Genetic and Pharmacologic Inhibition of Complement Impairs Endothelial Cell Function and Ablates Ovarian Cancer Neovascularization

Abstract

Complement activation plays a critical role in controlling inflammatory responses. To assess the role of complement during ovarian cancer progression, we crossed two strains of mice with genetic complement deficiencies with transgenic mice that develop epithelial ovarian cancer (TgMISIIR-TAg). TgMISIIR-TAg mice fully or partially deficient for complement factor 3 (C3) (Tg⁺C3^KO and Tg⁺C3^HET, respectively) or fully deficient for complement factor C5a receptor (C5aR) (Tg⁺C5aR^KO) develop either no ovarian tumors or tumors that were small and poorly vascularized compared to wild-type littermates (Tg⁺C3^WT, Tg⁺C5aR^WT). The percentage of tumor infiltrating immune cells in Tg⁺C3^HET tumors compared to Tg⁺C3^WT controls was either similar (macrophages, B cells, myeloid-derived suppressor cells), elevated (effector T cells), or decreased (regulatory T cells). Regardless of these ratios, cytokine production by immune cells taken from Tg⁺C3^HET tumors was reduced on stimulation compared to Tg⁺C3^WT controls. Interestingly, CD31⁺ endothelial cell (EC) function in angiogenesis was significantly impaired in both C3^KO and C5aR^KO mice. Further, using the C5aR antagonist PMX53, tube formation of ECs was shown to be C5a-dependent, possibly through interactions with the VEGF₁₆₅ but not VEGF₁₂₁ isoform. Finally, the mouse VEGF₁₆₄ transcript was underexpressed in C3^KO livers compare to C3^WT livers. Thus, we conclude that complement inhibition blocks tumor outgrowth by altering EC function and VEGF₁₆₅ expression.

Introduction

Tumor development is a multistep process of cumulative genetic alterations that lead to cell autonomy. Inflammatory mechanisms are thought to play a critical role in this process [1,2]. Lung, skin, gastrointestinal, liver, urinary, cervical, and ovarian cancers are all associated with chronic inflammation, and attenuating such inflammation has proved beneficial in the clinical setting [3–5]. Epithelial ovarian cancer (EOC), the fifth leading cause of cancer death among women in the United States, is also intimately related to inflammation. Incessant ovulation, a purported cause of malignant transformation in the ovarian surface epithelium, is associated with the activation of cytokine networks and repair mechanisms in ovarian stroma, whereas pelvic inflammatory conditions, including endometriosis, predict an increased risk of ovarian cancer [6,7]. An early increase in serum inflammatory cytokines is detected in ovarian cancer patients [8], and ovarian tumors and ascites are characterized by a brisk inflammatory milieu [9–11]. Finally, elevated levels of complement anaphylatoxins suggestive of local complement activation have been observed in ovarian cancer patients' ascites [12]. The complement system is comprised of serum proteins, membrane-bound receptors, and regulatory proteins [13,14]. Its effector functions in host defense and inflammation are mediated mainly through the sequential activation and proteolytic cleavage of a series of serum proteins. Complement activation occurs through three distinct activation routes, the alternative, classic, and lectin pathways, all of which converge at a critical step: the activation of complement factor 3 (C3) by C3 convertase-mediated cleavage [14]. Complement functions include pathogen opsonization, inflammation mediated by C3a and C5a complement anaphylatoxins, and cytolysis resulting from the assembly of the membrane attack complex on targeted cells. Eliminating C3 prevents complement cascade activation and the generation of complement effectors that mediate a wide array of functions [13–15]. We demonstrated a role for complement activation in promoting the growth of transplanted tumors through myeloid-derived suppressor cell (MDSC) recruitment and activation in mice [16], but the role of complement in early oncogenic events remains unknown.

Complement proteins are well established as important effectors in pathologic neovascularization in age-related macular degeneration (AMD [17]), diabetic retinopathy, and retinopathy of prematurity [18], as well as in the regulation of the angiogenic factors required for normal placental development [19,20]. AMD involves a process whereby inappropriate angiogenesis in the choroid causes vascular invasion into the adjacent retina (choroidal neovascularization); pre-clinical models of AMD directly link complement to this process as complement components C3a and C5a promote choroidal neovascularization [21] and C5a increases vascular endothelial growth factor (VEGF) secretion of human retinal pigment epithelial cells [17]. Neovascularization is also a critical contributor to solid tumor progression, including cases of ovarian cancer [22]. VEGF, first identified as a vascular permeability factor secreted by tumor cells [23], plays a principal role in angiogenesis by stimulating migration and proliferation of endothelial cells (ECs) and the expression of angiogenesis-related genes in ECs. Alternative splicing of the VEGF gene gives rise to multiple isoforms, including 121, 165, 189, and 206 amino acid long products that are differentially expressed in a variety of human tissues and tumors [24,25]. Individual VEGF isoforms may differentially contribute to tumor vascularization according to the gradient model of Grunstein et al. [26] with VEGF₁₂₀ (VEGF₁₂₁ mouse homologue) recruiting peripheral vessels, VEGF₁₈₈ (VEGF₁₈₉ mouse homologue) promoting tumor neo-angiogenesis, and VEGF₁₆₄ (VEGF₁₆₅ mouse homologue) affecting intermediate structures. The importance of VEGF in tumor-associated angiogenesis is well established: overexpression of membrane-bound VEGF₁₂₁ and VEGF₁₆₅ was linked to tumor-associated intracerebral hemorrhage [27–29]; in a breast cancer xenograft model, VEGF₁₂₁ was described as the most potent tumor angiogenic factor; and soluble VEGF₁₈₉ expression in human colon, renal, and lung cancers was strongly associated with increased microvessels, cancer metastasis, and poor prognosis [30,31]. Notably, high levels of VEGF₁₆₅ have been associated with reduced survival in ovarian cancer [32], and a phase 3 clinical trial showed that Bevacizumab, a humanized monoclonal antibody (mAb) that inhibits VEGF-A, improves progression-free survival in women with ovarian cancer [33]. Although both complement activation and neo-angiogenesis play critical roles for ovarian cancer development, direct effects of complement activation on tumor neo-angiogenesis remain poorly understood.

To address this gap in knowledge, we crossed C3-deficient (C3^KO) mice [34] or complement factor 5a receptor-deficient (C5aR^KO)mice [35] with C57BL/6 TgMISIIR-TAg transgenic mice that spontaneously develop EOC [36–38]. The immediate-early region of the SV40 virus that contains the oncogenic large and small T-antigen genes (TAg and tag, respectively) has been extensively used in the development of transgenic mouse models of cancer [39–42]. SV40-TAg models have revealed much about tumor biology, permitting significant advancements in the understanding of the “angiogenic switch” [43–45] and tumor progression and invasion [46]. Expression of TAg results in the functional inactivation of the critical tumor suppressor proteins p53 and Rb. In fact, mutation and/or loss of TP53 is the most frequent genetic alteration identified in EOC, particularly the serous subtype [47,48], and mutation and/or loss of Rb and its downstream signaling mediators is also common [36,37,49,50]. Consistent with this fact, the expression of SV40 under the transcriptional control of the Müllerian inhibiting substance type II receptor (MISIIR) gene promoter triggers spontaneous ovarian cancer development in transgenic mice. SV40 TAg-induced ovarian cancer closely resembles the histotype of serous EOC, the most common EOC histotype in humans [36]. We report that C3 and C5aR deficiencies resulted in profoundly impaired EOC growth, reduced tumor vascularization, and inhibition of specific VEGF isoforms, revealing a novel mechanism of VEGF regulation by complement anaphylatoxin C5a.

Materials and Methods

Mice

All mouse experiments were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania and conducted according to guidelines of the National Institutes of Health. C57BL/6 TgMISIIR-TAg transgenic mice [36], C3^KO (B6.129S4-C3^tm1Crr/J) mice [34], and C5aR^KO(C5ar1^tm1Cge/J) mice [35] have been described. Mice were housed in a barrier animal facility of the University of Pennsylvania on a 12-hour light/dark cycle. Water and standard rodent diet were provided ad libitum.

C57BL/6 TgMISIIR-TAg mice develop spontaneous bilateral serous ovarian tumors and have a mean life expectancy of 151.6 days (∼22 weeks) [36]. The ovarian cancer phenotype is transmitted as a dominant trait with 100% penetrance. TgMISIIR-TAg females are infertile; therefore, mice are propagated by crossing TgMISIIR-TAg males with wild-type (WT) littermate females.

Both C57BL/6 C3^KO (B6.129S4-C3^tm1Crr/J) and C5aR^KO (C5ar1^tm1Cge/J) mice are viable and fertile but have impaired immune responses.

To obtain TgMISIIR-TAg transgenic female mice deficient for C3 (Tg⁺C3^KO) or C5aR (Tg⁺C5aR^KO), we crossed TgMISIIR-TAg males with C3^KO or C5aR^KO females. TgMISIIR-TAg males heterozygous for C3 (Tg⁺C3^HET) or C5aR (Tg⁺C5aR^HET) were thencrossed with females homozygous for C3 or C5aR. The litters were sparse and small (less than or equal to five pups), consistent with an earlier report describing early pregnancy impairment in C3^KO mice [51]. Genotyping was performed by polymerase chain reaction (PCR) using previously described primers [35,36] (Figure W1, E and F).

Immunohistochemistry and Immunofluorescence

Ovaries and ovarian tumors were collected from 16-week-old mice and preserved in optimal cutting temperature (OCT) media embedding (Tissue-Tek; Sakura Finetek USA, Torrance, CA). Sections (8 µm) were cut, air-dried for 1 hour at room temperature, and fixed by immersion in cold 100% acetone for 5 minutes. After two washes in phosphate-buffered saline (PBS), slides were blocked with goat serum (Vector Labs, Burlingame, CA). ECs were stained with a rat anti-mouse CD31 mAb (BD Pharmingen, San Jose, CA) followed by biotinylated polyclonal goat anti-rat IgG (BD Pharmingen) and detected with the VectaStain ABC reagent and DAB (Vector Labs). Microvessels were defined as any dark-stained ECs or cell clusters clearly separated from adjacent structures, tumor cells, and other connective tissue elements. Micro-vascular density was expressed as the number of vessels per field of view. C3 deposition was detected with a rat anti-mouse C3b/iC3b/C3c mAb (2/11; HyCult) followed by Alexa 488-labeled goatanti-ratanti-IgG1 mAb (Invitrogen). The values reported reflect C3b/iC3b deposition per cell, averaged from five field-of-view tumors. Adjacent sections were stained with hematoxylin and eosin for histopathologic evaluation.

Cell Preparation and Antibodies

Ovaries and ovarian tumors from WT, TgMISIIR-TAg, Tg⁺C3^WT, Tg⁺C3^HET, Tg⁺C3^KO, Tg⁺C5aR^WT, Tg⁺C5aR^HET, and Tg⁺C5aR^KO mice were collected in ice-cold Dulbecco's modified Eagle's medium supplemented with 10% (vol/vol) FBS and 1% penicillin/streptomycin and mechanically dissociated using a cell strainer (BD Biosciences, San Jose, CA). Red blood cells were eliminated with ammonium chloride-potassium (ACK) lysing buffer (Gibco, Grand Island, NY). Freshly isolated single-cell (WT, Tg⁺C3^WT, Tg⁺C3^HET, and Tg⁺C3^KO) or frozen suspensions (95% FBS and 5% DMSO, stored at -80°C for up to 4 months; Tg⁺C5aR^WT, Tg⁺C5aR^HET, and Tg⁺C5aR^KO) were first incubated with rat anti-mouse CD16-CD32 (2.4G2) mAb to block Fc receptors, and mouse leukocyte labeling was performed by incubation with fluorochrome-conjugated rat anti-mouse CD3 (17A2), CD4 (L3T4), CD8 (53–6.7), CD11b (M1/70), CD45 (30-F11), CD45R (B220), CD19 (ID3), F480 (BM8), or Gr-1 (RB6-8C5) Abs, at 4°C for 30 minutes. Isotype-matched antibodies were used as negative controls. For intracellular analysis of cytokines, cells were permeabilized with Cytofix/Cytoperm and Perm/Wash buffers (BD Biosciences) and stained with fluorochrome-conjugated rat anti-mouse interleukin-10 (IL-10), IL-12 (p40/p70), interferon-gamma (IFN-γ), or FoxP3 (FJK-16a) Abs. All Abs were purchased from BD Biosciences.

For intracellular analysis of cytokines, single-cell suspensions of whole mouse tumor (5 x 10⁶) were first cultured for 24 hours at 37°C in 5% CO₂. Cells were then resuspended in fresh medium with 0.002% 2-β-mercaptoethanol (Sigma, St Louis, MO) for 4 hours in 5% CO2. In some cases, cells were stimulated with lipopolysaccharide (LPS; 1 µg/ml) and IFN-γ (10 ng/ml) plus monensin (4 µl/6 ml of cell culture). For T-cell activation, 20 µl of anti-CD3/anti-CD28 activation beads were added per milliliter of culture as recommended by the manufacturer (Miltenyi, Auburn, CA).

In Vivo Angiogenic Assays

Growth factor-reduced Matrigel was mixed at 4°C (liquid state) at a 1:1 ratio with 10 ng/ml VEGF-A human recombinant homodimer protein (Millipore, Billerica, MA) or PBS as negative control. C57BL/6 WT mice were injected with 100 µl of each mixture in the right and left flanks. Animals were sacrificed 2 weeks after injection. Matrigel plugs were excised and digested with 1 mg/ml collagenase II and 10,000 U/ml DNAse I (Sigma) for 30 minutes at room temperature, as described [52], then resuspended in PBS and analyzed by flow cytometry for the expression of CD31 and CD144.

Tube Formation

Tube formation assays measure the ability of ECs to form capillary-like tubes. Human mammary epithelial cells (HMECs; 2 x 10⁶) at passages 2 to 6 were cultured in a T75 cm² flask in HMEC medium [MDCB131 medium (Gibco) supplemented with 10% (vol/vol) FBS, 1% penicillin/streptomycin, 100 µg/ml endothelial cell growth supplement (ECGS), and heparin (Sigma)]. The following day, cells were starved for 1 hour in MDCB131 medium and then harvested, and 2x10⁵cells/well were incubated in Matrigel 24-well plates (BD) with 2 nM C5a native protein or human recombinant proteins VEGF-A (VEGF₁₆₅ homodimer) at 2 ng/ml, VEGF₁₆₅ single domain (35.0 MWt, Peprotech, Rocky Hill, NJ) at 2 ng/ml (57 nM) or 2.7 ng/ml (77 nM), or VEGF₁₂₁ single domain (25.4 MWt, Peprotech) at 2 ng/ml (78 nM), in the presence or absence of 100 x PMX53 (AcF[OPdChaWR]) or of a chemically similar control peptide (AcF[OPdChaA(d)R]) that lacks functional properties. Plates were incubated at 37°C for 4 hours and endothelial tube formation was examined using a light microscope (Nikon, Melville, NY) at x10 magnification. Tube formation per area was analyzed using ImagePro software (Media Cybernetics, Rockville, MD). Experiments were repeated three times.

VEGF Isoform-specific Reverse Transcription-PCR

RNAs were extracted using Trizol reagent from 2 x 10⁵ HMECs or from single-cell suspension of mouse organs. HMECs were cultured at 2 x 10⁵ cells/well in HMEC medium alone or supplemented with C5a native protein (10,400 MWt, CompTech, Tyler, TX) at 2 nM (20.8 ng/ml), and/or VEGF-A (1 ng/ml), with or without PMX53 C5a inhibitor or PMX53 control peptide in 12-well plates for 24 hours. Mouse spleen, liver, and bone marrow tissues were collected from C57BL/6, C3^KO, or C5aR^KO mice. mRNAs were reverse transcribed using M-MLV Reverse Transcriptase (Promega, Madison, WI). Human and mouse VEGFs, human β₂-microglobulin, and mouse glyceraldehyde-3-phosphate dehydrogenase (GAPDH) primers and PCR conditions were used as described [53]. Densitometry analysis was performed using ImageJ software (Image processing and analysis in Java) freely available from the National Institutes of Health at http://rsb.info.nih.gov/ij. Experiments were repeated three times.

Statistical Analyses

The data from experiments with two groups were analyzed using Student's t test and data from experiments with more than two groups were analyzed using one-way analysis of variance (ANOVA) using Satterthwaite and Welch adjustments for unequal variances, respectively. RNA expression and percentages of cells by type were transformed using the natural logarithm before their analysis. When the ANOVA was significant, pairwise comparisons were evaluated using Tukey procedure. Statistical significance was reached at P < .05. Analyses were performed using SAS software, version 9.2 (SAS Institute, Inc, Wayne, PA).

Results

Genetic C3 Deficiency Impairs Ovarian Tumor Development and Growth

In a transplanted tumor model, we have previously characterized the spontaneous activation of complement in the tumor vasculature and stroma, as evidenced by tumor-specific deposition of the C3 cleavage products C3b/iC3b in tumor-bearing mice [16]. To determine whether complement also plays a role during early spontaneous tumor formation, we crossed C3^KO mice with TgMISIIR-TAg transgenic mice to obtain Tg⁺C3^KO mice. Although tumor development in the TgMISIIR-TAg model is stochastic, longitudinal in vivo imaging and quantitative tumor growth analyses demonstrated that TgMISIIR-TAg ovarian tumors are apparent by immunohistochemical staining as early as 4 weeks; ovaries exhibit enlargement by 13 weeks with a doubling time of 7.3 days [37,38], and by 16 weeks, all mice develop solid tumors that consistently reach 0.7 ± 0.2 cm³ in volume. TgMISIIR-TAg tumors are frequently accompanied by peritoneal dissemination and ascites, and the average life span of these mice is ∼22 weeks [37]. Given the central role of C3 in the complement cascade, we hypothesized that Tg⁺C3^KO mice would be unable to generate complement effectors downstream of C3.

We compared tumor development of 16-week-old female littermates heterozygous for the TAg transgene and WT (Tg⁺C3^WT), heterozygous (Tg⁺C3^HET), or deficient (Tg⁺C3^KO) for C3. All Tg⁺C3^WT mice developed bilateral ovarian tumors. In contrast, 9 of 12 ovaries of Tg⁺C3^KO mice were histologically normal (Figure 1A) with no visible microscopic tumors (Figure W1). In addition, tumors that developed in Tg⁺C3^KO and Tg⁺C3^HET mice were significantly smaller than Tg⁺C3^WT tumors (both P ≤ .001), while Tg⁺C3^KO and Tg⁺C3^HET tumor volumes were not significantly different from each other (P = .279) (Figure 1A). Tg⁺C3^KO mice were unable to produce the convertases required for the generation of complement effectors downstream of C3 because of complement deficiency, and the deposition of iC3b was attenuated in heterozygous mice (Tg⁺C3^HET) when compared to littermates with intact complement (Tg⁺C3^WT) (Figure 1, B and C). These results indicate that complement activation is critical for ovarian tumor establishment and growth.

Figure 1.

Genetic complement deficiency impairs SV40⁺ ovarian tumor growth and development. (A) Tumors were collected from 16-week-old Tg⁺C3^WT, Tg⁺C3^HET, and Tg⁺C3^KO mice and measured with calipers. Volumes were calculated using the mathematical formula for the volume of an ellipsoid: V = 4/3(3.1416) abc, where a, b, and c represent the semi-axes of each diameter. P < .05. (B) Confocal microscopy of frozen tumor sections labeled with anti-C3b/iC3b/C3c antibody and 4′,6′-diamino-2-phenylindole (DAPI). Original magnification, x23. Results are representative of four Tg⁺C3^WT tumors, four Tg⁺C3^HET tumors, and two Tg⁺C3^KO tumors. (C) Quantification of C3b/iC3b/C3c deposition on tumor cells from 16-week-old Tg⁺C3^WT, Tg⁺C3^HET, and Tg⁺C3^KO mice (n/group = 5). Bars represent means ± standard error. (D) Cell composition of ovarian tumors in Tg⁺C3^WT (n = 3), Tg⁺C3^HET (n = 5), and Tg⁺C3^KO (n = 3) mice by flow cytometry. Tumor cells were identified by staining with anti-SV40 Ab and tumor-infiltrating leukocytes with anti-CD45 Ab.

C3 Deficiency Is Associated with Dampened Cellular Effector Mechanisms

The composition of tumor-infiltrating leukocytes was analyzed as a possible explanation for decreased or absent tumor growth in Tg⁺C3^HET and Tg⁺C3^KO mice. No difference in the percentages of tumor-infiltrating leukocytes were observed across the groups (Figure 1D, P = .355). Due to the fact that Tg⁺C3^KO mice bore no, or only very small, ovarian tumors with a sparse amount of tumor cell infiltrates (Figure 1D), we focused the rest of our studies on Tg⁺C3^WT and Tg⁺C3^HET tumor-infiltrating leukocytes. No significant differences in the frequency of tumor-infiltrating macrophages, B cells, and MDSCs were found between Tg⁺C3^WT and Tg⁺C3^HET groups at 16 weeks of age. However, Tg⁺C3^HET tumor infiltrates contained more CD8⁺T cells (P < .0001) and less FoxP3⁺CD4⁺ T cells (P =.029) than Tg⁺C3^WT tumor infiltrates (Figure 2A).

Figure 2.

Functional impairment of tumor-immune infiltrate in Tg⁺C3^HET mice. (A) Ovarian tumors from 16-week-old Tg⁺C3^WT mice (black bars; n = 4) and Tg⁺C3^HET mice (gray bars; n = 4) were dissociated and stained for CD45, CD11b, B220, F480, GR1, CD3, CD8, CD4, CD25, and FoxP3. CD45⁺ leukocytes were gated and macrophages were identified as CD11b⁺F480⁺, B cells as CD11b^-B220⁺ cells, MDSCs as CD11b⁺GR1⁺, CD8+ T cells as CD3⁺CD8⁺ cells, and regulatory T cells as CD4⁺CD25⁺FoxP3⁺. Bars represent percentages of CD45⁺ cells ± standard error. (B-E) Ovarian tumors were collected from 16-week-old Tg⁺C3^WT (n = 4) and Tg⁺C3^HET (n = 4) mice, mechanically dissociated to a single-cell suspension and incubated 4 hours in medium only (white bars), or activated for 4 hours with LPS and IFN-γ (black bars), or with anti-CD3 and anti-CD28 mAbs (gray bars). Cells were then stained for CD45, F480, and CD11b (macrophages) (B, C), CD45 and B220 (B cells) (D), or CD45, CD3, and CD8 (T cells) (E), washed and permeabilized for intracellular staining for IL-12 (B), IL-10 (C, D), or IFN-γ (E). Bars represent mean percentages ± standard errors.

Macrophages and T and B cells from Tg⁺C3^HET and Tg⁺C3^WT littermates were then functionally characterized by flow cytometry. In the absence of stimulation (Figure 2, B–E, white bars), no difference were noticed among the two groups in the percentages of macrophages expressing IL-12 (Figure 2B, P = .948) or IL-10 (Figure 2C, P = .447), of B cells expressing IL-10 (Figure 2D, P = .092), or of CD8⁺ T cells expressing IFN-γ (Figure 2E, P = .765). However, after a short stimulation with LPS and IFN-γ (Figure 2, B–D, black bars) or with anti-CD3 and anti-CD28 (Figure 2E, gray bars), the percentages of macrophages producing IL-12 (Figure 2B, P = .034) or IL-10 (Figure 2C, P = .001), of B cells producing IL-10 (Figure 2D, P = .0043), or of T cells producing IFN-γ (Figure 2E, P = .0038) were significantly lower in Tg⁺C3^HET mice than in Tg⁺C3^WT mice. We thus concluded that C3 deficiency was associated with dampened cellular effector mechanisms in the ovarian tumor microenvironment. This suggested that alternate mechanisms besides cell-mediated immune rejection were likely mainly responsible for the decrease in tumor incidence and growth observed in Tg⁺C3^KO mice.

Genetic C3 Deficiency Impairs Tumor Vascularization and In Vivo EC Function

In preclinical models of AMD, deficiencies of C3a or C5a receptors result in decreased choroidal neovascularization and VEGF levels. Similar observations were reported when WT mice were treated with C3aR and C5aR antagonists, or neutralizing antibodies against C3a and C5a [21], suggesting an important role for complement in modulating vascularization. In humans, the number of chronically activated tumor-infiltrating innate immune cells is positively correlated with blood vessel density [54], while the attenuation of innate immunity in premalignant tissues reduces angiogenesis and limits tumor development in murine models [55]. We therefore asked whether defective angiogenesis was responsible for the reduced tumor formation in our model. Tumor microvascular density (MVD) was assessed by CD31 staining and quantification of vessel number in Tg⁺C3^HET and Tg⁺C3^WT littermates. MVD was significantly decreased in tumors from Tg⁺C3^HET littermates (Figure 3A, P =.023), suggesting a functional role of complement for ECs. To address whether genetic complement deficiency could alter in vivo angiogenesis, we compared the in vivo colonization of Matrigel plugs admixed with VEGF in WT and C3^KO mice. The numbers of total cells infiltrating VEGF-Matrigel plugs after 2 weeks were significantly reduced in C3^KO mice compared to WT mice (Figure 3B, P = .0012), as was the number of CD31⁺ ECs (Figure 3C, P < .0001). However, the ratios of CD31⁺ ECs versus total cells were not significantly different in WT and C3_KO mouse VEGF-Matrigel plugs (Figure 3D, P = .149). These results supported the hypothesis that genetic C3 heterogeneity or deficiency impairs EC effector mechanisms in our model.

Figure 3.

Genetic complement deficiency impairs neo-angiogenesis. (A) Tumor microvasculature was evaluated by CD31 staining (dark blue staining) of frozen tumor sections from 16-week-old Tg⁺C3^WT and Tg⁺C3^HET mice. Values reflect the number of CD31⁺ vessels averaged from five high-power fields per slide per mouse. Original magnification, x10. Bars represent means of CD31⁺ microvessels ± standard error. (B-D) Bars represent the absolute number of total cells (B), or CD31⁺ ECs (C), or the ratios of CD31⁺ ECs and total cells (D) attracted to Matrigel plugs admixed with recombinant VEGF (10 ng/ml) and implanted in C57BL/6 mice (WT, gray bar, n/group = 5) or C3^KO mice (black bar, n/group = 5) ± standard error.

Pharmacological Inhibition of Complement Impairs VEGF₁₆₅-dependent Function of Human ECs

To assess whether our in vivo findings were relevant to human physiology, we sought to study the effects of complement pharmacological inhibition on the human EC line HMEC. The C5aR is expressed on ECs and upregulated during inflammation [56,57]. C5aR function can be inhibited by PMX53, a peptide that has been used in preclinical models to block neutrophil chemotaxis and clinically for treatment of rheumatoid arthritis and psoriasis [58]. We studied the effects of PMX53-mediated inhibition of C5aR on HMEC tube formation stimulated by VEGF-A (Figure 4, A1 and B1, white bar) and the commercially available isoforms VEGF₁₆₅ (Figure 4, A2 and B2, white bar) and VEGF₁₂₁ (Figure 4, A3 and B3, white bar) compared to a control peptide (Figure 4, A11–15 and B, gray bars). PMX53 (Figure 4B, black bars) impaired VEGF₁₆₅-mediated tube formation (Figure 4, A7, B2, and C) but not VEGF₁₂₁-mediated tube formation (Figure 4, A8, B3, and C, ANOVA P > .05). PMX53 also impaired HMEC tube formation mediated by equimolar concentrations of VEGF₁₆₅ and VEGF₁₂₁ (Figure 4C, P = .0232). As expected, treatment with control peptide did not affect tube formation. Of note, the difference observed between ECs activated by VEGF-A and treated with PMX53 was statistically significant but not meaningful (Figure 4, A6 and B1, P = .0127 for VEGF-A vs VEGF-A/PMX53 but P = .9368 for VEGF-A/PMX53 vs VEGF-A/control peptide). These results suggested that C5aR inhibition selectively affected tube formation mediated by the VEGF₁₆₅ isoform.

Figure 4.

Pharmacological inhibition of C5aR targets the VEGF₁₆₅ isoform. (A) Representative pictures of tube formation. Human micro-vascular ECs (2 x 10⁵) were seeded onto Matrigel and cultured in MCDB-131 culture medium supplemented with ECGS/heparin and 2 ng/ml of either VEGF-A (1, 6, 11), VEGF₁₆₅ (2, 7, 12), or VEGF₁₂₁ (3, 8, 13), or 2 nM C5a (4, 9, 14) in the presence of 200 ng/ml PMX53 (6–10) or PMX53 control peptide (Ctl peptide, 11–15). As negative controls, cells were cultured with medium supplemented with ECGS/heparin only (5, 10, 15). Tube formation was assayed after 4 hours by phase contrast microscopy and analyzed by measuring tube length per area using the ImagePro software. (B, C) Measures of tube formation of four fields per treatment (B) with 2 ng/ml VEGF-A, VEGF₁₆₅, or VEGF₁₂₁ or (C) with equimolar concentrations of VEGF₁₆₅ (2.7 ng/ml) and/or VEGF₁₂₁ (2 ng/ml). Bars represent means ± standard error from two independent experiments. (D) Representative picture of PCR amplification of VEGF isoform in the presence or absence of C5a, VEGF-A, and/or PMX53 (as indicated). PCR amplification of β₂-microglobulin was used as a control. (E, F) Densitometric analysis of transcript expression of VEFG165 (E) and VEFG₁₂₁ (F) isoforms in HMECs treated with 2 nM C5a (black bars) or 2 ng/ml VEGF-A (gray bars) in medium or in the presence of 200 ng/ml PMX53 or control peptide (as indicated). VEGF isoform values were normalized to those of β₂-microglobulin. Bars represent mean ± standard error from five independent experiments.

C5a Promotes Tube Formation in a VEGF₁₆₅-dependent Manner

To determine whether C5a had direct functional effects on ECs, HMECs were incubated with native C5a protein at 2 nM (20.8 ng/ml), which is within the plasma concentration range of C5a degradation product C5adesArg [59]. Tube formation similar to that mediated by VEGF-A and VEGF isoforms was observed (Figure 4, A4 and B4, white bar), while PMX53 inhibited C5a-mediated tube formation (Figure 4, A9 and B4, black bar, P < .0001). In addition, we found that PMX53 specifically inhibited the transcription of VEGF₁₆₅ (Figure 4, D–E, P = .050) but not of VEGF₁₂₁ in C5a-treated HMEC cells (Figure 4, D and F, P = .202). These results suggested a novel role for C5a in promoting EC tube formation through enhancing expression of VEGF₁₆₅.

Genetic C3 Deficiency Affects the Expression of VEGF₁₆₄ Isoform

As experimental evidence pointed toward a role for C5a in VEGF₁₆₅-mediated activation of human ECs, we performed a comparative analysis of the expression of VEGF isoforms in mice that are complement deficient or complement sufficient. We analyzed the expression of the mouse homologues of human VEGF₁₂₁ and VEGF₁₆₅ that are mouse isoforms VEGF₁₂₀ and VEGF₁₆₄, respectively, in the spleens, livers, and bone marrows isolated from WT and C3^KO mice. Figure 5 represents the relative expression of VEGF isoforms in WT and C3^KO mice after normalization against GAPDH. The expression of VEGF₁₆₄, but not of VEGF120, was significantly reduced in C3^KO livers compared to WT livers (Figure 5A, P = .048 and Figure 5B, P =.404).

Figure 5.

Altered expression of VEGF164 isoform in complement-deficient mice. Densitometry analysis of expression of mouse VEFG₁₆₄ (A) and VEFG₁₂₀ (B) isoforms in WT versus C3^KO spleens, livers, and bone marrows (as indicated). VEGF isoform values were normalized to those of GAPDH. Bars display mean ± standard error from three different mice per group.

Genetic C5aR Deficiency Impairs Tumor Growth and Neo-angiogenesis but Not Tumor Immune Infiltrate

To further assess the role of the C5a complement anaphylatoxin in ovarian cancer progression and angiogenesis, we crossed C5aR^KO mice [35] with C57BL/6 TgMISIIR-TAg transgenic mice. We compared tumor development in 16-week-old female littermates heterozygous for the TAg transgene and C5aR sufficient (Tg⁺C5aR^WT), C5aR heterozygous (Tg⁺C5aR^HET), or C5aR deficient (Tg⁺C5aR^KO). While Tg⁺C5aR^WT and Tg⁺C5aR^HET mice developed unilateral ovarian tumors, three of six ovaries of Tg⁺C5aR^KO mice were histologically normal and Tg⁺C5aR^KO tumors were significantly smaller than Tg⁺C5aR^WT and Tg⁺C5aR^HET tumors (Figure 6A, P = .001 and P = .002, respectively). In contrast with C3 deficiency, no difference in tumor sizes was observed between Tg⁺C5aR^WT and Tg⁺C5aR^HET mice (Figure 6A, P = .945), and no significant differences were found across the groups regarding tumor immune infiltrate (Figure 6B, all ANOVA P > .05). However, as seen in Tg⁺C3^HET mice, tumor MVD was significantly decreased in tumors from Tg⁺C5aR^HET mice as well as Tg⁺C5aR^KO mice compared to Tg⁺C5aR^WT littermates (Figure 6C, P =.0004 and P < .0001, respectively). Additionally, MVD in tumors from Tg⁺C5aR^KO mice was significantly decreased compared to Tg⁺C5aR^HET mice (Figure 6C, P < .0001). Finally, in vivo assays showed significantly impaired angiogenesis in C5aR^KO mice compared to WT mice (Figure 6D, P = .019). Collectively, these results supported the hypothesis that the tumor growth inhibition observed in mice with genetic complement deficiencies was associated with the alteration of EC function.

Figure 6.

Genetic C5aR deficiency also impairs ovarian cancer growth and vascularization but not tumor immune infiltrate. (A) Tumors were collected from 16-week-old Tg⁺C5aR^WT (n = 4), Tg⁺C5aR^HET (n = 4), and Tg⁺C5aR^KO mice (n = 3) and measured with calipers. Volumes were calculated using the mathematical formula for the volume of an ellipsoid: V = 4/3(3.1416) abc, where a, b, and c represent the semi-axes of each diameter. (B) Comparative composition of tumor-infiltrating leukocyte function in Tg⁺C5aR^WT, Tg⁺C5aR^HET, and Tg⁺C5aR^KO mice. Ovarian tumors from 16-week-old Tg⁺C5aR^WT mice (black bars, n = 3), Tg⁺ C5aR^HET mice (gray bars, n = 3), and Tg⁺C5aR^KO mice (white bars, n = 3) were dissociated and stained as described in Figure 2A. Bars represent mean percentages ± standard errors. (C) Tumor microvasculature was evaluated by CD31 staining and immunohistochemistry (IHC) (dark brown staining) of tumors from 16-week-old Tg⁺C5aR^WT, Tg⁺C5aR^HET, and Tg⁺C5aR^KO mice. Values reflect the number of CD31⁺vessels averaged from five high-power fields per slide per mouse. Original magnification, x10. Bars represent means of CD31⁺ microvessels ± standard errors from at least four tumors per group. (D) Bars represent CD31⁺ ECs attracted to Matrigel plugs admixed with VEGF (10 ng/ml) in C57BL/6 mice (WT) (n/group = 5) or C5aR^KO mice (n/group = 5) ± standard error.

Discussion

Cancer establishment and progression is a multistep process of cumulative genetic alterations that enable cells to acquire uncontrolled growth. It has recently become clear, however, that to initiate cancer growth, genetically induced cell autonomy is often not sufficient, and an inflammatory reaction of the underlying stroma is required. The idea that chronic inflammation is intimately related to malignant transformation and tumor progression has been substantiated by human epidemiologic data and genetic experiments in mice, but the molecular and cellular determinants of these processes have not been well characterized. Here, we showed that deficiency of C3 or C5aR dramatically attenuates the tumor phenotype, abrogating tumor establishment and progression. Thus, complement activation is a critical component of the inflammatory process that supports oncogene-driven carcinogenesis. However, complement inactivation did not reduce the inflammatory infiltrate in tumors; we found no difference in the percentage of tumor-infiltrating leukocytes in Tg⁺C3^HET when compared to Tg⁺C3^WT, with the exception of CD8⁺ T cells and CD4⁺FoxP3⁺ T cells. This suggests that, contrary to the acute inflammation process in which complement anaphylatoxins likely play a major role in orchestrating the inflammatory infiltrate, other factors, such as chemokines produced by tumor and stromal cells, recruit copious amounts of leukocytes to tumors, even in the absence of complement activation. The presence of more tumor-infiltrating effector T cells and less regulatory T cells in Tg⁺C3^HET compared to Tg⁺C3^WT was in agreement with our previous findings in a mouse model of subcutaneous TC-1 cervical cancer, in which impaired growth of tumor cells was found when C5aR was blocked by an antagonist peptide [16]; this effect was attributed to enhanced CD8⁺ T-cell infiltration in tumors and reduced recruitment of MDSCs [16]. Furthermore, we found no evidence for immune cell activation in Tg⁺C3^HET tumors but rather a general suppression of leukocyte function, including a global reduction of cytokine production in Tg⁺C3^HET mice when compared to those fully sufficient in C3. These findings thus showed only indirect evidence of immune-mediated tumor rejection in these mice, through the decrease of immune regulatory cells or cytokines. Likewise, no evidence of immune-mediated tumor rejection was found in C5aR-hemizygous or C5aR-deficient mice, despite the dramatic attenuation of the tumor phenotype. The lower percentage of tumor-infiltrating CD4⁺FoxP3⁺ T cells in Tg⁺C3^HET tumors could explain why the Tg⁺C3^HET mice, but not the Tg⁺C5aR^HET mice, developed smaller tumors than their WT littermates. However, this finding was not sufficient to explain the dramatic phenotype of Tg⁺C5aR^KO mice where no change in tumor-infiltrating CD4⁺FoxP3⁺ T cells was found. Thus, while our data support the notion that complement activation plays a critical role in activating tumor-infiltrating leukocytes, a process that is apparently essential to their tumor-promoting function, our data also indicate that increased activation of effector cells is not the major mechanism leading to suppressed ovarian oncogenesis in C3-hemizygous and C5aR-hemizygous or C5aR-deficient mice, and there is likely an alternate mechanism of tumor inhibition.

Because tumor-induced angiogenesis is largely dependent on VEGF, VEGF-neutralizing drugs are actively sought to inhibit cancer growth. Current VEGF-neutralizing agents inhibit tumor angiogenesis by blocking activation of VEGF receptors [60], but agents have already been developed to selectively target VEGF isoforms. For example, Pegaptanib targets VEGF₁₆₅ for the treatment of AMD [61], and the specific targeting of VEGF isoforms has been proposed as an effective cancer therapy [62]. We show here that the complement anaphylatoxin C5a present in the inflammatory milieu activates ECs and that the pharmacological inhibition of C5aR specifically targets the VEGF₁₆₅ isoform. Our results reveal a previously unknown role for complement activation in ovarian tumor vascularization and growth. However, these data should be considered within the context of our ovarian tumor model. While another study has shown that lack of C3 results in increased angiogenesis in a model of retinopathy of prematurity [63], here it was found that C5a stimulation of ECs mediates angiogenic activity. Thus, the specific pathology may determine how complement influences different cell types, as well as the outcomes of that influence. The fact that small, unilateral ovarian tumors were present in 3 of 12 Tg⁺C3^KO mouse ovaries might be due to phenotype variability or incomplete penetrance not related to variations in genetic background or levels in gene expression [64,65]. Because C3^KO mice produce C3 transcript and some of it can be translated into an inactive pro-form of C3 protein, one cannot eliminate the possibility that some active C3 can be produced in the Tg⁺C3^KO mice through alternate splicing or processing. Alternatively, some unrelated factors could have also been differentially activated in these mice, such as the tumor microenvironment, or a different microbiome that could affect factors related to tumor growth, or even epigenetics. Future studies are needed to determine the molecular mechanisms underlying Tg⁺C3^KO phenotype variability.

In conclusion, our findings have important implications for our understanding of the mechanisms underlying ovarian cancer development and for the design of future preventive and therapeutic strategies, such as combination therapies associating complement inhibitors with targeted anti-angiogenic therapeutic agents for the treatment of ovarian cancer.

Abbreviations

EOC

epithelial ovarian cancer

C3

complement factor 3

MDSC

myeloid-derived suppressor cell

AMD

age-related macular degeneration

ROP

retinopathy of prematurity

VEGF

vascular endothelial growth factor

ECs

endothelial cells

MISIIR

Müllerian inhibiting substance type II receptor

C5aR

complement factor 5a receptor

WT

wild type

C3^KO

C3 deficient

C5aR^KO

C5aR deficient

Tg⁺

TgMISIIR-Tag transgenic mice

Tg⁺C3^WT

TgMISIIR-Tag mice wild-type for C3

Tg⁺C5aR^WT

TgMISIIR-Tag mice wild-type for C5aR

Tg⁺C3^KO

TgMISIIR-Tag mice deficient for C3

Tg⁺C5aR^KO

TgMISIIR-Tag mice deficient for C5aR

Tg⁺C3^HET

TgMISIIR-Tag mice heterozygote for C3

Tg⁺C5aR^HET

TgMISIIR-Tag mice heterozygote for C5aR

mAb

monoclonal antibody

FOV

field of view

HMECs

human mammary epithelial cells

ECGS

endothelial cell growth supplement

GAPDH

glyceraldehyde-3-phosphate dehydrogenase