Preclinical Molecular PET-CT Imaging Targeting CDCP1 in Colorectal Cancer

Colorectal cancer (CRC) is the third most common malignancy in the world, with 22% of patients presenting with metastatic disease and a further 50% destined to develop metastasis. Molecular imaging uses antigen-specific ligands conjugated to radionuclides to detect and characterise primary cancer and metastases. Expression of the cell surface protein CDCP1 is increased in CRC, and here we sought to assess whether it is a suitable molecular imaging target for the detection of this cancer. CDCP1 expression was assessed in CRC cell lines and a patient-derived xenograft to identify models suitable for evaluation of radio-labelled 10D7, a CDCP1-targeted, high-affinity monoclonal antibody, for preclinical molecular imaging. Positron emission tomography-computed tomography was used to compare zirconium-89 (89Zr)-10D7 avidity to a nonspecific, isotype control 89Zr-labelled IgGκ1 antibody. The specificity of CDCP1-avidity was further confirmed using CDCP1 silencing and blocking models. Our data indicate high avidity and specificity for of 89Zr-10D7 in CDCP1 expressing tumors at. Significantly higher levels than normal organs and blood, with greatest tumor avidity observed at late imaging time points. Furthermore, relatively high avidity is detected in high CDCP1 expressing tumors, with reduced avidity where CDCP1 expression was knocked down or blocked. The study supports CDCP1 as a molecular imaging target for CRC in preclinical PET-CT models using the radioligand 89Zr-10D7.


Introduction
Colorectal cancer (CRC) is the third most common malignancy and the fourth leading cause of cancer-related death in the world [1]. Metastatic disease is the major cause of death with liver and lung the most common sites of metastasis in 50-60% and 10-30% of CRC patients, respectively [2][3][4][5]. Staging and assessment of treatment response are aided by imaging using contrast-enhanced chest, abdominal and pelvic computed tomography (CT) with or without magnetic resonance (MR) and 2-deoxy-2-[F-18] fluoroglucose (FDG) positron emission tomography (PET)-CT for improved anatomical localisation, lesion differentiation, and detection of small lesions [4,6,7].
Cancer diagnosis, risk stratification, therapy prognostication, and assessment of treatment efficacy can be improved by additional imaging approaches that noninvasively characterise and measure biological processes in vivo at the molecular level [8]. Molecular imaging integrates 2D or 3D imaging with cumulative quantification of cellular events using a variety of protocols including nuclear medicine radioligand imaging, MR imaging, MR spectroscopy, optical imaging, and ultrasound [9]. Radioligand molecular imaging employs ligands that incorporate a radionuclide conjugated to a peptide or antibody component that is specific for a protein enriched on the surface of malignant cells [10]. Systemically administered radioligands locate and bind to tumors and emit a radioactive signal to allow real-time detection using nuclear imaging modalities, such as PET, enabling determination of a range of clinically important parameters such as antigen biodistribution, anatomical location, pharmacokinetics, response to therapy, dose thresholds for malignant lesions, and off-target dosimetry [10][11][12][13][14][15][16][17][18]. As such, radioligand molecular imaging is a valuable tool for disease staging and guiding treatment decision making [2,19,20]. When combined with CT or MRI, both cellular and morphological features are acquired simultaneously [13].
An effective target protein for radioligand molecular imaging is expressed on the surface of tumor cells to be accessible to systemically delivered radioligands. To attain a high tumor-to-normal tissue ratio, candidate proteins should have homogeneous tumor expression with limited expression in normal tissue [17,21]. Complement C1r/ C1s, Uegf, Bmp1 domain-containing protein-1 (CDCP1) is a type I membrane-spanning glycoprotein with a 636 amino acid extracellular region, 20 amino acid transmembrane region, and 150 amino acid cytoplasmic domain [22,23]. It is also known as subtractive immunisation M + HEp3-associated 135 kDa protein (SIMA135), Transmembrane and Associated with Src Kinases (TRASK), and cluster of differentiation 318 (CD318) [22,23]. CDCP1 is expressed as a full-length 135 kDa protein and can also undergo proteolytic cleavage generating a 70 kDa membrane-spanning carboxyl-terminal fragment and a 65 kDa aminoterminal fragment that is either shed from the cell surface or remains bound to CDCP1 on the plasma membrane [24]. Previous studies have demonstrated the utility of mouse monoclonal 10D7 antibody that binds to the CDCP1 amino terminal, for delivery of Zirconium-89 ( 89 Zr) for PET-CT-based detection, and cytotoxins for treatment of preclinical models of ovarian [25] and pancreatic [24] cancer. In CRC, elevated CDCP1 correlates with poorer patient outcome [3,26]. Analysis of CDCP1 mRNA in a CRC cohort of 101 patients indicated that elevated levels correlate significantly with advanced stage, node metastasis, and diminished recurrence-free and overall survival [26]. Similarly, immunohistochemical analysis of 128 CRCs, including 38 cases without metastasis on presentation, 51 with liver metastasis, 35 with lung metastasis, and four with both liver and lung metastasis, elevated CDCP1 correlated significantly with tumor size, grade and stage, and decreased lung metastasis free survival [3]. e present study aimed to investigate CDCP1 as a potential radioligand molecular imaging target for PET-CT detection of CRC in cell line xenograft and patient-derived mouse models using 89 Zr-labelled 10D7. e specificity of 89 Zr-10D7 for CDCP1 expressing CRC is explored using unlabelled 10D7 to compete for antibody binding sites and via silencing of CDCP1 expression to reduce the number of antibody binding sites.

Flow
Cytometry. At 50% confluence, adherent cells were nonenzymatically detached in PBS/EDTA (0.48 mM) and counted. Cells (2.5 × 10 5 ) were washed twice in PBS, blocked in PBS containing 2% FCS for 30 minutes, and stained with mouse anti-CDCP1 aminoterminal antibody 10D7 for 1 hour at 4°C [25]. Cells were then washed twice in PBS containing 2% FCS for 30 minutes at 4°C before incubation with an Alexa Fluor 488-conjugated goat anti-mouse secondary antibody at 4°C for 30 minutes ( ermo Fisher Scientific). Cells were washed in PBS twice more, then events (20,000/condition) were recorded and analysed using a FACs Fortessa flow cytometer.

Subcutaneous Mouse Xenograft Models.
Experiments involving mice were approved by the University of Queensland Animal Ethics Committee (approval 112/17). Mice were housed in a pathogen-free environment with food and water provided ad libitum. Eight-week-old male NOD Cg-Prkdc scid IL2rg tm1Wjl /SzJ (NSG) mice ( e Jackson Laboratory, Bar Harbor, ME) were injected subcutaneously with HCT116 cells (1 × 10 6 ) or spheroids of CRC13, CRC13-shScr or CRC-shCDCP1 cells (0.1 g/mouse of pelleted cell slurry in PBS suspension). Tumors were grown for three weeks prior to the commencement of radioligand molecular imaging experiments.
irty-minute PET image acquisition was performed. CT imaging (10 minute acquisition) was performed for anatomical registration and attenuation correction (80 kV, 500 μA, 230 ms exposure time, 360°rotation with 180 rotation steps, binning factor of 4, low magnification position-producing an effective pixel size of 106 μm). CT images were reconstructed using the Feldkamp algorithm. PET images were reconstructed using an ordered subset expectation maximisation (OSEM2D) algorithm with CT attenuation correction. A conversion factor obtained from a cylindrical phantom filled with a known activity of 89 Zr was used to convert PET activity per voxel to becquerel (bq)/ cubic centimetre (cc). Image reconstruction and data analysis were performed using the Inveon Research Workspace (Siemens). In vivo tissue radioactivity was analysed at each time point within regions of interest (ROI) recorded as the percent of injected dose per unit volume (cubic centimetre) of tissue (%ID/cc). During measurement of each tumor deposit or organ, careful segmentation was performed only inclusive of the ROI. Ex vivo biodistribution of radioligand signal was performed after the 144-hour imaging time point. Mice were euthanised by CO 2 asphyxiation and a blood sample obtained by heart puncture. Tissues (tumor, heart, lungs, liver, kidneys, femur, muscle, tail, blood, and testes) were harvested and weighed. Radioactivity of ROIs was measured in a Wizard 2480 Automatic Gamma Counter (PerkinElmer) as the percent of the injected dose per unit weight (g) of tissue (%ID/g) calculated was corrected for decay and detector efficiency.

Statistical
Analysis. Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software, San Diego, CA, USA). Statistical analyses of in vivo and ex vivo datasets were carried out individually, using two-tailed Student's ttest and two-way ANOVA analysis of variance. Values represent the mean ± standard deviation (SD). A value of p < 0.05 was considered statistically significant. is paper was written in accordance with the ARRIVE guidelines [32].

Cell-Based and Mouse Xenograft Assays Identify HCT116 Cells as Suitable for Assessment of CDCP1-Directed Radioligand Molecular Imaging for CRC.
To identify a cell line suitable for CDCP1-targeted radioligand molecular imaging, flow cytometry assessing cell surface expression of CDCP1 was performed on the three CRC lines HT29, SW480, and HCT116, with ovarian cancer OVMZ6 cells used as a negative control and prostate cancer PC3 cells as a positive control [25,33]. As shown in Figure 1(a), cell surface levels of CDCP1 were the highest on HCT116 and SW480 cells, comparable to levels on prostate cancer PC3 cells, while levels were at least 50% lower on HT29 CRC cells. Western blot analysis of cell lysates indicated that CDCP1 is predominantly expressed by HCT116 cells as the full-length 135 kDa protein with only low levels of the 70 kDa carboxyl-terminal fragment generated by proteolysis ( Figure 1(b)). High levels of primarily full-length CDCP1 retained on the plasma membrane should act as a suitable candidate for radioligand molecular imaging of HCT116 cell xenografts in mice.

CDCP1-Targeted Molecular PET-CT Imaging Detects a CRC Cell Line
Xenograft. HCT116 cells grown as subcutaneous tumors in mice were employed to assess the ability of CDCP1-targeted molecular imaging to detect CRC in vivo. As summarized in Figure 2(a), after three weeks of HCT116 cell growth as xenografts, mice were injected i.v. with 89 Zr-10D7, unlabelled 10D7 followed by 89 Zr-10D7 after 60 minutes, or control 89 Zr-IgG1κ, and PET-CT imaging was performed at 1, 24, 48, 72, and 144 hours later. Histological analysis of representative untreated subcutaneous HCT116 tumors revealed that xenografts display adenocarcinoma histological features (Figure 2(b) left) and CDCP1 expression which is located predominantly on the surface of malignant cells (Figure 2(b) right). ese data confirm that HCT116 cell xenografts are suitable for assessment of CDCP1-directed molecular imaging of CRC in vivo.
In vivo PET analysis indicated that 89 Zr-10D7 signal increased up to 72 hours then plateaued (Figures 2(c) and 2(d); statistical analysis in Table S1). It revealed that accumulation of 89 Zr-10D7 in tumors (7.1 ± 1.4%ID/cc at 144 hours) was largely ablated by competition with unlabelled 10D7 (2.8 ± 0.7%ID/cc at 144 hours), while mice injected with control 89 Zr-IgG1κ had negligible tumor avidity (0.8 ± 0.1%ID/cc at 144 hours). is analysis also demonstrated that in contrast to increasing 89 Zr-10D7 signal in tumor sup to 72 hours, off-tumor 89 Zr signal in heart, lungs, and liver rapidly decreased in the first 24 hours and then plateaued in all groups of mice (Figure 2(d)).
At the end of the assay, quantitative radiometric gamma counter analysis was performed to further examine tumor and off-tumor radioactivity. e data are provided in Table S2, and statistically significant differences are shown in Table S3. e results demonstrate that endpoint radioactivity in tumors was 11.2 fold higher in mice administered 89 Zr-10D7 than 89 Zr-IgG1k (Figure 2(e); Table S2) confirming the CDCP1-specific binding and uptake apparent from the PET analysis. e specificity of 10D7 for CDCP1 expressing CRC cells in vivo was further confirmed by results showing that signal from xenograft tumors was 78% lower in mice coadministered unlabelled 10D7 and 89 Zr-10D7 compared with mice administered only 89 Zr-10D7 (Figure 2(e); Table S1).
Examination of off-tumor signal by ex vivo radiometric gamma counter analysis activity demonstrated the significantly higher hepatic activity of 89 Zr-IgG1κ in comparison to 89 Zr-10D7 coadministered with unlabelled 10D7, followed by 89 Zr-10D7. It also revealed relatively lower blood, heart, lung, and renal activity of 89 Zr-IgG1κ compared to 89 Zr-10D7 coadministered with unlabelled 10D7 followed by 89 Zr-10D7 (Figure 2(e)).
ese findings may be a consequence of reduced specific binding of 89 Zr-IgG1κ resulting in increased hepatic uptake and catabolism as a mononuclear phagocyte system-containing organ [34][35][36][37][38]. Furthermore, high femoral activity was also identified in mice that received 89 Zr-IgG1κ (Figure 2(e)). is finding further supports this hypothesis of increased 89 Zr-IgG1κ hepatic catabolism with unbound 89 Zr accumulating within unfused skeletal epiphyses of the skeletally immature mice used in these experiments, a known 89 Zr phenomenon [39,40]. High blood activity was detected in mice that received unlabelled 10D7 combined with 89 Zr-10D7 (7.5 ± 0.8%ID/g) compared to 89 Zr-10D7 and 89 Zr-IgG1κ (5.0 ± 0.6%ID/g and 0.5 ± 0.0%ID/g, respectively). We speculate that unlabelled 10D7 hepatic uptake and catabolism delayed the degradation of 89 Zr-10D7, resulting in greater free circulating activity within the blood at 144 hours. Statistically significant differences in ex vivo activity are provided in Table S3.

CDCP1-Targeted Molecular PET-CT Imaging Detects a CRC PDX.
To assess the effectiveness of a CDCP1-targeted agent against a more disease relevant model, PET-CT imaging was performed on PDX CRC13 [28].
is patientderived model retains the features of the original primary colon cancer displaying histology of a poorly differentiated adenocarcinoma (Figure 3(a) left) and prominent cell surface localisation of CDCP1 (Figure 3(a) right). After three weeks of CRC13 growth, PET-CT imaging of mice at 1, 24, 48, 72, and 144 hours after i.v. administration of radio-labelled antibodies indicated increasing accumulation of 89 Zr-10D7 signal in subcutaneous flank tumors with negligible tumor signal from mice injected with control 89 Zr-IgG1κ (Figure 3(b)). As was observed for the HCT116 cell xenograft model, both 89 Zr-10D7 and 89 Zr-IgG1κ accumulated in liver and the cardiopulmonary system with signal reducing significantly 24 hours after administration by in vivo PET analysis (Figure 3(c)). Quantitative analysis of signal from tumor, heart, lung, and liver indicated increasing 89 Zr-10D7 tumor avidity over the time course and the highest avidity at the 144-hour time point (6.8 ± 1.6%ID/cc) whereas signal from heart, lung, and liver reduced during this time period (Figure 3(c)). ese results were confirmed by quantitative ex vivo biodistribution analysis of 89 Zr-10D7 and 89 Zr-IgG1κ in recovered tumors, organs, and blood. As shown in Figure 3(d), markedly higher activity was detected in tumors from mice administered 89 Zr-10D7 (13.1 ± 1.7%ID/g) compared to 89 Zr-IgG1κ (4.9 ± 0.8%ID/g). Radioactivity was approximately the same in lung, tail, and testes from these mice, and levels were higher in liver and femur of mice administered 89 Zr-IgG1κ, while levels were marginally higher in heart, kidney, and muscle of mice administered 89 Zr-10D7 (Figure 3(d) and Table S4). Surprisingly, signal was considerably higher in the blood of mice administered 89 Zr-10D7 compared with 89 Zr-IgG1κ (Figure 3(d) and Table S4) likely as a result of reduced nonspecific accumulation of 89 Zr-10D7 in off-target tissues resulting in increased, free circulating 89 Zr-10D7 compared to 89 Zr-IgG1κ. Overall findings of this experiment correlate with those from the HCT116 CRC model presented earlier.
To examine the impact of reduced CDCP1 expression on 89 Zr-10D7 tumor avidity, PET-CT imaging results were compared from CRC13 xenografts stably transduced with a lentivirus CDCP1 silencing construct (CRC13-shCDCP1) or a scramble control construct (CRC13-shScr). As shown in Figure 4(a), histochemical analyses indicated that CDCP1 expression was markedly reduced in CRC13-shCDCP1 compared to CRC13-shSrc tumors (top) and the histology of xenografts was unaltered by reduced levels of this protein (bottom). Twice weekly measurement of xenograft volume indicated that tumor growth was unaffected by silencing of CDCP1. PET-CT imaging of mice administered 89 Zr-10D7 indicated that after 24 hours, CRC13-shCDCP1 tumors (blue circle) with reduced levels of CDCP1 had significantly lower avidity than control CRC13-shScr tumors (red circle) and the difference in avidity increased over time (Figure 4(b)). Quantitative imaging and radiometric analyses indicated that 89 Zr-10D7 tumor avidity was reduced by about 40% in CRC13-shCDCP1 (6.5 ± 1.0%ID/g) compared to CRC13-shSrc (10.8 ± 1.3%ID/g) tumors (Figures 4(c) and 4(d)). Although IHC analysis confirmed nearly 100% downregulation of CDCP1 expression in CRC13-shCDCP1 compared to CRC13-shSrc tumors, tumor avidity varied by 40%.

Discussion
Using cell line xenograft and patient-derived models in mice, our data indicate that the radio-labelled antibody agent, 89 Zr-10D7, directed against the receptor CDCP1, can be employed in PET-CT imaging to detect CRC in vivo. Combined with previous reports of elevated CDCP1 in CRC patient cohorts [3,26], our results suggest that clinical implementation of a CDCP1-directed PET-CT imaging agent could have utility in CRC including for staging and assessment of treatment response as an aid to existing modalities of CT, MR, and FDG PET-CT imaging.
We have previously shown that antibody-based CDCP1 directed agents are effective for PET-CT imaging and treatment of preclinical models of ovarian [25] and pancreatic [24] cancer. e present study demonstrates the utility of a 89 Zr-labelled CDCP1-directed agent (10D7) for CRC and also extends the ovarian and pancreatic cancer studies by demonstrating the selectivity of 89 Zr-10D7 for     e addition of high dose, unlabelled ligands is a recognized method to demonstrate in vivo target-specific blocking [41][42][43]. Our results indicate that unlabelled10D7 competes with 89 Zr-10D7 for CDCP1 binding sites, thereby reducing PET signal by ∼60%, comparable to similar experimental design in existing publications [42,44]. CRC13-shCDCP1 tumors had ∼40% reduced tumor avidity compared to CRC13-shSrc tumors. Comparable reductions are seen with other published silencing models [45]. Additionally, the peak avidity of CRC13-shCDCP1 tumors administered 89 Zr-10D7 and nontransfected CRC13 administered 89 Zr-IgG1κ was comparable. Residual radioligand uptake from silenced and nonspecific ligand experiments is attributed to nonspecific enhanced permeability (EPR) and retention effect [43], with exaggerated effects in most rapidly growing solid tumors [46] and in vivo small animal xenograft tumor models [47]. An underlying principle of radioligand molecular imaging is the provision of the lowest effective dose while maintaining diagnostically adequate spatial resolution [48].
e half-life of the chosen radionuclide must approximate that of the selected ligand [19]. 89 Zr has a long half-life of 78.4 hours, making it suitable for full-length antibodies, such as 10D7 [19,39]. However, the prolonged activity of such radionuclides results in increased off-target exposure [49]. As experienced in the in vivo models in this study, late imaging time points following the administration of a radiolabelled antibody are required to detect peak tumor-tobackground avidity, at the expense of increased radiation exposure [19]. Peak in vivo 89 Zr-10D7 tumor-to-background signal was detected at 144 hours. Antibodies appeal as ligands because of high target specificity and affinity [50]. However, ligand serum half-life is dependent on size and structure [19,39], with the relatively high molecular weight of ∼150 kDa of full-length antibodies such as 10D7 [24] requiring several days to reach peak tumor-to-background signal ratio due to initial blood pool and slower perfusion times compared to smaller vectors [40,[50][51][52]. e delay in achieving adequate tumor-to-background signal after agent administration may negatively impact clinical implementation of an antibody-based CDCP1-directed PET imaging agent for CRC and other cancers.
Another factor impacting implementation of a 89 Zr-labelled antibody against CDCP1 is in vivo radionuclide metabolism and dissociation which results in 5-10% of conjugated 89 Zr dissociating within 48 hours after administration. is is a clinical issue because unbound 89 Zr accumulates in radiosensitive bone and skeletal growth plates, reducing its diagnostic utility and increasing bone marrow radiation dose [39,40]. Another issue is the phenomena of in vivo transmetallation or transchelation of 89 Zr with metal complexing proteins such as transferrin and ceruloplasmin in the liver and kidneys which also increases off-tumor irradiation [49]. ese issues may be addressed by employing smaller CDCP1-targeted ligands including antibody fragments or peptides. Reducing the molecular size of full-length antibodies by altering the Fc receptor-binding domain can accelerate peak tumor-to-background ratio and improve blood clearance, tumor retention, and PET avidity [15]. Antibody fragments and peptides are smaller ligands with lower molecular weights and reduced serum half-lives which can more easily perfuse tissue [20,39]. Antibody fragments typically share the same high-affinity binding and specificity properties as full-length counterparts [50] and some peptides also exhibit similar binding affinities as full-length antibodies with the added advantage of rapid tissue penetration [50,52]. Peptides can also be resistant to protease hydrolysis, increasing in vivo stability [52]. In addition, low molecular weight ligands such as peptides are suited for radionuclides with short half-lives including 18 F (half-life 109.8 min) and 68 Ga (half-life 67.6 min) [19,39,50]. Such radionuclides are favoured for radioligand molecular imaging as imaging can be performed soon after administration in a clinically manageable timeframe with reduced effective dose and minimal residual radioactivity on patient discharge [53,54].

Conclusions
In summary, the preclinical murine models in this paper support CDCP1 as a radioligand molecular imaging target for CRC. Quantitative analysis by flow cytometry and semiquantitative analysis by immunohistochemistry confirm CDCP1 expression in a range of CRC cell lines and a PDX. Statistically significant high tumor uptake of the radiolabelled antibody 89 Zr-10D7 was found in two CDCP1expressing CRC mouse models at late imaging time points with specificity indicated on protein silencing and epitope blocking models. e findings support further work to examine CDCP1 as a radioligand molecular imaging target for CRC.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.  Table S1: two-way ANOVA of statistically significant in vivo PET avidity (%ID/cc) of HCT 116 tumors for imaging time points. Table S2: ex vivo radiometric gamma analysis activity (%ID/g mean ± SD) of HCT116 tumors at 144 hours. Table  S3: two-way ANOVA of statistically significant ex vivo radiometric gamma analysis activity (%ID/g) of HCT116 tumors at 144 hours. Table S4: ex vivo radiometric gamma analysis activity (%ID/g mean ± SD) of CRC13 at 144 hours. (Supplementary Materials)