Vascular Endothelial Growth Factor Secretion by Tumor-infiltrating Macrophages Essentially Supports Tumor Angiogenesis, and IgG Immune Complexes Potentiate the Process

Plasma concentrations of VEGF² are augmented in cancer patients in correlation with the extent of the disease (1, 2, 3, 4), which suggests that the source of this factor is the tumor tissue. Furthermore, experimental tumor models produced by injecting genetically manipulated tumor cells directly into the tissues have also suggested that tumor cells are the main source of VEGF (5, 6, 7). It is a common consensus that tumor cells are capable of producing VEGF that autostimulates neoangiogenesis of neoplastic tissue and tumor growth (8). However, similar models of in vivo tumor growth with no genetically manipulated tumor cells have shown that if host fibroblasts do not produce VEGF, tumor growth is impaired (9), suggesting that tumor cell-derived VEGF is important but not sufficient to automaintain tumor growth. Detailed pathology studies of tumors reveal that the source of VEGF is not exclusively tumor cells; fibroblasts also show important expression of this angiogenic factor (10). In glioblastomas, normal astrocytes located outside the tumor are the most important source of VEGF (11). It was suggested that the stimulant of VEGF production in tumors should be hypoxia (12), which can be produced when there is an overwhelming growth of the tumor tissue. However, it has been observed that hypoxia is not a key stimulant of VEGF production in the tumor (13). Micrometastatic growth, for example, can be inhibited by antiangiogenic factors such as angiostatin (14), but micrometastases do not have enough mass to justify hypoxia and are normally established surrounding capillaries (15). These observations, added to progressive analysis of tumor angiogenesis, are generating questions about the actual contribution of tumor tissue cellular components to tumor behavior (16, 17, 18).

Cellular sources of VEGF, other than tumor cells, are multiple. Most of these sources are involved in normal adult tissue repair and remodeling processes. Mast cells and muscle cells are an important source of VEGF (19, 20). Fibroblasts, PMN cells, and monocytes (MOs) are sources of VEGF in the wound-healing process (21, 22). Multiple factors, including hypoxia, nitric oxide, and inflammatory cytokines such as MCP-1 (23), interleukin 1, and interleukin 6, are involved in controlling VEGF expression in normal tissue repair and remodeling events (24, 25, 26).

Angiogenesis inhibition is a target for anticancer therapy. In this scenario, the definitions of the actual cellular producers of VEGF in quantities sufficient to promote tumor growth and of the molecular mechanisms involved in stimulating those cells to produce VEGF are fundamental.

This study is an analysis of the importance of MOs infiltrating the neoplastic tissue as a source of VEGF within the tumor structure and the importance of the antitumor immune response in sustaining MO contribution to tumor neoangiogenesis.

Normal C3H, nu/nu beige mice, and SCID beige mice were purchased from Harlan Sprague Dawley (Indianapolis, IN). Female 5–6-week-old mice were used for all experiments. The mice were maintained in a barrier-type facility approved by the American Association for Accreditation of Laboratory Animal Care and in accordance with current regulations and standards of the United States Department of Agriculture, Department of Health and Human Services, and NIH. STn-positive tumor cell lines T47D (human ductal breast carcinoma) and SW620 (human colon carcinoma) were obtained from ATCC (Manassas, VA), and the Met129 cell line (C3H murine methylcholanthrene-induced mammary carcinoma) was kindly provided by Dr. J. Vaage (Roswell Park Memorial Institute, Buffalo, NY). The hybridoma cell line B72.3 (derived from P3X63Ag8 plasmocytoma cells producing anti-sTn IgG1) was used as a source of anti-sTn IgG1. All cell lines were maintained in medium supplemented with 10% fetal bovine serum. MH-S mouse alveolar MOs were obtained from ATCC. PMN cells were obtained from peripheral blood of SCID beige mice at a final concentration of 1.5 × 10⁶ cells/ml.

The ELISA tests for sTn and anti-sTn IgG were made in the laboratory by combining mouse irrelevant IgG1 (Pierce, Rockford, IL), peroxidase-conjugated antimouse IgM and IgG (Sigma, St. Louis, MO), anti-sTn MAbs HB-sTn (DAKO, Carpinteria, CA), and peroxidase-conjugated B72.3. Purified B72.3 anti-sTn Ab was produced from the hybridoma cell line (ATCC). Peroxidase-conjugated B72.3 was synthesized with a Pierce kit, according to the manufacturer’s instructions. The ELISA kits for m-VEGF and h-VEGF were purchased from R&D Systems (Minneapolis, MN).

Intercellular interaction studies were performed in 24-well plates (Becton Dickinson, San Jose, CA). For coculture experiments, the concentrations of cells and test products were adjusted as follows: MH-S MO was diluted to a final concentration of 1.5 × 10⁶ cells/ml, equal amounts of PMN cells were adjusted (tested by Giemsa staining) and resuspended in RPMI 1640 to a final concentration of 10⁶ cells/ml, and tumor cells (T47D or SW620) were diluted to a final concentration of 1.5 × 10⁵ cells/ml. The quantity of the different additives used in the tests was as follows: (a) B72.3 anti-sTn Ab (purified from B72.3 hybridoma cell culture media) at 0.06 μg/well; (b) BSM (Sigma), a polyvalent sTn Ag, at 0.75 ng/well; (c) monovalent sTn Ag (sTn epitope; Calbiochem, San Diego, CA) at 30 ng/well; and (d) latex beads (Sigma) at 3 × 10⁶/well.

Blood samples were obtained from the tail vein (20 μl/sample). The cells were sedimented by centrifugation, and the serum was separated. The serum was diluted 1:50 in PBS containing 0.05% Tween. ELISA tests, using plates coated with sTn Ag (BSM) or anti-sTn Ab, were used to measure the serum content of specific anti-sTn IgG, anti-sTn IgM, and sTn Ag.

VEGF transcription modification by cellular interactions (direct contact or through soluble factors) was studied in cell cultures performed on polystyrene films using OptiCell (BioCrystal, Westerville, OH). To test the VEGF production of cocultured cells, tumor cells and MOs were seeded as follows: (a) 5 × 10⁵ T47D and 5 × 10⁵ MH-S cells on the same film; and (b) 5 × 10⁵ Met129 cells on one film and 5 × 10⁵ MH-S cells on the opposite film. The cells were cultured for 72 h in RPMI 1640 supplemented with 10% fetal bovine serum. After culture, the films were separated, and 1-cm² fragments were cut from each film and immunostained. The cell composition on each film was tested by fluorescence microscopy staining with PE-conjugated antimouse CD11b Ab (PharMingen, San Diego, CA) as a marker for MS-H cells. Total RNA from each individual film was extracted using 1 ml of Trizol reagent (Life Technologies, Inc., Carlsbad, CA). Each total RNA (800 ng) was reverse transcribed using the SuperScript II Pre-Amplification System (Life Technologies, Inc.). PCR primers for m-VEGF were constructed by Sigma Genosys (The Woodlands, TX). Primer sequences are as follows: (a) VEGF-A, 5′-TCT-CTT-GGG-TGC-ACT-GGA-C and 3′-TTT-AAC-TCA-AGC-TGC-CTC-GC; (b) VEGF-B, 5′-CCA-GCC-ACC-AGA-AGA-AAG and 3′-GAG-TGG-GAT-GGA-TGA-TGT-C; (c) VEGF-C, 5′-CTG-TGC-TTC-TTG-TCT-CTG-G and 3′-CAC-ATC-TAT-ACA-CAC-CTC-ACG; and (d) VEGF-D, GAA-GAA-TGG-CAG-AGG-ACC and 3′-CAT-GGT-GCT-TTA-CAG-ACG. G3PDH was used as an internal control. G3PDH primers were purchased from Clontech (Palo Alto, CA). The PCR cycling protocols for VEGF and G3PDH control were as follows: 35 cycles of 7 min at 95°C, 45 s at 94°C, 45 s at 55°C, 2 min at 72°C, 10 min at 72°C, and hold at 4°C.

Tumors were produced by s.c. injection of 10⁶ tumor cells, suspended in 100 μl of PBS. Tumor growth was monitored by daily external measurement of two crossed diameters with a caliper. Samples of blood were obtained 7 days after tumor implant and after sacrifice. The serum contents of anti-sTn IgG1, anti-sTn IgM, and free sTn Ag were measured by the use of ELISA. When one diameter of the tumor reached 25 mm, the mice were sacrificed by CO₂ asphyxiation. Liver, spleen, and lungs were then removed for study. Tumors were harvested and divided into halves: one half was snap-frozen in liquid nitrogen; and the other half was fixed in formalin and embedded in paraffin. The tissue content of anti-sTn IgG1 and the free sTn Ag was analyzed in extracts of homogenized tissue by ELISA. The tissue distributions of IgG1, IgM, and sTn Ag were assessed by immunohistochemistry (immunoperoxidase). The number of metastases was counted using microscopy. Histology was assessed using standard H&E staining. In all animal experiments, the NIH principles of laboratory animal care (Principles of Laboratory Animal Care, NIH Publication No. 85-23, revised 1985) were followed, as well as specific national laws (e.g., the current version of the German Law on the Protection of Animals), where applicable.

Blocks of paraffin-embedded human colonic carcinoma tissue were kindly provided by the Department of Pathology of Memorial Hospital of Union County (Marysville, OH). Serial 3-μm-thick sections of each tumor were immunostained with peroxidase-conjugated mouse antihuman CD11b (Chemicon International, Temecula, CA), peroxidase-conjugated B72.3 mouse anti-sTn MAb, or peroxidase-conjugated mouse antihuman IgG MAb (DAKO). All sections were counterstained with hematoxylin. Digital photographs of adjacent fields were obtained, and coincidence of different stains in specific cell clusters was assessed.

Tumors were disrupted using a combination of gentle mechanical dissociation followed by digestion with collagenase type IV (Sigma) and DNase I (Sigma) mixture. Spleens were mechanically dissociated. Spleen cells and tumor cells were isolated and immunostained with PE-conjugated anti-CD11b MAb (PharMingen). The relative amount of CD11b⁺ cells was assessed by flow cytometry using an EPICS system (Beckman Coulter, Inc., Miami, FL).

For each tumor, 10 cryostat sections (5-μm thick) were serially cut, 300 μm apart. Tumor sections were stained with both FITC-conjugated Abs to mouse CD31 (PharMingen) to visualize the vessels and PE-conjugated Abs to mouse CD11b (PharMingen) to identify MOs, and nuclei were counterstained with DAPI fluorochrome. The fluorescent and bright-field digital images of adjacent fields were recorded using a Zeiss photomicroscope (Carl Zeiss, Thornwood, NY) equipped with a TEC-470 video camera (Optronix, Goleta, CA). The digital images were analyzed with PhotoQuant image analysis software (NBG, Bilbao, Spain). Serial sections were used to determine associations between cell clusters and structures (MO clusters, tumor cell clusters, and vessels). The number of fluorescent MOs, total cells, and vessels was counted and expressed as the number of events/mm² tumor tissue surface area. Granulocytes were counted by nuclear morphology recognition (DAPI staining). The infiltration was expressed in cells/mm² tumor tissue surface area or by ratios (number of CD11b⁺:total DAPI-stained cell number, number of CD31⁺ events:total DAPI-stained cell number, and number of CD11b⁺ cells:number of CD31⁺ events).

All statistical analyses were performed using GraphPad Prism V.2.1 (GraphPad Software Inc., San Diego, CA). The statistical significance of the results was tested by Student’s t test (two-tailed P) or ANOVA, depending on the hypothesis to test. For subgroup analyses and to investigate the relationships between variables, Spearman’s correlation, χ² , Kruskal-Wallis, ANOVA, Mann-Whitney, and Student’s t tests were used as appropriate. P < 0.05 was considered statistically significant.

In an initial study, a limited number of archival human paraffin-embedded adenocarcinomas (five breast and five colon adenocarcinomas) were randomly selected. Serial sections of samples of each tumor were stained for CD11b, sTn Ag, or human IgG. In all tumors studied, abundant CD11b⁺ MOs, plasma cells and PMN cells were observed between the tumor cells and in the reactive stroma (Fig. 1,A). The higher concentration of these infiltrating cells was always in the interstitial stroma and in the interface tumor/normal connective tissue. A second staining using B72.3 MAb, which reacts exclusively with clustered sTn (27, 28), permitted the detection of cells containing the TAA TAG-72 (29, 30). This immunostaining (Fig. 1,B) showed that many MOs and PMN cells surrounding tumor cell clusters harbored tumor-derived TAG-72 molecules. Serial sections stained with antihuman IgG showed that most of the phagocytes infiltrating the tumor also contained IgG in the phagosomes (Fig. 1 C).

Eighteen C3H mice received a s.c. injection of 10⁶ Met129 isogeneic adenocarcinoma cells. Groups of six mice were sacrificed 1, 2, and 3 weeks after the tumor challenge, and the tumors were studied. The tumor sections were immunostained with PE-conjugated anti-CD11b MAb and counterstained with DAPI. Automatic cell counts of the different tumor sections, performed with the specific computer program, showed that CD11b⁺ cells were present throughout the entire tumor tissue and were mixed with tumor cells (Fig. 2,A). Moreover, the number of total cells integrating tumor tissue was proportional to the number of CD11b⁺ cells (Fig. 2,B). Considering a total of 300 frames analyzed, variations of the total cells:CD11b⁺ cells ratio were approximately 5:1 (Fig. 2,B). It was evident that this proportion was independent of tumor size or age. The flow cytometry analyses confirmed these data. The proportion of CD11b⁺ cells was 22–28% in all samples analyzed (30 total), independent of the size or age of tumors (Fig. 2 C).

Tumor cell-MO interactions in the tumor microenvironment may result in different consequences such as tumor cell death and/or the release of inflammatory cytokines and VEGF (31, 32). To test whether interactions of tumor cells and MOs changed the pattern of VEGF production, an in vitro assay was developed in which tumor cells and MOs were cocultured, and the VEGF release into the media was monitored within a 48-h period. In that model, the tumor cells were human (T47D breast adenocarcinoma and SW620 colon carcinoma), and the MOs were murine. h-VEGF and m-VEGF were differentiated using specific MAb with no cross-reactivity. That strategy allowed the discrimination of tumor cell or MO origin of the VEGF released into the media. Two types of cocultures were tested: (a) tumor cells and MOs (or others) in contact; and (b) tumor cells and MOs (or others) not in contact but sharing the media. These experiments demonstrated that the tumor cells secrete VEGF in detectable amounts (Fig. 3,A). We tested several murine cell types including MOs, latex-activated MOs, PMN cells, and the plasmocytoma cell line P3X63Ag8. Only the MOs secreted considerable amounts of VEGF even when cultured alone. In this experiment, the amount of VEGF secreted by the MOs was always 2–3 times higher than the amount of VEGF secreted by those specific tumor cells (Fig. 3, A and B). When MOs and tumor cells were cocultured in contact or sharing soluble factors, the tumor cells never altered their pattern of VEGF secretion (Fig. 3,C). However, the MOs in the presence of tumor cells always increased secretion of VEGF by a factor of 2 or 3 over the MO baseline (Fig. 3,D). The consequence of that increase was that the final VEGF produced by the MOs was 5–6 times the VEGF produced by tumor cells (Fig. 3, C and D). To test whether the PMN cells in the inflammatory infiltrate could affect the amount of VEGF release into the tumor microenvironment, proportional numbers of murine PMN cells were added to MO cultures and to MO/tumor cell cocultures. PMN cells alone did not secrete significant amounts of VEGF (Fig. 3, B and D). MOs augmented by a statistically significant 20–35% the amount of m-VEGF released into the media in the presence of PMN cells. Addition of tumor cells into cultures containing MOs and PMN cells increased the amount of m-VEGF released into the media by 400–500% over the MO baseline (Fig. 3 D).

To test whether tumor cell-MO interaction could alter patterns of VEGF mRNA expression of either cell type, several models of cocultures were tested. MOs cultured alone showed a repetitive pattern: VEGF-A and VEGF-B mRNA were present, but VEGF-C or VEGF-D were not (Fig. 4,A). MOs cocultured with tumor cells, in contact or separated, showed the consistent presence of VEGF-A and VEGF-B mRNA, but not VEGF-C or VEGF-D (Fig. 4, B–D). When cultured alone or in the presence of MOs, tumor cells showed all four types of VEGF mRNA (Fig. 4,E). However, in growing Met129 tumors, VEGF-C mRNA was not detectable (Fig. 4 F).

To test the possible association between MO infiltration and tumor neoangiogenesis, Met129 tumor cells were injected s.c. in groups of mice. These mice were sacrificed when the tumors were 5 and 10 mm in diameter. Tumors were excised together with surrounding tissues to maintain intact histology of tumor borders. Serial sections were triple-stained with DAPI (to use as a nuclear reference), PE-conjugated anti-CD11b MAb (to detect MOs), and FITC-conjugated anti-CD31 MAb (to detect endothelial cells). A computerized analysis was done on 300 color video images of 30 tumor sections. The red:blue fluorescence ratio was determined as an expression of the ratio of MOs to total cells. In early tumors (5 mm in diameter), the MOs were predominantly concentrated on tumor borders, especially at the interface between the growing mass of tumor cells and s.c. connective tissue (Fig. 5,A). CD31⁺ cell concentration was parallel to the MO concentration, forming capillaries with a predominant orientation toward the center of the tumor mass, in the same direction as the MO gradient (Fig. 5,B). In advanced tumors (10 mm in diameter), MOs were highly concentrated in the periphery of large vessels in perivascular connective tissue (Fig. 5,C), suggesting that this area assumed the role of the interface observed in early tumors. The CD31⁺ cells followed parallel distribution to MO clusters and were more concentrated in the periphery of large vessels, projecting capillaries toward the tumor center (Fig. 5 D).

Previously, we observed that IgG against TAAs could stimulate tumor invasion and metastasis (31), mediated by a cooperation of PMN cells and MOs activated by IgG⁺ Ag complexes (32). Based on these published data and the association of TIMs with TAAs and host IgG observed in human tumors, we studied whether both TAAs and Abs raised against these Ags could modulate the MO-derived VEGF release. Using an in vitro model, two human tumor cell lines (breast T47D and colon SW620 cells) that secrete sTn TAA were selected (Fig. 6,A). The secreted Ag was captured from the cell culture supernatant in a column or by immunoprecipitation using B72.3 MAb. Immobilized Ag was further detected by peroxidase-conjugated B72.3 Mab, showing, as known, the existence of multiple repeated epitopes on the same molecule/conglomerate of Ag released by the tumor cells (repeated epitope Ag). Additionally, it was confirmed that neither murine MOs nor murine PMN cells released such Ag (Fig. 6,A). Bovine salivary mucin, a repeated epitope sTn Ag, and a synthetic single epitope sTn Ag were used as controls for in vitro experiments. When tumor cells (secreting sTn) were cultured alone in the presence of B72.3 Mab, with the single epitope sTn Ag, or with both together, the amount of VEGF released into the media in 48 h was not changed. Therefore, the presence of anti-tumor-associated Ab did not modulate the tumor cell VEGF release (Fig. 6,B). The presence of single epitope sTn Ag or anti-sTn IgG did not change the release of MO-derived VEGF in 48 h compared with the controls. However, it was noted that MOs cultured in the presence of repeated epitope sTn Ag increased VEGF release in 48 h (Fig. 6,C). Addition of both repeated epitope sTn Ag and anti-sTn IgG produced the same effect as repeated epitope sTn Ag alone (Fig. 6,C). Cocultures containing MOs, PMN cells, and T47 D tumor cells showed that in the presence of tumor cells (tumor cells and sTn Ag secreted by the tumor cells), MOs increased the release of VEGF in 48 h (Fig. 6,D). Also consistent with previous experiments, the addition of anti-sTn IgG did not change the amount of MO VEGF released in the presence of tumor cells and sTn Ag (Fig. 6,D). Notably, the presence of tumor cells in cocultures of MOs and PMN cells dramatically increased the MO release of VEGF (Fig. 6,D). Significantly, when anti-sTn IgG was added to cultures of MOs in the presence of tumor cells with their own shed sTn Ag and PMN cells, in 48 h the amount of MO-secreted VEGF increased 500% over the normal production (Fig. 6 D).

To analyze the in vivo consequences of the data shown above, the density of CD11b⁺ cells and capillaries generated (using immunofluorescence staining with fluorescein-conjugated antimouse CD31) were measured in T47D tumors growing in SCID mice and nu/nu mice. In an experimental model of SCID mice growing T47-D sTn⁺ tumors, no anti-sTn IgG was present in serum. In this model, tumors grew slowly and showed very few new vessels (Fig. 7, A and B), and the number of MO clusters was parallel to the number of vessels and the number of tumor cells. In a group of six SCID mice of this model, s.c. injection of 150 μg of irrelevant IgG did not introduce differences in tumor angiogenesis or tumor growth. However, a group of these mice, which received a s.c. injection of 150 μg of B72.3 Mab, showed a very significant increase of serum anti-sTn IgG (exogenous), new tumor vessels (Fig. 7,A), and a proportional number of MOs. Those tumors showed a significantly faster tumor growth (Fig. 7,B). A second model of nu/nu mice growing T47D sTn⁺ tumors showed results consistent with the above data. In the serum of the nu/nu mice, there were significant titers of endogenously generated anti-sTn IgG (Fig. 7,A). Tumors of this model showed higher concentrations of new vessels and faster growth (Fig. 7, A and C). The s.c. injection of 150 μg of irrelevant IgG in these mice did not introduce significant changes in MO infiltration, vessel density, or tumor growth (Fig. 7,A). However, the group of mice that received s.c. injection of 150 μg of B72.3 MAb (exogenous anti-sTn IgG) showed a very significant increase of serum anti-sTn IgG, new tumor vessels (Fig. 7,A), and a proportional number of MOs (data not shown). The tumor growth was significantly faster during the last 7 days of the experiment (Fig. 7 C).

The results presented here suggest that the TIMs have an important role in promoting tumor angiogenesis, maintaining tumor growth, the invasive course of malignant tumors, and facilitating the metastasis cascade. In addition, formation of immune complexes in tumor connective matrix appears to be a complementary mechanism of MO and PMN cell attraction into the tumor tissue. Since the discovery of the capacity of tumor cells to produce VEGF, it has been broadly assumed that the tumor cell drives vessel formation in growing neoplastic tissue (33, 34).

Independently of this assumption, experiments inhibiting critical angiogenic factors such as VEGF demonstrate that neoangiogenesis is a bottleneck in the malignant process; by impairing angiogenesis, tumor growth could be totally suppressed (35, 36, 37). Because the majority of tumor cells express angiogenic factors, it can be assumed that the tumor cell is necessary and sufficient to sustain the neoangiogenesis that autoensures tumor growth. However, there is a certain discrepancy in accepting the predominant role of the tumor cell supplying sufficient vascular stimulating factors for effective tumor neoangiogenesis. Other normal cells involved in the tumor biology such as monocytes and platelets may import considerable amounts of VEGF into critical points of the tumor (38, 39, 40). MOs are capable of producing and releasing VEGF under the influence of certain factors such as hypoxia and nitric oxide, both common factors of the tumor environment (41). Actually, high VEGF expression has been frequently observed in TIMs (42), and increased tumor angiogenesis has been reported after MO recruitment into the tumor (23).

TIMs are common and abundant (43) and are frequently interpreted as representative of antitumor activity (44). However, here and in other published studies (45), our histology results cannot sustain an important tumor cytolysis due to the infiltrating MOs. Actually, our results show that the TIMs are mainly concentrated in the interphase between the tumor and normal tissue and, in general, are associated with phagocytosis of tumor cell debris and immune complexes containing TAAs. This suggests that in the tumor environment, the MO is acting more as a scavenger cell than as a cytotoxic/cytolytic cell. Our substantial morphometric and flow cytometry analyses show that the number of MOs infiltrating the tumor mass in experimental tumors is proportional to tumor volume and that during the exponential phase of tumor growth a ratio of tumor cells:MO remains practically constant. Considering both observations together strongly reinforces the concept that the MOs may play an essential role in promoting tumor growth (23). In vivo, MO recruitment into the tumor could be driven by multiple factors including hypoxia, MO or tumor cell production of chemokines (46, 47, 48), and accumulation of cell debris and/or immune complexes in the tissue microenvironment (49, 50). The repetitive presence of MOs with endocytosed material containing TAA and anti-TAA IgG in human tumors, together with our in vitro results, supports the possibility that in the in vivo tumor environment, both phagocytosis of cell debris and the presence of immune complexes (formed with TAA and anti-TAA IgG) may promote the secretion of very important amounts of VEGF by infiltrating MOs. The frequent association of MOs with PMN cell and plasma cell infiltration, mainly observed at the tumor boundaries, suggests that a concomitant inflammation follows MO recruitment into the tumor cell invasive front. Immune complexes involving shed TAA and antitumor IgG may activate this inflammatory cellular infiltrate, stimulating the local secretion of proinflammatory cytokines and establishing a feedback process of infiltration-activation (31). The in vitro tests also showed that tumor cells could stimulate the secretion of MO VEGF and not vice versa. No experiments were done to uncover the molecular nature of factors involved in this stimulus, but the results showed that tumor cell-MO contact is not needed to provoke this MO response. RT-PCR analysis of the MO mRNA before and after such stimulus showing the same pattern of VEGF gene expression (VEGF-A and -B were expressed, but not VEGF-C or -D) suggests that the VEGF MO response was quantitative but not qualitative. A tumor cell type capable of expressing VEGF-A, -B, -C, and -D in vitro inhibited the expression of VEGF-C, which is supposedly involved in promoting lymphatic angiogenesis, when growing in vivo (51, 52). This VEGF-C-specific inhibiting phenomenon could be anecdotal of this tumor cell type but essentially illustrates that tumor cells could alter their pattern of VEGF expression depending on microenvironmental factors. Importantly, the in vitro coculture experiments involving MO, PMN cells, and tumor cells strongly suggest that in the tumor, within the cellular inflammatory environment, each MO can secrete much more VEGF than certain tumor cells, which substantiates the importance of TIMs as a source of VEGF in the tumor tissue. In normal tissue repair, it has been demonstrated that the recruitment of MOs constitutes the main source of VEGF and the fundamental step for the establishment of effective angiogenesis (53). In experimental s.c. injected tumors, our results showed that in early stages of tumor growth, the MO concentration and capillary development were always coincidental (statistically significant, P < 0.001, positive correlation between number of CD11b cells/number CD31 structures/field). Actually, the images suggested that tumor cells were growing toward the MO/capillary-rich areas, in an expansive/invasive progression.

Consistent with this pattern, in advanced tumors (10 mm in diameter), large tumor vessels with well-formed walls assume the role of the s.c. connective tissue interface observed in the early tumors. The new capillary formation was always close to MO clusters, concentrated in the periphery of the large vessels and projecting radial branches toward the tumor center. Both sets of data emphasize that tumor cells grow, invading the well-vascularized reactive stroma (54) and not that the stroma infiltrates the tumor cell clusters.

If the reactive stroma can support tumor growth, the identification of mechanisms capable of triggering and sustaining a stroma reaction is very important. Cytokines and other factors released by the tumor cells could be initiators. In our experiments, molecules identified by the immune system as TAAs were one kind of these tumor-released factors. The example of the common tumor-associated sTn Ag was illustrative. Complexes of molecules with repeated sTn epitopes and B72.3 IgG activate the MO VEGF response, probably through the FcγRI receptor, as demonstrated previously (31, 32). B72.3 IgG alone did not increase the MO VEGF release, showing that the presence of monomeric IgG does not induce that MO response. However, immune complexes formed with sTn-presenting molecules and specific IgG act as moderate stimulators of the MO VEGF response and also as strong inducers of MO inflammatory response (32), which may attract PMN cell infiltration. The in vitro experiments suggest that the concurrence of tumor cells shedding antigenic debris, the specific antitumor Ag IgG, and infiltrating PMN cells can dramatically increase the MO release of VEGF. Therefore, the locally concentrated immune complexes composed by shed tumor Ags and the specific IgG could represent a fundamental primer and/or maintainer of stroma activation, which can support tumor growth (55). The results from the in vivo experiments are consistent with these interpretations. The number of capillaries/unit of tumor volume and the daily growth rate of tumors were related to the serum concentration of anti-sTn IgG. Tumors growing in SCID mice, which do not develop specific IgG response, had very low vascular network and very slow growth. However, tumors growing in SCID mice injected with exogenous B72.3 IgG, specific for the TAA sTn, showed much higher capillary density and faster tumor growth. Such a tumor-enhancing effect was not observed in SCID mice injected with the same doses of irrelevant IgG isotype, which proves that increase of angiogenesis is related to the specific recognition of TAAs by the specific IgG. Moreover, nu/nu mice capable of developing a B-cell response, mainly IgM based, produced a certain serum concentration of endogenous anti-sTn IgG, and these mice developed tumors with a significantly higher density of capillaries as compared with tumors developed in SCID mice. With identical numbers and classes of tumor cells implanted in nu/nu beige mice or in SCID beige mice, tumors grew faster in nu/nu mice. When nu/nu mice had an exogenous source of B72.3 Ab supplied by injection, the results showed a higher concentration of serum anti-sTn IgG and dramatically increased the tumor vascularization. That effect was also specific because the same carcinoma growing in nu/nu mice injected with irrelevant IgG had the same capillary density as the controls. Neither of these mouse types, SCID or nu/nu beige, have natural killer cells or effective T cells. Therefore, the difference in tumor growth cannot be attributed to a different tumor rejection capacity. The only possible effectors for tumor rejection in these mouse types are the MOs and the PMN cells. However, the microscopy examination showed no suspicion of efficient tumor rejection associated with stromal cell infiltration. On the contrary, the stromal cell infiltrate was always associated with neoangiogenesis. In vitro data suggest that immune complexes involving shed tumor Ags and specific IgG constitute a powerful enhancer of VEGF production by stromal cell infiltrate. Considering that those immune complexes can also induce the release of chemoattractant factors by stromal cells, it is understandable that the immune complexes could further sustain tumor infiltration by more stromal cells, establishing a feedback process of infiltration-inflammation-angiogenesis. All these data suggest that tumor growth could be maintained, at least in part, by the humoral immune response against shed TAAs that may promote recruitment of stromal cells in the tumor milieu and MO production of inflammatory and angiogenic factors including high amounts of VEGF (31, 32). To address possible therapeutic implications, it is key to know how important this tumor-promoting host response is in the biology of human tumors. Many clinical data suggest that the humoral immune response against tumor Ags does not represent an effective tumor cell rejection process, and historically, the value of such a response has been considered puzzling (56, 57).