Noninvasive assessment and quantification of tumor vascularization using [18F]FDG-PET/CT and CE-CT in a tumor model with modifiable angiogenesis—an animal experimental prospective cohort study

Two human non-small-cell lung cancer (NSCLC) cell lines (A549 and H1299) were examined and rat glomerular endothelial (RGE) cells were used to manipulate vascularization levels. Cells were cultivated in high-glucose Dulbecco’s modified Eagle medium (DMEM) with heat-inactivated 10% fetal calf serum and 1% non-essential amino acids at 37 °C in a humidified atmosphere containing 7% CO₂ as published before [21].

Four- to six-week-old athymic nude rats were used. Animals received food and water ad libitum. To allow standardized growth conditions by decreasing residual immune response, all rats received a uniform whole body irradiation of 4 Gy [22, 23]. The rats in the animal experimental prospective cohort study were assigned randomly to one of four experimental groups (Fig. 1). The tumor cells were transplanted 48 h after irradiation by a subcutaneous injection of 200 μl tumor cell solution (5 × 10⁶ tumor cells) into the rat’s right lower limb. In both groups with modified vascularity, this tumor cell solution contained additionally 2 × 10⁶ RGE cells, 160 ng recombinant human vascular endothelial growth factor 165 (rHu VEGF-165), and 320 ng recombinant human fibroblast growth factor b (rHu FGF-b; Fig. 1).

After transplantation, the animals were weighted every other day, the transplantation region was inspected, and the tumor’s length (a) and width (b) were measured by caliper (volume = 1/6 ∙ π ∙ a ∙ b²). As soon as any tumor had reached a maximal size of 20 mm in one dimension, the rat was subjected to the multimodal imaging protocol, but not later than 49 days after transplantation.

The approval by the local animal care committee was obtained in accordance with the institutional guidelines and the German animal welfare regulations (24D-9168.11-1/2007-3).

The imaging of tumors included [18F]FDG-PET/CT, CE-CT, and histological examinations.

For [18F]FDG-PET/CT and CE-CT the PET/CT, “biograph 16 Hi-Rez” (Siemens, Knoxville, TE, USA; Fig. 2a) and “syngo TrueD” software were used. After 24 h of fasting, the rats received an intraperitoneal (i.p.) injection of 11.1 MBq (SD 1.2 MBq) [18F]FDG followed by intraperitoneal injection of ketamine (90 mg/kg) and xylazine (10 mg/kg) 20 min later for anesthesia [24]. In this study, the i.p. way was used due to the very easy and safe injection compared to intravenous (i.v.) injection into the tail vein of the rats. Due to the equal blood concentration of [18F]FDG reached by the i.p. injection [25, 26], the authors neither expect a better nor a worse visualization of tumors in respect to injection technique. Subsequently, a catheter was implanted to the nude rat’s jugular vein (24 gauge peripheral venous catheter, B. Braun Melsungen AG, Melsungen, Germany). The catheter was used for i.v. application of contrast agent (CM) Ultravist 370, Bayer Schering Pharma AG (Bayer AG, Leverkusen, Germany) during CT imaging protocols. Accomplished CT protocols were attenuation correction CT, plain CT without CM, and venous CT with CM. After the plain CT scan, the injection of CM started and 1 ml of CM was applied in 30 s. CT scanning started 20 s after the injection start. All CT protocols were carried with the animal in the same position.

The following scan protocols and specifications were used. PET: scan start 30 min post intraperitoneal [18F]FDG injection; matrix 256 × 256; voxel size 2.05 × 2.05 × 1.5 mm; slices 109; CT for transmission correction: 120 kV; 100 mAs; matrix 512 × 512; voxel size 0.47 × 0.47 × 0.75 mm; CT: 80 kV, 100 mAs, matrix with 512 × 512 pixel, voxel resolution of 0.42 × 0.42 × 0.75 mm.

After PET and CT imaging, 200 μl of 2.4 x 10exp7 fluorescent beads (diameter 2.5 μm, excitation wave length 633 nm; G. Kisker GbR, Steinfurt, Germany) were injected via the jugular vein of the rats. The rats were sacrificed 2 min after injection. Subsequently, the organs and the tumor were removed and conserved in liquid nitrogen. Representative slices (thickness 10 μm) of organs were stained with hematoxylin-eosin (HE). Tumor slices were additionally stained with cluster of differentiation 31 (CD-31) antibody to quantify microvessel density (MVD).

Histology and bead counts (number of beads per field of view) served as standard of reference. The beads were examined at fivefold magnification using a fluorescent microscope in 20 fields of view (FOV). For CD-31 staining, five hot spots for each tumor were digitally recorded at 20-fold magnification and the number of CD-31 positively stained structures was evaluated as published before [27], with MVD estimated as the number of stained structures per mm². The estimated parameters were beads/FOV and MVD/mm².

For analysis of [18F]FDG-PET/CT data, tumors were visually assigned to one of two groups: heterogeneous or homogenous pattern of [18F]FDG distribution (Fig. 3). Contours were drawn, and SUVs were estimated for both the brain (Fig. 2e) and a representative part of the muscles in the left lower thigh. Contours were drawn surrounding the tumor in the CT images and the SUV in the tumor was estimated for three different lower limit values—A: 50% of brain SUVmean, B: 50% of tumor SUVmaximum, and C: 1.5 x muscle SUVmean [28, 29]. By applying each lower threshold (A, B, C), three corresponding metabolic active volumes were estimated (tumor volume exhibiting a SUV of the threshold or higher). Relative active volume was estimated by relation of the active volume to the gross tumor volume in CE-CT. In CE-CT series, an outline was drawn around the contrast-enhanced tumor in all slices, including outer edge of tumor, skin, and CM-lacunas but excluding bones and vessels without contact to the tumor. Limits for analysis of the entire tumor were − 50 ≤ Hounsfield Unit (HU) ≤ 350. By modifying these limits to + 50 ≤ HU ≤ 350 within the same outline, the characteristics of the non-necrotic (vital) parts of the tumors were investigated. Characteristics of the non-contrast-enhanced tumor were determined by copying this outline to the plain CT series.

Estimated parameters in [18F]FDG-PET/CT: minimum, maximum, and mean SUV; metabolic active volume; relative metabolic active volume; [18F]FDG-pattern; in CE-CT: tumor volume, vital and relative vital tumor volume, contrast enhancement (CE) in vital tumor volume in relation to maximum HU in plain CT series.

All statistical analyses were carried out with PASW 18 (Predictive Analytics SoftWare, IBM, Armonk, NY, USA) by comparisons between co-transplanted and non-co-transplanted tumors with and without regard of the tumor cell line. The Kolmogorov-Smirnov test was used to test for normality. The Mann-Whitney U test was used to analyze parameters that violated normality, whereas Student’s t test was used when normality was given. Spearman’s rank correlation coefficients were calculated to evaluate correlations. The Chi²-test was used to compare the results of [18F]FDG distribution pattern. A p value lower than 0.05 was considered to indicate a statistically significant difference.

The co-transplanted tumors revealed a significantly faster growth compared to the non-co-transplanted tumors. This holds true for both tumor cell lines H1299 and A549 (Table 1; Fig. 4). Both the MVD (Table 1; Fig. 5) and the number of beads (Table 1; Fig. 5) were higher in co-transplanted tumors, while the number of beads in the investigated organs did not differ between the groups. Histology revealed that the central regions of the co-transplanted tumors often showed clusters of necrotic regions next to vessels with larger diameters (Fig. 5). Notably, the latter were often closely connected to clouds of CD-31-positive small vessels (Fig. 5). Both necrotic areas and CD-31-positive vessels with large diameter were rarely found in the center of non-co-transplanted tumors (Fig. 5).

There were significant correlations between [18F]FDG-PET/CT, CE-CT, and histology. The correlation between the maximal SUV in tumors and the beads/FOV as assessed by fluorescent microscopy was found to be significant in all tumors (p = 0.005, r = 0.353, Fig. 8). Results are similar for the correlation between the maximal SUV and the microvessel density as assessed by CD-31 staining in immunohistology (p = 0.036, r = 0.294, Fig. 9). The correlation between both histological quantifications of vascularization (beads and CD-31) also was significant (p = 0.007, r = 0.376). The correlation between the metabolic active volumes in [18F]FDG-PET and the vital volumes in CE-CT was found to be significant in all tumors too (p < 0.001; r = 0.919, Fig. 10); calculation was done with SUV threshold A (50% of brain SUVmean). Furthermore, the correlation between the maximal SUV in tumors and the contrast enhancement in the vital part of the tumors was found to be significant in all tumors (p < 0.001; r = 0.635; Fig. 11).

Glucose and [18F]FDG accumulation in tumors dependent, among other things, on tumor vasculature (i.e., deliver the tracer to the region of interest), on glucose transporters (i.e., intracellular uptake of the tracer), on phosphorylation (i.e., trapping the trace into the cell), and on metabolism (i.e., need and use of the tracer) [33]. The amplified expression of various transmembrane glucose transporters (GLUTs) and intracellular enzymes (e.g., hexokinases, HK) are common features in a multitude of tumors. This seems necessary because of higher glucose consumption due to altered metabolism (aerobic glycolysis) in tumors [9, 34]. Both GLUTs (esp. GLUT-1, GLUT-3) and the high active hexokinases (esp. HK-II) are two reasons for the increased influx and trapping of [18F]FDG in tumor cells [35‐38]. All of this influences the uptake of [18F]FDG and thus the measured imaging parameters. This must be taken into consideration when interpreting [18F]FDG-PET/CT data for additional conclusions about vascularization.

A more focused approach assessing vasculature is the application of an intravascular tracer not being influenced by transporter proteins or metabolism. Such tracers are delivered with the blood in the vessels. This study used intravenous fluorescent beads as such robust tracer, with the need of post-mortem quantification. Linking these tracers to radioactive markers enables in vivo analysis. The preparation of such a tracer (e.g., [99mTc]Phytate, starch-based microparticles, [99mTc]Albumin) is rather complex and the behavior of these tracers in the body have to be learned in clinical studies [39‐41]. Of course, these methods are less influenced by confounders compared to above mentioned [18F]FDG analysis, but in return they can hardly provide any further information as [18F]FDG-PET/CT does. For example, they cannot image the activation of metabolic pathways which are connected to angiogenesis. Due to the molecular link between glucose uptake and angiogenesis, [18F]FDG-PET may do.

Another approach to observe angiogenesis is to identify hypoxic regions driven by the idea that hypoxia causes both angiogenesis and high glucose metabolism (anaerobic glycolysis) in tumors. The transcriptional activator hypoxia-inducible factor-1 (HIF-1) normally becomes stabilized in cells under hypoxic conditions, and then leads to angiogenesis. However, in cancer, some data do not show correlations between the expression of hypoxia-inducible factor (HIF-1) and regions defined as hypoxic [30, 42]. Imaging hypoxia then barely helps to capture angiogenesis. But what if connection between metabolism and angiogenesis is stronger than between hypoxia and angiogenesis? Finding hypoxic regions would then be less important [30]. Detecting glucose metabolism using [18F]FDG-PET/CT might then indicate for angiogenesis. This may refer to a closer connection between angiogenesis and glucose metabolism. In this study, the authors decided to explore this connection. They choose [18F]FDG as imaging tracer to investigate whether molecular links between glucose and vascularization are reflected in parameters derived in [18F]FDG-PET.

The molecular links between the pathways of angiogenesis and glucose uptake and metabolism provide the foundation for this study. Therefore, in the following, it is worth putting the spotlight briefly on the molecular level.

Mutated pathways in cancers often involve both metabolism and angiogenesis; the transcription factors HIF-1 and p53 are two important examples. Due to mutation, the overexpression of HIF-1 has been found in various cancers [12, 34, 43]. It leads to activation of glycolysis and to inhibition of oxidative phosphorylation, thus HIF-1 influences glucose metabolism [12, 34]. HIF-1 also drives the synthesis of VEGF, which promotes angiogenesis [34]. The mutation of the P53-gene in cancer cells lead to amplified expression of GLUT-1, GLUT-3, and GLUT-4 and consequently to increased glucose uptake in cancer cells and in consequence to a higher [18F]FDG uptake in PET [36, 37, 44]. In addition, the deficiency of p53 leads to the upregulation of proangiogenic and to the downregulation of antiangiogenic factors [45]. Thus the (functional) inactivation of p53 also influences both metabolism and angiogenesis [16, 34, 44‐46].

This exemplary reflects the close relation between angiogenesis and glucose (metabolism). So, [18F]FDG-PET should detect much more than just the metabolic activity or accumulation of tracers only due to perfusion. The present study demonstrated a significant influence of tumor vascularization on estimated parameters in [18F]FDG-PET.

An important condition for both fast and continuous tumor growth is the angiogenic switch. The tumor model used in this study increased the probability of a more intense angiogenic switch leading to a better vascularization in these tumors. The histological and immunohistological results support this interpretation, the co-transplanted tumors revealed the better vascularization (higher MVD and higher number of beads). Tumor vessels are not as functional as normal vessels, but they are perfused [47]. This is proven by the higher number of fluorescent beads in tumors with higher MVD, as beads can reach the tumor only by perfusion. Thus, co-transplantation of RGE cells and growth promoters lead to improved vascularization of the tumors. Because the cell line and the amount of tumor cells remained the same, the faster growth rates in the co-transplanted tumors can be explained by a better supply with oxygen and nutrients due to better vascularization. Some tumors in the non-co-transplantation group also reached comparable sizes; however, the number of tumors that did and their rate of growth were much smaller. This indicates an occurred but less intense angiogenic switch in these tumors.

In this study, [18F]FDG-PET and CE-CT parameters seem to be connected to tumor vascularization. Different explanations must be taken into consideration.

Maybe the regulation of both angiogenesis and glucose metabolism is a result of different pathways. However, [18F]FDG uptake seems to be an indicator for vascularization of malignant tumors. This study could successfully show correlations between glucose metabolism and vascularization. The results are consistent with similar studies [55‐57]. Other groups find contrary results for comparison of perfusion imaging and [18F]FDG-PET [58, 59]. Sauter et al. [60] found negative correlations for SUV and MVD but positive correlations for perfusion CT and MVD. Accordingly, imaging parameters of [18F]FDG-PET may be directly connected to the degree of vascularization of tumors. The correlations found between SUV and beads/FOV and SUV and CE in CT in vital tumor support this assumption. Xing et al. [55], Yokobori et al. [61], and Kaira et al. [57] could also show correlations between SUV and MVD in pulmonary tumors. Whereas the latter found only weak correlations in patients with secondary pulmonary tumors. Tateishi et al. [56] also found correlation between SUV in [18F]FDG-PET and MVD in pulmonary tumors. Interestingly, this was not true for benign lesions. Using only two tumor cell lines as in this study avoided the influence of cell characteristics on imaging. All these results may be allegeable if the regulation of both angiogenesis and glucose metabolism is discussed. As described before, both processes are influenced by the same regulator proteins, among others HIF and p53. Based on mutation or hypoxia, HIF in cells becomes stabilized. Both, angiogenesis via VEGF and glycolysis via induction of HKII, GLUT-1, GLUT-3, and GLUT-4 are promoted [12, 16, 34]. The mutation of the P53 gene, often found in cancer cells, rules both processes similarly. The absence of functional p53 (due to mutation) leads to both a stop of inhibition of glycolytic enzymes and to the promotion of angiogenesis [16, 34, 44‐46].

In tumors, vascularization is associated with higher aggressiveness. If high [18F]FDG-uptake is associated with higher vascularization, one perhaps can hypothesize with reservations that high [18F]FDG-uptake in some tumors may hint for higher aggressiveness. It was shown recently that local relapses of NSCLC tumors after radiation therapy are significantly more often found in pretherapeutic high [18F]FDG uptake tumor areas [62]. This raises the question which cancer cell traits in these areas lead to both the high glucose metabolism and the high malignant potential. In other tumors, the data for [18F]FDG-uptake and aggressiveness are contrary. Due to different imaging protocols, the measured SUV values are hard to compare. That makes it difficult to establish cut-off values for estimating tumor aggressiveness and prognosis.

The threshold for defining “metabolic active” is from tremendous meaning at this point. Above described results hold true for a lower threshold of 50% of the particular rat’s brain SUVmean. By using a threshold of 50% of the rat’s tumor SUVmax, both volumes change if tumors contain large parts with low glucose turnover. In non-co-transplanted tumors, this threshold is closer to tumor SUVmean than in co-transplanted tumors. In consequence, nearly the entire tumor volume is defined as “metabolic active.” The significant differences remain between both groups, but in the opposite direction (Table 2). This fact mirrors the finding of heterogeneous [18F]FDG pattern and coexistence of necroses and vital tumor tissue in co-transplanted tumors. The findings in this study also underscore the tremendous meaning of the level of thresholds selected. Their importance increases if they lay the foundation for determination of target volumes for radiation therapy.

Although in this study correlations between imaging parameters in [18F]FDG-PET/CT and vascularization were detected, it must be emphasized that correlations not necessarily mean causality. Especially in [18F]FDG-PET, the abovementioned factors influencing the tracer uptake must be considered. In the high complexity of cancer, conclusions should never be driven on the basis of one single method. In this way, [18F]FDG-PET/CT should be used when estimating vascularization in tumors, as a further piece for a very complex puzzle. A limitation of this study is due to the fact that human tumor xenografts were grown in animals. It changed the original microenvironment of the tumors, but it enabled an easier translation into clinic using human cancer cell lines, the exclusively use of clinically approved scanners and imaging protocols, and by alteration of microenvironment within the same tumor cell line. Although the technique for the estimation of the MVD is well established, it is limited because it only focuses on some hot spots of the tumors and may not represent the entire tumors’ vascularization. The random selection of evaluated tumor areas should have minimized this limitation. The analyzed tumor heterogeneity may not only help to characterize tumors, it may also influence other estimated parameters in an unknown manner. Furthermore, the partial volume effect in PET and CT analyses may hamper some results. Because it is clinically well established, this study only focuses on static SUV estimation. However, dynamic [18F]FDG-PET registration may add further information in subsequent studies. Already dual time point imaging can expand the information obtainable from [18F]FDG-PET imaging. Nakajima et al. were able to differentiate histological types of renal cell carcinomas by early imaging after [18F]FDG application and raise hints for aggressiveness by comparison of [18F]FDG accumulation over time [63]. Wu et al. could show that dual time point analyses in [18F]FDG-PET can be used for further evaluation of bone lesions [64]. Estimating PET parameters from tracer kinetic modeling uses dynamic PET for continuously collecting data during a certain timeline. In order to calculate parameters from the compartment model, an arterial input function is mandatory for calculating tracer kinetics. To avoid the gold standard of blood sampling, most studies used image derived input functions. [18F]FDG is irreversible trapped intracellular. This means [18F]FDG exhibits an unidirectional transfer from the blood into the cells; it is for this reason why for calculations of transfer constants the Patlak analysis can be used [65]. Compared to static or dual time point imaging, the dynamic [18F]FDG-PET much more deepen the information extracted from imaging by determining parameters from tracer kinetic [66]. They can help to further understand pathophysiological mechanisms or the dignity of tumors [66]. Although considering the complexity of the analysis, Wu et al. emphasize that dynamic [18F]FDG-PET is more helpful in discriminating between the dignity of bone lesions [64].

While morphologic imaging exhibits high reproducibility, the results of functional imaging methods like [18F]FDG-PET and perfusion imaging are dependent from the methods and deepness of the selected type of anesthesia. In a preliminary study, the authors evaluated the used imaging protocols that revealed a reliable distribution of [18F]FDG-PET through the whole animal. Several studies compared different application forms of [18F]FDG-PET (e.g., intravenous, i.v. and intraperitoneal, i.p.) and explored factors influencing tracer distribution in rodents [67, 68] and revealed a comparable increase in [18F]FDG-PET uptake in xenograft tumors after i.v. as well as i.p. application [69]. Furthermore, the type of anesthesia influences imaging results. The authors in this study used i.p. ketamine and xylazine after tracer application to minimize the effects on tracer metabolism. Ketamine is not supposed to suppress cardiac output significantly [25] and inotropic and chronotropic effects are dose dependent [70]. Other studies showed that different narcotics may cause worsening of cardiac function [70‐72]. Summarized, all anesthetics may cause worsening of cardiac function influence haemodynamics, all with their inherent advantages and disadvantages. Due to the used doses in this study, the authors believe, in line with the literature, that there were no relevant cardiodepressive effects that could have disturbed the distribution of [18F]FDG-PET in this study.