Preclinical 89Zr Immuno-PET of High-Grade Serous Ovarian Cancer and Lymph Node Metastasis.

mAb-B43.13 was conjugated to DFO using p -SCN-Bn-DFO and subsequently radiolabeled with 89 Zr 4+ to yield 89 Zr-DFO-mAb-B43.13 in greater than 99% radiochemical purity and with a specific activity of 203.5 ± 29.6MBq/mg (5.5 ± 0.8 mCi/mg) ( Supplemental Figs. 1 and 2; supplemental materials are available at http://jnm.snmjournals.org ). In vitro immunoreactivity experiments using CA125-expressing OVCAR3 epithelial ovarian cancer cells revealed an average immunoreactive fraction of approximately 0.92, and in vitro serum stability assays demonstrated that 89 Zr-DFO-mAb-B43.13 is greater than 96% stable to decomposition over 7 d at 37 °C in human serum ( Supplemental Figs. 3–5 ). Small-animal PET imaging experiments facilitated the visualization of the in vivo performance of 89 Zr-DFO-mAb-B43.13. In preliminary experiments, mice ( n = 5) bearing bilateral OVCAR3 (CA125-positive) and SKOV3 (CA125-negative) ovarian cancer xenografts were administered 89 Zr-DFO-mAb-B43.13 (10.2–12.0 MBq) intravenously and subsequently imaged daily from 24–120 h after injection. The images clearly illustrate the selective targeting of 89 Zr-DFO-mAb-B43.13 to the CA125-expressing OVCAR3 tumors, with high levels of the radioimmunoconjugate accumulating in the antigen-bearing tissue. In contrast, the CA125-negative SKOV3 tumors were characterized by a low uptake of the tracer, likely due to the non-specific enhanced permeability and retention effect ( Fig. 1 ) ( 17 ). In subsequent experiments, mice ( n = 4) bearing subcutaneous OVCAR3 tumors alone were injected with 89 Zr-DFO-mAb-B43.13 (10.2–12.0 MBq), and serial PET images were acquired from 24–120 h after injection. These images revealed the clear delineation of the CA125-positive OVCAR3 xenografts, even at early time points (i.e., 24 h; Fig. 2 ). These imaging observations were confirmed by biodistribution experiments. To this end, mice bearing subcutaneous OVCAR3 ovarian cancer xenografts ( n = 4 per group) were injected intravenously with 89 Zr-DFO-mAb-B43.13 (0.55–0.74 MBq, 4–6 µg) and euthanized at 24, 48, 72, 96, and 120 h after injection, followed by the collection and weighing of tissues and the assay of 89 Zr activity in each tissue ( Fig. 3 ; Supplemental Tables 1 and 2 ). The amount of radioactivity in the tumor was initially modest (7.2 ± 0.3 percentage injected dose per gram [%ID/g] at 24 h after injection) but increased significantly over the course of the experiment, reaching a maximum of 24.7 ± 7.5 %ID/g at 120 h after injection. As is typical for radioimmunoconjugates, a concomitant decrease in the activity concentration in the blood was also observed, falling to 4.0 ± 3.0 %ID/g at 120 h. The nontarget organ with the highest activity concentration was the liver, with approximately 15 %ID/g at all time points. All other organs—including the ovaries—displayed activity concentrations of less than 5 %ID/g. As a control, an additional cohort of mice ( n = 4) was coinjected with 480 µg of unlabeled DFO-mAb-B43.13 to selectively saturate the antigen in vivo. Critically, the blocking experiment lowered the uptake of the radioimmunoconjugate in the tumor from 22.3 ± 6.3 to 7.6 ± 3.2 %ID/g ( P = 0.0093) at 72 h after injection, clearly indicating that 89 Zr-DFO-mAb-B43.13 specifically targets the CA125 antigen. The most intriguing facet of the PET images collected using 89 Zr-DFO-mAb-B43.13 was the observation of high activity concentrations in the ipsilateral LNs. In all cases, high levels of uptake were observed in the axillary and brachial LNs proximal to the OVCAR3 xenograft ( Figs. 4A and 4B ; Supplemental Fig. 6 ). In a subset of mice, however, an extended LN involvement was observed, with high activity concentrations in the ipsilateral submandibular LN as well ( Fig. 4A ; Supplemental Fig. 7 ). To investigate the uptake of 89 Zr-DFO-mAb-B43.13 in the LNs more completely, additional biodistribution studies were performed in athymic nude mice bearing subcutaneous OVCAR3 ovarian cancer xenografts ( n = 4 per group). The animals were injected intravenously with 89 Zr-DFO-mAb-B43.13 or 89 Zr-DFO-IgG (0.5–0.7 MBq; 4–6 µg) and euthanized 72 h after administration of the radioimmunoconjugate. The tissues of interest were collected, weighed, and assayed to determine the amount of activity in each sample. Not surprisingly, the tumoral activity concentrations at 72 h after injection were significantly higher ( P = 0.0005) with 89 Zr-DFO-mAb-B43.13 (18.1 ± 3.4 %ID/g) than with 89 Zr-DFO-IgG (6.1 ± 0.4 %ID/g) ( Fig. 4C ; Supplemental Table 3 ). In addition, this study clearly demonstrated high uptake of 89 Zr-DFO-mAb-B43.13 in the ipsilateral axillary and brachial LNs, with activity concentrations of 104.8 ± 45.1 and 56.6 ± 39.4 %ID/g, respectively. This is in stark contrast to the level of uptake in the contralateral axillary (3.7 ± 3.2 %ID/g; P = 0.0042) and brachial (6.4 ± 3.7 %ID/g) nodes. For the nonspecific 89 Zr-DFO-IgG construct, the degree of uptake of the probe in LNs was drastically reduced ( P = 0.0039 in the ipsilateral axillary LNs), and—equally important—no difference in the activity concentrations in the ipsilateral and contralateral nodes was found. PET imaging studies with the isotype control IgG further confirmed these observations ( Supplemental Fig. 8 ). Most interestingly, we observed a progression of LN involvement further up along the ipsilateral chain in some of the OVCAR3 tumor–bearing animals that were serially imaged with 89 Zr-DFO-mAb-B43.13 at 3- to 4-wk intervals ( Fig. 5 ; Supplemental Figs. 9–12 ). The observation of uptake along the ipsilateral chain of LNs demanded further investigation to confirm the presence for metastatic spread from the primary xenograft. To obtain histopathologic evidence for metastasis, ex vivo analyses of the relevant tissues—the OVCAR3 xenografts as well as the ipsilateral and contralateral axillary, brachial, submandibular, and inguinal LNs—were performed. Hematoxylin and eosin staining of the frozen tissue sections revealed that the OVCAR3 tumors were positive for the expression of the target CA125 antigen and were well vascularized with few necrotic regions typical of xenografted tumors ( Fig. 6A ). Digital autoradiography revealed a heterogeneous distribution of 89 Zr-DFO-mAb-B43.13 in the tumors ( Fig. 6B ), with regions of higher uptake closely associated with the tumor vasculature ( Fig. 6C ). There was no radioactivity in necrotic regions of the tumors, as seen in the lower quartile of the tissue ( Fig. 6B ). Histopathologic evaluation of the formalin-fixed tumor sections revealed a multilobulated and invasive neoplasm within disorganized tufts of ovarian cancer cells surrounded by fibrovascular stroma ( Fig. 7C ; Supplemental Fig. 13 ). Evaluation of the ipsilateral axillary and brachial LNs revealed metastasis composed of the same cell population as the tumors in all the levels of the tissues examined. Neoplastic cells were also observed multifocally in subcapsular sinuses at all 5 levels. In contrast, the ipsilateral inguinal LNs showed no pathologic changes, whereas the contralateral axillary LNs showed mild medullary sinus edema and erythrophagocytosis but no evidence of metastasis. Immunohistochemical analyses of adjacent sections from the tumors ( Fig. 7C ) and ipsilateral submandibular ( Fig. 7A ) and axillary LNs ( Fig. 7B ) stained positive for CA125, thus reinforcing the presence of grossly metastasized ovarian carcinoma cells. Critically, the pathologic analysis of the ipsilateral LNs is in excellent agreement with the imaging results. To wit, in mice displaying high-activity concentrations of 89 Zr-DFO-mAb-B43.13 in the ipsilateral axillary and brachial nodes but not the ipsilateral submandibular node, it was found that the axillary and brachial nodes showed evidence of metastatic spread whereas the submandibular node did not ( Supplemental Fig. 14 ). In contrast, in mice displaying high-activity concentrations in all 3 ipsilateral LNs, evidence of metastases was found in the ipsilateral submandibular node as well ( Fig. 7A ), whereas the contralateral submandibular node showed no evidence of metastatic ovarian cancer. Taken together, these data strongly support the reliability and precision of this method for the delineation of LN involvement in our preclinical model of HGSOC.

Recent years have played witness to a surge in the use of radioimmunoconjugates for PET imaging of cancer, driven largely by the emergence of 89 Zr (half-life = 3.2 d) as a particularly well-suited radioisotope for this application ( 18 ). In the work at hand, we have extended the potential of immuno-PET to epithelial ovarian cancer by creating a 89 Zr-labeled, CA125-targeting radioimmunoconjugate: 89 Zr-DFO-mAb-B43.13. This study presents a significant advance from our previous work with a 64 Cu-labeled variant of mAb-B43.13, which was limited in its clinical potential due to the short half-life of 64 Cu (12.7 h) ( 14 ). Furthermore, our preclinical data suggest that targeting CA125 via 89 Zr-DFO-mAb-B43.13 outperforms 18 F-FDG, the radiotracer currently used for clinical PET imaging of ovarian cancer ( Supplemental Fig. 15 ) ( 14 ). The biodistribution data are a highlight of this study and provide vital insight into the in vivo behavior of 89 Zr-DFO-mAb-B43.13. The uptake of the radioimmunoconjugate in the OVCAR3 xenografts increases steadily over the course of the experiment and is offset by concomitant decreases in the activity in the blood pool and non-target organs. MUC16 is a cell surface glycoprotein, and previous studies using MUC16-targeted antibodies (including mAb-B43.13) have shown slow internalization of the antibody-antigen complex—a finding that may explain the relatively gradual accretion of activity in the OVCAR3 tumors ( 14 , 19 ). The persistence of radioactivity in the liver throughout the study may be attributed to the highly perfused nature of the organ or the hepatic clearance of potential high-molecular-weight immunocomplexes of 89 Zr-DFO-mAb-B43.13. The specificity of CA125 targeting by 89 Zr-DFO-mAb-B43.13 was established both by imaging experiments using mice bearing bilateral xenografts and by a biodistribution study in which mice bearing OVCAR3 tumors were administered a vast excess of unlabeled DFO-mAb-B43.13 with 89 Zr-DFO-mAb-B43.13. In the former, the CA125-positive OVCAR3 xenografts displayed progressively increasing uptake of the radioimmunoconjugate, whereas the CA125-negative SKOV3 tumors showed far lower activity concentrations ( Fig. 1 ). In the latter, a drastic reduction in the activity concentration within OVCAR3 tumors was observed along with a corresponding increase in the activity concentration in the blood, almost certainly resulting from the saturation of the antigen by the unlabeled immunoconjugate ( Fig. 3 ; Supplemental Table 1 ). In all cases, the CA125-expressing tumor tissue was delineated as early as 24 h after injection, with uptake values that steadily increased thereafter. Indeed, at 72 h after injection 89 Zr-DFO-mAb-B43.13 produces images with high contrast and high tumor-to-background ratios. By this time, maximum tumor accretion is achieved with minimal systemic background activity. This would certainly facilitate the visualization of residual and recurrent disease, which is often present as carcinomatous nodules within the peritoneum. The most intriguing aspect of the PET imaging experiments was the visualization of high concentrations of 89 Zr-DFO-mAb-B43.13 in the LNs proximal to the tumor. The specificity of this uptake was confirmed in biodistribution studies that used 89 Zr-DFO-IgG as an isotype control. A starkly different biodistribution profile was observed between the 2 radioimmunoconjugates, with 89 Zr-DFO-mAb-B43.13 exhibiting far higher activity concentrations in the CA125-positive OVCAR3 xenografts as well as the proximal LNs ( Supplemental Table 3 ). HGSOC patients commonly present with widespread intraperitoneal disease and LN involvement. The paraaortic region is the prime site for early LN metastasis, followed by pelvic LN involvement ( 10 , 20 ). This phenomenon poses a surgical challenge if the disease is present in the LNs above the renal hilum and in the chest, as these nodes are difficult to resect. There are also isolated clinical reports of axillary LN metastasis of HGSOC ( 21 ). Notably, sentinel LN procedures are generally not performed in the ovarian cancer setting because of the challenge posed by the local spread of the disease within the peritoneum. More recently, the presence of tumor cells in the lumens of the lymphatic capillaries (lymphovascular space invasion) has been identified as an early indicator of nodal metastasis in patients ( 22 ). In our studies, the high volume of metastatic disease confined within the small and densely packed LNs seemed to be the root cause for the high concentrations of 89 Zr-DFO-mAb-B43.13 within the ipsilateral nodes. Most importantly, the selectivity of this effect is illustrated by the correlation between the histopathologic analysis of the nodes and the PET images: the LNs delineated in the PET scans were pathologically confirmed to contain metastatic disease, whereas the LNs not visualized in PET scans proved negative for the presence of ovarian cancer metastasis on histopathologic analysis. In other words: no false-negatives were observed. In addition to its role as a marker for the peritoneal spread of ovarian cancer, high preoperative serum levels of CA125 also help identify candidates for surgical lymphadenectomy ( 8 , 23 ). Despite the shedding of CA125 by OVCAR3 cells in vitro, immunoassays did not reveal detectable levels of the antigen in the serum of tumor-bearing animals ( Supplemental Fig. 16B ). This may be a limitation of this subcutaneous xenograft model, as we observed the subcutaneous collection of CA125-rich lymphatic fluid surrounding the tumor ( Supplemental Figs. 16A and 17 ). That said, it is also possible that the amount of circulating CA125 in the tumor-bearing mice was below the detection limit of the enzymelinked immunosorbent assay. Clearly, further investigation of the utility of 89 Zr-DFO-mAb-B43.13 in more clinically and pathologically relevant animal models of HGSOC is warranted. In the interest of balance, it is important to note that despite its promise, 89 Zr-DFO-mAb-B43.13 does have some minor limitations. First, the persistence of radioactivity in the liver throughout the investigation may mask the detection of liver metastases and carcinomatous nodules present in its immediate vicinity. However, parenchymal liver metastases are rare at presentation and are more commonly found as surface capsular deposits on recurrence ( 24 ). Second, CA125 is also expressed in several non-gynecologic malignancies including breast, gastric, and lung cancers and in benign conditions such as pregnancy and menstruation and maladies such as endometriosis, liver disease, and congestive heart failure ( 16 ). Finally, mAb-B43.13 is a murine antibody and thus poses the risk of provoking a human anti-mouse antibody response in patients. However, variants of mAb-B43.13 have already been used in HGSOC patients for radioimmunoscintigraphy and demonstrated a pronounced immunotherapeutic effect ( 25 ). Nonetheless, the humanization of mAb-B43.13 is currently under way and will undoubtedly increase the potential of 89 Zr-DFO-mAb-B43.13 as a clinical immuno-PET probe.

The work at hand represents a significant step forward in our efforts toward the development of a CA125-targeted probe for the PET imaging of HGSOC. We strongly believe that 89 Zr-DFO-mAb-B43.13 has the potential to create roadmaps for the management of HGSOC by facilitating the selection of patients for surgery and treatment based on the site and status of LN involvement and the extent of the metastatic spread of the disease. Furthermore, 89 Zr-DFO-mAb-B43.13 could also prove valuable in the detection of metastases and recurrent disease, both of which affect treatment options and survival outcomes for ovarian cancer patients. Ultimately, we contend that 89 Zr-mAb-DFO-B43.13 is particularly well positioned to make a near-term impact on the clinical care of ovarian cancer not only because of the promising results described herein but also because of mAb-B43.13’s previous experience in the clinic as well as the growing clinical enthusiasm for 89 Zr immuno-PET.

PET imaging experiments were conducted on a microPET Focus rodent scanner (Concorde Microsystems). Mice bearing subcutaneous OVCAR3 (right shoulder) xenografts (150–200 mm 3 ) were administered 89 Zr-DFO-mAb-B43.13 (10.2–12.0 MBq in 200 µL of 0.9% sterile saline, 40–50 µg) via intravenous tail injection (t = 0). Approximately 5 min before PET imaging, mice were anesthetized by inhalation of a 2% isoflurane (Baxter Healthcare)/oxygen gas mixture and placed on the scanner bed; anesthesia was maintained using a 1% isoflurane/gas mixture. PET data for each mouse were recorded via static scans at time points between 24 and 120 h. Images were analyzed using ASIPro VM software (Concorde Microsystems). All data are expressed as mean ± SEM. Where applicable, statistical differences were analyzed by an unpaired Student t test using GraphPad Prism 6 software (GraphPad Software). Comparisons with P values of less than 0.05 were considered significant.