In this study we aimed to evaluate the use of Quantitative Susceptibility Mapping (QSM) in a mouse model of colorectal carcinoma metastasis to liver. Tumour blood flow and oxygenation are critical factors for successful cancer therapy[1], and there is currently no clinically accepted, non-invasive method of monitoring this[2]. We hypothesised that QSM could provide a novel mechanism to characterise the pathophysiology of the tumours based on differences in blood oxygenation. Results under hyperoxic and normoxic conditions were compared to assess the differential response of the tumours and normal liver. Lastly, a vascular disrupting agent (VDA) was administered to evaluate the effect of this vascular targeting therapy on the QSM measurement.MF1 nu/nu mice (n=3) were inoculated with 1x10^6 SW1222 colorectal liver metastases cells via intrasplenic injection. MRI was performed at approximately 4 weeks post-surgery. Mice were anaesthetised using 4% isoflurane in 100% O2. During scanning, respiratory rate was monitored and maintained at ~60-80 breaths per minute by varying isoflurane concentration between 1.5 and 2%. Hyperoxia was induced with 100% O2, followed by medical air to induce normoxic conditions. Fully flow-compensated, respiratory-gated, single-echo 2D T2*-weighted GRE data were acquired on an Agilent 9.4T scanner using a 39mm birdcage coil (Rapid biomedical, Rimpar, Germany). Matrix=136x136, 80 slices, 200µm isotropic resolution, TR/TE=1000/4ms, averages=4. A sample of purified water accompanied each mouse in the scanner to provide a reference for susceptibility measurements. Acquisition time was approximately 20 minutes for each gas. VDA (OXi4503) was administered i.v. (40 mg/kg) directly after the pre-drug acquisitions; post-drug acquisitions took place 48 hours later. A binary mask was manually drawn around the entire liver in each magnitude image using ITK-SNAP[3]. The corresponding phase data was unwrapped and the background field suppressed using a Laplacian based SHARP algorithm (TSVD threshold = 0.08, mask erode = 1)[4]. Inversion was carried out using a TKD algorithm (threshold = 5)[5]. Regions of interest (ROIs) were manually drawn on the magnitude images and then transferred to the susceptibility maps. ROIs were drawn in normal liver tissue and all visible tumours in each mouse on the magnitude images. A large individual tumour in one mouse was selected to examine the intra-tumoural response to the gas challenge and VDA. To assess regional differences within the tumour, the mean susceptibility in the tumour in each transverse slice was plotted against slice position.

Fig. 1 shows a susceptibility map of the liver vasculature overlaid with a susceptibility map of the tumours. The blue and green areas within the tumours represent differences in susceptibility that may be indicative of regional differences in blood oxygenation. Both tumour and normal liver were more paramagnetic under hyperoxia than normoxia (fig. 2). Susceptibility values measured in the liver tissue were compared to those within the tumours using an unpaired Student’s t-test, and a significant difference was measured between the liver tissue and tumours during administration of both gasses (p=0.0083 normoxic, p=0.014 hyperoxic). Within the individual tumour, the difference in susceptibility between norm- and hyperoxic states was most prominent in the tumour periphery, where, during normoxia, the tissue was relatively diamagnetic, compared to the more paramagnetic centre (fig 3). The difference in susceptibility between the tumour centre and periphery was attributed to differences in blood oxygenation between the regions. The periphery was less diamagnetic under hyperoxic conditions (fig 3), suggesting a reduction in the amount of oxyhaemoglobin present. In response to VDA therapy, the region at the centre of the tumour was more diffuse post-VDA (fig. 4). Furthermore, the prominent changes in susceptibility caused by the gas challenge, prior to VDA therapy, no longer occurred post-VDA, suggesting a reduction in vascular function by the drug (fig 3).

This study demonstrates the feasibility of using QSM to interrogate a mouse model of colorectal liver metastases. Differences in susceptibility were detected between liver and tumour tissue. Regional susceptibility variations were measured within a tumour and differences mediated by the gas challenge were also detected. Heterogeneous perfusion is a characteristic of tumours [6], and the differences in susceptibility between the centre and periphery of the tumour may be indicative of this. OXi4503 is a vascular disrupting agent that has been shown to be effective against the vasculature of several tumour types[7]. The modulation of susceptibility mediated by the gas challenge was not observed in the post-VDA images. This suggests that the gas challenge elicited a vasoactive response, and that QSM could be used to detect changes in the functional vascular volume within an individual tumour, that may inform on the efficacy of vascular disrupting agents.