Efficacy of Affibody-Based Ultrasound Molecular Imaging of Vascular B7-H3 for Breast Cancer Detection

Breast cancer is the second leading cause of cancer-related deaths and the most common site for cancer development at 30% of all new cancer cases in women in the United States, with an estimated 41,760 deaths and 268,600 new cases diagnosed in 2019 (1). This incidence is expected to grow by more than 50% by 2030 (2). Detection during early, localized stages of the disease significantly improves survival with 5-year survival rates of 99% compared with 27% for highly advanced stages (3).

Mammography is the first-line method in breast cancer screening programs. Clinically detected lesions have a median size of 2.6 cm, whereas those found with mammography screening have a median size of 1.5 cm (4), with digital mammogram analyses further increasing screening sensitivity (5). However, mammograms often result in overdiagnosis and unnecessary biopsies with one-half of the women experiencing false positives during the course of multiple screenings (6). Increased frequency of mammography exams along with factors such as age and breast density lowers the overall specificity of screening and results in more false positives. Women with extremely or heterogeneously dense breasts have a 4- to 6-fold greater chance of developing breast cancer compared with women with fatty breasts (7). Furthermore, women with dense breasts frequently have worse prognosis due to late-stage disease detection (8, 9). Mammographically dense tissues decrease detection of malignant lesions because dense tissues appear opaque, which can obscure or mimic malignant lesions and to some extent calcifications, and mislead even experienced radiologists.

Currently, alternative imaging tools are applied whenever mammography alone is insufficient for radiological detection of breast lesions. MRI is used to screen high-risk candidates with familial history of breast cancer, but is expensive, not readily available, and is not suitable for everyone. At present, supplemental screening by ultrasound (US) is recommended to assess suspicious lesions observed in mammograms and is available for high-risk patients with contraindications to MRI (10). US is widely available, portable, noninvasive, cost-effective, and free of ionizing radiation. In a report, supplemental screening with US detected 4.4 additional cancers per 1,000 exams in women with dense breasts and negative mammography, but it also increased false positives (11). Although US is helpful in detecting breast lesions, it is associated with low positive predictive value (5.6%–8.6%; refs. 12, 13) and a low sensitivity (as low as 17%; ref. 14). As a supplemental tool, molecular imaging with US contrast agents would enable highly accurate and sensitive noninvasive methods of differentiating cancer from benign lesions in women with dense breasts.

Contrast-mode US (also called contrast-enhanced US) using microbubbles (MB) as blood-pool agents is often applied to assess abnormal tissue perfusion patterns in tumors due to altered angiogenesis (15, 16). For example, Bar-Zion and colleagues used contrast-enhanced US imaging to monitor tumor perfusion changes after treatment by a vascular disrupting agent in a breast cancer mouse model (16). In addition, molecularly targeted contrast MB that bind to proteins expressed on the tumor neovasculature is an emerging imaging approach with large potential for improving diagnostic accuracy and characterization of focal breast lesions (17, 18).

MB are gas-filled lipid- or protein-stabilized particles and enhance US imaging contrast due to differential response to changes in acoustic pressure. Gas-filled MB that are a few micrometers in size (1–4 μm) remain within the blood vessel lumen, which renders molecularly targeted MB uniquely suitable for US imaging of molecular markers expressed on the neovasculature (18–21). It is of high importance to identify molecular markers that are differentially expressed on tumor-associated neovasculature compared with that of normal tissue or benign breast lesions. Although several cancer-specific vascular markers have been used in contrast-enhanced US imaging of tumors (18, 19), our group and others have extensively explored B7-H3 (CD276) as a promising breast cancer US molecular imaging target. B7-H3, a T-cell modulator, has highly specific overexpression on vascular endothelial cells of different subtypes of human breast cancer and high-grade ductal carcinoma in situ (DCIS) compared with normal breast tissue and benign lesions (22, 23). B7-H3 is a cell-surface receptor protein and highly correlated with tumor drug resistance, metastasis, and immune-regulation (24–26).

Monoclonal antibodies produced by hybridoma technology have primarily been used as target-binding ligands to functionalize the surface of MB (27), although small protein scaffolds have been used as well for other targets (28, 29). Previously, an anti–B7-H3 antibody-conjugated MB was utilized for preclinical US molecular imaging of mammary tumors (22). In this study, US molecular imaging signal with antibody-conjugated MB was shown to be highly correlated with pathology-based B7-H3 expression in the mammary tumors. Unfortunately, antibody-based ligands can be problematic for clinical translation due to inefficient and random conjugation, costly and time-consuming production, and potential immune response, especially with repeated dosing (30). Smaller protein scaffolds enable efficient, site-specific conjugation and prokaryotic production and can replace antibodies (∼150 kDa) for MB functionalization. Recently, affibodies (ABY) have been shown to be a promising platform for designing binding ligands for molecular imaging (31, 32). The ABY is a 58-amino-acid protein (∼7 kDa) derived from the three alpha helix bundle Z domain of Staphylococcus aureus protein A. Compared with antibodies, ABY exhibit faster renal clearance (33), greater stability at a wide range of physiologic pH and temperature, are cost-effective for larger-scale production, and enable site-specific conjugation (34). ABY molecules are an alternative to full-length antibodies in diagnostic applications and have been validated preclinically and translated to clinical trials (35, 36).

The purpose of this proof-of-principle study is to utilize a recently engineered B7-H3–specific ABY protein (ABY_B7-H3) conjugated to contrast MB (37), to validate their use in vitro for endothelial cell binding under flow shear stress conditions, and to assess specificity and sensitivity for US molecular imaging of tumor neovasculature in two different mouse models of breast cancer: human xenografts and transgenic models.

Human breast cancer samples for tissue staining were collected as described elsewhere (22) and were obtained with written-informed consent and Institutional Review Board approval. All studies involving human subjects were conducted in accordance with ethical guidelines such as the Declaration of Helsinki. RNA sequencing (RNA-seq) expression data for CD276 (B7-H3) and PECAM1 (CD31) in patient breast-invasive carcinoma were downloaded from The Cancer Genome Atlas (TCGA) database and compared for PAM50 gene-expression–based intrinsic breast cancer subtypes using UCSC Xena Browser (38).

ABY_B7-H3 ligand was recombinantly expressed in Escherichia coli and purified by a HisTrap FF column (GE Healthcare Biosciences). For further details on production of ABY and its B7-H3–binding activity, please refer to Supplementary Methods.

Commercially available perfluorocarbon-filled, lipid-shelled, streptavidin-coated contrast MB (VisualSonics) with a mean diameter of 1.5 μm (range, 1–3 μm) were reconstituted in 1 mL sterile saline (0.9% sodium chloride; ref. 39). Three types of MB were prepared: two MB with the ability to target both hB7-H3 (for xenograft tumor models) and mB7-H3 (for transgenic tumor models), made by either conjugating MB with biotinylated ABY_B7-H3 (MB_ABY-B7-H3) or a commercially available biotinylated anti–B7-H3 antibody (eBiosciences, clone M3.2D7, MB_Ab-B7-H3), and unconjugated MB (MB_Non-targeted).

To create each targeted MB, lyophilized streptavidin-coated MB were suspended in 1 mL of sterile saline according to the published recommendations (40). Six microgram of biotinylated ABY or antibody were incubated per 5 × 10⁷ MB in 50 μL for 10 minutes at room temperature and washed before use in in vitro and in vivo experiments (19). To separate unconjugated ligands, the MB solution vial was centrifuged at 300 g in 4°C for 2 minutes followed by removal of clear liquid from the bottom of the vial using a syringe needle without disturbing the top MB layer. MB were then resuspended in PBS and washed similarly two more times (19). For the confirmation of ABY conjugation to MB, biotin-ABY was also conjugated to AF647 NHS ester (ThermoFisher Scientific) dye, purified, and then incubated with streptavidin MB for 10 minutes. Nonbound ligands were removed by centrifugation. ABY-coated MB were then assessed for AF647 signal by flow cytometry (Guava easyCyte, Luminex). The average number of biotinylated-ABY or -antibody, when occupying all the streptavidin molecules in our tested concentrations, are approximately 7,600 per square micrometer MB surface (22).

Binding specificity of MB_ABY-B7-H3 and MB_Ab-B7-H3 to the target B7-H3 was first assessed in cell culture experiments under flow shear stress conditions simulating flow in blood capillaries by using a flow chamber experimental set-up. Please see Supplementary Methods for more details.

All experiments were approved by the Institutional Administrative Panel on Laboratory Animal Care. An orthotopic human breast cancer model with tumors from MDA-MB-231 cells stably expressing firefly luciferase (f-luc) reporter gene mixed in matrigel with either MS1_WT or MS1_hB7-H3 cells expressing renilla luciferase (r-luc) reporter gene was used for establishing tumors in the contralateral flanks of nude (nu/nu) mice (The Jackson Laboratory). Reporter expression details are provided in the Supplementary Document. Note that 1 × 10⁶ MDA-MB-231/f-luc cells were mixed with 5 × 10⁶ MS1/r-luc cells and implanted on the fourth mammary glands: MS1_hB7-H3/r-luc cell mixture on the left flank and MS1_WT/r-luc cell mixture on the right flank. Imaging was performed in the nude mice after 2 weeks of tumor cell engraftment with a mean size of 4 mm (range, 3–5 mm).

In addition, the transgenic mouse model of breast cancer development FVB/N-Tg(MMTV-PyMT)634Mul was used (The Jackson Laboratory; ref. 22). The mammary tissue of this transgenic mouse model progresses through four distinct histologic stages from normal mammary tissues, hyperplasia, ductal carcinoma in situ, and finally invasive breast carcinoma, which highly recapitulates human breast cancer. For this study, female mice (mean age, 7 weeks; range, 4–10 weeks) with 10 mammary glands with invasive breast carcinoma were imaged with a mean size of 7 mm (range 5–9 mm) by US molecular imaging. The litter mates with normal mammary glands were used as controls.

Mice coinjected with MDA-MB-231/f-luc cells and MS1/r-luc cells were tested for successful implantation by bioluminescence imaging in live animals after 2 weeks of engraftment. Mice were subjected to i.p. injection of 50 μL D-luciferin (30 mg/mL) substrate followed by anesthesia in 2% isoflurane in room air, and bioluminescence imaging (Lago in vivo Imaging System, Spectral Instruments Imaging) to confirm the growth of MDA-MB-231 tumors. After a 24-hour interval, anesthetized mice were injected with 150 μL of coelenterazine (5 mg/mL) substrate via tail vein injection followed by bioluminescence imaging to confirm the presence of MS1 cells within these tumors.

Contrast-enhanced US imaging was performed using a dedicated small-animal high-resolution US imaging system (Vevo 2100; VisualSonics). All mice were kept anesthetized with 2% isoflurane in room air at 2 L/min on a heated stage at 37°C throughout the US molecular imaging sessions. Image acquisition was performed in the transverse plane using a high-resolution transducer (MS250; center frequency, 18 MHz, lateral and axial resolution of 165 μm and 75 μm, respectively; focal length, 8 mm; transmit power, 10%; mechanical index, 0.2; dynamic range, 40 dB). Imaging was performed by fixing the transducer with a clamp and placing the acoustic focus at the center of the mammary tumors in the plane showing the largest transverse cross-section. The same US settings and equipment were used for all imaging experiments.

Recommendations for using targeted MB (MicroMarker, VisualSonics) for small animal imaging were followed as described elsewhere (40). US molecular signal (i.e., US contrast signal from vessel-bound MB_ABY-B7-H3) was obtained by the destruction–subtraction technique (41), and its signal specificity confirmed by comparisons with a positive control, MB coated with anti–B7-H3 antibody (MB_Ab-B7-H3), and a negative control, nonfunctionalized MB (MB_Non-targeted). Before i.v. injections with a catheter (12 cm PU tubing and 27G butterfly needle; VisualSonics), MB concentration was determined in a particle counter (Z2 Particle Counter, Beckman Coulter). A small sample volume of MB solution was diluted in 10 mL diluent (Isoton II; Beckman Coulter) and concentration determined based on a size range calibration of 1–5 μm. A total of 5 × 10⁷ MB were mixed in approximately 100 μL final volume/injection and bolus administered i.v. through the tail vein followed by another 20 μL saline flush. MB injection was performed manually at a low constant pressure (100 μL in 6 seconds as recommended by the manufacturer) and tissue contrast signal bias between animals minimized by random injection order of different MB constructs. Targeted MB were allowed to attach to B7-H3 on the tumor neovasculature. After 4 minutes, 200 imaging frames were captured and averaged to obtain imaging signal from adherent and freely circulating MB. This was followed by a 1-second continuous high-power destructive pulse of 3.7 MPa (transmit power, 100%; mechanical index, 0.63), which destroyed all MB within the image. After the destruction pulse, another 200 imaging frames were acquired and averaged to capture the signal from the replenishment of freely circulating MB into the tumor area. To determine US molecular imaging signal, the averaged images before and after bursting MB were subtracted as differential targeted enhancement (dTE). dTE image, representing B7-H3 signal, was displayed as a colored overlay on the contrast-mode. The time between injection and imaging with different MB constructs was 20 minutes to allow for freely circulating MB to clear from any previous injections. In addition, destruction–replenishment curves for MB-associated contrast signal before and after the ultrasonic destructive pulse were formed by perfusion modeling in which amplitude-related parameters are expressed relative to plotting as a function of time expressed in seconds.

All US data analyses, including breathing motion correction, defining the region of interest in mammary tissue, dTE, and destruction–replenishment curves were performed in Vevo LAB (Visual Sonics) software (22).

The Student t test was used to compare statistical significance between the experimental groups, and all data were expressed as mean ± SEM. Experiments were considered significantly different if the P value was less than 0.05. ROC curve was generated (Prism 8, GraphPad) to determine the diagnostic accuracy of MB_ABY-B7-H3 in differentiating normal tissue from malignant tumors by US molecular imaging.

To compare gene expression of CD276 (B7-H3) in patient breast cancer samples against a pan-endothelial marker, PECAM1 (CD31), RNA-seq expression data across PAM50 molecular breast cancer subtypes in the TCGA database were analyzed using the UCSC Xena web-based tool (38). Analysis showed higher CD276 expression in all breast cancer subtypes compared with the normal tissue, whereas the PECAM1 expression was higher in normal tissue compared with cancer tissue (Fig. 1A). As B7-H3 is expressed by both vascular endothelial cells and neoplastic cells, we tested its protein expression in patient tissue sections by IHC staining. In the representative samples, B7-H3 staining was observed in structures morphologically resembling blood vessels as well as the cancer cells independent of disease subtypes as seen by its increased staining in luminal, triple-negative, and Her2⁺ breast cancer samples compared with normal mammary tissue (Fig. 1B; ref. 22). Based on endothelial expression of B7-H3 in clinical tissue samples, experimental plans for clinically translatable targeted MB contrast agent development and its use in contrast-enhanced US imaging of mouse models of breast cancer were formulated as the work flow in Fig. 1C.

Development of a B7-H3–binding ligand from an ABY yeast display library is described elsewhere (37). Here, an ABY ligand clone termed ABY_B7-H3 with a K_d of 310 ± 100 nmol/L is used. ABY_B7-H3 ligand was recombinantly expressed in E. coli, purified, and functionalized with approximately 1:1 molar ratio of biotin (Supplementary Fig. S1). SDS-PAGE analysis of biologically produced ABY showed the predicted size of the ABY (expected: 7,490 Da), which was measured to be 7,546 Da by mass spectrometry (Fig. 2A). No impurities were observed in the purified ABY by SDS gel or mass spectrometry. In order to confirm binding specificity of recombinantly produced ABY_B7-H3 to its target, biotinylated-ABY was immobilized to streptavidin-magnetic beads and incubated with 3 μmol/L of recombinant soluble human B7-H3 ectodomain protein. Unconjugated beads (negative control) showed 1% binding compared with ABY_B7-H3–conjugated beads, which showed 68% binding to the soluble B7-H3. The binding of ABY_B7-H3 is comparable to the beads conjugated to biotinylated anti–B7-H3 antibody positive control, which had 81% binding (Fig. 2B).

ABY_B7-H3 binding to cells predominantly expressing 4Ig-B7-H3 (hB7-H3) isoform and mouse cells expressing 2Ig-B7-H3 (mB7-H3) isoform (42) was tested. B7-H3 total protein as well as cell-surface expression was confirmed in MS1 mouse endothelial cells engineered to overexpress hB7-H3 (MS1_hB7-H3) by Western blotting and flow cytometry (Fig. 3A and B); only MS1_hB7-H3 cells expressed B7-H3 protein, whereas both MS1_WT and MS1_hB7-H3 cells endogenously expressed CD31 protein, a common vascular endothelial marker, by Western blot. Biotinylated ABY_B7-H3 bound specifically to MS1_hB7-H3 cells but not to MS1_WT cells as detected by streptavidin-AF647 dye in flow cytometry (Fig. 3B). Similarly, ABY_B7-H3 binds to the mouse monocyte cell line, RAW264.7, which endogenously expresses moderate levels of cell-surface mB7-H3 (Fig. 3C). ABY_B7-H3 dose-dependent increases in binding signal in both MS1_hB7-H3 (0–10 μmol/L) and RAW264.7 (0–25 μmol/L) cells were observed, but not in the B7-H3–negative cells, MS1_WT (Supplementary Fig. S2).

Because ABY_B7-H3 binding to endogenously expressed hB7-H3 is critical for its clinical translation, a human monocytic cell line, THP1, was chemically induced for hB7-H3 expression by phorbol myristate acetate (PMA) to test for ABY binding (43). THP1 cells express a basal level of cell-surface hB7-H3, which was further increased upon stimulation by PMA (10 ng/mL) as detected by APC-conjugated anti–B7-H3 antibody using flow cytometry (Fig. 3D). The increase in PMA-induced B7-H3 expression correlated with increased biotinylated ABY_B7-H3 binding to THP1 cells as detected by streptavidin AF647 (Fig. 3D).

As posttranslational modifications, including glycosylation of tissue-expressed B7-H3 receptors (44), can limit accessibility of engineered ligands or antibodies to the receptor binding pockets, ABY binding was tested in human breast tissue sections by immunofluorescence staining. ABY_B7-H3 conjugated to AF647 dye stained positive for B7-H3 expressed on the cell surface on the tumor tissue section (Fig. 3E). In contrast, ABY_B7-H3 did not bind to normal breast tissue sections. In addition, ABY_B7-H3 IHC staining of human breast carcinoma tissue sections positively stained tumor epithelia and blood vessels that were consistent with anti–hB7-H3 antibody staining results (Supplementary Fig. S3).

To mimic the in vivo shear stress that would occur on bound MB by blood flow in the capillaries, an in vitro flow chamber cell attachment assay of MB using MS1_hB7-H3 cells was performed (Fig. 4A). Biotinylated ABY was conjugated to streptavidin MB (MB_ABY-B7-H3). A positive control group was included with biotinylated anti–B7-H3 antibody-conjugated MB (MB_Ab-B7-H3), which were validated in previous studies (22). Unconjugated MB (MB_Non-targeted) served as negative control for quantification of nonspecific binding to cells. The number of MB_ABY-B7-H3 (8.5 ± 1.4 MB/cell) and MB_Ab-B7-H3 (9.8 ± 1.3 MB/cell) attached to MS1_hB7-H3 cells was significantly higher (P < 0.0001) compared with the MB_Non-targeted (0.5 ± 0.1 MB/cell; Fig. 4B and C). In contrast, a low number of all MB constructs attached to MS1_WT cells (Supplementary Fig. S4). There was no significant difference between MB_ABY-B7-H3 (1.5 ± 0.3 MB/cell; P < 0.00001 vs. MS1_hB7-H3) and MB_Non-targeted (1.3 ± 0.3 MB/cell) attachment to the MS1_WT cells. Furthermore, to determine the specificity of MB_ABY-B7-H3 to the cell-surface B7-H3, a receptor-blocking study was performed by incubating MS1_hB7-H3 cells with free ABY_B7-H3 (5 μg/mL) prior to the cell attachment assay with MB_ABY-B7-H3. Blocking significantly (P < 0.009) decreased the MB_ABY-B7-H3 attachment to cells (10.9 ± 2.9 MB/cell) compared with cells without blocking (26.3 ± 4.4 MB/cell; Fig. 4D). Overall, these results suggest that MB_ABY-B7-H3 is capable of binding specifically to human B7-H3–expressing cells under flow shear stress conditions.

To evaluate binding of MB_ABY-B7-H3 to both human and murine B7-H3 in vivo, both an orthotopic human breast cancer xenograft expressing hB7-H3 on its vasculature as well as a transgenic mouse model of spontaneous breast cancer expressing mB7-H3 on its neovasculature were used. US molecular imaging of tumors was conducted with all MB constructs in mice. After 2 weeks of tumor and endothelial cell coengraftments in nude mice, in vivo bioluminescence imaging confirmed signal from MDA-MB-231/f-luc cancer cells from both flanks, MS1_hB7-H3/r-luc endothelial cells on the left flank, and MS1_WT/r-luc endothelial cells on the right flank (Fig. 5A). Luciferase reporter gene introduction into MS1 cells did not alter B7-H3 expression, and bioluminescence activity was confirmed in vitro prior to in vivo use (Supplementary Fig. S5). US molecular imaging of the left tumors, consisting of MS1_hB7-H3 cells, produced significantly higher molecular imaging signal with MB_ABY-B7-H3 (8.4 ± 3.3 a.u.; n = 12 tumors, *, P < 0.04) compared with the MB_Non-targeted (1.4 ± 1.0 a.u., n = 14 tumors). The positive control MB_Ab-B7-H3 also produced high molecular imaging signal (8.2 ± 1.3 a.u.; n = 6, *, P < 0.001 vs. MB_Non-targeted), similar to that from MB_ABY-B7-H3 (Fig. 5B and C). In order to confirm that the B7-H3 molecular imaging signal quantified by the dTE method was from the stationary MB attached to the tumor vasculature, we examined the destruction–replenishment curve for each contrast agent. MB-associated echo power after the ultrasonic bursts showed a decrease only in the MB_ABY-B7-H3 imaging group of tumors consisting of MS1_hB7-H3 cells (Fig. 5C). In the tumors consisting of MS1_WT cells on the right flanks, the molecular imaging signals with MB_ABY-B7-H3 (1.6 ± 0.6 a.u.), MB_Ab-B7-H3 (3.1 ± 0.9 a.u.), and MB_Non-targeted (0.6 ± 0.2 a.u.) were low. Of all the MB constructs tested for molecular imaging signal in the tumors with MS1_WT cells, MB_Ab-B7-H3 produced the highest background imaging signal (*, P < 0.001 vs. MB_Non-targeted). MB_ABY-B7-H3 (**, P < 0.05) and MB_Ab-B7-H3 (**, P < 0.008), but not MB_Non-targeted, produced higher imaging signal in tumors consisting of MS1_hB7-H3 cells compared with their respective MB constructs in the tumors consisting of MS1_WT cells (Fig. 5B and C). Immunofluorescence costaining with anti-mouse CD31 and anti-human B7-H3 antibodies confirmed the integration of MS1 cells on the mouse blood vessels as well as nonvascular compartments of the engrafted tumors (Fig. 5D). hB7-H3 expression was absent on the tumors engrafted with MS1_WT cells.

After evaluation in the xenograft breast cancer model, we also tested targeted MB in a MMTV-PyMT transgenic mouse model that spontaneously develops mammary tumors, which progress into highly invasive disease over time (22). MB_ABY-B7-H3 (9.6 ± 2.0 a.u.; n = 47 tumors, *, P < 0.0002) and MB_Ab-B7-H3 (7.2 ± 1.8 a.u.; n = 45 tumors, *, P < 0.001) produced significantly higher US molecular imaging signal from mammary tumors compared with the MB_Non-targeted (1.3 ± 0.3 a.u.; n = 41 tumors; Fig. 6A and C). In vivo blocking of the mB7-H3 receptors in the transgenic mice with free ABY (150 μg) 24 hours prior to tumor imaging resulted in significantly reduced molecular imaging signal (0.4 ± 0.2 a.u.; n = 5, *, P < 0.02) from tumors compared with the same animals imaged before blocking (4.3 ± 1.0 a.u.) with the MB_ABY-B7-H3 (Fig. 6B and C). Destruction–replenishment curve analysis showed a decrease in MB-associated echo power after the ultrasonic bursts only in the tumor imaging groups with MB_ABY-B7-H3 and MB_Ab-B7-H3 but not in the MB_Non-targeted or blocking groups (Fig. 6C). Also, as a negative control for nonangiogenic B7-H3–negative blood vessels, normal mammary glands in age-matched control littermates were scanned after intravenous administration of MB_ABY-B7-H3, MB_Ab-B7-H3, and MB_Non-targeted. US imaging signal in normal mammary gland was low with all MB constructs [MB_ABY-B7-H3 (2.3 ± 0.5 a.u.; n = 14 glands), MB_Ab-B7-H3 (2.5 ± 0.8 a.u.; n = 10 glands), and MB_Non-targeted (2.6 ± 0.6 a.u.; n = 15 glands); Fig. 6D]. B7-H3 expression on the CD31-positive vasculature of the transgenic mammary tumors was confirmed by immunofluorescence staining with anti–B7-H3 antibody (Supplementary Fig. S6). B7-H3 was not expressed in CD31-positive vasculature in the tissue sections derived from normal mammary glands. ROC analysis was performed to assess the ability of MB_ABY-B7-H3 to distinguish invasive cancer (n = 47) from the normal breast tissue (n = 44) by US molecular imaging. Overall, MB_ABY-B7-H3 contrast agent allowed the differentiation of normal tissue from breast cancer with high diagnostic accuracy, with an area under the ROC curve (AUC) of 0.90 [95% confidence interval (CI), 0.82–0.95; Fig. 6E]. These results confirm that ABY is specific to both human and murine B7-H3 expressed on tumor endothelial cells in vivo and that MB_ABY-B7-H3 provides B7-H3–specific molecular imaging signal of breast tumors compared with the normal mammary tissue.

The purpose of this study was to develop a clinically translatable contrast agent for US molecular imaging of breast cancer. An engineered ABY protein ligand for a breast cancer–associated vascular marker, B7-H3, was applied to MB-based US molecular imaging for mammary tumor detection in mouse models. In vitro, ABY_B7-H3 binds specifically to its soluble and cell-surface target overexpressed on the endothelial cells and mouse RAW264.7 or human THP1 monocyte cells that are known to endogenously express B7-H3. ABY_B7-H3 also binds specifically to B7-H3–positive tumor epithelial and endothelial cells of human breast carcinoma samples. Targeting of MB contrast agent by biofunctionalization with surface conjugation of ABY_B7-H3 significantly increased their attachment to endothelial cells under flow shear stress conditions. In vivo, B7-H3 targeting of MB (MB_ABY-B7-H3) significantly improved the blood vessel–associated US molecular imaging signal in tumors but not in the normal mammary glands. In vivo blocking of B7-H3 receptor significantly reduced the molecular imaging signal achieved with MB_ABY-B7-H3 further validating its molecular specificity. The sensitivity and specificity of this molecular contrast agent is encouraging for clinical translation for US detection of vascular B7-H3–expressing breast tumors, which could increase the sensitivity of US as a complementary imaging modality for accurate diagnosis of breast cancer in women, including those with mammographically dense breasts.

In recent years, US molecular imaging of tumor angiogenesis and inflammatory processes using MB-based contrast agents has made significant progress in biomarker-based detection of underlying pathology (45). For this study, B7-H3 (CD276) was selected as a biomarker for US molecular imaging as the expression of B7-H3 in tumor endothelial and epithelial compartments of breast cancer has been extensively reported in the literature (22, 23, 26, 46). In our analysis, B7-H3 expression in the blood vessels as well as the cancer cells in human tissue sections of various breast cancer subtypes was upregulated compared with normal breast tissues (22). As B7-H3 expression is a subtype-independent biomarker of breast cancer angiogenesis and is downregulated in benign tumors of the breast tissue, it is an ideal biomarker for the development of a clinically translatable US contrast agent. As an example, a targeted MB contrast agent against another validated vascular marker, VEGFR2/KDR, has been used in antiangiogenic therapy monitoring of colon cancer (47) and recently tested in the first-in-human trials (48) with US signal correlating with histologic VEGFR2 expression of patient tissue sections of breast and ovarian cancer. The current work aims to expand the capacity for molecular profiling of breast cancer.

Previously, an US contrast agent composed of MB functionalized with anti–B7-H3 antibody improved US imaging signal of mammary tumors expressing B7-H3 in the vasculature (22). Instead of an antibody, the use of ABY is economical for large-scale synthetic production, which is a major advantage over the costs associated with development and production of humanized monoclonal antibodies. ABY have high stability, solubility, and ability to withstand high temperatures (90°C) or acidic and alkaline conditions (pH 2.5 and pH 11, respectively; ref. 37). Moreover, the small, single-domain architectures of ABY allow for efficient site-specific chemical conjugations via incorporation of terminal amino acids such as a cysteine to functionalize contrast MB. ABY_B7-H3 was developed based on multiple criteria encompassing affinity, specificity, solubility, and thermal stability that are crucial for its optimal function in vivo. ABY scaffolds are of clinical value for biomarker detection and safe for imaging use in humans (32). ABY_B7-H3 recognized both exogenously overexpressed human B7-H3 in endothelial cells as well as endogenously expressed mouse/human B7-H3 by monocyte cell lines in our in vitro experiments. ABY_B7-H3 recognized B7-H3 expressed in tissue sections of patients with breast cancer. MB_ABY-B7-H3 bound specifically to the endothelial cells expressing human B7-H3 under flow shear stress conditions in vitro and tumor blood vessels of breast cancer mouse models. A crucial requirement for targeted MB is that they not only bind to the desired target but bind under shear stress from the expected forces on the MB from blood flow in capillaries. We have shown in an in vivo murine model of breast cancer that MB_ABY-B7-H3 enhanced US molecular imaging signal of MDA-MB-231 orthotopic tumors consisting of MS1_hB7-H3 but not those consisting of MS1_WT cells. MB_ABY-B7-H3 showed lower nonspecific imaging signal compared with MB_Ab-B7-H3 in tumors implanted with MS1_WT cells suggesting its ability to reduce background contrast signal in US. Immunofluorescence staining of tumor sections showed hB7-H3 staining in the CD31-positive tumor endothelium indicating incorporation of MS1_hB7-H3 cells in angiogenic vessels during the growth of orthotopic tumors. Furthermore, MB_ABY-B7-H3 significantly increased imaging signal of spontaneously developed mammary tumors in transgenic mice but not the normal glands of control mice. Immunostaining of B7-H3 showed expression in CD31-positive blood vessels of mammary tumor tissue sections, while staining was negative in normal glandular tissue. Anecdotal evidence suggested that the in vivo B7-H3 receptor blocking with excess ABY_B7-H3 prior to imaging with MB_ABY-B7-H3 significantly reduced US molecular imaging signal from tumors. These results indicate that MB_ABY-B7-H3 not only differentiates a malignant tumor within the normal mammary tissue but also generates a highly specific molecular B7-H3 signal in tumors. Due to its affinity for both mouse and human isoforms of B7-H3, ABY_B7-H3 allows for simultaneous optimization of MB contrast development for human use and testing in translational mouse models of breast cancer. However, additional imaging studies are necessary in a larger number of animals expressing human B7-H3 in the tumor-associated blood vessels to determine sensitivity and specificity of MB_ABY-B7-H3.

Although a clinical grade MB_ABY-B7-H3 contrast agent may be suitable for breast cancer screening, it will be an ideal development for ultrasonic detection of mammographically occult malignancy in women with dense breast tissue. Full clinical translation of MB_ABY-B7-H3 for breast cancer detection and screening needs additional work in many other fronts, such as safety testing of targeted MB use in patients, fine tuning the acoustic parameters for the MB constructs, and image processing or pulse sequencing for use in newer clinical US systems, such as the automated breast volume scanners that are capable of producing 3D images and lowering operator dependency or software beamforming systems that are capable of integrating new and more sensitive molecular imaging techniques. To date, the use of targeted and nontargeted contrast MB has shown a very low number of adverse events in humans and had no nephrotoxic effects, which also means a patient is not required to perform renal function tests prior to MB administration as required before CT and MRI imaging methods (45, 48). In addition, MB show a low systemic toxicity profile in human subjects based on recent clinical trials on inoperable pancreatic cancer patients (49). In the first-in-human trial with MB targeted to VEGFR2 (BR55; ClinicalTrials.gov identifiers: NCT01253213, NCT02142608, EudraCT2012-000699-40), 12 of 45 patients reported 15 adverse events that were mild in intensity with no serious complications (48).

Limitations of the current contrast agent and imaging technique need to be addressed for full clinical translation. In the current study, MB were conjugated to ABY_B7-H3 with biotin–streptavidin chemistry. This conjugation chemistry is immunogenic in humans, and streptavidin can also bind nonspecifically to biotin present in the human body. Less immunogenic chemical approaches are available for conjugating ABY_B7-H3 to the MB such as cysteine–maleimide conjugation used in the FDA-approved antibody–drug conjugates (50) and synthesis of MB from preformed lipid–ligand conjugates increasing their shelf life. Also, although this study was conducted with an adequate affinity ABY, a higher affinity ABY (K_d = 0.9 ± 0.6 nmol/L) was developed (37), which can further increase the US molecular signal sensitivity and specificity. In preclinical US molecular imaging, high-frequency US is used to delineate targeted imaging signal, which cannot be integrated in low-frequency US systems used in clinic. Preclinical US systems are also poor in distinguishing background tissue signal from MB signal due to high echogenicity of tissue in a heterogeneous tumor microenvironment. Clinical-frequency, real-time molecular imaging methods that do not require the destruction–subtraction technique used here and techniques to improve the sensitivity of bound MB signals to background tissue signal and nonbound MB signals are necessary for full clinical translation.

Our work is one step closer to achieving clinically translatable targeted MB for accurate breast cancer detection by US molecular imaging. The use of B7-H3–targeted US imaging can be expanded to monitoring breast cancer antiangiogenic therapy, determining disease progression with noninvasive quantification of vascular B7-H3 expression as a proxy for tumor pathologic state, and creating multitargeted MB with other relevant disease markers for high contrast clinical images. More immediately, MB_ABY-B7-H3 will aid in the earlier cancer detection of breast cancer as a supporting tool for mammography.

R. Bam and L. Abou-Elkacem are listed as coinventors on a provisional patent application on Affibody Proteins Specific for B7-H3 (CD276) that is owned by Stanford University. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Bam, P.S. Lown, K.E. Wilson, A.M. Lutz, R. Paulmurugan, B.J. Hackel, L. Abou-Elkacem

Development of methodology: R. Bam, A.M. Lutz, R. Paulmurugan, L. Abou-Elkacem

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Bam, K. Sharma, K.E. Wilson, L. Abou-Elkacem

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Bam, P.S. Lown, L.A. Stern, K. Sharma, K.E. Wilson, G.R. Bean, B.J. Hackel, J. Dahl, L. Abou-Elkacem

Writing, review, and/or revision of the manuscript: R. Bam, P.S. Lown, L.A. Stern, K.E. Wilson, G.R. Bean, A.M. Lutz, R. Paulmurugan, B.J. Hackel, J. Dahl, L. Abou-Elkacem

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Bam, P.S. Lown, A.M. Lutz, R. Paulmurugan, L. Abou-Elkacem

Study supervision: R. Bam, R. Paulmurugan, B.J. Hackel, J. Dahl, L. Abou-Elkacem

Other (I am the first author of this work): R. Bam

We acknowledge the guidance and mentorship of the late Dr. Juergen K Willmann. We thank the Preclinical Imaging Core Facility and Proteomics Resource Facility at the Canary Center for Cancer Early Detection, Stanford University. IHC staining of human breast cancer specimens was performed at Human Pathology/Histology Service Center, Stanford University. This study was funded by the NCI under grant number R41-CA213544 and a Stanford Women's Cancer Center Innovation Award from the Stanford Cancer Institute, a NCI-designated Comprehensive Cancer Center.

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.