The Host Environment Promotes the Constitutive Activation of Nuclear Factor-κB and Proinflammatory Cytokine Expression during Metastatic Tumor Progression of Murine Squamous Cell Carcinoma¹

We and others have previously detected elevated concentrations of proinflammatory cytokines IL³-1α, IL-6, IL-8, and GM-CSF in cell line supernatants (1, 2, 3, 4, 5, 6), tumor specimens (2, 4, 7), and serum (7, 8, 9, 10) of patients with SCC arising from different sites. Squamous carcinoma cells are an important source of proinflammatory cytokines in situ, and the expression of these cytokines has been associated with increased potential for growth and metastasis of SCC and other types of cancer. IL-1, IL-6, and IL-8 have been linked with increased growth (11, 12, 13, 14, 15, 16), and production of IL-6, IL-8, and GM-CSF has been associated with increased spread and metastasis (5, 8, 9, 15, 16, 17). These observations suggest that constitutive activation of cytokine genes may play a role in tumor progression. However, the mechanism(s) involved in activation of cytokine gene expression during tumor progression are poorly defined.

We previously established cell lines from lymph node (LY) and lung (LU) metastases (16, 17) of the murine SCC line Pam 212 (18) and observed that these metastatic cell lines exhibit an increase in growth and metastatic potential in association with a increase in constitutive expression of proinflammatory cytokines IL-1α, IL-6, GM-CSF, and KC (the murine homologue of Gro-α, which is a member of the C-X-C chemokine superfamily that includes IL-8; Ref. 16). The constitutive expression of this repertoire of proinflammatory cytokines in the metastatic variants led us to hypothesize that gene expression of these cytokines may occur as a result of activation of a common transcriptional regulatory mechanism.

During the early response to injury, transient expression of proinflammatory cytokines can be induced by cytokines such as TNF-α through activation of transcription factor NF-κB. NF-κB binding sites identified in the promoter regions of many cytokine and immunoregulatory genes have been shown to serve as major control elements for cytokine expression (19, 20). NF-κB is a heterodimeric transcription factor composed of members of the Rel family of proteins, including p65 (RelA), p50 (NF-κB1), p52 (NF-κB2), RelB, and c-Rel (21). In many types of cells, NF-κB/RelA is retained in the cytoplasm in an inactive form due to binding by IκBs proteins. Upon activation by TNF-α or other signals, these inhibitor proteins are phosphorylated and undergo degradation by an ubiquitin-dependent pathway, releasing NF-κB for nuclear localization and activation of target genes (21). Activation of NF-κB target genes, including cytokines, has been implicated in the promotion of transformation and survival of tumor cells (22, 23, 24, 25).

In the present study, we compared the constitutive and TNF-α-induced expression of proinflammatory cytokines in parental Pam 212 and metastatic LY-2 and LY-8 cell lines and determined the relationship of cytokine expression to activation of NF-κB. We report that metastatic cell lines selected in the host environment exhibit an increase in constitutive and TNF-α-inducible expression of proinflammatory cytokines when compared with parental Pam 212 cells. The increased cytokine expression was associated with an increase in constitutive and TNF-inducible activation of NF-κB. Constitutive nuclear localization of NF-κB p65 was observed in LY-2 and LY-8 cells in culture and in vivo. TNF-α inducible but not constitutive activation of NF-κB was inhibited by the addition of proteasome inhibitor or overexpression of hIκBαM in Pam 212 cells. The data indicate that activation of NF-κB is an important molecular controlling mechanism for proinflammatory cytokine expression during metastatic tumor progression of SCC.

The origin and characterization of the parental Pam 212 and metastatic LY-2 and LY-8 cells has been described previously (16, 17, 18). The Pam 212 cell line is a spontaneously transformed cell line derived from neonatal BALB/c keratinocytes in vitro, which forms SCCs in vivo (18), and was provided by Dr. Stuart Yuspa of the National Cancer Institute. The metastatic LY cell lines were isolated from lymph node metastases that formed after s.c. inoculation of Pam 212 tumor fragments in BALB/c mice (16, 17). The three cell lines used in the present study retained cytokeratin markers. The Pam 212, LY-2, and LY-8 cells were grown in EMEM plus 10% FCS and penicillin, streptomycin, and glutamine. The cell lines were tested and found to be free of Mycoplasma.

Antibodies to p65 and p50 were purchased from Rockland (Gilbertsville, PA) and Santa Cruz (Santa Cruz, CA), respectively. Polyclonal rabbit anti-mouse cytokeratin K6 antibody was a kind gift of Dr. Stuart Yuspa (National Cancer Institute, NIH, Bethesda, MD). Plasmid IgκB-Luc containing two copies of the NF-κB binding site upstream of the minimal promoter fused with the luciferase gene was described previously (26). pCMVLacZ was made by Dr. Giovana Thomas in our laboratory and consists of a LacZ gene inserted between the CMV promoter and BGH poly(A) signal sequence in pcDNA3 (Invitrogen, Carlsbad, CA). pCMVKC was made by ligation of a 300-bp cDNA containing the entire coding sequence of KC into pcDNA3. The 300-bp KC cDNA was generated by 5′ rapid amplification of cDNA ends of mRNA from LY-1, and the entire sequence was confirmed with published KC sequence (Ref. 27; data not shown). pMT2-IκBαM was described by Brown et al. (28) and includes a cDNA for hIκBαM cloned downstream of the adenoviral major late promoter. In hIκBαM, substitution mutations were introduced to replace coding sequences of serines 32 and 36 in human IκBα. pMT2 expression vector and pSV40neo were described previously (29). CPI, TPCK, and PDTC were purchased from Calbiochem (La Jolla, CA) and prepared according to manufacturer’s suggestions; aliquots were stored in −20°C.

Log phase grown Pam 212, LY-2, and LY-8 were transfected as described in cell transfection section with either pMT2 or pMT2-hIκBαM along with a 10-fold less amount of pSV40neo. Forty-eight h after transfection, cells were changed into media containing 400 μg/ml of G418 (Life Technologies, Inc., Gaithersburg, MD). Ten to 14 days after selection in G418, surviving cells were cloned by limiting dilution. Individual clones were expanded and screened for the expression of hIκBαM by Western blotting with antibody to human IκBα as described below.

ELISA kits for murine IL-1α, IL-6, GM-CSF, and TNF-α were purchased from Endogen (Cambridge, MA), and murine KC was purchased from R&D systems (Minneapolis, MN); the ELISA assay was carried out according to the manufacturer’s protocol. Cells were cultured in a T-25 flask starting with 5 × 10⁵ cells/flask and incubated overnight. The next day, culture medium was changed to medium with or without TNF-α (10,000 units/ml) for an additional 48 h. The supernatants were collected and centrifuged at 4000 rpm for 10 min to remove the cell debris. Each sample was tested in duplicate in each of two or more replicate experiments. After development of the colorimetric reaction, the absorbance at 450 nm was quantitated by an 8-channel microplate autoreader (Biotek Systems, Winooski, VT), and the absorbance readings were converted to pg/ml based upon standard curves obtained with recombinant cytokine in each assay. If the absorbance readings exceeded the linear range of the standard curves, the ELISA assay was repeated after serial dilution of the supernatants. The sensitivity and the linear range of each cytokine tested were : IL-1α, <6 pg/ml (15.6–1000 pg/ml); IL-6, <15 pg/ml (50–1250 pg/ml); KC, <2 pg/ml (15.6–1000 pg/ml); GM-CSF, <5 pg/ml (10–250 pg/ml); and TNF-α, <10 pg/ml (50–2450 pg/ml).

Mini-scale nuclear extracts were made according to the method of Beg et al. (30). To make whole-cell extracts, we followed a procedure from the laboratory of Dr. U. Siebenlist (National Institute of Allergy and Infectious Diseases, NIH) with minor modifications. Briefly, 1 × 10⁶ cells were rinsed with PBS and harvested from tissue culture flasks by gentle scrapping. After spinning down the cells and removing PBS, an equivalent volume of lysis buffer [50 mm Tris (pH 7.4), 100 mm NaCl, 50 mm NaF, 30 mm sodium PP_i, and 0.5% NP40] was added to resuspend the cell pellet. A protease inhibitor cocktail tablet (Complete, Mini; Boehringer Mannheim, Mannheim, Germany) was added per 10 ml of lysis buffer before use. Samples were then spun at 18,000 × g for 20 min at 4°C. Supernatants were aliquoted, snap frozen, and stored at −80°C. Protein concentrations were determined using a BCA protein assay kit following the manufacturer’s instructions (Pierce, Rockford, Illinois).

EMSA was performed as described previously (31) with minor modifications. Briefly, 5–10 μg of whole-cell extracts were incubated with 1 μg of poly(deoxyinosinic-deoxycytidylic acid) (Pharmacia Biotech, Piscataway, NJ) alone or with unlabeled wild-type or mutant DNA or antibodies in 20 μl of buffered binding mixture [20 mm HEPES (pH 7.9), 5 mm MgCl, 60 mm KCl, 1 mm DTT, 0.1% NP40, and 10% glycerol] for 10 min at room temperature. ³²P-labeled probe (20,000 cpm) was then added, and the reaction mixture was incubated for another 30 min at room temperature. Each reaction mixture was loaded immediately onto a 5% nondenaturing polyacrylamide gel made in 0.25× TBE [0.22 m Tris-Borate, 0.0005 m EDTA]. Gels were run at 200V for 2 h. After being dried, gels were either scanned with InstantImager (Packard, Meriden, CT) or subject to autoradiography.

Immunostaining of NF-κB p65 was performed following standard immunohistochemical methods (32). For staining of Pam 212, LY-2, and LY-8 cell lines, 5 × 10³ cells/well were grown on 8-well chamber slides (Lab-Tek, Naperville, IL) in complete media or media supplemented with TNF-α for 60 min at 37°C prior to staining. For staining of tumor specimens, tissues were preserved by freezing in OCT and sectioned in 6–8-μm sections. The cells or tissues were fixed with 3.7% formaldehyde in PBS at room for 5 min and permeabilized by 0.2% Triton X-100. After washing and treatment with 3% H₂O₂, samples were incubated sequentially in 10% goat serum (Vector Laboratory, Inc., Burlingame, CA) as the blocking reagent for 30 min, polyclonal rabbit anti-NF-κB antibody (1:2000–4000) or isotype control antibody (rabbit IgG; Cappel, West Chester, PA) for 60 min, and goat anti-rabbit antibody (Vector Laboratory) at 1:200 dilution in PBS containing 1.5% goat serum for 30 min. As a positive control, rabbit anti-mouse cytokeratin K6 serum (kindly provided by Dr. Yuspa, National Cancer Institute, NIH, Bethesda, MD) was used at 1:2000 dilution. The samples were incubated with biotin/avidin horseradish peroxidase conjugates and chromogen 3,3′-diaminobenzidine (Vectastain Elite ABC kit; Vector Laboratory) according to the manufacturer’s instructions. After counterstaining with hematoxylin (Gill’s formula; Vector Laboratory), the slides were mounted with Permount (Fisher Scientific, Pittsburgh, PA) or glycerol (Sigma Chemical Co., St. Louis, MO), and photomicrographs were obtained immediately at a magnification of ×400.

Pam 212, LY-2, and LY-8 cells (2 × 10⁵) each were grown in six-well plates in triplicate the day before transfection. For transfection, cells were incubated with 2 μg of IgκB-Luc and 0.1 μg of pCMVLacZ DNA mixed with 16 μl of LipofectAMINE in OPTI-MEM medium (Life Technologies) for 5 h at 37°C. The transfection media was then replaced with EMEM plus 10% serum, and incubation was continued for 48 h. The cells were harvested, reporter gene activities were assayed using the Dual-Light Luciferase and β-Galactosidase Reporter Gene Assay System (Tropix, Bedford, MA), and chemoluminescence was measured by a Mono- light 2010 luminometer (Analytical Luminescence Lab, San Diego, CA). The relative light units were calculated as follows:

For Northern analysis, total RNA was isolated from cultured tumor cells when 80–90% confluent in 25-cm² tissue culture flasks using Trizol reagent according to the manufacturer’s instructions (Life Technologies). RNA concentration, purity, and integrity were determined by a spectrophotometer at 260 and 280 nm and by RNA agarose gel electrophoresis. Twenty μg of total RNA from each cell line were resolved by a 1.2% formaldehyde agarose gel electrophoresis, and Northern blot was performed as described by Sambrook et al. (33). RNA was transferred overnight to uncharged Hybond-N 0.45-μm nylon membrane (Amersham Life Science, Arlington Heights, IL) and UV cross-linked onto the membrane (UV Stratalinker 2400, 120 mJ; Stratagene, La Jolla, CA). cDNA containing the murine KC coding sequence was excised from pCMVKC by restriction enzyme digestion and purified with Gene Clean II kit (Bio 101, Vistas, CA). Twenty-five to 50 ng of each purified DNA fragment were added to a random priming reaction using Prime-It RmT kit (Stratagene), and the labeled DNA probes were purified using a G-50 column (5′→3′ Prime, Inc., Boulder, Co). Labeled probe (10⁶ cpm/ml) was used for hybridization in QuikHyb (Stratagene). After hybridization and washing under high stringency conditions, the membrane was subjected to autoradiography.

For Western blot analysis, 50 μg of cell extracts prepared as above were resolved on 11 × 14-cm denaturing SDS-PAGE. After gel transfer, nitrocellulose membrane (Bio-Rad, Hercules, CA) was incubated sequentially with 5% dry milk in PBS for 1 h, polyclonal rabbit antibody to IκBα (Santa Cruz Biotechnology, Santa Cruz, CA) for 1 h, and goat anti-rabbit IgG-HRP conjugate (Bio-Rad) for 1 h. Membranes were developed in chemiluminescent substrate (Pierce, Rockford, IL) and exposed to X-OMAT film.

Our laboratory reported previously that metastatic SCC cell lines LY-2 and LY-8 produced higher concentrations of proinflammatory cytokines IL-1α, IL-6, KC, and GM-CSF than the parental Pam 212 line (16). In the present study, we examined whether these cytokines could be induced by TNF-α, a known stimulus of proinflammatory cytokine expression. Conditioned media were harvested from untreated cells or cells treated with human and murine TNF-α for 48 h and assayed by ELISA. In preliminary experiments, human and murine TNF-α were titrated by their ability to induce KC production in Pam 212, LY-2, and LY-8 cells. The maximal induction was found at 10,000 units/ml with either human or murine TNF-α (data not shown). Fig. 1 shows maximal stimulation of cytokine production by 10,000 units/ml of human TNF-α. As observed previously, unstimulated LY-2 and LY-8 cell lines produced several of the cytokines at higher concentrations than cultures of the parental Pam 212 cell line. Untreated LY-2 produced IL-1α (Fig. 1,A), KC (Fig. 1,C), and GM-CSF (Fig. 1,D), and LY-8 produced IL-6 (Fig. 1,B) and KC (Fig. 1 C). Upon the addition of TNF-α, a further increase in the concentration of cytokines was observed in LY-2 cells (IL-1α, KC, and GM-CSF) and LY-8 cells (IL-1α, IL-6, KC, and GM-CSF). Treatment of Pam 212 cells with hTNF-α induced KC only. The Pam and metastatic variants were resistant to cytotoxicity or growth inhibition by concentrations of up to 10,000 units/ml of hTNF-α in 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay.⁴

Because NF-κB transcription factor binding sites are present in the promoter region of all four proinflammatory cytokine genes (19), we compared the NF-κB binding activity in the cell lines by EMSA. Fig. 2 A shows the presence of a major and a minor complex, which bind NF-κB sequence in LY-2 (Lane 5) and LY-8 (Lane 8) cell extracts, whereas only a weak signal is observed in Pam 212 cells (Lane 2). The difference in signal represents a 2–5-fold increase in NF-κB binding activity in the extracts from LY-2 and LY-8 when compared with Pam 212 (data not shown). The specificity of NF-κB binding activities was confirmed by competitive inhibition with a 100-fold excess of unlabeled wild-type (Lanes 3, 6, and 9), but not mutant, oligonucleotides (Lanes 4, 7, and 10).

The composition of the NF-κB complexes was determined by super-shift with antibodies specific to p65 and p50 NF-κB subunits with the extracts from LY-2 (Fig. 2,A, Lanes 11 and 12) and LY-8 (Lanes 13 and 14). Preincubating the cell extracts with antibody to p65 subunit shifted the major NF-κB complex (Lanes 11 and 13). Antibody to p50 subunit shifted the major and minor NF-κB complexes (Fig. 2,A, Lanes 12 and 14). Using a recombinant p50 as a control in an independent EMSA experiment, we confirmed that the minor NF-κB complex comigrated with the complex formed by this purified p50 protein (data not shown). Neither antibody shifted the migration of another more rapidly migrating complex, which was determined to be nonspecific, based upon the lack of reactivity with NF-κB/Rel-specific antibodies, inconsistent appearance among different cell extracts, and nonspecific competition by mutant oligonucleotides in independent experiments (data not shown). We conclude the NF-κB complexes identified in these SCC cell lines include p65 and p50 Rel proteins. Oct-1 binding activities were also tested in these cell lines by EMSA (Fig. 2,B). No difference in Oct-1 binding activity was observed among the three cell lines, indicating that the constitutive NF-κB binding activity detected in metastatic SCC lines LY-2 and LY-8 was not the result of a general transactivation or unequal loading. These results demonstrate that constitutive NF-κB binding activity in metastatic SCC lines LY-2 and LY-8 is increased relative to the parental cell line Pam 212, consistent with the differences in constitutive production of proinflammatory cytokines exhibited by the metastatic cell lines in Fig. 1.

To establish whether differences in constitutive NF-κB binding activities observed in Pam 212, LY-2, and LY-8 represent differences in the functional activation of NF-κB, we performed a transient transfection assay using an NF-κB luciferase reporter gene in Pam 212, LY-2, and LY-8 cells as shown in Fig. 3. A 2–5-fold increase in reporter gene activity was detected in LY-2 and LY-8 cells relative to the Pam 212 cell line after normalization to β-galactosidase reporter activity. These results were consistent with the differences in NF-κB binding activities observed in Pam 212, LY-2, and LY-8 by EMSA in Fig. 2. Because TNF-α was shown to induce cytokine production in all three SCC lines by ELISA, we also tested the effect of TNF-α in reporter gene assay. As shown in Fig. 3, TNF-α induced a significant increase of luciferase activity in all cell lines tested. These results confirm that constitutive and TNF-α-induced production of cytokines correlates with functional activation of NF-κB.

We examined the distribution of NF-κB p65 localization in either unstimulated or TNF-α-treated Pam 212, LY-2, and LY-8 cells (Fig. 4,A) by immunostaining. In untreated Pam 212 cells, the majority of cells stained positive for p65 in cytoplasm (Fig. 4A, a). Sixty min after TNF-α treatment, anti-p65 staining was localized in the nuclei (Fig. 4,A, b). In contrast, in unstimulated LY-2 and LY-8 cells, constitutive nuclear localization of p65 was observed (Fig. 4,A, d and g), along with cytoplasmic staining. The nuclear staining of NF-κB p65 is more abundant in LY-8 cells (Fig. 4,A, g) than in LY-2 cells (Fig. 4,A, d), consistent with the results of the luciferase reporter gene assay (Fig. 3). TNF-α treatment of LY-2 and LY-8 only slightly enhanced nuclear staining in these cells (Fig. 4 A, e and h). Similar staining patterns of NF-κB in these cells were also confirmed using immunofluorescence (data not shown). The immunostaining results demonstrated that, without stimulation, nuclear NF-κB p65 was more abundant in LY-2 and LY-8 than in Pam 212 cells and confirmed that a constitutively active form of NF-κB detected by EMSA and the reporter gene assay was localized within the nuclear compartment.

To confirm that constitutive activation of NF-κB observed is not limited to cells in culture, we performed immunohistochemical analysis on tumor specimens of Pam 212, LY-2, and LY-8. Fig. 4,B shows immunostaining of Pam 212, LY-2, and LY-8 tumor sections by antibodies to NF-κB p65, keratin 6, and isotype control. All three tumor specimens showed cytoplasmic and nuclear staining for p65 (Fig. 4,B, a, d, and g). In Pam 212 tumors (Fig. 4,B, a), p65-specific staining was weaker than LY-2 and LY-8 tumors (Fig. 4,B, d and g), whereas normal skin cells were mostly negative (data not shown). The expression of squamous epithelial markers by the NF-κB-positive cells was confirmed by anti-keratin 6 staining in the adjacent sections (Fig. 4 B, b, e, and h). Thus the constitutive NF-κB observed in SCC tumor cells in vitro was also observed in vivo and is associated with tumor progression.

Activation of NF-κB involves phosphorylation and ubiquitination of IκB, followed by their degradation through proteasome activity (21). Protease inhibitors such as CPI and TPCK have been used as efficient blocking agents of NF-κB activation (21, 34, 35). The antioxidant PDTC has also been used to block NF-κB activation and subsequent induction of cytokine production, suggesting that oxidation plays a major role in regulating the activities of NF-κB proteins (36). To explore the mechanisms of the constitutive NF-κB activities observed in Pam 212, LY-2, and LY-8, cells were preincubated with CPI, TPCK, or PDTC before the addition of TNF-α, and cell extracts were assayed by EMSA. The addition of TNF-α led to an increase in the intensity of NF-κB complex in Pam 212 (Fig. 5,A, Lane 2). Pretreatment of Pam 212 cells with each of these three inhibitors for 15 min prior to TNF-α treatment completely blocked the TNF-α-inducible activation of NF-κB (Fig. 5,A, Lanes 3–5). The inhibition was not observed for Oct-1 binding activities (Fig. 5,B), suggesting that the diminished NF-κB induction by TNF-α was specific and not due to the cytotoxicity by the inhibitors used. In LY-2 and LY-8 cells, constitutive NF-κB binding activity was enhanced further by TNF-α (Fig. 5,A, Lanes 10 and 18). The use of inhibitors also blocked NF-κB induced by TNF-α in these two cell lines (Fig. 5,A, Lanes 11–13 and 19–21). In contrast, constitutive NF-κB binding activities present in LY-2 and LY-8 lines were not affected by CPI (Fig. 5,A, Lanes 11 and 19) or PDTC (Fig. 5,A, Lanes 13 and 21) but only partially reduced by TPCK (Fig. 5 A, Lanes 12 and 20) under the experimental conditions used. These results provide further evidence that all three SCC lines tested retained the ability to respond to TNF-α activation. The induction of NF-κB binding by TNF-α was efficiently blocked by using protease inhibitors, consistent with the hypothesis that phosphorylation and degradation of IκB contributes to induction of NF-κB activation in SCC. The results with PDTC suggest that oxidation may play a role in the induction of NF-κB in SCC. That TPCK, but not CPI and PDTC, was able to reduce the constitutive NF-κB suggested that the constitutive NF-κB in these cell lines involve protease activities and/or other signal transduction processes that may be distinct from those controlling TNF-induced NF-κB activity.

In the experiments above, we showed that both constitutive and TNF-α-induced KC production in Pam 212, LY-2, and LY-8 lines were correlated with the activation of NF-κB, and both constitutive and TNF-α-induced NF-κB were blocked by TPCK. To determine whether modulation of NF-κB with TPCK would affect KC expression, we treated Pam 212, LY-2, and LY-8 lines with TPCK 15 min before the addition of TNF-α as in Fig. 5 and examined KC mRNA expression by Northern blot analysis. As in Fig. 6, KC mRNA was more abundant in LY-2 and LY-8 than in Pam 212 cells. TNF-α treatment increased expression of KC mRNA in all three cells lines, and the induction was more robust in LY-8, consistent with the increase in KC protein induced under the same conditions in the experiment shown in Fig. 1. TNF-α-induced KC mRNA expression was inhibited by the addition of TPCK, at a concentration as low as 12.5 μm. The constitutive level of KC mRNA in LY-2 and LY-8 was also reduced by TPCK. We conclude that NF-κB plays a role in controlling constitutive and inducible KC production in these SCC cells.

Inhibition of NF-κB activity by proteasome/protease inhibitors suggested that NF-κB activities in SCC are subject to the modulation by IκBs. To directly address whether IκBα is involved in modulating constitutive or TNF-α-induced activation of NF-κB and expression of cytokine KC, we stably transfected Pam 212, LY-2, and LY-8 lines with an expression vector containing cDNA for a human IκBα super suppressor protein (hIκBαM), where substitution of serine 32 and 36 in IκBα renders the mutant protein resistant to phosphorylation-induced activation and degradation, thus providing a dominant suppressor specific for NF-κB. After selection in G418 for stable transfected cells, individual clones were isolated, expanded, and assayed for the expression of hIκBαM by Western blot. Among 14 Pam 212 clones assayed, 5 (36%) produced variable levels of hIκBαM. In clones F3 and C10, the level of hIκBaM expression was at least three to five times higher than the endogenous IκBa protein (Fig. 7 B). Surprisingly, only 2 of 30 (7%) clones in LY-2 and 1 of 12 (8%) in LY-8 showed positive hIκBαM production. The level of hIκBαM protein detected in these clones was much lower compared with that of endogenous murine IκBα proteins. No significant difference in transfection efficiency was observed among three cell lines as judged by the numbers of blue cells after transfection with CMV-lacZ. Therefore, our inability to obtain stable transfection of hIκBαM clones in metastatic reisolates LY-2 and LY-8 may suggest that NF-κB plays a role in cell growth or survival in these SCC cells.

Effects of hIκBαM on the TNF-α induced NF-κB DNA binding activity and KC production were assessed in the stable Pam 212 transfectants F3 and C10 by EMSA and ELISA. As shown in the EMSA in Fig. 7,A, NF-κB binding activity in control vector-transfected cells was low but induced by TNF-α, consistent with results in untransfected parental Pam 212 cells. In hIκBαM-transfected clone F3, TNF-α induction was largely impaired, whereas in clone C10, this induction was completely blocked. In contrast, AP-1 binding activities showed equal intensity among three cell lines tested and were not affected by TNF-α treatment (Fig. 7,A). Concomitant with reduced NF-κB binding activities, TNF-α-induced KC production in F3 and C10 was reduced by ELISA, with greater decrease in the latter (Fig. 7 C). Clones of LY-2 and LY-8, which showed weak expression of hIκBαM as compared with endogenous IκBα, were tested in ELISA and produced more KC protein than the parental lines (data not shown). KC induction by TNF-α was only partially inhibited (data not shown). These results provided direct evidence indicating that TNF-α-induced KC production in Pam 212 cells is highly dependent on the activation of NF-κB.

In this study, we demonstrated that metastatic variants LY-2 and LY-8 selected in the host environment exhibited higher levels of proinflammatory cytokine secretion and constitutive activation of NF-κB than the parental Pam 212 cell lines derived in vitro (Figs. 1,2,3and 4,A), and the increased constitutive activation of NF-κB in the metastatic LY-8 and LY-2 cells was confirmed in vivo (Fig. 4,B). A correlation between proinflammatory cytokine production and NF-κB activation was established by the induction of NF-κB in response to TNF-α, as well as by use of inhibitors of NF-κB activation. Induction of NF-κB activation by TNF-α was demonstrated by EMSA (Fig. 5), reporter gene assay (Fig. 3), and immunohistochemical staining (Fig. 4,A), consistent with the TNF-α-induced increase in cytokine production (Fig. 1). The protease inhibitor TPCK inhibited both constitutive and TNF-α-inducible NF-κB activation (Fig. 5), as well as KC mRNA expression (Fig. 6). Direct involvement of IκBα in modulating TNF-α-induced NF-κB activation and KC production was observed in Pam 212 cells when hIκBαM was overexpressed. Therefore, we have provided evidence that an increase in expression of NF-κB-dependent cytokines and constitutive activation of NF-κB occurs with metastatic tumor progression of SCC, and that the host environment may favor outgrowth of tumors where NF-κB is activated.

Expression of proinflammatory cytokines has been reported to promote tumor cell proliferation, host angiogenesis, inflammation, and catabolism in animal models and in cancer patients (5, 8, 9, 11, 12, 13, 14, 15, 16, 17), and increased expression of cytokines in these tumors has been associated with increased activation of NF-κB (37, 38, 39). In breast cancer, constitutive activation of NF-κB was found in three of three estrogen receptor-negative breast cancer cell lines, which produced functional IL-1α constitutively (39, 40). Sovak et al. (41) have detected NF-κB/Rel activity in two human breast cancer cell lines, in multiple human breast cancer specimens, and in carcinogen-induced primary rat mammary tumors cell lines. A human melanoma cell line, Hs294T, established from lymph node metastasis, also showed constitutive activation of NF-κB in association with expression of a chemokine MGSA/GRO-α, which serves as an autocrine growth factor for Hs294T melanoma cells (38, 42). In our murine SCC model, the constitutive activation of NF-κB associated with the more aggressive phenotype in metastatic lines LY-2 and LY-8 is also correlated with the higher production of proinflammatory cytokines, IL-1α, IL-6, KC, and GM-CSF by these cell lines (16, 17). Treatment of Pam 212 cells with IL-1α increased cell proliferation in the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay, whereas overexpression of KC in Pam 212 enhanced metastasis in vivo,⁵ suggesting one of the functions of NF-κB in tumorigenesis may be attributed to regulation of cytokine genes.

In addition to the potential effects of NF-κB-dependent proinflammatory and proangiogenic cytokines upon the pathogenesis of cancers, NF-κB has been implicated in tumor transformation, survival, and metastasis (22, 23, 40). NF-κB activation is induced during viral oncoprotein-mediated transformation in both mouse and human keratinocyte models (43, 44). Recently, constitutive activation of NF-κB has also been associated with overexpression of urokinase in pancreatic adenocarcinoma with increased metastatic potential (45), increased cell adhesion in Caco-2 colonic cancer cells (46), and promotion of tumor-transforming and growth activity in several solid tumors (47). We and other laboratories have also shown that activation of NF-κB can promote resistance of tumor cells to radiation (24, 48, 49, 50).

Although NF-κB appears to be involved in several critical biological events affecting malignancy, the mechanism of the constitutive activation of this factor in tumor cells is not clear. The development of constitutive activation of NF-κB with tumor progression in the host suggests that constitutive activation of NF-κB may involve host-selective factors. Pam 212 cells were established by culturing neonatal murine keratinocytes in vitro without exposure to host factors (18), and LY-2 and LY-8 cell lines were isolated from metastatic lymph nodes after tumor progression in vivo (17). Despite a similar growth pattern in vitro, LY-2 and LY-8 showed a dramatic increase in growth in vivo, suggesting that acquisition of increased growth may result from acquisition of a selective growth advantage or resistance within the host. For instance, we have observed increasing nuclear immunohistochemical staining of NF-κB toward the center of LY-2 and LY-8 tumors, which is the region where greater hypoxia is likely to occur (data not shown). In support of this hypothesis, we observed a strong cellular inflammatory and angiogenesis response in tumors formed from metastatic variants of Pam 212,⁶ suggesting that host cells could serve as a source of mitogenic cytokines and increased blood supply in vivo.

Constitutive activation of NF-κB could be due to autocrine regulatory mechanisms involving the cytokines produced by tumor cells. SCC in this and other models have been shown to produce cytokines/growth factors with NF-κB-activating activity, such as TNF-α, IL-1α, and platelet-derived growth factor. LY-2 and LY-8 produce a detectable amount of IL-1α (16), which is a potent inducer of NF-κB activation and KC production in Pam 212 cells.⁷ The role of IL-1α in induction of NF-κB is presently under investigation. All three cell lines produced minimal detectable levels of TNF-α between 5 and 15 pg/ml by ELISA. No reduction of constitutive NF-κB binding activity was observed in LY-2 cells after cells were incubated with an anti-mouse TNF-α neutralizing antibody for 24 h, whereas the amount of TNF-α antibody used was sufficient to block the induction of NF-κB in Pam 212 cells treated with 2.5 ng/ml murine recombinant TNF-α, suggesting that the constitutive NF-κB observed in LY-2 is unlikely to be due to the autocrine stimulation by TNF-α.⁸

Constitutive activation of NF-κB could be the result of oncogene activation within pathways that mediate phosphorylation and degradation of IκBs that regulate NF-κB. Decreased steady-state levels of IκBα and β proteins were observed in two of three estrogen receptor-negative breast cancer cell lines with constitutive activation of NF-κB (40), and rapid degradation of IκBα has also been reported in Hs294 melanoma cells (42), human colon cancer cell lines (51), and murine WEHI 231 lymphoma cells (52). Phosphorylation of IκBα and IκBβ at NH₂-terminal serines by the IκB kinase complex or at COOH-terminal PEST sequences by CKII renders these phosphorylated proteins as substrates for proteasome-dependent and -independent degradation (21). CPI is a specific proteasome inhibitor and has been used to efficiently block the induction of NF-κB by preventing the degradation of phosphorylated IκBs; PDTC is routinely used as antioxidant agent, which likely prevents NF-κB activation through antioxidation mechanisms (53); and TPCK is an inhibitor of serine protease that affects both the phosphorylation and degradation of IκBα as well as directly inhibiting NF-κB binding in vitro (34, 54). The mechanism for NF-κB inhibition by these inhibitors is not fully understood. In Pam 212 cells, all inhibitors tested were able to block the TNF-α-induced NF-κB activation. But in LY-2 and LY-8 cells, only TPCK inhibited constitutive activation of NF-κB. The selective inhibitory effects of protease inhibitors CPI and TPCK, and the antioxidant agent PDTC on the constitutive and inducible activation of NF-κB, indicate that distinct regulatory mechanisms could be involved in these activation pathways (Fig. 5). In our tumor model, we have not detected a decrease in the levels of IkBα and β proteins in LY-2 and LY-8 cells by Western blot analysis (data not shown). Additional experiments are required to compare the kinetics of phosphorylation and degradation of IκB proteins during NF-κB activation among these cell lines, which may shed light on the mechanisms involved in the constitutive activation of NF-κB regulated by IκBs.

The mechanisms for constitutive production of Gro-α, the human homologue of cytokine KC, have been shown to reside at the transcriptional level (55). NF-κB, Sp1, and high mobility group proteins HMGI(Y) are involved in constitutive as well as inducible expression of Gro-α in human melanoma cells (56). Ohmori et al. (57) studied lipopolysaccharide-induced KC expression in macrophages and demonstrated that NF-κB sites are necessary for the induction. The mechanism(s) for constitutive expression of KC is unknown. In our present study, constitutive as well as TNF-induced KC production is correlated with NF-κB but not AP-1. Modulation of NF-κB by either chemical inhibitors or IκBαM overexpression paralleled the level of reduction of KC. These observations argue that NF-κB, but not AP-1, plays an important role in controlling KC expression in SCC, and expression of cytokine KC may thus be a good indicator for NF-κB activity in our model.

We believe that NF-κB may promote tumor progression of SCC through enhancement of tumor survival and expression of cytokines and other genes that contribute to the malignant phenotype. This murine SCC model may provide an excellent system for studying the molecular regulatory events associated with NF-κB during the tumor progression. The constitutive activation and nuclear localization of NF-κB observed in this study suggests that NF-κB may be a useful marker for tumor aggressiveness and progression, as well as a target for therapy. Preclinical investigation using NF-κB inhibitors are under way to evaluate the importance of the constitutive activation of NF-κB in tumor progression.

We are grateful for the generous gift of the plasmid IgκB-Luc from Drs. Brown and Siebenlist of National Institute of Allergy and Infectious Diseases and their critical reading and comments on the manuscript. We also thank Dr. Yuspa of the National Cancer Institute for reading the manuscript, Dr. Giovana Thomas for providing plasmid pCMVLacZ for use in this study, and Dr. Melvin Spigelman of Knoll Pharmaceuticals for providing recombinant human TNF-α.