NF-κB, CHEMOKINE GENE TRANSCRIPTION AND TUMOUR GROWTH

Abstract

The constitutive expression of angiogenic and tumorigenic chemokines by tumour cells facilitates the growth of tumours. The transcription of these angiogenic and tumorigenic chemokine genes is modulated, in part, by the nuclear factor-κB (NF-κB) family of transcription factors. In some tumours, there is constitutive activation of the kinases that modulate the activity of inhibitor of NF-κB (IκB) kinase (IKK), which leads to the constitutive activation of members of the NF-κB family. This activation of NF-κB is associated with the dysregulation of transcription of genes that encode cytokines, chemokines, adhesion factors and inhibitors of apoptosis. In this review, I discuss the factors that lie upstream of the NF-κB cascade that are activated during tumorigenesis and the role of the putative NF-κB enhanceosome in constitutive chemokine gene transcription during tumorigenesis.

Chemokines are small, pro-inflammatory peptides; they are often expressed in response to the induction of expression of nuclear factor-κB (NF-κB) by cytokines or other stimuli. Chemokines regulate the transport, activation and, sometimes, proliferation of several cell types, including myeloid, lymphoid, endothelial and epithelial cells^1,2. There are four chemokine subfamilies — CXC, C, CX₃C and CC — based on the positions of conserved cysteine residues near the amino terminus of the proteins¹ (TABLE 1). Melanoma growth-stimulatory activity (CXCL1) and interleukin-8 (IL-8, CXCL8) are members of the CXC-chemokine family, which also includes interferon-γ (IFN-γ)-inducible gene 10 (CXCL10), IFN-inducible T-cell α-chemoattractant (CXCL11), monocyte induced by IFN-γ (CXCL9), epithelial-derived neutrophil-activating peptide 78 (ENA-78, CXCL5), granulocyte chemotactic protein 2 (CXCL6), neutrophil activating peptide 2 (CXCL7), a mitogen for B-cell progenitors known as stromal-derived factor (CXCL12) and B-cell-activating factor 1/B-lymphocyte chemoattractant (BCA1/BLC, CXCL13)¹. The expression of various members of the chemokine superfamily is induced by several cytokines, such as IL-1 and tumour-necrosis factor (TNF), through an NF-κB-mediated event; by IFN-γ through the Janus-activated kinase (JAK)–signal transducer and activator of transcription (STAT) pathway; or by activating protein 1 (AP1)-mediated transcription. The transcription of chemokine genes is often inhibited by transforming growth factor-β (TGF-β) and glucocorticoids¹.

Table 1.

CXC-, C-, CX₃C- and CC-chemokine and receptor families

Systematic name	Chromosome	Human ligand	Mouse ligand	Chemokine receptor(s)
CXC-chemokine/receptor family
CXCL1	4q21.1	GROα/MGSAα	GRO/MIP2/KC?	CXCR2>CXCR1
CXCL2	4q21.1	GROβ/MGSAβ	GRO/MIP2/KC?	CXCR2
CXCL3	4q21.1	GROγ/MGSAγ	GRO/MIP2/KC?	CXCR2
CXCL4	4q21.1	PF4	PF4	Unknown
CXCL5	4q21.1	ENA78	GCP2/LIX?	CXCR2
CXCL6	4q21.1	GCP2	GCP2/LIX?	CXCR1, CXCR2
CXCL7	4q21.1	NAP2	Unknown	CXCR2
CXCL8	4q21.1	IL-8	Unknown	CXCR1, CXCR2
CXCL9	4q21.1	MIG	MIG	CXCR3 (CD183)
CXCL10	4q21.1	IP10	IP10/CRG2	CXCR3 (CD183)
CXCL11	4q21.1	ITAC	ITAC	CXCR3 (CD183)
CXCL12	10q11.21	SDF1α/β	SDF1/PBSF	CXCR4 (CD184)
CXCL13	4q21.1	BCA1	BLC	CXCR5
CXCL14	5q31.1	BRAK/Bolckine	BRAK	Unknown
(CXCL15)	Unknown	Unknown	Lungkine/WECHE	Unknown
CXCL16	17p13	Unknown	Unknown	CXCR6
C-chemokine/receptor family
XCL1	1q24.2	Lymphotactin/SCM1α/ATAC	Lymphotactin	XCR1
CX3C-chemokine/receptor family
CX3CL1	16q13	Fractalkine	Neurotactin/ABCD3	CX3CR1
CC-chemokine/receptor family
CCL1	17q11.2	I309	TCA3/P500	CCR3
CCL2	17q11.2	MCP1/MCAF/TDCF	JE?	CCR2
CCL3	17q12	MIP1α/LD78α	MIP1α	CCR1, CCR5
CCL3L1	17q12	LD78β	Unknown	CCR1, CCR5
CCL4	17q12	MIP1β	MIP1β	CCR5 (CD195)
CCL5	17q12	RANTES	RANTES	CCR1, CCR3, CCR5 (CD195)
(CCL6)		Unknown	C10/MRP1	Unknown
CCL7	17q11.2	MCP3	MARC?	CCR1, CCR2, CCR3
CCL8	17q11.2	MCP2	MCP2?	CCR3, CCR5 (CD195)
(CCL9/10)		Unknown	MRP2/CCF18/MIP1γ	CCR1
CCL11	17q11.2	Eotaxin	Eotaxin	CCR3
(CCL12)		Unknown	MCP5	CCR3
CCL13	17q11.2	MCP4	Unknown	CCR2, CCR3
CCL14	17q12	HCC1	Unknown	CCR1, CCR5
CCL15	17q12	HCC/LKN1/MIP1δ	Unknown	CCR1, CCR3
CCL16	17q12	HCC4/LEC/LCC1	Unknown	CCR1, CCR2
CCL17	16q13	TARC	TARC/ABCD2	CCR4
CCL18	17q12	DC-CK1/PARC/AMAC1	Unknown	Unknown
CCL19	9p13.3	MIP3β/ELC/exodus-3	MIP3β/ELC/exodus-3	CCR7 (CD197)
CCL20	2q36.3	MIP3α/LARC/exodus-1	MIP3α/LARC/exodus-1	CCR6
CCL21	9p13.3	6Ckine/SLC/exodus-2	6Ckine/SLC/exodus-2/TCA4	CCR7 (CD197)
CCL22	16q13	MDC/STCP1	ABCD1	CCR4
CCL23	17q12	MPIF1/CKβ8/CKβ8-1	Unknown	CCR1
CCL24	7q11.23	Eotaxin-2/MPIF2	MPIF2	CCR3
CCL25	19p13.3	TECK	TECK	CCR9
CCL26	7q11.23	Eotaxin-3	Unknown	CCR3
CCL27	9p13.3	CTACK/ILC	ALP/CTACK/ILC/ESkine	CCR10
CCL28	5p12	MEC		CCR3/CCR10

Reproduced, with permission, from REF. 123 (© 2001 The Society for Leukocyte Biology).

As the CXC-chemokines CXCL1 and CXCL8 have been associated with tumour growth, metastasis and angiogenesis, I concentrate on these important chemokines. The biological functions of chemokines are mediated through seven-transmembrane G-protein-coupled receptors. CXCL8 and CXCL6 bind to the chemokine receptors CXCR1 and CXCR2, whereas CXCL1 and other CXC-chemokines that have a Glu-Leu-Arg (ELR) motif at their amino terminus (CXCL2, -3 and-5) bind to and activate CXCR2 only². There are also viral receptors and binding proteins for these chemokines³. Receptor binding initiates a series of downstream signals, including calcium mobilization and the activation of phospholipase-Cβ (PLCβ), phosphatidylinositol 3-kinase (PI3K), RAS, the RHO family of GTPases, p21-activated kinase (PAK), extracellular signal-regulated kinases 1 and 2 (ERK1 and ERK2), p38 mitogen-activated protein kinase (p38 MAPK) and NF-κB^1,4,5.

Both the CXC-chemokines and their receptors, and the CC-chemokines and their receptors, are involved in tumorigenesis and metastasis⁶. Expression of CXC-chemokines that have an ELR motif at the amino terminus is associated with enhanced tumorigenic capacity and metastasis, whereas expression of ELR-negative CXC-chemokines, C-chemokines and CC-chemokines by tumours is often associated with the inhibition of tumour growth or regression^3,7. Much research into the dysregulated expression of chemokines during tumour formation has used the melanoma model, in which CXCL1, CXCL8 and regulated on activation, normal T-cell expressed and secreted (RANTES, CCL5) are expressed frequently at high levels in metastatic lesions⁸. This constitutive expression of chemokines occurs due to transcriptional activation of the encoding genes. In part, the endogenous transcription of chemokine genes in melanoma cells is due to altered NF-κB activation. Tumour angiogenesis, growth and metastasis are facilitated by the NF-κB-modulated transcription of chemokine genes. In this review, I explore the mechanisms by which constitutively activated kinases that function upstream of the NF-κB cascade facilitate chemokine-mediated tumorigenesis. I also examine the role of NF-κB-interacting factors — which are thought to form an ENHANCEOSOME-like structure — in the dysregulation of transcription of chemokine genes during tumour development.

Chemokines and cancer

Constitutive secretion of chemokines in melanoma

CXCL1, CXCL8 and CCL5 have been shown to be expressed constitutively in human melanoma cell lines and tumours (see REF. 8 for a full review). The over-expression of these chemokines in melanocytes and melanoma tumour cells is associated with enhanced tumour-forming capacity and metastatic potential. Antibodies that are specific for these ligands or their receptors slow the growth of melanoma tumours in mice⁸. Human melanoma lesions are reported to express several chemokine receptors^6,8–11.

A role for CXCL1 and CXCL8 in tumorigenesis is supported further by data showing that the overexpression of one of the mouse homologues of CXCL1 — known as KC (on the basis of the ordinate and abscissa of the cloning position) — enhanced the growth and metastasis of KC-deficient PAM212 squamous-cell carcinoma cells¹². KC-expressing PAM212 cells injected into BALB/c mice formed tumours that grew more rapidly than PAM212 parental cells, but this growth advantage associated with the expression of KC was not observed in BALB/c Cxcr2 ^−/− mice. When BALB/c mice were injected with PAM–Ly2 cells — a metastatic variant of PAM212 that expresses KC — and the tumour-bearing mice were treated with the NF-κB inhibitor PS-341, which blocks the proteasomal degradation of inhibitor of NF-κB (IκB), the growth of the tumours was inhibited by 50%. The mice that were treated with PS-341 had blocked expression of KC and vascular endothelial growth factor (VEGF), which inhibited tumour growth and angiogenesis¹². Similar experiments carried out with xenografts of human squamous-cell carcinoma cell lines into severe combined immunodeficient (SCID) mice gave similar results. In general, the greater the constitutive activation of NF-κB, the greater the sensitivity to PS-341-induced apoptosis. Using DIFFERENTIAL DISPLAY and complementary DNA MICROARRAYS to investigate differences in gene expression between PAM212 cells that metastasize and those that do not, Dong et al.¹³ found that of the genes that are differentially expressed in the metastatic cells, 10 out of 22 are NF-κB dependent in their expression. CXCL1, certain signalling molecules (c-Met, Syk and Yes-associated protein), proteins involved in cell growth (p21, p27 and cyclin D) and proteins involved in apoptosis (cellular inhibitor of apoptosis 1, Iap1; the phosphoprotein enriched in astrocytes of 15 kDa, Pea15; and Fas ligand, FasL) were among the proteins encoded by these NF-κB-inducible genes. So, genes that are regulated by NF-κB, including chemokine genes, modulate squamous-cell carcinoma and melanoma tumorigenesis.

Regulation of expression of chemokine receptors

The expression of chemokine receptors is also modulated by NF-κB. The expression of CXCR2 is regulated through an NF-κB element, and the expression of this receptor is induced in response to granulocyte–macrophage colony-stimulatory factor (GM-CSF), IL-4 and IL-13, whereas lipopolysaccharide (LPS) induces the loss of expression of CXCR2 (REFS ^14–17). The expression of CXCR1 and CXCR2 has been detected in melanoma tumours, which indicates that melanoma cells might be targets for autocrine stimulation by CXCL8 (REF. ⁹). However, some melanoma tumour cell lines express CXCR4, CCR7 and CCR10, but no detectable levels of other chemokine receptors⁶. The expression of CXCR3 and CXCR4 has been noted in BLM and Sk-Mel37 human melanoma cells¹⁰. The ligand for CXCR4 (CXCL12) is produced by lymph node, liver, lung and bone marrow. Ligands for these receptors⁶ stimulate small GTPases and induce reorganization of the actin cytoskeleton, which indicates that these receptors might be involved in the migration and, perhaps, metastasis of these tumour cells. The activation of NF-κB also occurs as a consequence of activation of chemokine receptors. The melanoma cell lines that express moderate levels of CXCR3 and CXCR4 (BLM and SK-Mel37) have barely detectable levels of immunoreactive CXCR1 and CXCR2. When B16 melanoma cells were transduced with a retroviral expression vector for CCR7, enhanced lymph-node metastasis was observed, which could be blocked by antibodies specific for secondary lymphoid tissue chemokine (CCL21), the ligand for CCR7 (REF. ¹¹). Together, these studies indicate that there is differential expression of chemokine receptors between melanoma cell lines and melanoma tumours. The expression of these receptors can often be correlated with the metastatic capacity of the tumour cells and their organ-specific metastasis.

Box 1 | Members of the REL family and their regulators

Members that activate transcription

• v-Rel — an avian retroviral protein

• c-REL — the cellular homologue of v-Rel

• RELA/p65 — the main REL protein

• RELB — a less commonly expressed REL protein

• Dorsal — a Drosophila Rel-like protein

• Dorsal-related immunity factor (Dif) — a Drosophila Rel-like protein

• p50 subunit of nuclear factor-κB (NF-κB) — derived from processing of p105

• p52 subunit of NF-κB — derived from processing of p100

Members that sequester NF-κB in the cytoplasm

• Inhibitor of NF-κB, α (IκBα)

• IκBβ

• IκBε— lacks the PEST domain (a sequence that is rich in proline, glutamate, serine and threonine residues)

• IκBγ

• Drosophila Relish

Members that enhance transcription or restrain NF-κB nuclear localization

• B-cell leukaemia/lymphoma 3 (BCL3)

Kinases that phosphorylate IκB

• IκB kinase-α (IKKα) — phosphorylates IκB and facilitates the processing of p105

• IKKβ— phosphorylates IκBα at Ser32 and Ser36, and IκBβ at Ser19 and Ser23

Adaptor molecules for the IKKs

• IKKγ (NF-κB essential modulator, NEMO) — a regulatory factor for IKKα and IKKβ

CXCL8 and its receptors, CXCR1 and CXCR2, are expressed in ovarian-cancer cells¹⁸. CXCR1 is expressed largely on the luminal side of the epithelial tumour cells, but there is no such polar distribution of CXCR2. CXCL8 and both CXCR1 and CXCR2 are expressed in human colon-carcinoma cells, and antibodies specific for CXCR1, CXCR2 or CXCL8 inhibited the invasive capacity of colon-carcinoma cells¹⁹. Antibodies specific for CXCR1 or CXCR2 also produced a moderate inhibition of the proliferation of the colon-cancer cells and of the adhesion of colon-carcinoma cells to endothelial cells.

NF-κB activation and cancer

Several tumour systems have been shown to have constitutive activation of NF-κB. This enhanced NF-κB activity is thought to allow tumour cells to constitutively express both angiogenic and angiostatic chemokines, cytokines such as VEGF, IL-1 and IL-6, and other factors that are reported to affect melanoma growth and escape from apoptosis. The expression of CCL5 in melanoma and breast cancer has been shown to correlate with the metastatic capacity of the tumour^20,21. Because the expression of CCL5 is induced by NF-κB²², down-modulation of CCL5 through the inhibition of NF-κB offers promise for therapeutic intervention. We have observed that once the expression of CXCL1 and CXCL8 has been induced, these secreted chemokines feed back on the activation of NF-κB to keep the cycle going. The endogenous activation of NF-κB in melanoma tumour cells can be inhibited by ~50% with an antibody specific for CXCL1, which indicates that chemokine autocrine loops might be important for perpetuating the constitutive activation of NF-κB in melanoma, but that additional cytokines or alterations in the signal-transduction pathway also contribute to the dysregulation of NF-κB in melanoma²³.

CXCL8, anoxia and tumour aggressiveness

Treatment of human malignant melanoma cells with ionizing radiation induces the coordinated activation of NF-κB and other regulators of transcription²⁴. The activation of NF-κB correlates with CXCL8 expression, growth, angiogenesis and metastasis of human melanoma cells^25,26. The induction of transcription of CXCL8 in necrotic/ANOXIC areas of melanoma tumours is dependent on NF-κB and AP1 activity²⁷. Transfection of melanoma cells with a dominant-negative inhibitor of NF-κB reduced tumour growth, metastatic potential, CXCL8 production and growth of blood vessels into the tumour²⁸. These data support an association between the endogenous activation of NF-κB and the enhanced metastatic potential of malignant melanoma cells, and indicate that targeting NF-κB might have some potential therapeutic effects in clinical trials. In support of this concept, the adenoviral-mediated transfer of IκB to melanoma cells renders them sensitive to TNF-induced cytotoxicity²⁹. Using a cell-culture system, we have shown that delivery of the super-repressor form of IκB — which cannot be phosphorylated on Ser32 or Ser36 and, therefore, is not targeted for ubiquitylation and degradation by the proteasome — to melanoma cells blocks FOCUS FORMATION and targets tumour cells to undergo apoptosis³⁰.

Mechanism of dysregulation of NF-κB

The NF-κB proteins constitute a family of proteins that are homologous to the chicken oncogene rel. These proteins are regulated at several levels — transcription, translation and post-translational processing (BOX 1). Normally, NF-κB (a p65–p50 or p65–p52 heterodimer) is sequestered in the cytoplasm, owing to the binding of IκB to the nuclear-localization sequence of the p65/RELA NF-κB protein. Activating signals lead to the phosphorylation and ubiquitylation of IκB and its degradation by the 26S proteasome, thereby freeing the NF-κB complex to move to the nucleus and bind specific DNA promoter sequences³¹. Much work has been done to clarify the role of IκB phosphorylation and ubiquitylation in its targeting to the proteasome^32–34. Inducible pathways for the regulation of these events involve activation of the IκB kinase (IKK) complex — comprising IKKα, IKKβ and the adaptor molecule IKKγ (NEMO) — which leads to the phosphorylation of specific serine residues of IκB (Ser32 and Ser36 of IκBα, or Ser19 and Ser23 of IκBβ). In addition, there are non-inducible pathways for the ubiquitylation and degradation of IκB, which result in the targeting of this protein for degradation by the proteasome³⁵.

Several members of the mitogen-activated protein kinase kinase kinase (MAP3K) family, such as NF-κB-inducing kinase (NIK) and MEKK1, affect the activation of NF-κB (FIG. 1). NIK was thought initially to phosphorylate and activate IKKα preferentially, whereas MEKK1 was proposed to phosphorylate and activate IKKβ preferentially^36,37. However, recent work with Nik^−/− and Mekk1^−/− mice indicates that NIK and MEKK1 are not essential for TNF- and IL-1-mediated induction of NF-κB, but are essential for the activation of NF-κB by other factors. For example, NIK is required for the lymphotoxin-β (LTβ)-mediated activation of NF-κB³⁸. Once in the nucleus, NF-κB binds DNA; then, additional proteins, such as CREB-binding protein (CBP)/p300, bind to NF-κB³⁹ and stabilize transactivation through the formation of an enhanceosome-like structure.

Figure 1. NF-κB is activated by plasma-membrane receptors that transduce signals to kinases such as PI3K.

Phosphatidylinositol 3-kinase (PI3K) can activate AKT, which potentially affects the phosphorylation of both inhibitor of NF-κB (IκB) kinase (IKK; a complex of IKKα, IKKβ and NEMO) and p65 (RELA). In addition, linker proteins, such as tumour-necrosis factor (TNF)-receptor-associated factors (TRAFs), participate in the activation of NF-κB-inducing kinase (NIK), as well as transforming-growth-factor-β-activated kinase 1 (TAK1). NIK-mediated enhancement of NF-κB activity might be facilitated by its activation of mitogen-activated protein (MAP) kinases, such as extracellular-signal-regulated kinase 1 (ERK1) and ERK2. The activation of IKK leads to enhanced phosphorylation of IκB, followed by its ubiquitylation then degradation by the 26S proteasome. This frees the NF-κB p50 and p65 subunits to be transported to the nucleus. Nuclear NF-κB p50–p65 heterodimers that are not phosphorylated on the p65 subunit can only partially activate gene transcription, whereas if p65 is phosphorylated, the enhancement of transcription is greatly magnified. LTβ, lymphotoxin-β; LTβR, lymphotoxin-β receptor; NEMO, NF-κB essential modulator; PtdInsP₃, phosphatidylinositol-3,4,5-trisphosphate; PTEN, phosphatase and tensin homologue.

The dysregulation of NF-κB in haematopoietic tumours, as well as solid tumours, has gradually become established as a paradigm³¹ (BOX 2). This alteration in NF-κB activity is often accompanied by altered expression of chemokines, growth factors and cytokines, oncogenes and tumour suppressors^30,40–46, adhesion proteins and proteases⁴⁷, and inhibitors of apoptosis⁴⁸. Increased expression of p50 or p100 (the precursor of p52) has been reported in biopsies of non-small-cell lung cancer and human breast cancer, as well as head and neck cancer cells^49–52. During TPA/DMBA (tetradecanoylphorbol 13-acetate/7,12-dimethylbenz(a)anthracene)-mediated induction of mouse skin cancer, there is an increase in the nuclear localization of p50 and B-cell leukaemia/lymphoma 3 (Bcl3) in the developing tumours⁵³. The transformation of human thyroid cancer cell lines is an NF-κB-related event⁵⁴. Nuclear localization of NF-κB is associated with breast cancer and might be related to the escape of these tumour cells from apoptosis⁵⁵. Malignant melanoma cells have elevated constitutive activation of IKK compared with normal human epidermal melanocytes, which results in the more rapid degradation of IκB, nuclear localization of p50–p65 NF-κB and enhanced transactivating capacity of the NF-κB complex^23,56,57. As the activation of NF-κB induces the transcription of inhibitors of apoptosis, as well as factors that are associated with tumour angiogenesis, metastasis and growth, we have come to understand that NF-κB could be an important potential therapeutic target in cancer, particularly if we can identify which of the upstream kinases that affect NF-κB become dysregulated during tumorigenesis. Moreover, it has been shown clearly that blocking the activity of NF-κB targets tumour cells for apoptosis^12,58. In cell models in which the activation of NF-κB facilitates the escape of tumour cells from apoptosis^59,60, the co-administration of an inhibitor of NF-κB, such as the proteasome inhibitor PS-341, would be expected to increase tumour kill after irradiation¹⁴.

NIK-mediated modulation of NF-κB

NIK was first identified on the basis of its association with TNF-RECEPTOR-ASSOCIATED FACTOR 2 (TRAF2)⁶¹. NIK associates with, and is reported to activate the phosphorylation of, IKKα and IKKβ. When overexpressed, NIK enhances NF-κB activity, and the expression of kinase-defective NIK blocks NF-κB activation in response to most inducers. However, NIK-deficient mice still respond to TNF and IL-1 with enhanced NF-κB activity, which indicates that NIK is not required for the activation of NF-κB by these cytokines. NIK is involved in the processing of p100 to the NF-κB p52 subunit⁶². This p52 subunit either homodimerizes or interacts with p65 to form the NF-κB p65–p52 heterodimer. So, the enhanced NIK activity in tumours⁵⁷ could be associated with the enhanced level of nuclear p52 that is reported to be present in breast cancers⁶³. NIK is associated also with the phosphorylation and activation of IKKα, which can affect the transactivation potential of p65 (REF. ⁶⁴). Mice that have a mutation in NIK (aly/aly mice) have a loss of signalling in response to the chemokines CCL21 and CXCL13 (REF. ⁶⁴). The chemotactic response is lost, as is the activation of NF-κB in response to CCL21-mediated activation of CCR7 (REF. ⁶⁵).

Box 2 | Tumours that have elevated NF-κB

• Melanoma^{23,26,28,40–41,56}

• Mammary carcinoma^49,55,63,85

• Non-small-cell lung carcinoma⁵¹

• Colorectal carcinoma^52,58

• Squamous-cell carcinoma^50,53

• Leukaemia^47,81,82,114

• Lymphoma^47,116

• Thyroid carcinoma⁵⁴

• Fibrosarcoma¹¹⁷

• Pancreatic cancer¹¹⁸

• Prostate cancer¹¹⁹

• Multiple myeloma¹¹³

• Ovarian cancer¹⁸

Dysregulation of NIK NF-κB activity in melanoma

IKK-associated NIK activity is enhanced in melanoma cell lines compared with normal human epidermal melanocytes (NHEMs)⁵⁷. A catalytically inactive form of NIK blocks constitutive NF-κB or CXCL1 promoter activity in melanoma cells, but not in control NHEMs. Transient overexpression of wild-type NIK results in increased phosphorylation of ERK1 and ERK2. This phosphorylation is inhibited in a concentration-dependent manner by PD98059, an inhibitor of p42/p44 MAPK. Moreover, the NF-κB promoter activity in melanoma cells decreases with the over-expression of dominant-negative ERK⁵⁷. Although these data support the hypothesis that NIK-mediated ERK activation contributes to NF-κB activity, they should not be over-interpreted because transfection/overexpression models were used.

AKT modulates NF-κB transactivation

Once IκB is degraded by the proteasome, the nuclear-localization sequence of the p50–p65 or p52–p65 NF-κB complex is exposed and, consequently, this complex can move to the nucleus and bind to the NF-κB element of the promoter of responsive genes. However, simply binding to the DNA is not sufficient. The p65 component needs to be phosphorylated in its transactivation domain to be fully active. Zhong et al.⁶⁶ showed that phosphorylation of p65 on Ser276 by protein kinase A (PKA) is required for its transactivation functions. In addition, other kinases, including casein kinase II (CKII), phosphorylate p65 at serine residues that differ from those that are phosphorylated by PKA, and this enhances the transcription activity of NF-κB⁶⁷. The p65 phosphorylation event might also be mediated by IKKβ⁶⁸ or, possibly, IKKα⁶⁹. In addition, it has been proposed recently that the serine/threonine protein kinase AKT, also known as PKB — which was isolated from the AKR mouse thymoma — might also facilitate the phosphorylation of p65 on Ser529 and Ser536 (REFS ^69–73). In response to the activation of PI3K, AKT is recruited to the cell membrane by phosphatidylinositol-3,4,5-trisphosphate (PtdInsP₃) or phosphatidylinositol-3,4-bisphosphate, where it becomes phosphorylated by 3′-phosphoinositide-dependent kinase 1 (PDK1) on Thr308 in the kinase domain and by PDK2 on Ser473. AKT phosphorylates several substrates at an Arg-Xaa-Arg-Xaa-Xaa-Ser/Thr motif, some of which mediate escape from apoptosis and contribute to the activation of nuclear NF-κB. Indeed, IKKα has a perfect consensus sequence for AKT-mediated phosphorylation, and Li and Stark have suggested that AKT-mediated activation of IKKα could occur upstream of the phosphorylation of p65 by IKK⁷⁰. However, work from Delhase et al. does not support this possibility⁷¹. Moreover, phosphorylation of the p50 NF-κB subunit by the PI3K–AKT pathway is also reported to increase the binding of the NF-κB complex to DNA⁷².

The lipid phosphatase known as phosphatase and tensin homologue (PTEN) is known to switch off the activation of AKT by dephosphorylating PtdInsP₃, which blocks the membrane localization of AKT. PTEN is also reported to block NF-κB activation without affecting the IκB degradation pathway⁷². PTEN is often deleted or mutated in melanoma and other types of tumour^74,75. The activation of AKT that we observe in some human melanoma cells in vitro is accompanied by the loss or reduction of expression of PTEN (P. Dhawan, D. Ellis and R. A., unpublished observations). The loss of PTEN with consequent enhanced activation of AKT has also been reported for several other types of tumour, which indicates that this is a frequent step associated with transformation.

RAS-mediated activation of NF-κB

The activation of NF-κB can suppress apoptosis induced by oncogenic RAS⁷⁶. The RAS superfamily is comprised of small GTPase proteins. One family member, HRAS (named after the Harvey-Ras oncogene), is associated with transformation. Paradoxically, when cells are transformed with HRAS, transcriptional activity is mediated through an NF-κB element, although this does not result in an increase in the nuclear level of p50–p65 (REF. ⁷⁷). Oncogenic RAS activates NF-κB transcriptional activity⁷⁷. Moreover, inhibitors of PI3K block the endogenous NF-κB activity of malignant melanoma cells, although these inhibitors will not block IKK-mediated phosphorylation of IκBα (P. Dhawan, D. Ellis and R.A., unpublished observations). Embryonic fibroblasts from both IKKβ- and IKKα-null mice cannot induce NF-κB transactivation. However, the restoration of IKKβ in IKKαβ-null mice will not restore cytokine-stimulated activation of RELA/p65 (REF. ⁶⁹). Both IKKα and IKKβ are required for the activation of the transactivating activity of the RELA/p65 subunit of NF-κB, presumably because IKKα is required for the phosphorylation and transactivation of p65 (REF. ⁶⁹). So, one can envision a pathway whereby RAS activates PI3K, which activates AKT, which activates IKKα-mediated phosphorylation of p65.

Other kinases can modulate NF-κB

Although many kinases have been shown to contribute to the activation of NF-κB, it has been difficult to show clearly the requirement for an activating kinase upstream of the IKK complex. NIK is a MAP3K, as are TGF-β-activated kinase 1 (TAK1), MEKK1 and MEKK3 (see REF. 70 for a review). Deletion of NIK and MEKK1 does not block TNF- or IL-1-mediated induction of IKK activity, but the possibility that TAK1 and MEKK3 are potential modulators of IL-1- and TNF-mediated induction of IKK remains open. NIK is required for LTβ-receptor-mediated activation of IKK³⁸. In addition to the MAP3Ks and AKT, several other indirect modulators of NF-κB activation have been described, including RAS, RAF and TAX^78–83. Protein kinase C (PKC) has also been noted as a potential modulator of IKK activity, and PKC-mediated activation of IKK is linked to tumorigenesis^84–88. So, when trying to discern why there is dysregulation of NF-κB in tumour cells, it is important to look at several upstream signal-transduction pathways.

Chemokines and NF-κB

NF-κB has been shown to be an essential modulator of transcription of the following chemokines: CXCL1, -2, -3, -5, -8, -9, -10 and -12, and CCL2, -3, -4, -5, -11 and -17. However, the context of the NF-κB element in the promoter of the 5′ regulatory regions of these genes varies considerably. This finding probably indicates that there are more-subtle mechanisms for the modulation of transcription that involve other co-regulatory factors. The constitutive expression of chemokines in tumour cells has been documented extensively. However, even though the expression of many of the chemokines is regulated by NF-κB and tumours have constitutive activation of NF-κB, there is variability in the expression of specific chemokines between tumours. As the ELR-motif-containing CXC-chemokines CXCL1 and CXCL8 are known for their modulation of tumour angiogenesis and tumour growth, I concentrate here on the modulation of the transcription of these chemokines by NF-κB.

Activation of NF-κB enhances chemokine transcription

The constitutive activation of NF-κB facilitates the immortalization of melanoma and mammary tumour cells, together with their escape from apoptosis⁸⁹. One consequence of this dysregulated NF-κB activity^{23,55–58,61–68} is that there is constitutive transcription of the NF-κB-dependent chemokines CXCL1 and CXCL8 in human tumours^90–96. Constitutive expression of these chemokines can transform immortalized melanocytes and is associated with increased metastasis in colon carcinoma, squamous-cell carcinoma and melanoma^19,28,91,97.

Transcriptional regulation of CXCL1 and CXCL8

Although cytokine-mediated activation of NF-κB is the main mechanism for the transcriptional induction of chemokines, NF-κB does not act alone in this induction. Moreover, the factors that interact with NF-κB to modulate transcription differ between CXCL1 and CXCL8 (FIG. 2). Whereas cooperation between NF-κB and either AP1 or the CCAAT/enhancer-binding protein (C/EBP) element is required for the transcriptional modulation of CXCL8 (REF. ⁹⁶), for CXCL1, NF-κB cooperates with a transcription factor that binds to GC-rich complexes (SP1), the high-mobility group protein HMGIY (as determined by an ELECTROPHORETIC MOBILITY-SHIFT ASSAY and immediate upstream region (IUR) elements to modulate the transcription of this chemokine⁴². SP1 and IUR elements are involved in basal and cytokine-induced transcription of the CXCL1 gene. The factors that bind to these elements might form an enhanceosome-like transcriptional response element. The IUR binds both positive and negative regulatory factors. The 115-kDa poly-ADP ribose polymerase (PARP) binds to the TCGATC sequence of the IUR⁹⁸, and thereby positively modulates CXCL1 transcription, perhaps by displacing the binding of the negative modulatory factor CCAAT-displacement protein (CDP)⁹⁹. Inhibitors of the ADP-ribose-polymerase function of PARP reduce CXCL1 promoter activity, which indicates that not only DNA binding, but also the enzymatic activity of PARP, might be required for the co-activating function of PARP in CXCL1 transcription^98,99.

Figure 2. Similarities between the proposed enhanceosomes for IL-6, CXCL8 and CXCL1.

Similar key regulatory elements participate in the modulation of transcription of CXCL8, interleukin-6 (IL-6) and CXCL1. The transcription of CXCL8 is largely modulated through a nuclear factor-κB (NF-κB) element, which works in concert with adjacent activating protein 1 (AP1) and/or C/EBP elements to modulate the expression of this chemokine. The transcription of CXCL8 is also modulated by an NF-κB-repressing factor (NRF)¹²⁰. This negative regulatory element has a dual role; in the absence of stimulation, NRF inhibits the transcription of CXCL8, but after an IL-1 signal, the NRF is required for the full induction of CXCL8 transcription. An enhanceosome model of cytokine gene expression, analogous to the CXCL8 paradigm, has been proposed for the regulation of the IL-6 promoter. Both CXCL8 and IL-6 promoters have binding sequences for NF-κB, C/EBP and the TATA BOX¹²¹. There is competition between NF-κB and an inhibitory factor, RBP-Jκ for the NF-κB promotor element. Promoter activation relies on the p65 NF-κB subunit, which recruits CREB-binding protein (CBP/p300) to the site¹²². The binding of CBP/p300, which has histone acetylase (HAT) activity, will stabilize transcription from these promoters. In an independent study, CCAAT displacement protein (CDP) has been shown to interact physically with CBP/p300 and is a target for acetylation at specific residues near the homeodomain. So, CDP and CBP seem to have antagonistic roles in the regulation of IL-6 and CXCL8 transcription. The transcriptional repression of CDP is probably associated with its ability to recruit the histone-deacetylase activity of HDAC1. HDAC1 deacetylates the histones and, as a result, the chromatin structure is closed and unavailable for transcription, with the end result being the consequent silencing of the chromatin. We propose that similar interactions between NF-κB, CBP and CDP might be involved in the transcription of CXCL1. c/EBP, CCAAT, enhancer-binding protein; CRE, cyclic AMP response element; IRE, interferon response element; IUR, immediate upstream region; PARP, poly-ADP ribose polymerase; RBP-Jκ, recombination signal binding protein-Jκ.

Within the IUR element, there is an overlapping consensus sequence (GGGATCGATC) that binds the 170–180-kDa CDP protein. Mutations in the IUR element that map to the putative CDP-binding site inhibit the binding of CDP to the IUR element and favour increased transcription from the CXCL1 promoter. CDP can negatively regulate the expression of CXCL1, whereas PARP can potentiate the transcription of this gene^99–101. CDP has been shown previously to repress actively the transcription of several other genes^101–105 either by displacing positive transactivating factors or through the recruitment of histone DEACETYLASE (HDAC), which would affect the open/closed state of the chromatin and thereby affect transcription. The action of CDP might be antagonized by the binding of CBP/p300, which has been shown to bind NF-κB and SP1, and —through its histone-acetylase activity — to stabilize the enhanceosome and enhance transcription^25,106,107.

The transcription of CXCL8 is modulated mainly through an NF-κB element that works in concert with adjacent AP1 and nuclear factor induced by IL-6 (NF-IL-6) elements. There is some cell-type specificity with regard to the requirement for transactivation through either the AP1 and NF-κB elements or the NF-IL-6 and NF-κB elements⁹⁷. The transcription of CXCL8 is also modulated by an NF-κB-repressing factor (NRF).

Inhibitors of NF-κB for tumour therapy

It is now clear that many types of tumour have constitutive activation of NF-κB (BOX 2). This activation of NF-κB leads to the endogenous expression of factors that are associated with tumorigenesis (VEGF, CXCL1 and CXCL8), escape from apoptosis (inhibitor of apoptosis proteins, IAPs) and metastasis (metalloproteinases). However, whereas the expression of CXCL1, CXCL8 and CCL2 is associated with increased tumorigenicity and metastasis — CCL1, CXCL1 and CXCL8 in melanoma and colon adenocarcinoma^19,108,109 — the treatment of various types of tumour with CXCL9, CXCL10, CCL1, CCL3, CCL5, CCL16, CCL19, CCL20, CCL21 and XCL1 is associated with tumour regression and increased survival, possibly due to the recruitment of infiltrating leukocytes and dendritic cells (DCs). As the expression of many of these chemokines is modulated by NF-κB also, the inhibition of NF-κB presents the risk of inhibiting the migration of natural killer (NK) cells, tumour-infiltrating lymphocytes and DCs into the developing tumour. However, this negative effect is, apparently, offset by the enhanced apoptosis of the tumour cells due to the inhibition of NF-κB. Indeed, treatment with NON-STEROIDAL ANTI-INFLAMMATORY DRUGS (NSAIDS) has been shown to inhibit both NF-κB and tumour growth, without detrimental effects on the host³¹. So, reagents such as sulindac or other NSAIDs, the proteasome inhibitor PS-341 and the IKK adaptor protein NEMO-binding peptide are potential therapeutic reagents for tumours^60,110–112 (FIG. 3). Indeed, initial preclinical and clinical data with PS-341 indicate that it will provide a powerful means of killing tumour cells by targeting them for apoptosis^113–115.

Figure 3. Model of potential components involved in the constitutive activation of NF-κB and enhanced expression of CXCL1 and CXCL8 in melanoma.

We postulate that the constitutive activation of nuclear factor-κB (NF-κB) is modulated by receptor-mediated signals, such as those mediated through chemokine receptors and the LTβ receptor, which result in the persistent activation of NF-κB-inducing kinase (NIK), AKT and, potentially, mitogen-activated protein (MAP) kinase kinase kinase 1 (MEKK1). This activation of the inhibitor of NF-κB (IκB) kinase (IKK; a complex of IKKα, IKKβ and NEMO) leads to the phosphorylation and degradation of IκB and the phosphorylation of RELA/p65, which can, in turn, lead to constitutive expression of chemokines. We postulate that chemokine expression can be blocked by treatment with non-steroidal anti-inflammatory drugs (NSAIDs), NEMO-binding peptide and PS-341, which would target tumour cells for apoptosis. CBP, CREB-binding protein; CDP, CCAAT displacement protein; HAT, histone acetyltransferase; HDAC1, histone deacetylase 1; IUR, immediate upstream region; LTβ, lymphotoxin-β; LTβR, lymphotoxin-β receptor; NEMO, NF-κB essential modulator; PARP, poly-ADP ribose polymerase; PI3K, phosphatidylinositol 3-kinase; PtdInsP₃, phosphatidylinositol-3,4,5-trisphosphate; PTEN, phosphatase and tensin homologue; TRAF, tumour-necrosis-factor-receptor-associated factor.

Conclusions

It has become clear that the dysregulation of NF-κB is a potential target for therapeutic intervention for cancers. This is due, in part, to the role of NF-κB in the modulation of expression of chemokines and inflammatory cytokines, and also to the role of NF-κB in the regulation of expression of inhibitors of apoptosis. An important point for consideration, however, is that NF-κB also modulates the expression of the angiostatic chemokines that participate in the recruitment of tumour-infiltrating lymphocytes and NK cells into the developing tumour. So, blocking NF-κB could be a double-edged sword, were it not for the fact that NF-κB does not act alone, but cooperates with several other transcription factors to modulate gene expression. The angiostatic chemokines are transcriptionally activated by IFN-γ, in addition to NF-κB. Clearly, characterizing the enhanceosome for the angiostatic chemokines, the chemokines that suppress tumorigenesis, and those that stimulate tumour growth and angiogenesis will allow investigators to develop more-precise ways to block the growth of tumours while maintaining full participation of the immune system in this blocking process.

ENHANCEOSOME

Gene transcription is achieved by the assembly of higher-order, three-dimensional transcription factor/enhancer DNA complexes, termed enhanceosomes. Enhanceosomes activate transcription by recruiting chromatin-modifying activities and basal transcription factors to the nearby promoters

SCID

(Severe combined immunodeficiency). Mice with this defect in their immune system do not have B or T cells and can, therefore, accept tumour cells from another species without rejection

DIFFERENTIAL DISPLAY

This is a powerful tool for the comparison of gene expression between two or more messenger RNA populations

MICROARRAYS

This technique allows the screening of messenger RNA extracted from cells against DNA from many thousands of genes. The DNA from each gene is positioned on a solid support in a highly ordered array

ANOXIA

Limited oxygen supply

FOCUS FORMATION

The ability of tumour cells to grow in an anchorage-independent manner by adhering to one another, thereby forming ‘foci’ of tumour cells

NEMO

(NF-κB essential modulator). This is the regulatory component of the inhibitor of NF-κB (IκB) kinase (IKK) complex, also known as IKKγ

TNF-RECEPTOR-ASSOCIATED FACTORS

(TRAFs). A term that originated with proteins that were found to bind to the cytoplasmic domain of the tumour-necrosis factor (TNF) receptor in a yeast two-hybrid screen

TATA BOX

A DNA motif that binds several factors (TATA-binding proteins, TBPs and TBP-associated factors TAFs) that facilitate the initiation of transcription

ELECTROPHORETIC MOBILITY-SHIFT ASSAY

A technique for detecting DNA–protein complex formation. It involves the incubation of nuclear extracts with a radiolabelled oligonucleotide probe, then separating the probe that has bound to nuclear proteins from the free radiolabelled probe by gel electrophoresis, followed by autoradiography

DEACETYLATION

Acetylation is a post-translational modification of chromatin components, particularly histones. Histone deacetylases have been identified as components of nuclear co-repressor complexes

NON-STEROIDAL ANTI-INFLAMMATORY DRUGS

(NSAIDs). Drugs, such as aspirin, that are used to ablate the inflammatory response. These drugs can stimulate apoptosis and inhibit angiogenesis, thereby suppressing malignant transformation and tumour growth. They work, in part, by suppressing NF-κB activation