Nuclear Factor-κB is an Important Modulator of the Altered Gene Expression Profile and Malignant Phenotype in Squamous Cell Carcinoma

Many cancers share a complex set of phenotypic traits, suggesting that common molecular regulatory mechanisms underlie their development (1). These traits include inhibition of programmed cell death and increased proliferation, migration, new blood vessel formation (angiogenesis), and inflammation. The resemblance of these complex phenotypic changes in cancer to those occurring transiently during the response to injury and infection suggests the hypothesis that oncogenic activation of injury response pathways and their target genes may be important in the pathogenesis of cancer. Consistent with this hypothesis, our investigations of molecular alterations in squamous cell carcinomas (SCCs) that arise from the skin and upper aerodigestive tract have revealed constitutive expression of multiple integrin cell adhesion molecules and proinflammatory cytokines that are conditionally expressed in response to injury (2, 3). A wider molecular comparison between normal, transformed, and metastatic keratinocytes by mRNA differential display and cDNA microarray revealed altered expression of an even greater diversity of genes, including genes functionally related to growth, apoptosis, adhesion, angiogenesis, and inflammation (4). We noted that a number of these cytokines, cell adhesion molecules, and other genes expressed in SCC have been associated with activation or as targets of the early injury response transcription factor, nuclear factor (NF)-κB (4, 5, 6, 7).

NF-κB is expressed in an inactive form in most cells, composed of NF-κB p50 and Rel A p65 subunits, and bound to an inhibitory protein, IκBα (8). In response to DNA damage, cytokine, or growth factor signaling, IκBα is phosphorylated, ubiquitinated, and degraded by the proteasome, enabling nuclear localization and binding of p50/p65 to the promoter region of target genes. We established that p50/p65 is the predominant form of NF-κB constitutively activated in human head and neck and murine SCCs and that NF-κB promotes the overexpression of the proinflammatory and angiogenesis factor homologues interleukin-8 and growth regulated oncogene-1 [Gro-1 (5, 6)]. Inhibition of NF-κB by a dominant negative IκBα phosphorylation mutant (IκBαM) or pharmacological antagonists inhibited tumor cell survival, cytokine expression, angiogenesis, and tumorigenesis by human and murine SCC (5, 7, 9, 10).

Molecular profiling of changes in gene expression with stepwise transformation and metastatic progression of SCC in the syngeneic murine Pam 212 model revealed altered expression of a diversity of genes, including a cluster of overexpressed genes in lymph node metastases (Pam LY cells) related to the NF-κB pathway (4). However, we observed that the inhibitory effects on cell survival of expression of IκBαM under a constitutive promoter made it difficult to obtain stable transfectants of Pam LY and other SCCs for the study of the role of NF-κB in gene expression and the malignant phenotype (7).

Here, we examined the overall differences in gene expression between normal, transformed, and metastatic keratinocytes using 15,000 cDNA microarrays, and we explored the role of NF-κB in the genetic and phenotypic changes acquired during tumor progression in the metastatic cells by genetically switching off NF-κB with IκBαΜ under control of a doxycycline [DOX (Tet)]-inducible promoter. We obtained evidence that NF-κB modulates a broad program of genes differentially expressed with tumor progression of this SCC. We further established that inhibition of NF-κB inhibited proliferation, cell survival, migration, angiogenesis, and tumorigenesis of SCC cells. These results provide evidence that NF-κB is an essential molecular switch for the gene expression program and malignant phenotype in SCC.

The murine SCC line Pam 212 was derived from spontaneously transformed BALB/c keratinocytes, and LY-2 was isolated from lymph node metastases that developed in BALB/c mice after inoculation of the parental PAM 212 SCC line (11). LY-2 has been shown to grow more aggressively and have a higher constitutive NF-κB activity than the PAM 212 line (5, 12). The SCC lines were maintained in Eagle’s Minimal Essential Medium (EMEM) (Life Technologies, Inc., Grand Island, NY) with 10% fetal calf serum, 1% glutamine, and 0.5% penicillin/streptomycin at 37°C in 5% CO₂. Cell lines were determined to be Mycoplasma-free by nested reverse transcription-polymerase chain reaction.

The Tet-On expression system used was obtained from Clontech Laboratories, Inc., (Palo Alto, CA). LY-2 cells were cotransfected with the Tet-On plasmid system containing the Tet-responsive transcriptional activator that binds the Tet-responsive element (TRE) and the gentamicin resistance gene (neo), together with the tetR plasmid that binds the TRE in the absence of DOX that reduces background levels of transcription. Cells (1.25 × 10⁶) were plated into 100-mm dishes, incubated overnight, and transfected with 1.6 μg of the Tet-On plasmid, 14.4 μg of tetR, and 128 μL per reaction of LipofectAMINE (Life Technologies, Inc., Gaithersburg, MD) in Opti-MEM (Life Technologies, Inc., Grand Island, NY). Clones containing the Tet-On plasmid were selected by limiting dilution in 96-well plates using selection medium containing 600 μg/mL G418. Clones were expanded and screened for responsiveness to DOX using a TRE-luciferase assay. Fourteen clones with low background and high fold induction of luciferase reporter activity after the addition of 2 μg/mL DOX were selected and pooled as controls (LY-2P) and for transfection by the response plasmid containing IκBαM. The 14 LY-2 clones containing the Tet-On plasmid were transferred into 6-well plates at a concentration of 2 × 10⁵ cells/well and incubated overnight at 37°C. The cells were transfected with 1 μg of TRE-IκBαM plasmid (pTRE-Igs32s36 from Clontech Laboratories, Inc.), 4 μL of LipofectAMINE, and 6 μL of PLUS reagent (Life Technologies, Inc., Gaithersburg, MD) per reaction in Opti-MEM. Reactions were incubated for 5 hours at 37°C in 5% CO₂. The medium was changed to complete EMEM (Life Technologies, Inc.), and cells were allowed to recover overnight. To select subclones containing both the Tet-On and TRE-IκBαM plasmids, transfectants were diluted 1:40 for subcloning using one 200-mm dish for each clone. After 7 days in EMEM containing 600 μg/mL G418, colonies were selected using cloning cylinders (Genechoice, Frederick, MD), and four subclones from each dish, labeled A to D, were transferred to a 24-well plate and expanded for screening for Tet-inducible expression of IκBαM by Western blot, as described previously (5).

Four clones selected for expression of IκBαΜ by Western blot and the LY-2P parent cells were transferred to triplicate wells at a concentration of 2 × 10⁵ cells/well in 6-well tissue culture plates. Cells were incubated overnight at 37°C and then transfected with 0.95 μg of NF-κB luciferase plasmid, 4 μL of LipofectAMINE, and 6 μL of PLUS reagent in 1 mL of Opti-MEM per well for 5 hours at 37°C. The transfection medium was replaced with complete EMEM, and cells were incubated until they reached 80% confluence. NF-κB-luciferase assays were performed as described previously (10).

Double-stranded DNA probes for NF-κB (5′-AGTTGAGGGGACTTTC-CCAGGC-3′) and Oct-1 (5′-TGTCGAATGCAAATCACTAGAA-3′) were commercially obtained (Promega, Madison, WI). Probes were labeled with [γ-³²P]ATP (6000 Ci/mmol; Amersham, Arlington Heights, IL). Nuclear extracts were prepared from DOX-treated and untreated LY-2 IκBαΜ clones, and DNA binding assays were performed with 5 μg of nuclear extracts using standard methods described previously (5).

Total RNA was isolated from primary BALB/c keratinocytes, Pam 212, and LY-2 cells and the LY-2P and the four inducible LY-2 IκBαΜ clones (23D, 24C, 25D, and 26C) with and without DOX treatment. RNA was collected 72 hours after DOX treatment in three independent experiments. The 15,000-element murine microarray developed by the Radiation Oncology Sciences Program, National Cancer Institute, using a cDNA library from the National Institute of Aging (Mm-ROSP-NIA15K), was used.⁴ Sequential arrays were obtained from two different printings. Experiments were performed in triplicate. The protocols used for RNA isolation, hybridization, and microarray analysis were those available from the National Human Genomics Research Institute.⁵ Arrays were read on a Genepix 4000B Scanner, and data were collected using Genepix Pro 4.0 software (Axon Instruments, Foster City, CA).

Array elements exhibiting mean signal differences of ≥2-fold (99% confidence interval) for triplicate comparisons among cDNAs from normal keratinocytes, transformed Pam 212, and/or metastatic Pam LY-2 cells were scored as being associated with tumor progression. Array elements exhibiting reversion of mean differences of ≥2-fold (99% confidence interval) for comparisons among cDNAs from metastatic Pam LY-2P and four LY-2 IκBαM clones ± DOX were scored as putative NF-κB-modulated genes. Analysis of the collected data was performed using the Cluster and Treeview software developed by Michael Eisen (Stanford University), as described previously (4). Genes were assigned to functional families based on information from LocusLink and PubMed. Sequence data generated through the use of the Celera Discovery System and verified in public databases were used for detection of NF-κB–related motifs. The sequences from 2 kb 5′ and the full-length of each gene found to be differentially expressed by microarray were examined for NF-κB binding motifs (12) by Genomatix software and for ACTACAG motifs involved in NF-κB–mediated down-regulation of MyoD and Sox9 mRNA (13) by BLAST search available from the National Center for Bioinformatics, National Institutes of Health.

RNA isolated for microarray analysis was subjected to electrophoresis and Northern blot as described previously (4). Blots were probed with full-length ³²P-labeled DNA probes for the cyclin D1, Trp53, inhibitor of apoptosis-1 (IAP-1), β-catenin, and β-actin genes and washed under high stringency conditions.

Whole cell lysates were collected from DOX-treated and untreated IκBαΜ clones in 6-well plates after washing twice with PBS and addition of 300 μL of lysis buffer [100 mmol/L potassium phosphate, 0.2% Triton X-100, and 0.05% dithiothreitol (Tropix; PE Biosystems, Bedford, MA)] to each well. Plates were held at 4°C for 1 hour and then subjected to three freeze-thaw cycles. Lysates were transferred to 2-mL microtubes and centrifuged for 5 minutes at 13,000 rpm. Lysates were diluted 1:10 in distilled H₂O and analyzed for protein content by the BCA protein assay kit (Pierce, Rockford, IL). Samples were stored at −80°C until use. Twenty micrograms of protein were loaded onto precast 12% Tris-glycine gels (Novex, San Diego, CA). Gels were run for 2 hours at 100 V, and then proteins were transferred to nitrocellulose membranes for 90 minutes at 20 V. Membranes were blocked in 5% milk in Tris-buffered saline at 4°C overnight. For selection of clones expressing IκBαΜ after DOX treatment, blots were incubated for 2 hours at 4°C in a 1:200 dilution of anti-IκBα (rabbit polyclonal antibody; Santa Cruz Biotechnology, Santa Cruz, CA). Antibodies recognizing cyclin D1, Trp53, IAP-1, β-catenin, and β-tubulin proteins were obtained from Santa Cruz Biotechnology. Membranes were washed and incubated for 1 hour at room temperature in a 1:3,000 dilution of goat antirabbit antibody labeled with horseradish peroxidase (Santa Cruz Biotechnology). Membranes were washed, enhanced with SuperSignal chemiluminescence reagent (Pierce), placed in heat-sealable pouches (Kapak, Minneapolis, MN), and exposed to film.

Proliferation of LY-P and the IκBαΜ clones was measured using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) cell proliferation kit (Roche Molecular Biochemicals, Mannheim, Germany). Cells were transferred in triplicate to a 96-well, flat-bottomed plate at a concentration of 5 × 10³ cells/well in 100 μL of complete EMEM with or without DOX. Cells were incubated at 37°C in 5% CO₂ for 1, 3, or 5 days, and growth rates were analyzed after the addition of MTT reagent to the cultured cells and lysis according to the manufacturer’s instructions. Absorbance was determined using multichannel spectrophotometry at a wavelength of 570 nm.

LY-2P, 23D, and 24C cells were transferred to 6-well plates at a concentration of 5 × 10⁵ cells/well and treated with 2 μg/mL DOX or left untreated. After 96 hours, floating and adherent cells were collected and counted. One million cells were washed twice in buffer solution from the CycleTest Plus DNA Reagent Kit (Becton Dickinson, San Jose, CA) and then resuspended in 250 μL of trypsin buffer. Cells were incubated for 10 minutes at room temperature. Two hundred microliters of trypsin inhibitor/RNase buffer were added to each tube, and cells were incubated for 10 minutes at room temperature. Two hundred microliters of cold propidium iodide staining solution were added. Cells were incubated for 10 minutes in the dark on ice, and within 3 hours of staining, quantification of DNA was determined on a FACScan flow cytometer using CellQuest software (Becton Dickinson).

Invasion assays were performed using a synthetic basement membrane cell invasion assay kit (Chemicon, Temecula, CA). Cells were plated at a concentration of 0.5 × 10⁶ cells/mL in serum-free EMEM into the inserts of a cell invasion assay plate. Plates were incubated at 37°C in 5% CO₂ for 72 hours. Cells on the interior of the inserts were removed by swabbing, and the exterior of the inserts was stained for evidence of cell migration through the synthetic basement membrane. Stained cells were solubilized in 10% acetic acid, and the intensity of staining was quantified by transfer of 100-μL aliquots to a 96-well plate for spectrophotometry at a wavelength of 570 nm.

LY-2P cells (5 × 10⁶) as a control and LY-2 24C cells were inoculated in 200 μL of serum-free EMEM subcutaneously into the flanks of congenic male BALB/c severe combined immunodeficient (SCID) mice (5 mice per group). Animals received water treated with 2 μg/mL DOX beginning at week 7. Tumor measurements were recorded once a week for the duration of tumor growth to 10 weeks. After euthanasia, tumors were harvested, prepared for frozen section, and stained with hematoxylin and eosin (H&E) or antibody specific for CD31 as described previously (9).

A stepwise model of cancer development and progression was established previously from syngeneic BALB/c mice consisting of normal keratinocytes (KER), a transformed SCC cell line (Pam 212), and SCC lymph node metastases, designated Pam LY (4, 11). Pam 212 cells produce slowly growing SCC that rarely metastasizes, and Pam LY cells isolated from rare lymph node metastases of Pam 212 are variants that form cancers that grow and metastasize rapidly. Molecular comparisons between these normal, transformed, and metastatic keratinocytes by mRNA differential display and a first-generation 4,000-element cDNA microarray provided evidence for altered expression of multiple genes related to growth, apoptosis, angiogenesis, inflammation, and the NF-κB pathway (4).

To directly explore the role of NF-κB in the accumulated changes in gene expression acquired with tumor progression of Pam LY, we sequentially transfected plasmids containing a TRE and the IκB signal mutant IκBαM under the control of a DOX-inducible promoter into the Pam LY-2 line, a representative metastatic line that exhibits constitutive activation of NF-κB (5). Multiple LY-2 cell clones that showed low background and high inducibility by the Tet derivative DOX in a Tet reporter assay after first-stage transfection were pooled for transfection by the Tet-IκBαM construct and for use as controls. Expression of IκBαM by LY-2 clones was detected by Western blot as described previously (5), and four clones that exhibited low basal and strong DOX-induced expression of IκBαM were selected to examine effects on NF-κB activation. The effect of DOX-induced IκBαM expression on constitutive NF-κB-luciferase reporter gene activity in the four LY-2 subclones and LY-2P control lacking Tet-IκBαM is shown in Fig. 1,A. IκBαM inhibited NF-κB functional promoter activity as determined using NF-κB-luciferase reporter assay (Fig. 1,A). Progressive inhibition of constitutive NF-κB reporter activity to <95% of baseline was observed by day 4 after addition of DOX to LY-2 IκBαM clones. NF-κB reporter activity in control LY-2P cells lacking Tet-IκBαM was not suppressed below control levels (Fig. 1,A). IκBαM also suppressed NF-κB DNA binding activity in nuclear extracts of LY-2 clones in electromobility shift assay (Fig. 1 B). The specificity and composition of constitutive NF-κBp50/Rel Ap65 DNA binding activity in LY-2 cells were reported previously (5).

To directly explore the alterations and role of NF-κB in gene expression with tumor progression, we used the National Cancer Institute 15,000-element cDNA microarray expressing known genes to examine the diversity of genes differentially expressed among keratinocytes, transformed Pam 212, and metastatic Pam LY-2 (Met) SCC cells, and we determined the effect of conditionally inhibiting NF-κB in Pam LY-2 cells by expressing IκBαM. Complementary DNA was prepared from mRNA from three independent preparations of keratinocytes, Pam 212, and Pam LY-2 cells and from four Pam LY-2 clones expressing IκBαM after culture with DOX for 72 hours. Messenger RNA was isolated at 72 hours based on the ≥75% inhibition of NF-κB reporter activity detected in Fig. 1 and to minimize cell death or RNA degradation, which was observed beyond 96 hours (Figs. 3 and 5; data not shown). Fig. 2 shows the cluster analysis for 308 genes (horizontal rows) that exhibited ≥2-fold difference in expression among keratinocytes (Ker), transformed Pam 212 (Trans), or metastatic Pam LY-2 (Met) cells and Pam LY-2 cells expressing IκBαM after DOX (vertical columns). Overall, the differential pattern of expression detected for 308 of 15,000 of the gene elements on the array represents about 2% of the genes compared. Three dominant patterns were detected that clustered according to the stepwise differences in phenotype among KER, Pam 212, and Pam LY-2 cells. Cluster 1 genes showed similar expression in normal and transformed Pam 212 cells but reduced expression in metastatic LY-2 cells. Cluster 2 genes showed lower expression in normal cells but increased expression in transformed and metastatic cells. Cluster 3 genes showed increased expression primarily in LY-2 metastatic cells. Overall, 141 genes in the upper cluster showed a decrease, and 167 genes in clusters 2 and 3 were increased in association with transformation, tumor progression, and activation of NF-κB.

Tables 1 and 2 show a selected list from the 308 genes detected by microarray and cluster analysis that exhibited ≥2-fold difference in hybridization with malignant progression and/or inhibition of NF-κB (the full list is available online).⁶ Table 1 includes the genes that show increased hybridization, and Table 2 includes the genes that show decreased hybridization by cDNA of Pam 212 and/or LY-2 cells. The genes were classified according to putative function and published associations with NF-κB as determined by search of National Center for Biotechnology Information PubMed and LocusLink. The ratio of LY-2 to keratinocyte hybridization, representing the overall change with malignant progression, and the ratio of LY-2 to LY-2 IκBαM, representing the effects of expression of IκBαM, are shown. The genes that are modulated by ≥2-fold by IκBαM are shown in bold.

Remarkably, Fig. 2 and Tables 1 and 2 reveal that expression of IκBαM in Pam LY-2 results in restoration of the pattern of expression observed for keratinocytes for many of the genes up- and down-regulated between keratinocytes and Pam LY-2 metastatic cells. Of the genes up-regulated in LY-2 cells, hybrization of 115 of 167 (69%) decreased by ≥2-fold toward normal levels after expression of IκBαM. Fully 88 of 141 (62%) genes down-regulated in LY-2 cells were restored by ≥2-fold back toward the level observed in keratinocytes after expression of IκBαM. The effect of IκBαM to up-modulate as well as down-modulate mRNAs provides evidence that the wide effect on gene expression is not merely due to degradation of mRNA during the process of cell death, and we verified the viability of cells and integrity of mRNA after 72 hours of exposure to DOX (data not shown).

As an initial assessment of the consistency and validity of the differential gene expression profile obtained for LY-2-expressing cells using the 15,000-element cDNA microarray, we identified genes that were previously found to be differentially expressed in multiple LY metastasis cell lines by a first-generation 4,000-element murine cDNA microarray and confirmed by reverse transcription-polymerase chain reaction and/or Northern blot analysis (4). Several genes in Table 1 that were classified as increased in LY-2 cells by ≥2-fold were included and detected in both array studies. These include the increase in expression of proliferation and antiapoptosis-related genes cyclin D1 (Ccnd1), proliferating cell nuclear antigen (Pcna), inhibitor of apoptosis-1 (IAP-1/Birc-2), and growth arrest-specific-5 (Gas5); inflammatory factors and receptors growth-regulated oncogene 1 (Gro-1/KC), colony-stimulating factor-1 (csf1), complement 3 (C3), and lymphocyte antigen 6e (Ly6e); metabolism and drug resistance genes cytochrome P450 (Cyp1b1) and glutathione S-transferase (Gstm1); signal molecules protein tyrosine phosphatase (Ptpn13) and Yes-associated protein 65 kDa (Yap65); and chromatin-associated high mobility group protein [Hmga1 (4)]. Of these, differential expression in LY-2 cells and two other LY lines was also confirmed previously by Northern blot for Gas5, Gro-1/KC, Ly6, Gstm1, and Yap65 (4, 14). Several genes in Table 2 that were classified as decreased in LY-2 cells by ≥2-fold were also detected using both arrays. These include cyclin-dependent kinases 1 and 4 (Cdkn1 and Cdk4), calmodulin (Calm), cadherin 1 (Cdh 1), prothymosin b4 (Ptmb4), mucin (muc18), procollagens type I and IVα (Col1a2 and Col4a2); ATPase (Atp11a), and creatine kinase (Ckmt1). Decreased expression of Ptmb4 was confirmed previously by Northern blot (4). Two elements showed discordant expression between the two studies, tropomyosin α (Tpm1) and A disintegrin/MMP (Adamts-1). Overall, a consistent pattern of expression was observed for the majority of elements found to be differentially expressed by cDNA hybridization in these independent studies and for six of eight genes confirmed by Northern blot (4).

Changes in expression of several of the genes classified by microarray as differentially expressed in LY2 cells and modulated in LY2 IκBαM-expressing cells have been detected with oncogenesis and metastatic progression of SCC. Cyclin D1, IAP-1, and mutant Trp53 have been reported to be overexpressed in advanced SCC, and β-catenin expression has been reported to be decreased in advanced SCC, consistent with the expression profile for these genes in LY-2 cells (Tables 1 and 2). Cyclin D1, Trp53, and IAP-1 may be up-regulated with NF-κB in other cancer lines, and β-catenin may be down-regulated with NF-κB in other cancer lines (15, 16, 17, 18, 19), consistent with increased activation of NF-κB (5) in LY-2 SCC cells (Tables 1 and 2).

To validate the expression profile of cyclin D1, Trp53, IAP-1, and β-catenin detected in LY-2 cells and that obtained after modulation of NF-κB by IκBαM, expression was compared by Western and Northern blot analysis using protein and mRNA extracts from samples obtained after 72 hours of culture ± DOX, concurrent with array experiments shown in Fig. 1. Fig. 3 A shows Western blots comparing cyclin D1, Trp53, IAP-1, and β-catenin protein expression in the low NF-κB Pam 212, high NF-κB LY-2 cells and in three LY-2 IκBαM clones ± DOX. Compared with Pam 212, LY-2 cells and most of the LY-2 IκBαM clones without DOX showed increased cyclin D1, IAP-1, and Trp53 and decreased β-catenin protein expression levels, consistent with the profiles obtained by array. Conversely, with inhibition of NF-κB by IκBαM in the LY-2 clones + DOX, decreased expression of cyclin D1, Trp53, and IAP-1 and increased expression of β-catenin proteins are observed in most clones, consistent with modulation of the genes encoding these proteins. There were no differences due to differences in loading or NF-κB activation between cell lines when blots were probed for the constitutively expressed protein β-tubulin. Thus, the Western blot data are consistent with the role of NF-κB in the modulation of these genes in LY-2 cells detected by microarray.

To establish whether cyclin D1, Trp53, IAP-1, and β-catenin mRNA expression was modulated with NF-κB in LY2 IκBαM-expressing cells, Northern blot analysis was performed to compare mRNA expression by the LY2 IκBαM clone 24C cultured ± DOX. The Northern blots in Fig. 3,B show that expression of IκBαM in LY-2 24C cells + DOX resulted in decreased mRNA signal for cyclin D1, Trp53, and IAP-1 compared with LY-2 24C cells without DOX. In contrast, β-catenin mRNA expression was increased after expression of IκBαM in LY-2 24C cells + DOX. No nonspecific induction or degradation of mRNA or differences in loading were detected when blots were probed with β-actin. The differences in protein and mRNA expression for the LY2 IκBαM clone 24C cultured ± DOX are quantified and compared by densitometry in Fig. 3 C and D. The direction of change for the proteins and genes studied is consistent for all four genes. The difference in protein and mRNA expression was similar for cyclin D1 and β-catenin, whereas the percentage difference for protein exceeds that observed for mRNA at the 72-hour time interval for IAP-1 and p53. Together, the Western and Northern blot data validate the expression profile detected by array for these known cancer-related genes in LY-2 cells and confirm that modulation of NF-κB by IκBαM results in up- or down-modulation of several NF-κB–regulated genes.

Because expression of wild-type Trp53 has been reported to have proapoptotic rather than antiapoptotic effects, we performed a sequence analysis of Trp53 in LY-2. This established the presence of nonconserved mutations in exons 6, codon 81, CCG (Pro) to CTG (Leu), and exon 7, codon 121, ATA (Ile) to ATG (Met), in Trp53 in LY-2 SCC cells (data not shown). Thus, NF-κB appears to contribute to the overexpression of Trp53 in LY-2 SCC cells.

To examine whether these and the larger list of genes modulated by IκBαM as detected by microarray contain DNA sequence motifs that are putative targets for NF-κB regulation, we performed a search of the DNA that included 2 kb of the promoter region and the full-length sequence of each gene. As annotated in Tables 1 and 2, multiple genes were identified that contain sequences for NF-κB DNA binding motifs (∗) and/or the ACTACAG 7-nucleotide motifs (#), recently shown to mediate posttranscriptional down-regulation of the genes Myo D and Sox9 by NF-κB (13). Fully 105 of 167 (63%) of the genes classified as increased with tumor progression contained putative NF-κB promoter DNA binding sequences, approximating the percentage of ∼69% found to be down-modulated by IκBαM (see Table 1, Genes Increased with Tumor Progression). These include cyclin D1, Trp53, and IAP-1 confirmed by Western and Northern blot analysis. Of the genes decreased with tumor progression (Table 2), 47 of 141 (33%) also contained NF-κB DNA binding motifs, including β-catenin. The DNA binding motif associated with p65 RelA (12) expressed in LY-2 and other SCCs was present in 42% of the genes showing increased expression but in only 14% of the genes showing decreased expression. Overall, NF-κB (63%) and p65 (42%) DNA binding motifs were increased in LY-2 cells with tumor progression when compared with NF-κB motifs detected in 25% of 900 genes randomly selected from the array list, indicating that the increase in representation of these motifs in genes expressed is unlikely to have occurred by chance (χ², P < 0.001).

The motif ACTACAG shown to be involved in posttranscriptional down-regulation of the genes MyoD and Sox9 by NF-κB (13) was detected in the promoter, exon, or intronic DNA sequence of over 50% of genes exhibiting decreased expression with tumor progression, but these motifs were present at a similar frequency in the DNA of genes showing increased expression and in 10⁴ genes selected randomly and analyzed by the National Center for Bioinformatics (χ², P > 0.05). The 7-mer motif was present in the coding portion of 30 of 167 (17%) and 18 of 141 (13%) genes showing increased or decreased expression, respectively. When only those genes that contain the 7-mer within the coding sequence and in-frame as found in MyoD and Sox9 (13) are included, we detected this motif in 14 of 167 genes increased with tumor progression and 11 of 141 genes decreased with tumor progression, a frequency of ∼8%.

Fig. 4 shows the alignment of the ACTACAG motif and adjacent DNA sequences for genes that showed conserved nucleotides with those of MyoD and Sox9 and/or within the same region of their human homologues (∗). Cytosines (C) located 5 and/or 8 nucleotides 5′ and after the 7-mer and C or G nucleotides (S) located 10, 13, 14, and 16 nucleotides 3′ to the motif were most common. Although these homologies suggest a possible relationship to the mRNA regulatory motifs found in MyoD and Sox9, all but Atp11a lacked multiple repeats, and Fn1 contained a motif not found in its human homologue. Thus, whereas the NF-κB DNA binding sequences involved in transcriptional regulation are significantly increased in genes differentially expressed in SCC, the significance and functional role of ACTACAG motifs found in the genes detected remain to be determined.

The putative functional role of the validated genes and other genes in Tables 1 and 2 in cell proliferation and death, inflammation/angiogenesis, migration/metastasis, metabolism, stress, signaling, gene and protein expression, and structure is consistent with the complex phenotypic differences between keratinocytes and highly metastatic SCC cells. To establish whether the modulation by NF-κB is associated with effects on the malignant phenotype, we examined the effect of modulation of IκBαM on several features of malignancy in LY-2 cells. DOX-induced expression of IκBαM had minor effects on proliferation beginning at day 3 as measured by MTT densitometric assay (Fig. 5,A), corresponding to the interval when DOX-induced inhibition of NF-κB activation was detected (Fig. 1,A). No changes in cell morphology, trypan blue uptake, or detachment were evident on microscopy (data not shown). To determine whether the decreased density observed by day 5 was related to cell death, cell cultures were examined by microscopy and by DNA cytofluorometry for cell cycle changes and subcellular DNA fragmentation that are produced on cell death. Cells expressing IκBαM began to show blebbing, fragmentation and detachment (data not shown). This was accompanied by the loss of DNA from G, S, and M phases and shift of ∼50% of the DNA to the sub–G₀-G₁ fraction, consistent with cell death and fragmentation (Fig. 5 B).

Because a number of genes potentially involved in cell adhesion, migration, and invasion were identified by array, we compared the migration and invasion of Pam 212 and LY-2 cells and examined the effect of inhibition of NF-κB on migration and invasion through Matrigel matrix in a two-chamber assay between 24 and 72 hours after addition of DOX, before significant cell loss or death was detected. Fig. 6 A shows that control LY-2 cells ± DOX exhibit a similar increase in invasion of matrix relative to Pam 212. However, the invasiveness of four of four LY-2 clones conditionally expressing IκBαM was reduced to levels at or below those observed for Pam 212, suggesting that the increase in NF-κB observed in LY-2 cells plays a role in the increased migration and invasiveness of these metastatic cells.

Because NF-κB regulates Gro-1, a factor that promotes tumorigenesis and host angiogenesis (5, 14), the effect of inhibition of NF-κB by IκBαM on tumorigenesis and angiogenesis was examined. Fig. 6,B shows the tumor growth of control LY-2 cells and the LY-2 clone 24C before and after addition of DOX to the drinking water of congenic BALB/c SCID mice. On addition of DOX (arrow), the parental line continued to grow, whereas LY-2 24C tumor rapidly regressed with inhibition of NF-κB by IκBαM. No regrowth was detected in mice undergoing complete regression, but four mice retained small palpable nodules. To determine whether the inhibition was associated with decreased angiogenesis, the density of CD31-expressing blood vessels in the control and four mice bearing residual 24C tumors was compared. Fig. 6 C illustrates the difference in tumor size in H&E-stained specimens and the associated difference in density of CD31 staining vessels. The microvessel density was significantly reduced in 24C tumors compared with control LY-2 tumors treated with DOX (53 ± 15 versus 175 ± 23, mean number of vessels per high-powered field ± SD; n = 4; P < 0.05). The clone transfected with IκBαM grew more slowly than the parental control, even in the absence of DOX. This slower tumor formation was associated with a ∼10% to 30% lower expression of NF-κB-modulated angiogenesis factors Gro-1 and vascular endothelial growth factor in 24C and other clones compared with parental LY-2 cells, suggesting a degree of “leakiness” and weak suppression of NF-κB-inducible genes by the Tet-inducible promoter, even in the absence of DOX (data not shown).

In this study, we obtained evidence for differential expression of a diversity of known NF-κB–related genes and provide evidence that modulation of the NF-κB pathway directly or indirectly modulates a significant portion of these and other genes in the molecular profile altered with malignant progression in a syngeneic model of SCC. The array profile obtained with the expanded 15,000-element microarray reproduced differences in expression of a number of NF-κB–related genes detected and validated by Northern blot in our previous study, as well as other genes reported to be differentially expressed and important in oncogenesis of SCC. We confirmed that IκBαM expression suppressed NF-κB reporter and DNA binding activity and modulated protein and mRNA expression of several putative NF-κB target genes differentially expressed in SCC, including cyclin D1, Trp53, IAP-1, and β-catenin (15, 16, 17, 18, 19). Bioinformatic analysis revealed that the sequence of these and many of the other putative NF-κB–modulated genes contain promoter or coding sequences homologous to the NF-κB DNA binding motif or an ACTACAG 7-nucleotide motif recently shown to mediate posttranscriptional down-regulation by NF-κB of MyoD and Sox9 mRNA expression (13). Consistent with the diversity and putative function of many of these genes, inhibition of NF-κB was found to inhibit proliferation, cell survival, migration, angiogenesis, and tumorigenesis. These results provide evidence that NF-κB is an important modulator of the gene expression profile and malignant phenotype in SCC and provide an expanded list of candidate genes potentially modulated directly or indirectly by NF-κB to be investigated in future studies.

Homologues of several of the genes detected with malignant progression and modulated by NF-κB in this murine SCC model have been found to be differentially expressed and important in oncogenesis in human SCCs. Cyclin D1 is overexpressed in the majority of human head and neck squamous cell carcinoma (HNSCCs) and promotes proliferation and tumorigenesis (20). Cyclin D1 overexpression results from gene amplification in only a minority of SCCs, and the results of this study provide evidence that NF-κB activation can contribute to the overexpression of cyclin D1 in SCCs. The tumor suppressor Trp53 is mutated and overexpressed in approximately half of human HNSCCs and plays an important role in cell cycle arrest and DNA repair or apoptosis (21). The overexpression of Trp53 in LY-2 and inhibition after inactivation of NF-κB suggest that NF-κB may contribute to overexpression of Trp53. Increased expression of IAP-1 (Birc2) was previously detected in metastatic SCC (4). The inhibition of IAP-1 expression with inhibition of NF-κB in this study is consistent with the effect of inhibition of NF-κB in sensitizing LY-2 cells to apoptosis and HNSCC to tumor necrosis factor-induced caspase-mediated cell death (22). Decreased expression of β-catenin has been observed and associated with metastasis of human oral SCC (23). The low expression of β-catenin in metastatic LY-2 cells relative to keratinocytes and restoration of increased mRNA and protein levels after expression of IκBαM suggest that repression of β-catenin expression can result from NF-κB activation in SCC. The effects of modulation of NF-κB on expression of cyclin D1, Trp53, IAP-1, and β-catenin observed in this study are consistent with those observed with modulation of NF-κB in other malignancies (15, 16, 17, 18, 19).

When the expression of the four genes was compared by microarray and Northern and Western blots, the direction of change for the proteins and genes studied was consistent for all four genes (Fig. 3 C and D). The percentage change was similar for protein and mRNA for cyclin D1 and β-catenin, whereas the percentage difference for protein was greater than that observed for mRNA at the 72-hour time interval for IAP-1 and p53. These differences between protein and mRNA could be due to a variety of mechanisms for individual genes. These include differences in kinetics related to other IκBs (24), kinetics of modulation of RNA and protein, differences in stability and posttranscriptional processing of RNA, or differences in efficiency of translation. It will be important to determine whether differences among genes are due to different regulatory mechanisms in future studies.

Bioinformatic analyses of the four validated genes for NF-κB–related motifs reveal that the sequences of the cyclin D1, Trp53, IAP-1, and β-catenin promoters contain NF-κB DNA binding motifs. The sequence of β-catenin also contains an ACTACAG 7-nucleotide motif in the noncoding portion of the gene. The presence of this motif in the coding region has recently been shown to mediate posttranscriptional down-regulation by NF-κB of MyoD and Sox9 mRNA expression (13), but the role of the motif located in the noncoding portion of β-catenin remains to be determined. Although we have obtained evidence that NF-κB modulates these and a wider profile of genes containing these motifs in Tables 1 and 2, it will be necessary to determine individually for each gene by mutagenesis whether gene expression is regulated directly by NF-κB or indirectly by other transcription factors or regulatory molecules also modulated during the 72-hour interval required for significant DOX-induced expression of IκBαM and inhibition of NF-κB.

Gro-1, one of the genes consistently detected in LY-2 and aggressive SCC in this and previous studies, was also previously shown to be modulated by NF-κB and to contribute to angiogenesis, tumorigenesis, and metastasis (5, 14). We showed that enforced expression of Gro-1 in low-Gro-1–expressing Pam 212 cells converted these cells to the LY phenotype and promoted increased angiogenesis, tumorigenesis, and metastasis, indicating that this NF-κB–regulated gene contributes directly to the malignant phenotype of SCC in this model (14). Conversely, we show in this study that inhibition of NF-κB and Gro-1 expression in LY-2 cells overexpressing Gro-1 is associated with decreased angiogenesis and tumorigenesis. Human cytokine GRO1 and its homologue, interleukin-8, have been found to be overexpressed in the majority of HNSCCs and many other cancers (3, 25). Consistent with the effects of inhibition of NF-κB by IκBα on gene expression and phenotype observed in this study, constitutive expression of an IκBα mutant in human HNSCC inhibited expression of interleukin-8 and tumorigenesis of xenografts in mice (7). Although the results suggest an important functional role of NF-κB in angiogenesis and tumorigenesis, the decrease in angiogenesis observed could be due in part to apoptosis and reduction in tumor mass producing angiogenesis factors, as well as decreased transcription.

Our previous studies and the present molecular profiling study of malignant progression reveal differences in expression of a variety of other putative NF-κB–modulated genes that are consistent with expression patterns observed more broadly in SCC. Glutathione S-transferase was increased in LY-2 and inhibited with suppression of NF-κB activation. Overexpression of glutathione S-transferases involved in metal metabolism has been detected and shown to be important in resistance of SCC to cis-platinum and other chemotherapy agents (26). Expression of the cell adhesion molecules integrin α₃β₁ and decreased expression of cadherin 1 and β-catenin protein are often observed and associated with poor prognosis in SCC and other epithelial cancers (2, 23). Decreased expression of the cell-cell receptor Ly-6 was also confirmed previously in LY SCC cells (27). Finally, the detection of several classes of molecules previously detected as tumor specific and tumor-associated antigens in SCC such as Ly-6, integrins, DEAD-box helicases, and ribosomal components (2, 27, 28, 29) suggests that NF-κB may contribute to expression of antigenic molecules detected by antibodies and T cells. Identification of NF-κB–regulated genes could help determine the identity of antigens for immune therapy as well as their function in the malignant phenotype. The differences in expression of these candidates implicated in regulation of resistance to chemotherapy-induced apoptosis, adhesion, and immune antigenicity in this model are therefore consistent with broader expression in SCC and make the role of NF-κB in their regulation of broader potential relevance.

We noted in the previous study with a first-generation 4,000-element murine array that a significant portion of the genes overexpressed in multiple metastatic reisolates were reported to be related to the NF-κB pathway. Consistent with this, a search of PubMed revealed that >40% of the up-regulated genes and ∼28% of the down-regulated genes detected in LY-2 with the 15,000-element array in the present study were related previously to NF-κB. The detection of both known and additional up- and down-regulated genes with modulation of NF-κB prompted us to undertake an analysis to determine whether the genes identified contained common regulatory motifs related to NF-κB. Bioinformatic analysis of the sequence of genes detected with tumor progression of LY-2 in this study revealed that 63% of the genes increased in LY-2 cells contained NF-κB binding motifs in the promoter region. Of the candidates showing lower expression, 33% contained NF-κB binding motifs, and ∼8% contained one or more motifs found in MyoD and Sox9 mRNA implicated in NF-κB–mediated posttranscriptional down-regulation (13). The detection of sequences homologous to the 7-nucleotide motif detected in MyoD and Sox9 mRNA suggests that this NF-κB–regulated mechanism may be involved in down-regulation of other genes besides MyoD. The restoration of expression of these genes back up to the levels observed in keratinocytes after expression of IκBαM in LY-2 provides support for this hypothesis. The mechanism underlying NF-κB modulation of MyoD and Sox9 mRNA and relevance to expression of other genes in cancer remain to be established.

Cumulative evidence in head and neck, lymphoid, breast, gastric, colorectal, and prostate cancers is consistent with the hypothesis that NF-κB is constitutively activated and a major “culprit” in the pathogenesis of cellular and host alterations in these cancers (reviewed in refs. 25 and 30). NF-κB has been implicated in the oncogenesis of a number of other malignancies, indicating that NF-κB activation may be a common pathway of broad importance in cancer (30). Viral and cellular members of the NF-κB family have been shown to be transforming (31, 32), and oncogenic activation of RAS, ABL, LMP-1, interleukin-1, and epidermal growth factor receptor has been shown to contribute to activation of NF-κB in different cancers (10, 33, 34, 35, 36). Constitutive activation of NF-κB and NF-κB–inducible genes has been detected in prostate and breast cancers, melanomas, and lymphomas (25, 37, 38, 39, 40, 41). NF-κB has been implicated in promoting expression of the phenotypic changes and genes involved in inhibition of programmed and therapeutic cell death (18, 22, 41, 42, 43, 44), proliferation (15), tumorigenesis (7, 25, 34), angiogenesis, invasiveness, and metastatic potential (9, 25, 38).

Furthermore, molecular profiling has revealed that NF-κB may regulate a diverse repertoire of genes in SCC (this study) and in Hodgkin’s and non-Hodgkin’s lymphomas (39, 40, 41). We have provided evidence that NF-κB modulates many of the genes and phenotypic features differentially regulated during malignant progression in this stepwise model of SCC. The broader significance of these results is independently supported by our earlier results showing that NF-κB is activated in multiple murine and human SCCs (2, 5, 6) and that constitutive inhibition of NF-κB in human SCC inhibits cell survival, angiogenesis factor expression, and growth of xenografts (7, 9). The genes detected represent programs that are functionally involved in many of the phenotypic features that constitute the hallmarks of cancer (1). These results suggest that NF-κB is a key molecular switch of the alterations in genotype and phenotype in malignant progression of SCC. Increasing appreciation of the role of NF-κB in regulation of expression of a diversity of genes in cancer is consistent with the evolutionarily importance of this transcription factor in regulating programmed or stereotyped gene responses to injury and pathogens (25, 30).