Adiponectin Deficiency Limits Tumor Vascularization in the MMTV-PyV-mT Mouse Model of Mammary Cancer Free

Purpose: High levels of the fat-secreted cytokine adiponectin (APN) are present in the circulation of healthy people, whereas low levels correlate with an increased incidence of breast cancer in women. The current study experimentally probes the physiologic functions of APN in mammary cancer in a newly generated genetic mouse model.

Experimental Design: We established an APN null mouse model of mammary cancer by introducing the polyoma virus middle T (PyV-mT) oncogene expressed from mouse mammary tumor virus (MMTV) regulatory elements into APN null mice. MMTV-PyV-mT–induced tumors resemble ErbB2–amplified human breast cancers. We monitored tumor onset, kinetics, and animal survival, and analyzed vascular coverage, apoptosis, and hypoxia in sections from the primary tumors. Metastatic spreading was evaluated by analyses of the lungs.

Results: APN prominently localized to the vasculature in human and mouse mammary tumors. In APN null mice, MMTV-PyV-mT–induced tumors appeared with delayed onset and exhibited reduced growth rates. Affected animals survived control tumor-bearing mice by an average of 21 days. Pathologic analyses revealed reduced vascularization of APN null tumors along with increased hypoxia and apoptosis. At the experimental end point, APN null transgenic mice showed increased frequency of pulmonary metastases.

Conclusion: The current work identifies a proangiogenic contribution of APN in mammary cancer that, in turn, affects tumor progression. APN interactions with vascular receptors may be useful targets for developing therapies aimed at controlling tumor vascularization in cancer patients.

The dependence of tumor growth and progression on the stromal microenvironment has inspired an intense body of research into the identification of factors that regulate these processes (1). Whereas much of this work has concentrated on vascular factors and inflammatory responses, less is known about the role of adipocyte-derived signals on cancer progression. Adiponectin (APN; also known as ACRP30, AdipoQ, and GBP28) is a circulating adipocytokine that is implicated in metabolic and vascular regulation. Serum levels are inversely correlated with obesity-related metabolic disorders, insulin resistance, hypertension, coronary heart disease, and stroke (2–5). Clinical studies correlate low APN levels in women with an increased risk of mammary cancer (6, 7) and a metastatic tumor phenotype (8). These observations have led to the suggestion that APN exerts a cancer protective function. Indeed, APN treatment reduces the proliferation of select human mammary cancer cell lines in vitro and in vivo (9). On the other hand, APN affects the vasculature that critically supports tumor growth and progression by providing nutrients and oxygen (10). APN induces blood vessel infiltration into matrigel plaques (11) and regulates angiogenesis in the experimental hind limb ischemia model (12). A similar proangiogenic role may play into tumor progression. APN-deficient mice develop normally until early adulthood and show no overt phenotype as judged by numerous physiologic and metabolic variables (12, 13).

To distinguish the cellular functions of APN in mammary cancer, we have generated an APN null mouse genetic model that mimics human disease. Thirty percent of human breast cancers are characterized by overexpression of the receptor tyrosine kinase ErbB2 (14). The viral polyoma middle T (PyV-mT) oncogene is a membrane-associated surrogate for ErbB2 (Her2/neu) heterodimers (15). Transgenic PyV-mT expression through mouse mammary tumor virus (MMTV) regulatory elements leads to ligand-independent activation of mitogen-activated protein kinase and phosphatidylinositol 3-kinase cascades in the mammary epithelium (15). The same signaling pathways are stimulated in Neu/ErbB2 transgenic murine mammary carcinomas and in human breast cancers overexpressing ErbB2 (14). Gene expression profiles show similarities of MMTV-PyV-mT–driven tumors to human luminal type B breast cancers (16). The PyV-mT model reiterates the stages of human carcinoma progression both by morphology and biomarker expression (17), including the loss of the luminal marker Gata-3 (18–20). In humans, poor clinical outcome is linked to ErbB2 overexpression and loss of estrogen receptor gene expression (21–23), and similarly, advanced murine MMTV-PyV-mT–induced tumors display ErbB2 amplification and estrogen receptor loss (17, 24). Finally, MMTV-PyV-mT transgenic mice develop metastatic cancer mimicking the clinically advanced stage of human mammary carcinoma progression (25).

We have used this clinically relevant breast cancer model to probe the functions of APN in tumorigenesis. By generating and analyzing APN null MMTV-PyV-mT transgenic mice, we reveal a novel and surprising contribution of APN to mammary tumor vascularization. This phenotype mimics that of MMTV-PyV-mT mice lacking the APN-binding protein T-cadherin (26) and implicates a functional link between these proteins in the crosstalk between tumor cells and the stromal microenvironment in mammary cancer.

Mice. C57BL/6 APN knockout (APN-KO) mice were generated in Dr. Yuji Matsuzawa's laboratory (Osaka University, Osaka, Japan; ref. 13). Dr. William Muller (McGill University, Montreal, Canada) first generated the MMTV-PyV-mT transgenic mice. For experiments reported in this study, we used mice with the MMTV-PyV-mT transgene in the C57BL/6 background provided by Dr. Leslie Ellies (University of California, San Diego, La Jolla, CA). For the generation of APN-KO MMTV-PyV-mT mice, MMTV-PyV-mT males were crossed with homozygous APN-KO females. The male APN heterozygous MMTV-PyV-mT offspring was bred with APN-KO or wild-type (WT) females, yielding MMTV-PyV-mT APN-WT and APN-KO females. Genotypes were determined by PCR. APN-for 5′-GGAACTTGTGCAGGTTGGAT-3′ and APNrev 5′-CAGTGCAAGCTCCAAGATGA-3′ amplified the WT 266-bp DNA fragment. APN-KO-for 5′-ATACTTTCTCGGCAGGAGCA-3′ and APNrev amplified the 900-bp KO DNA fragment. We confirmed the lack of APN protein in APN-KO animals by ELISA (R&D; Supplementary Table S1). The animals were fed with a normal diet.

Tumor formation and analysis. Tumor onset was monitored twice weekly by palpation. Tumor sizes were measured with digital calipers twice weekly, and the volume was calculated as (length × width²)/2. Animals were sacrificed when the tumor burden visibly affected the host or when the tumors reached the institutionally set limit of 20 mm along one axis. Statistical analysis of tumor onset and survival was done in Prism with the use of a log rank test, and growth kinetics was analyzed with the use of linear regression.

Immunoblotting. Mammary gland or tumor tissue was mechanically dissociated in lysis buffer [50 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 5 mmol/L EDTA, 1 mmol/L dithiothreitol, 1/100 protease inhibitor cocktail (Sigma), 0.1 mmol/L phenylmethylsulfonyl fluoride, and 1% NP40]. Protein (10 μg per lane) was separated by reducing SDS-PAGE and transferred to polyvinylidene fluoride membranes. T-cadherin, APN, and β-tubulin were detected with the use of specific antibodies and the ECL Detection kit (Amersham).

Quantitative real-time PCR. Total RNA was extracted from normal mammary gland or tumor tissue with the use of Trizol (Invitrogen). Equal amounts were reverse transcribed with the use of oligo(dT)₁₈ and random hexamer primers (Transcriptor First Strand cDNA Synthesis kit; Roche). Real-time PCR analysis was done with SYBR green with the use of the Stratagene Mx3000P instrument. T-cadherin (5′-catcgaagctcaagatatgg-3′, 5′-gatttccattgatgatggtg-3′), AdipoR1 (5′- acacagagactggcaacatc-3′, 5′-gagcaatccctgaatagtcc-3′), and AdipoR2 (5′- tggacacatctcctaggttg-3′, 5′-tagagaagagtcgggagacc-3′) cDNAs were detected and normalized to glyceraldehyde-3-phosphate dehydrogenase (5′-ccagtatgactccactcacg-3′, 5′-gactccacgacatactcagc-3′).

Tissue preparation and histology. Tumors and lungs were fixed overnight in 10% zinc formalin (Fisher), embedded in paraffin, and then sectioned and stained with H&E. For pathologic evaluation, Dr. Cardiff examined 33 tumors from 17 WT mice and 16 tumors from 31 APN-KO mice, as well as the lungs from 10 WT and 12 APN-KO mice. Snap-frozen fresh mammary glands were sectioned and fixed with ice-cold acetone.

Immunohistochemistry. Tumor sections were deparaffinized and incubated with 3% hydrogen peroxide in PBS. For CD31 staining, sections were digested with 0.1% trypsin (Zymed). For proliferating cell nuclear antigen (PCNA) staining, sections were treated with target retrieval solution (Dako). Endogenous biotin was blocked with the avidin-biotin kit (Vector Labs). Sections were incubated with antibodies for CD31 (Pharmingen) or PCNA (Chemicon), and staining was detected with the respective biotinylated secondary IgGs (Pharmingen), streptavidin/horseradish peroxidase (Vector Labs; ABC standard), and 3,3′-diaminobenzidine (Dako). CD31-labeled sections were counterstained with hematoxylin, and PCNA-stained sections with methyl green. Apoptotic cells were identified by the terminal deoxyribonucleotidyl transferase–mediated dUTP nick end label (TUNEL, Millipore). Hypoxic tumor areas were visualized with the use of the Hypoxyprobe-1 Plus (Natural Pharmacia International). TUNEL and hypoxyprobe-stained sections were counterstained with methyl green.

Acetone-fixed mouse or human mammary gland sections were air-dried and blocked in PBS containing 10% FCS, 1% bovine serum albumin, and 0.2% Triton X-100. Sections were then incubated with APN (PA1-054; Affinity BioReagents and R&D), CD31 (MEC 13.3; Pharmingen), T-cadherin (26) antibodies. Alexa 488 and Alexa 594 fluorescent conjugates (Molecular Probes) were used to detect the respective primary antibodies. The negative controls were slides processed in parallel without primary antibody. For T-cadherin and APN, tissues from KO animals were used as additional controls.

Mammary gland whole mounts. Mammary glands were processed as described previously (26).

Image acquisition and analysis. Whole mounts were digitized with an Epson Perfection 4490 photo scanner and analyzed with the use of ImagePro and Prism (unpaired t test). Images of fluorescently stained mammary glands were acquired with the use of a FluoView 1000 confocal microscope (Olympus). For quantitative analyses, sections stained with CD31, PCNA, TUNEL, hypoxyprobe, and H&E were scanned with the use of the Aperio ScanScope XT system (27) at ×20 magnification (resolution of 0.5 μm/pixel). This automated microscope generates high-resolution whole slide images of the full face of tissue sections and does not restrict analysis to representative images. The two largest tumors from each mouse were used for analyses; one section was analyzed from each tumor. Image analysis was done with the use of the web-based digital pathology information management Aperio Spectrum software (www.aperio.com). The number of lung metastases (foci of metastatic adenocarcinomas lodged in the lung parenchyma) was counted manually and blinded with the use of at least five serial sections from each of the three lobes of the right lungs. CD31 and hypoxyprobe staining was quantified with the use of a brown-versus-blue deconvolution algorithm yielding the ratio of positive signal area versus whole tumor area. An algorithm detecting the ratio of positive nuclei versus all nuclei in a complete tumor section was used for analysis of PCNA and TUNEL staining. Statistical analyses were done in Prism with the use of unpaired t test or nonparametric Mann-Whitney test.

Corneal vascularization assay. Slow-release hydron pellets (Sigma; 0.4 × 0.4 × 0.2 mm) contained 180 ng vascular endothelial growth factor (VEGF; R&D) or 500 ng high–molecular weight (HMW) APN or PBS. Adult male WT mice were anesthetized with Avertin (0.015 mL/g), and the pellet was implanted into the corneal stroma of one or both eyes at a distance of 2 mm from the corneoscleral limbus. Clock hours of neovascularization (CN) and maximal vessel length (VL) were measured after 10 d; the angiogenic area was calculated as 0.2 × π × VL (mm) × CN (mm) (28).

Generation of recombinant HMW APN. APN was isolated as described previously (29).

APN associates with the mammary gland vasculature. Because little is known about the localization and cellular binding sites for APN in the mouse mammary gland, we first examined the distribution of APN in normal tissue. Immunohistochemistry of fat pads from 60-day-old virgin females detected APN in prominent association with blood vessels identified by endothelial cell marker CD31 expression (Fig. 1A). The vascular APN signal was abolished in the fat pads from APN null (APN-KO) mice (Fig. 1A), supporting the specificity of the antibody reagents. No overt differences in vascular density were noted between genotypes in the normal mammary gland. Vascular endothelial cells profusely express the APN-binding protein T-cadherin (26). To determine if APN codistributes with glycosylphosphatidylinositol–linked T-cadherin on endothelial cell membranes, we did double-labeling immunohistochemisty with goat anti-APN and rabbit anti–T-cadherin antibodies (26). APN precisely mirrored the punctate pattern of T-cadherin on the capillaries in mammary fat pads (Fig. 1A, g-i). No specific APN staining could be identified in the ductal epithelium (not shown). To validate the mouse model, we compared the distribution of T-cadherin and APN in the human normal breast and in a mammary ductal invasive carcinoma biopsy. As in the mouse model, the capillaries were positive for T-cadherin and APN in human tissue (Supplementary Fig. S1). Together with our previous finding that the vascular APN association depends on T-cadherin expression (26), these data show that the vasculature is a major site for APN binding in the human and mouse mammary gland.

To address a possible contribution of APN to mammary gland development, we compared the ductal patterning of whole-mounted, carmine-stained mammary glands from virgin WT and APN-KO females at 21, 35, 56, and 84 days of age. Ductal growth and branching showed no differences between genotypes (Supplementary Fig. S2). Moreover, APN-KO females nourished multiple normal-size litters (data not shown). Thus, normal mammary gland development and function appear to occur independent of APN.

APN and APN receptor expression in the MMTV-PyV-mT mouse mammary tumor model. The transgenic expression of PyV-mT from MMTV promotor/enhancer elements produces tumors progressing from premalignant to malignant stages with distant metastases in all female mice (25). In MMTV-PyV-mT–induced neoplastic areas of the mammary glands of 120-day-old mice and in tumors of 170-day-old mice, APN was specifically associated with the vasculature (26).⁴

Because some breast cancers in women correlate with low APN levels (6, 7), we investigated if serum APN levels show a similar decrease in the MMTV-PyV-mT model. ELISA analyses of serum from normal and tumor-bearing female mice at 85 and 150 days of age identified similar APN concentrations in tumor-free (5.6 ± 0.11 μg/mL) and tumor-bearing (5.41 ± 0.31 μg/mL) animals (n = 6 per group; Supplementary Table S1). No serum APN was detected in APN-KO mice. Thus, in the MMTV-PyV-mT model, tumor formation is not linked to reduced serum APN concentrations.

To determine how APN association in the mammary gland correlates with tumorigenesis, we compared APN levels in the fat pads from normal and tumor-bearing mice by Western blotting. In transgenic MMTV-PyV-mT tumors from 150-day-old females, the levels of APN and the APN-interacting protein T-cadherin were dramatically reduced compared with normal virgin fat pads from 85-day-old mice (Fig. 1B). No APN was detected in normal or tumor tissue from APN-KO mice (Fig. 1B). APN-KO PyV-mT tumors displayed reduced T-cadherin levels compared with the normal mammary gland.

In addition to T-cadherin, APN interacts with the two seven-pass transmembrane APN receptors 1 and 2 [AdipoR1 and AdipoR2; ref. 30]. To correlate the expression levels of these receptors with tumor progression, we examined the mRNA expression levels in normal and tumorigenic mammary glands by quantitative real-time PCR (Fig. 1C). The T-cadherin mRNA was reduced in tumors compared with normal breast tissue, confirming the protein expression data. In addition, the mRNA levels for AdipoR2 were significantly diminished in PyV-mT tumors, whereas the mRNA for AdipoR1 remained unchanged (Fig. 1C). These combined data suggest the coordinate reduction of APN, AdipoR2, and T-cadherin in PyV-mT mammary tumors and raise the possibility of their concerted regulation or interactions in the mammary gland.

Restricted MMTV-PyV-mT mammary tumor growth in APN null mice. To investigate the contributions of APN in mammary tumorigenesis, we introduced the APN null mutation into the syngeneic C57BL/6 MMTV-PyV-mT transgenic mice and generated homozygous APN-KO and WT female offspring in the second generation. Both the WT and APN-KO MMTV-PyV-mT mice developed tumors. Tumor onset was defined as the age of a mouse when the first tumor could be palpated. WT females presented the first tumor at a median age of 73 days, whereas tumors were first detected at 94 days in APN-KO mice (n = 17 for WT and n = 33 for APN-KO). Thus, tumor onset was delayed by an average of 21 days in APN-KO mice compared with WT controls (Fig. 2A).

To probe if APN affected early tumor development, we investigated MMTV-PyV-mT–induced neoplasias in number 4 mammary gland whole-mount preparations from WT and APN-KO mice. At 84 days of age, we registered a 2.2-fold reduction in neoplastic area in APN-KO animals (5.41% ± 0.94%) compared with WT controls (11.73% ± 1.72%; Fig. 2B). The reduction of neoplastic growth at early stages is in line with the later occurrence of mammary tumors in the APN-KO mice.

The reduced tumor growth in APN-KO MMTV-PyV-mT transgenic mice extended life span. Whereas WT PyV-mT mice survived for an average of 130 days, tumor-bearing APN-KO mice carried on for a median of 160 days (Fig. 2C; n = 17 for WT and n = 33 for APN-KO). The institutionally determined experimental end point was when the largest tumor reached the maximal size of 20 mm over one axis or when the animal became moribund. The extended survival rates of APN-KO mice reflect the retardation in tumor onset and a possible reduction in the tumor growth kinetics. Indeed, a linear regression analysis of the tumor growth in APN-KO and WT MMTV-PyV-mT mice revealed significantly reduced growth rates of APN-KO versus WT tumors (Fig. 2D). In combination, these data show that loss of APN leads to delayed onset and slower growth of mammary tumors, which results in increased survival of affected animals.

Pathologic analysis of WT and APN-KO tumors revealed the presence of solid or papillary adenocarcinomas and adenosquamous carcinomas without clear histologic differences between genotypes. This suggests that the pathology of APN-KO tumors remained within the normal range of the MMTV-PyV-mT phenotype (31). Tumor growth depends on sufficient oxygen and nutrient supply through blood vessels. To determine if APN affects tumor angiogenesis, we examined the vascular coverage of APN-KO and WT PyV-mT tumors. Vessel density determined by CD31 immunostaining was reduced by 47% in APN-KO tumors versus WT (WT, 5.81% ± 0.83% versus APN-KO, 3.09% ± 0.54%; P < 0.01; Fig. 3A; n = 34 tumors for WT and n = 35 tumors for APN-KO). These data suggest an angiogenic contribution of APN to tumorigenesis.

To gain further insight into the cellular alterations of APN-KO mammary tumors, we examined the degree of hypoxia and apoptotic cell death in tumors from both genotypes. Hypoxyprobe identified 39.86% ± 5.59% of the total APN-KO tumor area as hypoxic versus 28.08% ± 1.97% in WT tumors (P < 0.05). This reflects a 42% increase of hypoxic tissue in APN-KO over WT tumors (Fig. 3B). Furthermore, TUNEL staining indicated a 2.4-fold significant increase of apoptotic cells in APN-KO tumors versus controls (WT, 1.6% ± 0.23% versus APN-KO, 3.85% ± 1.02%; P < 0.05; Fig. 3C). Proliferation rates measured by counting PCNA-positive nuclei showed no statistically significant difference between genotypes, although a trend toward reduced proliferation in APN-KO animals was noted (WT, 13.3% ± 2.01% and n = 5 versus APN-KO, 9.97% ± 1.05% and n = 6; P > 0.05; data not shown). These combined data suggest a proangiogenic contribution of APN with increased hypoxia and apoptosis in APN-KO mammary tumors.

Increased metastatic rates in APN-KO MMTV-PyV-mT tumors. Tumor cells respond to a hypoxic tumor microenvironment with changes in tumor behavior. Although we could not identify pathologic changes in APN-KO tumors versus WT, analyses of right lung serial sections at the experimental end point (Fig. 3D) revealed an increase in the number of pulmonary metastases in the APN-KO condition (6.8 ± 2.08; n = 12 animals) versus WT (1.9 ± 0.63; n = 10 animals; P < 0.05). Because APN is not detected in association with tumor cells in WT mouse MMTV-PyV-mT tumors, we attribute the higher metastatic rates in the APN-KO to the limitation of tumor angiogenesis that leads to hypoxia and, in turn, supports a metastatic phenotype.

HMW APN induces angiogenesis in the cornea. APN is prominently present in the serum as the HMW isoform (32). HMW APN binds T-cadherin and is suggested to be the physiologically most relevant isoform (33). To test if HMW APN is sufficient to induce angiogenesis, we tested HMW APN–induced angiogenesis into the normally avascular cornea of WT mice. Slow-release pellets containing eukaryotically produced, recombinant HMW APN (500 ng), VEGF (180 ng), or PBS were implanted close to the center of the cornea. Blood vessel growth toward the implants was monitored on a daily basis. Ten days after surgery, we observed a comparable degree of blood vessel growth toward the VEGF and APN implants, whereas no vascularization was detected in the negative controls (Fig. 4). The quantification of angiogenic areas from at least four mice in each condition established that the supply of HMW APN in the vicinity of blood vessels is sufficient to induce angiogenesis. These data provide evidence in support of the suggestion that APN in the tumor microenvironment contributes to tumor vascularization.

Tumor growth and progression are regulated by both intrinsic factors and interactions with the stromal microenvironment. By generating and analyzing an APN null mouse genetic model of mammary cancer, we provide evidence for a proangiogenic contribution of APN to tumor vascularization that, in turn, promotes tumor growth.

APN strongly associates with the vasculature in both the normal and the tumorigenic mouse and human mammary gland colocalizing with vascular T-cadherin [Fig. 1; Supplementary Fig. S1; ref. 26]. Loss of APN from the tumor vasculature restricts vessel density, thereby limiting oxygen and nutrient supply. PyV-mT tumors are critically dependent on angiogenesis as PyV-mT–driven tumors expressing MMTV-controlled VEGF show significantly earlier tumor onset than their WT counterparts (34). Vice versa, limited vascularization, as observed in T-cadherin–deficient mice, retards tumor growth (26). Thus, vascular starvation of the tumors in APN-KO mice seems a major cause for the tumor growth retardation and extension of animal survival.

Functions for APN in vascular cells are well established in vitro and in vivo: APN promotes endothelial cell migration and differentiation into tube-like structures in vitro (11), and in vivo accelerates neovascularization after hind limb ischemia (12). We show here in the corneal assay that the major HMW APN isoform is physiologically active and sufficient to stimulate angiogenesis similar to VEGF alone (Fig. 4). APN binding to endothelial cells in vitro activates downstream signal transduction cascades involving adenosine monophosphate–activated protein kinase (11). The specific membrane receptors and associated signaling intermediates in the vasculature, however, remain to be elucidated. Two seven-span APN receptors, AdipoR1 and AdipoR2 (30), and the glycosylphosphatidylinositol–anchored cadherin-type APN-binding protein T-cadherin (35) serve in associating APN with the cellular plasma membrane (33, 36).

We note striking parallels between the APN-KO tumor phenotype with the T-cadherin null MMTV-PyV-mT model (26). T-cadherin is sufficient to bind APN (33) and is required for the vascular association of APN in mammary tumors (26). In T-cadherin null mice, the loss of vascular APN association dramatically increases APN levels in the circulation (26), implicating a critical role for T-cadherin in the cellular association of APN. One possibility is that glycosylphosphatidylinositol–linked T-cadherin is part of an oligomeric protein complex that in combination with AdipoRs serves in APN binding and signaling. The coincident reduction of T-cadherin and AdipoR2 in MMTV-PyV-mT tumors (Fig. 1) is consistent with this suggestion. Alternatively, T-cadherin may be necessary for docking circulating APN to the vasculature, thereby regulating its availability to other vascular receptors that, in turn, activate associated downstream signaling cascades. Although evidence for the direct mechanistic link between APN and T-cadherin in vivo is still lacking, the combined analyses of APN-deficient and T-cadherin–deficient MMTV-PyV-mT mouse models yield strong parallels supporting concerted roles in mammary tumor angiogenesis.

Despite the reduced overall tumor burden and the prolonged life span of APN-KO PyV-mT transgenic mice, autopsies at the experimental end point show increased metastatic spreading compared with controls (Figs. 2C and 3D). One possible interpretation of this result is that pulmonary metastases increase at the same rate in both WT and APN-KO mice, but frequency is enhanced in the KO condition due to longer animal survival. An alternative explanation is that the challenging tumor microenvironment supports an aggressive and lethal tumor phenotype after a honeymoon phase of disease improvement. Indeed, multiple in vitro studies have linked hypoxia to increased tumor cell invasiveness (37–39). Moreover, MMTV-PyV-mT tumors lacking hypoxia-inducible factor–α (HIF-1α) display reduced pulmonary metastases (40). In APN-KO mice, PyV-mT mammary tumors show significantly larger hypoxic areas than in WT animals, thus linking enhanced hypoxic signaling to increased pulmonary metastases. These observations relate to clinical findings that drug inhibitors of the proangiogenic VEGF pathway typically lead to only temporary improvements, and breast cancers resist such treatment after a limited time frame (41). The identification of the alternative APN-induced proangiogenic pathway reported here might open new opportunities for drug development that may help overcome this limitation.

MMTV-PyV-mT transgenic mice reliably represent ErbB2-overexpressing human breast cancer. Numerous studies have shown that MMTV-PyV-mT–induced tumors mimic human ErbB2 carcinomas in terms of biomarker expression, stages of tumor progression, distant metastases, and activation of downstream signaling pathways (15–25). We detect APN solely in association with the tumor vasculature in both the mouse model and in human mammary carcinomas, implicating tumor blood vessels as primary APN targets. The current mouse genetic model consistently identifies a prominent proangiogenic contribution of APN in tumorigenesis. Epidemiologic studies have implicated a tumor protective role of APN in humans (6, 7). The PyV-mT mouse mammary tumor model, however, does not support a cancer protective function. Instead, the current study provides evidence for a novel and unexpected contribution of APN to tumor angiogenesis. Future work will need to address if APN exerts dual functions in tumors that may weigh differently depending on the tumor type. It will also remain an important task to identify changes in the cellular interactions between tumor cells and the stromal microenvironment in relation to serum APN levels. The multiple contributions of APN in cancer and metabolic disease certainly warrant the detailed understanding of the cellular and molecular underpinnings regulating these functions.

No potential conflicts of interest were disclosed.

We thank Dr. Robert Oshima (Burnham Institute for Medical Research, La Jolla, CA) for helpful discussions and suggestions throughout this work and comments on the manuscript; Drs. Norikazu Maeda, Tohru Funahashi, and Yuji Matsuzawa (Osaka University, Osaka, Japan) for generating and generously providing the APN-KO mice; Dr. Simone Codeluppi (Burnham Institute for Medical Research, La Jolla, CA) for advice on the statistical analyses; Dr. Stan Krajewski (Burnham Institute for Medical Research, La Jolla, CA) for assistance with the algorithms for histologic tumor analysis; and Adriana Charbono (Animal Facility, Burnham Institute for Medical Research, La Jolla, CA), and Robbin Newlin, Karen Teofilo, and Ronald Torres (Histology and Molecular Pathology Facility, Burnham Institute for Medical Research, La Jolla, CA) for outstanding technical support to this work.