|Home | About | Journals | Submit | Contact Us | Français|
Activation of transcription factor NF-κB can affect the expression of several hundred genes, many of which are involved in inflammation and immunity. The proper NF-κB transcriptional response is primarily regulated by post-translational modification of NF-κB signaling constituents. Herein, we review the accumulating evidence suggesting that alternative splicing of NF-κB signaling components is another means of controlling NF-κB signaling. Several alternative splicing events in both the tumor necrosis factor and Toll/interleukin-1 NF-κB signaling pathways can inhibit the NF-κB response, whereas others enhance NF-κB signaling. Alternative splicing of mRNAs encoding some NF-κB signaling components can be induced by prolonged exposure to an NF-κB-activating signal, such as lipopolysaccharide, suggesting a mechanism for negative feedback to dampen excessive NF-κB signaling. Moreover, some NF-κB alternative splicing events appear to be specific for certain diseases, and could serve as therapeutic targets or biomarkers.
Activation of the nuclear factor-κB (NF-κB) signal transduction pathway occurs in response to a wide range of stimuli and results in nuclear accumulation of NF-κB transcription factors causing changes in the expression of target genes involved in innate and adaptive immunity, inflammation, cell survival, hematopoiesis, and lymphoid development (www.nf-kb.org). Identification of human disease mutations in genes encoding NF-κB signaling components and the varied phenotypes seen in NF-κB knockout mouse models have highlighted the importance of precise regulation of the NF-κB pathway (Courtois and Gilmore, 2006; Gerondakis et al., 2006). In this review, we describe alternative splicing events that occur in genes involved in mouse and human NF-κB signal transduction pathways, and discuss the functional significance these splice variants may have for regulation of NF-κB.
Initiation of the NF-κB signaling cascade often begins with cellular recognition of extracellular signals. A large number of ligand-receptor interactions can lead to activation of NF-κB. Two of the most studied receptor families involved in recognition of NF-κB activating signals are the tumor necrosis factor receptor (TNFR) family and the Toll/interleukin-1 (IL-1) (TIR) family (Figure 1).
Tumor necrosis factor (TNF) is a cytokine employed by hematopeitic cells for cell-cell communication and immune responses. TNF is generated by many different cell types, most notably macrophages and monocytes in response to cellular stresses including infection and inflammation. There are two TNF receptors: TNFR1 (p55) which is expressed in a broad range of cell types and the more cell type-specific TNFR2 (p75). The TNFR1 signaling pathway to NF-κB (Figure 1) is known in more detail than the TNFR2 pathway (Bradley, 2008; Wertz and Dixit, 2008). Interaction of TNF with TNFR1 leads to receptor trimerization. Upon trimerization, TNFR1 undergoes a conformational change enabling its cytoplasmic domain to interact with the TNFRSF1A-associated via death domain (TRADD) adaptor protein. In the classical NF-κB signaling pathway, TRADD recruits TNFR-associated factor (TRAF) proteins via TRAF’s C-terminal domain to the receptor complex. TRAF then ubiquitinates the TNF receptor interacting protein (RIP1), which enables the inhibitor of κB kinase (IKK) complex to be recruited to the TNFR complex through binding of the IKK subunit NEMO to ubiquitinated RIP. Activated IKK then phosphorylates inhibitor of κB (IκB), signaling IκB for degradation, which enables NF-κB nuclear localization and activation of pro-survival target genes. The pathway is turned off through the action of deqbiquitinases such as A20 and CYLD, and by activation of the IκBα gene by NF-κB.
The Toll/interleukin-1 (IL-1) receptor family (TIR) consists of Toll-like receptors (TLRs) and IL-1 receptors (IL-1Rs). While the extracellular sequences of the Toll-like and IL-1R receptors differ, their cytoplasmic domains are related by the presence of TIR domains (Hayden et al., 2006). Like TNF, IL-1 is a pro-inflammatory cytokine produced by monocytes and macrophages in response to cell stress and infection. Signaling by IL-1 involves binding to the IL-1R and heterodimerization of IL-1R with the IL-1 receptor accessory protein (IL-1RAcP) (Huang et al., 1997). Signaling through TLRs occurs via recognition of a wide range of ligands that contain pathogen-associated molecular patterns (PAMPs), thus providing a means for specific TLR recognition of both gram-negative and gram-positive bacteria, fungi, and viruses. For example, lipopolysaccharide (LPS) is a component of gram-negative bacterial membranes that signals pro-inflammatory cellular responses through TLR4.
Upon ligand binding and multimerization of the IL-1R and TLR proteins, their cytoplasmic TIR domains recruit the same protein components. In the most common pathway, adaptor proteins including myeloid differentiation primary response gene (88) (MYD88), TIR-domain-containing adapter-inducing interferon-β (TRIF), and TIR domain-containing adaptor protein (TIRAP) are recruited to the receptor complex via the receptor’s TIR domains. MyD88 associates with interleukin-1 receptor-associated kinase 1 (IRAK1) through the adaptor Tollip. IRAK4 then phosphorylates IRAK1, triggering IRAK1 autophosphorylation. Phosphorylated IRAK1 dissociates from the receptor complex and binds TRAF6. The IRAK1/TRAF6 complex activates the TAK1/TAB1 complex, which in turn activates IKK and thus, induces NF-κB. Nuclear NF-κB transcription factors then regulate target genes involved in cellular defense against the inducing immunological or inflammatory challenge.
The IKK complex contains the two kinase subunits, IKKα and IKKβ,and the regulatory NEMO (aka IKKγ) subunit. In the classical NF-κB signaling pathway, the activated IKKβ subunit phosphorylates the IκB protein, signaling it for ubiquitination and proteasomal degradation. The alternative NF-κB pathway proceeds via IKKα-mediated phosphorylation of C-terminal IκB-like sequences in p100. The phosphorylated C-terminal sequences of p100 are then processed by the proteasome, yielding the mature N-terminal p52 DNA-binding subunit. In both pathways, NF-κB dimers, now liberated from the inhibitory effects of IκB, then enter the nucleus and regulate transcription of specific target genes.
In addition to the primary IKK/IκB-mediated regulation of NF-κB, the activities of proteins in these signaling cascades undergo a multitude of post-translational modifications including phosphorylation, ubiquitination, sumoylation, S-nitrosylation, acetylation, and cysteine oxidization that can further impact NF-κB signaling (Perkins, 2006). Alternative splicing is a critical cellular process that can generate a large array of mRNA transcripts and protein isoforms from a limited number of genes. Several studies have indicated that alternative splicing of human genes is pervasive, and estimates are that 30–75% of all human genes undergo alternative splicing (Mironov et al., 1999; Brett et al., 2000; Kan et al., 2001; Modrek et al., 2001; Johnson et al., 2003). Alternative splicing can provide a means of changing protein functionality by deletion or insertion of protein components. Not surprisingly, these alternatively spliced protein isoforms often show protein-protein interactions, subcellular localization, stability, DNA binding, and enzymatic properties that differ from their normally transcribed/translated counterparts (López et al., 1995; Stamm et al., 2004). Furthermore, changes in protein function due to alternative splicing have been implicated in a number of human diseases including cystic fibrosis, hemophilia, breast cancer, and prostate cancer (Cáceres et al., 2002; Brinkman, 2004; Skotheim and Nees, 2007). In this review, we describe several alternative splicing events that may regulate NF-κB signaling.
TNFR2 plays a crucial role in the proliferation of several hematopoietic cell types including natural killer (NK) cells, B cells, and mature T cells. While physiological levels of TNFR1 can induce NF-κB, the role of TNFR2 in NF-κB signaling is less clear (Bradley, 2008), as activation of NF-κB is only seen with overexpressed TNFR2. An mRNA variant of human TNFR2, termed icp75TNFR, is generated via an alternative transcriptional start site resulting in mRNA lacking the normal 5’ untranslated region (UTR) and exon 1 (Seitz et al., 2001). As such, icp75TNFR has an alternative first exon, termed 1a, which encodes a distinct 27 amino acid (aa) N terminus with the remaining sequences having identity to the normal TNFR2. This alternative splice removes the normal N-terminal signal sequence and, as a consequence, the icp75TNFR protein shows a cytoplasmic localization, which differs from the transmembrane localization of TNFR2. Expression of icp75TNFR mRNA is lower than TNFR2 and appears to be cell type-specific: icp75TNFR is found in HeLa, Kym-1, THP-1, HMEC cells, but not in HEK293 and HepG2 cells. Furthermore, expression of icpTNFR is increased by LPS treatment in both HUVEC and U937 cells. The identification of endogenous icp75TNFR protein has been difficult, due to its low mRNA expression level and the small difference in expected protein size from wild-type (wt) TNFR2. In contrast to wt TNFR2-transfected cells, TNF can induce potent activation of NF-κB in icp75TNFR-transfected cells (Seitz et al., 2001), suggesting that the icp75TNFR has enhanced NF-κB-inducing activity.
In mice and humans, the TRAF family consists of six members, TRAF1-6. TRAF proteins contain a RING finger domain that functions as a ubiquitin ligase, several zinc fingers, and a C-terminal TRAF domain that mediates TRAFs interaction with TNFR, TRADD, and RIP. TRAF2 and TRAF6 participate in both the JNK and NF-κB signaling pathways. The murine Traf2 gene undergoes alternative splicing (Brink and Lodish, 1998). By RT-PCR analysis of mouse kidney cDNA, a Traf2 variant cDNA (Traf2A) containing a 7-aa insertion within the RING finger domain was identified; however, RT-PCR analysis of murine tissues revealed that Traf2A is also expressed in a number of other tissues. Traf2A expression is highest in the brain and spleen; in fact, in the spleen wt Traf2 and Traf2A are expressed at approximately equal levels. In contrast to wt Traf2, overexpression of Traf2A is incapable of inducing NF-κB activity; thus, Traf2A behaves much like a previously described dominant-negative Traf2 RING finger deletion mutant (Rothe et al., 1995). Furthermore, the Traf2A protein has a shorter half-life than wt Traf2 (Grech et al., 2000). One model for Traf2A function is that during periods of excessive TNF signaling, Traf2A acts as a dominant-negative adaptor to reduce normal Traf2-mediated activation of NF-κB. Traf2A cDNA has not, however, been detected in either rat or human tissue (Grech et al., 2000). Given the potential regulatory role of Traf2A in TNF signaling in mouse cells, the identification of other alternative splicing events within the Traf2 RING finger domain in the human TRAF2 gene would be of interest.
TRAF3 is an adaptor that interacts with the TNFR-like receptor CD40. The exact role of TRAF3 in normal CD40-based activation of NF-κB signaling is unclear. Traf3−/− B cells show a normal proliferative response to CD40 ligand which is generally thought to require NF-κB (Xu et al., 1996), and overexpression of wt TRAF3 fails to induce NF-κB in reporter gene assays (Rothe et al., 1995; Takeuchi et al., 1996; Dadgostar and Cheng, 1998). Nevertheless, some of the phenotypes of Traf3 knockout mice suggest that Traf3 is involved in NF-κB signaling (Xu et al., 1996).
There are several human TRAF3 splice variants, suggesting that the cellular regulation of TRAF3 protein function is quite complex. Eight TRAF3 splice variants, all with alterations in the zinc-finger domain were cloned from human Jurkat T-cell mRNA (Sato et al., 1995; Krajewski et al., 1997; Eyndhoven et al., 1998; Eyndhoven et al., 1999). RT-PCR analyses of tonsilar B cells and a panel of human B- and T-cell lymphoma lines show expression of all eight TRAF3 splice variants in a variety of lymphoma cell lines and at a wide range of expression levels. However, proteins corresponding to only two alternatively spliced forms of TRAF3 have been detected (Gamper et al., 2001). In contrast to wt TRAF3, overexpression of all TRAF3 variants, except for the variant lacking the entire zinc-finger domain, can activate an NF-κB-dependent luciferase reporter in A293 cells, and three of the eight are able to induce NF-κB in BJAB human B-lymphoma cells (Gamper et al., 2001). Given that several TRAF3 splice variants can induce NF-κB, it will be of interest to determine whether expression of these variants is a product of aberrant splicing machinery in lymphomas, many of which have high constitutive levels of NF-κB activity (Courtois and Gilmore, 2006), or is a regulated cellular or developmental event.
Mutations in CYLD cause a recessive human disease called cylindromatosis, characterized by benign tumors of the head (Bignell et al., 2000). CYLD is a tumor suppressor and deubiquitinating enzyme whose activity is directed towards lysine-63 (K63) linkages. CYLD can inhibit NF-κB activity by deubiquitinating and inactivating TRAF2 and TRAF6, RIP, NEMO, and the NF-κB co-activator BCL-3 (Trompouki et al., 2003; Massoumi and Paus, 2007).
RT-PCR analysis in one CLYD knockout mouse model revealed the presence of an alternatively spliced CYLD transcript lacking both exons 7 and 8 (sCYLD) (Hovelmeyer et al., 2007). sCYLD mRNA and protein were detected in both CLYD knockout and wt B cells. Interestingly, sCYLD protein represents only a minor product in wt B cells, but is present in the CYLD knockout cells at levels comparable to full-length CYLD in wt B cells; that is, the knockout mouse lacked wt CYLD and expressed only the sCYLD splice variant mRNA/protein. Although lacking both TRAF2 and NEMO binding sites, sCYLD exhibits deubiquitinase activity as demonstrated by its ability to remove K63-linked ubiquitin from BCL-3. Mice expressing solely sCYLD show increased expansion of mature B lymphocyte populations in peripheral lymphoid organs (Massoumi et al., 2006; Hovelmeyer et al., 2007). Furthermore, the sCYLD mice show elevated levels of TRAF, IκBα, p100, and RelB proteins, with no alterations in their mRNA expression, suggesting that the B-cell abnormalities in B-cell expansion are a result of protein stability and not mRNA expression. This mouse knockout model that shows expression of a minor, alternatively spliced CYLD variant (sCYLD) in the absence of full-length CYLD provides a unique in vivo opportunity to study the role that this alternative splice variant plays. For example, given that the knockout mice expressing solely the alternatively spliced sCYLD protein show an altered B-cell expansion profile and have more stable NF-κB proteins, one can predict that sCYLD performs functions in B-cell regulation which are distinct from full-length CYLD.
There have been several reports of alternative splicing of TIR signaling components. Many of these studies suggest that alternative splicing of TIR pathway components serves to down-regulate NF-κB signaling, possibly preventing tissue damage that could arise from persistent inflammation.
Immunity to Mycobacterium tuberculosis is dependent on coupling of the innate immune response to T cell-mediated adaptive immunity. TLR2 heterodimerization with TLR1 or TLR6 occurs after receptor recognition of gram-positive bacteria or soluble tuberculosis factors of M. tuberculosis. Receptor dimerization drives two processes necessary for the innate immune response and T-cell activation. First, this TLR heterodimerization induces the innate immune response by activating NF-κB signaling. Second, TLR dimerization leads to antigen producing cell (APC) maturation and expression of stimulatory molecules such as B7, to enhance T-cell activation. As such, appropriate expression of TLRs upon M. tuberculosis infection is imperative for a positive clinical outcome.
RT-PCR analysis of TLRs was used to quantify their mRNA expression levels in patients with pulmonary tuberculosis (TB). A splice variant of TLR1, hsTLR1, showed the greatest change in expression levels in TB-infected tissue as compared to non-infected tissue. Therefore, the level of expression of hsTLR1 could serve as a clinical marker for TB status. Sequence analysis revealed that hsTLR1 lacks the normal non-translated exon 2 of TLR1, which may play a role in TLR1 mRNA stability (Chang et al., 2006). Since proper TLR2 function is important for bacterial clearance, it is possible that defective TLR1 splicing alters the relative TLR1 mRNA levels and stability, resulting in alterations in dimerization levels of TLR2 with either TLR1 or TLR6, which could affect the course of the disease.
TLR2 plays a key role in resolving gram-positive infections in mice, as TLR2−/− mice are highly susceptible to gram-positive infection yet show a normal gram-negative response to LPS (Takeuchi et al., 1999). Murine and human TLR2 mRNAs are derived from 3 exons: exons 1 and 2 are untranslated, while exon 3 contains the entire open reading frame. However, TLR2 mRNA expression patterns are quite different in mice and humans, suggesting that TLR2 may play different roles in these two species. The difference in TLR2 mRNA expression patterns between mice and humans might be the result of different TLR2 transcriptional regulation as suggested by the extensive sequence divergence between the mouse and human TLR2 promoters. Analysis of TLR2 cDNA sequences from human monocytes revealed five alternatively spliced variants (Haehnel et al., 2002). In all five cases, the TLR2 open reading frame (ORF) does not change, as the alternative splicing events occur in the non-coding 5’ UTR. One variant lacks exon 2, while the other four contain parts of exon 2 but with different splice acceptor and donor sites. RT-PCR analysis of human monocyte cDNA revealed that TLR2 splice variant expression varied between donors, with the shortest TLR2 mRNA present only in freshly harvested monocytes and the longer variants occurring in monocytes grown in prolonged culture conditions. Given that all five variants encode the same TLR2 protein, the function of the alternative splicing events is unclear. One proposal is that alterations in the 5’ UTR provide RNA secondary structures that result in variations in mRNA stability or translatability.
TLR4, via association with the extracellular MD-2 protein, is considered the major LPS signaling receptor. Normal murine Tlr4 is encoded by 3 exons, however a splice variant of murine Tlr4 has a 144-bp insertion between coding exons 2 and 3, resulting in a premature stop in translation (Iwami et al., 2000). This variant encodes a 122 aa protein (86 aa from Tlr4 with an additional 36 aa) lacking transmembrane sequences. Because the Tlr4 splice variant protein is predicted to be soluble, it was termed smTLR4; nevertheless, smTlr4 protein has been detected in both the culture supernatant and the membrane fraction of CHO cells, perhaps due to a post-translational modification that results in membrane localization. RT-PCR analysis of mouse macrophage RAW264.7 cells stimulated with LPS demonstrated that the smTLR4 splice variant is induced at 4 hours post treatment, with mRNA being detected for at least 24 hours; this is in contrast to the normal Tlr4 that is not inducible by LPS. Furthermore, transfection of RAW264.7 cells with smTlr4 cDNA blocked the ability of LPS to stimulate both NF-κB luciferase reporter gene activity and NF-κB-induced TNFα production. Such results suggest that prolonged exposure to LPS results in alternative splicing of Tlr4 to yield smTlr4 as a means of negatively regulating NF-κB induction and perhaps sustained TNFα production.
Four alternative splicing events of human TLR4 have been described wherein the second and/or third exon of TLR4 is skipped. The two human TLR4 splice variants that contain exon 3 are induced by LPS treatment, while the two TLR4 variants lacking exon 3 are not induced by LPS (Jaresová et al., 2007). Although TLR4 splice variant protein has not yet been detected, it is possible that, like the murine smTlr4 splice variant, induction of certain human TLR4 splice variants serves to attenuate prolonged LPS signaling.
Resolution of the inflammatory response by down regulation of the IL-1R can occur by expression of the decoy receptor (Type II IL-1R) or by competitive inhibition of ligand binding by the IL-1R antagonist (IL-1Ra) (Arend, 2002). IL-1Ra binds IL-1R, blocks IL-1R association with the accessory protein IL-1RAcP, and thereby interferes with IL-1 binding to the receptor. IL-1Ra exists in four forms arising from alternative splicing in mice and humans. One isoform is secreted (sIL-1Ra) and three isoforms are cytoplasmic as they lack a signal sequence (icIL-1Ra1, 2, 3) (Muzio et al., 1995; Gabay et al., 1997; Malyak, et al., 1998; Weissbach et al., 1998; Arend and Guthridge, 2000). All IL-1Ra splice variants are transcribed from the same 7 exon gene, and all four variants contain exons 4–7 (however, the cytoplasmic variants contain a truncated exon 4). The icIL-1Ra isoforms differ in the composition of exons 1–3: icIL-1Ra1 lacks exons 2 and 3; icIL-1Ra2 lacks exon 3; and icIL-1Ra3 lacks exon 2. The sIL-1Ra variant is not detected in unstimulated mouse tissue, but is up-regulated in the lung, spleen and liver after LPS injection (Gabay et al., 1997), while the icIL-1Ra variants are constitutively expressed in murine skin and in LPS-stimulated RAW 264.7 cells.
Inhibition of IL-R1 and NF-κB by sIL-Ra is well established to occur via competitive inhibition of ligand binding, but it is unclear whether icIL-1Ra1, located only in the cytoplasm, is able to inhibit IL-1 signaling. Nevertheless, introduced expression of cytoplasmic icIL-1Ra1 can inhibit IL-1-induced production of IL-6 and IL-8 in Caco-2 cells, which is known to be an NF-κB-dependent process (Garat and Arend, 2003). However, in HeLa cells, transfection of icIL-1Ra1 did not inhibit NF-κB luciferase reporter gene activity (Evans et al., 2006). Despite the contrasting reports of icIL-1Ra’s role in the cell, the LPS-induced expression of the inhibitory sIL-1Ra splice variant provides a mechanism for titration of IL-1 ligand and down-regulation of IL-1-induced NF-κB.
Following binding of IL-1 to the IL-1R, IL-1RAc is recruited to the receptor complex at the plasma membrane (Greenfeder et al., 1995; Huang et al., 1997), and as outlined above, NF-κB is ultimately activated. IL-1RAc is a 12 exon gene that encodes a 570 aa protein that consists of three extracellular immunoglobulin (Ig) domains involved in interaction with the IL-1R, a transmembrane domain, and the intracellular TIR domain which binds adaptor proteins such as Tollip and MyD88.
Three IL-1RAc alternatively spliced mRNAs have been identified. The first, sIL-1RAc, encodes a 356 aa soluble form of IL-1RAc (sIL-1RAc) derived from the first 9 exons of IL-1RAc, including an alternate 9b exon. The sIL-1RAc protein lacks both the transmembrane and intracellular TIR domains (Jensen et al., 2000). The sIL-1RAc protein is expressed across a range of human tissues, but is most prominent in human blood serum and the sIL-1RAc mRNA is expressed at high levels in liver cells; furthermore, the sIL-IRAc mRNA can be induced to levels higher than wt IL-1RAc in HepG2 hepatoma cells in response to phorbol ester treatment. sIL-1RAc expression is also found in both mouse and rat cells (Greenfeder et al., 1995; Gayle et al., 1997; Plata-Salamán and Ilyin, 1997). Expression of a membrane-bound sIL-1RAc-MHC fusion protein can inhibit NF-κB signaling. Additionally, sIL-1RAc increases the type II IL-1RAc decoy receptor’s affinity for IL-1, providing a second means by which this alternatively spliced variant can inhibit NF-κB activation (Smith et al., 2003).
The second IL-1RAc splice variant, IL-1RAcβ, encodes a 346 aa protein that is also a soluble form of IL-1RAc which lacks the C-terminal transmembrane and intracellular TIR domain sequences (Jensen and Whitehead, 2003). IL-1RAcβ lacks exon 9 sequences and contains a novel 45 aa C terminus that results in a frame-shift translation of exons 10–12. IL-1RAcβ is only expressed in cells treated with staurosporine, an inducer of apoptosis. While the physiological role of IL-1RAcβ remains unclear, the structural similarities it shares with sIL-1RAc suggest that it also plays an inhibitory role in NF-κB signaling. As such, the inducible expression of two soluble IL-1RAc splice variants that can inhibit NF-κB signaling represents another mechanism to down-regulate the inflammatory response through IL-1R signaling.
The third IL-1RAc splice variant encodes a 687 aa protein termed mIL-1RAcP687 (Lu et al., 2008). mIL-1RAcP687 is encoded by IL-1RAc exons 1-11 and an alternate 13th exon. Therefore, aa 1–448 of mIL-1RAcP687 are identical to wt IL-1RAc, but its 239 aa C-terminal aa are distinct, such that it lacks box 3 of the wt IL-1RAc TIR domain. Despite these TIR domain sequence differences, mIL-1RAcP687 is still membrane-bound, can interact with IL-1R, Tollip, and MyDD, and can induce NF-κB.
MyD88 is an adaptor protein involved in Toll/interleukin-1 receptor signaling to NF-κB. MyD88 contains a C-terminal TIR domain that facilitates MyD88 interaction with TIR receptors and an N-terminal death domain (DD) which mediates its interaction with IRAK proteins. The association of MyD88 with both the receptor and IRAK proteins leads to activation NF-κB.
The MyD88 DD and TIR are separated by a stretch of residues of unknown function termed the intermediate domain (ID). An alternatively spliced MyD88 variant, MyD88s, lacks exon 2 resulting in an in-frame deletion of the ID. Myd88s mRNA has been identified in the mouse macrophage cell line Mf4/4, in mouse spleen and brain tissue, and in the human monocytic cell line THP-1 after LPS treatment (Janssens et al., 2002). Furthermore, induction of the MyD88s variant correlates with defective NF-κB signaling as measured by reduced degradation of IκBα after LPS treatment. Deletion of the ID in MyD88s does not affect its binding to IL-1R and IRAK1, but abolishes the ability of MyD88 to induce IRAK1 phosphorylation and NF-κB activation. Under normal circumstances, MyD88 facilitates the interaction of IRAK1 with its kinase IRAK4, resulting in IRAK1 phosphorylation. However, MyD88s cannot bind to IRAK4, and therefore expression of MyD88s disrupts IRAK1 phosphorylation and NF-κB activation (Burns et al., 2003). Once again, the ability of LPS to induce the alternative splicing of a TIR pathway member (MyD88) to a new form (MyD88s) that is defective in NF-κB signaling suggests the importance of tempering the NF-κB response. Interestingly, MyD88s can support normal JNK signaling (Janssens et al., 2003), suggesting that induction of the NF-κB and JNK signaling pathways diverges at MyD88 and that the ID sequences of MyD88 are necessary for NF-κB signaling but not JNK signaling.
There are four members of the IRAK family of serine/threonine kinases: IRAK1 and 4 have kinase activity, whereas IRAKM and IRAK2 do not. As discussed above, following IRAK1 association with MyD88, IRAK1 is phosphorylated by IRAK4, inducing IRAK1 autophosphorylation. Phosphorylated IRAK1 can then leave the receptor complex and activate TRAF6. Three IRAK1 splice variants have been detected in mice and humans (Jensen and Whitehead, 2001). IRAK1b uses a different 5’-acceptor splice site within exon 12, resulting in an in-frame deletion of 30 aa. The IRAK1b protein lacks kinase activity, resulting in its prolonged association with the receptor complex and extended half-life. Not surprisingly, the kinase-dead IRAK1b protein shows greatly reduced NF-κB-inducing activity as compared to wt IRAK1 in transient transfection assays. Therefore, the more stable IRAK1b protein could promote prolonged, but low level, NF-κB activation.
Irak1s is a second alternatively spliced form of Irak1 found specifically in mice (Yanagisawa et al., 2003). Irak1s is generated by utilization of a splice acceptor site within the normal exon 12, resulting in a transcript shorter than the normal Irak1 mRNA by 271 nucleotides. The 537 aa Irak1s protein consists of 513 N-terminal aa identical to wt Irak1, followed by 24 unique C-terminal aa. Irak1s has no kinase activity, but (unlike IRAK1b), expression of IRAK1s in A293 cells can still induce NF-κB activity.
IRAK1c is an alternatively spliced variant of IRAK1 that lacks exon 11. IRAK1c expression is found in a wide range of human tissues and cell lines, and represents the major IRAK1 transcript in the brain (Rao et al., 2005). Similar to induction of the negative regulator IRAKM by LPS, expression of IRAK1c is induced in macrophages stimulated with LPS. IRAK1c is devoid of kinase activity, yet retains the ability to bind the TIR complex proteins IRAK2, Tollip, MyD88, and TRAF6. IRAK1c cannot be phosphorylated by IRAK4 and cannot activate NF-κB. Thus, IRAK1c may represent a natural dominant-negative inducible form of IRAK1, as IRAK1c can be induced by LPS, can bind the TLR4 receptor, but is incapable of activating NF-κB.
Taken together, these findings suggest that expression of alternative IRAK1 splice variants provides another level of regulation of NF-κB signaling. The expression of IRAK1b or IRAK1c during times of excessive LPS signaling may attenuate signaling to NF-κB. In contrast, overexpression experiments with IRAK1s suggest that its role is to upregulate NF-κB; however, at normal physiological levels, IRAK1s could serve a different role.
IRAK2 is likely an enzymatically inactive member of the IRAK family of proteins. Nevertheless, IRAK2 may influence the activity of some other enzymatically active TIR signaling molecule, perhaps IRAK4, to effect NF-κB activation, as Irak2 knockout mice are defective for sustained NF-κB activation in response to LPS. Moreover, only Irak1/Irak2 double knockout mice show profound NF-κB signaling defects (Kawagoe et al., 2008). Four alternatively spliced versions of murine Irak2 have been identified (Irak2a−d): two activate NF-κB when overexpressed (Irak2a+b) and two inhibit NF-κB (Irak2c+d). Irak2a is considered the full-length transcript and contains all 13 exons, while Irak2b lacks exon 3, Irak2c lacks the first three exons, and Irak2d lacks exon 2 and 10 aa from exon 12 due to use of an alternative 3’ splice acceptor site (Hardy and O’Neill, 2004). LPS treatment of RAW264.7 cells induces the inhibitory Irak2c transcript. Therefore, as with LPS-induced alternative splicing of IRAK1, alternative splicing of Irak2 may reduce the level of NF-κB signaling.
TGF-β activated kinase 1 (TAK1) and TAK1 binding protein (TAB1) are downstream signaling targets of TRAF6. Following IL-1 stimulation and consequent TRAF6 activation, TAB2 translocates from the membrane to the cytoplasm where it bridges an interaction of TRAF6 with the TAK1/TAB1 complex. TRAF6 is then able to activate TAK1. The association of TAB1 with TAK1 enhances TAK1’s ability to phosphorylate IKKβ and activate the IKK complex.
The N terminus of TAB1 shows some homology to protein phosphatase 2C (PP2C), a serine/threonine phosphatase, while the C terminus contains TAK1 binding sequences. TAB1β is an alternatively spliced form of TAB1 that lacks exons 11 and 12 but instead contains an alternate 27 aa exon termed exon β,resulting in a TAB1 protein isoform that is unable to bind or activate TAK1 (Ge et al., 2003). TAB1 and TAB1β mRNAs are both expressed in several human cell lines and two TAB1-specific protein bands have been identified by Western blotting, suggesting that both splice variant proteins are expressed. Although the effect of TAB1β on NF-κB signaling has not been determined, the inability of TAB1β to bind TAK1 suggests that it would be impaired in its ability to signal to NF-κB.
The normal TAK1 protein arises from a 16 exon mRNA (called TAK1a) and this TAK1 protein is a protein kinase that can activate IKK and induce NF-κB activity. The TAK1 N terminus contains sequences necessary for TAB1 binding and for TAK1 catalytic activity, while the C terminus is important for interaction with TAB2. Three alternatively spliced TAK1 variants have been identified that produce alterations of the C terminus (Kondo et al., 1998; Sakurai et al., 1998; Dempsey et al., 2000): TAK1b contains an alternate exon (AE1) between exons 11 and 12; TAK1c contains AE1 but lacks exon 15; and TAK1d lacks both AE1 and exon 15. Deletion of exon 15 in both TAK1c and TAK1d results in a reading frame shift and inclusion of an alternate 10-residue C terminus. Expression of mRNA of each TAK1 variant has been detected across a panel of human tissue types; however, there is significant variation in their abundance. TAK1a mRNA was most abundant in 6/11 tissues, TAK1b in 4/11, TAK1c in 1/11, and TAK1d was present in 7/11 tissues but was never the most abundant (Dempsey et al., 2003). Overexpression of TAB1 with TAK1a, b, or c results in activation of NF-κB as measured by luciferase reporter gene analysis, DNA binding and IκBα/β degradation (Sakurai et al., 1998). Therefore, it is not clear how or if the splice variant-encoded forms of TAK1 differ from the normal TAK1a protein. For example, it remains to be determined whether the C-terminal alterations affect TAK1 interaction with TRAF6 via TAB2.
Almost all signals that activate NF-κB proceed through activation of the IKK complex. IKK consists of two catalytic kinases, IKKα and IKKβ, and the regulatory component NEMO (IKKγ). Ubiquitination of NEMO and phosphorylation of at least one IKK kinase subunit is necessary for IKK activation and for proper IκB phosphorylation/degradation and NF-κB signaling.
NEMO, the regulatory subunit of the IKK complex, contains two coiled-coil domains, a leucine-zipper, and a C-terminal zinc-finger domain. In the IKK complex, NEMO is normally a dimer, which is required for NF-κB activation. A human NEMO splice variant (IKKγ-Δ) lacking exon 5 has been identified (Hai et al., 2006). IKKγ-Δ mRNA has been detected in all cell lines and tissue samples examined, and the expression of IKKγ-Δ protein has been confirmed by 2D gel electrophoresis. Despite loss of the coiled-coil sequences encoded by exon 5, IKKγ-Δ can interact with both wt NEMO and IKKβ, and can function in activation of NF-κB in response to TNF. However, unlike wt NEMO, IKKγ-Δ cannot function in HTLV-1 Tax-induced NF-κB signaling. This Tax signaling defect is not surprising, however, because deletion of NEMO residues 196–419, which partially overlap the exon 5 sequences missing in IKKγ-Δ (aa 174–224), creates a dominant-negative mutant for Tax-induced activation of NF-κB (Iha et al., 2003).
A splice donor-site mutation that results in expression of an mRNA encoding a NEMO protein lacking exons 4, 5, and 6 sequences has been identified in a family with a possible immunodeficiency syndrome (Ørstavik et al., 2006). The 35 kDa mutant NEMO protein lacks 153 aa that comprise coiled-coil 1 domain, but still has the coiled-coil 2 domain, the leucine-zipper domain, and zinc-finger domain. This NEMO splicing mutation yields a protein that is defective for NF-κB signaling and resembles a well-characterized NEMO deletion mutant lacking exons 4–10 that results in a severe immunodeficiency disease termed incontinentia pigmenti (Courtois and Gilmore, 2006). While expression of the NEMO disease mutant lacking exon 4, 5, and 6 sequences is not a normal alternative splicing event, its expression highlights the fact that aberrant splicing can affect NF-κB signaling and can have severe biological consequences.
IκB proteins generally bind and sequester NF-κB dimers in the cytoplasm. All IκBs have 6–8 copies of an ankyrin repeat sequence that is required for binding to NF-κB dimers, and several IκBs have N-terminal regulatory domains containing IKK phosphorylation sites, and C-terminal domains with PEST degradation sequences. One well-studied member of this family is IκBα, which is commonly associated with the NF-κB dimer p50/p65. However, IκBα has overlapping NF-κB inhibitory roles with IκBβ and IκBε. Upon phosphorylation by IKK, IκB is targeted for proteolytic degradation by the proteasome, freeing the NF-κB dimer to enter the nucleus and activate transcription of target genes. In addition, some IκB family members (Bcl-3, IκBζ) can function as co-activators for NF-κB transcription.
The functional difference between IκBα and IκBβ is unclear. Mice with a knock-in of IκBβ into the IκBα locus show no defect in activation of NF-κB. Nevertheless, several cell culture experiments have suggested that there are differences between IκBα and IκBβ in their binding preferences for NF-κB dimers, rates of degradation, and rates of resynthesis (Thompson et al., 1995; Hirano et al., 1998; Hoffmann et al., 2002). For example, IκBβ is degraded much slower than IκBα in response to phosphorylation by IKK (Thompson et al., 1995).
Two IκBβ mRNAs (IκBβ1 and IκBβ2) from human cells have been isolated (Hirano et al., 1998). These mRNAs encode two distinct protein isoforms: IκBβ1 is a 43 kDa protein of 361 aa, whereas IκBβ2 is a 41 kDa protein with a C-terminal truncation in the PEST domain. IκBβ2 represents the major IκBβ isoform in humans as it is more abundant than IκBβ1 in most cell types, whereas IκBβ1 is the only IκBβ mRNA that has been detected in mouse cells. The IκBβ1 protein is quickly degraded in response to TNF-α, IL-1β, and PMA. Furthermore, in unstimulated cells, IκBβ1 is found in the nucleus. In contrast, IκBβ2 is the more abundant of the two human isoforms and is only weakly degraded following IKK phosphorylation. Despite these differences, both variants are indistinguishable in their binding to NF-κB dimers and show a strong preference for NF-κB dimers containing p65. Furthermore, both IκBβ variants show delayed kinetics of resynthesis after NF-κB inducation as compared to IκBα. It remains unclear whether there are circumstances where one IκBβ variant is expressed at higher than normal levels. The reduced degradation of IκBβ2 in response to a variety of stimuli suggests that its expression would render cells insensitive to NF-κB signaling stimuli that proceed through an IκBβ-dependent pathway.
The inhibitory IκBγ protein contains 7 ankyrin repeats and corresponds to the C-terminal sequences of p105. In mice, IκBγ mRNA is generated by transcription from an internal promoter within p105 as well as by limited processing of p105 by the proteasome (Inoue et al., 1992). IκBγ is a cytoplasmic protein that preferentially binds the NF-κB transcription factors p50, RelA, and c-Rel, inhibiting their nuclear localization. Two cDNAs have been identified in mice that correspond to alternatively spliced forms of IκBγ (Grumont and Gerondakis, 1994). The protein encoded by one splice variant, IκBγ1, lacks 59 aa C-terminal to ankyrin repeat 7 and has a distinct 35 aa C terminus. The 63 kDa IκBγ1, isoform shows both cytoplasmic and nuclear subcellular localization. The second splice variant, IκBγ2, lacks 190 C-terminal aa and encodes a 55 kDa, nuclear protein. In contrast to the normal IκBγ protein, both alternatively spliced variants show a greatly decreased ability to inhibit several types of NF-κB complexes. The nuclear localization and greatly reduced inhibition of certain NF-κB complexes by both IκBγ1, and IκBγ2, as compared to wt IκBγ, suggests that these alternatively spliced IκBγ isoforms would allow certain NF-κB-regulated genes to escape inhibition by IκBγ.
Like other IκB proteins, IκBζ contains C-terminal ankyrin repeat motifs that mediate association with NF-κB transcription factors, however, IκBζ also has a large, approximately 450 aa N-terminal domain not found in other IκBs (Yamazaki et al., 2001). IκBζ mRNA is inducibly expressed and the IκBζ protein is localized to the nucleus in response to both LPS and IL-1, but not TNF (Yamamoto et al., 2004). IκBζ has both negative and positive effects on NF-κB dimers. Under some conditions, IκBζ can inhibit NF-κB DNA binding (Yamakazi et al., 2001). However, when fused to the GAL4 DNA-binding domain, the N-terminal IκBζ residues 329–402 exhibit transactivation activity, and co-transfection of IκBζ with NF-κB subunits p50 or p65ΔTAD (lacking transactivation domain sequences) results in increased NF-κB reporter gene activity (Motoyama et al., 2005). Furthermore, binding of IκBζ to NF-κB p50 is essential for induction of IL-6 transcription upon stimulation with LPS or IL-1β (Motoyama et al., 2005). Two splice variants of mouse IκBζ have been characterized (Motoyama et al., 2005).
An LPS-inducible, alternatively spliced variant of IκBζ, named IκBζ (S), lacks the N-terminal 99 aa (Motoyama et al., 2005). Like wt κBζ, a GAL4 fusion protein containing the N-terminal domain of IκBζ(S) shows transactivation activity. Additionally, LPS treatment of cells expressing either wt IκBζ or IκBζ(S) leads to enhanced IL-6 production. Thus, it is not clear how IκBζ (S) differs from wt IκBζ.
A second alternatively spliced IκBζ variant is expressed in mice at very low levels. This splice variant, IκBζ(D), lacks exon 7 (aa 236–429), which encodes the sequences that contain transactivating activity. Not surprisingly, GAL4-IκBζ(D) fusion proteins are unable to activate transcription, and full-length IκBζ (D) cannot enhance p50 or p65ΔTAD NF-κB reporter gene activity. Therefore, it is possible that IκBζ(D) represents a dominant-negative form of IκBζ. It would be interesting to determine at what level and under what circumstances these alternatively spliced variants are induced and how they contribute to IκBζ-mediated NF-κB gene activation or repression.
The five members of the mammalian NF-κB transcription factor family contain an approximately 300 aa N-terminal Rel homology domain (RHD) that is responsible for DNA binding, dimerization, nuclear localization, and IκB binding. The NF-κB transcription factors c-Rel, RelA, and RelB all contain C-terminal transactivation domains. In contrast, p100 and p105 contain inhibitory C-terminal ankyrin repeat domains that can be proteolytically cleaved to yield the mature p52 and p50 DNA-binding proteins, respectively. NF-κB transcriptions factors form homo- and heterodimers in almost all combinations. As described above, activation of the TNF or Toll/interleukin receptor pathway culminates in the nuclear localization of NF-κB dimers and regulation of specific target gene transcription.
Given that alternative splicing events have been identified in many upstream components of the NF-κB signaling pathway, it is no surprise that the genes encoding NF-κB transcription factors also undergo alternative splicing. In addition to the mammalian splicing events described below, there are several alternative splicing events in insects that affect NF-κB transcription factors (Gross et al., 1999; Shin et al., 2002; Kawai et al., 2003).
Under normal circumstances, p100 (or p105) is a cytoplasmic protein, because nuclear import is inhibited via masking of the nuclear localization sequence (NLS) by the C-terminal ankyrin repeat domain. Proteolytic cleavage of the p100 (or p105) C-terminal ankyrin repeat domain liberates the RHD-containing p52 protein (or p50) such that it can enter the nucleus.
Alternative splicing in mice generates two p105 isoforms that show differences in subcellular localization, expression patterns, and function (Grumont et al., 1994). The first alternatively spliced p105 isoform, p98NF-κB1, lacks 111 C-terminal aa and has 35 unique C-terminal aa. The second p105 isoform, termed p84NF-κB1, lacks 190 C-terminal aa including a portion of ankyrin repeat 7. p84 is inducibly expressed in 3T3 cells in response to phorbol ester. In contrast to wt p105, p98 and p84 can both bind weakly to DNA and show some nuclear localization. Distinct from both p105 and p84, p98 is exhibits some transactivation potential as measured in κB site-containing reporter assays; however, it is unclear if this transactivation ability is intrinsic to p98 or occurs via its association with p50, RelA, or c-Rel. Interestingly, as a GAL4-fusion protein in yeast, p98 sequences corresponding to p105 aa 727–806 can activate transcription (Morin and Gilmore, 1992), suggesting that p98 has inherent transcriptional activity. Although both alternatively spliced p105 isoforms have altered DNA binding and transcriptional functions, their extremely low expression levels compared to wt p105 raise questions as to their physiological relevance. Of note, C-terminal truncations in p100 occur in several types of human lymphoma (Courtois and Gilmore, 2006), indicating that such splice variants of p105 may need to be tightly regulated.
Alternative splicing of NF-κB transcription factors as a result of a misregulated LPS signaling has been detected in CD14 knockout mice (Phan et al., 2006). Under normal conditions, CD14 associates with TLR4 following LPS stimulation, and signals from this receptor complex lead to activation of IKK. To examine the role of CD14 in LPS signaling, a CD14 knockout mouse was subjected to burning to induce LPS signaling. Under these conditions, nine alternative splicing events of genes encoding NF-κB subunits were identified, including ones that produce large deletions within the RHD and the C termini of RelA, RelB and p100. Taken together, the large number of splicing events seen with the LPS signaling-deficient, CD14 knockout mouse model suggests that alternative splicing of NF-κB transcription factor genes could be regulated by the intensity of the LPS signal.
RelA (also referred to as p65) is generally found in a dimer with p50 in the classic NF-κB complex. rela−/− mice die between E14.5-E16 due to excessive TNF-induced hepatocyte apoptosis (Beg et al., 1995b). Furthermore, RelA is essential for TNF-mediated pro-survival signaling to NF-κB as RelA knockout mouse embryonic fibroblasts rapidly undergo apoptosis in response to TNF (Beg and Baltimore, 1996), due to a failure to be able to induce NF-κB anti-apoptosis target genes (Wang et al., 1998).
Three alternative splice variants of RelA have been identified: p65Δ, p65Δ2, and p65Δ3. p65Δ arises by use of an alternative splice acceptor site located 30 nucleotides into exon 8 in mouse and human cells, and p65Δ encodes a protein lacking RHD aa 222–231, which are required for association with p50 and for DNA binding. p65Δ is the prevalent transcript in certain cell lineages (S17, pre-B, and erythroid) in which there is an almost complete absence of full-length p65 mRNA, suggesting that alternative splicing of p65 is a regulated normal event (Narayanan et al, 1992). While p65Δ does not form dimers with p50 or itself, p65Δ can weakly form heterodimers with full-length p65, but these p65Δ/p65 dimers have a greatly diminished DNA-binding ability. As such, p65Δ /p65 dimers could represent non-functional dimers that negatively regulate NF-κB (Ruben et al., 1992).
The p65Δ2 and p65Δ3 splice variants have not been extensively characterized, either in terms of expression profile or activity. p65Δ2 lacks N-terminal aa 13–22, but has an intact RHD (Lyle et al., 1994). p65Δ3 was identified in a non-small cell lung carcinoma cell line as a RelA splice variant lacking exons 7 and 8 (RHD aa 187–293) (Maxwell and Mukhopadhyay, 1995). It is likely that p65Δ3 is also defective for dimerization and DNA binding given that p65Δ3 lacks aa 222–231, which are the residues absent in the better characterized p65Δ splice variant.
Like p65, c-Rel is an NF-κB transcription factor that contains a C-terminal transactivation domain. c-Rel is important for normal B-cell proliferation, as c-Rel−/− mice exhibit defects in B-cell proliferation and survival after treatment with LPS, α-IgM, or CD40 (Grumont et al, 1998). c-Rel also has a role in B-cell oncogenesis (Courtois and Gilmore, 2006): for example, the REL gene is amplified in many B-cell lymphomas and overexpression of REL can malignantly transform chicken B-lymphoid cells (Gilmore et al., 2001). Furthermore, some C-terminal deletions and point mutations enhance REL’s transforming activity (Starczynowski et al., 2003; Starczynowski et al., 2005; Starczynowski et al., 2007).
Human c-Rel (REL) was originally cloned as two cDNAs from the human lymphoma Daudi cell line (Brownell et al., 1989). One cDNA contains the normal 587 aa REL coding sequence. The other REL cDNA clone contains an additional 96 base pairs, arising from an exonized Alu sequence (REL+Alu) between REL exons 8 and 9. REL+Alu encodes a 619 aa protein (Brownell et al., 1989). A second alternatively spliced REL variant (RELΔ9) lacks the entire exon 9, and encodes a 564 aa protein. As a consequence, both REL+Alu and RELΔ9 have alterations in the sequence located between the RHD DNA-binding domain and the C-terminal transactivation domain; these changes result in REL+Alu and RELΔ9 having enhanced transactivation and DNA-binding abilities as compared to wt REL (Leeman et al., 2008). As such, the region where the splice alterations occur has been termed RID (REL Inhibitory Domain). The endogenous REL+Alu or RELΔ9 proteins have not been detected, most likely due to the lower expression of their transcripts as compared to the normal REL transcript. While the REL+Alu transcript is ubiquitously expressed, the RELΔ9 transcript shows elevated expression in primary lymphoma cells and cell lines, but little or no expression in normal lymphoid tissue or non-lymphoma cell lines. Therefore, RELΔ9 might represent a useful marker for the diagnosis of certain lymphomas.
Alternative splicing events have been described at virtually every level of NF-κB signaling, from the initiating receptors to the transcription factors regulating target gene expression (Table 1). In many cases, the mRNAs encoding these alternative splice protein isoforms can be induced (Table 1), and their proteins display altered structures and function (Fig. 2). As such, expression of these variants may serve to regulate proper NF-κB signaling. As shown in Fig. 1, some alternative splicing events can enhance the NF-κB response, while others can act inhibit NF-κB signaling. In this review we focus on alternative splicing in the TNF and TLR pathways. However, alternatively splicing of signaling components also occurs in NF-κB pathways initiated by other important receptors, such as the T- and B-cell receptor pathways where alternative splicing is known to occur with Fyn (Davidson et al., 1992; Goldsmith et al., 2002), Lyn (Yi et al., 1991), Syk (Rowley et al., 1995; Goodman et al., 2001; Wang et al., 2003), PKCβ/θ (Niino et al., 2001; Kawakami et al., 2002; Sledge and Gökmen-Polar, 2006) and BCL10 (Grimwade et al., 2000).
The critical role NF-κB plays in the immune and inflammatory responses is indisputable as evidenced by that phenotypes seen in mouse knockout models and the range of human immunological and inflammatory diseases attributed to inappropriate NF-κB signaling. Because alternative splicing of NF-κB signaling genes appears to play a role in modulating NF-κB activity, aberrant expression of some alternatively spliced NF-κB signaling transcripts may also have a role in certain diseases. While it is unclear whether alternative splicing of these genes plays a causative role in disease or is simply a consequence of a stressed cell, alternatively spliced NF-κB regulatory genes may in some cases serve as markers for disease progression and outcome, as well as being possible therapeutic targets.
Surely, this review of alternative splicing in the NF-κB pathway is incomplete; the identification of alternative splicing events is sparse and has generally occurred serendipitously. More comprehensive EST database analysis and evolving microarray and sequencing technologies will make further identification of alternatively spliced mRNAs easier. Indeed, by searching the NCBI mouse and human EST databases, we have found several additional splice variants in the TNF and Toll/IL-1 receptor pathways (Table 2). Molecular analysis of these alternatively spliced mRNAs and protein isoforms will undoubtedly reveal a more complete picture of NF-κB signaling.
We thank Jim Deshler for helpful discussions. Research in our laboratory on NF-κB is supported by NIH grant CA47763.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.