Eugene Kim1, Jana Cebulla1, Astrid Jullumstrø Feuerherm2, Berit Johansen2, Olav Engebråten3, Gunhild Mari Mælandsmo3, Tone Frost Bathen1, and Siver Andreas Moestue1
1MR Cancer Group, Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway, 2Avexxin AS, Department of Biology, Norwegian University of Science and Technology, Trondheim, Norway, 3Department of Tumor Biology, Institute for Cancer Research, Oslo University Hospital, Oslo, Norway
Synopsis
This
study investigated the relationship between tumor vascular function (DCE-MRI)
and structure (ex vivo micro-CT). Control tumors did not exhibit any
significant correlations between micro-CT and DCE-MRI parameters. Tumors
treated with bevacizumab or a cPLA2 inhibitor (AVX235), both anti-angiogenic
drugs, displayed reduced perfusion and vascularization. But interestingly,
there was a significant positive correlation between vascular surface area and
Ktrans in AVX235-treated tumors, whereas the corresponding
correlation was negative in bevacizumab-treated tumors. This suggests that different
therapies can differentially modulate the vascular structure-function relationship,
which highlights the challenge in interpreting DCE-MRI measurements and
adopting them as clinical biomarkers of therapeutic response. Purpose
To compare the effects of two anti-angiogenic drugs -- bevacizumab
(Avastin®, Genentech) and a cytosolic phospholipase A2 (cPLA2) inhibitor
(AVX235, Avexxin AS) [1] -- on vascular function using DCE-MRI, vascular
structure using micro-CT, and the relationship between vascular structure and
function in a patient-derived xenograft model of breast cancer.
Methods
Basal-like
breast cancer xenografts were orthotopically transplanted into 26 female nude
mice. Baseline MRI was performed on all mice (day 0). Afterward, one group
(n=9) received bevacizumab (5 mg/kg) on days 0 and 3; another group (n=9)
received daily doses of AVX235 (45 mg/kg) for one week; and a third group (n=8)
received daily, volume-matched injections of DMSO vehicle. All mice were scanned
again on day 4, after which the bevacizumab-treated mice were sacrificed;
AVX235-treated and control mice were scanned again on day 7 before being
sacrificed.
In
each imaging session, DCE-MRI was performed on a 7T Bruker Biospec. Baseline T1
maps and a dynamic series of 200 T1w images were acquired. The enhancing voxel
fraction (EVF, signal enhancement>50% after one minute) and maps of the initial
area under the signal enhancement curve (AUC1min) and the extended
Tofts model parameters were computed from the DCE-MRI data. Tumor-wise
parameter medians were computed from the maps, which were masked to exclude non-enhancing
voxels and non-tumor tissue. Further details on the MRI
protocol and analysis can be found in [2].
Immediately after the final MRI session, mice were
sacrificed by perfusion fixation and perfused with Microfil® (Flow Tech, Inc.),
a vascular casting agent. Then, tumors were excised and scanned on a SkyScan 1176
micro-CT system (Bruker microCT) at 9-μm isotropic resolution. One control and
one bevacizumab-treated tumor were excluded due to incomplete Microfil®
perfusion. Tumor blood vessels were segmented from the micro-CT images using a
Hessian-based filtering method [3]. From the segmented vessels and manually drawn
tumor masks, fractional blood volume (FBV), vessel surface area normalized to
tumor volume (VSA), median and 90th percentile distance to the
nearest vessel (DNV and DNV90), and median and 90th percentile vessel
caliber (VC and VC90) were computed for each tumor.
Two-tailed Wilcoxon signed-rank tests were performed
to test for significant intra-group changes in DCE-MRI parameters from baseline
to final time point. Kruskal-Wallis one-way ANOVA and Tukey’s HSD tests were
used for inter-group comparisons of DCE-MRI and micro-CT parameters. Pearson
correlation coefficients (r) were
computed to measure the correlation between DCE-MRI and micro-CT parameters.
For all tests, α=0.05.
Results
AUC1min
decreased in all bevacizumab-treated tumors, 7/9 AVX235-treated tumors, and 3/8
control tumors. The change in AUC1min was significantly greater in
bevacizumab-treated tumors compared to controls (Fig. 1b). Post-treatment Ktrans was significantly lower
in bevacizumab-treated tumors compared to control and AVX235-treated tumors (Fig. 1c).
VSA,
FBV, and VC were significantly smaller, and DNV significantly larger in
bevacizumab-treated tumors compared to controls. VC90 was significantly smaller
and DNV90 significantly larger in AVX235-treated tumors compared to controls (Fig. 2).
Pooled
r values are presented in Table 1 and show that DCE-MRI
parameters correlate with VSA, FBV, and DNV90, but not DNV or VC(90).
Interestingly, none of the correlations between DCE-MRI and micro-CT parameters
were significant for the control group, and the correlation coefficients of the
treatment groups had opposite signs (Fig.
3).
Discussion
The
decrease in median AUC1min and/or Ktrans observed in 88%
of treated tumors is a common effect of angiogenesis inhibition [4]. The
decreased vessel density and caliber are also expected after anti-angiogenic
treatment. However, the correlation analysis revealed that the two drugs had
very different effects on the tumor vasculature.
The pooled
correlations indicate that DCE-MRI measurements (i.e. contrast agent delivery)
depend on vascular structure (particularly surface area) as well as flow and
permeability. But the intra-group correlations suggest that this relationship
is not constant. The lack of strong correlations between micro-CT and DCE-MRI
in the control tumors suggests an uncoupling between vascular structure and
function. While AVX235-treated tumors were not significantly different from
controls based on individual DCE-MRI measurements, the significantly positive
correlations between VSA and AUC1min and Ktrans suggest more normally functioning vasculature after cPLA2 inhibition. In contrast, the
less vascularized bevacizumab-treated tumors were better perfused (Fig. 4), which could reflect improved
flow as a result of vascular remodeling and pruning.
Conclusion
These
results demonstrate that DCE-MRI measurements can highly depend on vascular
surface area, which is largely overlooked compared to flow and permeability
when discussing DCE-MRI pharmacokinetics; and that the relative contributions
of these three factors change with treatment and may differ from tumor to
tumor. This highlights the challenge of developing DCE-MRI parameters into
clinically validated biomarkers.
Acknowledgements
This work was funded by the
liaison committee between the Central Norway Regional Health Authority and the
Norwegian University of Science and Technology (NTNU) (grant no.
652510-46056806). The authors would like to thank Alexandr Kristian at the
Department of Oncology and Department of Tumor Biology, Oslo University
Hospital, for performing the xenograft transplants. References
1. Kokotos G et al. J Med Chem 2014;57(18):7523-35.
2. Cebulla J et al. Br J Cancer 2015;112(3):504-13.
3. Kim E et al. Magn Reson Med 2013;70(4):1106-16.
4. O’Connor et al. Nat Rev Clin Oncol 2012;9(3):167-77.