Despite advances in diagnostic clinical imaging, there is a critical need for new imaging agents and new methods that allow non-invasive characterization of lung lesions in the sub-centimeter range. Engineered knottin peptides have recently been validated by our group to specifically image integrin-expressing tumors in xenograft mouse models (17
). Here we expand on our previous work and show that we can characterize spontaneous lung tumors developing in a conditional transgenic mouse model using 64
Cu-DOTA-knottin 2.5F and combined small animal PET/CT. These transgenic mouse models provide a unique opportunity to study tumorigenesis in a lung cancer model that more closely resembles the “natural” development and progression of human cancer. Since there is always an underlying concern using artificial xenografts models (28
), these studies provide additional information which will aid in clinical translation of newly developed molecular imaging probes.
The performance of knottin 2.5F, a peptide engineered to target integrin receptors, was compared to that of 18
F-FDG. We were able to obtain a higher tumor to background ratio with 64
Cu-DOTA-knottin 2.5F compared to 18
F-FDG (6.01±0.61 versus 4.36 ±0.68, p<0.05) due to a low background signal in the area of interest. Is has previously been reported that anesthesia does not influence the tumor uptake of FDG (30
). It is unlikely that the difference in tumor to background ratio can be addressed to the difference in anesthetic regiment between the two probes during PET imaging. Lung lesions imaged with 64
Cu-DOTA-knottin 2.5F are clearly visible allowing for easier image interpretation easier compared to 18
F-FDG due to large spillover effect from uptake in the heart. However, it is important to note that in a few cases the quantification of 18
F-FDG uptake is biased towards a higher tumor uptake because of this spillover from the heart to the adjacent lung tumors. The presence of the αv
integrin imaging target on the tumor neovasculature was confirmed by ex vivo
fluorescence microscopy. Furthermore, in a previous study we determined the specificity of integrin-binding knottin probes by competition with an excess of unlabeled peptide (19
). Collectively, our results indicate that engineered knottin peptides are able to provide detailed molecular information about receptors expressed on vascular endothelium and the surface of tumors. Studies using dynamic PET imaging and compartment modeling are in progress to investigate the fraction of the signal originating from knottins bound to integrins on the endothelium compared to the surface of the tumor.
Clinical studies with radiolabeled RGD peptides (18
F-galacto-RGD and 18
F-AH111585) for PET imaging have demonstrated the feasibility of imaging integrin receptors expressed on tumor cells and tumor neovasculature in cancer patients (17
). In addition, recent studies show that the pharmacokinetics of integrin targeting peptides may easily be fine-tuned (19
). For example, in an orthotopic lung cancer mouse model, PET imaging with 64
Cu-DOTA labeled PEGylated dimeric cyclic RGD peptide showed better tumor delineation compared to imaging with 18
F-FDG. These studies demonstrate the potential of integrin imaging for early diagnosis and primary staging of lung cancer (15
An important characteristic of any imaging probe is its uptake efficiency at the tumor site and its lack of accumulation at non-tumor sites. The optimal probe will have high affinity for the target of interest leading to high uptake and at the same time show minimal background accumulation (38
). Significantly work has been put into engineering multimeric RDG peptide probes with increased affinity towards integrins. Multimeric RGD peptide probes show higher tumor accumulation and retention in vivo
, but are also accompanied by and increase uptake in non-tumor tissues (36
). In contrast, engineering of the knottin scaffold with the monomeric RGD binding motif shows comparable affinity towards integrins as the tetrameric RGD peptide 64
, while maintaining a low accumulation in normal tissues leading to a higher tumor to non-tumor tissues ratio (Supplementary Table 1
A common criterion for diagnosing malignancy with FDG-PET clinically is the standardized uptake value (SUV), which is a measure of the absolute uptake in a given region normalized for injected dose and body mass. However, it has been reported that using the tumor to background ratio improves the sensitivity of FDG-PET for diagnosis small pulmonary nodules (7
). In the current study, absolute radiotracer uptake in the lung tumors was higher for FDG than for knottin 2.5F. This example illustrates that the comparatively better contrast demonstrated by knottin 2.5F was due to lower background signal in the thorax for 64
Cu-DOTA-knottin 2.5F compared to 18
F-FDG. Together, the favorable biodistribution and tumor accumulation merits further investigation to see if these findings translate to the clinical setting.
Knottins have previously been shown to be non-immunogenic and synthetic version of the knottin MVIIA (Ziconotide) is approved for treatment of chronic pain (19
). We have successfully performed up to 6 imaging sessions over 8 months with 64
Cu-DOTA-knottin 2.5F in immunocompetent mice without the observation of an acute immune response (data not shown). This demonstrates that repeated imaging with 64
Cu-DOTA-knottin 2.5F in mice is possible. However, rigorous toxicity studies will be required before clinical translation to minimize the possibility of an unexpected immunologic reaction against 64
The transgenic mouse model allows us to investigate if imaging of integrin expression with 64
Cu-DOTA-knottin 2.5F can be used to characterize the lesions identified on CT as malignant. However, the difficulties of generating adequate numbers of transgenic mice with the right genotype and the inability to control the exact latency of tumor development after administration of doxycycline made a rigorous investigation the tumor development at the earliest time impossible. Therefore, we were not able to investigate in detail the minimal detectable size of lung nodules that can be characterized as malignant by integrin imaging, but based on our data we were able to characterize nodules as small as 3 mm in diameter as malignant (). Some of the mice (N=2) had lesions visible on CT smaller than 3 mm in diameter (N=5). 64
Cu-DOTA-knottin 2.5F were able to delineate one lesion with a diameter of 2.5 mm but missed the other small lesions, whereas FDG missed all of the lesions smaller than 3 mm in diameter. Assuming that all lesions detected by CT are true positive findings, the sensitivity of 64
Cu-DOTA-knottin 2.5F and FDG for all lesions is 73.3% and 66.7% respectively, and 100% for both probes for lesions larger than 3 mm in diameter. Although we do not have histology to confirm the malignancy of the smaller lesions we know from previous work that lung lesions detected by CT correspond with lung cancer (20
). A possible explanation for failed detection of those small lesions by PET could be loss of contrast in the PET images due to blurring originating from respiratory motion during the PET acquisitions and PVE. Unlike CT, no respiratory gating was applied during the PET acquisitions.
Because of the relative poor resolution of clinical PET scanners the PVE becomes substantial for tumors in the sub-centimeter range. For small tumors, PVE will lead to a under estimation of the true uptake and decreased sensitivity; however, using mathematical models, it is possible to correct for PVE. Using a combination of morphological information obtained from CT images, correction for respiratory motion, and low background signal from engineered peptides such 64
Cu-DOTA-knottin 2.5F, PVE-corrective algorithms will likely lead to better estimations of true uptake in lung nodules (42
). Furthermore, clinical scanners with improved spatial resolution continue to be developed, which if married to the right tracers has significant potential for improved cancer diagnostics/management.
The clinical utility of 64Cu-DOTA-knottin 2.5F imaging will have to be carefully studied in patients with different types of lung cancer. If knottin 2.5F imaging is able to outperform FDG it could replace FDG-PET. Alternatively, if the knottin 2.5F imaging agent is not able to perform as well as FDG, then it could be used in cases where FDG uptake is borderline as indicated by an intermediate SUV. In these cases, FDG is unable to distinguish inflammation vs. tumor and the knottin 2.5F may help to do so.
In summary our results demonstrate that engineered peptides, such as 64Cu-DOTA-knottin 2.5F, have the potential to be used in early detection of lung cancer. However, further studies are needed in order to address the smallest detectable tumor size by 64Cu-DOTA-knottin 2.5F. Further studies will also be needed to understand 64Cu-DOTA-knottin 2.5F uptake in sites of pulmonary infection and inflammation, which could lead to up-regulation of integrins during the inflammatory process. Together with improvements in scanner spatial resolution/sensitivity, PVE correction, and further engineering of knottins to improve the tracer performance, PET imaging with engineered knottins might prove to be more sensitive than FDG-PET for primary diagnosis of lung lesions.