Applying PET to broaden the diagnostic utility of the clinically validated CA19.9 serum biomarker for oncology.

Before radiolabeling 5B1, its binding specificity was probed by staining selected malignant and normal tissue microarrays. 5B1 reactivity was restricted to malignancies and occasional normal tissues previously known to overexpress CA19.9. Most normal tissues showed negative reactivity to 5B1 ( Supplemental Table 1 ; supplemental materials are available at http://jnm.snmjournals.org ) as previously described ( 11 ). In contrast, strong positive staining was found in 21 of 34 colon adenocarcinomas (62%), 33 of 57 adenocarcinoma metastases to the ovary (58%), and 7 of 9 pancreatic ductal cancers (78%) at various stages ( Supplemental Table 2 ). As shown in Figure 1 , the typical reactivity observed was diffuse cytoplasmic staining, with some tumor cells clearly showing distinct staining of the cell membrane. In addition, some signet ring-ovarian cancers, urinary bladder tumors, and lymph node metastasis were also found to be strongly positive ( Supplemental Fig. 1 ). In contrast, only 4 of 43 (9%) prostate cancer samples and 0 of 51 (0%) gastrointestinal stromal tumor cases were positive (data not shown). Modification of 5B1 with the benzyl-isothiocyanate analog of desferrioxamine (the standard chelation moiety for 89 Zr) was performed using previously reported methods ( 13 , 20 ), and radiolabeling of the adduct with 89 Zr proceeded at room temperature and a pH of 7.0–7.2. High specific activities (12.1 ± 1.1 mCi/mg) were consistently obtained, and radiochemical purities of more than 95% were ensured before use. Immunoreactivity assays displayed retention of activity for CA19.9 in vitro (72.4% ± 1.1%). Stability studies in bovine serum albumin at 37°C showed that more than 95% of the radiotracer remained intact after 5 d ( Supplemental Fig. 2 ), whereas in saline, more than 75% of the radiotracer remained intact after 5 d at 37°C. Collectively, these results suggested that 89 Zr-5B1 was a suitable adduct for in vivo studies. We first asked if 89 Zr-5B1 could target a tumor type known to produce CA19.9 in a clinical setting. Small-animal PET imaging and biodistribution studies were conducted using female mice bearing a subcutaneous BxPC3 pancreas cancer xenograft on the left hind leg. PET imaging confirmed impressive delineation of the tumor-associated CA19.9 by 89 Zr-5B1 ( Fig. 2A ). ROIs drawn on the tumor from the PET images displayed an uptake of 5.0 ± 0.4 %ID/g (2 h), 16.2 ± 2.5 %ID/g (24 h), 23.8 ± 4.7 %ID/g (48 h), 36.8 ± 6.1 %ID/g (96 h), and 49.5 ± 7.7 %ID/g (120 h). Blood-pool and normal tissue binding activity appears to clear at 24 h after injection. Biodistribution data corroborated the PET data ( Fig. 2B ), as high tumor localization of 89 Zr-5B1 at 24 h (84.7 ± 12.3 %ID/g, n = 4) was observed and tumor uptake increased over time out to 120 h after injection (114.1 ± 23.1 %ID/g, n = 4). Furthermore, we found the tumor uptake of 89 Zr-5B1 at 24 h after injection to be 10-fold higher than that of the nonspecific IgG at similar time points ( Fig. 2B , inset). Also pointing to the specificity of uptake, coinjection of 89 Zr-5B1 with an excess (~250 µg) of nonradiolabeled 5B1 blocked the radiotracer accumulation in the tumor. In all studies, trivial uptake of the 89 Zr-5B1 in the normal pancreas was observed (2.78 ± 0.66 % ID/g at 24 h, 1.73 ± 0.80 %ID/g at 48 h, and 0.96 ± 0.38 %ID/g at 120 h after injection). After establishing proof-of-concept with the subcutaneous model, we next set out to validate 89 Zr-5B1 in a more clinically relevant tissue microenvironment. To this end, BxPC3-luc (the parental line stably over-expressing a luciferase construct) was injected directly into the pancreas of female SCID mice ( n = 10), and tumor growth was monitored via bioluminescent imaging. Two weeks after inoculation, tumor onset was confirmed by bioluminescence ( Fig. 3A ). MR imaging studies were performed to determine tumor volumes and verify the presence of a tumor ( Fig. 3B ). Tumor volumes ranged from 125 to 270 mm 3 based on MR imaging. The entire cohort then received 18 F-FDG PET, after which, the mice were randomly selected for received 89 Zr-5B1 PET (only after complete decay of 18 F-FDG; Supplemental Fig. 3 ). Serial PET images of transplanted BxPC3 tumor xenografts from 2 to 120 h after injection demonstrate tumor uptake and clearance from normal tissues of 89 Zr-5B1 in a time-dependent fashion ( Supplemental Fig. 4 ). ROIs ( Fig. 3C , left) drawn from the tumor delineated by 18 F-FDG displayed a mean uptake value of 4.8 ± 1.3 %ID/g, with a tumor-to-muscle ratio of 3.5 ± 1.1. 89 Zr-5B1 exhibited a mean tumor uptake of 11.1 ± 5.0 %ID/g as early as 2 h after injection, reaching 23.3 ± 10.0 %ID/g at 24 h after injection with the maximum peak uptake of 30.7 ± 6.6 %ID/g at 48 h. Tumor uptake at 96 and 120 h after injection was determined to be 31.9 ± 7.1 and 33.5 ± 6.5 %ID/g, respectively. Distinct signal-to-noise contrasts were observed on comparing 18 F-FDG PET with 89 Zr-5B1 ( Fig. 3C , right). Tumor-to-muscle ratios were determined at 6.2 ± 2.3 at 2 h after injection, doubling to 11.0 ± 1.9 at 24 h after injection. This contrast essentially peaked at 48 h after injection with 20.9 ± 3.9. PET/CT images were also coregistered to more clearly interpret the PET data ( Figs. 3D and 3E ). To confirm authentic tumor uptake of 89 Zr-5B1, we excised the pancreas-bearing tumor including surrounding normal tissues (i.e., spleen, liver, stomach, and lungs) approximately 5 min after intraperitoneal administration of luciferin. Optical imaging of the tumor and the organs showed positive bioluminescence signal in the luciferase-transfected BxPC3 pancreatic tumor ( Supplemental Fig. 5 ). PET imaging of the organs validated the tumor accumulation of 89 Zr-5B1. Coregistered autoradiography and histologic staining of frozen sections obtained from the excised tumors demonstrated preferential localization of 89 Zr-5B1 to regions of viable tumor and significantly lower uptake in regions of necrotic tumor or normal pancreas ( Fig. 4 ; Supplemental Fig. 6 ). These data corroborate the specificity of 89 Zr-5B1 for the target tissue and also imply that this radioconjugate could be used to directly determine the effect of treatment on viable tumor. To assess the ability of 89 Zr-5B1 to detect other CA19.9-expressing adenocarcinomas, we examined this radiotracer in lung and colon cancer models, 2 tumor types for which some attempt has been made to monitor clinical disease with serum CA19.9. Small-animal experiments were conducted using DMS79 small cell lung cancer cells and Colo205 colon cancer cells implanted subcutaneously on the right hind leg of female SCID mice. PET images were acquired at 24–120 h after injection. Heterogeneous DMS79 tumor uptake was demonstrated with 38.15 ± 2.12 %ID/g as early as 24 h after injection, with excellent signal against background ( Fig. 5A ). An increase in tracer tumor accumulation resulted at 48 h after injection (44.60 ± 6.47 %ID/g), with retention at 120 h after injection (41.97 ± 12.23 %ID/g). As expected, nonspecific 89 Zr-5B1 accumulation cleared rapidly from normal tissues, with minimal to no background uptake at 48 h after injection. In addition, tumor delineation was observed in the Colo205 xenografts as shown in Figure 5B at 24–120 h after injection. The ROIs displayed tumor accumulation with 10.5 ± 0.76, 23.5 ± 2.7, 24.8 ± 4.0, 18.4 ± 4.7, and 16.5 ± 2.3 %ID/g at 2, 24, 48, 96, and 120 h, respectively. An observable increase in liver accumulation resulted over time with consequent decrease in tumor uptake as shown in the ROIs drawn from the PET images ( Fig. 5C ). Data generated from the biodistribution studies ( Supplemental Tables 3– 4 ) correlate well with the observed PET results. Collectively, these results highlight the broad avidity of 89 Zr-5B1 for CA19.9-positive tumors, regardless of tissue of origin.

In this report, we have developed a lead agent to clinically validate the hypothesis that a cognate PET imaging tool can be applied to supplement the empirically annotated limitations of a serum-based biomarker. Surveying the literature, we found the antigen CA19.9 to be uniquely positioned (both for clinical and technologic reasons) to rapidly address this hypothesis in humans and therefore prepared and conducted proof-of-concept studies with the novel radiotracer 89 Zr-5B1 in preclinical cancer models. 89 Zr-5B1 specifically localized to several tumor types and produced PET images of exceptional contrast, including for a PDAC lesion established within the pancreas itself. Finally, gross tumor uptake and the tumor–to–normal tissue ratios conferred by 89 Zr-5B1 exceeded that observed with 18 F-FDG, the community’s current gold standard radiotracer. Despite our data, prior reports attempting to develop immunoPET reagents targeting CA19.9 yielded somewhat discouraging results. For instance, studies applying a 124 I-labeled chimeric mAb ( 21 ) or fragmented derivatives ( 10 ) showed significantly poorer tumor uptake in PDAC models in vivo (tumor uptake in BxPC3 and Capan-2 were ~1.1 and 0.5 %ID/g, respectively). Although one possible explanation for the large discrepancy in tumor uptake values between studies for BxPC3 tumors could be the species of the antibodies—prior studies invoked a murine rather than the human mAb used in this study—the authors did report that the murine 124 I-labeled mAbs bore high affinity for the human epitope in vitro, raising the possibility that the choice of 89 Zr as the radiolabel is crucial. Indeed, at least 1 prior pairwise comparison between a 89 Zr-labeled versus an 124 I- or 131 I-labeled mAb (the chimeric antibody U36) showed clearly higher tumor uptake in vivo associated with the 89 Zr-labeled mAb ( 22 ). It is also worth noting that the antibody–antigen complex is internalized in vitro at late time points (~24 h, data not shown). In this regard, a radionuclide such as 124 I may be suboptimal, as prior reports have shown that after intracellular radiotracer metabolism, free 124 I is rapidly extruded from a cell ( 89 Zr-desferrioxamine complexes are not thought to share this property ( 23 , 24 )). Although conventional wisdom might argue against developing a radiotracer to a cell surface epitope that is also present within an abundant circulating molecule (i.e., the circulating antigen could sequester the radiotracer in serum to prevent tumor accumulation), our own experience and previous literature accounts suggest this need not necessarily be a concern for radiotracer development programs. For instance, in this study, 89 Zr-5B1 successfully localized to Colo205 tumors, despite supraphysiologic levels of circulating CA19.9 in the serum (>1318 U/mL, trivial serum CA19.9 was detected in animals bearing BxPC3 or DMS79 tumors; Table 1 ). Moreover, our prior work with a radiolabeled antibody ( 89 Zr-5A10) showed that we could effectively target free prostate-specific antigen in prostate cancer tumors, despite circulating antigen in the bloodstream ( 12 ). Finally, several venerable clinical and preclinical studies from independent groups have shown that cell surface antigens with abundantly secreted isoforms can nevertheless be targeted for PET imaging, most notably, carbonic anhydrase 9 with 124 I-cG250 and carcinoembryonic antigen with several radiolabeled antibody fragments ( 25 – 27 ). On the basis of our current body of preclinical data, there is good reason to express optimism that 89 Zr-5B1 may be an improvement over the standard of care, particularly with respect to the management of PDAC, where 1 major unmet clinical need is the detection of occult metastatic foci. Indeed, partial pancreaticoduodenectomy is rarely curative, underscoring the urgent need to advance highly sensitive and specific diagnostic tools into the clinic to detect the disease. Because the current methods of detection (e.g., endoscopic ultrasound, MR imaging, and CT) have well-documented limitations in resolution or sensitivity ( 28 ), clinicians have begun to apply 18 F-FDG PET to detect PDAC, and numerous reports have supported an incremental benefit in this context ( 29 , 30 ). However, 18 F-FDG cannot resolve nonmalignant pathologies of the pancreas from PDAC, and the effective staging of small primary lesions (<7 mm) or liver metastases (<1 cm) is controversial ( 31 – 33 ). In this regard, because 89 Zr-5B1 consistently outperformed 18 F-FDG in the preclinical models used in this study, it is reasonable to speculate that (pending its clinical translation) 89 Zr-5B1 may uncover occult lesions not currently detectable in humans. Perhaps in further support of this hypothesis, the fact that 89 Zr-5B1 could detect CA19.9-positive models that do not produce circulating antigen (BxPC3, DMS79; Table 1 ) suggests that imaging may better reveal the percentage of clinical disease sharing this property.

With 5B1 showing potent bioactivity against CA19.9-positive tumors, 89 Zr-5B1 can potentially become an insightful selection marker to identify a priori patients most likely to respond to therapy based on pretreatment retention of the radiotracer in the tumor. Indeed, for PDAC, it is well known that a significant minority of patients will be CA19.9-negative, and a tool such as 89 Zr-5B1 could spare these patients the burden of unnecessary therapy and more effectively curate patient populations in the context of clinical trial accrual.

Recombinant 5B1 antibodies were prepared and purified as described previously ( 11 ). The 5B1 antibodies and a nonspecific human IgG were functionalized with p -isothiocyanatobenzyl-desferrioxamine (DFO-Bz-NCS; Macrocyclics, Inc.) with a 1:4 mAb:DFO-Bz-NCS ratio. For example, to 300 µL of 5B1 (1.23 mg in 0.9% saline, pH ~9), a volume of 7.2 µL of DFO-Bz-NCS (4.25 mM in dimethyl sulfoxide) was added. The reaction was incubated at 37°C for 1–1.5 h. The DFO-modified antibodies were purified via either a PD10 desalting column (GE Healthcare) or a 10-kDa centrifugal filter (Amicon). 89 Zr was produced through proton beam bombardment of yttrium foil and isolated in high purity as 89 Zr-oxalate at Memorial Sloan-Kettering Cancer Center according to a previously established procedure ( 16 ). Antibodies were labeled via methods established by Holland et al. ( 13 ). In general, 89 Zr-oxalate was neutralized to pH 7.0–7.2 with 1 M Na 2 CO 3 . The DFO antibodies were then added. The reaction was incubated at room temperature for 1–2 h. Subsequent purification was conducted using a PD10 desalting column with 0.9% saline. Purity and radiolabeling yields were quantified through instant thin-layer chromatography. The small cell lung cancer DMS79 and BxPC3 pancreas cancer cells were obtained from the American Type Culture Collection. Colo205-luc colorectal cancer cells (Bioware Ultra) were purchased from Caliper Life Sciences. All cells were grown according to the recommendations of American Type Culture Collection and Caliper Life Sciences under 37°C with 5% CO 2 humidified atmosphere. All tissue culture manipulations were performed following sterile techniques. 89 Zr-5B1 was investigated for stability in vitro in 0.9% saline and in 1% bovine serum albumin for 5 d at 37°C. Changes in radiochemical purity were monitored at 0–5 d via radio–instant thin-layer chromatography with 50 mM diethylenetriaminepentaacetic acid as a mobile phase. In vitro immunoreactivity assays were performed according to the protocol established by Lindmo et al. ( 17 ) to demonstrate the integrity of the 89 Zr-radiolabeled antibodies. All animal studies were conducted in accordance with the guidelines set by the Institutional Animal Care and Use Committee. Female CB17SC-F severe combined immunodeficient (SCID) mice (Jackson Laboratories; 6–8 wk, 20–22 g) were induced with tumors on the hind legs. All cell lines were inoculated subcutaneously (~3 × 10 6 cells) in 200 µL of 1:1 medium:Matrigel (BD Biosciences) solution and grown to a maximum tumor volume of 300 mm 3 before use. BxPC3 with the luciferin-luciferase reporter gene (BxPC3-luc) was suspended in a 1:1 mixture of fresh medium and Matrigel (BD Biosciences). Approximately 3 × 10 5 BxPC3-luc cells in a 30-µL volume were orthotopically transplanted in the pancreas via surgery in female SCID mouse (CB17SC-F, 6–8 wk old; Taconic). Mice were anesthetized with 1.5%–2% isoflurane (Baxter Healthcare) in oxygen. Mice were administered 200 µL of d -Luciferin (3 mg) intraperitoneally. Approximately 10–20 min later, bioluminescent images were acquired with the IVIS 200 system (Caliper Life Sciences, Inc.) according to the manufacturer’s protocol. MR images of the mouse pancreatic tumors were acquired on a 300-MHz Bruker 7T Biospec scanner equipped with a 560 mT/m inner diameter 12-cm gradient (Bruker Biospin MR imaging GmbH; Resonance Research, Inc.). A custom-built quadrature birdcage resonator with an inner diameter of 36 mm (Stark Contrast MR Imaging Coils Research Inc.) was used for radiofrequency acquisition. The mice were immobilized using 1% isoflurane (Baxter Healthcare Corp.) gas in oxygen. Animal respiration was monitored with a small-animal physiologic monitoring system (SA Instruments, Inc.). Scout images along 3 orthogonal orientations were first acquired for animal positioning. For mouse pancreatic imaging, respiratory gated T2-weighted images with a fast spin-echo RARE (rapid acquisition with relaxation enhancement) sequence were used to acquire axial images of the pancreas with a repetition time of 1.73 s, echo time of 47 ms, RARE factor of 8, slice thickness of 0.8 mm, field of view of 30 mm, in-plane resolution of 117 × 156 mm, and 8 averages. Tumor volume was calculated by tracing out and measuring the tumor areas, Si, in each slice and using the formula Vol = sum(Si × di) − (S1 × d1 + Sn × dn)/2, where di is the distance between the center of 2 adjacent slices. PET imaging experiments were accomplished with a microPET Focus 120 or R4 scanner (Concorde Microsystems). Mice ( n = 3–5) were administered 89 Zr-labeled antibodies (7.4–11.1 MBq, 15–25 µg) in 100–200 µL of 0.9% saline formulations via lateral tail vein injections. PET whole-body acquisitions were recorded for mice at 24–120 h after injection while anesthetized with 1.5%–2.0% isoflurane (Baxter Healthcare) in oxygen. The images were analyzed using ASIPro VM software (Concorde Microsystems). Regions of interest (ROI) were drawn and plotted versus time. CT scans were obtained using a small-animal Siemens/CTI micro-CAT II CT scanner (Siemens Medical Solutions) with an 8.5-cm axial by 5.0-cm transaxial field of view. The acquired CT images were coregistered using AMIRA software (Visage Imaging) following a previously published protocol ( 18 ). For digital autoradiography and immunohistochemistry experiments, orthotopic pancreatic tumor xenografts were harvested and flash-frozen, and 10-µm frozen sections were cut. Digital autoradiography images were obtained as described elsewhere ( 19 ), with slight modifications. Sections were then exposed to a phosphor plate for 24 h and subsequently read out at 50-µm resolution using a BAS-1800II Bio-Imaging Analyzer (FujiFilm Medical Systems). The same sections were then stained with hematoxylin and eosin and imaged using an Olympus BX-60 microscope equipped with a CC12 Soft Imaging Systems camera (Olympus America). Images were registered manually using Adobe Photoshop software (version CS5; Adobe Systems). Biodistribution studies were performed on several cohorts of mice bearing separate subcutaneous Colo205-luc colorectal, BxPC3 pancreas, and DMS79 small cell lung xenografts ( n = 3–5). 89 Zr mAbs (0.37–0.74 MBq, 1–2 µg) in 100 µL of 0.9% saline were administered in the lateral vein. Additional cold mAb (10–50 µg) was coinjected along with the tracer. A blocking study with a 250-µg excess of cold mAb was performed to address specificity of the antibody to CA19.9 in a cohort of mice. After each time point (24, 48, and 120 h after injection), the mice were euthanized by asphyxiation with CO 2 . Blood was collected immediately via cardiac puncture while the tumor along with chosen organs was harvested. The wet weights of each tissue were calculated. The radioactivity bound to each organ was counted using a Wizard 2 2480 γ-counter (Perkin Elmer). The percentage of tracer uptake expressed as percentage injected dose per gram (%ID/g) was calculated as the activity bound to the tissue per organ weight per actual injected dose decay-corrected to the time of counting. The CA19.9 levels in the sera of mice were measured using the ST AIA-PACK CA19.9 kit (catalog no. 025271; TOSOH Bioscience Inc.). The principle of the assay is based on the 2-site immunoenzyme-metric assay. The analysis was performed as described in the manufacturer’s instruction manual. The optical density of immunoassay plates were measured using an AIA2000 Automated Immunoassay Analyzer (TOSOH Bioscience, Inc.). Data values were expressed as the mean ± SD unless otherwise stated. Statistical analysis was performed with GraphPad Prism (version 5.03; GraphPad Software) using 1-way ANOVA followed by Dunnett test. A P value of less than 0.05 was considered statistically significant.