Micro-PET imaging of angiogenesis based on 18F-RGD for assessment liver metastasis in colorectal cancer

Background: This study aimed to explore the feasibility of 18 F-AIF-NOTA-E[PEG4-c(RGDfk)]2 (denoted as 18 F-RGD) PET quantitative parameters to distinguish the angiogenesis in colorectal cancer (CRC) mice which has different metastatic potential. Methods: Animal models of CRC liver metastases were established by implanttation of human CRC cell lines LoVo and LS174T via intrasplenic injection. Radiotracer-based micro-positron emission tomography imaging of animal model was performed and the uptake of 18 F-RGD tracer in the tumor tissues were quantied as tumor-to-liver maximum or mean standardized uptake values ratio. Pearson correlation was used to analyze the relationship between radioactive parameters and tumor markers. Results: The SUVmax ratio and SUVmean ratio of LoVo model was signicantly higher than LS174T ones in both liver metastasis and primary tumor lesions (P (cid:0) 0.05 ). A signicantly difference was observed in both VEGF and Ki67 expression between LoVo and LS174T primary tumors (P (cid:0) 0.05 ). The T/L SUVmean or SUVmax ratio of 18 F-RGD showed a signicantly correlation with VEGF expression, but weakly correlated with Ki67 expression. The areas under the ROC curves of 18 F-RGD SUVmean ratio for differentiate LoVo from LS174T tumor was 0.801. Conclusions: The T/L SUVmean ratio of 18 F-RGD is a promising parameters for tumor imaging and monitoring angiogenesis process in CRC xenograft mice model.


Introduction
Colorectal carcinoma (CRC) is the third most commonly diagnosed malignancy worldwide and liver metastasis is one of the main cause of CRC-related death [1,2]. Tumor metastases is the culprit associated with cancer-related deaths, represent the end-products of cell-biological process [3]. Although the progression-free survival of multiple cancer has signi cantly increased pro t from early detection and improved by individualized therapeutic regimens of modern clinical medicine, metastasis remains as the ultimate obstacle in our ght against cancer. A majority of CRC patients with liver metastasis face with poor prognosis and low overall survival rate.
The accumulation of tumor metastasis molecular mechanisms over the past decade has given us an indepth understanding of the biological behavior of metastatic progression in various kinds of cancers [4].
Angiogenesis is de ned as the formation of new blood vessels from the pre-existing microvasculature network. The supplement of neovascular is the key factor in malignant tumor growth, progression, and metastasis [5,6]. The neovascularization in tumor is complex and disorganized. Vascular endothelial growth factor(VEGF)has been identi ed as the most important regulator of tumor angiogenesis, which widely expressed in a variety of tumors. The overexpressed VEGF in tumor cells or tumor microenvironment can promote CRC tumor growth and hematogenous metastasis by stimulating angiogenesis [7,8]. Integrins are cell adhesion receptors for extracelluar matrix (ECM) proteins, and highly expressed on the surface of various cancer cells and neovascular endothelial cells. Tumor cells can migrate effectively on ECM substrates, and the multiple integrin functioning contributes to this process and plays an important role in tumor angiogenesis. Among all integrins, integrin αvβ3 probably the most strongly involved in the regulation of angiogenesis [9]. The speci cal interaction between integrin αvβ3 and vascular endothelial growth factor receptors (VEGFRs) is considered to be crucial for tumor growth and metastasis [10].
Positron emission tomography (PET) imaging is a non-invasively method to visualize and quantify the tumor microenvironment, such as proliferation and angiogenesis et al. Arginine-glycine-aspartic acid (RGD)-based peptides are well-known to speci cally bind with integrin αvβ3 and considered to be a promising positron emission tomography (PET) tracer for monitoring the status of tumor angiogenesis [11]. However, there are few reports on the prediction of tumor metastasis potential based on assessment of tumor angiogenesis status. Therefore, in this study, we investigated the feasibility of dimeric RGD peptides-based PET quantitative parameters to distinguish the angiogenesis in CRC mice which showed different metastatic potential and to predict the metastatic potential of CRC in mice model. These ndings may contribute to the better understanding neovascular microenvironment of CRC noninvasively and to yield useful radioactive markers, which are necessary to guide personalized therapeutic regimens.

Cell Culture
The human colorectal cancer cell lines LoVo and LS174T were purchased from the Type Culture Collection of the Chinese Academy of Sciences, Shanghai, China. LS174T cell line was derived from the primary lesion of colorectal adenocarcinoma, Dukes' type B; LoVo cell line was derived from the left supraclavicular metastatic site of colorectal adenocarcinoma, Dukes' type C. The LoVo and LS174T cell lines were maintained in Dulbecco's Modi ed Eagle Medium (DMEM, Gibco Corporation, USA) supplemented with 10% fetal bovine serum (Hyclone, USA) and 1% penicillin-streptomycin (Beyotime Biotechnology, China). All cells were cultured in a humidi ed atmosphere containing 5% CO 2 air at 37ºC.

Animal model
Five-week-old female BALA/C nude mice (weight, 16-18 g) were purchased from Animal Laboratory of Cavens Corporate of Changzhou (Changzhou, China). Colorectal cancer liver metastases (CLM) xenograft models were established by injecting LoVo or LS174T cells (2.0 × 10 7 cells in 0.15 mL of phosphatebuffered saline) into spleen which anesthetized by intraperitoneal injection of 10% chloral hydrate, respectively (n=20). All the animals were housed in an environment with temperature of 22 ± 1 ºC, relative humidity of 50 ± 1% and a light/dark cycle of 12/12 hr. All animal experiments were conducted in compliance with the protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Jiangsu Institute of Nuclear Medicine. The body weight of nude mice was recorded every three days.
Radiopharmaceutical preparation 18 F-AIF-NOTA-E[PEG4-c(RGDfk)]2 (denoted as 18 F-RGD) was acquired by Jiangsu Key Laboratory of Molecular Nuclear Medicine, Jiangsu Institute of Nuclear Medicine. Synthesis of 18 F-RGD has been described previously [12]. 18 F-RGD was a glutamic acid linked dimeric RGD labeled with NOTA-18 F-AIF that has been proved to be safe and stable for PET imaging [11].

Cellular uptake in vitro
LoVo and LS174T cells were maintained in corresponding medium for 24 hours before 18 F-RGD cellular uptake. For normal group, the cells, 18 F-RGD and buffer (DMEM containing 0.2% bovine serum albumin) were mixed into a glass test tube (2ml). For block group, the cells, 18 F-RGD and inhibitor buffer were mixed into a glass test tube (2ml). All the glass test tube were incubated at 37°C for 1 hour. The tubes were divided in three groups: Group O was used as a control tube, composed of 100μl radionuclides and 200μl buffer (DMEM buffer with 0.2% BSA); Group T was used to measure the radionuclide dose, contained 100μl radionuclides; Group X, was composed of 100μl radionuclides, 100μl cell suspension and 100μl corresponding buffer. The cellular uptake was normalized to 5×10 5 cells/tube; each cell line was tested three times at each time point. The radioactivity of each cell line at 1 hour was accurately measured using Automatic Gamma Counter (PerkinElmer, 2480, USA). The cellular uptake ratio was calculated using the following formula: X(cpm)-0(cpm)/T(cpm)%.
Micro-PET Imaging and analysis 18 F-RGD PET imaging of LoVo and LS174T-CLM mice models (n=20) were performed seven weeks after implantation. Six-minute static 18 F-RGD (About 9.25MBq, 250μCi) was acquired 1 hours (radiopharmaceutical administration time in mice) after injection via tail vein. All the mice were anesthetized with 2% iso urane in 100% oxygen with a ow rate of 2 L/min prior to imaging. Micro-PET scans were acquired in 3-dimensional mode using an Inveon micro-PET scanner (Siemens Medical Solutions) with an ordered-subset expectation maximization/maximum: matrix, 128×128×159; pixel size, 0.86×0.86×0.8mm; β-value, 1.5, with uniform resolution. PET images were reconstructed and postprocessed using Inveon Acquisition Workplace software (version 2.0, Siemens Preclinical Solutions).
Regions of interests (ROIs) were drawn on images around the entire liver metastasis or primary tumor lesions and normal liver tissue using ASI Pro VM 6.8.6.9 software (Concorde Microsystems, LLC). The maximum/mean standardized uptake value (SUVmax/mean) ratio was calculated as the tumor-to-liver SUVmax/mean (T/L ratio). All Micro-PET imaging procedures were conducted according to protocol approved by the Jiangsu Institute of Nuclear Medicine Animal Care and Use Committee.

Immunohistochemical staining
The tumor specimens were xed in 10% formalin for 48 h, para n-embedded, and cut into 3 μm-thick sections. Immunohistochemical staining was performed as previously described [13]. Brie y, the slides were incubated with anti-Ki67 (1:100, ab16667, Abcam) or anti-VEGF (1:200, ab46154, Abcam) at 4°C overnight. Next, the slides were incubated with horseradish peroxidase-labeled goat anti-mouse or antirabbit secondary antibody (Boster, Wuhan, China) at room temperature followed by counterstaining with hematoxylin. The staining was observed under a BX53 Olympus microscope (Olympus, Japan) at magni cation 200×. A brown-yellow staining was de ned as positive expression. Ki67 or VEGF protein were quantitated by Image-J software (NIH, Bethesda, MD, USA).

Statistical analysis
All data were expressed as mean±standard deviation. Statistical analyses were performed using R version 3.6.1 (www.R-project.org). The difference between two groups was assessed using Student's unpaired t-test. The Fisher exact test was used for comparing differences in liver metastatic potential in vivo. The correlation between the RGD parameters and tumor marker was analyzed using Pearson correlation analysis. The receiver operating characteristic (ROC) curve was used to differentiate LoVo tumor from LS174T tumor. A P value <0.05 was considered statistically signi cant.

Comparison of metastatic potential and malignancy between the two models
Firstly, we calculated the tumor formation rate (including primary tumor and liver metastasis) of CLM model mice to evaluate the in vivo metastatic potential of LoVo and LS174T colorectal cancer cells (Table 1). Liver metastasis rate was de ned as: the number of mice with liver metastasis divide by the number of mice with primary tumor in spleen. The liver metastasis rate of LoVo-CLM mice (66.67%) was signi cantly higher than LS174T ones (41.67%) (X 2 = 6.559, P = 0.039). In addition, the median survival time of LoVo and LS174T CLM models were 8 and 11.5 weeks respectively (P = 0.0006; Fig. 1A). The results show that LoVo-CLM mice has a shorter survival time compared to LS174T-CLM mice. In addition, the body weight of LoVo and LS174T CLM models shown in Fig. 1B also con rmed that LoVo cells demonstrated a higher metastatic potential than LS174T cells in vivo. Our next experiments will be based on these two cells with differences in metastatic potential.

Comparison of 18 F-RGD parameters between LoVo and LS174T models
We next compared the difference of 18 F-RGD parameters between LoVo and LS174T CLM models in vivo.
The 18 F-RGD parameters of liver metastasis and primary tumor were measured 1 h post 18 F-RGD injection. The liver metastasis tumor and primary tumor in spleen were con rmed by HE staining ( Fig. 2A). The 6 minutes static scan of whole-body 18 F-RGD PET were performed in Fig. 2B. Our results showed that LoVo CLM model had a signi cantly higher SUVmean ratio and SUVmax ratio values in both liver metastasis and primary tumor than LS174T ones (P<0.05). The 18 F-RGD parameters of both SUVmean ratio and SUVmax ratio in primary tumor were higher than corresponding liver metastasis tumor ( Table 2). Our results indicated that tumor with high metastasis potential prone to high 18 F-RGD uptake in vivo. spleen. Immunohistochemical staining demonstrated that VEGF protein was mainly expressed in the cytoplasm and occasionally in the nucleus, whereas Ki67 protein was expressed in cell nucleus (Fig. 3A). Both VEGF and Ki67 expression in LoVo tumor were higher than that in LS174T tumor with a signi cant difference (P<0.05), as shown in Fig. 3B.

Correlation Analysis
We next investigated whether 18 F-RGD parameters can re ect the tumor angiogenesis or tumor proliferation in vivo. In Fig. 4, the results showed that no correlation was seen between RGD SUVmax ratio and Ki67 expression (P = 0.0718), however, a weak correlation was found between 18 F-RGD SUVmean ratio and Ki67 expression (P = 0.0438). On the contrary, the signi cant correlation was found between 18 F-RGD SUVmax ratio or SUVmean ratio and VEGF expression in primary tumor (P = 0.001 and P<0.0001, respectively). Our results demonstrated that the parameters of 18 F-RGD can re ect tumor angiogenesis.

Diagnostic performance of 18 F-RGD parameters to differentiate LoVo from LS174T tumors
Further, we investigated whether 18 F-RGD parameters in primary CRC can distinguish LoVo tumor from LS174T tumor, and can assess the tumor metastatic potential from the primary lesion. The ROC curve analysis were summarized in Table 3 and Fig. 5. The areas under the ROC curves (AUC) of 18 F-RGD SUVmean ratio, SUVmax ratio and SUVmean ratio combined with SUVmax ratio for differentiate LoVo from LS174T tumor was 0.801, 0.759 and 0.787 respectively. The optimal cut-off value to differentiate LoVo from LS174T tumor was 1.551 for SUVmean ratio, 1.629 for SUVmax ratio and 0.759 for SUVmean ratio combined with SUVmax ratio. In addition, the sensitivity and speci city of SUVmean ratio to differentiate LoVo from LS174T tumors in mice models were 100% and 61.1%, respectively, the same as SUVmean combined with SUVmax ratio. Our results demonstrated that SUVmean ratio with a good AUC is a suitable parameter for predicting the metastatic potential of CRC in animal models.

Discussion
Tumor metastasis is the leading cause of CRC-related death. In patients with CRC metastasis, distant metastasis is usually con ned to an isolated organ. Liver is the most common site of CRC metastasis [14]. Understanding the biological characteristics of tumor metastasis is the essential prerequisite for predicting CRC metastasis. Angiogenesis provides favorable conditions for tumor growth and metastasis,Folkman J et al suggested that tumors transformed into metastatic potential must undergo an "angiogenic switch" by perturbing the balance of proangiogenic and antiangiogenic factors [15]. Tumors with VEGF-overexpressed were more prone to occur "angiogenic switch", consequently converting to metastatic potential.
In recent years, advances in the eld of anti-angiogenesis therapy in oncology have provided more options for the treatment of cancer patients. More and more preclinical and clinical researches were focusing on how to monitor the tumor angiogenesis and treatment response by using multiple imaging modalities such as PET, SPECT, molecular MRI, targeted ultrasound, or optical imaging. Integrins are strongly involved in mediating adhesion events during tumor metastasis by activating many cellular signaling pathways, and the expression of integrin αvβ3 is mostly correlated with tumor metastatic potential [16,17]. RGD peptides-based PET imaging for evaluation tumor angiogenesis and proliferation is a promising non-invasive method for understanding the tumor biological behavior [18]. In addition, integrin αvβ3 plays an important role in regulating tumor angiogenesis and proliferation. Our results evidenced that a signi cant correlation between the 18F-RGD tracer parameters and VEGF expression in primary CRC tumor. Moreover, the SUVmean ratio of 18 F-RGD in primary tumor had the ability to distinguish LoVo tumor from LS174T tumor in mice model, which were with different tumor metastatic potential.
In this trail, we rstly compared the liver metastatic potential of LoVo and LS174T cells in vivo by establish a liver metastasis model. The results showed that LoVo-CLM models had a higher liver metastatic rate than LS174T ones (66.67% vs 41.67%), in addition, LoVo-CLM mice has a shorter survival time compared to LS174T-CLM mice (8 vs 11.5 weeks), suggesting that LoVo cells exhibited a stronger metastatic capability than LS174T ones. The result was same with our previous study [19].
The 18 F-RGD tracer was selected for assessing integrin αvβ3 expression and used its radiological parameters to quantify the angiogenesis of CRC in mice model. For the cellular uptake of 18 F-RGD in vitro, a low RGD uptake was found in two cells and there was no statistical difference between LoVo and LS174T cells at 1 h post-incubation. The reason for this phenomenon can be explained as an inactive or "off" state of integrins expressed on cell surfaces, in which they do not bind ligands and do not signal [20]. In order to explore the expression levels of integrins in solid tumor neovascularization of CRC mice, we perform 18 F-RGD PET imaging 1 hours post-injection and compare the radiological parameters between LoVo and LS174T mice. We delineated the ROI in liver metastases and primary splenic tumors according to corresponding anatomical locations. In further in vivo PET experiments, the encouraging data showed that the expression of integrin αvβ3 in tissue was superior to cells, and the parameters of 18 F-RGD SUVmean/SUVmax ratio was capable of distinguish the differences in angiogenesis between LoVo and LS174T mice. Integrin αvβ3 is mainly expressed on vascular endothelial cells,and therefore exhibits good 18 F-RGD uptake in tumor tissues with neovascular networks [21]. Our results demonstrated that tumors with high metastatic potential have high 18 F-RGD uptake capacity, acquisition of angiogenic properties accelerate the change from a quiescent to an invasive phenotype which known as "angiogenic switch" [22].
VEGF, which is an important factor in tumor angiogenesis, and Ki67, which is a nucleolar protein widely appreciated as a cell proliferation, are both associated with tumor growth and metastasis [23,24]. Our quanti cation of immunohistochemical staining in primary tumor showed that the expression of VEGF and Ki67 in LoVo tissues were stronger than LS174T ones, combined with the results of correlation analysis, a weak correlation between 18 F-RGD SUVmean ratio and Ki67 expression (P = 0.0438) and a strongly signi cant correlation between RGD SUVmax ratio/SUVmean ratio and VEGF expression in primary tumor (P = 0.001 and P 0.0001, respectively), indicating that the SUVmax ratio/SUVmean ratio of 18 F-RGD may be recognized as an ideal radiation maker to re ect tumor angiogenesis in CLM mice model. Our viewpoint was consistent with previous reviews, that 18 F-RGD is a promising tracer for tumor angiogenesis imaging and monitoring anti-angiogenesis therapy in solid tumor [25,26]. In Sibel lsal's study, they also veri ed that the expression of integrin αvβ3, which are targeted by RGD-based peptides tracer, was associated with cell proliferation in glioblastoma [27].
We aimed in this study to predict tumor metastatic potential in primary CRC tumor using radiolabeled-RGD peptides. According to the ROC analysis, 18 F-RGD parameters of SUVmean ratio and SUVmax ratio has the capacity to distinguish LoVo tumor from LS174T tumor, among these SUVmean ratio shown the highest areas under the ROC curves (AUC) with the sensitivity 100% and speci city 61.1%. A SUVmean ratio of ≥ 1.551, a SUVmax ratio of ≥ 1.629 and a SUVmean ratio combined with SUVmax ratio of ≥ 0.759 identi ed as the best cut-off value to determine LoVo tumor from LS174T tumor. Our results were in line with the available published reference, as recently research showed that the contrast-to-noise (CNR) indexes of the Gd-RGD tracers in MRI was suitable for differentiating hepatocellular carcinoma (HCC) tissues with high metastatic potential from those with low metastatic potential [28]. However, the present study also had its limitation due to only two kinds of CRC cell lines with different metastatic potentials included in the comparative study. In the future, we will compare the predict potential of 18 F-RGD on various types of CRC cell with different metastatic potential.
In conclusion, we rstly investigated the correlation between the tumor makers VEGF/Ki67 expression and 18 F-RGD uptake, and the results suggested that 18 F-RGD is a promising tracer for tumor imaging and monitoring angiogenesis microenvironment in CRC xenograft mice model. SUVmean ratio of 18 F-RGD in PET imaging can be used for differentiating LoVo tumor with high metastatic potential from LS174T tumor with low metastatic potential. Therefore, SUVmean ratio of 18 F-RGD is a suitable parameter for predicting the metastatic potential of CRC in animal models.

Declarations
Authors' contributions ZMY, JHJ, ZRJ and WZC contributed to conception and design; ZMY, JHJ, ZRJ, JH and WZC contributed to acquisition of data, or analysis and interpretation of data; ZMY, JHJ, JH, and WZC have been involved in drafting the manuscript or revising it critically for important intellectual content; all authors have given nal approval of the version to be published.

Figure 1
Survival analysis of LoVo and LS174T liver metastasis model. n=20.

Figure 2
In vivo whole-body Micro-PET images of 18F-RGD. A: HE staining of liver metastasis tumor and primary tumor. B: White arrows indicate the uptake of 18F-RGD in liver metastasis tumor. Yellow arrows indicate the uptake of 18F-RGD of the primary tumor in spleen.

Figure 4
Correlation analysis between RGD parameters and biomarkers. A: correlation of 18F-RGD SUVmax ratio between Ki67 expression. B: correlation of 18F-RGD SUVmean ratio between Ki67 expression. C: correlation of 18F-RGD SUVmax ratio between VEGF expression. D: correlation of 18F-RGD SUVmean ratio between VEGF expression. ROC curves analysis of different RGD parameters to distinguish LoVo tumor from LS174T tumor.