Introduction

Tumour vessels in tumour angiogenesis are structurally and functionally abnormal, and are essential for tumour growth, progression, and metastasis. The inhibition of tumour vasculature can be a valuable target for cancer therapy. Tumour vascular targeting agents (VTAs) [1, 2] can be classified into either angiogenesis inhibitors [3, 4] such as antiangiogenic drugs, which inhibit new vessel formation by interrupting essential aspects of angiogenesis, or vascular disrupting agents [5], which cause the damage of the previously established tumour endothelium and vascular shutdown, subsequently resulting in tumour ischemia and necrosis.

AG-013736 (Axitinib), the small molecule tyrosine kinase inhibitor that targets all known vascular endothelial growth factor (VEGF) receptors, potently inhibits angiogenesis and has single-agent clinical activity in a range of solid tumours [6–8]. The development of novel therapeutic agents targeting angiogenesis is challenging without a validated non-invasive biomarker of an anti-angiogenic effect in humans. A suitable biomarker may increase the probability of detecting a true antitumor effect, as well as predict clinical activity after treatment in a short course. Computed tomography (CT) imaging-based therapy monitoring has gained an important role in oncologic imaging [9–11]. Tumour size change in conventional CT has been widely used to assess antitumor response [11–13]. However, it cannot be relied upon for estimating tumour response to due to its insensitivity or its delayed identification [12, 14]. The use of perfusion CT functional imaging in tumour angiogenesis has also been limited in clinical practice owing to low spatial resolution [15] and poor image quality caused by the low tube voltage [16], posing new demands on CT functional imaging techniques.

Dual-energy spectral CT using a single-detector, single-source with rapid-kV-switching between two peak voltage settings(140 kVp and 80 kVp), allows the reconstruction of accurate material-decomposition (MD) images (e.g., water- and iodine-based MD images) for iodine concentration (in milligrams per millilitre) derivation. Multiple studies have demonstrated the validity of iodine concentration quantification in material differentiation, material identification and tumour response after radiotherapy or chemotherapy [17–21]. CT spectral imaging parameters might be useful in repeated monitoring of the microenvironmental change of tumours after administration of antiangiogenic agents, because antiangiogenic agents would affect the tumour vasculature and result in necrosis of the tumour, causing the change of iodine concentration in tumours after iodinated contrast agent administration. However, to date, there are few studies dealing with the use of spectral CT for evaluating therapeutic response of AG-013736.

The current study aims to assess the capacity of using CT spectral imaging parameters to evaluate the therapeutic response of AG-013736 in inhibiting VX2 liver tumours in rabbits.

Materials and methods

Animal model

The ethics review committee of the life sciences in our institution approved the animal experiments. New Zealand male rabbits (Experimental Animal Center of Zhengzhou University) weighing 2.5–3.5 kg were anesthetized with intramuscular injection of xylazine hydrochloride ( Shengda Co., Jilin, China) at a dose of 0.2 ml/kg body weight. The rabbit VX2 tumour model was used because of its hepatic artery blood supply [22] and biological progression [23] similar to that of human hepatocellular carcinoma. The intramuscular growth of VX2 tumour-bearing rabbit was harvested under the same anaesthesia when tumour portions grew approximately 4 cm3 in volume. Chunks of the tumour were minced into 2.0 mm3 cubes and implanted into the left liver lobes through a 17 gauge catheter under CT (Discovery CT590 RT, GE Healthcare, Milwaukee, Wisconsin, USA) guidance (Y.W.P and S.L), inducing liver solitary tumours in 14 days with a diameter of 1–2 centimetres in the longest dimension, which would have relatively high vascularity and not be too small for imaging [24].

AG-013736 preparation and study protocol

AG-013736 (Selleckchem.LLC, Texas, USA), a receptor kinase inhibitor of all known VEGF receptors, was prepared in a 0.5 % carboxymethyl cellulose suspension vehicle, and given orally by gavage at a dose of 1, 6 or 12 mg/kg ( twice daily in the fasted state). The dose was selected based on the conversion of effective dose for mice studies (30 mg/kg twice daily) [25, 26] and the maximum tolerated dose (10 mg twice daily) of AG-013736 defined during the phase II trial in patients with advanced hepatocellular carcinoma [27].

Twenty-nine qualified tumour-carrying rabbits were randomly assigned by means of random number lists into the control group (n = 6; group 1) and AG-013736–treated groups at different dosage regimens of 1 mg/kg (n = 8; group 2), 6 mg/kg (n = 9; group 3), and 12 mg/kg (n = 6; group 4). Spectral CT scan were performed in all rabbits before (baseline) and at 2 days, 4 days, 7 days,10 days and 14 days after AG-013736 treatment. During the 14-day follow-up, six rabbits (two in group 2, three in group 3 and one in group 4) in the treated group died during the experimental period, which may be related to the factors such as laparotomy procedure, repeated anaesthesia or imaging studies, and the toxicity of the agent. Finally, six rabbits in the control group (group 1) and 17 in the treated group (six each in groups 2 and 3, five in group 4) were included in this study.

CT scanning

Triple-phase spectral CT scans (i.e. unenhanced and two-phase contrast enhanced CT examinations) on all animals were performed on a Discovery CT750HD system (GE Healthcare, Milwaukee, Wisconsin, USA). Under the same anaesthesia, the rabbits were fixed on a board in supine position with abdominal bandage to further reduce the respiratory movement.

Two-phase contrast-enhanced CT images covering the whole liver were acquired in the spectral imaging mode (Gemstone Spectral Imaging (GSI) protocol 32, GSI-32) with a single tube, rapid dual kVp (80 kVp and 140 kVp) switching technique. Other parameters were as follows: collimation, 64 × 0.625 mm; tube current, 375 mA; rotation speed, 0.6 second; helical pitch, 0.984:1. Animals were injected with 5.0 ml of non-ionic contrast medium (Ioversol Injection, Optiray 320; Jiangsu Hengrui Medicine Co., China) through auricular venous access at a rate of 1 mL/s followed by 10 mL of saline administered at the same rate using a power injector (Envision CT injector, Medrad, Georgia, United States). The arterial and portal venous phase imaging began 15 s and 35 s after the administration of contrast agent.

The spectral CT images were reconstructed by using projection-based material decomposition software and a standard reconstruction kernel with slice thickness and reconstruction interval of 1.25 mm. Water- and iodine-based MD images as basis material pairs were reconstructed from the single spectral CT acquisition for analysis.

Image analysis

Two radiologists (H.G and J.B.G) with 18-32 years’ experience in abdominal imaging performed the quantitative measurements in consensus for analyzing the iodine-based MD images by using the GSI viewer software on a commercially available workstation (Advantage Windows, version 4.4; GE Healthcare, Wisconsin, USA). The change in tumour size was defined at the longest dimension measured on the iodine-based MD images to determine the effect of tumour treatment. Percentage change in tumour size (△TS) [12] on all rabbits between baseline and 14-day follow-up was calculated as following: ΔTS = (LD14days − LDbaseline)/LDbaseline * 100 %, where LD is the longest dimension of the tumour.

At each time point, an iodine concentration (IC) of the tumour was measured on the iodine-based MD image by using an operator-defined region of interest (ROI) (mean value 206 mm2, range 102– 409 mm2) around the entire tumour [13, 28] (Fig. 1). In order to minimize variations between rabbits, the IC of tumour was normalized to the iodine concentration in the aorta (ICaorta) to derive a normalized iodine concentration (nIC): nIC = IC tumour /IC aorta. The difference for the normalized iodine concentration (nICD) between a given time and baseline were calculated as follows: nICD = nICgiven time - nICbaseline. For all measurements, an average of three separate ROIs was obtained to ensure consistency and the copy-and-paste function was applied to make sure the ROIs were at the same place in the two phases.

Fig. 1
figure 1

An axial iodine-based material decomposition image in rabbit shows placement of region of interest around the entire tumour (green outline)

Full size image

Histological analysis

The rabbits were sacrificed immediately after spectral CT scan by the intravenous injection of 10 ml of 10 % kalium chloratum while they were deeply anesthetized. After fixation in 10 % buffered formalin, the tumour specimens excised from the rabbit livers were sliced at approximately 1 mm intervals in the transverse plane to match the CT images. Hematoxylin-eosin (HE) staining at 3 μm section was performed to evaluate the necrotic fraction (NF) of the tumour by using the following formula: NF = Areanecrosis / Areatotal tumor [12]. The blood vessel marker CD31 expressions were detected to analyze the microvessel density (MVD) of the tumours by using immunohistochemical methods for the specific endothelial antigen CD31. To obtain the MVD of the tumour, hot spots (highest vascular density area within the lesions) were identified at low magnification (×100), and the three areas with highest CD31-stained vessels were counted at high magnification (×200; 0.458 mm2/field) to obtain the mean value.

Statistical analysis

All statistical analyses were performed by using SAS (SAS, version 9.1.3; SAS Institute, Cary, NC). To evaluate the reproducibility of CT spectral imaging parameters, tumour nIC values at baseline in the control group (n = 6) were compared with the nIC values measured again at 6 hours later in the arterial phase (AP) and portal phase (PP) to calculate the corresponding coefficients of variation (CVs) within rabbits (good reproducibility, CVs of 10 % or less; moderate, 10 %–25 %; poor, 25 % or greater) [29]. Considering the skewed distribution of data probably because of the small number of samples, the Mann-Whitney test or the Kruskal-Wallis test with post-hoc comparison was used to compare the tumour size change and CT spectral imaging values between the control and treated groups or among the treated groups with different dosage regimens. Serial changes in nICs of the tumours in the treated group (n =17) at different time points were evaluated by using the Friedman test, with post-hoc Wilcoxon signed rank test used for further comparisons in cases of statistical significance. The correlation between nICDs and the tumour size changes at 14 days, and the correlation between nICs at 14 days and corresponding histological features such as NF and MVD were assessed using Spearman rank correlation test due to the skewed or ordinal distribution of data. All statistical analyses were performed with P < 0.05 indicating a statistically significant difference. For the case of the post-hoc Wilcoxon signed rank test (12 paired comparisons at six points), P < 0.0042 (0.05/12) was considered to indicate significance for a Bonferroni correction.

Results

Tumour growth

The Rabbits carrying VX2 tumours were first treated with 1, 6, or 12 mg/kg/bid AG-013736 to define a dose that would not stop but slow tumour growth (Fig. 2).The percentage increase in the tumour size was significantly smaller in the treated group (median, 58.62 %) than that in the control group (median, 100.1 %) (P = 0.024). This size increase of the tumour in the treated groups tended to be smaller with higher dosages (median, 75.73 %, 65.69 %, and 42.85 % in group 2, 3, and 4, respectively), with statistically significant difference between group 4 and the control group (P = 0.004). However, there was no statistically significant difference among the three kinds of dosage regimens (P = 0.247).

Fig. 2
figure 2

The tumour longest diameter showed as mean ± standard deviation at different time points, and five to six tumours were included per group. The tumours in the control and treated groups (three different doses of AG-013736: 1 mg/kg bid in group 2, 6 mg/kg bid in group 3, and 12 mg/kg bid in group 4) all grew gradually as a function of days following initiation of treatment, but grew more slowly in the treated group since D7(P < 0.05)

Full size image

Reproducibility of CT spectral imaging parameters

In the control group, nICs of the tumours demonstrated good reproducibility, with within-subject CVs of 6.10 % and 8.56 % in AP and PP respectively.

Comparison between the control and treated groups

In general, the differences of the normalized iodine concentrations (nICDs) of the tumours relative to baseline for each group at each time point in AP and PP were lower in the treated group as compared with the control group (Fig. 3). nICDs of the tumours were significantly lower in the treated group than in the control group at 2 days (-0.031 ± 0.055 vs. 0.067 ± 0.086, P = 0.015) in AP; and at 2 days (-0.166 ± 0.239 vs. 0.041 ± 0.148, P = 0.032) and 4 days (-0.144 ± 0.240 vs. 0.054 ± 0.122, P = 0.020) in PP after treatment.

Fig. 3
figure 3

Box-and-whisker plots (box: 25 %, 75 %; centreline: medium; whisker: min, max) show normalized iodine concentration differences (nICD) values of the tumours between the given time and baseline during the AP (a) and PP (b). P < 0.05 (*) indicates a statistically significant difference between the control and treated groups by using the Mann-Whitney test

Full size image

All parameters between the subgroups 2 and 3 (treated with 1 mg/kg and 6 mg/kg dose of AG-013736, respectively) and the control group (Table 1 and Fig. 4) showed no statistically significant difference at any follow-up time points. Compared with the control group, group 4 showed significantly lower nICDs at 2 days (P = 0.015 and 0.004 for AP and PP respectively) , almost similar results at 4 days and 7 days, but almost lower values at 10 days and similar (AP) or lower values (PP) at 14 days after treatment. For all nICDs of the tumours at any follow-up time points, there were no statistically significant differences among the three subgroups in the treated group.

Table 1 Comparisons of normalized iodine concentration difference (nICD) between the control group and group 4 (AG-013736 of 12 mg/kg bid) at different time points
Full size table
Fig. 4
figure 4

Normalized iodine concentration differences (nICD) values of the tumours in the control and treated groups during AP (a) and PP (b) as a function of days following initiation of treatment. Comparisons between control group and treated groups with three different doses of AG-013736 (1 mg/kg bid in group 2, 6 mg/kg bid in group 3, and 12 mg/kg bid in group 4) at each time point were performed by using the Mann-Whitney test (* = P < 0.05). During AP, tumours nICDs in group 4 at day 2 and day 10 were significantly lower than in the control group. During PP, tumour nICDs in group 4 at each time point except at day 7 decreased significantly compared to the control group

Full size image

Serial measurements in the treated group

Results for serial measurements of nICs of the tumours in the treated group are summarized in Table 2. Compared with the baseline, nICs in the treated group at 2-day follow-up decreased significantly (P < 0.0042), but recovered at 4-day and 7-day follow-ups with no significant differences, whereas nICs decreased slightly at 10 days and significantly at 14 days after treatment (P <0.001 and 0.001 for AP and PP respectively) (Fig. 5). nICs significantly decreased at 14 days as compared with 4 days and 7 days (All P values < 0.0042).

Table 2 Comparisons of normalized iodine concentration (nIC) for the tumour between different time points in the treated Group (n = 17)
Full size table
Fig. 5
figure 5

Iodine–based material decomposition images at different time points in group 4 (AG-013736 treated with 12 mg/kg bid) during AP and PP. 2 days after treatment, the iodine concentration of tumour is sharply decreased, but obvious hyperintense peritumoral rims (red arrow) in AP and slightly hyperintense centres (white arrow) relative to that of day 2 in PP were noticed at 4 days and 7 days .The iodine concentration gradual decreases from 10 days to 14 days

Full size image

Prediction of tumour response

There was statistically significant positive correlation with tumour response for nICDs of the tumours at 2-day follow-up in both AP (r = 0.69, P =0.002) and PP (r = 0.64, P =0.006) after treatment, which meant that a greater increase in nICD at 2 days would predict a smaller increase in tumour size at 14 days.

Histological analysis

The correlation between nICs of the tumours at 14-day follow-up and corresponding histological features, including NF and MVD were showed in Table 3. NF was significantly correlated with nICs of tumours negatively in both AP and PP. MVD showed positive correlation with nICs of the tumours in both AP and PP.

Table 3 Correlation between histological features and normalized iodine concentration (nIC) of the tumour at 14–day follow up
Full size table

Discussion

AG-013736 has been shown to inhibit angiogenesis, vascular permeability, blood flow and antimetastatic activity in previous nonclinical and clinical studies [6, 7, 30]. In this study, AG-013736 was administered in rabbits with VX2 liver tumours, which are most commonly used large animal model in imitating primary or secondary liver cancers in humans [13, 31, 32]. A number of studies about the effect of AG-013736 on breast cancer, prostate cancer, neuroblastoma and cholangiocarcinoma have been tested previously in rat and mouse models [7, 25, 33]. This is the first time the anti-angiogenic agent of AG-013736 has been tested for antitumor efficacy in this model.

In this study, the change of tumour size was evaluated by using the longest dimension of the tumour instead of tumour volume, based on the methodologies of previous literatures [12, 34–36] and the fact that our all tumours showed interval growth indicating progressive disease. The results on tumour size demonstrated that tumour growth was significantly suppressed by AG-013736 compared with the control, tended to be smaller with higher dosages, similar to the results of Liu G et al. that the effect of AG-013736 is dose dependent [6]. The ability of quantitative analysis for CT spectral imaging parameters in our study has been proved to be feasible in depicting the serial changes after administration of AG-013736 in a rabbit VX2 liver tumour model, with good reproducibility.

CT with energy information, in particular spectral CT, extends the capabilities of conventional CT. It provides the material-decomposition basis pair of water and iodine in medical diagnostic imaging for quantification [20, 37] in elemental and molecular composition and contrast materials based on their attenuation properties [38]. The quantitative IC in the tumour derived from the iodine-based MD image is an indication of its blood flow [19, 39], depending on the amount of iodinated contrast agent administration. The quantitative measurements obtained in this study showed that nICDs of the treated group were lower as compared with the control group, although not statistically significant in all time points and all parameters. Interestingly, in the treated group, nICs of the tumours decreased at 2 days, recovered at 4 days and 7 days, and further decreased from 10 days to 14 days after the administration of AG-013736. This different responses as function of time after treatment may be explained by the intrinsic mechanism of VEGF inhibitor. It can selectively prune the chaotic and inefficient vasculature commonly found in tumours and improved vasculature by reducing vessel counts and increased coverage of pericytes [40], resulting in the reduction of blood flow to the tumours, which would be followed by ischemic changes such as cellular edema and intratumoral necrosis [12]. However, the peripheral viable tumour usually survives and regrows, appearing to be weaker than the centre of the tumour [13] in the treatment response, reflected by the serial changes of iodine concentrations in the peripheral and central regions of the tumour in our study, similar to other studies [12, 13] about the therapeutic efficacy of VTAs in rabbit VX2 liver tumours. But further repeated drug administration at time intervals have been demonstrated effective in improving therapeutic effects by our and other studies [41].

This study showed that relative changes of iodine concentration in tumours had the potential to predict favourable tumour response after AG-013736 treatment. nICDs of the tumours at 2 days in both AP and PP after treatment showed a statistically significant correlation with the final changes in tumour sizes (P = 0.002 and P =0.006, respectively). Consistent with this result, Liu G et al. [6] utilized magnetic resonance imaging to study pharmacodynamic response in AG-013736-treated patients with advanced solid tumours and found day 2 vascular response seemed to be a useful indicator of drug pharmacology. It should be noted that tumour size of VX2 tumours in the treated group grow slowly but progressively, indicating that AG-013736 s are not expected to reduce tumour size, consistent with other studies [6, 25, 42]. The ability of nICDs in reflecting the angiogenesis of the tumour can be explained by the positively correlation between nICs of the tumours and MVD. An increase in nICDs at 4 days after treatment can also be explained by the increase in intratumoral necrosis, considering the significant correlation between nICs and NF at 14 days.

For the assessment of tumour response, there have been few studies evaluating the usefulness of CT spectral imaging. Although Tang L et al. [18] and Hu S et al. [19] have demonstrated the value of spectral CT in evaluating the therapeutic effect of neoadjuvant chemotherapy on gastric carcinomas and 125I interstitial brachytherapy on pancreatic carcinoma, the potential role of CT spectral imaging parameters as early predictors of tumour response after VTA treatments has not been previously well established in the literature. Dual-source dual-energy CT (DECT) has also been demonstrated to be effective in early prediction of tumour response in human study of Meyer M et al. [43], however, this study has used the first available DECT scan instead of prior treatment similar to our study as baseline evaluation, which may lead to a long term evaluation bias. In contrast to the relatively short period of 4 days in the animal study of Knobloch G et al. [44] to assess the early regorafenib treatment effects, the evaluation of AG-013736 treatment in our study was performed during the relatively long period of 14 day so that therapy-related changes in the tumour microvasculature and tumour size could be performed. The study of Lee SH et al. [45] has assessed the value of iodine maps in the evaluation of therapeutic response to radiofrequency ablation (RFA) for hepatic tumours but without evaluating it as early predictors of tumour response.

The present study did have limitations. First, we didn’t perform a power analysis in the study, as no prior data existed and assumed that five or more rabbits per group would be sufficient for this animal study. Therefore larger prospective investigations are needed to confirm our findings. Second, the correlation of the CT spectral imaging parameters with histological features at each time point were not evaluated except at 14 days after the treatment, which would not explain the early changes of histological features. Third, the accuracy of quantitative evaluation on VTAs with CT spectral imaging was not validated by other imaging modalities such as perfusion CT, Doppler ultrasound and dynamic contrast-enhanced MR imaging. These shortcomings will be addressed in our future research.

In conclusion, the therapeutic effects of AG-013736 in inhibiting VX2 liver tumours in rabbits can be noninvasively and quantitatively monitored with CT spectral imaging parameters. The day 2 vascular response measured using nICDs from material decomposition images in spectral CT can be used as suitable markers for the early prediction of tumour response.