et al., 1998). RIP recently has been shown to recruit the IKK signalsome to the TNF receptor I by binding IKK, directly recruiting the cytosolic IKK to it (Zhang et al., 2000). IKK and IKKß then are phosphorylated, releasing the activated IKK complex back into the cytosolic fraction (Poyet et al., 2000; Zhang et al., 2000).
In the IKK complex, the catalytic IKKß subunit is largely thought to be responsible for the site-specific serine phosphorylation of IB in its NH2-terminal regulatory domain, resulting in the first rate limiting in the process of NF-B activation. PhosphoIB is specifically bound by the Skp1-cullin-F-box-type E3 ubiquitin ligase, E3RS, initiating IB ubiquitination and proteolysis through the proteasome (Brown et al., 1995; Karin, 1999; Karin and Ben Neriah, 2000). A parallel pathway important in viral infection that produces IB degradation through cytoplasmic calpains also has been identified (Jamaluddin et al., 1998; Han et al., 1999b). Following IB proteolysis, liberated NF-B enters the nucleus to activate target gene transcription.
B. MECHANISMS FOR NF-B-DEPENDENT TRANSACTIVATION
Liberated, cytoplasmic NF-B rapidly enters the nucleus through specific nuclear-importing signals located in its NH2-terminal Rel homology domain and bind to specific DNA sites in the regulatory regions of inducible promoters. A nonbiased polymerase chain reaction (PCR)-binding site-selection assay has elucidated the high-affinity binding sites for the homodimeric NF-B1 (p50) and Rel A (p65) proteins as 5'-GGGGATYCC-3' and 5'-GGGRNTTTCC-3', respectively (Kunsch et al., 1992). These binding sites match the empirically defined consensus binding sites of acute-phase reactant (angiotensinogen) (Brasier et al., 1996), cytokine (IL-8) (Garofalo et al., 1996; Brasier et al., 1998), and chemokine (RANTES, an acronym for regulated upon activation, normally T-cell expressed and presumably secreted) (Casola et al., 2000b) promoters.
Promoter-bound NF-B activates transcription from inducible regulatory elements through the assembly of larger nucleoprotein complexes termed enhanceosomes. Pioneering studies on the interferon (IFN)-ß promoters have revealed that inducible transcriptional regulation is a multistep process involving the cooperative assembly of architectural (DNA remodeling), sequence-specific transcription factors, and coactivator/bridging proteins on a target enhancer. The phenomenon of binding cooperativity allows highly inducible genes to be expressed in the setting of limiting concentrations of transcription factors (reviewed in Carey, 1998). The IFN-ß promoter contains a virus-inducible enhancer with three essential domains, termed positive regulatory domains (PRDs). In vitro, PRDII binds NF-B, PRDIV binds activating transcription factor (ATF)-2/c-Jun, and PRDIII-I binds interferon regulatory factor-1 (IRF-1). Gene-transfer studies have shown that stimulus-specific activation of IFNß is accomplished through initial recruitment of the architectural protein high-mobility group (HMG) I(Y), a minor groove DNA-binding protein recognizing AT-rich elements contained in the PRDII and PRDIV domains (Yie et al., 1999b). HMG I(Y) binding bends the DNA into a conformation that allows cooperative binding of the transcription factors NF-B and ATF-2/c-Jun to their respective motifs in PRDII and PRDIV domains. Additional protein-protein interactions recruit IRF, general transcription factors, and the coactivator/pol holoenzyme cyclic AMP response binding protein (CBP)/polII into the complex (Munshi et al., 1998). Additional modifications to DNA and histones result in an increased rate of preinitiation complex formation, allowing multiple rounds of transcription and reinitiation of the IFNß transcripts (Yie et al., 1999b). Once assembled, the enhanceosome is an extremely stable nucleoprotein structure through protein-protein and protein-DNA interactions, resistant to oligonucleotide competition and detergent treatment (Yie et al., 1999a).
We have made similar observations on the NF-B-dependent, highly inducible IL-8 and RANTES promoters (Casola et al., 2000a,b). For example, in the case of IL-8, although NF-B is absolutely required for inducible transcriptional activation (mutation of the NF-B-binding site renders the promoter inert to any stimulation), an additional upstream site containing the IRF-1/7 complex is required for activation of the gene (Casola et al., 2000a). Interestingly, gene-transfer experiments also demonstrate that the requirement for "ancillary" cis elements (e.g., the IRF-binding site in viral stimulation) is distinct from those required for activation by TNF (Casola et al., 2000a). Together, these data indicate that the actions of NF-B are determined by presence of ancillary binding sites in the target promoter and constrained by the stimulus applied.
Enhanceosome formation involves coactivator recruitment, proteins required for control of core promoter activity at a distance. Coactivators are non-DNA binding proteins that allow transcription factors to couple with the basal transcriptional apparatus and induce chromatin remodeling (Blobel, 2000). Chromatin remodeling is an active process occurring as a first step in promoter activation, known as "de-repression" (Jones and Kadonaga, 2000). This adenosine triphosphate (ATP)-dependent event is mediated through enzymatic (histone acetyltransferase, HAT) activity contained in the p300/CBP and p300/CBP-associated factor (p/CAF) coactivators. HATs acetylate basic residues on core histones, weakening their binding, thereby allowing other components of the transcriptional machinery to access the target promoter (Korzus et al., 1998; Blobel, 2000). Although there is redundancy in coactivator activity, certain types of coactivators are known to preferentially mediate genomic actions of transcription factor classes. Of relevance here, Rel A recruits the p300/CBP coactivator through interaction with its COOH transactivation domain (Perkins et al., 1997; Wadgaonkar et al., 1999). Antibody-injection experiments have shown that the HAT activity of the related coactivator p/CAF is also important in NF-B transactivation. Moreover, the mechanism for coactivator recruitment appears to be distinct between NF-B1 and Rel A-complexed DNA-binding sites. The steroid receptor coactivator 1 (SRC-1), originally thought to be a specific nuclear receptor coactivator, transactivates NF-B1 binding sites in synergy with p300/CBP (Na et al., 1998). SRC-1 specifically binds NF-B1 but not Rel A (Na et al., 1998), strongly suggesting that alternative pathways for coactivator recruitment and promoter activation are utilized, depending on the composition of the NF-B complexes assembled on a specific site. The individual roles of the various coactivators have yet to be determined.
Some recent intriguing work in macrophages has indicated that chromatin organization controls the temporal access of NF-B to its target promoters (Saccani et al., 2001). Chromatin ImmunoPrecipitation (ChIP) assays involve the specific immunoprecipitation of genes reversibly cross-linked to NF-B that are detected by PCR. Using ChIP assays on lipopolysaccharide-stimulated macrophages, two distinct waves of NF-B recruitment to target genes were observed. NF-B was recruited rapidly (i.e., within 20 minutes of stimulation) to the IB, MIP-2, and manganese superoxide dismutase gene promoters, constituting the "constitutively and immediately accessible" (CIA) genes. CIA genes are stably associated with acetylated histone H4. Conversely, a second wave of NF-B recruitment was observed on the RANTES, MCP, and IL-6 promoter, constituting the "regulated late accessibility" (RLA) genes. The RLA genes were not constitutively bound by acetylated histone H4 but were induced to bind acetylated H4 prior to detectable NF-B recruitment (Saccani et al., 2001). Genes in the RLA group, although activated by NF-B, apparently had to undergo chromatin remodeling event prior to NF-B binding. The mechanisms and pathways controlling chromatin remodeling are incompletely understood; however, these workers observed that recruitment of activator protein-1 (AP-1) complexes temporally preceded that of NF-B, suggesting that AP-1 may be inducing chromatin remodeling in a subset of RLA promoters. These observations suggest unanticipated complexity in NF-B action by regulating its accessibility to target promoters to certain chromatin domains.
|
II. Approach for Experimental Identification of Genetic Networks Downstream of NF-B
|
|---|
Although NF-B-binding sites now can be identified with some degree of reliability, it is clear that the actions of NF-B are controlled by the promoter context in which it binds, the stimulus applied, and the chromatin structure of the target promoter. Our general strategy was to employ a nonbiased, high-throughput methodology to identify genes downstream of NF-B in response to a complex biological stimulus, schematically diagrammed in Figure 1. We reasoned that comparing high-density oligonucleotide arrays on control or stimulated cells in the absence or presence of a specific inhibitor of NF-B activation would identify those genes downstream of NF-B.
For this approach, we required a cell that could be manipulated to express a dominant-negative NF-B inhibitor. Because NF-B is required for normal cell growth and survival (NF-B Rel A deficiency in mice is embryonic lethal due to massive hepatic apoptosis) (Beg et al., 1995) and adenovirus-mediated transduction could activate signaling pathways in infected cells, we developed a tetracycline-regulated cell system (Tet-Off) (Gossen and Bujard, 1992). This cell system expresses the nondegradable, epitope-tagged IB (IB Ser32Ala/Ser36Ala, termed FLAG-IB Mut) under control of the Tet Operator sequences (TetO). FLAG-IB Mut contains site mutations in the serine phosphoacceptor sites of IKKß and functions as a potent dominant-negative inhibitor of NF-B activation (Thomas et al., 1998b). In the Tet-Off stable cells, we were able to inhibit expression of FLAG-IB Mut during the selection process by including low concentrations of doxycycline (Dox) in the selection medium. Dox binds the tetracycline transactivator (tTA) and inhibits its binding. Upon Dox withdrawal, tTA binds DNA (TetO) and activates expression of the target gene (Figure 2). Using this strategy, we successfully isolated stable clonal cell lines where FLAG-IB Mut expression was induced strongly after Dox withdrawal from the culture medium. For example, in Figure 3, a western immunoblot of cytoplasmic extracts was performed using antibody recognizing the FLAG epitope tag of the FLAG-IB Mut fusion protein. Highly inducible and tightly regulated expression of FLAG-IB Mut was seen with early passages after stable clones were isolated. Importantly, in the short durations used for expression and stimulation, these cells did not have detectable differences in spontaneous or viral-induced apoptosis (Tian et al., 2002). FLAG-IB Mut expression potently inhibited NF-B-dependent transcription in response to the prototypical NF-B-activating cytokine, TNF, and in response to respiratory syncytial virus (RSV) infection (Figure 4). Figure 4A shows an electrophoretic gel mobility-shift assay (EMSA) of nuclear extracts taken from cells stimulated with TNF in the absence of the dominant-negative inhibitor (+ Dox) vs. those expressing the dominant-negative inhibitor (-Dox). In these cells, TNF is a potent inducer of Rel A NF-B1 binding (Garofalo et al., 1996; Brasier et al., 1998; Tian et al., 2002). (These sequence-specific binding complexes previously have been shown to be subunit specific using supershift assays and microaffinity isolation.) Figure 4B shows a western immunoblot of cytoplasmic extracts from the same cells (as used in the EMSA in Figure 4A) staining with antibody that recognizes IB. Expression of the dominant-negative IB (the slower migrating band) is induced strongly in the absence of Dox at levels comparable to endogenous IB. Similarly, although endogenous IB is degraded proteolytically by TNF, the FLAG-IB Mut protein is not. Finally, RNAse protection assays demonstrated that sufficient NF-B inhibition was achieved to interfere with expression of known NF-B-dependent target genes (Figure 5). In this experiment, cells cultured in the absence or presence of Dox were infected with RSV and steady-state levels of RANTES, IP-10, and IL-8 were measured simultaneously by ribonuclease protection assay (RPA). The strong viral-inducible and time-dependent induction of RANTES, IP-10, and IL-8 was inhibited significantly in cells cultured in the absence of Dox. Control experiments, published earlier, indicated that similar levels of viral transcription was occurring in both cells (Tian et al., 2002). Together, these data indicated that we could achieve a robust and highly reproducible inhibition of NF-B-dependent translocation and nuclear actions in response to a variety of stimuli.
View larger version (73K):
[in this window]
[in a new window]
|
FIG. 4. Dox-dependent inhibition of NF-B activation and IB proteolysis. (A) Time course of NF-B binding in electrophoretic mobility shift assay (EMSA). Sucrose cushion-purified nuclear extracts were prepared from tumor necrosis factor alpha (TNF)-stimulated cells cultured in the absence or presence of Dox for the time indicated. EMSA was performed using the 32P-labeled acute-phase response element, wild-type (APRE WT) duplex oligonucleotide probe (Brasier et al., 2001). The complexes were fractionated on 6% native polyacrylamide gels; an autoradiographic exposure is shown. The relative migration of the RelA/NF-B1 ("p50") heterodimer and the p50 homodimer (p50)2 complexes are indicated (Jamaluddin et al., 1998). Expression of FLAG-IB Mut completely inhibits RelA/NF-B1 binding. (B) Resistance of FLAG-IB Mut to TNF-induced proteolysis. Western blot of cytoplasmic extracts prepared from cells cultured in the presence or absence of Dox and stimulated for the indicated times with TNF. Western blot was probed with antibody to IB, detecting the FLAG-IB Mut (top band) and endogenous IB (bottom band). FLAG-IB Mut inhibits the expression of endogenous IB and is not proteolyzed by TNF treatment. Control for protein loading was performed by reprobing the membrane for ß-actin (bottom). [Data adapted from Tian B, Zhang Y, Luxon BA, Garofalo RP, Casola A, Sinha M, Brasier AR 2002 Identification of NF-B dependent gene networks in respiratory syncytial virus-infected cells. J Virol 76:6800–6814.]
|
|
View larger version (55K):
[in this window]
[in a new window]
|
FIG. 5. Effect of FLAG-IB Mut expression on NF-B-dependent transcription. Cells were cultured in the absence or presence of Doxy for 7 days and infected with purified respiratory syncytial virus (RSV) (multiplicity of infection (MOI 1)) for indicated times (in hours) prior to harvest of total cellular RNA. mRNA abundance was determined in each sample by multiprobe ribonuclease protection assay (RPA). Shown is an autoradiographic exposure after denaturing gel electrophoresis. Location of protected fragments for RANTES (regulated upon activation, normally T-cell expressed and presumably secreted), IP-10, interleukin (IL)-8, and the two housekeeping genes, L32 and glyceraldehye-3-phosphate dehydrogenase (GAPDH), is indicated at left. FLAG-IB Mut inhibits viral-inducible RANTES, IP-10, and IL-8 expression. [Data adapted from Tian B, Zhang Y, Luxon BA, Garofalo RP, Casola A, Sinha M, Brasier AR 2002 Identification of NF-B dependent gene networks in respiratory syncytial virus-infected cells. J Virol 76:6800–6814.]
|
|
With this information, high-density oligonucleotide arrays were used to assay for profiles of gene expression in response to RSV infection in the presence of FLAG-IB Mut vs. those induced in its absence.
|
III. Identification of NF-B-dependent Gene Network in Viral Infection
|
|---|
We were interested in identifying NF-B-dependent genes in response to a complex biological stimulus, mucosal viral infection with RSV. RSV, a negative-sense RNA virus of the paramyxoviridae family, is the leading cause of epidemic bronchiolitis and pneumonia in children (Ruuskanen and Ogra, 1993). Lacking an effective vaccine, infection with this ubiquitous virus causes 40–60% of the bronchiolitis and 15–25% of the pneumonia cases in hospitalized children (Shay et al., 1999), accounting for 100,000 hospitalizations and 500 deaths annually in the United States (Shay et al., 2001). RSV productively replicates at high levels in a variety of airway epithelial cells, where it potently alters host gene expression. Our work using high-density oligonucleotide arrays has shown that at least 16 different C, CC, CXC, and CX3C chemokines are expressed by RSV-infected respiratory epithelial cells in three general expression profiles (Garofalo et al., 1996; Casola et al., 2000b; Zhang et al., 2001). Of these, several are known to be induced through an NF-B-dependent mechanism involving IB proteolysis (Garofalo et al., 1996; Jamaluddin et al., 1998). Importantly, inhibition of NF-B prevents RSV-induced airway inflammation in experimental models of infection (Haeberle et al., 2002). Together, we interpret these data to indicate that the actions of NF-B are to induce the expression of genes important in airway inflammation.
To identify these potential pathology-inducing genes, we subjected RNA for high-density oligonucleotide array analysis extracted from control or infected cells (± Dox treatment). For comparison of the fluorescent intensity (average difference) values among multiple experiments, we normalized the data relative to the 2% trimmed mean, a global hybridization metric to correct for changes in hybridization efficiency (Zhang et al., 2001; Tian et al., 2002). The normalized average difference values were then subjected to a two-way analysis of variance (ANOVA with replications) to determine which genes were significantly influenced by either the RSV or Dox treatment. We found that RSV infection significantly changed the abundance of 1359 mRNAs, while FLAG-IB Mut expression influenced only 380 gene-expression profiles. Comparison of the two groups found only 144 genes common to the two treatments. The proteins encoded by the genetic group sensitive to RSV + FLAG-IB Mut expression were classified by their putative biological pathways (Table I). Notably, members of a wide variety of biological pathways were identified. Numerically, the largest groups included membrane proteins, metabolic enzymes, signaling molecules (i.e., kinases, phosphatases), transcription factors, and those of unknown function.
To better visualize the gene-expression changes in the 144-member group regulated by both RSV and FLAG-IB Mut expression, hierarchical clustering was performed. In this technique, each gene-expression profile is grouped with its nearest neighbor and the mathematical proximity of this gene-expression profile is indicated by the height of a common line that connects the two nodes. A representative analysis is shown in Figure 6, where each treatment condition represents the average of the normalized average difference values from three independent experiments. From visual inspection of the dendogram, the gene-expression patterns are divided into two large clusters. The top cluster contains a group of genes whose constitutive (unstimulated) expression is high and falls with RSV infection. In this group, inhibition of NF-B reduces the viral-induced inhibition of their expression. The bottom cluster contains a group of genes whose constitutive expression is low and is strongly increased by RSV infection. In this group, inhibition of NF-B blocks the viral-induced activation of their expression.
The genes with high constitutive activity whose expression fell with RSV treatment were classified into biochemical pathways to determine whether a single process was being influenced by NF-B action. Disappointingly, approximately half of the genes were unknown and, of those that could be identified, no signal pathway emerged. For example, we found secreted peptides (growth hormone, Genbank Accession No. J03071), intracellular signaling proteins (tyrosine phosphatase, X54131; serine/threonine kinase, AB018324; RAB4, M28211), nucleic acid binding proteins (RNPA0, U23803), metabolic enzymes (enolase, X56832; selenium donor, U34044), transcription factors (micro-opthalmia-associated transcription factor, AB006909), extracellular structural proteins (type XI collagen, J04177; 1-type XVI collagen, M92642), and others (Tian et al., 2002). Control experiments also indicated that expression of other genes in this list was influenced by Dox treatment. The mechanisms for NF-B-dependent gene downregulation will require independent analysis and further study.
A similar analysis was conducted for the genes whose constitutive activity was low and whose induction by RSV was blocked by FLAG-IB Mut expression. We identified secreted complement factors (complement B), transcription factors (the proteolytically processed NF-B member, NF-B2; the signal transducer and activator of transcription, STAT-1; the interferon regulatory factors, IRF-9 and IRF-7), metabolic enzymes (5'-aminolevulinate synthase); growth factor binding proteins (insulin-like growth factor binding protein 6), kinases in translational control (human GCN1), cytokines (RANTES, IL-8), and others (B94). The group indicated by the parentheses in Figure 6 is also tabulated (Table II). Again, no single biological pathway is easily seen, although identifying some of the previously known NF-B-dependent genes (e.g., IL-8, RANTES) was an important validation of our system and analysis (Figure 5). Together, these data suggest that NF-B plays a multifaceted role in control of constitutive gene expression, involvement in RSV-mediated downregulation of genes, as well as in mediating upregulation of a network of RSV-inducible genes.
|
IV. Insights into Biological Roles of NF-B Through the Functions of Its Downstream Genes
|
|---|
RSV is a negative-sense RNA virus that is replicated in the cytoplasm through its own RNA-dependent RNA polymerase. In RSV infection, therefore, NF-B has no role in viral transcription and its purpose is to activate expression of the host’s innate immune and inflammatory responses. A number of highly inducible genes encoding cytokine, chemokine, acute-phase reactant, and adhesion molecules contain NF-B-binding sites in their proximal promoters and, in a certain number of cases, are induced by RSV (Li and Brasier, 1996; Brasier et al., 1998; Thomas et al., 1998a; Casola et al., 2001; reviewed in Zhang and Ghosh, 2001). However, whether NF-B is truly a master regulator of gene expression in mucosal viral infection and which genes it controls have not been subjected to direct experimental investigation. A surprising finding from our study is the relatively small number of genes that were sensitive to FLAG-IB Mut expression. Of the 1359 RSV-inducible genes, only about 10% were shown to be NF-B dependent. It is important to emphasize that our analysis is based on whether a given gene is induced by virus for us to identify it as an NF-B target. In this regard, a study using a similar approach to identify NF-B-dependent genes in rat pancreatic ß cells identified 66 genes under NF-B control (Cardozo et al., 2001). The NF-B-dependent targets that this study identified in pancreatic ß cells were markedly different from those found in our study. In ß cells, the most highly upregulated genes are inducible peptide 10 kDa (IP-10), IL-15, cluster of differentiation-40 (CD-40), and major histocompatibility complex (MHC)-II-associated invariant chain. Like our findings, a number of genes were downregulated in an NF-B-dependent manner, including the Pdx-1 transcription factor, a ß-cell differentiation factor that controls the expression of glucose transporters (Cardozo et al., 2001). These surprising differences between our studies indicate that the "NF-B network" is under cell type-specific control. Also, species-specific effects come into play as, for example, rats do not have an IL-8 gene.
In human epithelial-derived cells, our studies show that NF-B is important in RSV-inducible gene expression for a number of genes with apparently diverse functions, including chemokines, transcription factors (interferon regulatory factor (IRF)/signal transducers and activators of transcription (STATs) and NF-B/IB members), proteins controlling translation/proteolysis, secreted proteins, cytoskeletal elements, and signaling proteins in those whose function is known (Table I). Clearly, these observations will have to be extended for other NF-B-activating stimuli and into other cell types. Our observations further suggest that not only does NF-B play a role in viral-dependent gene activation, it also plays a role in constitutive and viral-dependent inhibition of distinct genetic elements. In the following paragraphs, we comment on some of the biological insights that this study suggested (see Figure 7 for a schematic overview).
View larger version (27K):
[in this window]
[in a new window]
|
FIG. 7. Putative biological functions for NF-B-dependent genes. Abbreviations: GRO, growth-regulated oncogene; IAP, inhibitor of apoptosis; IGFBP, insulin-like growth factor binding protein; IRF, interferon regulatory factor; ISGF, interferon-stimulated gene factor; TNFR, tumor necrosis factor receptor. [See text for discussion.]
|
|
A. CONTROL OF NEUTROPHILIC AND EOSINOPHILIC CHEMOKINE EXPRESSION
Cellular recruitment into the virally infected lung is a multistep process involving adherence of circulating leukocytes to an activated endothelial surface, followed by their diapedesis into the inflamed tissue and migration toward chemical gradients of chemoattractant peptides or antigens (reviewed in Springer, 1994). Recent attention has focused on the important role of chemokines in mediating leukocyte chemotaxis into the airways. Chemokines are a superfamily of proteins divided into four distinct groups — C, CC, CXC, and CX3C — based on the number and spacing of highly conserved NH2-terminal cystine residues (reviewed in Oppenheim et al., 1991; Baggiolini et al., 1997). This grouping has functional significance because the spacing of the NH2-terminal cystine residues influences the type of cell-surface leukocyte receptors that they bind to produce leukocyte activation and chemotaxis. Although a number of cell types can inducibly secrete chemokines, the airway epithelium is thought to play a central role in initiating pulmonary inflammation by RSV because the epithelium is the first target for RSV replication and the mucosa is the only tissue that allows productive viral replication. Our recent studies using high-density oligonucleotide arrays have shown that at least 16 different C, CC, CXC, and CX3C chemokines are expressed by RSV-infected respiratory epithelial cells in three general expression profiles (Zhang et al., 2001). The molecular mechanisms controlling expression of CXC- and CC-type chemokines in RSV-infected airway epithelial cells have been investigated extensively (Garofalo et al., 1996; Brasier et al., 1998; Thomas et al., 1998a; Casola et al., 2000b). Our studies here show that the NF-B-dependent chemokine group contains the CXC chemokines GRO- and IL-8 and the CC chemokine RANTES. The CXC chemokine class is potent chemotactic cytokines for activated neutrophils and is functionally subdivided into two groups, based on the presence or absence of a signature ELR (Glu-Leu-Arg) motif upstream of the canonical CXC motif. The ELR-containing group of CXC chemokines includes IL-8, GRO-, and ENA78. These cytokines primarily activate the bacteriocidal activity and chemotaxis of neutrophils, an abundant cell type found in the bronchoalveolar lavage of intubated, RSV-infected children (Everard et al., 1994). The ELR CXC chemokines can activate other target cells. For example, IL-8 activates T cells and eosinophils (Matsushima and Oppenheim, 1989; Murphy, 1994), while GRO-/ß/ activate basophils (Geiser et al., 1993). This may account for some of the spectrum of cellular infiltration in RSV-infected lungs and presence of cell-specific degranulation products in nasopharyngeal secretions in patients with naturally acquired RSV infection (Garofalo et al., 1992).
The CC chemokine RANTES, originally described as a T cell-derived chemokine, is now known to be highly expressed in stimulated airway epithelial cells, where it may serve to recruit T lymphocytes, monocytes, basophils, and eosinophils into the virally infected lung (Alam et al., 1994). RANTES has been shown to be highly NF-B dependent by our studies and those of others (Thomas et al., 1998a; Casola et al., 2000b). Of relevance, RANTES has been shown to be present in nasal washes of children infected with RSV, where its levels and those of its cellular target, eosinophilic degranulation products, correlate with disease severity (Garofalo et al., 1994; Teran et al., 1996). Our studies suggest that NF-B controls the expression of chemokines important in neutrophilic and eosinophilic recruitment into the airways.
B. CONTROL OF IFN SIGNALING PATHWAY
We were surprised that the NF-B-dependent gene list contained important members of two transcription factor families, STAT-1 and IRF-9 and IRF-7B. STATs are cytosolic proteins contained in high molecular mass complexes ("statosomes") (Ndubuisi et al., 1999) activated by tyrosine phosphorylation mediated by the interferon and growth factor receptor-associated kinases, Jak and Tyk (reviewed in Schindler and Darnell, 1995). Activated STATs form homo- or heterodimers through intermolecular SH2 phosphotyrosine interactions and subsequently are translocated into the nucleus in distinct binding complexes dictated by the nature of the stimulus and target promoters (Taniguchi et al., 2001). The IRFs are a family of nine constitutive and inducible helix-turn-helix transcription factors important in innate defense through their ability to mediate viral-inducible transcription of cytokine and chemokine genes (reviewed in Taniguchi et al., 2001). IRF-9 is a DNA-binding component of the ISGF3 complex induced by -IFN. ISGF3 contains STATs — 1 and 2 — as well as IRF-9. The observation that two important components of ISGF3 are controlled by NF-B highlights an interdependent relationship between these transcription factor families. Perhaps not surprisingly, others have shown the IRF-7 promoter contains an NF-B-binding site, mediating its response to cytokine stimulation (Lu et al., 2002). STAT1, to our knowledge, is not known to be NF-B dependent. However, our findings that the STAT/IRF family members are downstream of NF-B in viral infection suggests an important mechanism for how NF-B plays an important role in innate immunity by facilitating signaling through the interferon pathway.
C. NF-B AUTOREGULATION
Autoregulation recently has been recognized to be an important controller of network stability. Studies of synthetic gene circuits have shown, for example, that negative feedback reduces the variability in target gene expression by limiting the range over which the concentrations of the signaling molecules operate (Becskei and Serrano, 2000). Remarkably, in our analysis, members of the NF-B activation pathway themselves are NF-B dependent. Of the members of the NF-B pathway identified in this study, the inhibitor of apoptosis (IAP-1)/cIAP-2 is an upstream regulator of caspase activation that associates with the tumor necrosis factor receptor 1 (TNFR1) and death domain-containing receptors. Inducible expression of IAP-1 may be one mechanism through which NF-B exerts its antiapoptotic effect. In support of our analysis, IAP-1 previously has been shown to be transcriptionally regulated by NF-B (Hong et al., 2000). The DNA-binding subunit, NF-B2, is an NF-B1 homolog encoded by a large, 105-kDa precursor that must be processed into its 50-kDa binding form. Like IAP-1, NF-B2 is known to be regulated by NF-B (Liptay et al., 1994). Although both NF-B1 and NF-B2 are inducible genes, in RSV-infected epithelial cells, expression of NF-B2 is predominately NF-B dependent. Unlike NF-B1, NF-B2 is a potent transcriptional activator in heterodimeric form with Rel A (Schmid et al., 1991). These observations suggest that epithelial stimulation by NF-B-activating agents changes the distribution of Rel A heterodimers, allowing the cell to respond differently to subsequent signals. Another well-established NF-B-dependent target is that of the IB inhibitors themselves (Han et al., 1999a; Brasier et al., 2001). Activation of the IB members is an autoregulatory feedback loop where NF-B induces the synthesis of its own inhibitor to terminate its action. Previously, we showed that the NF-B-IB autoregulatory loop was impaired in RSV-infected cells. This finding explains why IB was not identified in the present analysis (Jamaluddin et al., 1998). IB
, in contrast to IB, is strongly upregulated by RSV infection and suggests stimulus-specific differences in the NF-B-dependent expression control of individual IB subunits. Notwithstanding, the induction of IB may compensate for a relative deficiency in IB. The existence of multiple, independent NF-B-IB and NF-B-BCL-3 inhibitory loops suggests that unregulated NF-B activation is highly deleterious (Jamaluddin et al., 1998; Han et al., 1999a; Brasier et al., 2001).
D. OTHER FUNCTIONS
Our study has identified a number of other highly NF-B-dependent genes that do not easily fit into a single biological pathway (Table I). The functional consequences of enhanced expression of 5' aminolevulinate synthase, a rate-limiting enzyme in heme biosynthesis, is unknown to us but suggests a role for NF-B in viral-regulated heme metabolism in nonerythroid cells. Conversely, the NF-B dependence of the E2-ubiquitin-conjugating enzyme suggests that NF-B activation may have an important role in determining cellular capacity to break down proteins regulated through the ubiquitin-proteasome pathway, a process important in cell-surface presentation of viral antigens in the context of MHC class I molecules. We were surprised to identify a human homolog of the Saccharomyces cerevisiae GCN1 gene, a protein controlling translational efficiency through modifying upstream activation of the eIF2 protein kinase (Marton et al., 1997). Gene expression/regulation studies of hGCN1 in the setting of viral infection have not been reported to our knowledge, although viral infections are known to profoundly influence translational regulation. Complement factor B, a hepatic acute-phase response factor important in the alternative complement pathway, is well known to be NF-B inducible (Nonaka and Huang, 1990). However, viral induction of the alternative complement pathway and its role in response to infection have not been investigated. The induction of insulin-like growth factor binding protein 6 (IGF-BP6) suggests that viral-infected cells exert paracrine control on the mitogenic actions of IGF-II (Gabbitas and Canalis, 1997). Perhaps IGF-BP6 expression is beneficial to prevent local cellular proliferation in the presence of an infecting viral agent. Upregulation of the subunit of the IL-15 receptor may suggest that viral-infected cells have distinct signaling phenotypes to cytokines as a result of NF-B action. Finally, B94 was identified as a highly cytokine- and lipopolysaccharide (LPS)-inducible transcript (Sarma et al., 1992) that may play a role in angiogenesis, spermatogenesis, or myelogenesis. Our data suggest that B94 is also a viral-inducible transcript in epithelial cells through an NF-B-dependent mechanism. The function of B94 in this context is unknown.
Recent work suggests that NF-B has genomic actions, even in the absence of exogenous stimuli. This "constitutive" NF-B activation may be important to inhibit apoptosis (reviewed in Barkett and Gilmore, 1999). For example, NF-B appears to be required to maintain low levels of the Bcl-2 protein, A1, at levels to prevent loss of mitochondrial transmembrane potential and apoptosis in macrophages (Pagliari et al., 2000). Alternatively, constitutive NF-B activity may be important in cellular immortalization (Arsura et al., 2000). Although our experimental design cannot distinguish these or other potential roles for constitutive NF-B activity, we interpret our data to mean that constitutive NF-B appears to downregulate expression of the collagen genes and others. The mechanism (transcriptional or post-transcriptional) by which NF-B influences the abundance of these genes will require further investigation. In this regard, we note that a recent study has implicated NF-B in the post-transcriptional control of MyoD mRNA abundance (Guttridge et al., 2000); perhaps collagen is regulated in a similar way.
|
V. Expression Kinetics of the NF-B-inducible Genetic Network
|
|---|
We preliminarily analyzed the kinetics of gene expression for the highly upregulated NF-B-dependent genes taken from Table I. For this analysis, microarray data from our previously established database of RSV infection was taken corresponding to type II-like alveolar epithelial cells (A549) (Zhang et al., 2001) and laryngeal carcinoma cells (Hep2) (B. Tian, unpublished data). A hierarchical clustering/heat map analysis was performed and is presented in Figure 8. In Figure 8A, the A549 time course shows that several of the NF-B-dependent genes exhibit anomalous behavior. For example, an "unknown" gene, cytochrome P (CYP)11B1, melanoma antigen-encoding gene (GAGE)-2, and tyrosine phosphatase were not expressed by A549 cells. In addition, the gene encoding a putative membrane protein was paradoxically regulated, being downregulated by RSV infection in A549 cells. The remaining genes were activated from 6–12 hours after RSV infection, a time when induction of NF-B binding can be demonstrated by EMSA and immunofluorescence analysis (Garofalo et al., 1996; Tian et al., 2002). Figure 8B shows a similar analysis for the kinetics of NF-B-dependent gene induction in Hep2 cells. Like A549 cells, the genes "unknown," p450, and GAGE-2 were not expressed; unlike A549 cells, however, the melanoma growth stimulatory activity/growth-regulated oncogene (MGSA/GRO)- and cholesterol hydroxylase genes were not expressed in the Hep2 cells. The remaining genes were regulated by RSV infection in a manner quite similar to that seen in A549 (compare Figure 8B with 8A). Together, these data indicate that cell type has a strong influence on the composition of a genetic network.
View larger version (50K):
[in this window]
[in a new window]
|
FIG. 8. Kinetics of NF-B-dependent gene expression in epithelial cells. (A) A549 cells. Clustering and heat map analysis of the kinetics of NF-B-dependent gene expression in type II-like alveolar cells following RSV infection. Data represent the mean of three independent time courses analyzed by high-density microarrays. Agglomerative hierarchical clustering was performed on average fluorescence intensity values from three independent experiments using the UPGMA technique. A heat map for each gene for the three independent experimental datapoints is shown at right. The color gradient used is dark gray (minimum value of 5 scaled fluorescence intensity units), light gray (middle value of 5000 scaled units), and white (maximum value of 10,000 scaled units). CYP, cytochrome P; GAGE, melanoma antigen-encoding gene; MGSA, melanoma growth stimulatory activity. [Data from Zhang Y, Luxon BA, Casola A, Garofalo RP, Jamaluddin M, Brasier AR 2001 Expression of RSV-induced chemokine gene networks in lower airway epithelial cells revealed by cDNA microarrays. J Virol 75:9044–9058.]
|
|
View larger version (38K):
[in this window]
[in a new window]
|
FIG. 8. (B). Hep2 cells. Clustering and heat map analysis of the kinetics of NF-B-dependent gene expression in Hep2 larygeal carcinoma cells following RSV infection. [Data analysis as in Figure 8A.]
|
|
|
VI. Promoter Analysis of NF-B-inducible Genes
|
|---|
The identification of NF-B-inducible genes allows us to glean insights into the actions of this master regulator of inflammation. That these genes are all downstream of the same transcription factor and have similar patterns of expression indicated to us that further analysis of their promoter organizations may yield some unifying insights into the mechanism for NF-B-dependent gene control. One successful approach to identification of common regulatory sequences has been to analyze empirically groups of promoters based on similarities in expression patterns (Tavazoie et al., 1999; Pilpel et al., 2001). In this example, a K-means algorithm was used to cluster groups of genes whose expression was similar in data derived from cell-cycle progression in Saccharomyces. A local alignment tool identified 17 motifs in 12 different clusters, over half of which corresponded to know DNA-binding sites, suggesting that expression coclustering identifies promoters with common regulatory motifs (Tavazoie et al., 1999). To address whether similarities could be found in the NF-B-inducible genes, we retrieved the promoter sequences of the most tightly coclustering expression group (Figures 8A and B). The promoter sequences were annotated and analyzed for the presence of transcription factor-binding sites. Although a number of methods for identifying the presence of transcription factor-binding sites exist, recent algorithms employing positional weight matrices (PWM) have significantly improved the reliability of computational prediction of transcription factor-binding sites. PWMs have improved the accuracy of prediction of specific transcription factor-binding sites by accommodating for the large, base-to-base variability in target binding sites. PWMs assign a weighted score for each nucleotide in the promoter sequence and generate an overall score as the sum of the matrix values. Moreover, the weighted matrix score is related to the Gibbs free energy of binding (Berg and von Hippel, 1987; Stormo, 2000) and therefore has biological significance.
For analysis of factors binding the proximal promoters, binding sites for all eukayrotic transcription factors were predicted using the high-stringency cutoff, to minimize the false positive and negative matches from the TRANSFAC database (Heinemeyer et al., 1999). To correlate these data with experimental determination of binding sites, we compared the TRANSFAC output with our in vitro binding studies on the IL-8 and RANTES promoters. For example, the analysis correctly predicted the known AP-1 (Vlahopoulos et al., 1999) and NF-B binding sites (Brasier et al., 1998) of the IL-8 gene. Similarly, the known NF-IL6 and NF-B binding sites from the RANTES promoter were predicted (Casola et al., 2000b). Computer-identified NF-B binding sites are displayed in Table II. Here, we also searched the literature for previously identified NF-B binding sites for experimental verification where possible. We were able to confirm the putative NF-B-binding sites with the published literature for the GRO- (Wood and Richmond, 1995), IL-8 (Brasier et al., 1998), RANTES (Casola et al., 2000b), IRF-7B (Lu et al., 2002), IAP-1 (Hong et al., 2000), NF-B2 (Liptay et al., 1994), and IL-15R (Mariner et al., 2001) promoters. In the other genes, the absence of NF-B sites may indicate that the binding site is located outside of the 700 bp of 5' flanking sequence that was analyzed or that the gene is indirectly regulated by NF-B. More investigation will have to be done to distinguish between these possibilities.
We then analyzed matches for all predicted eukaryotic transcription factor-binding sites using the most stringent cutoffs for matrix similarity score in the TRANSFAC database. In the sequences analyzed, the predicted binding site density was one transcription factor for approximately every eight base pairs (7.9 ± 1.8). Unfortunately, this density of transcription factor-binding data made it difficult to see relationships between the various promoters. To help identify and visualize binding patterns, we developed an interactive, versatile data visualization/mining tool that we call GeneRep. GeneRep displays the location of transcription factor-binding sites as colored rectangles along their linear sequence. The closeness of the match (the "similarity score") is indicated by the height of the rectangle. To help identify the location of the NF-B binding sites, they appear as solid black in Figure 9. The NF-B2, IL-8, IGF-BP6, and Gro- promoters all contained high-affinity binding sites in the proximal 100 bp of the TATA box. Preliminary analysis did not reveal any consistent relationship of the NF-B-binding site with binding sites for AP-1, IRF, or NF-IL6. More systematic analysis will be required to identify any common features for this subclass of inducible promoters.
|
VII. Conclusions
|
|---|
NF-B is a highly inducible transcription factor that controls hepatic acute-phase response, innate and adaptive immunity, and cellular survival through the induction of genetic networks. We reviewed our work using a high-density microarray analysis of a tightly regulated cell model expressing a specific NF-B inhibitor to identify its downstream gene network. Our findings suggest that NF-B regulates expression of distinct genetic networks of constitutive genes whose expression are inhibited further by viral infection as well as controlling a distinct subset of those that are viral inducible. Moreover, these data suggest that NF-B is a upstream regulator of RSV-inducible gene expression through controlling expression of the proteins involved in interferon signaling (STAT/IRF), perhaps providing insights into mechanisms of how NF-B controls the innate immune response. More work will be required to understand the cell type-specific influences on expression of NF-B networks, so that members of this network can be comprehensively identified. Further analysis is required to understand whether a common promoter architecture of an NF-B-dependent gene can be identified, which genes are indirectly controlled by NF-B, and, in these, the mechanisms for this regulation.
|
ACKNOWLEDGEMENTS
|
|---|
TOP
ABSTRACT
I. Introduction
