Introduction
The microenvironment that exists during chronic inflammation can contribute to cancer progression [
1,
2]. Therefore, anti-inflammatory strategies are being investigated as potential cancer therapies. However, such approaches may have undesired effects on host immune responses. Potentially, these negative effects could be limited by targeting interventions to specific cell types, but little is known about the contribution of individual inflammatory cell types to the progression of cancer.
The transcription factor nuclear factor kappa B (NF-κB) regulates inflammatory status and plays key roles in immune responses. NF-κB is a dimer formed from a multi-subunit family consisting of p65 (Rel A), Rel B, c-Rel, p105/p50 (NF-κB1), and p100/p52 (NF-κB2) [
3,
4]. In the classical NF-κB signaling pathway, the p50/p65 subunits are held in the cytosol by the inhibitory IκBα [
5,
6]. During activation, IκB kinase (IKK) 2 phosphorylates IκBα leading to ubiquitination and degradation. P50/p65 then translocates into the nucleus, binds to consensus DNA sequences activating expression of target genes [
3,
6,
7]. We and others have modulated this signaling using a constitutive form of IKK2 to activate or a mutant form of IκBα to inhibit NF-κB activity [
8‐
10].
Macrophages are key mediators of the interaction between inflammation, immunity and cancer. The role of macrophages in cancer has received attention due to the discovery of their tumor-promoting effects [
11‐
13]. Efforts have been made to classify macrophages according to phenotype. The M1 macrophage phenotype is associated with production of reactive oxygen species (ROS), presentation of antigens, release of inflammatory cytokines and cytotoxic effects on pathogens and tumor cells. In contrast, the M2 macrophage phenotype has a reduced ability to present antigens and produce ROS, expresses scavenger receptors, promotes angiogenesis and wound healing, and is associated with tumor-promoting effects [
14,
15]. However, tumor-associated macrophages (TAMs) can express genes associated with both the M1 and M2 phenotypes highlighting the need for further investigation into macrophage phenotype during tumorigenesis [
16].
Recent studies have investigated the role of NF-κB in macrophages during tumorigenesis. A LysM-Cre/floxed IKK2 transgenic, resulting in deletion of IKK2 in cells of the myeloid lineage (macrophages and granulocytes), led to a reduction in colon tumor incidence and size in a colitis-associated cancer model [
17]. An IκB-super repressor targeted to Kupffer cells (liver macrophages) led to a similar reduction in tumor incidence in a murine model of hepatocellular carcinoma [
18]. It has been reported that TAMs can induce invasive behavior of ovarian tumor cells via NF-κB [
19,
20]. In established orthotopic ovarian tumors, the introduction of macrophages with inhibited IKK2 led to a reduction in tumor burden. These effects were associated with a switch from M2 to M1 phenotype upon NF-κB inhibition [
21]. These studies suggest a tumor-promoting effect of NF-κB in myeloid cells, including macrophages.
Despite these data pointing to a pro-tumor role for NF-κB in macrophages, genes regulated by NF-κB could also lead to an anti-tumor phenotype, suggesting that effects may be more complex. We have generated transgenic mouse models to modulate NF-κB in specific tissues by introducing doxycycline (dox) in drinking water [
8,
10]. To investigate NF-κB signaling within macrophages during metastasis we used a bi-transgenic system in which the colony stimulating factor 1 receptor promoter (cfms) drives the monocyte/macrophage-specific expression of the reverse tetracycline transactivator (rtTA). To activate NF-κB, the cfms-rtTA mouse was crossed with a second transgenic in which constitutively active IKK2 (cIKK2) was controlled by the tet operon (termed IKFM). To inhibit NF-κB, cfms-rtTA is crossed with a transgenic line in which dominant negative IκBα is controlled by the tet operon (termed DNFM). We used these models in a mouse mammary tumor cell tail vein-injection metastasis model that is extensively used as a methodology to investigate the later stages of the metastatic process from the point at which circulating tumor cells are present. We find that activation of NF-κB in macrophages during a short window around the time of cell injection leads to a reduction in lung metastasis of mammary tumor cells. The mechanism involves induction of an anti-tumor M1 phenotype that rapidly clears tumor cells. Two days after injection of tumor cells this window of opportunity closes and from this point activation of NF-κB no longer exhibits anti-tumor effects.
Materials and methods
Tumor cells
PyVT R221A and PYG 129 polyoma tumor cells were isolated from PyVT mammary gland tumors and cultured as previously described [
9,
22].
Mouse strains
All animal experiments were approved by the Vanderbilt University Institutional Animal Care and Use Committee. All mice were on an FVB strain background, except the TG/CRE/FMR mice, which were mixed background (C57BL6 and FVB).
To generate the cfms-rtTA transgenic, the 7.2 kb mouse
cfms promoter region was used to drive the expression of
rtTA-M2 [
23]. The
cfms-rtTA-M2 transgenic construct was microinjected into mouse embryonic stem cells by standard methods. Progeny were screened for incorporation of transgene by southern blot and founder lines identified. Macrophage specific expression of rtTA was determined in mice transgenic for cfms-rtTA, tet-O-Cre and the ROSA26 LSL-lacZ allele [
24]. cfms-rtTA mice were crossed with mice containing the NF-κB inhibiting (tet-O)
7-IκBαDN-Myc-His construct or the NF-κB activating (tet-O)
7-FLAG-cIKK2 construct [
8,
10]. Double transgenics were termed DNFM or IKFM, respectively. Littermates lacking one or both transgenes were used as controls.
To generate TG/Cre/FMR mice, Tomato Red/GFP reporter mice [
25] were crossed with tet-O-Cre transgenic mice [
26]. Double transgenic progeny were then crossed with cfms-rtTA mice to produce triple transgenic offspring containing all three transgenes. Littermates lacking the tet-O-Cre were used as controls.
To induce transgene expression mice were treated with 2 g/L dox (Sigma, St Louis, MO, USA) in drinking water. Sucrose (5%) was added to decrease the bitter taste of dox water. A red bottle was used to prevent light-induced degradation and water was replaced twice weekly. For metastasis studies, 1 × 106 polyoma tumor cells (PyVT R221A or PYG 129) in PBS were injected via the tail vein. Mice were treated with dox either one week prior to cells until sacrifice, one week prior to cells until two days post injection, or two days post injection of cells until time of sacrifice. Mice were sacrificed at two weeks (PyVT R221A) or five weeks (PYG129) post cell injection for surface lung tumor quantification. Mouse lungs were inflated with Bouin's fixative (RICCA Chemical, Arlington, TX, USA). After 24 hours of fixation, surface lung tumors were counted.
For seeding studies, mice were treated with dox for one week then 1 × 106 PyVT R221A tumor cells were injected via the tail vein. Mice were sacrificed one or six hours post cell injection and lungs harvested for real-time PCR and western blot analysis.
RNA isolation and RT-PCR
Tissue was homogenized in Trizol (Invitrogen, Carlsbad, CA, USA) at sacrifice. For characterization, IKFM mice and controls were treated with dox for one week followed by intraperitoneal injection of 1.5 ml of 4% thioglycolate in PBS. At sacrifice, three days post injection, the peritoneal cavity was lavaged with PBS and cells pelleted.
RNA isolation and reverse transcription reactions were performed as described [
10] for all studies. For studies utilizing real-time PCR analysis, an Applied Biosystems Stepone Plus Real-Time PCR system and SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) were used. Primers: polyoma middle T [
27], TNF-α [
10], CXCL9: FOR 5' GTGGTGAAATAAAAAGATCAGGGC 3', REV 5' AAGAGAGAAATGGGTTCCCTG 3'; CCL3: FOR 5' TGCCCTTGCTGTTCTTCTCT 3', REV 5' GATGAATTGGCGTGGAATCT 3'; mannose receptor: FOR 5' CAAGGAAGGTTGGCATTTGT 3', REV 5' CCTTTCAGTCCTTTGCAAGC 3'; arginase-1: FOR 5' ATGGAAGAGACCTTCAGCTAC 3', REV 5' GCTGTCTTCCCAAGAGTTGGG 3'; GAPDH: FOR 5' TGAGGACCAGGTTGTCTCCT 3', REV 5' CCCTGTTGCTGTAGCCGTAT 3'.
Bone marrow derived macrophages
IKFM and DNFM mice were treated with dox (2 g/L) for one week and bone marrow-derived macrophages (BMDMs) isolated as described [
28]. BMDMs were treated with dox (1 μg/ml) for 16 hours. RNA was isolated as above, and RT-PCR was completed with the following primers: IKK: FOR 5' GGAGCTCCACCGCGGTGCGG 3', REV 5' TCAGGGACATCTCGGGCAGC 3', and DN: FOR 5' CCTGGCTGTTGTCGAATACC-3', REV 5' GGTGATGGTGATGATGACCGG 3'. BMDMs were grown on microscope slides and treated with dox as above. Cells were fixed in 3% paraformaldehyde and stained for F4-80 (Invitrogen, Carlsbad, CA, USA), with DAPI (Sigma, St Louis, MO, USA).
Western blotting
Whole cell lung extracts were prepared and western blot performed as previously described [
10]. Antibody: Cleaved-Caspase-3 (Cell Signaling Technology, Beverly, MA, USA).
Luminol imaging
Mice were treated with dox (2 g/L) for one week. L-012 (1.25 mg in 100 μl PBS, Wako Chemicals, Richmond, VA, USA) was intravenously injected into anesthetized mice. Mice were imaged with an intensified charge-coupled device camera (IVIS 200; Xenogen, Hopkinton, MA, USA). Light emission was detected as photon counts and analyzed by defining a standard area over the chest and determining total integrated photon intensity (Living Image software; Xenogen, Hopkinton, MA, USA).
Flow cytometry
Mice were sacrificed, lungs perfused with cold PBS, minced and incubated in RPMI media containing 0.7 mg/ml collagenase XI (Sigma, St Louis, MO, USA) and 30 μg/ml DNAse I (Sigma, St Louis, MO, USA) for 40 minutes at 37°C. Digests were strained through a 70 micron filter. Cells were pelleted and treated with 1 ml red blood cell (RBC) lysis buffer (ACK buffer), washed, and re-suspended in PBS. Cells were blocked with anti-mouse CD16/CD32 antibody (eBioscience, San Diego, CA, USA) before staining with anti-mouse antibodies: CD45 (30-F11), Gr-1 (Ly-6G), CD11b (M1/70), CD11c (N418), and CD19 (all eBioscience, San Diego, CA, USA); F4/80 and B220 (Invitrogen, Carlsbad, CA, USA); CD4 and NK1.1 (BD Pharmingen, San Diego, CA, USA). Analysis was performed on an LSRII cytometer with DIVA software (BD Biosciences, Franklin Lakes, NJ, USA).
CD11b positive cell isolation
Mice were treated with dox (2 g/L) for one week prior to sacrifice. Lungs were digested as described above. Cell suspensions were incubated with MACS CDllb MicroBeads (Miltenyi Biotec, Auburn, CA, USA) and separated by positive selection using MS columns.
Immunofluorescence
For whole lung tissue staining of IKFM and control mice, lungs were first perfused with PBS, then inflated and fixed in formalin overnight followed by paraffin embedding and sectioning. For cytospin staining, lungs were dissociated and CD11b+ cells positively selected as above. Cells were spun onto microscope slides then fixed in 3% paraformaldehyde. Antigen was unmasked by the sodium citrate method. Primary antibody: NF-κB phospho-p65 (Ser536) (Cell Signaling Technology, Beverly, MA, USA). Secondary antibody: goat anti-rabbit Alexa Fluor 594 (Invitrogen, Carlsbad, CA, USA). DAPI (Sigma, St Louis, MO, USA) was used for nuclei staining. Staining was visualized by Zeiss microscope and analyzed by MetaMorph software (Molecular Devices, Sunnyvale, CA).
TG/Cre/FMR lungs were perfused with cold PBS and then inflated and fixed for four to six hours in 4% paraformaldehyde before being paraffin embedded and sectioned. DAPI (Sigma, St Louis, MO, USA) counterstain was used as above.
Statistical analyses
Statistical analyses were performed using Graph Pad Prism (GraphPad Software Inc., La Jolla, CA, USA). All data are plotted graphically with vertical bars representing standard error. Unpaired Student's t test was used to assess differences between experimental conditions. A probability (P) value of less than 0.05 was taken as an appropriate level of significance.
Discussion
We generated a transgenic mouse model to investigate the role of classical NF-κB signaling in macrophages during tumorigenesis. We show that activation of NF-κB in macrophages in a mammary tumor tail vein metastasis model leads to a reduction in lung tumor formation with effects observed only when NF-κB is modulated prior to tumor cell introduction during the early seeding phase. Investigations into the lung phenotype associated with this effect show that activation of NF-κB leads to a shift in macrophage populations in the lung with a higher percentage of cells that are Gr1+/Cd11b+ and Gr1-/CD11b+ and a lower percentage of Gr1-/Cd11c+ cells. This shift in macrophage population and reduction in lung tumor numbers occurs in lungs that express increased levels of markers of the M1 anti-tumor macrophage phenotype. In agreement with this, we see a decrease in mammary tumor cell seeding following cell injection, an increase in apoptosis and enhanced formation of ROS and the cytokine CXCL9. Our data suggest that activation of NF-κB in macrophages within the lung during the seeding of lung metastases has an anti-tumor effect.
Our data are in contrast with other
in vivo studies highlighting a tumor-promoting role for macrophages [
17,
21]. This may be due to the different murine models and cancer types and the timing of NF-κB activation. These recently published studies have used the strategy of Cre-mediated deletion of the IKK2 to inhibit NF-κB signaling in myeloid cells. Our strategy employs specific activation of the signaling pathway by expression of a constitutive activator or direct inhibition by expression of a dominant form of inhibitor, either of which is regulated in an inducible manner. In our model, activation of NF-κB in macrophages before tumor cell arrival generates a hostile environment for tumor seeding and growth. This environment is similar to the M1-type macrophage response seen with a bacterial challenge [
33]. The tail vein metastasis model is inherently an acute model as mice develop a heavy metastatic load in the lungs at two weeks and need to be euthanized for humane reasons. This likely precludes observing the pro-tumor effects that have been reported in models of established tumors. Future studies to address effects on established tumors will need to be performed using orthotopic models or those in which spontaneous tumors arise, such as the polyoma transgenic model, to address this issue.
Recent studies highlight the complexity of NF-κB signaling during tumorigenesis. Tumor-associated macrophages isolated from an orthotopic fibrosarcoma model show a defective activation of NF-κB in response to lipopolysaccharide [
16]. Further investigation reveals that this lack of activation is due to p50 subunit homodimers and knockout of p50 restores the M1 macrophage phenotype and reduces tumor burden [
35]. Although this study suggests that defective NF-κB activation within macrophages leads to the development of an M2, pro-tumor phenotype, others find that NF-kappaB signaling, specifically IKK2, is necessary to maintain the M2 phenotype in an ovarian cancer model [
21]. These apparent contradictions were recently reviewed by Hagemann
et. al [
36] who propose that NF-κB regulation of TAMs is both context- and gene-dependent.
IKFM mice showed a significant shift in macrophage populations when NF-κB was activated for one week before tumors were established. Interestingly, when cells were analyzed at the third time point (where metastases are evident), a different pattern of surface marker expression was observed. This pattern was more similar to control populations. This agrees with recent reports showing the plasticity of the macrophage population during tumorigenesis with a different set of markers expressed at each stage [
37].
The cell marker phenotype in the IKFM mice when an anti-tumor effect was observed correlates with the reduction in tumor burden. For example, GR1+/CD11b+ myeloid cells are a source of ROS, the levels of which were seen to increase [
29]. CXCL9 was significantly up-regulated in IKFM lungs after one week of dox treatment. NF-κB activation leading to CXCL9 production within macrophages has previously been reported [
38]. The injection of a mammary tumor cell line stably expressing CXCL9 cDNA into mice resulted in smaller tumors as well as fewer lung metastases. This was found to be a T-cell-mediated effect, in that the CXCL9 overexpressing tumors had increased levels of infiltrating CD4 T cells as compared with vector controls [
34]. Although we see increased CXCL9 expression in our model, we observed a concurrent decrease in CD4 T cells as assessed by flow cytometry. This could imply a separate, non-T-cell-mediated, anti-tumor effect associated with CXCL9, or suggest that the time points analyzed were not optimal for analysis of lymphocyte infiltration. In either case, further investigation into the role of CXCL9 in this model is required.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
LC performed initial characterization of the model, completed various trials of the tail vein studies, and drafted the manuscript. WB participated in ongoing breeding of the transgenic mice, completed tail vein trials and carried out real-time PCR, western-blot analysis and CDllb bead separation experiments. HO completed immunoflourescent staining, flow-cytometry experiments and analysis, CDllb bead separation experiments - RT-PCR and immunofluorescence, and carried out tail vein injections. LChen participated in the breeding, genotyping, treatment, and collection of mice. TS carried out tail vein injections and aided in lung analysis. TZ and MO created the cfms-rtTA transgenic mouse line. TB was involved in data interpretation and critically edited the manuscript. FE conceived of the study, participated in its design and coordination and helped draft the manuscript. All authors read and approved the final manuscript.