A preclinical PET dual-tracer imaging protocol for ER and HER2 phenotyping in breast cancer xenografts

MCF-7 was obtained from ATCC. JIMT-1 cells were a generous gift from Dr. Heikki Joensuu (University of Helsinki) [25]. Cell lines were grown in DMEM media supplemented with 1% amphotericin B, 1% penicillin/streptomycin, and 10% FBS (reagents supplied by Wisent, Canada).

Mice were handled in accordance with our institution’s Ethics Committee for Animal Experiments guidelines. Tumor implantation was performed under anesthesia with a mixture of 1.5% isoflurane and 2 L/min oxygen flow on female athymic nude mice (Charles River Laboratories, Wilmington, MA, USA). For ⁸⁹Zr-T imaging optimization and pharmacokinetic evaluation, 5 × 10⁶ JIMT-1 cells were implanted subcutaneously in mice for longitudinal PET imaging and for biodistribution at multiple time points. A cohort of mice (n = 4) was implanted subcutaneously with 5 × 10⁶ MCF7 and JIMT-1 cells on each shoulder. For orthotopic tumors, 5 × 10⁶ MCF-7 (n = 4) or JIMT-1(n = 5) cells were implanted in a thoracic mammary pad. At the time of first imaging sessions, tumor volumes were 60–100 mm³, and by the end of the imaging sequences, no tumor had a volume of > 310 mm³.

4FMFES radiosynthesis, purification, and activity were performed as already described [24]. For ⁸⁹Zr-T preparation, trastuzumab was obtained from the clinical pharmacy at the Sherbrooke Medical Center. Trastuzumab (10 mg) was diluted in 0.1 M Na₂HCO₃ (pH 9.0) and reacted with a 10-fold molar excess of p-isothiocyanatophenyldeferoxamine (p-SCN-DFO) active ester (Macrocyclics, USA). After 30 min, the reacted trastuzumab was placed into Amicon Ultra 0.5-mL centrifugal filter (50 kDa cut-off) tubes (Millipore-Sigma, Canada) and centrifuged and buffer exchanged with PBS (pH 7.0). ⁸⁹Zr-oxalate was produced as per the method by Alnahwi et al. [26]. One hundred megabecquerel of ⁸⁹Zr-oxalate was neutralized with 2 M Na₂HCO₃ (pH 9.0) slowly while stirring. When the pH reached ≥ 6.5, 1 mL of 1 M HEPES buffer (pH 7.2) was added to the reaction tube. DFO-conjugated trastuzumab (1 mg) was introduced into the ⁸⁹Zr-oxalate solution and incubated for 30 min at room temperature. ⁸⁹Zr-T was purified using centrifugal filter tubes.

4FMFES radiochemical purity was measured as previously described [27]. A sample of 1 μg of ⁸⁹Zr-T was evaluated by SDS-PAGE (4–15% gradient polyacrylamide gel) followed by autoradiography (Additional File 1a). In addition, ⁸⁹Zr-T radiochemical purity was measured by instant thin-layer chromatography with 0.1 M DTPA as eluant (Additional File 1).

PET imaging sessions were performed on a LabPET8/Triumph small animal PET platform (Trifoil, CA, USA). JIMT-1 tumor-bearing female nude mice (n = 4) were intravenously injected with 2.0 ± 0.4 MBq of ⁸⁹Zr-T and then imaged at 24 h, 48 h, 72 h, 144 h, and 168 h p.i., with each imaging sessions lasting for 15 min. Acquisition data were reconstructed using 20 iterations of a 3D Maximum Likelihood Expectation Maximization algorithm implementing a physical description of the detectors in the system matrix. Tracer distribution on the PET images was analyzed using the AMIDE software. A cylindrical phantom (24.8 mL) containing 1.4 ± 0.5 MBq of ⁸⁹Zr at day 0 (with the same phantom was re-measured each subsequent ⁸⁹Zr-T imaging day) or 2.4 ± 0.5 MBq of ¹⁸F for 4FMFES scans (for use in later sections) was used to obtain a calibration factor for converting the radioactive counts per second into percent injected activity/gram (%IA/g). Regions-of-interests (ROIs) were drawn for tumors, liver, muscle, heart, and bone, which were readily visible, as previously described [28]. Radioactivity uptake in the knee joint and the heart was used to determine general uptake for the “bone” and “blood,” respectively.

JIMT-1 tumor-bearing mice were injected with 1.8 ± 0.4 MBq ⁸⁹Zr-T. Mouse cohorts (n = 3/group) had blood samples taken at days 1, 2, 3, 5, 6, and 8 h p.i. to calculate blood clearance rate. Mice were also euthanized by CO₂ asphyxiation under deep isoflurane anesthesia, and biodistribution performed at 24 h, 72 h, 144 h, and 168 h. Organs of interest and tumors were collected and counted in a Packard Cobra II gamma-counter (GMI, Ramsay, MN, USA) for 1 min per tube with a 15–1000 keV energy window, background corrected and converted into %IA/g. Bone uptake reflected mostly the bone of the femur shaft, but there were portions of the knee joint included as separation from the lower limb involved slicing through the knee joint.

The imaging protocol commenced with the intravenous (i.v.) injection of tumor-bearing mice with 2.6 ± 0.3 MBq 4FMFES followed by PET scans 45 min p.i. After the scan mice were immediately administered i.v. 3.9 ± 0.4 MBq ⁸⁹Zr-T (50 μg) then returned to their cages. PET acquisitions were performed at 48 h and 144 h p.i. of ⁸⁹Zr-T for the subcutaneous model and at 144 h only for the orthotopic model. Acquisition scan times of 15 min were performed for 4FMFES and 15 min and 30 min for ⁸⁹Zr-T at 48 h and 144 h p.i., respectively.

Tissue uptake and tumor-to-background ratios for each group were reported as mean ± standard deviation. A Shapiro-Wilk normality test was performed on every dataset, which were all above the probability threshold set a priori at p < 0.05. Significance testing between radiotracers was performed using a 1-way ANOVA with Tukey’s multiple comparisons test, with a probability threshold set at p < 0.05.

Longitudinal PET imaging of ⁸⁹Zr-T injected (2.0 ± 0.4 MBq/mouse [n = 4]) in JIMT-1 bearing mice was performed to follow progression of the tumor uptake and to evaluate off-target accumulation over time. As shown in Fig. 1a, and as expected, image quality progressively improved over time. There was radioactivity present in the central cavity relative to radioactivity accumulated in tumors at 24 h. However, by 48 h, the radioactivity in the central cavity was noticeably reduced. The tumor uptake was strong in the tumor from 48 h to the final 168 h time point. As the background radioactivity reduced at 72 h and 144 h, the tumor contrast also proportionally increased. There was no visual difference in tumor contrast from 144 to 168 h. From a visual standpoint, sharp tumor images were evident at 48 h.

Tumor uptake increased steadily through most of the 168 h study. The tumor uptake at 48 h was 15.4 ± 4.4 %IA/g and increased to 16.3 ± 3.5 %IA/g at 72 h and peaked at 22.2 ± 6.0 %IA/g by 144 h (Fig. 1b). Radioactivity in the blood, as measured by ROI in the heart cavity, had a maximum uptake of 11.0 ± 1.3 %IA/g at 24 h and steadily decreased with values of 8.9 ± 1.4 %IA/g and 7.4 ± 1.8 %IA/g at 48 h and 72 h, respectively. The blood radioactivity decreased to a low of 3.2 ± 0.3 %IA/g by 168 h. The liver, muscle, and bone uptake remained steady and significantly lower than that of the tumor at all time points. Interestingly, a 2-fold uptake spike in the bone at 168 h was observed and most likely was caused by free ⁸⁹Zr accumulation in the knee joint (Fig. 1b). The tumor uptake was significantly greater than the liver (p < 0.05), muscle (p < 0.001), and blood (p < 0.01) starting at the 48 h time point.

Tumor-to-muscle (T/M) ratios were ~ 7.5 at the early time points of 48 h and 72 h. The T/M ratios increased to ~ 15.0 at the later time points of 144 h and 168 h (Fig. 1c). In contrast, the tumor-to-blood (T/B) ratios started at 1.1 ± 0.3 at 24 h and increased steadily peaking to 4.3 ± 0.8 by 144 h (Fig. 1c).

Taken together, tumor uptake of approximately 15 %IA/g that resulted in T/M and T/B ratios of ~ 7.5 and 3.0, respectively, produced high contrast images of JIMT-1 tumors for ⁸⁹Zr-T in a time window of 48–72 h p.i. As anticipated with intact antibodies, beyond 72 h, as tumor uptake increased and non-target uptake decreased, tumor contrast increased accordingly.

In parallel to the mice previously analyzed by ROI, biodistribution was also performed on different tumor-bearing mouse cohorts. Physical blood sampling of ⁸⁹Zr-T radioactivity in the blood gradually decreased according to an exponential fit with a half-life of 75 h (R² = 0.92; Fig. 2a). The biological (whole-body residency) half-life, as determined by serial measurements of whole-body mice activity in a counting well compared to calculated radioactive decay of the injected activity following a linear regression (R² = 0.91), was estimated at 234.4 h.

To determine whether the previous ROI measurements accurately reflected the circulating blood radioactivity, the heart was also dissected and contained radioactivity measured by gamma-counting. Uptake in the actual blood had a maximum uptake of 18.0 ± 4.7 %IA/g at 24 h and dropped to 12.0 ± 2.9 %IA/g and to 10.6 ± 2.4 %IA/g at 48 h and 72 h, respectively (Fig. 2b). At 120 h, the blood radioactivity fell nearly 3-fold to 3.7 ± 0.7 %IA/g and remained stable at time points 144 h, 168 h, and 192 h (Fig. 2a, b). Apart from the 24 h time point the ROI- and biodistribution-derived blood uptake values were comparable.

In contrast, the uptake in the heart derived from biodistribution did not match the uptake from the actual blood. The heart had a maximum uptake value of 5.3 ± 0.5 %IA/g at 24 h, which was markedly lower than the radioactivity in the actual blood at the same time point (Fig. 2b). By 144 h, uptake in the heart (1.6 ± 0.4 %IA/g) was decreased almost to background levels observed in the muscle (1.3 ± 0.5 %IA/g). Thus, the uptake in the blood, as measured on the heart by ROI, was in line, albeit with slightly reduced uptake values, with the radioactivity levels obtained from gamma counting on actual blood.

The biodistribution data of ⁸⁹Zr-T revealed high and specific uptake in JIMT-1 tumors, which increased over time (Fig. 2b). JIMT-1 uptake was 14.2 ± 1.5 %IA/g at 24 h and increased to 15.9 ± 0.9 %IA/g and 16.4 ± 1.7 at 72 and 144 h, respectively. At 192 h p.i., the tumor accumulation was highest at 20.3 ± 4.3 %IA/g. Compared to ROI analysis, biodistribution showed that the tumor uptake curve was still rising through the final 192 h time point. In contrast, ROI showed that tumor uptake peaked at 144 h and then shouldered off at the final assessment time point of 168 h p.i. As with the ROI analysis, biodistribution also showed that tumor uptake was significantly increased compared to the liver (p < 0.05), muscle (p < 0.001), and blood (p < 0.05) starting at 72 h p.i.

Biodistribution-derived uptakes in the muscle, liver, and bone were very similar to the uptake values derived from ROI analysis. Interestingly, there was a spike in bone uptake from 72 h to 144 h and then dropped back down at 192 h. This pattern of bone uptake was discerned by both methods. Bone uptake at 168 h measured by ROI was 7.9 ± 1.7 %IA/g whereas by biodistribution uptake peaked to 5.1 ± 1.5 %IA/g at 144 h and then dropped to 2.9 ± 0.2 %IA/g at 192 h. This was most likely due to differences of bone sampling between both methods. Whereas ROI-derived uptake was determined by measuring radioactivity in the knee joint, biodistribution measured mostly the bone from the femur shaft. The knee joint which typically has the highest radiotracer content relative to the limb bones for ⁸⁹Zr-T was most likely why the uptake values were higher by ROI analysis.

The T/M ratios calculated from biodistribution were 10.6 ± 1.6, 15.0 ± 0.9, 12.9 ± 3.2, and 32.6 ± 8.8 for 24 h, 72 h, 144 h, and 192 h, respectively. Although the T/M ratio was much lower by ROI analysis at 24 h (4.5 ± 1.0) (Fig. 1c), the ratios were comparable from 72 h and beyond. Importantly, at 72 h, the T/M by ROI of 7.5 matched closely with the T/M from the biodistribution data. The T/B ratios calculated from biodistribution were 0.8 ± 0.1, 1.5 ± 0.2, 4.2 ± 0.5, and 5.4 ± 0.7 for 24 h, 72 h, 144 h, and 192 h time points, respectively. This also matched well with the ROI-derived T/B ratios in Fig. 1c. The biodistribution-derived T/heart ratios were markedly increased at all time points relative to the T/B ratios, with values of 2.7 ± 0.7, 4.2 ± 0.8, 10.5 ± 1.7, and 13.4 ± 1.6 at 24 h, 72 h, 144 h, and 192 h, respectively. These results further support that the radioactivity measured by ROI in the heart region reflected the continual blood flow through this organ and not myocardium uptake.

The comparison between ROI- and biodistribution-derived uptake values and T/M and T/B ratios revealed that the two methods were comparable for determining ⁸⁹Zr-T targeting of JIMT-1 tumors. However, there are slight discrepancies, but they can be explained by the differences between the two methodologies and further complicated by inter-group biological variations. Differences are most likely not caused by ⁸⁹Zr-T synthesis as this radiotracer was repeatedly produced with high specific activity and purity. More importantly, these studies validated the use of ROI for moving forward and investigating the dual-tracer sequential-imaging approach and its ability to detect the ER+/HER2− and ER−/HER2+ phenotypes. In addition, this study identified that 48 h p.i. was suitable for ⁸⁹Zr-T to produce high contrast tumor images, and we could integrate this time point into the dual-tracer protocol.

Sequential injection and imaging by 4FMFES-PET followed by ⁸⁹Zr-T-PET selectively discriminated ER+/HER2− MCF-7 and ER−/HER2+ JIMT-1 tumors, respectively (Fig. 3a). 4FMFES-PET detected the ER+/HER2− MCF-7 tumors but not ER−/HER2+ JIMT-1 tumors (Fig. 3a). By ROI analysis, the uptake value for MCF-7 tumors (2.3 %IA/g ± 1.2 %IA/g) was significantly (p < 0.005) increased over JIMT-1 tumors and muscle by factors of 2.6 and 4.6, respectively (Fig. 3d). Uptake in the intestines was observed and indicated normal hepatobiliary elimination for estradiol-based tracers in mice and humans [23, 24]. Accordingly, the liver uptake value was 4.0 %IA/g ± 0.6 %IA/g. The uptake in the MCF-7 xenografts but not JIMT-1 tumors was significantly (p < 0.005) increased over uptake in the muscle (Fig. 3d).

Immediately following 4FMFES-PET, mice were intravenously injected with ⁸⁹Zr-T. Selecting the 48 h p.i. time point from Fig. 1a, ⁸⁹Zr-T-PET revealed strong tumor signals in ER−/HER2+ JIMT-1 but not in ER+/HER2− MCF-7 tumors (Fig. 3b). The ROI-derived uptake value in the JIMT-1 tumors was 16.3 %IA/g ± 2.5 %IA/g, which was significantly (p < 0.001) increased over MCF-7 by a factor of 2.4 at 48 h (Fig. 3e). As anticipated, high levels of circulating ⁸⁹Zr-T were observed but well below the JIMT-1 tumor PET signal intensities. As a result, JIMT-1 tumor uptake was also significantly (p < 0.001) increased relative to uptake in the muscle, liver, and blood. The late time point of 144 h revealed increased signals from JIMT-1 relative to the 48 h time point (Fig. 3c). The radioactivity from circulating ⁸⁹Zr-T and in all tissues was relatively low, with the exception of the liver, which presented a discernable uptake. The uptake value in JIMT-1 tumors increased from 13.7 ± 1.9 %IA/g at 48 h to 19.0 ± 1.0 %IA/g at 144 h, whereas the uptake value in MCF-7 tumors decreased in the same time interval (5.9 ± 2.3 %IA/g to 5.0 ± 1.4 %IA/g) and could not be discerned in the image (Fig. 3f). ⁸⁹Zr-T uptake in the muscle and liver remained stable. Radioactivity in the blood decreased from 7.3 ± 1.1 %IA/g to 2.7 ± 0.9 %IA/g for 48 h and 144 h, respectively. ⁸⁹Zr-T achieved significantly increased (p < 0.001) uptake margins between JIMT-1 tumors and the MCF-7 tumors and healthy organs at the later time point. The most commonly used chelator for ⁸⁹Zr is DFO; the ⁸⁹Zr-DFO complex is partly unstable, and as a result, there can be substantial nonspecific accumulation in the bone in preclinical tumor models at later time points [29‐31]. Nonetheless, the uptake in bone at 144 h was reasonably low at 5.1 ± 1.5 %IA/g compared to previously published reports (Fig. 3f). Importantly, uptake values in the tumor and organs for ⁸⁹Zr-T matched well with the ROI-derived uptake values during the imaging optimization studies with ⁸⁹Zr-T only (Fig. 1).

Comparing the analyzed T/background, radioactivity uptake ratios in tumors and healthy organs provided increased information on the feasibility of 4FMFES and ⁸⁹Zr-T to specifically image the ER+/HER2− and ER−/HER2+ phenotypes, respectively. 4FMFES-PET revealed MCF-7 and JIMT-1 T/M ratios of 3.7 and 1.4, respectively. ⁸⁹Zr-T-PET at 48 h p.i. revealed a JIMT-1 T/M ratio of 6.7, and T/liver and T/blood ratios were both 2.2. The JIMT-1 T/M, T/liver, and T/B ratios increased proportionally at 144 h to values of 8.5, 3.5, and 7.1, respectively, whereas the T/bone ratio reached 2.0 (Fig. 3f). The MCF-7 T/M and T/B ratios at 144 h were between 1.8 and 2.8, respectively, and significantly lower than those observed for JIMT-1 tumors (p < 0.001). Hence, the PET images, uptake values, and T/background ratios collectively demonstrated the feasibility of the dual-tracer protocol to provide ER+/HER2− and ER−/HER2+ tumor targeting specificity within a week and as early as 72 h from start to finish.

To further evaluate the protocol capacity to discriminate between ER+/HER2− and ER−/HER2+ tumors with high contrast visualization in a more patient-reflective microenvironment, the location of the tumors was changed from heterotopic to orthotopic. In addition, mammary fat pads were chosen for tumor growth that were in close anatomical proximity to the hepatobiliary system, which is the major metabolic sink for 4FMFES and ⁸⁹Zr-T.

Mice injected with 4FMFES produced PET images where signals from the MCF-7 tumors were well above surrounding muscle tissue and could be discerned from the high background signals from the hepatobiliary system (Fig. 4a). In contrast, the JIMT-1 tumors produced a low background signal equivalent to the signals coming from the surrounding muscle tissue. The uptake value for MCF-7 tumors (1.6 ± 0.6 %IA/g) was significantly (p < 0.005) increased over JIMT-1 tumors (0.6 ± 0.2 %IA/g) (Fig. 4b). The uptake in the MCF-7 xenografts but not JIMT-1 tumors was significantly (p < 0.001) increased over uptake in the muscle.

At 144 h, ⁸⁹Zr-T-PET showed strong PET signals from ER−/HER2+ JIMT-1 tumors with minimal background including the liver (Fig. 4c). Unlike the heterotopic model, PET images did show signals from ER+/HER2− MCF-7 tumors that had marginal increased intensity relative to signals from the liver, which was in close proximity. The uptake value in the JIMT-1 tumors was 13.7 ± 1.9 %IA/g, which was significantly (p < 0.001) increased over MCF-7 by a factor of 2.3 (Fig. 4e). The radioactivity from circulating ⁸⁹Zr-T and in all tissues with the exception of the liver had scant radioactivity. ⁸⁹Zr-T accumulation in JIMT-1 tumors was significantly increased relative to the MCF-7 tumor, liver, muscle, blood, and bone (Fig. 4d).

The tumor uptake values obtained for both radiotracers in the orthotopic model were slightly decreased compared to the uptake values from the heterotopic tumor model (Fig. 3a–c; Fig. 4a, b). ¹⁸F-4MFES uptake also decreased, albeit non-significantly (p = 0.35) in both tumors compared to the heterotopic tumor model. The uptake for ⁸⁹Zr-T in target JIMT-1 tumors was reduced by ~ 4% at 6 days p.i. However, uptake in healthy tissues for both radiotracers was also reduced, most likely because of inter-group variability. Hence, the T/background ratios were similar between both models. For instance, MCF-7 T/M ratios for 4FMFES were 3.7 and 3.5 for the heterotopic and orthotopic models, respectively (Fig. 3d and Fig. 4b). JIMT-1 T/M ratios for ⁸⁹Zr-T were 6.7 and 6.4 for heterotopic and orthotopic models, respectively (Fig. 3f and Fig. 4d). Thus, the PET images, uptake values, and T/background ratios in the orthotopic model were comparable to the results obtain in the heterotopic tumor model.