Monitoring therapeutic efficacy of sunitinib using [¹⁸F]FDG and [¹⁸F]FMISO PET in an immunocompetent model of luminal B (HER2-positive)-type mammary carcinoma

FVB/N-Tg (MMTV-PyMT)634Mul/J (PyMT) 12-weeks-old mice were used as tumour donor. Aseptically collected mammary tumours from PyMT mice were minced and immersed in cold Dulbecco's Modified Eagle's Medium (Sigma, USA). Mechanical cell dissociation was performed using Medicon disposable chambers (BD bioscience, USA). The cell suspension was then progressively filtered using Filcon filters with pore sizes of 500 μm, 200 μm and 70 μm (BD bioscience). Finally, cells were aliquoted in freezing medium (Life Technologies, USA) and stored in liquid nitrogen.

After freezing medium removal and enumeration, the tumour cells were directly inoculated, without any in vitro culture step, in the mammary fat pad of the posterior nipple in FVB mice. The tumour volumes were calculated using calliper measurements and the approximated formula for a prolate ellipsoid, given by:

To evaluate drug toxicity, body animal weights were also monitored.

Two sets of mice were used in this study. For the main set A, 7 animals were implanted with 3 million viable cells. PyMT tumours were allowed to grow for 21 days. The mice were randomized into treated (n = 4) and control (n = 3) groups. The treated one received per os a daily dose of sunitinib at 40 mg/kg in 20 mM dimethyl sulfoxyde (DMSO, Sigma). The control group received only the DMSO solution. Drug administration was performed during 9 consecutive days.

To further explore the neoadjuvant therapy effects on the metastatic incidence, we extended primary tumour growth and treatment times before mammary tumour surgical resection. Thus, a secondary set B of treated (n = 4) and control (n = 4) mice was obtained implanting 400 000 viable cells. Treatment began at day 25 post implantation. It continued for 14 days until resection at day 39. For both sets A and B, the primary tumours were surgically removed after treatment and the mice were kept alive for 60 supplementary days before euthanasia to analyse the lungs for metastasis content.

[¹⁸F]FDG and [¹⁸F]FMISO PET scans were performed on mice from set A at days 0 and −1 respectively prior to treatment and at days 5 and 6 of treatment. 15 min long PET acquisitions were performed 60 min after [¹⁸F]FDG injection and 90 min after [¹⁸F]FMISO injection. PET data were corrected for attenuation, scatter and radioactive decay and reconstructed using a two dimensional ordered-subset expectation maximization (2D-OSEM) algorithm after Fourier rebinning, with a voxel size of 0.5 × 0.5 × 0.8 mm³ (sofware ASIPro VM™, CTI Concorde Microsystems). Radioactivity uptake in regions of interest (ROIs) was measured using BrainVISA 4.0 and Anatomist 4.0.2 (CEA/Neurospin/SHFJ, France) and expressed in Standardized Uptake Value (SUV) calculated using:

Primary tumours from set A of animals were fixed in zinc solution (BD bioscience) and included in paraffin. Series of tissue sections were sequentially cut. For blood vessels, macrophages and cellular hypoxia sensor labelling, the slides were immersed in toluene and progressively rehydrated. Endogenous peroxidases and biotin were blocked with 3 % hydrogen peroxide solution (Sigma) and biotin blocking kit (Life technologies) respectively. Rat anti CD31 (Pharmingen, USA), rat anti F4/80 (Caltag, UK) and rabbit anti Hypoxia Inducible Factor 1 alpha (HIF1 alpha, LSBio, USA) were used as primary antibodies for each labelling respectively. Biotin-goat anti rat IgG (Life technologies) was used as secondary antibody for vascular and macrophage staining. The tyramide signal amplification (TSA) system (Perkin Elmer, USA) was then used following manufacturer’s instructions. For HIF1 alpha labelling, HRP-goat anti rabbit IgG (Life technologies) was incubated as secondary antibody. For cellular proliferation and Glucose transporter 1 (GLUT1) expression labelling, paraffin removal was performed using heated PT module buffer pH8 (Fischer Scientific, USA). As above, after the blocking steps, the slides were incubated with goat anti Ki-67 (Santa Cruz, USA) or rabbit anti GLUT1 (Neomarker, USA) for each labelling respectively. Secondary detection reagents were biotin-rabbit anti goat IgG (Life technologies) followed by TSA system for Ki-67 or HRP-goat anti rabbit IgG (Life technologies) for GLUT1. After 3-3'–diamino-benzidine (DAB, Sigma) revelation, counterstaining was performed with hematoxylin (Sigma) and slides were mounted with Eukitt (Sigma). For late apoptosis staining, terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL, Promega, USA) was used according to the manufacturer’s protocol. Slides were then mounted with ProLong Gold Antifade Reagent containing 4',6'-diamidino-2-phenylindole (DAPI, Life technologies).

The set of tissue sections uniformly sampling the whole volume of each tumour was entirely scanned at high resolution (0.37 μm per pixel) using an AxiObserver Z1 (Zeiss, Germany). The resulting brightfield image series were analysed using the CellProfiler software [20]. After a colour deconvolution step, the segmentation of each structure of interest was based on a constant labelling-dependent threshold. A filtering step was added for size-based vessel clustering. Logic diagrams of the processing pipelines are available as supplementary data (see Additional files 1 and 2). The DAB-labelled surface areas and whole hematoxylin areas were measured by the software. The consistency of the automatic segmentation was controlled visually on the original images supplemented with the outlines of identified objects. Whole tissue sections fluorescently labelled with the TUNEL method were acquired using two excitation/emission filter sets: 365/445 nm for DAPI and 470/525 nm for TUNEL staining. TIF-format images were processed using the ImageJ software [21], yielding the total area corresponding to fluorescent pixels above a given constant threshold. The measured areas were multiplied by the distance between each tissue slide to get volume estimates.

The whole-lung tissue ribonucleic acids (RNA) were extracted using the total RNA isolation kit (Macherey-Nagel, Germany) following manufacturer’s instructions. RNA was reverse transcribed using SuperScript II (Life technologies) with random primer hexamers. On a LightCycler 1.5 (Roche, Switzerland), a subsequence of the PyMT cDNA was amplified in Master SYBR Green I mix (Roche) using the previously described primers [22]. The housekeeping myelin protein zero (P0, MPZ) gene was used as an internal control. A relative quantification analysis was performed applying the delta-delta Ct method.

For statistical analysis, unpaired Student t-tests were performed using GraphPad Prism software. A p-value of 0.05 or less was interpreted as statistically significant. In all graphs, values are reported as mean ± one standard deviation (SD).

In set A of mice, the mean tumour volume measured by calliper was 209 ± 38 mm³ (n = 7) just before treatment (day 21). During the treatment phase until day 30, the tumours of the control group continued to grow up to 418 ± 62 mm³ (n = 3), while the size of the treated tumours decreased down to 109 ± 24 mm³ (n = 4) (Fig. 1a, p < 0.001). In set B of mice, the mean tumour volume measured by calliper was 115 ± 9 mm³ when treatment started (day 25, n = 8). At resection (day 39), tumour volumes were 282 ± 43 mm³ in the control group (n = 4) and 57 ± 11 mm³ in the treated group (n = 4) (Fig. 1b, p < 0.0001). The mouse weights corrected for their tumour weight (Fig. 1c-d) were not significantly different between the treated and control arms.

PET monitoring was only performed for set A of mice. In the reconstructed images, the signal appeared more prominent in the tumour, bladder and heart compared to the rest of the body (Fig. 2a). At randomization, [¹⁸F]FDG uptakes expressed in SUV were similar in both groups. In the control group, the tumour [¹⁸F]FDG uptake was 1.4 greater after 5 days than at randomization, from 1.2 ± 0.1 to 1.6 ± 0.3 (n = 3) in SUV, although the difference was not statistically significant (NS). In the treated group, the [¹⁸F]FDG uptake remained stable during sunitinib administration, from 1.1 ± 0.2 to 1.1 ± 0.2 (n = 4) in SUV (NS). As a result, after 5 days of treatment, the [¹⁸F]FDG uptake was significantly lower, by a factor of 1.5 (p < 0.05), in the treated versus the control tumours (Fig. 2b).

The [¹⁸F]FMISO PET images exhibited an enhanced contrast in the tumour and intestine regions (Fig. 2c). At randomization, [¹⁸F]FMISO tumour uptake was not significantly different between the group to be treated (1.2 ± 0.2 in SUV, n = 3) and the control group (0.8 ± 0.1 in SUV, n = 3), although lower in the control group. After 6 days of sunitinib or vehicle only administration, the tracer uptake remained stable in the control group (0.9 ± 0.2 in SUV, n = 3) and decreased in the treated one (0.9 ± 0.4 in SUV, n = 3), thus reducing the initial differences between the two groups (Fig. 2d).

Digital microscopy analysis was exclusively performed on set A of mice. In the control and treated groups, the mean hematoxylin volumes estimated using digital microscopy were 126 ± 6 (n = 3) and 45 ± 20 mm³ (n = 4) respectively (p < 0.01). The nine day sunitinib treatment induced a 4.9 fold regression of the vascular volume reported to the hematoxylin volume, with values of 3.6 ± 1.8 % (n = 3) and 0.7 ± 0.05 % (n = 4) in the control and treated groups respectively (p < 0.05) (Fig. 3a). The sunitinib treatment induced a reduction of the large vessel proportion (from 31.3 ± 15.6 % in the control group to 8.3 ± 4.0 % in the treated group, p < 0.05), an increase of the small size vessels (from 33.4 ± 16.5 % in the control group to 61.0 ± 11.1 % in the treated group, p < 0.05) and no evolution for medium size vessels (Fig. 4). Interestingly, the vascular volume decrease did not induce a global increase in HIF1 alpha protein expression (Figs. 3b and 5a-b). On the contrary, sunitinib therapy led to a twofold reduction of HIF1 alpha labelling, from 11.5 ± 1.0 % (n = 3) in control tumours to 5.7 ± 3.5 % in treated ones (n = 4, p < 0.05). TUNEL labelling revealed a high induction of apoptosis by a factor of 17.7. Indeed, in nutrient supplied expanding tumours, very low programmed cell death was observed, representing 4.1 ± 0.3 % in volume (n = 3), while this value was 72.6 ± 10.7 % in treated tumours (n = 4, p < 0.001) (Figs. 3c and 5c-d). A mean pool of 1339 ± 248 cells per mm³ of tumour (n = 3) were over-expressing Ki-67 in non-treated tumours and the therapeutic agent reduced this population to 730 ± 334 proliferating cells per mm³ (n = 4, p < 0.05). This represents a 1.8 fold reduction of the tumour proliferation process (Figs. 3d and 5e-f). Tumours grown in the control conditions presented a mean density of 933 ± 212 macrophage cells (F4/80 positive) per mm³ of viable tumour tissue (n = 3). In the sunitinib treated mice, this value was at 546 ± 169 macrophages per mm³ (n = 4) (Figs. 3e and 5g-h) (p < 0.05 compared to the control mice). Finally, GLUT1 whole tumour expression was enhanced by a factor of 2.57 in the sunitinib treated group when compared to control (Figs. 3f and 5i-j). Indeed, transporter labelling represented 15.5 ± 2.6 % of control hematoxylin volume (n = 3), and reached 40.0 ± 12.3 % after the 9 days-long treatment (n = 4, p < 0.05).

In set A, with a primary tumour growth duration of 30 days, comprising a 9-day sunitinib or DMSO administration, no lung metastasis was detected 60 days after tumour resection in the treated (n = 4) and control (n = 3) groups. In set B with a 39 days tumour growth duration, including 14 days of sunitinib or DMSO treatment, the incidence of lung metastasis was of 50 % in the treated (n = 4) and control (n = 4) groups (Fig. 6).