Preliminary Clinical Application of RGD-Containing Peptides as PET Radiotracers for Imaging Tumors

Angiogenesis is a common feature of many physiological processes and pathological conditions. RGD-containing peptides can strongly bind to integrin αvβ3 expressed on endothelial cells in neovessels and several tumor cells with high specificity, making them promising molecular agents for imaging angiogenesis. Although studies of RGD-containing peptides combined with radionuclides, namely, 18F, 64Cu, and 68Ga for positron emission tomography (PET) imaging have shown high spatial resolution and accurate quantification of tracer uptake, only a few of these radiotracers have been successfully translated into clinical use. This review summarizes the RGD-based tracers in terms of accumulation in tumors and adjacent tissues, and comparison with traditional 18F-fluorodeoxyglucose (FDG) imaging. The value of RGD-based tracers for diagnosis, differential diagnosis, tumor subvolume delineation, and therapeutic response prediction is mainly discussed. Very low RGD accumulation, in contrast to high FDG metabolism, was found in normal brain tissue, indicating that RGD-based imaging provides an excellent tumor-to-background ratio for improved brain tumor imaging. However, the intensity of the RGD-based tracers is much higher than FDG in normal liver tissue, which could lead to underestimation of primary or metastatic lesions in liver. In multiple studies, RGD-based imaging successfully realized the diagnosis and differential diagnosis of solid tumors and also the prediction of chemoradiotherapy response, providing complementary rather than similar information relative to FDG imaging. Of most interest, baseline RGD uptake values can not only be used to predict the tumor efficacy of antiangiogenic therapy, but also to monitor the occurrence of adverse events in normal organs. This unique dual predictive value in antiangiogenic therapy may be better than that of FDG-based imaging.


INTRODUCTION
Angiogenesis, a well-known process by which new blood vessels sprout from the pre-existing capillaries, includes vascular endothelial cell activation, the degradation of vascular basement membrane during the activation, proliferation and migration of endothelial cells, and the construction of new blood vessels and vascular networks. It is associated with a long list of physiological and pathological events (1,2). Under normal physiological conditions, such as embryonic development, wound healing, and the menstrual cycle, angiogenesis serves primarily to ensure adequate nutrient and oxygen supply and waste removal. Of course, normal angiogenesis also occurs in pathological conditions, such as tissue damage after reperfusion of ischemic tissue or cardiac failure (3,4). Unlike the normal angiogenesis described above, both solid and hematologic tumors feature a high dependence on persistent abnormal angiogenesis (5)(6)(7). Solid tumors initially form as avascular masses that rely on the vascular system in the host microenvironment, but upon reaching a diameter of 1-2 mm, these tumors exhibit excessive dysfunctional angiogenesis to obtain essential oxygen and nutrients (8)(9)(10)(11). The initial recognition of angiogenesis as an interesting process in oncology began in the early 1970s, when Drs. Folkman and Denekamp put forward the idea that highly immature neovascularized tumors are vulnerable via their blood supply. The immature vessels are characterized by an undifferentiated endothelium and a lack of smooth muscles, and act as the vehicles for peripheral sprouting of new capillaries, invasion and distant metastasis (12)(13)(14)(15). These mechanisms are facilitated by the synergistic effects of transmission signals from the extracellular matrix (ECM) to endothelial cells, among which integrin binding to endothelial cell recognition sites is the first process to occur. Importantly, the tripeptide sequence of Arg-Gly-Asp (RGD) was identified as a common recognition site in the ECM and blood, and can bind to a variety of adhesion proteins and integrins (16).
The contribution of the RGD tripeptide sequence to cell attachment was fi rst recognized in fi bronectin by Pierschbacher and Ruoslahti in 1984 and later confirmed in more ligands, namely, vitronectin, fibrinogen, decorsin, von Willebrand factor, thrombospondin, disintegrins, laminin, entactin, tenascin, prothrombin, osteopontin, bone sialoprotein, adenovirus penton base protein, and collagens (16)(17)(18). Combinations of different alpha (a) and beta (b) subunits, such as avb3, avb5, avb8, a5b1 and a8b1, act as receptors for many of the above-mentioned proteins, mainly acting on angiogenesis, inflammation, thrombosis, lung development, cell differentiation, and other pathophysiological processes (19)(20)(21)(22). Among these, avb3 is highly expressed on angiogenic endothelial cells and several tumor cells, but is expressed only at low levels by resting endothelial cells of normal tissues (23)(24)(25)(26). The avb3 expression level has also been reported to be a promising factor in tumor diagnosis, staging and therapy (27)(28)(29). Due to the high specific affinity between RGD and avb3, radiolabeled RGD peptides have been intensively studied for their ability to represent avb3 expression density and the tumor angiogenesis state noninvasively (30,31). Combinations of RGD-containing peptides with radionuclides such as 18 F, 64 Cu, and 68 Ga, have been introduced successfully in positron emission tomography/computed tomography (PET/ CT) imaging, among which 18 F has near ideal decay characteristics for PET (t 1/2 = 110 min, b + = 0.64 MeV). A linear RGD peptide was first labeled with 18 F via solid-phase synthesis in 2001, but this radiotracer showed nonspecific accumulation within tumor tissue and early breakdown (32). Then, the cyclic pentapeptide cyclo (Arg-Gly-Asp-D-Phe-Val) (denoted as cyclic RGD peptide) was prepared and found to be a highly potent, stable and selective inhibitor of integrin signaling (33,34). In addition, based on the polyvalency effect, different types of dimeric and multimeric cyclic RGD peptides were prepared to improve avb3-binding affinities for intense tumor accumulation and increased tumor/background ratios compared with the monomeric compounds (35)(36)(37). During the optimization of radiotracers, strategies such as cyclization, glycosylation, pegylation, multimerization, and insertion of spacers/linkers have been adopted to optimize the in vitro and in vivo behaviors of radiolabeled RGD peptides (38)(39)(40)(41). However, few of these RDG peptide labeling methods have been successfully translated from the bench to clinical use.
Therefore, in this review, we focus on the preliminary clinical applications of PET radionuclide-conjugated RGD-containing peptides that target RGD-binding integrins (mainly avb3 integrin), discussing their distribution characteristics and potential use in diagnosis, delineation of tumor subvolume, evaluation response of chemoradiotherapy or antiangiogenic therapy, and the comparative results achieved with traditional 18 F-fluorodeoxyglucose ( 18 F-FDG) imaging agent in the context of functional tumor imaging.

COMPARATIVE CHARACTERISTICS OF RGD-BASED AND FDG TRACERS IN CLINICAL APPLICATION
Monomeric and multimeric RGD-based PET tracers that have been investigated in preliminary clinical studies are listed in Supplementary Table 1 and Figure 1. 18 F-Galacto-RGD, which is a glycosylated RGD-peptide, was initially reported by Haubner et al. in 2005 and was the first monomeric integrin-specific PET tracer used in patients (56). It is well tolerated, as all studies reported neither drug-related adverse events nor important clinical indicators. The dose distribution results were consistent for a variety of tumor types, with similar evidence of predominant renal excretion, fast blood elimination, rapid tumor tracer accumulation, and low background concentration in most organs, namely, the lung, muscle tissue and brain, allowing imaging of tumor lesions with good contrast (56)(57)(58). A relatively small amount of 18 F-Galacto-RGD is also excreted via the hepatobiliary pathway, and thus, in some patients, tracer uptake is seen in the gallbladder. Similar biodistribution results were observed for the other monomeric radionuclide tracers such as 18 F-fluciclatide (denoted as 18 F-AH111585) (59-61) and 18 F-RGD-K5 (43). Consistent with the polyvalency effect caused by dimerization and multimerization of cyclic RGD peptides, series of RGD radiotracers were found to show improved radiosynthesis yields, relatively higher tumor integrin-specific accumulation and favorable in vivo pharmacokinetic properties compared with monomeric ones in both preclinical and clinical studies (44,45,59,(62)(63)(64)(65)(66)(67)(68). However, the multistep, timeconsuming and low-yield synthetic procedures make an automated production process, which is mandatory for routine clinical use, extremely difficult, and thus, the commercial clinical viability of this tracer is limited. Chen et al. successfully established a simplified labeling procedure for 18 F-fluoridealuminum complex-labeled RGD tracer 18 F-AlF-NOTA-PRGD2 (denoted as 18 F-Alfatide) and 18 F-AlF-NOTA-E [PEG4-c(RGDfk)]2 (denoted as 18 F-Alfatide II) in one single step of radiosynthesis (46,54,69). Upon successful introduction of a simple, one-step, lyophilized kit for 18 F-alfatide preparation, with which complete radiosynthesis including purification can be accomplished within 30 min with a radiochemical purity of more than 95%, the potential for clinical transformation of this radiotracer was greatly improved. The pharmacokinetics and imaging properties of 18 F-Alfatide were comparable to those of 18 F-FP-PEG3-E[c(RGDyK)]2 (denoted as 18 F-FPPRGD2) (54,70). In addition, the positive effect of cyclic mono or multimeric peptides was also successfully developed into 68 Ga-labeled RGD analogs, which have shown promise for imaging integrin avb3 expression (48,(71)(72)(73)(74)(75)(76)(77)(78)(79).
Unlike the phenomenon of decreasing standard uptake values (SUVs) in normal organs over time, tumor retention of RGDbased PET tracers did not stop increasing until a plateau period was reached, which showed no obvious fluctuation within nearly 40-60 min after tracer injection. It appeared as either a distinct increase in primary tumors and lung, pleura, bone, and lymph node metastases compared with surrounding normal tissue or as moderate or reduced accumulation in the cases of kidneys, bladder, liver, spleen, and intestine. Sixty minutes after injection was recommended as the appropriate imaging time for optimal lesion-to-background contrast, according to the data from RGD-based fluorine-labeled monomeric or dimeric radiotracer studies (44,45,57,59,60). For biodistribution in normal tissues, RGD tracers can be used to image brain tumors or brain metastases with a clear background, due to their inability to pass through the healthy blood-brain barrier (BBB) but ability to cross the damaged BBB created by the tumor (43,46,61). Primary high uptakes were shown in organs such as the kidneys, bladder, liver, spleen and intestine organs, along with diffuse uptake in the bilateral ventricles and thyroid. High uptake in the kidneys and bladder resulted predominantly from urinary excretion. The uptake values of liver and spleen were thought to be physiological, high liver background may result from the densely vascularized nature of the organ and metabolism of RGD-based imaging tracers by the liver (60). Prominent accumulation of RGD peptide ligands in the liver reported in mice and humans also supported this assertion (80,81). The uptake values in the intestine maybe associated with physiological expression of avb3 on intestinal smooth muscle cells, because the uptake of these tracers did not change significantly during an observation period of 90 min (57,58,82). While an initial 18 F-FPPRGD2 increase followed by a pattern of decreasing uptake in the intestine was also observed, this may reflect initial dimeric RGD accumulation was cleared from hepatobiliary excretion (45,83). Overall, this finding indicates the unsuitability of RGD imaging for hepatocellular carcinoma (HCC) tumors or diseases with liver metastases.
In addition to integrin avb3 expression, glucose metabolism is also believed to correlate with tumor aggressiveness and progression (84,85). In tumor lesions, RGD-based tracer uptake is significantly lower than that of FDG, as commonly observed in studies comparing FDG and RGD tracers (52,63,65,66,(86)(87)(88)(89)(90)(91)(92), but this difference was altered in some cases, such as with the higher maximum SUV (SUV max ) and tumor-to-blood ratio (TBR) of 68 Ga-NOTA-3P-TATE-RGD compared with 18 F-FDG in neuroendocrine tumor (NET) cases and also in both human epidermal growth factor receptor 2 negative [HER2 (-)] and estrogen receptor positive [ER (+)] breast cancer (52,88). One possible reason responsible for this difference may be related to the different binding mechanisms of the tracers. 18 F-FDG mainly depends on the hypermetabolism of glucose in malignant tumor cells, whereas RGD PET imaging predominantly binds to avb3 highly expressed in tumor neovascular endothelial cells. Also, within each lesion, the number of endothelial cells expressing integrin avb3 is smaller than the abundant number of tumor cells containing glucose transporters (93). In addition, both tracers showed strong uptake in inflammatory lesions similar to malignancies. RGD was thought to be more beneficial than FDG for differentiating inflammation, as it demonstrated significantly lower retention than FDG after comparable high uptake levels in inflammation (56,(94)(95)(96). With regard to the correlation between RGD and FDG data, inconsistent results were observed when tumor lesions overall were considered. Moderate or significant correlations between RGD and FDG parameters were reported in locally advanced rectal cancer (LARC) and squamous cell carcinoma of head and neck (HNSCC) (87,88,90,97), and for subgroups of tumors, namely, non-small cell lung cancer (NSCLC), triple-negative breast cancer (TNBC), and FDG-avid tumors (86,89). However, other studies failed to demonstrate a correlation between RGD and FDG uptake (65,91,(98)(99)(100).
These results overall may indicate the coexistence of interrelated pathophysiological phenomena within the tumor, such as cell proliferation and neoangiogenesis, but the lack of a significant correlation between RGD and FDG uptake suggests that these imaging modalities provide complementary rather than similar information (101). RGD imaging offers indispensable value in terms of evaluating tumor angiogenesis and planning avb3-based therapy. As the next logical step, many research groups focused on exploring the clinical characteristics of RGD-based imaging agents.

PRELIMINARY DIAGNOSTIC VALUE OF RGD-CONTAINING PEPTIDE TRACERS
In a pan-cancer study with 9 patients (5 with malignant melanomas, 2 with sarcomas, 1 with osseous metastasis from renal cell carcinoma [RCC], and 1 with villonodular synovitis), 18 F-Galacto-RGD SUVs showed inter-and intra-individual variances ranging from 1.2 to 10.0 (56). Consistently, Beer et al. reported the heterogeneous accumulation of 18 F-Galacto-RGD from 1.2 to 9.0 in 19 patients. Distribution volume values for tumor tissue (1.5 ± 0.8) were 4 times higher than those for muscle tissue (0.4 ± 0.1), suggesting intense specific receptor binding in tumor tissues and only minimal free and bound (specific or nonspecific) tracer in muscle tissue (57). Similarly, studies of other imaging agents such as 18 F-AH111585 and 68 Ga-NODAGA-E[(cRGDyK) 2 ] (denoted as 68 Ga-NODAGA-RGD2) also reported heterogeneous tracer uptake in tumor and metastatic sites and superior tumor-to-background ratios in patients with malignant disease (58,61,76,86,102). However, the sensitivity of 18 F-Galacto-RGD (76%) for lesion identification was lower than that of conventional staging with contrastenhanced CT and 18 F-FDG PET in a comparative study (86). This is not surprising because, unlike the glycolytic effect of entire tumor cell, avb3 is expressed at elevated levels on angiogenic blood vessels and on malignant tumors (103,104). Interestingly, the related research demonstrated high inter-and intraindividual variances in tracer accumulation in different types of lesions, indicating great diversity in receptor expression.
According to the literature, the expression patterns of avb3 are specific on malignant carcinomas and benign tumors or inflammatory tissues (95,105,106). In benign lesions (e.g., pigmented villonodular synovitis, neurofibroma, and inflammatory tissue), avb3 expression is located only on the vasculature. However, in malignant lesions, avb3 is expressed not only on the endothelial cells of neovasculature but also on the surface of some tumor cells. In HNSCC, avb3 is located predominantly on the neovasculature but shows some low expression on tumor cells (56,102,107). Immunohistochemical analyses confirmed avb3 expression predominantly on microvessels (5/5) and, to a lesser extent, on tumor cells (3/5) in invasive ductal breast cancer (IDC) (28,108). In malignant melanoma and clear cell RCC (ccRCC), avb3 is highly located on the surface of tumor cells (61,102,109,110). In the other solid tumors, namely, glioma, lung cancer, differentiated thyroid cancer, and breast cancer (56,102,(108)(109)(110)(111)(112), avb3 expression is located on the neovasculature and on the tumor cells at varying degrees. For example, a significant correlation between avb3 expression and radiotracer uptake was reported by a study employing 18 F-Galacto-RGD PET in melanoma and sarcoma patients (56). Specially, in ccRCC cases, the 18 F-FB-mini-PEG-E[c(RGDyK)]2 (denoted as 18F-FPRGD2) PET signal correlated with integrin avb3 expression on tumor cells, and immunohistochemistry results confirmed high expression of avb3 on ccRCC cells. In contrast, in papillary RCC (pRCC) cases, the 18 F-FPRGD2 PET signal correlated with the integrin avb3 level on tumor vessels (67). These results demonstrate that RGD-based tracer uptake can reflect integrin avb3 expression in renal tumors, but represents angiogenesis only when tumor cells do not express integrin avb3. In other words, RGD-based PET may be regarded as a surrogate marker of angiogenesis for the visualization and noninvasive quantitation of integrin expression when avb3 expression is predominantly confined to the tumor vasculature as in HNSCC (107). Overall, the association between RGD SUVs and TBR and the corresponding avb3 expression level demonstrated the value of noninvasive techniques for appropriately diagnosing tumor lesions, although a positive correlation was not always clear (56,61,107,(113)(114)(115)(116). The heterogeneous distribution pattern among different lesions described above may explain this phenomenon. Next, we provide a review of reported clinical investigations related to the application of RGD-based tracers for diagnosis, definition of tumor subvolume, and therapeutic response prediction according to specific tumor types ( Table 1).

Glioblastoma
Tumor tracer uptake was heterogeneously restricted to the periphery of the tumor in glioma patients, with no activity in the necrotic center of the lesions and normal brain tissue. The correlation between tracer accumulation and immunohistochemical staining intensity for avb3 expression in 21 tissue samples was weak but significant (r = 0.463, p = 0.034), which provided evidence that 18 F-Galacto-RGD PET is a promising tool for noninvasively monitoring avb3 integrin expression in patients with malignant gliomas (113). However, the characteristics revealed in this study mainly related to diagnosed glioblastoma multiforme (GBM) and recurrent GBM after external beam radiation or chemotherapy, which may reduce the reliability of the results. Li et al. further explored patients with newly diagnosed GBM of different grades in a prospective 68 Ga-NOTA-PRGD2 (denoted as 68 Ga-PRGD2) PET/ CT imaging study. The results from 12 GBM patients also showed significant accumulation in the tumors, but not in the white matter or cortical gray matter, with the exception of the choroid plexus. The maximum TBR (TBR max ) of 68 Ga-PRGD2 PET/CT was superior for differentiating high-grade glioma (HGG) from lowgrade glioma (LGG) compared with 18 F-FDG PET/CT (sensitivity, 100% vs. 88%; specificity, 100% vs. 75%). This clinical case-by-case evaluation provides a sufficient basis for applying RGD-based imaging to determine the boundary and stage of glioma (114). Data obtained by Liu et al. also supported this phenomenon of HGG exhibiting a more intense 18 F-Alfatide SUV max , mean SUV (SUV mean ), and tumor-to-normal ratio (TNR) than LGG (121). These results could be explained by the higher avb3 expression level in HGG compared with LGG (122). Overall, this research demonstrates that RGD-based tracer imaging might be superior to 18 F-FDG PET/CT for GBM imaging, as the tracer molecules only specifically accumulated in the tumors, while not accumulating in the normal brain tissue as characterized by high 18 F-FDG uptake.
For distinguishing meningioma and HGG lesions, RGD also provided considerable added value to 18 F-FDG PET/CT imaging. Li et al. showed that higher SUV max (4.23 ± 2.48) of 68 Ga-PRGD2 was observed in the uncommon meningiomas than in the HGG lesions (1.57 ± 0.33, p = 0.0047). The uptake ratios of 68 Ga-PRGD2 over 18 F-FDG normalized as lg100 * SUV max (RGD/ FDG) were significantly higher in the uncommon meningiomas than those in HGG (1.87 ± 1.36 vs. 1.04 ± 0.87, p = 0.0001) with a defined cutoff value of 1.58. However, no significant differences were detected by 18 F-FDG PET between the rare subtypes of meningioma and HGG (115).

Head and Neck Squamous Cell Carcinoma (HNSCC)
Given the predominant avb3 expression on the neovasculature in HNSCC, angiogenesis maybe the optimal surrogate variable in such cases (56,110). Eleven patients with HNSCC were examined by 18 F-Galacto-RGD static emissions scanning, it successfully identified 83.3% (10/12) of tumor-bearing cases and 33.3% (2/6) of pathological metastatic lymph nodes (107). When comparing with 18 F-FDG uptake, Durante et al. demonstrated that all primary tumors, lymph nodes and distant metastases could be seen using both tracers, with higher tumor uptake was observed in 18 F-FDG than in 68 Ga-NODAGA-c(RGDyK) ( 68 Ga-NODAGA-RGD) imaging. Notably, the SUV mean for RGD showed a positive correlation  with that for 18 F-FDG (Spearman's r = 0.89, p = 0.0068) (87). The feasibility of 68 Ga-DOTA-E-[c(RGDfK)]2 ( 68 Ga-DOTA-RGD2) PET/CT for visualizing angiogenesis was also validated in patients with oral squamous cell carcinoma with SUV max ranging from 4.0 to 12.7 (77).

Thyroid Cancer
In radioiodine refractoryIn differentiated thyroid cancer (RAIR-DTC) patients, 18 F-FDG PET/CT showed higher SUV max compared to RGD PET/CT whether in primary tumor lesions or lymph node metastasis (91,92). These two tracers showed similar sensitivities of 82.3% in detecting lesions, but 68 Ga-DOTA-RGD2 PET/CT had a higher specificity and accuracy of 100 and 86.4% compared to 50 and 75% for 18 F-FDG PET/CT (92). No significant differences were observed between RAIR lesions and benign lesions by 18 F-Alfatide and 18 F-FDG imaging.
Notably, 18 F-Alfatide uptake was significantly higher in lesions with a diameter larger than 1.5 cm than lesions smaller than 1.5 cm (91). These results supported feasibility of RGD PET/CT not only on the diagnosis of RAIR-DTC but also on the evaluation of treatment response of 177 Lu-DOTA-RGD in these patients, especially in cases deemed negative/suspicious on 18 F-FDG PET/CT (123).

Lung Cancer
In lung cancer, RGD has been widely explored in tumor diagnosis and differential diagnosis. 18 (124,125). In another study using 68 Ga-NOTA-PRGD2 PET/CT, benign and malignant tumors and lymph nodes were successfully identified in 91 patients with suspected lung lesions. Both the SUV mean and SUV max of primary malignancies were higher than those of benign lesions, and the SUV max of metastatic lymph nodes was higher than that of non-metastatic nodes. In addition, comparative results of 68 Ga-NOTA-PRGD2 and 18 68 Ga-NOTA-PRGD2 PET/CT offered similar sensitivity and higher specificity for detecting lung lesions compared with 18 F-FDG PET/CT, but remarkable improvement for the assessment of lymph nodes metastasis with a higher specificity (100,126).
Consistent results were also obtained for the ability of RGDbased radiotracers to differentiate tuberculosis and lung cancer, with studies reporting that tracer uptake values for 18 F-Alfatide II was significantly higher in malignant lesions than in tuberculosis, which indicated that RGD-based PET/CT might differentiate lung cancer from tuberculosis (117). The 68 Ga-NOTA-RGD SUV max allowed for significant delineation of a solitary pulmonary nodule from non-malignant lung tissue (3.7 ± 0.7 vs. 0.4-0.9, p <0.001) (60). Consistent with the different avb3 expression patterns in NSCLC and small cell lung cancer (SCLC), NSCLC samples showed positive avb3 expression in both tumor cells and the vasculature epithelium, whereas SCLC exhibited inconspicuous avb3 expression. Tracer accumulation of 68 Ga-DOTA-RGD2 was higher in NSCLC (ranging from 5.7 to 1.2) than in SCLC (3.3 to 1.1) for both primary lesions and metastatic lymph nodes. In contrast, synchronous 18 F-FDG PET/CT examination did not show the differences between NSCLC and SCLC groups (111). 18 F-Alfatide imaging also showed statistically higher peak SUV (SUV peak ), SUV mean and total lesion angiogenesis (TLA) for squamous cell carcinoma than for adenocarcinoma (127). Given the low levels of avb3 expression in SCLC, 68 Ga-NOTA-3P-TATE-RGD was developed to overcome the insufficiencies of TATE imaging in NSCLC and RGD imaging in SCLC. The TBR with 68 Ga-NOTA-3P-TATE-RGD was significantly higher than that with 68 Ga-NOTA-RGD (6.06 ± 6.09 vs. 2.65 ± 1.19, p = 0.0344) in SCLC, or that with 18 F-FDG in neuroendocrine tumors (2.91 ± 1.71, p = 0.0234) (52). It is important to note that lung cancer and inflammatory lesions or inflammatory pseudotumors cannot be distinguished with RGD imaging, nor with FDG imaging (124). Based on these data, RGDrelated imaging may be an appropriate strategy for avb3associated therapy and imaging, which showed advantages on distinguishing between NSCLC and SCLC.

Esophageal Cancer
In a case report for a moderately differentiated carcinoma of the gastro-esophageal junction, 68 Ga-NODAGA-RGD tracer uptake was clearly localized in non-FDG-avid perilesional structures, demonstrating that angiogenesis imaging might be a crucial tool for early disease identification and localization, metastasis evaluation, and efficacy prediction (128). Dong

Breast Cancer
The RGD accumulation was thought to represent the sum of uptake in neovasculature and tumor cells according to the immunohistochemical staining in IDC patients (108). 18 F-AH111585 tracer uptake either homogeneously distributed in tumors or appearing within the tumor rim in metastatic breast cancer patients, which was consistent with the general principles of angiogenesis localization within the outer zones of breast tumors, where viable tumors exist (129)(130)(131)(132). Vatsa et al. reported the results of 68 Ga-DOTA-RGD2 PET/CT in patients with locally advanced breast carcinoma (LABC) and found both 68 Ga-DOTA-RGD2 and 18 F-FDG were similarly able to diagnose primary and metastatic lymph nodes, and RGD-based imaging showed additional value for distinguishing nonspecific 18 F-FDG uptake due to an infectious or inflammatory etiology (133). The sensitivity and specificity of 18 F-FPPRGD2 PET for identifying either breast lesions or metastatic lesions were higher than those for 18 F-FDG PET/CT (95.7% vs. 87.0%, and 100% vs. 57.1%) (63). The relative uptake values of glucose metabolism and angiogenesis imaging were also investigated. In newly diagnosed or recurrent breast cancer, all SUV max for 18 F-FPPRGD2 were lower than those for 18 F-FDG, although the differences were not statistically significant (63). Again, the 18 F-Alfatide II SUV max was also lower than the 18 F-FDG SUV max (3.77 ± 1.78 vs. 7.37 ± 4.48) in breast cancer lesions in the study of Wu et al., and the difference was significant (p <0.05). However, in HER2 (-) and ER(+) breast cancer lesions, 18 F-Alfatide II uptake was higher than 18 F-FDG uptake (88). It is worthy noting that RGD PET may neglect the distant metastasis in liver; for example, Kumar et al. found that 68 Ga-DOTA-RGD was not retained in liver metastases in 9 patients (16%) that were discovered with 18 F-FDG imaging (119,133). The high background activity of liver regions may contribute to this phenomenon (60).
The relative uptake values of RGD and FDG varied with the pathological or molecular phenotypes of breast cancer. Higher uptake values were observed in malignant compared with benign breast lesions on both 18 F-Alfatide II and 18 F-FDG imaging (all p <0.05). However, 18 F-Alfatide II imaging showed a higher sensitivity (97.6% vs. 78.6%), lower specificity (54.5% vs. 81.8%), lower Youden index (52.1% vs. 60.4%), and lower area under the curve (0.738 vs. 0.838) on SUV max analysis. These results indicate that 18 F-Alfatide II has comparable diagnostic value to 18 F-FDG, but may not be superior for identification of breast cancer (88). Concerning the molecular phenotypes in breast cancer patients, the SUV max from 68 Ga-NOTA-RGD-BBN ( 68 Ga-BBN-RGD) PET was lower in ER(−) primary lesions than in ER(+) primary lesions (p <0.01). Additionally, 18 F-Alfatide II uptake in the triple-negative subtype was significantly lower than that those in the luminal A and luminal B subtypes. However, no significant differences in uptake of either 18 F-Alfatide II or 68 Ga-BBN-RGD were observed between HER2(+) and HER2(−) subtypes in either patients with suspected breast cancer on screening mammography or patients who underwent breast cancer radical mastectomy (p = 0.6589) (88,116). In patients with large or locally advanced IDC, the RGD SUV max and SUV mean differed significantly between the HER2(+) and HER2(−) groups (2.90 ± 0.75 vs. 2.42 ± 0.59, p = 0.04 for SUV max ; 1.95 ± 0.53 vs. 1.60 ± 0.38, p = 0.04 for SUV mean ) and showed significantly higher values in the ER/PR(+), HER2(−) and ER/PR(+), HER2 (+) subgroups versus the ER/PR(−), HER2(−) subgroup in large or locally advanced IDC. 18 F-FDG uptake was higher in HER2 (+) breast cancer lesions than in the other 3 subtypes (triplenegative subtype, luminal A and luminal B subtypes) (88,89). Both RGD-based tracers offer great performance for identifying breast cancer, but not superior to 18 F-FDG. However, RGDbased tracers may be superior to 18 F-FDG in detecting breast cancer that is strongly ER(+) and HER2(−). This implies that differing HER2 status, even if the hormonal receptor status is the same, can cause a difference in FDG uptake. At the same time, different hormonal receptor status can cause a difference in RGD uptake, even if the HER2 status is the same.

Hepatocellular Carcinoma
As described above for the biodistribution of RGD-based tracers in the liver, two out of nine patients showed lower tracer accumulation in the HCC lesion as compared to the remaining liver parenchyma (72,73). A similar lack of RGD-based tracer uptake was reported in liver metastases in breast cancer patients (60,119). These findings were unexpected based on results demonstrating that 77% of investigated HCC specimens were highly vascularized with integrin avb3 expression compared with only 22% of normal liver tissue showing detectable avb3 expression (134), supporting the assumption that RGD uptake should be relatively higher in HCC than in normal liver tissue. The contradictory hypothesis that high RGD uptakes in HCC and later observations of hypointense tumors in the liver needs to be further clarified. The relatively clear background for 18 F-FDG imaging in the liver appears to result from elevated glucose-6phosphatase levels in liver cells. Overall, the research to date indicates the unsuitability and associated challenges of RGDbased imaging for HCC tumors or diseases with frequent liver metastases.
In the study by Withofs et al., 18 F-FPRGD2 PET/CT allowed estimation of both intra-and interindividual variability of integrin avb3 expression successfully, the correlation between 18 F-FPRGD2 uptake and avb3 expression varied among different cases. In whole tumor samples, 18 F-FPRGD2 uptake correlated significantly with avb3 expression. In the subgroup analysis of ccRCC, the 18 F-FPRGD2 PET signal correlated with integrin avb3 expression on tumor cells, whereas in the pRCC group, the signal correlated with integrin avb3 expression on blood vessels. Immunohistochemical examination on tumor cells showed that integrin avb3 expression was significantly higher on ccRCC than on pRCC (3.6 ± 2 vs. 2.14 ± 1.8, p = 0.0099). However, for 18 F-FPRGD2 uptake as a reflection of integrin avb3 staining on entire tumors, the mean SUV max did not different between ccRCC (4.1 ± 1.2) and pRCC (3.3 ± 0.7) (67). These results indicated that integrin avb3 may highly expressed on tumor cells in ccRCC and new blood vessels (endothelial cells) in pRCC. 18 F-FPRGD2 PET/CT can reliably depict avb3 expression in renal tumors but is only representative of angiogenesis when tumor cells do not significantly express integrin avb3.

Prostate Cancer
Neoplastic cells of human prostate cancer are characterized by the overexpression of both integrin avb3 and gastrin-releasing peptide receptor (GRPR) (135). 68 Ga-BBN-RGD targets both avb3 and GRPR and thus was tested for its ability to improve the detection rate of primary prostate cancer. 68 Ga-BBN-RGD PET/ CT could successfully revealed 3 of 4 (75%) lesions and all 14 metastasized lymph nodes. It could also detect 20 bone lesions in seven patients, among which 11 showed a positive signal on 99m Tc-MDP bone scans and only 5 of these lesions were detectable with MRI (74). In a recently published clinical study in prostate cancer patients using 18 F-Galacto-RGD PET, 58 of 74 (78.4%) bone lesions and 2 of 5 metastatic lymph nodes were identified. Thus, RGD imaging was feasible for achieving good visualization and detection of bone metastases due to low background activity in the surrounding normal bone tissue (136).
Importantly, prostate-specific membrane antigen (PSMA), a well-known target in prostate cancer, is not expressed in nonprostate tumors nor in healthy vasculature, but is expressed in the endothelial cells of tumor-associated neovasculature (137). PSMA expression appears to be prognostic biomarker in HNSCC, osteosarcoma, colorectal cancer, adenocarcinoma of the pancreas, lung cancer and others (138). In prostate cancer patients, the correlation of 18 F-Galacto-RGD uptake and PSMA expression was also investigated. For all patients including one outlier or not with a very high PSMA value of 1,935 ng/ml, a significant and inverse correlation was identical between PSMA expression and 18 F-Galacto-RGD SUV mean . Thus, RGD PET may serve as a new method for monitoring the status of PSMA.

Brain or Bone Metastasis
Almost all RGD-containing imaging agents showed no or very low tracer retention in normal brain tissue, as reported by preliminary studies and case reports, indicating that RGDbased radiotracer does not penetrate the BBB but can localize in tumor tissue through the disruption of the BBB (43,61). In a study of RGD-based imaging of brain metastasis, 18 F-Alfatide II PET could visualize all 20 brain lesions successfully, while only 10 lesions were visualized by 18 F-FDG imaging, and 13 by CT. Similar to uptake at other tumor sites, 18 F-FDG showed an overall higher SUV max (10.0 ± 5.7) than 18 F-Alfatide II (1.8 ± 1.1) in brain tumor lesions. However, 18 F-Alfatide II showed a much better TNR (18.9 ± 14.1) than 18 F-FDG (1.5 ± 0.5) in these brain lesions (47). A possible explanation is that the high physiologic glucose uptake in the surrounding normal brain tissues further enhanced RGD uptake, and low 18 F-Alfatide II PET uptake in normal tissue may allow clear diagnosis and evaluation of tumor metastases in the brain (139). Another possible explanation for these results is that angiogenesis, the target of 18 F-Alfatide II, is one of the key pathophysiologic processes in brain metastases (47,140).
In a study of RGD-based imaging of bone metastasis in breast cancer patients, 18 lesions (33%) were detected with both 68 Ga-DOTA-RGD and 18 F-FDG imaging in (119). In another larger prospective study comparing 18 F-Alfatide II and 18 F-FDG PET, bone metastases were divided into four groups: osteolytic, osteoblastic, mixed and bone marrow based on their characteristics (141). In 11 patients with a total of 126 bone metastasis lesions, 18 F-Alfatide II PET showed a higher positive detection rate than 18 F-FDG PET (92% vs. 77%). In the subgroup analysis, 18 F-Alfatide II PET was comparable to 18 F-FDG PET/ CT for the detection of osteolytic (100% vs. 90%) and mixed bone metastases (100% vs. 90%), but superior to 18 F-FDG PET/ CT for detecting osteoblastic (70% vs. 53%) and bone marrow lesions (98% vs. 77%). The sensitivity of 18 F-Alfatide II PET/CT for osteolytic, mixed, and bone marrow lesions was nearly 100% (120). The higher positive detection rate of 18 F-Alfatide II PET for bone marrow metastatic lesions may be due to either subclonal selection of integrin avb3-expressing tumor cell populations or upregulation of integrin avb3 in the bone microenvironment during the early phase of bone metastasis (142). Interestingly, RGD-containing tracer uptake was not lower than that of FDG in bone metastatic sites (4.27 ± 2.42 vs. 4.18 ± 2.58, p >0.05), which may be related to abundant integrin avb3 expression on the metastatic tumor cells, endothelial cells, and osteoclasts, whereas glycolysis is increased only within metastatic tumor cells and not in endothelial cells or osteoclasts (143,144).
Both types of tracers showed significantly lower accumulation in osteoblastic lesions than that in osteolytic and mixed lesions (p <0.001 for all comparisons). The SUV max of 18 F-FDG in osteoblastic lesions was also significantly lower than that in bone marrow metastases (p <0.05). However, the 18 F-Alfatide II SUV max did not differ between osteoblastic lesions and bone marrow metastases. Another case report with 68 Ga-DOTA-RGD2 imaging, it clearly depicted a left-sided frontal tumor without 18 F-FDG PET/CT uptake. Considering the limitations of 18 F-FDG PET in a setting of low glycolytic activity, 68 Ga-DOTA-RGD2 PET/CT may be a clinically useful imaging modality for early detection of recurrent osteosarcoma (145).

DEFINITION OF TUMOR SUBVOLUMES USING RGD-CONTAINING PEPTIDES AS RADIOTRACERS
For radiation treatment planning, imaging quality is vitally important for biological target volume definition. Beer et al. fused images from 18 F-Galacto-RGD PET and MRI and/or multislice CT (MSCT) scanning in HNSCC patients. The mean gross tumor volume (GTV) was 26.4 ± 20.2 cm 3 , and when a SUV threshold >3.0 was applied, the mean tumor subvolume was only 3.4 ± 3.4 cm 3 , accounting for 13.1 ± 11.7% of the GTV (107). These results confirm the imaging quality and heterogeneous subvolume achieved with RGD-containing peptide tracers, which allows for adequate manual fusion with MRI or MSCT using anatomic landmarks. As displayed in a HNSCC patient population by Durante et al., 68 Ga-NODAGA-RGD and 18 F-FDG PET/CT individually led to inconsistent tumor subvolume delineation. The tracer avid tumor volume (TATV) calculated from RGD imaging tended to be larger than the metabolic tumor volume (MTV) calculated from 18 F-FDG imaging when a 42% SUV max fixed threshold similarly to MTV delineation was used (p = 0.085) (87). That difference in tumor delineation is assumed to be attributable to the difference in tracer targeting between FDG and RGD, as mostly homogeneous FDG uptake is significantly correlated with glucose metabolism and heterogeneous RGD uptake is associated with avb3 on angiogenesis. In addition, the physical properties of radionuclides fluorine-18 and gallium-68 also should be considered. At the same time, the 42% SUV max of RGD was much lower than that of 18 F-FDG, which may also have contributed to the calculation of a larger volume for the same tumor. In LARC cases examined with 18 F-FPRGD2 imaging, the baseline integrin tumor volume (ITV) 70% (ITV 70% , delineated using a threshold of 70% of the SUV max ) was moderately correlated with the MTV 40% (delineated using a threshold of 40% of the SUV max ) and total lesion glycolysis (calculated as SUV mean × MTV 40% ) (66).
However, such results have not been reproduced in Lewis lung carcinoma (LLC) tumor-bearing C57BL/6 mice. The GTV determined from 18 F-Alfatide PET was smaller and much closer to the pathological volume than that from 18 F-FDG PET although a similar 40% of the SUV max threshold was used (146). Several points are worth discussing regarding this difference. First, the expression pattern of avb3 varies among cancer types; it is mainly expressed on the neovasculature of HNSCC but on both the neovasculature and tumor cells in lung cancer (107,111,147). Secondly, different radionuclides were used in the two studies with 18 F-Alfatide and 68 Ga-NODAGA-RGD. Additionally, differences in metabolism between animals and humans may also influence the results. Importantly, it is unscientific to define the cutoff value for tumor subvolume delineation in reference to the established threshold for FDG. Instead, threshold optimization for estimating tumor volume based on a spatial comparison of avb3 expression on wholetumor histological slices needs to be performed.

ASSESSMENT OF THERAPEUTIC RESPONSE AND ADVERSE EVENTS USING RGD-CONTAINING PEPTIDES AS RADIOTRACERS
Because RGD is a marker of avb3 expression on angiogenesis which is of paramount importance for tumor oxygenation, RGD-containing PET tracers have been considered as a potentially valuable tool for monitoring treatment response and identifying patients less likely to respond to therapy. Several clinical pilot studies have explored their value in assessing treatment efficacy and adverse events, with encouraging results for both antiangiogenic therapy and chemoradiotherapy. Knowledge of the precision of repeated imaging parameters is a prerequisite for serial response measurements and, therefore, response to treatment. In the study by Sharma et al., solid tumor patients underwent two 18 F-AH111585 imaging sessions within an interval of 10 days before commencement of therapy. No significant change in the SUV peak was observed between the two scans, and the 95% reference range of spontaneous fluctuations was 35-39% (148). Measurements of 18 F-FPPRGD2 uptake in normal organs also showed no significant changes between pre-, 1-week post-and 6-week post-therapy scans (149). Thus, with their protocol, the serial reproducibility of RGD-based tracer uptake supports further investigation of these novel tracers as biomarkers of treatment response ( Table 2).
Significant differences in RGD-based tracer uptake before and after antiangiogenic therapy have been reported in several studies (90,97,149). In GBM patients receiving bevacizumabcontaining therapy, decreases of less than 15% in the SUV max and angiogenic volumes from 18 F-FPPRGD2 imaging at 1 week predicted recurrence and poor prognosis, while decreases of more than 50% predicted good outcomes to treatment (149). In another study including ovarian and cervical cancer, patients who experienced clinical disease progression, complete response, and partial response showed decreases in the lesional 18 F-FPPRGD2 SUV mean of 1.6, 25.2, and 7.9%, respectively, with similar decreases in the 18 F-FDG SUV mean of 9.4, 71.8, and 76.4%, respectively (90). The above studies mainly evaluated the changes in tracer uptake with the application of different antiangiogenic combination therapies, which may confound our understanding of RGD-based imaging for monitoring the response. Li et al. successfully investigated the relationships of 18 F-Alfatide with the response to single antiangiogenic therapy in solid malignancies, higher pretreatment SUV peak and SUV mean were seen in responding tumors compared with non-responding tumors (4.98 ± 2.34 vs. 3.59 ± 1.44, p = 0.048; 3.71 ± 1.15 vs. 2.95 ± 0.49, p = 0.036). Cases with an 18 F-Alfatide SUV mean above 3.82 showed longer progression-free survival (PFS) than those with a SUV mean less than 3.82. These results provided clinical evidence of the effectiveness of RGD-based PET at baseline for predicting the response to antiangiogenic therapy, suggesting this approach could potentially replace indicators based on change ratios in most previous studies (150). In addition, they also reported the predictive value of RGD-based imaging for antiangiogenic adverse events for the first time. Patients with low maximum 18 F-Alfatide SUVs in the liver, spleen, and cardiac tissue were more likely to experience fatigue, hypertension, and nausea during Apatinib treatment (151). Thus, higher 18 F-Alfatide uptake in specific normal organs predicted less adverse events, whereas higher uptake in tumor lesions predicted a better response to therapy. Higher accumulation in normal organs may indicate more regular, functional vasculature and sufficient blood flow, and higher accumulation within tumors may indicate greater expression of avb3 integrins (155,156).
Results suggesting RGD-based tracer uptake at baseline can predict the response to antiangiogenic therapy have also been obtained in patients with platinum-resistant/refractory ovarian cancer (152). Of interest, a negative trend was observed between the baseline SUV 60,mean and PFS. There are several possible explanations for this contradictory observation. First, unlike the study by Li et al., the patients in this study received antiangiogenic combination therapy. It is well established that the addition of chemotherapy also enables a reduction in tumor size, which may account for the lack of association. Secondly, the PET response was assessed after 1 week, while the RECIST 1.1 response was evaluated after 3 months. Finally, preclinical work suggests that normalization of vessels occurs within hours and lasts 7-10 days, it is possible that performing the second PET scan at 7 days may have missed the neovascularization window (157). Overall, further research is needed to explore the potential of RGD-based tracers as biomarkers for response to antiangiogenic and antiangiogenic combination therapies. The predictive value of RGD-based imaging for the response to chemoradiotherapy was also studied. Twenty-five newly diagnosed GBM patients who were ready for concurrent chemoradiotherapy (CCRT) 3-5 weeks after surgical resection were prospectively enrolled in a study by Zhang et al. The baseline 18 F-Alfatide SUV max , the 3-week SUV max , and the TNR were found to be predictive of sensitivity to CCRT. The 3-week SUV max (AUC = 0.846) was shown to be the best predictive parameter, with a 3-week SUV max higher than 1.35 predicting worse efficacy (153). Further, Luan et al. explored the similar potential value of 18 F-Alfatide PET/CT in locally advanced NSCLC. From their analysis, the pretreatment SUV max , SUV peak , TNR, TBR, and TMR were all significant parameters, with higher SUVs in non-responders than responders, and the TNR was shown to be an independent predictor (nonresponders vs. responders, 8.31 ± 0.61 vs. 6.53 ± 0.78, p <0.001) to CCRT (154). Chen et al. prospectively investigated the potential usefulness of RGD-K5 PET in HNSCC, patients whose primary tumors had lower accumulation of both tracers at 3 months achieved response to CCRT successfully (99). For LARC patients treated with CCRT, baseline 18 F-FPRGD2 SUV max and SUV mean were negatively correlated with the tumor regression grade (TRG). The baseline SUV max of both RGD and FDG imaging were significantly higher in TRG 0 cases than in the other patients. A cut-off value of 5.6 for RGD SUV max identified TRG 0 status with 100% sensitivity and 66.7% specificity (AUC = 0.84) (66)..
From the research described above, it is interesting to note that RGD-based tracer uptake follows different trends according to differing treatment modalities. For antiangiogenic therapy, the higher uptake values may indicate a high density of effective target receptors and a better response to apatinib therapy. For CCRT, higher uptake values may represent high malignancy and severe hypoxia, which could contribute to increased chemoradiation resistance in tumors (158)(159)(160).

CONCLUSION AND OTHER PERSPECTIVES
A variety of radiolabeled, RGD-based PET imaging agents have been developed and extensively investigated in terms of their value for tumor detection and diagnosis, patient stratification, and monitoring of treatment response. According to promising results, these PET imaging peptides were transferred successfully from preclinical animal studies to clinical human trials. Different RGD-based PET imaging can be used in patients with glioma, lung cancer, esophageal cancer, breast cancer, renal carcinoma and other tumors to differentiate benign and malignant lesions and also pathological types and stages, and also can further distinguish molecular subtypes of breast cancer, delineate tumor volume, and predict the efficacy of radiochemotherapy and antiangiogenic therapy. The results regarding their diagnostic and predictive value are mostly derived from preliminary studies, and continued research is necessary to guarantee their development for clinical application.
In addition, RGD-based imaging was useful as a noninvasive surrogate parameter of cardiovascular disease. Normal angiogenesis under pathological conditions in cardiovascular diseases such as carotid atherosclerosis, myocardial infarction or reperfusion were also characterized by the overexpression of avb3 (3,4,161). In atherosclerotic lesions, highly expressed avb3 were found in both macrophages and activated endothelial cells (105,162). The uptakes of 68 Ga-DOTA-RGD and 18 F-GalactoRGD PET increased in the atherosclerotic plaques in mice (95,163). Clinically, RGD uptakes were heterogeneously distributed and significantly correlated with avb3 integrin staining score in human atherosclerotic carotid plaques (164,165). For ischemia tissue repair after myocardial infarction, avb3 is one of the necessary integrins involved in the formation of new capillaries (25,166). RGD-based imaging of myocardial infarction observed consistent findings from studies of rats, specific RGD tracer accumulation uptake increased in vivo corresponding to infarcted and peri-infarct myocardial regions, and decreased after reperfusion (167)(168)(169)(170). Tracer uptakes were expected to be used to monitor the process of myocardial repair (171)(172)(173). Makowski et al. confirmed the feasibility of 18 F-Galacto-RGD to monitor vascular remodeling during evolving MI in human studies; tracer uptake was significantly correlated with infarct size and impairment of myocardial blood flow (174). Thus, RGD-based imaging of avb3 integrin expression could represent a novel promising imaging approach in determining the angiogenesis of carotid atherosclerosis and myocardial repair, which is potentially affecting the plaque vulnerability and left ventricular remodeling.
The RGD peptides also served as the new target for tumor treatment. RGD peptides could significantly reduce the functional vascular density, inhibit the adhesion and induce apoptosis of tumor cells according to inhibiting the interactions between integrins and their ligands (175)(176)(177). The RGD modified gene carriers and target drugs also showed special interest in delivery (178,179). RGD peptides for imaging or tumor treatment all need to be further investigations for clinical translation.
It is important to note that contradictory reports about the correlation between RGD tracer uptake and avb3 expression level exist and that the ultimate meaning of avb3 expression in the context of angiogenesis is very complex and requires further characterization. The highly abnormal and dysfunctional vasculature created tumor hypoxic environment and impaired the ability of immune effector cells to penetrate solid tumors. In turn, stimulation of immune cell could also help normalize tumor vessels (180). Other studies have demonstrated the expression of integrin avb3 on activated macrophages by different methods (162,181,182). Therefore, RGD-based imaging targeting tumor neovascularization may offer a strategy to monitor the tumor microenvironment status and therapeutic response during immune checkpoint blockade or combination treatment. Moreover, the development of the clinically available applications of RGD-based tracers has mainly focus on radionuclide-labeled tracers, approaches to develop molecularly targeted nanoparticles for molecular imaging should be explored. Additionally, the simultaneous acquisition of functional parameters by MRI and PET may help compensate for the shortcomings of these different approaches. A thorough and complete assessment of different aspects of angiogenesis observed via multimodal imaging and biomomics characteristics should be implemented as part of the concept of personalized functional medicine.

AUTHOR CONTRIBUTIONS
SY proposed the conception and edited the manuscript. LL participated and drafted the manuscript. XC and JY participated and edited the manuscript. All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.