Background
DARC was originally identified as a blood group antigen, and later as a coreceptor for malaria [
1‐
3]. More recently, it was demonstrated that DARC binds a variety of CXC and CC chemokines [
4‐
7]. In particular, DARC binds to the CC chemokines monocyte chemotatic protein-1 (MCP-1; CCL2) and regulated upon activation normal T expressed and secreted (RANTES; CCL5), and the CXC chemokines interleukin-8 (IL-8, CXCL8), growth related gene alpha (GRO-α, CXCL1), and neutrophil activating peptide-2 (NAP-2, CXCL7). Expression of DARC
in vivo is not limited to the red blood cell, and DARC expression has been detected in endothelial cells of the kidney, spleen and brain and in large vessel and post-capillary venules [
1,
2]. It has also been detected in epithelial cells lining the ducts of the kidney, in pulmonary alveoli [
8], and in a subset of neurons [
9]. Interestingly, DARC is also expressed in endothelial cell types in individuals who lack expression of DARC (Duffy antigen negative) on their red blood cells [
6,
10]. Therefore, the expression of DARC by cells other than red blood cells suggests that its presence on these cells may be important in modifying the biological behavior of specific chemokines.
Similar to other receptors identified to date that are known to bind chemokines, DARC is a seven-transmembrane receptor, however, unlike other chemokine receptors, ligand binding by DARC does not induce G-protein coupled signal transduction nor a Ca
2+-flux [
11]. Amino acid alignment with other seven-transmembrane G-protein coupled receptors indicates that DARC lacks a highly conserved DRY motif in the second intracellular loop of the protein that is known to be associated with G-protein signaling [
12]. Despite the lack of evidence for DARC signal transduction, it has been shown in DARC-transfected cells that DARC is internalized following ligand binding [
10]. These results have led to the hypothesis that expression of DARC on the surface of red blood cells, endothelial, neuronal cells, and epithelial cells may act as a sponge and provide a mechanism by which inflammatory chemokines may be removed from circulation as well as their concentration modified in the local environment.
The members of the CXC chemokine family that bind DARC are also known to induce angiogenic responses in a variety of assays. The ELR+ CXC chemokines IL-8, GRO-α and epithelial neutrophil activating protein 78 (ENA-78, CXCL5) have all been shown to be important mediators of tumor-derived angiogenesis
in vivo [
13‐
19]. Neutralization of ELR+ CXC chemokines in SCID mouse models of human non-small cell lung cancer (NSCLC), as well as human prostate cancer has demonstrated that these molecules significantly contribute to tumor-associated neovascularization, tumor progression and metastasis [
17‐
19]. Interestingly, the angiostatic members of the CXC chemokine family, namely the ELR- CXC chemokines interferon inducible protein-10 (IP-10, CXCL10) and monokine induced by γ-interferon (MIG, CXCL9) do not bind to DARC [
7]. Thus DARC demonstrates a disparate activity in its ability to bind to angiogenic versus angiostatic chemokines, and thus might contribute to the net angiogenic activity within the local tumor microenvironment.
We hypothesized that over-expression of DARC in a tumor could lead to sequestering of angiogenic chemokines from the local vascular environment and thus inhibit tumor-associated neovascularization and tumor progression. To test this hypothesis, we chose a model of human NSCLC, A549 cells, whose growth and metastatic potential
in vivo has been previously shown to be dependent on the production of angiogenic ELR+ CXC chemokines by the tumor cells [
17,
18]. We predicted that expression of DARC by the A549 tumor cells would lead to the binding of angiogenic chemokines by the tumor cells themselves and would result in the sequestering of angiogenic factors away from the tumor-associated endothelial cells. When A549 cells were transfected to stably over-express human DARC, we found that over-expression of DARC by two independently generated tumor cell clones resulted in increased tumor size compared to control transfected cells. Although tumor size was increased, the overexpression of DARC within the tumor was associated with marked tumor necrosis, reduced tumor cellularity, reduced vascularity, and impaired metastatic potential, as compared to the control tumors. These results indicate that DARC expression within the tumor microenvironment alters the characteristics of tumor growth and interferes with tumor-derived neovascularization, tumor cellularity and metastatic potential.
Discussion
In this study, we sought to take advantage of the fact that the Duffy antigen receptor for chemokines binds multiple members of the ELR+ CXC family of chemokines that induce angiogenesis
in vivo. The importance of these molecules in tumor-associated neovascularization has been previously shown in NSCLC, prostate carcinoma, melanoma, bladder cancer, ovarian and gastric carcinomas [
17,
19,
20,
22‐
26]. As DARC has not been shown to be associated with the induction of signal transduction events following ligand binding, we hypothesized that the expression of DARC within the local tumor microenvironment would lead to sequestering of angiogenic CXC chemokines, and thus compete for their binding to signal-inducing chemokine receptors on endothelial cells. We generated human NSCLC A549 cells that stably expressed human DARC, and assessed their ability to form tumors in SCID mice. Unexpectedly, we observed that the tumors derived from DARC-expressing A549 cells had an increased tumor volume as compared to control tumors. This increased tumor volume observed
in vivo however, was not associated with an enhanced ability to induce neovascularization, as tumors derived from DARC-expressing A549 cells showed increased necrosis and a reduced degree of neovascularization as compared to control tumors. It is also unlikely that DARC-A549 cells expressed increased levels of CXC chemokines
in vivo, as ELISAs for IL-8, ENA-78 or GRO-α from tumor homogenates at 7 weeks post-injection showed no significant differences in these protein levels between any of the tumors examined (data not shown). Taken together, these results indicate that overexpression of DARC in the tumor microenvironment likely sequesters ELR+ CXC chemokines within cells thus interfering with the normal chemotactic gradients that need to be formed and thus having profound effects on tumor growth, neovascularization and metastatic potential.
Our results support the hypothesis that DARC functions as a "sink"
in vivo, whereby it binds ELR+ CXC chemokines and prevents them from associating with other receptors that would lead to the induction of biological responses, such as angiogenesis. Normally, chemokines form a concentration gradient via a combination of soluble and extracellular matrix bound protein that results in activation and chemotaxis of endothelial and immune cells along that gradient. We observed no differences in the amounts of ELR+ CXC chemokines found in tumors in vivo suggesting that DARC-expressing and control transfected A549 cells made similar amounts of these proteins, however we did observed decreased amounts of these chemokines in conditioned cell supernatants in vitro. This contrast is likely a reflection of the manner in which the experiments were performed as ELISA analysis of the levels of chemokines in vivo tumors was derived from frozen-thawed homogenates of tumor biopsies and thus the chemokines that would be bound to the surface or trapped within the golgi or endoplasmic reticulum of tumor cells would be free to be measured. In contrast, the ELISA data generated from DARC or control transfected tumor cells in vitro was derived on conditioned cell supernatants only, thus would not measure the amount of chemokine bound to the surface of cells nor sequestered within cells due to inhibition of processing and secretion through the endoplasmic reticulum and golgi apparatus following binding to DARC in DARC-expressing cells. Moreover, our immunohistochemical data from tumor sections suggests increased amounts of IL-8 on or in tumor cells expressing DARC as compared to control tumors
in vivo and supports the notion that perhaps the chemotactic gradients of chemokines required to induce appropriate neovascular responses are not formed and instead the chemokines are sequestered by the tumor cells themselves. We cannot however exclude the possibility that DARC may mediate an as yet uncharacterized signal transduction event. DARC is known to lack the highly conserved DRY amino acid motif that interacts with G-proteins in its second intracellular loop [
12], however it remains possible that DARC may interact with other cellular receptors or membrane components and elicit specific responses to ligand binding and thus may be contributing to the phenotype we observed via this mechanism.
Although DARC can be detected on a variety of different cell types, including epithelial cells, expression of DARC
in vivo is primarily localized to endothelial cells in a variety of organs and tissues. The fact that DARC is expressed on the endothelial cell surface in people who are negative for DARC expression on their erythrocytes [
10] implies that DARC plays an important role in modulating endothelial cell biology. It is of interest to note that the pattern of endothelial cells that express high levels of DARC is similar to areas that are key in leukocyte trafficking and extravasation [
27‐
29]. This observation suggests that DARC may be involved in the maintenance of chemotactic gradients across the endothelium. As mentioned our data supports the notion that expression of DARC in the tumor microenvironment sequesters chemokines thus preventing the formation of the appropriate gradients to which various cell types would migrate toward. Indeed our results suggest that DARC-expressing tumors have an impaired ability for NK and neutrophil extravasation into the tumors. DARC is also strongly expressed on post-capillary venules [
1], the same cell type that is primarily involved in the initiation of neovascular responses. More recently it has been demonstrated that transgenic mice expressing DARC under an endothelial specific promoter have reduced angiogenic responses in the corneal micropocket assay in response to MIP-2 a CC chemokine that is also known to bind DARC [
30]. Thus DARC may play a significant role in the modulation of neovascular responses, and our data would support the idea that DARC plays a significant role in tumor-associated angiogenesis.
We observed an increase in tumor necrosis and a significant decrease in tumor-associated angiogenesis in tumors expressing DARC which may be attributed to DARC acting as a "sink"
in vivo and disrupting chemotactic gradients that would be required for angiogenesis to occur. More recently it has been demonstrated that DARC knockout mice undergo modified infiltration of leukocytes in response to inflammatory stimuli such as lipopolysaccharide suggesting that DARC may play a significant role in leukocyte-endothelial cell adhesion and migration [
31,
32]. Moreover, immobilization of IL-8 is required for neutrophil emigration into tissue and it has been suggested that this occurs via a cooperative mechanism involving IL-8 binding to DARC and heparan sulfate [
33]. Thus the decreases in immunological cell types present in DARC tumors also supports the disruption of appropriate chemotactic gradients along the endothelial surface as DARC expressed on tumor cells would act as a "sink" and sequester chemotactic chemokines within the tumor itself. This may partially explain the increase in size of DARC expressing tumors as normally dying tumor cells in necrotic areas would be removed by infiltrating immune cells leading to resolution of the necrosis and this process might not occur if these cells cannot extravasate into the tumor space, leading to increases in the necrotic areas and hence overall size of the tumor. Impaired infiltration of immune cells might also contribute to the reduced neovascularization observed in DARC-expressing tumors as it has been shown that the presence of macrophages may contribute to tumor growth and progression and induction of tumor neovascularization [
34,
35]. As we observed a slight reduction in the number of tumor-associated macrophages in DARC-A549 tumors as compared to control tumors it remains possible that in addition to DARC directly binding angiogenic chemokines and preventing them from directly affecting the neovasculature, that reduction in infiltrating immune cells may also contribute to the reduced neovascularization associated with DARC-A549 tumor growth.
As mentioned, we hypothesized that DARC expression by the tumor cells would bind the angiogenic chemokines and sequester them from endothelial cells, thus preventing endothelial cell chemotaxis and tumor-associated angiogenesis. The lack of tumor-associated angiogenic activity would lead to significant increases in tumor necrosis, which is what was observed in our DARC-A549 generated tumors. The lack of endothelial cell recruitment into the tumor may partly explain why we observed a reduction in the spontaneous metastasis of the DARC-A549 tumors, since metastasis is thought to occur by tumor cell invasion of the vasculature. The ability of DARC to inhibit neovascular responses could result in the inhibition of the establishment and growth of micrometastases. This hypothesis is supported by our data that DARC2F5 cells have fewer and significantly smaller metastatic nodules in the lung, as compared to the control A549 cells.
Materials and methods
Reagents
Polyclonal rabbit anti-human IL-8 sera was produced by immunization of rabbits with IL-8, (Peprotech, Rocky Hill, NJ) in multiple intradermal sites with complete Freund's adjuvant. This antibody has been previously well characterized for its neutralizing capacity [
17]. The IL-8 anti-serum specificity has been confirmed by Western blot analysis against recombinant human IL-8 and was found to neutralize 30 ng of IL-8 at a dilution of 1:1000 [
17]. Furthermore, in a sandwich ELISA, this antibody is specific for IL-8 without cross-reactivity to a panel of 12 other recombinant human cytokines or the murine chemokines KC and MIP-2 [
17].
Human cell lines
The A549 human adenocarcinoma cell line (ATCC CCL 185) was purchased from the American Type Culture Collection (Rockville MD). A549 cells were maintained in RPMI 1640 medium supplemented with 1 mM glutamine, 25 mM Hepes buffer, 100 U/ml penicillin, 100 ng/ml streptomycin and 10% fetal bovine serum (all from Whitaker Biomedical Products, Whitaker CA).
Generation of stably transfected NSCLC cell lines
The cDNA for human DARC was amplified by RT-PCR from total RNA isolated from K562-DARC cells that had been stably transfected to express DARC using the following primer pair: Forward – 5' CAT CTG AAT TCC TGC AGA GAC CTT GTT C 3' and Reverse – 5' AAA GGA TCC GTC TAG ACT TTA ATT CAG GTT CAG 3'. RT-PCR was performed using the Access RT-PCR kit according to the manufacturer's directions (Promega Corp., Madison WI). The reverse transcriptase reaction was performed at 48°C for 45 min, followed by denaturation at 94°C for 2 min. PCR amplification was then performed using the following cycle conditions: 94°C for 1 min, 56°C for 1 min, 68°C for 2 min for a total of 40 cycles, followed by a 7 min extension cycle at 68°C. The resulting 1.2 kb DNA band was gel purified using the Wizard PCR purification kit and cloned into the pTARGET mammalian expression plasmid system (both from Promega Corp., Madison WI). A549 cells were transfected with varying quantities of plasmid DNA using the calcium phosphate-mediated transfection technique [
36]. Following transfection, cells were grown in the presence of 400 μg/ml of G418 (Geneticin, Gibco BRL, Grand Island, NY), and G418 resistant colonies were isolated and expanded in culture. One empty pTARGET control vector transfected line was chosen for further comparisons and from herein will be referred to as "control". Two independently generated DARC-expressing A549 cell lines were chosen for further analysis, DARC1A6 and DARC2F5.
ELISA analysis
The quantity of IL-8, ENA-78, and GRO-α, GRO-γ, IP-10 and MIG present in conditioned cell supernatants was determined by specific ELISA, using a modification of the double ligand method as previously described [
13,
37]. Briefly, flat-bottomed 96-well microtiter plates were coated with 50 μl/well of specific polyclonal rabbit anti-CXC chemokine antibodies (1 μg/ml in 0.6 M NaCl, 0.26 M H
3BO
3, 0.08 N NaOH, pH 9.6) for 24 hr at 4°C, and then washed with PBS pH 7.5 plus 0.05% Tween-20 (wash buffer). Plates were blocked with 2% BSA in PBS for 1 h at 37°C and then washed three times with wash buffer. 50 μl of sample (1:10 and neat) was added and the plates were incubated at 37°C for 1 h. Plates were washed three times, 50 μl of appropriate biotinylated polyclonal anti-CXC chemokine antibodies (3.5 ng/μl in PBS, pH 7.5, 0.05% Tween-20 and 2% FBS) were added, and plates were incubated at 37°C for 45 min. Plates were washed three times, streptavidin-peroxidase conjugate was added, and the plates were incubated for 30 min at 37°C. Plates were washed again and 100 μl of 3, 3', 5, 5'-tetramethylbenzidine (TMB) chromogenic substrate was added. Plates were incubated at room temperature to the desired extinction and the reactions were terminated by the addition of 100 μl/well of 1 M H
3PO
4. Plates were read at 450 nm in an automated microtiter plate reader and the amount of CXC chemokine present was determined by interpolation of a standard curve generated by known amounts of recombinant CXC chemokine protein. The sensitivity for the specific chemokine ELISAs was >50 pg/ml and these assays failed to cross-react with a panel of other known cytokines and chemokines.
Northern blot analysis
Expression of DARC was detected at the message level by Northern blot analysis. Total RNA was extracted from cells in Trizol (Gibco BRL, Rockville MD), and 10 μg of RNA was subjected to formaldehyde agarose gel electrophoresis. The RNA was then transferred overnight by capillary transfer in 10 × SSC to nylon membranes (Roche, Indianapolis IN) and the membranes were subsequently UV cross-linked. For mRNA detection of the northern blot, dioxygenin (DIG)-labeled probes were generated following random priming reactions of a gel purified cDNA fragment using the DIG-High Prime kit according to the manufacturer's recommendations (Roche, Indianapolis IN). Hybridization and detection of the DIG-labeled probes was performed according to the manufacturer's directions (Roche, Indianapolis IN). Ethidium bromide staining of the blot for 28S and 18S rRNA was used as a control to ensure equal loading of the gel.
Detection of DARC on transfected cells
DARC protein was detected on transfected cells by FACS analysis. Briefly, cells were removed from flasks in 1 mM ice cold EDTA in PBS and then incubated at 4°C for 1 hr with 5 μg/ml Fy6 monoclonal antibody to human DARC or with isotype control antibody. Following two washes with ice-cold PBS containing 0.2% FBS, cells were incubated with FITC-conjugated anti-mouse antibody and incubated for an additional 1 hr at 4°C. Cells were again washed two times, and were then fixed with 2% paraformaldehyde prior to FACS analysis.
Binding of IL-8 by DARC transfected cells
Transfected cells were removed from flasks following incubation in 1 mM ice cold EDTA in PBS. Cells were then incubated in the presence of biotinylated IL-8, and bound ligand was detected using FITC-conjugated streptavidin according to the manufacturer's protocol (Pharmingen, San Diego CA). Bound IL-8 ligand was then detected by FACS analysis.
Human NSCLC-SCID mouse chimeras
Six to eight week old CB-17 SCID mice were injected subcutaneously (sc) in each flank with 1 × 106 A549 NSCLC cells in 100 μl. On a weekly basis once palpable tumors were visible, tumors were measured in three dimensions with engineer's calipers. Tumor volume was calculated using the equation axbxc where 'a' is the longest diameter, 'b' is its perpendicular, and 'c' is the depth of the tumor. At 3, 4 and 7 weeks post-tumor cell injection, animals were euthanized by ketamine overdose, the tumors were removed and measured as described above, and a portion was fixed in 4% paraformaldehyde for histological analyses. Another portion of the tumor was snap frozen and stored at -70°C for generation of protein extracts. Total protein extracts were generated by subsequent homogenization and sonication in Complete™ antiprotease buffer (Roche Molecular Diagnostics, Indianapolis, IN) and filtration through a 1.2 μm filter. Protein concentration was determined using the BCA Protein Assay Kit (Pierce, Rockford IL), and specific chemokine ELISAs were performed as described above. All chemokine concentrations were normalized to total protein.
FACS analysis
Two 6 mm tumor punches were minced in 5 ml of Dispase® (Becton Dickinson, San Jose, CA) solution and incubated with agitation for 1 h in a 37°C water bath. Cells were further separated by repeatedly aspirating the cell suspension through a 20 ml syringe, and filtration through gauze. Cells were then pelleted at 600 × g for 10 min, resuspended in sterile water for 30 s to lyse remaining red blood cells, and washed in 1 × PBS. Cells were counted and transferred at a concentration of 5 × 106 cells/ml to fluorescent antibody solution (1% FA buffer, 1% FBS, and 0.1% sodium azide), and maintained at 4°C for the remainder of the staining procedure. 100 μl of the cell suspension was labeled with FITC-conjugated anti-mouse CD31 to detect mouse endothelial cells, or with PE-conjugated anti-human CD49b to detect human A549 tumor cells (both from Pharmingen, San Diego CA). All samples were simultaneously incubated with Tri-color-conjugated anti-mouse CD45 to detect endogenous immunological cell types that were then gated out. These results are therefore presented as the percentage of positive cells following the exclusion of murine CD45 positive cell populations. Immune cells were also detected in parallel samples using PE-conjugated anti-mouse DX5 (for NK cells, Pharmingen, San Diego CA), FITC-conjugated Ly-6G (for neutrophils, Pharmingen, San Diego CA), and FITC-conjugated Moma-2 (for macrophages, Serotec, Raleigh NC). The unbound antibody was washed with FA buffer and the cell suspension subsequently analyzed by FACS.
In situ detection of apoptotic tumor cells
For each tumor group, three paraffin-embedded tumor sections were taken a minimum of 60 μm apart from 5 different tumors (n = 15 per group). For detection of apoptosis, DNA fragmentation was determined using the In situ cell death detection kit (Roche, Indianapolis IN). The percentage of cells staining positive by TUNEL was determined in 10 random fields of view at 200× magnification. Fields of view that were completely necrotic were excluded for these purposes.
Immunohistochemical detection of IL-8
Tumor sections were subjected to immunohistochemical staining for IL-8 as previously described [
17,
38]. Briefly, tumor sections were incubated in a 1:500 dilution of rabbit anti-human IL-8 serum or rabbit pre-immune serum. Primary antibody was detected with biotinylated goat anti-rabbit antibodies followed by incubation with streptavidin-conjugated HRP and colorimetric detection with 3,3'-diaminobenzidine (DAB) (all detection reagents from Vector Laboratories, Burlingame CA). Sections were counterstained in Mayer's hematoxylin.
Statistical and morphometric analyses
All statistical analysis of data was performed using the Statview 4.5 software program (SAS Institute Inc., Cary NC). Groups of data that appeared statistically different were compared by Students t test for comparison of means, and were considered statistically significant if p-values of <0.05 were obtained. The Mann-Whitney U test was used to compare groups of observations that did not appear normally distributed, and were considered statistically significant if p-values of <0.05 were obtained. For analysis of lung metastases and necrotic areas, morphometric analysis was performed on at least 16 separate hematoxylin and eosin stained sections taken 60 μm apart from each of 5 different animals under 200× magnification. An Olympus BH-2 microscope coupled to a Sony 3CCD camera was used to capture images that were then analyzed for total area of metastatic lesions using the NIH Image 1.55 software. For determination of tumor necrosis, the average percentage of necrotic area per high power field (200×) for each tumor was determined by morphometric analysis, and this number was then multiplied by the tumor volume yielding the average necrotic volume for each tumor.
Acknowledgements
This work has been supported, in part, by grants from the National Institutes of Health (CA87879, HL66027, P50HL67665, and P50CA90388 for R.M.S. and KO8HL04493 for J.A.B.). C.L.A. is a research fellow of the National Cancer Institute of Canada supported with funds provided by the Terry Fox Run.