The CCL6 Chemokine Is Differentially Regulated by c-Myc and L-Myc, and Promotes Tumorigenesis and Metastasis¹

Three members of the MYC oncogene family (C-MYC, N-MYC, and L-MYC) are dysregulated frequently in human cancers. Some of the most common examples are seen in Burkitt’s and AIDS-related lymphomas, where a large fraction of tumors are associated with translocations between C-MYC and one of the immunoglobulin loci; in neuroblastoma, which is often associated with N-MYC amplification; and in SCLC,³ where L-MYC amplification may occur concurrently with that of C-MYC and/or N-MYC (1).

Myc oncoproteins belong to the helix-loop-helix leucine zipper family of transcription factors (2). After their heterodimerization with another helix-loop-helix leucine zipper protein, Max, Myc proteins bind to specific DNA elements (E-boxes) of the sequence CAC/TGTG (3). Transcriptional activation of the bound gene is then mediated by an NH₂-terminal segment of the Myc protein termed the TAD (4). Two segments of the TAD, termed MBI and MBII, are highly conserved among the three Myc proteins, and play important roles in target gene regulation (5).

Myc proteins can also participate in transcriptional repression, although this appears to be mediated primarily through different target gene sequences, most notably initiator or NFY-B sites (6). Several studies have demonstrated that down-regulation of certain genes is particularly dependent on MBII (5, 6). Until now, repression mediated subsequent to E-box binding has not been described.

A major goal in the study of Myc proteins has been to identify their target genes. A number of these have now been characterized, mostly as a result of the application of DNA microarray technology (6, 7, 8, 9, 10, 11, 12). Many of the target genes encode proteins involved in cell cycle control, DNA synthesis, metabolism, apoptosis, and extracellular matrix formation. Some are also transforming when overexpressed (5, 6).

A related goal that has received little attention concerns the degree of overlap among the transcriptional targets for the three MYC genes. Indirect evidence suggesting that some, if not all, of these targets are the same, derives from studies demonstrating that the embryonic lethality of c-Myc−/− mice can be rescued by N-Myc (13), and that each of the MYC genes can rescue the growth deficit of c-Myc−/− fibroblasts and promote apoptosis (14). On the other hand, indications that the target genes may differ is provided by observations that the transforming ability of L-Myc is 10–100 times less potent than c-Myc or N-Myc, and that L-Myc−/− animals are phenotypically normal (15, 16).

In the current work, we provide evidence that CCL6, a member of the CC or β chemokine family (also known as C10/SCYA6), is directly up-regulated by L-Myc but is repressed by c-Myc in a MBII-dependent manner. Although the exogenous addition of CCL6 is well tolerated by some cell types, it is markedly proapoptotic for others. CCL6 overexpression also dramatically alters the in vitro and in vivo behavior of several cell types, either by relieving them of their growth factor dependence and/or by markedly enhancing their tumorigenic and metastatic potential.

32D cells transfected with wild-type murine c-Myc (32D-c-Myc cells) and the control, parental vector (32D-neo cells), have been described previously (17) as have 32D cells expressing c-Myc with deletions in the TAD (10). These latter included 32D-c-MycΔ2–42, 32D-c-MycΔMBI (deletion of amino acids 42–66), 32D-c-MycΔ70–130, 32D-c-MycΔMBII (deletion of amino acids 133–147), and 32D-c-MycΔTAD (deletion of amino acids 2–147). All of the latter lines produced their respective mutant proteins at levels comparable with that of the wild-type protein in 32D-c-Myc cells (10). All of the 32D-derived cell lines were grown in RPMI 1640 plus 10% FBS (Life Technologies, Inc., Grand Island, NY). Unless otherwise stated, cell lines also contained 10% conditioned medium from IL-3-producing WEHI-3B cells and 400 μg/ml G-418 (Life Technologies, Inc.). Most other cell lines were grown either in RPMI 1640 or DMEM containing 10% FBS or supplemented calf serum (Life Technologies, Inc.). Human SCLC lines were kindly provided by Dr. Michael Birrer (National Cancer Institute).

All of the antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). These included the 9E10 anti-Myc epitope tag monoclonal antibody (sc-40), rabbit polyclonal anti-c-Myc (sc-764), rabbit polyclonal anti-L-Myc (sc-790), and murine monoclonal antiactin antibody (sc-8432). A polyclonal antibody raised against bacterially expressed, full-length max, has been described previously and used for immunoprecipitations (18). Horseradish peroxidase-conjugated secondary antibodies were also purchased from Santa Cruz and were used according to the directions of the supplier.

Polyadenylated mRNA was isolated from 32D-c-Myc and 32D-l-Myc cells as described previously (10). Reverse transcription, labeling with Cy3 and Cy5 chromophores, and hybridization to murine cDNA microarrays (GEM Microarrays; Incyte Pharmaceutical, St. Louis, MO), were performed as described previously (10). Northern blots were performed with 5 μg of total RNA/lane, blotted to Nytran membranes (Schleicher & Schuell, Keene, NH), and hybridized in UltraHyb solution (Ambion, Austin, TX).

A cDNA clone encoding the entire CCL6 coding region was obtained from Incyte Genomics, Inc. The coding region was amplified with the following primers: forward: 5′-CGC GTC GAC ACC ATG AGA AAC TCC AAG ACT-3′ and reverse: 5′-CGC CTC GAG AGC AAT GAC CTT GTT CCC AGA TC-3′. In these sequences, italicized nts indicate engineered “GC clamps,” and SalI and XhoI restriction enzyme sites, respectively, and underlined triplets indicate the initiator methionine codon and the final triplet in the coding region. After digestion of the amplified product with SalI and XhoI, it was isolated by agarose gel electrophoresis and cloned into the XhoI site of the pSVL-3′MT-puro cloning vector (18). The resultant clone contained the CCL6 cDNA cloned in-frame at its 3′ end to a c-Myc epitope tag. The vector was linearized in the plasmid backbone for electroporation into 32D cells as described previously (17). Alternatively, supercoiled plasmid DNA was transfected into A549 or HeLa cells using Lipofectamine (Life Technologies, Inc.). Stable transfectants were selected in 1 μg/ml of puromycin (Sigma, St. Louis, MO).

The murine CCL6 promoter region between nts −1423 and +49 relative to the start of mRNA transcription was amplified from bacterial artificial chromosome DNA that was kindly supplied by Dr. Hisayuki Nomiyama (Kinki University School of Medicine, Osaka, Japan). PCR primers consisted of the following sequences: forward, 5′-CGC CTC GAG TAA GTT GGG CAT GAA CTG GGC A-3′ and reverse, 5′-CGC AAG CTT CCT CCT GGA GGA CAG GCC C-3′, where italicized bases denote engineered “GC clamps,” and XhoI and HindIII sites, respectively, to facilitate unidirectional cloning. After amplification, the fragments were digested with XhoI and HindIII, isolated by agarose gel electrophoresis, and cloned into the XhoI + HindIII-digested luciferase reporter vector pGL2-puro. This vector was constructed by ligating a ∼1 kb blunt-ended EcoRI-NheI DNA fragment containing a puromycin resistance cassette from the pAPuro vector (19) into the blunted-ended unique SalI site of the pGL2 luciferase reporter vector (Promega, Madison, WI).

SDS-PAGE, Western blotting, and enhanced chemiluminescence detection were performed as described previously (17, 18). In most cases, 25 μg of whole cell lysate/lane was used. To detect secreted, epitope-tagged CCL6, 32D cells were maintained for 24 h in tissue culture medium lacking serum. The supernatant was then concentrated 10-fold using PAGE-Perfect (Geno Technology, Inc., St. Louis, MO) before SDS-PAGE.

These were performed essentially as described (20) except that Staph A cells were replaced by Protein A-Agarose (Affigel; Bio-Rad, Hercules, CA). Isolated chromatin from ∼6 × 10⁶ cell equivalents was used for each point. To precipitate sheared, formaldehyde cross-linked chromatin, antibodies were added to a final dilution of 1:500. One-tenth of the final precipitated DNA (2 μl) was used in each PCR reaction. PCR primers were synthesized to allow amplification of a 353-bp region from the murine CCL6 promoter (nt −1416 to −1063) that contained all six of the E-box elements depicted in Fig. 2 A. The PCR primer sequences were: forward, 5′-GGG CAT GAA CTG GGC ACT GCT CCA-3′ and reverse, 5′-AAC TGC CAG GAA CTC CAG CTG CAG-3′. Other PCR primers were designed to amplify a 301-bp fragment from the 3′-untranslated region of the CCL6 gene (GenBank accession no. AB051897), located ∼6 kb downstream of the promoter (forward 5′-GGC AGG CAT TGT CAC CCA CTT TCT-3′ and reverse 5′-CAT CTC AGC TAG ATA TTA CCC GGC-3′); a 295-bp segment flanking both E-box elements located in the first intron of the murine ODC gene (GenBank accession no. JO3733.1: forward 5′-CAG CCA GCA GCT CGG CGC CAC CT-3′ and reverse CGA GGT CCA GGA GCA GCT GCC TTC-3′); and a 308-bp segment spanning nt 5252–5560 of the murine junB promoter (GenBank accession no. U20735: forward 5′-GTC GCC AGG CCA GCG TAG GAG G-3′ and reverse 5′-CTG TTC CAT TTT CGT GCA CAT CCG-3′). In addition to the above input PCR primer pairs, PCR reactions (100 μl) contained 1× PCR buffer (Promega), variable concentrations of MgCl₂, 100 μm of each deoxynucleoside triphosphate, and 2.5 units of Taq DNA polymerase (Promega). A total of 40 cycles were performed in a DNA Thermal Cycler 480 (Perkin-Elmer, Inc., Foster City, CA). The resultant amplified fragments were then resolved by 2% agarose gel electrophoresis. Confirmation that they corresponded to the regions of interest was obtained by hybridization of the blotted fragments with an end-labeled oligonucleotide synthesized to an internal sequence or by restriction digestion with enzymes recognizing internal sites.

These were performed as described previously (14).

To identify transcripts that were differentially regulated by c-Myc and l-Myc, we compared the DNA microarray profiles of 32D-c-Myc and 32D-L-Myc cells. These cell lines were derived from the IL-3-dependent 32D murine myeloid cell line after stable transfection with c-Myc or L-Myc expression vectors. Pooled transfectants expressed high levels of the respective oncoprotein and showed similar rates of accelerated apoptosis in response to IL-3 withdrawal (17).

Most transcripts that are differentially expressed between 32D-c-Myc and control cells (10) did not register in the current comparison between 32D-c-Myc and 32D-L-Myc cells, thus indicating similar transcriptional regulation by the two oncoproteins (data not shown). One exception was that of the CC chemokine CCL6, which showed a 7.5-fold greater relative up-regulation in 32D-L-Myc cells.

To confirm the differential expression of CCL6 detected by microarray analysis, we performed Northern analysis on RNAs extracted from 32D-c-Myc or 32D-L-Myc cells, either during logarithmic growth, or after a 12-h period of IL-3 deprivation. RNAs from control 32D cells, transfected with the empty parental vector (32D-neo cells), were also included in the analysis. As seen in Fig. 1 A, CCL6 transcripts were detected only in 32D-L-Myc cells, thus confirming the results of the DNA microarray analysis. Two distinct transcripts were noted, which may reflect the utilization of alternate consensus polyadenylation signals at positions 998 and 1308 of the murine cDNA sequence in the GenBank database (accession no. NH009139). The withdrawal of IL-3 had no discernible effect on the abundance of CCL6 transcripts.

The failure to detect CCL6 transcripts in 32D-neo and 32D-c-Myc cells could indicate that c-Myc either lacked the ability to up-regulate the CCL6 gene or that that the gene was actively suppressed by c-Myc. To distinguish between these possibilities, we performed Northern analyses on RNAs extracted from 32D cells stably expressing c-Myc proteins with defined deletions in the TAD (10). These included MBI (amino acids 43–69), MBII (amino acids 131–147), and the less conserved inter-Myc Box regions comprising amino acids 2–42 and 70–130. We have shown previously that these mutant proteins are expressed at levels equivalent to that of full-length, wild-type c-Myc, and that each of these regions is responsible for controlling the expression of distinct subsets of c-Myc-responsive genes in 32D cells (10). As seen in Fig. 1,B, cells expressing the c-MycΔMBII mutant showed a marked up-regulation of CCL-6. Taken together with the data presented in Fig. 1 A, these results indicate that the expression of CCL6 is markedly up-regulated by L-Myc in 32D cells, that the chemokine in actively suppressed by wild-type c-Myc, and that the MBII domain plays an important role in this suppression.

To determine whether the CCL6 gene was a direct transcriptional target for L-Myc and c-Myc, we first sequenced ∼2.0 kb of the murine CCL6 promoter. Examination of this region showed that it contained six variant E-box elements (CATGTG; Ref. 3) located between −1130 and −1380 relative to the presumed start of mRNA transcription, with five of the sites residing within a 110 nt span between −1130 and −1220 (Fig. 2,A). To determine the responsiveness of the promoter to various c-Myc and L-Myc proteins, the region between −1425 and +50 nt was amplified and cloned into the pGL2-Puro luciferase reporter plasmid. The vector was then linearized in its backbone and cotransfected along with pCMV-β-gal into each of the 32D cell lines described above. After selection in puromycin, the stably transfected pooled clones were assayed for expression of luciferase and β-galactosidase, with the latter being used to control for differences in transfection efficiencies (<two-fold in all of the cases). As seen in Fig. 2 B, significant increases in luciferase activity relative to those of control 32D-neo cells were seen only in 32D-L-Myc and 32D-c-MycΔMBII cells.

To determine whether the effects of L-Myc and c-MycΔMBII on the CCL6 promoter were direct, we performed ChIP studies on several of the above cell lines (20). In these experiments, in vivo cross-linking of DNA-bound protein was achieved by the direct addition of formaldehyde to cells suspended in tissue culture medium. Whole chromatin was then isolated from the cells, sheared to an average size of 300–500 bp, and immunoprecipitated by the addition of anti-c-Myc, anti-L-Myc, or control antibodies. After the reversal of cross-links, the precipitated DNAs were initially used as PCR templates with primers that spanned the E-box-containing CCL6 promoter region. Analysis of the PCR products (Fig. 2,C, top panel) revealed several features. First, the CCL6 promoter region was not amplified from any of the cell lines after control chromatin precipitation performed in the absence of antibody or with a nonspecific antibody. Second, the promoter region was amplified from 32D-neo, 32D-c-Myc, and 32D-c-MycΔMBII-derived chromatin precipitated with anti-c-Myc antibody but not from the same cells precipitated with anti-L-myc antibody. In several independent experiments, the intensity of these latter two bands ranged between 1.6 and 3 times that of the former. This was in good agreement with the Western blots shown in Fig. 1,C and our previous reported results (10) indicating that 32D-c-Myc and 32D-c-MycMBII cells contained about three times as much c-Myc protein and control 32D-neo cells. Third, after ChIP with anti-L-Myc antibody, the promoter region was amplified only from 32D-L-Myc cells. Finally, the promoter region was not amplified from 32D-L-Myc-derived chromatin after immunoprecipitation with anti-c-Myc antibody. Taken together with the findings presented in Fig. 1 and Fig. 2 A, these findings suggest that the CCL6 promoter, while capable of binding either endogenous or overexpressed c-Myc, remains unresponsive to the oncoprotein unless its repression is eliminated by the removal of MBII. The inability to detect bound c-Myc protein in 32D-L-Myc cells additionally suggests that c-Myc and L-Myc proteins compete for the same binding site(s) in the CCL6 promoter, but that the latter, when expressed at high levels, can displace the former. This is consistent with the finding that endogenous c-Myc protein can still be detected in 32D-L-Myc cells.⁴ Therefore, the absence of CCL6 promoter-bound c-Myc is not because of a silencing of the endogenous c-Myc gene and the subsequent decay of c-Myc.

Various control ChIP experiments were also performed (Fig. 2, C and D). In these experiments, and using the same chromatin precipitates, we were unable to amplify a region from the 3′-untranslated region of the CCL6 gene, which resides ∼6 kb downstream of the promoter region and contains no known E-box elements. In contrast, The precipitation pattern for the first intron of the ornithine decarboxylase gene, which contains two c-Myc-responsive E-box elements (21), mimicked that of the CCL6 promoter. In other experiments, none of the anti-Myc antibodies was able to precipitate chromatin containing the promoter region of the junB gene, which is not known to be a Myc target gene. Finally (Fig. 2 D), an anti-Max antibody (22) was able to precipitate the CCL6 promoter region from all of the cell lines, consistent with the notion that all of the Myc proteins bind to their cognate sites as Myc-Max heterodimers.

Finally, to demonstrate that c-Myc and L-Myc transcriptionally compete for the CCL6 promoter, we performed reciprocal competitive transient transfection experiments. In the first, 32D-L-Myc cells were transfected with a constant amount of the CCL6 promoter-luciferase reporter vector described above and increasing amounts of a c-Myc expression vector. As seen in Fig. 2 E (right panel), luciferase activity decreased as a function of c-Myc input. In the second study, 32D-c-Myc cells were transiently transfected with a constant amount of CCL6-luciferase reporter and increasing amounts of an L-Myc expression vector. As seen in the left panel this resulted in a dose-dependent enhancement of luciferase activity. Both sets of results are consistent with the idea that each of the transiently expressed Myc proteins competes with its stably expressed relative for regulation of the CCL6 promoter.

To gain a better understanding for the function of CCL6, it was stably expressed as an epitope-tagged fusion protein in several of the above 32D cell lines. Pooled clones were then examined for both cell-associated and secreted CCL6. As seen in Fig. 3, A and B, all of the cell lines transfected with the CCL6 expression vector expressed and secreted high levels of the chemokine. CCL6 was not detected in cells or supernatants derived from empty vector control transfectants. Each of the c-Myc proteins was also present in these cells at levels comparable to those as described previously by us (Fig. 3 C; Ref. 10).

Because some chemokines have been reported to affect the survival of certain cell types (23), we assessed the ability of each of the above 32D cell lines to survive and proliferate in the absence of IL-3. As seen in Fig. 3 C, neither the enforced expression of CCL6 alone, nor the combination of L-Myc+CCL6 nor of c-MycΔMBII+CCL6 was sufficient to abrogate IL-3 dependence. Indeed, as in the case of control 32D cell, all of these lines succumbed to apoptotic death within 72 h of removal of the cytokine. However, the combination of c-Myc+CCL6 not only prevented apoptotic death, but allowed for continued cell proliferation after IL-3 withdrawal. The 32D-c-Myc+CCL6 cells doubled approximately five times in the course of the first 4 days of the study, thus indicating that the assumption of IL-3-independent growth was immediate and thus a property of all of the cells in the culture. However, in the presence of IL-3, these cells did not proliferate any more quickly than did control cells (data not shown). Thus, we conclude that CCL6 not only overcomes the potent c-Myc-mediated proapoptotic phenotype associated with growth factor removal (24), but actually cooperates with c-Myc to impart a new behavior to the cells, namely that of growth factor (IL-3)-independent proliferation.

Despite the ability of CCL6 to impart IL-3-independent growth of 32D-c-Myc cells, we encountered considerable difficulty in establishing stable CCL6-expressing clones of certain other cell types. This suggested that CCL6 might exert pleiotropic effects and that in some cells it might actually induce apoptosis. To test this, we either cocultivated various known tumorigenic and nontumorigenic cell lines with CCL6-overexpressing 32D cells or grew them in CCL6-conditioned medium. As seen in Fig. 4, A and B, several of the cell lines tested showed enhanced apoptosis in the presence of CCL6 as confirmed by trypan blue staining and TUNEL assay. Although our survey was not exhaustive, it suggested that nontransformed cells are more sensitive to the proapoptotic effects of CCL6 than are transformed cells. However, exceptions were seen in both cases.

The IL-3 independence of 32D-c-Myc+CCL6 cells (Fig. 3) prompted us to investigate their in vivo survival. At the same time, we wished to study the consequences of enforced CCL6 overexpression on the behavior of known tumorigenic cell lines. Therefore, based on the apoptotic profiles of the cell lines depicted in Fig. 4,A, demonstrating that A549 lung cancer cells and HeLa cells were resistant to the proapoptotic effects of CCL6, we established several CCL6-overexpressing clones from each line. Control clones were also established from cells transfected with the empty parental vector. As seen in Fig. 5 A, several of the clones examined from each of the former transfections expressed and secreted high levels of CCL6. Under standard culture conditions, all of the CCL6-overexpressing clones grew at rates that were indistinguishable from those of their vector control counterparts (data not shown).

One clone of each of the above cell types, as well as 32D cells described previously, were next inoculated s.c. into nude mice, which were then monitored weekly for tumor development. Of the various 32D cell lines, only 32D-c-Myc+CCL6 cells were tumorigenic (Fig. 5,B). These formed readily detectable tumors within 2–3 weeks and grew rapidly over the ensuing 3–4 weeks. Similarly, A549-CCL6 cells grew at a much faster rate than their control A549-puro counterparts (Fig. 5,C). In the former case, tumors first became detectable within 4–5 weeks of inoculation, compared with 6–7 weeks for A549-puro tumors. In contrast to our findings with 32D and A549 cells, the growth rates of HeLa-CCL6 tumors were virtually indistinguishable from those of control HeLa-puro tumors (Fig. 5 D). We have subsequently obtained similar results in smaller groups of animals with two additional A549-CCL6 clones and one additional HeLa-CCL6 clone (data not shown).

Histological examination of tumors derived from 32D-c-Myc+CCL6 cells (Fig. 6, A and B), showed them to be comprised almost exclusively of immature myeloid elements that could readily be distinguished from neighboring cells by their strong myeloperoxidase positivity (Fig. 6 C). These tumors readily breached the surrounding capsule and invaded adjacent normal structures such as s.c. fat and muscle. Distant metastases were also identified, and involved lung, liver, heart, and kidney. In some animals, complete replacement of the bone marrow was observed. Cells derived from primary tumors retained their IL-3 independence when recultured in vitro and expressed CCL6 protein at levels comparable with those observed before in vivo proagation (data not shown).

Given the proapoptotic effects of CCL6 on certain cell types observed previously, we also performed TUNEL assays on some of the tissue sections taken from the above animals. Considerable numbers of apoptotic cells were seen adjacent to the invading tumor cells in most samples (Fig. 6 D).

Histological examination of tumors derived from A549-CCL6 cells also showed dramatic differences from A549-puro cell-derived tumors of comparable size. Whereas the latter nearly always remained confined to their capsules or pseudocapsules, the former aggressively invaded surrounding tissues (Fig. 6,E). Indeed, in many cases, it was difficult or impossible to discern a capsule. Metastases to liver, lung, and spleen were also observed frequently (Fig. 6, F and G). TUNEL assays revealed the presence of robust apoptosis in tissues near the periphery of the infiltrating tumors and within the tumor capsules (Fig. 6,H). In contrast, TUNEL assays performed on A549-puro tumor-bearing animals showed a notable lack of apoptosis in the tumor capsule or adjacent tissues (Fig. 6 N).

Although the expression of CCL6 did not influence the actual growth rates of HeLa cell-derived tumors (Fig. 5,D), microscopic examination nonetheless provided evidence for their more aggressive behavior. Whereas neoplasms originating from HeLa-puro clones did infiltrate adjacent tissue, they seldom showed evidence of distant metastatic spread, and little evidence for apoptosis of adjacent normal tissues was observed (Fig. 6, O and P). In contrast, HeLa-CCL6-derived tumors showed a much more aggressive pattern of local tissue invasion and distal metastases (Fig. 6, I–K). In virtually all of the cases, this was accompanied by a marked apoptotic response of normal tissues adjacent to the tumor edges (Fig. 6 L).

Taken together, our results indicate that CCL6 can cause tumorigenic conversion of otherwise growth factor-dependent and nontumorigenic cells, and can greatly augment the in vivo growth rates and/or invasive/metastatic capabilities of other tumor cell lines established previously. In each of these cases, this is accompanied by a marked apoptotic reaction of normal cells in the path of the advancing tumor.

Human SCLC is the only known human neoplasm associated consistently with amplification of the L-MYC locus, which occurs in 10–15% of such cases (1). In some instances, this is associated with a coamplification of C-MYC and/or N-MYC (1). We examined the expression of each of these genes, as well as CCL6 in 8 SCLC cell lines, some of which had been determined previously to overexpress L-Myc.⁵ As seen in Fig. 7, 4 of the lines (N691, H510, H1836, and H889) expressed high levels of L-Myc transcripts, 2 (H378B and H1694) expressed considerably lower levels, and 2 (H298 and H345) expressed an undetectable amount. High levels of c-Myc expression were seen in 3 cases (H510, H298, and H378B), lower levels in a fourth (H345), and barely detectable, basal levels in the remainder. Finally, a single cell line (H345) expressed fairly high levels of N-Myc.

Rehybridization of these blots with the murine CCL-6 probe showed high level expression of the chemokine in one cell line (N691). Interestingly, this was one of the two lines that did not also express c-Myc or N-Myc, and of these two, it expressed higher levels of L-Myc. These results indicate that CCL6 expression may be deregulated in SCLC cell lines in association with L-MYC gene amplification and overexpression, but that this is not always the case.

One of the questions concerning the MYC oncogene family is why its members have been evolutionarily conserved. Many studies have supported the general idea that the expression of c-Myc, N-Myc, and L-Myc is differentially regulated, both at the tissue level and during ontogeny (1, 25). Their association with specific tumor types (1) additionally suggests highly specialized roles for each, as does the finding that homozygous deletion of either c-Myc or N-Myc results in embryonic lethality (26, 27), whereas deletion of L-Myc does not (28). Despite the evidence suggesting differential functions of these genes, little information has emerged regarding the molecular basis for this.

Chemokines are a large group of low molecular weight chemotactic cytokines that were originally identified as mediators of leukocyte trafficking (23, 29, 30, 31). More recently, they have been shown to exert a variety of effects on cellular proliferation, apoptosis, and neoplastic cell migration. With regard to the latter, it has been shown that breast cancer cells frequently express the chemokine receptors CXCR4 and CCR7 on their surfaces, and that their predilection to metastasize to liver, lung, and bone reflects the expression of specific chemokines by these organs (32). Similarly, infiltration of lymph nodes by CLL cells, a prognostically significant feature of the disease, may be associated with expression of CCR7 by the CLL cells. This allows their migration through high endothelial venules of the lymph nodes, which express the CCR7 ligands CCL19 and CCL21 (33). In several experimental model systems, enforced expression of individual chemokines has been reported to either enhance or inhibit the metastatic and/or invasive properties of tumors (reviewed in Refs. 29, 30).

In the current work, we have identified CCL6, a member of the CC or β chemokine family, as a specific and direct positive target for L-Myc. Thus, CCL6 represents the first reported target for a Myc oncoprotein that is not similarly regulated by other members of the family. Indeed, our data are compatible with a model in which c-Myc, rather than exerting no effect on the CCL6 gene, actually represses it. The basis for this conclusion rests on several findings. First, overexpression of c-MycΔMBII is associated with an up-regulation of CCL6 comparable with that seen with L-Myc. This suggests that c-MycΔMBII behaves in a dominant-negative fashion. Second, ChIP experiments indicate that endogenous c-Myc, as well as ectopically expressed c-Myc and L-Myc occupy the same region(s) of the CCL6 promoter. Third, overexpression of L-Myc results in the displacement of endogenous c-Myc from the CCL6 promoter. The observation that the CCL6 gene remains repressed after a prolonged period of IL-3 starvation, during which time endogenous c-Myc levels should be markedly reduced (10), suggests that other alterations of CCL6-associated chromatin are likely to be necessary for the activation of the gene.

The binding of c-Myc to its positively regulated target genes typically occurs at consensus E-box elements located within the promoter or first intron (6, 7). Negative regulation by c-Myc is much less well understood, although it appears more complex and until now has not been identified as being mediated by E-box binding (6). Rather, in several cases thus far described, it involves the interaction of c-Myc with non-E-box elements of the promoter, particularly initiator or NF-YB binding sites (6, 34, 35). Indeed, it is not completely clear that this regulation actually involves the direct binding of c-Myc to the gene (36, 37, 38). The work reported here suggests that E-boxes can, at least in the case of the CCL6 gene, serve as either positive and negative elements, and that the ultimate mode of regulation depends on the nature and disposition of the bound Myc oncoprotein. However, we cannot rule out the possibility that c-Myc may be also be repressing the CCL6 gene through other, non-E-box-dependent mechanisms.

The means by which c-Myc activates the transcription of its target genes include the recruitment of histone acetylase complexes, ATP-dependent chromatin remodeling, and promoter clearance (reviewed in Ref. 6). In the first case, considerable evidence indicates that MBII plays a crucial role in the interaction of c-Myc with TRRAP, a large protein that serves as a nucleation center for histone acetylases and their associated proteins (5, 39). On the other hand, proteins such as Bin1, MM-1, and pag have been shown to interact with the MBII region of c-Myc, and, in some cases, to serve as negative regulators of certain c-Myc target genes and biological functions (6, 40). Taken together, these observations suggest that the MBII region of c-Myc may be involved in both transcriptional activation and repression as we have suggested previously (10). Which of these activities predominates may depend on the relative abundance of these various MBII-interacting proteins, tissue-specific factors, and non-MBII regions of the TAD. Despite the high degree of similarity between the MBII regions of c-Myc and L-Myc, they nonetheless manifest distinctly different functions and divergent sequences outside of MBI and MBII. Therefore, the differences between c-Myc and L-Myc on the CCL6 gene may well involve regions of the TAD encompassing more than their MBII regions.

A surprising feature of CCL6 is its ability to confer IL-3-independent growth and aggressive tumorigenic behavior to 32D cells. This appears to be a novel property of CCL6, which has been suggested previously to play roles in wound healing and cardiac allograft rejection (41, 42). However, whereas CCL6 expression was necessary for these properties, it was not sufficient, being seen only when cells concurrently overexpressed c-Myc. Indeed, neither L-Myc nor c-MycΔMBII was able to substitute for c-Myc. Given that L-Myc is a much weaker transforming agent than c-Myc (15) and that the MBII domain is required for the transforming abilities of c-Myc (5, 39, 43), it is tempting to speculate that the ability of CCL6 to cooperate with c-Myc to impart growth factor independence and alter tumor behavior is related in some way to its activity as a potent transforming oncoprotein.

We were unable to compare the tumorigenic behavior of 32D-c-Myc+CCL6 cells to that of any other 32D cell line because only the former demonstrated IL-3 independence and autonomous tumor growth. However, the observation that CCL6 expression altered the growth and/or invasive behaviors of two other already tumorigenic cell lines (Fig. 5 and Fig. 6) argues that the chemokine functions in ways other than to promote cell survival. Indeed, the ability of CCL6 to induce apoptosis of certain cell types both in vitro and in vivo suggests that one of its major effects is to enhance tumor invasion by disrupting local tissue barriers. Overexpression of CCL6 may also allow for cellular migration into organs that express the CCL6 receptor, precisely the inverse effect that has been described for chemokine receptor-expressing mammary carcinoma and CLL cells (32, 33).

One of the best known attributes of c-Myc is its ability to induce apoptosis (24, 44, 45, 46). It has been argued that this may represent a cellular response that recognizes the impending neoplastic catastrophe arising as a consequence of c-Myc deregulation, and that eliminates such incipient transformed cells in favor of host survival (41). Numerous in vitro and in vivo studies have supported this model by demonstrating the dire consequences of suppressing these apoptotic pathways (46, 47, 48, 49, 50). However, additional cellular safeguards, which suppress the emergence of other properties essential to the transformed state, such as tissue invasion, have not been as intensively investigated. Given the strong synergy that exists between c-Myc and CCL6 with regard to tumorigenesis and metastatic spread, it is tempting to speculate that the active suppression of CCL6 in the face of c-Myc overexpression may represent an additional cellular mechanism to ensure against additional tumorigenic evolution, even after proapoptotic responses have been circumvented.

In conclusion, the studies reported here have identified the CCL6 chemokine gene as a direct positive transcriptional target for L-Myc and a direct negative target for c-Myc. The opposite actions of these two oncoproteins appear to be the result of their binding to identical or closely adjacent E-box elements in the CCL6 promoter. In turn, the level of positive or negative regulation appears to be determined by a balance between these two related Myc oncoproteins. An implication of these findings is that either of these proteins might by itself be capable of regulating CCL6 expression provided that MBII interactions with various cofactors was permitted. The consequences of combined CCL6 and c-Myc deregulation are to transform non-neoplastic cells to a tumorigenic and highly invasive phenotype, and to markedly enhance the tumorigenic behavior of cells transformed previously. These studies characterize for the first time differences in the transcriptional repertoire between c-Myc and L-Myc. In addition, they identify novel chemokine properties that bear investigating in other contexts.

We thank Dr. Hisayuki Nomiyama for supplying the CCL6 bacterial artificial chromosome, Dr. Michael Birrer for SCLC lines, Amy Drzal for microarray data analysis, Linnette Grove for DNA sequencing, and Sean Ward for advice on nude mice.