Secreted frizzled-related protein 1 overexpression in gastric cancer: Relationship with radiological findings of dual-energy spectral CT and PET-CT

We explored the role of secreted frizzled-related protein 1 (sFRP1) overexpression in gastric cancer and its relationship with radiological findings from dual-energy spectral CT(DEsCT) and positron emission tomography/computed tomography (PET/CT). We established mouse metastatic models using the SGC-7901/sFRP1 gastric cancer cell line. A control group was established using the SGC-7901/vector cell line. The models were then scanned with dual-energy spectral CT and PET-CT. Subsequent analysis, including immunohistochemistry and Transferase-mediated deoxyuridine triphosphate-biotin nick end labelling (TUNEL), was performed to confirm the role of sFRP1. Transwell chamber and angiogenesis assays were conducted to verify the effect of sFRP1 in vitro. We found that the control group showed negative radiological performance with successful implantation. Concurrently, the treated group showed visible lesions, a higher FDG uptake and increasing enhancement. The immunological and histological analysis confirmed the positive radiological performance with larger size, increasing proliferation, more microvessels and less apoptosis. The angiogenic up-regulation of sFRP1 overexpression were further verified with in vitro cell models. This preliminary study demonstrates that sFRP1 overexpression in gastric cancer cells leads to increased cell proliferation and angiogenesis, which may, in turn, contribute to positive PET/CT and CT performances.

The increasing incidence and mortality have spurred researchers to identify its molecular mechanisms. Secreted frizzled-related protein 1 (sFRP1) contains a frizzled (FRZ)-type cysteine-rich domain (CRD), possessing 30-50% sequence similarity to those of Wnt receptor frizzled proteins. As the Wnt signalling pathway is a prerequisite for the process of embryo development, proliferation, differentiation and apoptosis in adult tissues, the dysregulated activation of the Wnt pathway may induce tumourigenesis [11][12][13][14] . Therefore, sFRP1 has been traditionally known as a Wnt antagonist by interacting with Wnt ligand 11,15 . Previous studies have shown the transcriptional silencing of sFRPs in various cancers. There was aberrant hypermethylation of the sFRP genes leading to sFRP1 inactivation in the early stages of gastric carcinogenesis 11,16 . However, emerging evidence suggests that sFRP1 may also promote tumour growth. Masaki et al. determined that some metastatic gastric cancers showed especially high expression of sFRP1, with the gene expression ratio between metastatic carcinoma cells: primary carcinoma cells > 2.0 17 . Fifty-four percent of gastric patients in the previous study had highly expressed sFRP1, reflecting the controversial influence of sFRP1. In a previous study by our group, TGFβ signalling activation was considered to be the mechanism by which sFRP1 promotes the formation of tumours. sFRP1-overexpressing cells showed an increase in the expression of TGFβ 1, its downstream targets, and TGFβ -mediated EMT 18 .
Studies on tumour imaging have focused on exploring the role of PET-CT in early response monitoring and seeking for the promising radiotracers, aiming at higher tumour-targeting efficacy and specificity based on mouse models 9,[19][20][21] . There has been little research with respect to tumour imaging on the basis of altering tumour-related protein expression. Therefore, the combination of DEsCT and PET/CT was utilized in the current study, in order to assess tumours in terms of both morphology and function, reflecting tumour aggressiveness. In this study, we first explored imaging findings induced by the sFRP1-overexpressing gastric cancer cell in vivo using mouse models, with PET/CT and DEsCT. Further analysis, including immunohistochemistry, transferase-mediated deoxyuridine triphosphate-biotin nick end labelling (TUNEL), was performed to confirm the role of sFRP1. Then, transwell chamber and angiogenesis assays were conducted to verify the effect of sFRP1 in vitro.

Results
To investigate the effect of sFRP1 up-regulation on tumour imaging, we exploited the SGC-7901/vector to establish a treated group and SGC-7901/sFRP1 cells to establish a control, including T-1 (the first mouse in the treated group, pulmonary metastasis), T-2 to T-7 (the other 6 mice in the treated group, peritoneal seeding), C-1 (the first mouse in the control group, pulmonary metastasis) and C-2 to C-7 (the other 6 mice in the control group, peritoneal seeding).
Overexpression of sFRP1 increased PET/CT and spectral CT performance. There were no obvious lesions, abnormal enhancement or high FDG uptake observed in the lungs of the C-1 mouse [ Fig. 1]. The corresponding SUV max was approximately 0.34. Two metastatic nodules in the bilateral lower lungs of T-1 were clearly revealed and were approximately 2.75 × 4.27 mm and 3.17 × 3.45 mm in diameter on PET-CT. Slight enhancement was visible in the lesions despite no obvious increased FDG uptake identified, with the SUV max approximating 0.22 and 0.34 [ Fig. 2].
Similarly, C-2 to C-7 [C3 shown in Supplementary Fig. 1] did not show any obvious positive performance with PET/CT and spectral CT [ Fig. 3]. The SUV max was approximately 0.3. However, radiological images of T-2 revealed one subcutaneous metastasis, which showed mild or moderate peripheral enhancement on DEsCT and much higher FDG uptake, with an SUV max close to 1.2. The pathology-identified peritoneal tumours of T-2 were not differentiated from surrounding structures by DEsCT, whereas PET/CT coronal fused images depicted the focal abnormal uptake of metastases, with an SUV max close to 0.98 [ Fig. 4]. With successful peritoneal implantation, the results in the other mice (T-3 to T-7) were analogous to that of T-2, with SUV max values of approximately 0.8, 1.06, 0.99, 0.83 and 0.96, respectively [T-3 to T-6 shown in Supplementary Figures 1,3,4 and 5]. A significant difference in the SUV max of peritoneal metastasis was observed between the treated and control groups (P < 0.001). sFRP1 overexpression increased cell proliferation and angiogenesis. Gross specimen analysis demonstrated a nodule in C-1 that was 0.8 mm in diameter [ Fig. 1]. Two metastatic nodules in the T-1 gross specimen were 4.1 and 3.2 mm in diameter, respectively, which was consistent with findings using PET-CT images (4.27 mm and 3.45 mm, respectively). The lesion in the inferior lobe of right lung is shown in Fig. 2, whereas the other lesion on the left side was not visible in this image. C-2 to C-7 showed four to ten nodules on the peritoneum, with 3.1 mm being the largest in size by gross specimen. The average size of peritoneal nodules from C-2 to C-7 was 1.13 mm (ranging from 0.5 mm to 3.1 mm). T-2 to T-7 revealed 8 to 21 peritoneal lesions with 6.11 mm as the largest size confirmed by gross specimen (mean 2.66 mm, ranging from 0.6 mm to 6.11 mm). Additionally, T-2 showed another subcutaneous lesion, with 13.6 mm as the largest size. No significant difference in tumour  Scientific RepoRts | 7:42020 | DOI: 10.1038/srep42020 number was shown between the treated and control groups (P = 0.528). For the peritoneal tumour size, there was a significant difference between the treated and control groups (P < 0.001). Therefore, the treated group showed an increasing average size of tumours in C-3 and other positive radiological images in T-3 to T-6 [see supplementary results].
After imaging, all tumours were dissected for further analysis. Histology and TUNEL were performed to measure apoptosis in the tumour tissue. TUNEL analysis showed that there were fewer apoptotic cells in tumours formed by sFRP1-overexpressing cells than by control cells [ Fig. 5C]. Immunohistochemical staining revealed that tumours derived from sFRP1-overexpressing cells contained more proliferative cells (assayed by Ki-67 staining) and more microvessels (assayed by CD34 staining). We next tested whether the supernatant of sFRP1-overexpressing cell cultures promoted angiogenesis in vitro. We found that the supernatant of sFRP1-overexpressing cells indeed induced more tubular formation of HUVECs compared with control supernatant, demonstrating the angiogenic effect of sFRP1 in vitro [ Fig. 6A]. Likewise, supernatant from sFRP1-overexpressing cells induced more HUVEC migration compared to supernatant from vector control cells.
[ Figure 6B]. sFRP1 22 and ID1 23 have been reported to increase tumour vessel density as well as the expression of vascular endothelial growth factor (VEGF), a well-known angiogenic factor induced by tumour growth factor β signalling 24 . Indeed, higher levels of VEGF were found in the culture supernatants of sFRP1-overexpressing cells compared to control cells [ Fig. 6C, left].
We therefore demonstrated that the overexpression of sFRP1 in SGC-7901 increases tumourigenesis and induces positive performance by PET/CT and spectral CT.

Discussion
Mice in the control group after successful implantation displayed no visible lesions, suspicious FDG uptake or enhancement. Compared with the control group, sFRP1 overexpression positively induced visibly larger nodules with increasing enhancement in lung metastasis and higher FDG uptake in peritoneal tumours. Simultaneously, subcutaneous metastases in the treated group showed positive functional performance both on DEsCT and PET/ CT, i.e., increasing enhancement and FDG uptake, in addition to the visible size of the lesion. The SUV max confirmed the increased FDG uptake. However, no obvious increased FDG uptake was identified in lung metastasis and no peritoneal tumours were differentiated in either control or treated groups by DEsCT. As verified by immunological and histological analysis, tumours from the treated group contained a greater number of proliferative cells, more microvessels and fewer apoptotic cells. Additionally, we demonstrated that sFRP1 contributed to in vitro angiogenesis and in vivo microvessel formation. Therefore, we argue that sFRP1 may contribute to positive performance in PET/CT and dual CT.  Great efforts have been put forth to show that FDG uptake is associated with tumour aggressiveness 25 . Progressive gastric carcinomas, represented as the depth of invasion, lymphatic permeation, vascular invasion and tumour size, showed higher FDG uptake 26 . In this study, the increasing number and larger size of tumours in the treated group demonstrated an increase in tumour aggressiveness. Consistently, the treated group with higher aggressiveness presented a positive performance, in contrast with the control group. In terms of quantitative evaluation, studies suggested that the SUV max (one of the most popular candidates for semi-quantitative analysis of tumour glucose metabolism) has a positive correlation with proliferation in various malignancies 27,28 . Ki-67 was utilized as a quantitative biomarker for tumour aggressiveness, reflecting invasiveness and metastatic potential 29,30 . A significant moderate correlation coefficient was observed between SUV max and the Ki-67 proliferation index (PI) 25,31 . The generation of a lethal tumour mass requires both tumour cell proliferation and angiogenesis 32 . Glucose metabolism was increased to supply sufficient energy for proliferation, leading to increased FDG accumulation in tumours with high growth rates, which explained the mechanism at cellular level 33 . In our study, the immunological analysis consistently demonstrated that sFRP1 overexpression resulted in increased cell proliferation and decreased apoptosis.
As described above, sFRP1-overexpressing cells showed an increase in the expression of TGFβ 1, its downstream targets, and TGFβ -mediated EMT. TGFβ signalling activation may be the mechanism by which sFRP1 promotes tumourigenesis 18 . Previous research has also demonstrated the MCF-7 breast cancer cells cultured with TGFβ showed increased EMT, 18 F-FDG uptake and glucose transporter (GLUT1) expression 34 . As a member of the multi-membrane-spanning facilitative transporter family, GLUT1 is considered to be the main functional transporter of glucose in most transformed cells 35 . Therefore, we hypothesized that TGFβ signalling activation promotes not only tumourigenesis but also GLUT1 expression and that both contribute to the observed positive performance in PET/CT and DEsCT.
Angiogenesis plays critical roles in the growth of cancer 36 . Angiogenesis is related to increased perfusion, blood volume and permeability, leading to increased contrast enhancement, which indicates that the development of new vessels within tumours determines their performance on contrast enhanced CT. Put another way, the intensity of neovascularization determines the contrast enhancement 37 . The degree of enhancement in the corticomedullary phase has been shown to correlate with microvessel density, in terms of renal parenchymal neoplasm 38 . A similar result was found in a study of pancreatic carcinoma in which the level of CT enhancement depended on the MVD number 39 . The enhancement of lung nodules appears to be an indicator of vascularity as well 40 . Consistent with these previous studies, a sFRP1 up-regulation model (T-1) showed increased numbers of microvessels assayed by CD34 staining and demonstrated visibly abnormal CT enhancement compared with a control group (C-1), which showed no abnormal enhancement.
In view of the spatial resolution of 64-slice CT, we considered the increased tumour size may partly contribute to visible nodules with CT, especially for nodules in T-1 (4.27 mm and 3.45 mm for the average size, respectively), which were much larger than C-1 (0.8 mm). As a result, a subcutaneous tumour with a much larger size (13.6 mm) was visible and presented peripheral enhancement, compared to other invisible peritoneal tumours (1.6 mm for the average size) in T2 by DEsCT. With limited resolution, peritoneal tumours with insufficiently increasing size for visibility showed similar enhancement with the surrounding structures on DEsCT, which may contribute to the negative performance observed in T2 to T7. Tumour cell proliferation and angiogenesis were both considered to be necessary in the development of a detectable tumour 32 . Nevertheless, PET-CT provides functional imaging for tumours rather than mere morphology. Peritoneal nodules showed positive performance with higher SUV max on PET/CT. The reason why the visible lung metastases from T-1, with a larger size, showed similar FDG uptake to those from C-1 in this study is currently unclear. We believe that size was not the determinant that changed the imaging appearance of tumours.
In this study, we explored imaging findings induced by sFRP1-overexpressing gastric cancer cells in vivo using mouse models with PET/CT and DEsCT. Additionally, we demonstrated that sFRP1 overexpression in gastric cancer induces positive performance in imaging modalities. Nevertheless, there are several limitations to our study. First, GLUT1 expression was not measured in this study. Second, the sample size used in our study was relatively small. More quantitative studies should be carried out to better understand this relationship. Future studies should focus on how sFRP1 promotes endothelial cell migration and invasion.
In conclusion, sFRP1 overexpression in gastric cancer cells leads to more angiogenesis and increasing proliferation, which may, in turn, induce positive PET/CT and CT performance.

Methods
Ethics statement. This study was performed in strict accordance with the standards established by the Guidelines for the Care and Use of Laboratory Animals of Shanghai Jiao Tong University and was approved by the laboratory Animal Ethics Committee of Ruijin Hospital (Permit Number: 112). Mice were anaesthetized with an intraperitoneal injection of sodium pentobarbital (50 mg/kg) anaesthesia to minimize suffering.

Cell lines and cell transfection.
The moderately differentiated SGC-7901 human gastric cancer cell line was purchased from the Chinese Academy of Sciences in Shanghai. Cells were propagated in RPMI-1640 supplemented with 10% foetal bovine serum, 100 units/ml streptomycin, and 100 μ g/ml penicillin, at 37 °C under 5% CO 2 . SFRP1-overexpressing models were generated with the SGC-7901 cell line, which normally expresses low or undetectable levels of sFRP1.
Plasmid containing Myc-DDK-tagged ORF clone of Homo sFRP1 was purchased from Origene (Origene Technologies, Rockville, MD). Transfection was carried out using Lipofectamine ® 2000 (Invitrogen, Carlsbad, CA) in accordance with the manufacturer's instructions.
Scientific RepoRts | 7:42020 | DOI: 10.1038/srep42020 Animal model. Six-week-old male athymic BALB/c (nu/nu) mice, ranging in weight from 25 g to 30 g, were obtained from the Animal Center (CAS, Shanghai, China) and housed under specific pathogen-free conditions in animal facility. The mice were randomly assigned to two groups, i.e., the treated group and control group. Then, 100-μ l (1 × 10 6 cells/ml) suspensions of SGC-7901/sFRP1 cells were administered via the tail vein and 150-μ l (2 × 10 6 cells/ml) suspensions were administered via the abdominal cavity to establish xenograft models of pulmonary metastasis and peritoneal seeding, respectively. Similarly, SGC-7901/vector cells were administered to establish the control group. After a period of 28 days of growth, tumour imaging and analysis were performed.
Tumour imaging. PET/CT SCAN. 18 F-FDG PET/CT scan was performed on a micro-PET/CT (Inveon mPET/CT; Siemens Preclinical Solution, Knoxville, Tennessee). The scope of scanning was from the head to the tail. Mice were fasted for 12 hours before 18 F-FDG PET scans but allowed free access to water. They were imaged in a supine position after general anaesthesia, and the CT component was performed before the emission component. The scanning parameters were a tube voltage of 80 kV and a tube current of 0.5 mA (with PET images taken at the same scope). Sixty minutes after an i.v. injection of 7.4 MBq of 18 F-FDG, PET scans were acquired. The maximum standardized uptake value (SUV max ) was calculated as the semi-quantitative analysis indicator. All PET images were corrected for emission scatter and attenuation.  40 . The image reconstruction thickness was 0.625 mm at an interval of 0.625 mm. Images were reconstructed with a 25-cm display field of view (DFOV) and a 512 × 512 reconstruction matrix size. Three types of images were reconstructed from the single dual-energy spectral CT acquisition for analysis: a set of polychromatic images corresponding to the conventional 140-kVp imaging, water-and iodine-based material-decomposition images, and 101 sets of monochromatic images corresponding to photon energies ranging from 40 to 140 keV. The spectral CT images were analysed with the Gemstone Spectral Imaging (GSI) Viewer software (GE Healthcare, Waukesha, Wisconsin), with a standard soft-tissue display window preset (WL 40 and WW 400). From the monochromatic and material decomposition image sets, cross-section, multi-planar reformat (MPR) and colour-scaled images were obtained. Two radiologists conducted the analysis. All the lesions were confirmed on monochromatic images of DEsCT, combined with the corresponding iodine-based material decomposition data. Metabolic information was further evaluated on PET/CT images. We then compared results from two modalities between 2 groups.
Immunological and histological analysis of tumours in xenografts. Tumours were then harvested and processed for immunohistochemistry. Consecutive tissue sections were hydrated for antigen retrieval carried out in citrate buffer (pH 6.0) at 95 °C. After peroxidase blocking, sections were incubated with Ki67 and CD34 rabbit polyclonal antibodies (Santa Cruz biotechnology Inc., Santa Cruz, CA) followed by HRP-conjugated secondary antibodies (Dako Cytomation, Carpinteria, CA). For quantification of Ki67, the percentage of positive cells was scored in five fields per slide. Microvessels were counted as CD34-positive endothelial cells in five fields per slide under microscopy.
TUNEL assays. The TUNEL assays were performed using a FragEL ™ DNA Fragmentation Detection kit (Calbiochem, La Jolla, CA) in accordance with the manufacturer's instructions. The negative control was generated by substituting H 2 O for the TdT in the reaction mixture during the labelling step. The positive control was generated by covering the entire specimen with 1 μ g/μ l of DNase I in 1 × TBS/1 mM MgSO 4 following proteinase K treatment for 20 min. All other steps were performed as previously described.
In vitro angiogenesis assays. Angiogenesis assays were performed according to the manufacturer's instructions. Briefly, 50 μ L of ECMatrix TM solution was transferred to each well of a precooled 96-well tissue culture plate on ice. After incubation at 37 °C for 1 hour to allow the matrix solution to solidify, human umbilical vein endothelial cells (HUVECs) were harvested, resuspended and seeded at 5 × 10 3 cells per well onto the surface of the polymerized ECMatrix TM . Cells were incubated with conditioned medium from SGC-7901/vector and SGC-7901/sFRP1 cells at 37 °C for 12 hours. The ability to form tubes was evaluated by counting the tubular number, the tubular length and tubular intersecting nods in five random fields.
Endothelial cell migration assays. Cell migration was analysed by a transwell chamber assay. Ten percent FCS or condition medium was used as the chemoattractant. Endothelial cells migrated through the transwell membrane and those that attached onto the lower surface of the insert were fixed and stained followed by counting under a light microscope.

Statistical analysis.
Values represented the mean ± standard deviation (SD) of samples measured in triplicate. Each experiment was repeated three times. Student's t-test and two-tailed distribution were conducted on all quantitative data. A P-value < 0.05 was taken as the level of significance.