Introduction
There is increasing evidence that monocyte derived myeloid cells expressing vascular markers such as Tie2 or VE-Cadherin support tumor growth [
1‐
5]. These cells are recruited to regions of hypoxia and promote angiogenesis and vasculogenesis [
6,
7]. Myeloid cell recruitment to the tumor bed appears to precede or coincide with the 'angiogenic switch'[
8,
9]. In an established tumor, myeloid cells appear to be a primary source of resistance to anti-VEGF therapy, suggesting a critical role for these cells in tumor angiogenesis [
5].
The exact mechanism of action of myeloid cells remains contentious. These cells can clearly promote angiogenesis through local production of angiogenic factors[
1,
10‐
13]. Some studies have suggested that these cells may be able to trans-differentiate to assume an endothelial cell fate, incorporate into vessel lumens, and contribute to vasculogenesis[
3,
14‐
17].
While the exact function of these proangiogenic myeloid cells remains controversial, murine studies confirm a critical role for these cells in tumorigenesis and indicate that these cells may be novel therapeutic targets for solid tumor therapy. Genetic manipulations to inhibit or eliminate these cells in both spontaneous and xenograft murine tumor models can severely restrict tumor growth [
3,
7,
9,
18]. Similarly, therapeutics targeting these cells reduce microvascular density and restrict tumor growth [
15,
19].
Proangiogenic myeloid cells similar to those found in mice have also been identified in human tumors. Myelo-monocytic cells expressing the hematopoietic marker CD14 and various vascular markers such as Tie2 (Tie2
+ Monocytes), VE-Cadherin, and VEGFR2 have been reported to take part in both ischemia-associated and tumor-associated angiogenesis [
17,
20]. We reported the presence of a proangiogenic myeloid cell population, expressing numerous myeloid (CD14, CD45, CD11c, CD11b) and vascular (VE-Cadherin, CD31, CD146) surface markers, in ovarian cancer [
21]. Given the dual phenotype of these cells, expressing both myeloid and vascular specific markers, and an angiogenic phenotype, we have termed these cells vascular leukocytes (VLC) [
15,
21]. VLC represent 10–70% of host cells and up to 30% of all cells in ovarian cancer ([
21] and unpublished data. In vitro and in vivo studies indicate VLC play a role in tumor angiogenesis. Increased recruitment of VLC to tumors by the chemokine B-Defensin-29 significantly increased murine tumor growth [
15]. Similarly, the direct addition of VLC to human tumor xenografts increased tumor microvascular density. VLC produce numerous pro-angiogenic factors such as TGF-β, VEGF, and Interleukin-8. VLC promote endothelial tubulogenesis and participate in perfusable vascular structures in matrigel in vivo [
15,
21,
22]. Importantly, inhibiting or eliminating VLC or similar myeloid cells in mice inhibits angiogenesis and severely restricts tumor growth [
15,
19].
Similar to VLC, proangiogenic CD14
+/Tie2
+ monocytes have recently been reported to be present in human tumors [
20]. Tie2
+ monocytes were identified in low numbers in the peripheral blood of cancer patients. Like VLC, Tie2
+ monocytes are present in high numbers in tumor tissue, but are rare in normal tissue. Also similar to VLC, the addition of Tie2
+ monocytes (but not Tie2-depleted monocytes) to tumor xenografts enhanced tumor microvascular density [
23]. Tie2
+ monocytes were described in many solid tumors including colon, lung, renal and breast cancer.
As animal studies indicate that VLC and Tie2+ monocytes are potentially legitimate therapeutic targets for solid tumor therapy, we sought to determine the relationship of VLC and Tie2+ monocytes. Furthermore, we attempted to identify an anti-VLC therapeutic for use in human cancers. We demonstrate here that many VLC appear to be a subset of Tie2+ monocytes. We identify the expression the hematopoietic antigen CD52, the target of the immunotherapeutic Alemtuzumab, on human VLC and Tie2+ monocytes. We show that Alemtuzumab is capable of inducing complement-mediated VLC killing. Finally, anti-VLC therapy with an anti-CD52 immunotoxin significantly restricted ovarian tumor growth in a murine ovarian tumor model. These studies provide important pre-clinical data supporting the use of Alemtuzumab as a therapeutic agent for ovarian cancer patients.
Materials and methods
Tissues
Stage III epithelial ovarian cancer (n = 10), and ductal breast cancer specimens (n = 1), non-small cell lung carcinoma (n = 3) (provided by Dr. Steven M. Albelda and Dr. Doug Arenberg) and melanoma (n = 3) (provided by Dr. David Elder), normal ovary (n = 2) and normal endometrium (n = 2) were collected at the University of Pennsylvania or the University of Michigan. After obtaining informed patient consent, ascites was collected either intraoperatively or at the time of therapeutic paracentesis. All specimens were processed in compliance with IRB and HIPAA requirements.
Tumor Processing
Freshly harvested solid tumors were mechanically dissected into 1–2 mm pieces and then further isolated to single cells using the Medi-machine (BD Pharmingen). Cell suspensions were then passed through a 40 um filter and finally isolated on ficoll gradient as previously described [
21].
Ascites Processing
For FACS characterization of VLC, ascites associated cells were concentrated by centrifugation and then red blood cells were lysed using ACK buffer (lonza, Walkersville, MD. Host cells were then isolated using a Ficoll gradient. Cells were then passed through a 40 um filter followed by 4 passes through a 28G needle to isolate single cells for FACS. For Alemtuzumab induced cytotoxicity assays in whole ascites, after red cell lysis, whole cell pellets were resuspended in 1/20th of the original volume of ascites supernatant and used directly in cytotoxicity assays.
FACS
Human CD45+/VE-Cadherin+ (CD144) vascular leukocytes and CD45(-)/VE-Cadherin+ tumor endothelial cells were FACS isolated from the ficoll isolated cells using APC anti-CD45 (BD Pharmingen, San Diego, CA) and PE-mouse anti-human CD144 antibody (eBioscience, San Diego, CA). CD52 expression was confirmed with using FITC-anti-human CD52 (GeneTex San Antonio, TX). For qRT-PCR experiments, a second vascular marker CD146 (P1H12-eBiosciences), was used in conjunction with CD45 and VE-Cadherin to increase purity.
Tie2 expression was confirmed using biotin-anti-human-Tie2 (Abcam Cambridge, MA) coupled with streptavidin-FITC. Tie2 monocytes were characterized using mouse anti-CD14-FITC (BD Pharmingen) and Mouse anti-human Tie2-APC (R&D Systems Minneapolis, MN). VE-Cadherin expression on Tie2 monocytes was confirmed using anti-VE-Cadherin-PE antibody. In order to avoid nonspecific antibody binding, PBS containing 10% normal murine serum (Sigma, St. Louis, MO) and 25 μg/ml anti-mouse Fc receptor (2.4G2 BD Pharmingen) were added prior to incubation. Mouse VLC were characterized using anti-CD45-APC (BD Pharmingen), anti-CD14-FITC and anti-CD14-PE (BD Pharmingen), anti-VE-Cadherin-biotin (Bender-Medsystems), and anti-CD52-PE (MBL, Cambridge, MA).
VLC FACS-isolated from ovarian tumor as described above were incubated with 10 μg/ml of Alemtuzumab (Genzyme Cambridge, MA) for thirty minutes. Isolated VLC were washed and incubated with 10% human serum or heat inactivated serum at 37°C for one hour (human serum was inactivated by incubating at 60°C for thirty minutes immediately prior to use). CD3+ peripheral blood lymphocytes were used as a positive control. Cells were then stained with Annexin-FITC (BD Pharmingen) and propidium iodide (BD Pharmingen) per manufacturer's protocol. To assure cellular viability throughout the assay, an aliquot of untreated VLCs were maintained in culture for the duration of the experiment. These untreated VLCs were stained for Annexin-V/PI in parallel with Alemtuzumab treated cells +/- inactivated serum. Cells negative for both Annexin V and PI were deemed viable cells.
A single cell suspension of whole ascites cells (host and tumor cells) suspended in ascites fluid was incubated for 90 minutes with 10 μg/ml Alemtuzumab or heat inactivated Alemtuzumab (heated at 80°C for 30 minutes). Cells were then immediately labeled with anti-CD45-APC (BD Pharmingen) and anti-VE-Cadherin-PE (eBioscience), or Annexin-FITC and 7-Amino Actinomycin D (7-AAD BD Pharmingen) and analyzed by FACS. Once again to assess cellular viability an aliquot of cells which receive no treatment were maintained at 37C in the ascites fluid throughout the course of the experiment. Viability of this control aliquot was then assessed with AnnexinV and 7AAD. AnnexinV(-)/PI(-) cells were considered viable
Quantitative RT-PCR
RNA was isolated from fresh VLC using the TRIzol method. RNA was reverse-transcribed into cDNA using superscript III per manufacturer's directions (Invitrogen Carlsbad, CA) and quantitative PCR was performed using 2 ng of total cDNA and SYBRgreen (Applied Biosystem; CD52, 5'primer CTTCCTCCTACTCACCATCAGC, 3'primer CCACGAAGAAAAGGAAAATGC).
Histology
Immunofluorescence was performed on fresh frozen, acetone fixed tissue using an anti-CD52 antibody (1:100 GeneTex, Inc) and anti-VE-Cadherin FITC antibody (1:200 Bender MedSystems). Immunohistochemistry was performed on murine tumors with anti-CD31 antibody (1:800 BD Pharmingen) and vecta-stain (Vector Labs Burlingame, CA) per protocol as described by the manufacturer.
CD52 Immunotoxin Development
Anti-CD52 antibodies (MBL Cambridge, MA) were biotinylated per protocol (Pierce). Biotinylation was confirmed by FACS analysis of murine splenocytes using biotinylated anti-CD52 antibody coupled with streptavidin-PE conjugate (BD Pharmingen). After biotinylation was confirmed, streptavidin-saporin (Advances Targeting Systems, San Diego, CA) was incubated with biotin labeled anti-CD52 antibodies in a 1.5:1 molar concentration. 2 μg/ml anti-CD52-saporin conjugate was then incubated with isolated ascites-associated cells for 36 hours in vitro and cytotoxicity confirmed by trypan blue and FACS staining (data not shown). To confirm in vivo toxicity, tumor bearing animals were treated twice-weekly with 2 ug of anti-CD52-saporin antibodies (n = 5) or control antibody (n = 3). After three weeks peripheral blood was collected, RBCs were lysed with ACK buffer, and then PBMCs were analyzed by FACS. Similarly tumors were resected, processed into single cells as described above and analyzed for VLC by FACS. Finally tumor ascites-bearing animals were treated with 2 μg of CD52-saporin or control IgG-saporin (n = 5 per group) daily for 48 hours and then ascites cells were harvested, red cells were lysed using ACK buffer, and whole ascites cell samples were analyzed for VLC by FACS.
Treatment of Flank Tumors
20 × 10
6 ID8-VEGF cells were injected subcutaneously into the flanks of C57BL6 mice and the tumors were allowed to grow for two weeks. The animals were then treated twice weekly with 2 μg of anti-CD52-saporin immunotoxin, or rat-IgG-saporin or immunopurified rabbit IgG-saporin control (a total n = 10, n = 5 and n = 5 respectively, in two independent experiments). Immunotoxins were administered intraperitoneally twice-weekly for three weeks. Rat and rabbit immunoglobulin controls revealed similar results and are presented as pooled data. Tumor growth curves were analyzed using ANOVA and Student's t-test At the time of sacrifice a subset of animals were perfused with biotinylated lycopersicon esculentum (tomato) lectin as previously described [
21].
Treatment of Intraperitoneal Tumors
10 × 106 ID8 cells were injected intraperitoneally into C57BL6 mice randomized by weight. Starting one week after the injection of tumor cells, mice were treated with 2 μg of anti-CD52-saporin immunotoxin or rat-IgG-saporin (n = 10 per group in two independent experiments) twice-weekly for three weeks. Animals were weighed to assess tumor growth. Animals were euthanized when they demonstrated 10 gm of weight gain secondary to ascites or animals appeared moribund. Survival curves were compared with the log-rank statistic.
Microvascular Density Analysis
CD31 IHC was performed simultaneously on four representative sections from 4 flank tumors in the treatment and control groups. Each section was systematically photographed in neighboring 40× fields such that 80–100% of each tumor section was photographed. Total CD31 stain area, as defined by pixel density and hue, was assessed using Olympus Microsuite Biological Suite software. Area of staining was then compared between control and treatment groups using a two-sided student's t-test.
Discussion
Our study adds to a growing body of literature indicating myeloid cells are legitimate therapeutic targets in the treatment of solid tumors. Several studies have used transgenic mice to demonstrate the importance of various myeloid cell populations. MMP-9 knockout mice were used to demonstrate a role for Gr
+/CD11b
+ cells in tumor vascularization [
3]. In fact, MMP-9 producing bone marrow derived cells have been implicated in both tumor angiogenesis and vasculogenesis [
7,
18]. Similarly, a transgenic suicide gene approach was used to demonstrate a potent anti-tumor effect of eliminating Tie+ monocytes [
23]. While representing important proofs of concept, these techniques obviously cannot be applied to humans. Other studies have utilized immunotherapeutic approaches; antibody therapeutics targeting chemokine receptor-6 and scavenger receptor-A on VLC each demonstrated restricted tumor growth and reduced vascular density [
15,
19]. However, these antibodies were against murine antigens, and therefore not directly translatable to humans.
The therapeutic effect seen with anti-CD52 therapy of ovarian tumors in mice is consistent with the aforementioned murine studies that indicate that myeloid cells promote tumor angiogenesis, vasculogenesis, and tumor growth. We observed a clear reduction in microvascular density in tumors treated with anti-CD52 therapy. This reduction in microvascular density correlated with a reduction of tumor vascular perfusion.
The observation that Alemtuzumab therapy can potently kill ovarian cancer VLC identifies a bona-fide therapeutic with which to test the importance of anti-VLC/myeloid cell therapy in human solid tumors. It is important that these studies confirmed the ability of Alemtuzumab to induce complement-mediated VLC killing within tumor ascites, an environment that closely resembles the in vivo tumor microenvironment. This would suggest that treatment effect will not be minimized by tumor-associated immunosuppressive elements or complement inhibitors. In addition, as Alemtuzumab killing is complement-mediated rather that cell-mediated, Alemtuzumab killing is less likely to be negatively impacted by dysfunctional cellular immunity. This is consistent with the activity of Alemtuzumab seen in chronic lymphocytic leukemia (CLL).
We demonstrated that Alemtuzumab can effectively kill VLC. We also observed that VLC appear to be a subset of Tie2
+ monocytes. Therefore, Alemtuzumab is capable of killing at least a subset of Tie2
+ monocytes. Furthermore, CD52 expression was identified on the vast majority of ovarian tumor-associated Tie2
+ monocytes, independent of their relationship to VLC, suggesting Alemtuzumab can target the majority of Tie2
+ monocytes. Tie2
+ monocytes have been reported in several solid tumor types including colorectal, breast, gastric, pancreatic, and lung carcinomas[
20] Consistent with this finding, we observed VLC in melanoma, breast, lung, and ovarian cancer. Taken together, these observations suggest that Alemtuzumab may be an effective therapeutic agent targeting VLC/Tie2+ monocytes in not just ovarian cancer but various other solid tumors as well. Use of Alemtuzumab could be restricted in heavily pretreated cancer patients as the primary side effect associated with Alemtuzumab therapy is immune-suppression. However, given the unique disposition of ovarian cancer to grow in a manner restricted to the peritoneal cavity it is possible that systemic side-effects could be minimized by intraperitoneal delivery of the drug.
Despite being a well-documented therapeutic target, the exact function of CD52 remains unknown. In the ID8-VEGF tumors, VLC account for the vast majority of tumor-associated host hematopoietic cells, thus the majority of the impact is likely attributable to an anti-VLC effect. Human tumors, in contrast, are significantly more complex. CD52 expression is observed on numerous tumor infiltrating host cells including lymphocytes, neutrophils, and mast cells. Therefore it is possible Alemtuzumab could have multiple different effects via this broad targeting. In addition to the expected effects on angiogenesis based on the elimination of VLC, Alemtuzumab may also inhibit angiogenesis via the elimination of B cells and mast cells from the tumor microenvironment; both of these cell types have also been implicated in promoting angiogenesis and tumor growth [
28‐
30].
Eliminating VLC may also impact anti-tumor immunity. Recent studies indicate that elimination of CD11c
+ cells, a population of cells that would include VLC, from the tumor microenvironment can actually enhance anti-tumor immunity [
31]. This is consistent with an immunosuppressive phenotype of VLC [
32]. Alemtuzumab could also promote anti-tumor immunity by eliminating regulatory T cells (T regs). T regs have been reported to accumulate in late stage ovarian tumors and to be a negative prognostic factor [
33]. In fact, in ovarian cancer T-regs may be induced by cancer-associated myeloid cells such as VLC [
34]. There is a potential detrimental immune-modulatory effect of Alemtuzumab via the elimination of anti-tumor T cells, or other inflammation mediated anti-tumor effects. However, at least in late stage tumors the impact of this anti-tumor immunity seems minimal.
Competing interests
The University of Michigan and RJB have submitted a patent regarding the use of Alemtuzumab as an anti-angiogenic agent in ovarian cancer. This was submitted after the completion of the described work.
Authors' contributions
HP: Performed experiments, wrote manuscript, GS: Performed experiments, IS: Performed experiments, KM: Performed experiments, AK: Contributed research material, RKR: Contributed research material, GC: Contributed research material, critical reading of manuscript, JCG: Contributed research material, critical reading of manuscript, RJB: Designed and performed experiments, wrote manuscript. All authors have read and approved the final manuscript.