AEG-1 Promoter–Mediated Imaging of Prostate Cancer

A transcription-based imaging, therapeutic, or theranostic system can be considered for clinical translation if it meets certain criteria such as high tumor specificity, broad application, and minimal toxicity (1, 2). The first two criteria can be met through the choice of a strong and tumor-specific promoter. For example, cancer-specific gene therapy with the osteocalcin promoter, delivered through intralesional administration of an adenoviral vector, caused apoptosis in a subset of patients with prostate cancer (2, 3). We have shown that cancer-specific imaging could be accomplished in experimental models of human melanoma and breast cancer by systemic delivery of imaging reporters under the transcriptional control of the progression elevated gene-3 promoter (PEG-Prom; refs. 1, 4). Here, we describe a nanoparticle-based, molecular–genetic imaging system using the astrocyte-elevated gene-1 promoter (AEG-Prom; ref. 5) for detecting metastases due to prostate cancer, including to bone, for which there is no reliable clinical imaging agent.

AEG-1 was first identified using subtraction hybridization as an upregulated gene in primary human fetal astrocytes infected with HIV-1 (6, 7). Subsequent studies identified AEG-1 as a metastasis-associated gene in the mouse, called metadherin (MTDH; ref. 8), and as a lysine-rich CEACAM1 coisolated gene in the rat, called LYRIC (9). Recent studies in multiple cancers confirm a significant role for AEG-1 as an oncogene (10) implicated in cancer development and progression in many organ sites (11). On the basis of the diverse roles of AEG-1 in tumor progression, including transformation, growth regulation, cell survival, prevention of apoptosis, cell migration and invasion, metastasis, angiogenesis, and resistance to chemotherapy (12), this gene may provide a viable target for developing therapies for diverse cancers. Expression of AEG-1 involves transcriptional regulation through defined sites in its promoter (5). A minimal promoter region of AEG-1 was identified by its association with oncogenic Ha-ras–induced transformation (5). AEG-1 is a downstream target of the Ha-ras and c-myc oncogenes, accounting, in part, for its tumor-specific expression. We have previously shown that AEG-Prom is activated by the binding of the transcription factors c-Myc and its partner Max to the 2 E-box elements of the promoter in Ha-ras–transformed rodent and immortalized transformed astrocyte cell lines (5). AEG-1 interacts with PLZF, the transcriptional repressor that regulates the expression of the genes involved in cell growth and apoptosis (13).

Although molecular–genetic imaging with AEG-Prom should be generally applicable to a variety of malignancies, our initial study performed here was, in part, to demonstrate the utility of this system in a relevant and challenging application, namely, for molecular imaging of prostate cancer. We also focus on prostate cancer because PET with [¹⁸F]fluorodeoxyglucose (FDG), which is the current clinical standard for a wide variety of malignancies, does not work particularly well for this disease (14). Although a variety of new molecular imaging agents for PET with CT (PET/CT) of prostate cancer are emerging, such as [¹⁸F]NaF (NaF) (15, 16), [¹¹C]- and [¹⁸F]choline (17–19), [¹⁸F]FDHT (20), anti-[¹⁸F]FACBC (21), and [¹⁸F]DCFBC (22), some are limited to detecting bone lesions (NaF), have significant overlap with normal prostate tissue (the cholines), or have not yet been extensively tested in the clinic. To maintain relevance to clinical translation, we used a linear polyethyleneimine (l-PEI) nanoparticle to deliver the construct systemically. Nanoparticles composed of l-PEI are being used in a variety of ongoing clinical trials (23–25). We describe AEG-Prom–mediated imaging in tumors derived from PC3-ML cells, a human androgen-independent invasive and metastatic model of prostate cancer (26–28). We show that imaging with AEG-Prom delineates lesions from prostate cancer as well as or with higher sensitivity than FDG- or NaF-PET/CT in this model system.

pPEG-Luc and pAEG-Luc were generated as described previously (4, 29). The firefly luciferase-encoding gene in pAEG-Luc was replaced by the HSV1-tk-encoding sequence from pORF-HSVtk plasmid (InvivoGen) to generate pAEG-HSV1tk. Details of cloning by restriction enzyme digestion and other conditions are available upon request. The plasmid DNA was purified with the EndoFree Plasmid Kit (Qiagen). Endotoxin level was ensured as <2.5 endotoxin units per mg of plasmid DNA.

PC3-ML-Luc (stable transfectants) and PC3-ML were provided by Dr. Mauricio Reginato (Drexel University, Philadelphia, PA). These were cultured in DMEM (Cellgro) supplemented with 10% (vol/vol) FBS and 1% (vol/vol) antimycotic solution (Sigma-Aldrich) and incubated at 37°C, 5% CO_2. PrEC (normal prostate epithelium) cells were provided by Dr. John T. Isaacs (Johns Hopkins School of Medicine, Baltimore, MD). These were grown in keratinocyte serum-free medium (total [Ca²⁺] is 75 ± 2 μmol/L) supplemented with bovine pituitary extract and recombinant EGF (Invitrogen Life Technologies).

The prostate cancer cell lines PC3, PC3-ML, LNCaP, DU-145, ARCaP-E, ARCaP-M, RWPE-1 (primary cells immortalized with HPV-18), and PrEC (primary cells) were plated in 6-well plates (BD Biosciences) at 180 × 10³ to 200 × 10³ cells. Cells were transfected using in-vitro jetPRIME (Polyplus-transfection) according to the manufacturer's instructions. The indicated cells were transfected with Luc under the experimental promoters AEG-Prom, PEG-Prom, and a promoterless empty vector (control) as a pDNA-PEI polyplex. Luminescence was normalized for transfection efficiency by cotransfection with a vector expressing Renilla luciferase, pGL4.74[hRluc/TK] (Promega). After 48 hours of transfection, the expression level of the Luc reporter was measured by the Dual Luciferase Reporter Assay Kit (Promega). Luminescence was normalized for cell number (by μg total protein) using the BCA Protein Assay Kit (Pierce Biotechnology).

The mEbox1 and mEbox2 sites were mutated in the wild-type pAEG-Luc plasmid to generate the pAEG-mEbox1&2-Luc plasmid. The consensus E-box sequences, CACGTG, for mEbox1 and mEbox2, were mutated into AGAGTG and AGATTG, respectively, using the QuikChange Lightening Site-Directed Mutagenesis Kit (Agilent Technologies). The sequences of the forward (F) and reverse (R) primers used for mutagenesis were F: 5′ CCCCGCCCGCCCCAGAGTGACGCCCA and R: 5′ GGACGACCGTGGGTCAATCTGGCGCC. The mutated E-box sequences and the luciferase sequence were confirmed by sequencing (Macrogen USA). PC3-ML cells were transiently transfected with the wild-type and mutated plasmid for the subsequent luciferase assay, as described above.

The plated cells were harvested and lysed using cell lysis buffer (Cell Signaling Technology) supplemented with 1 mmol/L phenylmethylsulfonylfluoride (PMSF; Sigma-Aldrich) with Protease cocktail inhibitor, phosphatase inhibitor (Roche). The whole cellular proteins were separated using 10% SDS-PAGE. For Western blotting, the primary antibodies used were rabbit monoclonal anti-c-Myc (1:1,000; Cell Signaling Technology) and mouse monoclonal anti-β-actin (1:2,000; Sigma-Aldrich). The secondary antibodies used were horseradish peroxidase (HRP)–conjugated polyclonal goat anti-mouse IgG (1:1,000; Dako) and polyclonal swine anti-rabbit IgG (1:3,000; Dako).

Protocols involving the use of animals were approved by the Johns Hopkins Animal Care and Use Committee. Four- to 6-week-old male NOG (NOD/Shi-scid/IL-2Rγ^null) mice were purchased from the Sidney Kimmel Comprehensive Cancer Center's Animal Resources Core (Johns Hopkins School of Medicine). PC3-ML and PC3-ML-Luc cells were expanded over 3 to 5 passages. Cells were harvested and diluted in sterile Dulbecco's PBS lacking Ca²⁺ and Mg²⁺ (Invitrogen Life Technologies). For intravenous injection, mice were administered 1 × 10⁶ PC3-ML cells in 100 μL of sterile Dulbecco's PBS via the tail vein. To ensure hematogenous dissemination, including to the bone, the cells were injected into the left ventricle of the heart (27, 28). For this, intracardiac model mice were anesthetized with ketamine (100 mg/kg) and xylazine (20 mg/kg) and inoculated into the left ventricle with 5 × 10⁴ PC3-ML-Luc enriched or PC3-ML cells in a total volume of 100 μL of sterile Dulbecco's PBS using a 26^3/4-gauge needle. To image the PC3-ML-Luc cells with bioluminescence imaging (BLI), mice were injected intraperitoneally (i.p.) with 100 μL of 25 mg/mL of d-luciferin solution (Caliper LifeSciences), and BLI was performed 20 minutes after the intracardiac injection to detect the distribution of cells. Mice were imaged weekly. Images were acquired on an IVIS Spectrum small animal imaging system (Caliper Life Sciences), and results were analyzed using Living Image software (Caliper Life Sciences). A group of age-matched healthy NOG mice served as a negative control for the prostate cancer metastasis model.

The PC3-ML-Luc cells were further selected for bone-homing tendency. Mice bearing PC3-ML-Luc tumors developed through the intracardiac injection method were monitored for tumor formation by BLI. After 5 weeks, following euthanasia, the femur and tibia of the regions demonstrating clear signal were aseptically dissected. The tumor cells were established in culture by mincing the epiphysis and flushing the bone marrow with 1× PBS (Invitrogen Life Technologies) as described previously (26). The subpopulations of cells selected using a Transwell migration chamber with an 8-μm pore size (BD Falcon) were tested and confirmed for Luc expression as described previously (30), but using 1 mmol/L of d-luciferin, potassium salt (Gold Biotechnology). The radionuclide imaging experiments were performed with the enriched PC3-ML-Luc cell lines.

Low-molecular-weight l-PEI–based cationic polymer, in vivo-jetPEI (Polyplus Transfection), was used for gene delivery. The DNA-PEI polyplex was formed according to the manufacturer's instructions. For systemic delivery, 40 μg of DNA and 4.8 μL of 150 mmol/L in vivo-jetPEI were diluted in endotoxin-free 5% (wt/vol) glucose separately. The glucose solutions of DNA and l-PEI polymer were then mixed together to give an N:P ratio (the number of nitrogen residues of vivo-jetPEI per number of phosphate groups of DNA) of 6:1 in a total volume of 400 μL. The DNA-PEI polyplex was injected intravenously as two 200 μL injections with a 5-minute interval.

In vivo BLI was conducted at 24 and 48 hours after the systemic delivery of reporter genes. Mice were imaged with the IVIS Spectrum. For each imaging session, mice were injected intraperitoneally with 150 mg/kg d-luciferin, potassium salt under anesthesia using a 2.0% isoflurane/oxygen mixture. Ex vivo BLI was conducted within 10 minutes of necropsy. Living Image 2.5 and Living Image 3.1 software were used for image acquisition and analysis.

At 48 hours after injection of pAEG-HSV1tk-PEI polyplex, animals were injected intravenously with 37.0 MBq (1.0 mCi) of [¹²⁵I]FIAU. At 18 to 20 hours after radiotracer injection, imaging data were acquired with the X-SPECT small-animal SPECT-CT system (Gamma Medica Ideas) using the low-energy single-pinhole collimator (1.0-mm aperture). Focused lung and liver imaging were acquired with a radius of rotation of 3.35 cm, and whole-body imaging was undertaken with a radius of rotation of 7.00 cm. Mice were imaged in 64 projections at 45 seconds of acquisition per projection. SPECT images were coregistered with the corresponding 512-slice CT images. Tomographic image datasets were reconstructed with the 2-dimensional (2D) ordered subsets-expectation maximum (OS-EM) algorithm. AMIDE (31) and PMOD (v3.3, PMOD Technologies Ltd.) software were used for image quantification and analysis.

About 9.25 MBq (0.25 mCi) of each imaging agent was injected via the tail vein. Animals were placed on a heating pad and were allowed mobility during the 1-hour radiotracer uptake period. The animals were then subjected to isoflurane anesthesia. Whole-body images were acquired with the eXplore Vista small animal PET scanner (GE Healthcare) using the 250 to 700 keV energy window. Acquisition time was 30 minutes (2 bed positions, 15 minutes per bed position). Mice were fasted for 6 to 12 h before receiving FDG to minimize radiotracer accumulation in non-tumor tissues. FDG and NaF imaging was done between 4 and 5 weeks after injection of PC3-ML-Luc cells. PET images were coregistered with the corresponding 512-slice CT images. Tomographic image datasets were reconstructed with the 3D OS-EM algorithm with 3 iterations and 12 subsets and were analyzed with AMIDE software (31).

After BLI data acquisition at 48 hours after pAEG-Luc-PEI delivery, each organ demonstrating expression of Luc was collected and fixed in 10% neutral buffered formalin. Tissues were embedded in paraffin blocks. Serial paraffin longitudinal sections were stained with goat anti-luciferase polyclonal antibody (Promega) or rabbit anti-Myc polyclonal antibody (Epitomics). HRP-conjugated polyclonal rabbit anti-goat antibody was used as a secondary antibody. HRP activity was detected with 3,3′-diaminobenzidine (DAB) substrate chromogen (EnVision+ Kit, Dako). Consecutive sections of each tissue sample were stained with hematoxylin and eosin (H&E) and were photographed with a Zeiss photomicroscope III.

After imaging experiments, animals were euthanized and their lung and liver tissue were harvested and snap frozen. Total DNA was extracted by using DNeasy Blood & Tissue Kit (Qiagen) following the manufacturer's instructions. One hundred nanograms of purified total DNA form each animal was used as a template. Quantitative real-time PCR was performed in triplicate per template using the inventoried TaqMan Gene Expression Assays (cat. #4331182, Life Technologies) with the FAM dye–labeled primer set for Luc. Reaction conditions were set as 50°C for 2 minutes, 95°C for 10 minutes and 50 cycles of 95°C for 15 seconds, 60°C for 1 minute followed by the disassociation step of 95°C for 15 seconds, 60°C for 15 seconds, 95°C for 15 seconds in a Bio-Rad iQ5 Multicolor Real-Time PCR Detection system (Bio-Rad Laboratories). Data were analyzed by the absolute quantification method using a standard curve by iQ5 v2.0 software (Bio-Rad). Quantified data were normalized relative to the amplification of mouse Gapdh DNA.

A Faxitron MX20 Specimen X-ray system (Faxitron Corp.) with digital exposures of 25 kV, 17 seconds was used. Films were obtained on Kodak Portal Pack Oncology X-ray film (25.4 × 30.5 cm²) for 22 kV, 15 seconds. For gross pathology, bone tissues were fixed in 10% neutral buffered formalin and were decalcified for 2 hours in Decal (Decal Chemical Corp.) and cut in thin slices.

For BLI, error bars in graphical data represent mean ± SD. P < 0.05 was considered to be statistically significant.

To examine the cancer-specific activity of AEG-Prom, we constructed 2 plasmids, pAEG-Luc, expressing firefly luciferase, and pAEG-HSV1tk, expressing the herpes simplex virus type I thymidine kinase (Supplementary Fig. S1). AEG-Prom drives the expression of the imaging reporter genes firefly luciferase (Luc) and HSV1-tk, which enable BLI and radionuclide-based techniques, respectively. Given the high sensitivity and ease of BLI, our initial studies used this modality for proof of concept. The HSV1-tk reporter gene was used, as before (1), to provide a method that has a clear path to clinical translation. The PEG-Prom construct, namely, pPEG-Luc, was generated previously (1) and was used as a comparison for some of the current studies.

Using BLI, we tested the cancer specificity of AEG-Prom and PEG-Prom in different prostate cancer cell lines, including HPV-18–transformed normal immortal prostate epithelial cells (RWPE-1), PC3, PC3-ML, LNCaP, DU-145, AR-CaP-E (metastatic-resistant epithelial clone), AR-CaP-M (metastatic-prone mesenchymal clone), and in the nonmalignant counterpart cells of prostate epithelium. Robust expression from AEG-Prom and PEG-Prom was observed only in the malignant cell lines, whereas promoter activity was negligible in the normal PrEC (Fig. 1A). To elucidate further the role of c-Myc in the activation of AEG-Prom, we engineered an AEG-Prom containing mutations in the 2 E-box elements, to which the c-Myc transcription factor was hypothesized to bind in prostate cancer cells, to produce pAEG-mEbox1&2-Luc, similar to the one reported in the study by Lee and colleagues (Fig. 1B; ref. 5). The mutant pAEG construct, pAEG-mEbox1&2-Luc, consists of AGAGTG and AGATTG in lieu of the consensus CACGTG in the Ebox1 and 2 regions of the promoter, respectively. Those constructs were transiently transfected into PC3-ML cells, and the promoter activities were compared with those of the wild-type AEG-Prom construct, pAEG-Luc. As shown in Fig. 1B, the pAEG-mEbox1&2-Luc is still active in the PC3-ML cell line, although there was a 3-fold reduction in the extent of activation of the AEG-Prom construct. Furthermore, we note that the AEG-Prom activity increases by 6-fold as it goes from a no–c-myc state in the AR-CaP-E cells to a substantial c-myc state in the AR-CaP-M clone (Fig. 1A; Supplementary Fig. S2 shows c-Myc levels). These results indicate that AEG-Prom activity is regulated primarily, but not exclusively, by the c-Myc transcription factor in these cancer cells.

We then tested and compared the specificity of AEG-Prom and PEG-Prom in vivo in a relevant experimental model of prostate cancer. To develop this model, we used 2 human prostate cancer sublines selected from initial metastases of the parental human PC3 cells that targeted the murine lumbar vertebrae, hence ML (metastasis lumbar). We used PC3-ML cells and the luciferase-tagged version of the PC3-ML cells, namely, PC3-ML-Luc (26–28), which were injected either intravenously or directly into the left ventricle of the heart to ensure widespread dissemination, including to the bone. BLI confirmed the presence of widespread metastases to liver, kidney, lung and bone after intravenous injection of PC3-ML-Luc cells (Supplementary Fig. S3). We assumed a similar time course for the development of metastases from the PC3-ML cells that did not express Luc so that we could use them in conjunction with the AEG-Prom–driven system to identify metastatic lesions by BLI. Mice received an intravenous dose of pAEG-Luc-PEI and pPEG-Luc-PEI polyplexes (Fig. 2; Supplementary Fig. S4). Twenty-four and 48 hours after plasmid DNA delivery, BLI revealed AEG-Prom- and PEG-Prom–driven gene expression above background only in the model demonstrating metastasis (Fig. 2B and D) and not in the healthy control group (Fig. 2A and C).

Histologic analysis of the photon-emitting regions within lung for the animals treated with pAEG-Luc or pPEG-Luc showed the presence of tumor and the correlative Luc expression in the cancer models, but not in the controls (Fig. 2E, F, I and J). In the lungs, Luc expression was detected by immunohistochemistry (IHC) from uniformly scattered tumor cells with some forming large, nodular aggregates. Lesions infiltrate capillaries, interstitium, septae, and larger blood vessels (Fig. 2F and J). The kidneys also demonstrated multiple metastases. Tumor replaced all normal tissue except individual glomeruli (indicated by “G” in Fig. 2G and K). Tumor cells in the liver formed multifocal nodules that in some cases demonstrated adjacent necrosis. Necrotic centers (indicated by “N” in Fig. 2H and L) correlated with a lack of Luc expression. We have also shown that expression of c-myc correlates with AEG-Prom–driven Luc expression within tumor (Fig. 2E–G). Similar expression of the c-myc or the Luc genes was not evident in the healthy, control mice.

BLI signal intensity was significantly higher in the prostate cancer group than in controls within lung at both the 24- and 48-hour time points (after administration of pAEG-Luc and pPEG-Luc; P < 0.0001; Fig. 2M). Moreover, at the 48-hour time point, we observed an approximately 2-fold higher Luc gene expression from the AEG-Prom group as compared with the PEG-Prom group (Fig. 2M). To compare the transfection efficiency between the lungs in pAEG-Luc and pPEG-Luc–treated PCa-1-3 (Fig. 2M; Supplementary Fig. S5), we quantified the amount of plasmid DNA delivered to each of the lung tissues. We performed quantitative real-time PCR with a primer set designed to amplify a region of the Luc-encoding gene in the pAEG-Luc and pPEG-Luc plasmids (Supplementary Fig. S5). We used total DNA extracted from the lungs as a template. The difference in transfection efficiency in the prostate cancer lungs between the pAEG-Luc and the pPEG-Luc group was not significant. That confirms that Luc expression from the pAEG-luc–treated prostate cancer models was due to the higher tumor-specific activity of AEG-Prom rather than higher transfection efficiency to malignant tissues. A possible reason for elevated expression using AEG-Prom in vivo includes that a human gene may demonstrate more productive interaction with the human proteins of the Ras signaling pathway such as c-Myc (PEG-Prom is of rat origin and expression is not dependent on these signaling pathways; ref. 4). Alternatively, AEG-Prom expression might be elevated further in vivo as a consequence of epithelial-to-mesenchymal transition (EMT). Further experimentation will be required to determine whether that hypothesis is correct.

To enable reliable formation of metastasis to bone, a tissue prominently involved in human metastatic prostate cancer, we injected PC3-ML-Luc and PC3-ML cells through an intracardiac route (Supplementary Fig. S3A). Once timing for the development of metastases was determined for the luciferase-expressing cells, we then studied metastases due to PC3-ML cells using the pAEG-Luc-PEI polyplex. At 48 hours after plasmid delivery, we observed AEG-Prom–mediated expression of Luc from the PC3-ML models, as shown for PCa-4 and PCa-2 and not from controls (Figs. 3 and 4, respectively). For PCa-4, when imaged 7 weeks after cell injection, BLI was able to detect cancer cells in the left tibia (Fig. 3B), as confirmed by histologic analysis (Fig. 3C). The BLI signal intensity, from deep within the bone, was weak in vivo, likely due to attenuation by living tissues (32).

Ex vivo BLI of PCa-2, when imaged 5 weeks after cell injection, showed the presence of tumor in the lungs, liver, adrenals, and kidneys, as also confirmed by gross pathology, histologic analysis, and Luc IHC (Fig. 4C and D; Supplementary Fig. S6). To study the transfection efficiencies of systemically delivered construct within lung and liver (Fig. 4E; Supplementary Figs. S6 and S7), we quantified the amount of plasmid DNA delivered to each of these tissues. We performed quantitative real-time PCR with a primer set designed to amplify a region of the Luc-encoding gene in the pAEG-Luc plasmid. We used total DNA extracted from the lung and liver as a template. The differences in transfection efficiency between areas of high tumor burden versus those of low tumor burden within liver in same animal, as well as between areas of high tumor burden within liver versus normal liver, were significant at P < 0.0005 and P = 0.0078, respectively. Lower transfection efficiency in diseased versus normal liver was likely due to lower delivery of plasmid to the diseased tissue, as demonstrated previously for PEG-Prom and lung replete with metastases (1). PCR was also performed in tissues from control animals after pAEG-Luc delivery. Differences in transfection efficiency within lungs and liver between the control group and the prostate cancer group were not significant. That confirms that higher visualized Luc expression from the prostate cancer models is due to the tumor-specific activity of AEG-Prom rather than higher transfection efficiency to malignant tissues.

BLI is limited to preclinical studies due to the dependence of signal on tissue depth, the need for administration of exogenous d-luciferin substrate at relatively high concentration for light emission, rapid consumption of d-luciferin leading to unstable signal, and low anatomic resolution (1). Accordingly, we cloned pAEG-HSV1tk (Supplementary Fig. S1B), which can be detected by the radionuclide-based techniques of PET or single-photon emission computed tomography (SPECT), upon administration of a suitably radiolabeled nucleoside analog (33). We examined the SPECT-CT imaging capabilities of pAEG-HSV1tk for detection of bone and soft tissue metastases in the PC3-ML model. Approximately 5 weeks after receiving an intracardiac administration of PC3-ML-Luc cells, the prostate cancer group and the corresponding controls received pAEG-HSV1tk–PEI polyplex. Forty-eight hours after plasmid delivery, mice received the known HSV1-TK substrate, 2′-fluoro-2′-deoxy-β-d-5-[¹²⁵I]iodouracil-arabinofuranoside ([¹²⁵I]FIAU) (29), and were imaged at 18 to 20 hours after injection of radiotracer. Figure 5 shows a representative example, PCa-3, for which we were able to detect the presence of multiple metastatic lesions with the pAEG-HSV1tk system. We then compared the sensitivity of the AEG-Prom imaging system with the current clinical PET-based methods for detecting soft tissue (FDG-PET) and bony (NaF-PET) metastatic lesions due to prostate cancer. Figure 5A and B show representative examples, PCa-3, and a healthy control, Ctrl-1, imaged with each method. Because NaF is a bone-seeking agent, there is substantial uptake within the normal skeleton (34), which may obscure lesions within bone (Fig. 5A). Moreover, on NaF, bone scan skeletal metastases are seen indirectly, depending on the reaction of bone to the lesion, whereas the AEG-Prom polyplex images tumor directly. The NaF-PET/CT study for PCa-3 appears similar to that for Ctrl-1. NaF-PET/CT failed to identify the metastases within the tibia and axial skeleton. The same mouse also underwent FDG-PET/CT, which was only able to identify a lesion in the left scapula (L1, Fig. 5B).

BLI performed ex vivo and gross pathology of lesions within the right humerus, dorsal thoracic wall, ribs, sternum, and the heart confirmed that tumor was the source of signal seen on the living images (Fig. 5C–E and Supplementary Fig. S8, respectively). We were able to identify a 3-mm tumor nodule on the heart (L2, Supplementary Fig. S8B), a 5-mm lesion in the dorsal thoracic wall adjacent to the mid-spine (L3, Fig. 5D), and a 1-mm lesion in the ventral midline of the sternum (L4, Fig. 5D). Furthermore, the macroscopic picture confirmed metastases in the bone marrow within the proximal tibia (L5, Fig. 5E, red dotted circles). An ex vivo plain film image of the pelvis of PCa-3 did not delineate bone lesions clearly (Supplementary Fig. S8C), suggesting that advanced changes in skeletal morphology might be needed for detection with conventional imaging modalities.

Our goal was to develop a systemically deliverable construct for molecular–genetic imaging of metastatic lesions within both soft tissue and bone in a relevant model of prostate cancer. Others have developed in vivo molecular–genetic imaging agents for prostate cancer using adenoviral-mediated, prostate-specific regulatory elements. Native androgen-dependent promoters/enhancers derived from prostate-specific antigen (PSA; ref. 35), probasin (36), human glandular kallikrein2 (37), and the prostate-specific membrane antigen (PSMA; ref. 38) have been used to drive transgene expression, but in a tissue-restricted, rather than a tumor-specific, manner (39). In addition, promoters such as PSES (PSA promoter/enhancer) have been improved by incorporating the TSTA system, a 2-step transcriptional amplification mechanism using the Gal4-VP16 fusion protein to enhance the transcriptional activity of weak PSES (40).

In comparison to the above-mentioned prostate-specific promoters, the tumor-specific promoters of PEG-3 and AEG-1 have certain features that might render them more specific and selective while at the same time instill them with greater use, namely, to use them in a variety of cancers beyond prostate cancer. (i) PEG-Prom and AEG-Prom maintain universal cancer specificity regardless of the tissue of origin, (ii) PEG-Prom and AEG-Prom do not require amplification to achieve high sensitivity. (iii) PEG-Prom and AEG-Prom are systemically delivered using a non-viral delivery vehicle. We note that the expression levels of both the AEG-Prom and PEG-Prom increase in the mesenchymal clone of the ARCaP cell line compared with the epithelial clone of the same cell line. (Fig. 1A), indicating a possible level of involvement in EMT.

To recapitulate the clinical characteristics of prostate cancer metastasis, we implemented a bone metastatic model of prostate cancer, which occurs spontaneously after the intravenous injection of prostate cancer cells without orthotopic injection directly to bone. In animals showing tibial lesions using BLI of PC3-ML-Luc cells, subsequent SPECT/CT was able to detect these lesions in all animals tested (Fig. 5B). In perhaps the example closest to ours, Wu and colleagues previously used the PSES-TSTA bioluminescent vector to identify tibial bone marrow metastases that could not be detected by FDG- or NaF-PET/CT (38). In addition to using a tissue-specific promoter, that study also differed from ours in that an orthotopic tibial prostate cancer model was used, an adenoviral vector served as the delivery vehicle, and TSTA amplification was used to boost the promoter activity by several orders of magnitude as compared with the parental PSES vector (40).

By using a biodegradable polymer, in vivo-jetPEI, we tried to avoid certain problems that may arise when using viral vectors, such as immune-mediated toxicity, inflammation and liver tropism (41). We checked the ability of the non-viral delivery vehicle to provide widespread, systemic dissemination of plasmid by conducting quantitative PCR on sections of lung and liver and compared the transfection efficiency between controls and animals affected with prostate cancer (Fig. 4). Group differences in gene delivery between lung and liver were insignificant. This study also confirmed our earlier results that nanoparticle delivery is most efficient to well-vascularized tissues (1). Liver tissue sections with a high tumor burden had significantly lower (P < 0.005) delivery of plasmid DNA, possibly due to the reduced vasculature of this tissue, which was also likely under high hydrostatic pressure and was adjacent to necrotic areas. Although imaging was mediated through activation of AEG-Prom, delivery was, in part, mediated through the enhanced permeability and retention (EPR) effect, interaction of positively charged DNA–PEI polyplex with the cell membrane followed by endocystosis, release from endosomes, and entry into the nucleus (42).

The methods used in this report are intended to enable rapid clinical translation. Accordingly, for systemic delivery, we used a non-viral delivery vehicle, which has seen clinical use (l-PEI; ClinicalTrials.gov #NCT01435720), and have used an imaging reporter gene/probe pair (HSV1-tk/FIAU) that has previously been used in patients (33). Nevertheless, several hurdles must be overcome, arguably the most significant of which is the delivery of the nanoparticles to the malignant tissue. Several excellent reviews on that topic have recently been published (43–45), with recognized obstacles including tumor heterogeneity, elevated interstitial fluid pressure, shifting properties of the microenvironment, and the difficulty of translating optimized conditions for animal systems to the clinic (46, 47). Strategies for enhancing tumor delivery include surface functionalization by affinity agents, particularly agonists that promote internalization (48), and surface coating with polyethylene glycol to increase circulation times. Other aspects requiring optimization include the reporter gene/probe pair, assuring that the gene is nonimmunogenic and that the probe has pharmacokinetics suitable for detection using widely available imaging modalities (49).

AEG-Prom enables a sensitive method for molecular–genetic imaging of prostate cancer in vivo. From mutational analysis of AEG-Prom, we have shown that its activation relies significantly on c-Myc binding to the 2 E-box elements as discussed above. As Ras-mediated c-Myc signal transduction is a pathway present in nearly all malignancies yet is absent in normal tissue (50), we anticipate that AEG-Prom will enable imaging of a wide variety of cancers directly and specifically.

No potential conflicts of interest were disclosed.

Conception and design: A. Bhatnagar, Y. Wang, P.B. Fisher, M.G. Pomper

Development of methodology: A. Bhatnagar, Y. Wang, M. Gabrielson, I. Minn

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Bhatnagar, R.C. Mease, M. Gabrielson, P. Sysa, I. Minn, G. Green, B. Simmons, K. Gabrielson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Bhatnagar, Y. Wang, M. Gabrielson, I. Minn, K. Gabrielson, S. Sarkar, M.G. Pomper

Writing, review, and/or revision of the manuscript: A. Bhatnagar, Y. Wang, R.C. Mease, M. Gabrielson, P. Sysa, I. Minn, K. Gabrielson, S. Sarkar, P.B. Fisher, M.G. Pomper

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Bhatnagar, R.C. Mease

Study supervision: A. Bhatnagar, I. Minn, K. Gabrielson, M.G. Pomper

The authors thank S. Nimmagadda and X. Guo for technical support (Johns Hopkins University).

The study was supported by the A. David Mazzone Research Awards Program of the Prostate Cancer Foundation (P.B. Fisher, M.G. Pomper, George Sgouros), CA151838 (M.G. Pomper), CA058236 (William Nelson), Patrick C. Walsh Foundation (M.G. Pomper), the National Foundation for Cancer Research (P.B. Fisher).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.