The mouse model of spontaneous developing glioblastoma was sourced from a colony at the Queensland Brain Institute, University of Queensland. The model was prepared by crossing Gt(ROSA)26So^{rtm14(CAG-tdTomato)Hze} 20023653 with Pten^tm2MAK; Rb1^tm2Brn; Trp53^tm1Brn; Tg(GFAP-cre/Esr1*,- lacZ)BSbk (alleles) and backcrossed six generations to latter mice to generate the high grade glioma mouse model [2, 10]. All experiments were approved by the University of Queensland Animal Ethics Committee and followed the Australian Code of Practice for Use of Animals for Scientific Purposes (AIBN/142/19/UQ).

Anaesthetised mice were cannulated in the tail vein using 55 cm of tygon tubing (ID 0.01 inch, OD 0.03 inch) with just the metal needle from a 30G needle on one end and an inserted 30G needle on the other end. Imaging was performed with a combined MR-PET system, comprising a 300 mm bore 7T ClinScan MR scanner (Bruker, Germany), running Siemens VB17, with a removable PET insert (Bruker, Germany) containing 3 rings of 16 detector blocks with 15 × 15 LSO crystals (1.6 × 1.6 × 10 mm) per block, located at the centre of the magnet bore. The scanner was operated under Siemens Inveon Acquisition Workplace (IAW, version 1.5.0.28). The animal was placed on an animal bed warmed with recirculating water and a 23 mm ID MRI RF coil inside the PET ring was used to acquire MR images simultaneously with the PET acquisition. A single syringe was prepared and connected to the cannula containing approximately 10 MBq of [¹⁸F]DPA-714 [9] tracer solution, 50 µL Gadovist (Bayer, Germany) and saline/10% ethanol to give a total volume of 200 µL.

Following MRI localiser images, a 60 min PET acquisition was started simultaneously with dynamic spoiled gradient echo (GRE) image acquisition with the following parameters: pulse angle = 45°, 6 × 1 mm slices, FOV = 20 × 20 mm, in-plane resolution = 313 × 313 μm, TR = 75 ms, TE = 3 and 6 ms, acquisition time = 3.6 s, 200 measurements. After a 2 min baseline period the combined PET tracer and Gadovist solution was injected by hand over approximately 5 s, with no flush.

Following the dynamic GRE images, high resolution transverse and sagittal T₂ turbo spin echo (TSE) anatomical images were acquired with the following parameters: 19 × 0.6 mm slices, FOV = 25 × 18.74 mm, in-plane resolution = 98 × 98 μm, TR = 2,200 ms, TE = 35 ms, averages = 2, acquisition time = 4:11 and then coronal with the following parameters: 23 × 0.6 mm slices, FOV = 20 × 20 mm, in-plane resolution = 104 × 104 μm, TR = 2,650 ms, TE = 35 ms, averages = 2, acquisition time = 5:02. Another block of 20 GRE dynamic measurements was then acquired.

Next, anatomical transverse and sagittal T₁ weighted TSE images with the following parameters: 19 × 0.6 mm slices, FOV = 19 × 18.74 mm, in-plane resolution = 78 × 78 μm, TR = 600 ms, TE = 13 ms, turbo factor = 7, averages = 3, acquisition time = 5:21 and coronal images with parameters: 25 × 0.6 mm slices, FOV = 25 × 20.75 mm in-plane resolution = 78 × 78 μm, TR = 600 ms, TE = 13 ms, turbo factor = 7, averages = 3, acquisition time = 8:07 were acquired with the same geometric positioning as the T₂ TSE images.

Following acquisition of fl3d_swi and fl3d_dixon images (not reported in this paper), a final set of 20 GRE dynamic measurements was then acquired. Therefore, dynamic scans extended for 60 min post injection.

At the end of the imaging, the cannula radioactivity was measured and decay corrected to allow calculation of the injected dose. PET data was reconstructed using a dedicated PET reconstruction software developed by the University of Tübingen. PET images with a matrix of 89 × 128 × 128 and field-of-view = 68 × 90 × 90 mm were reconstructed using the ordered-subset expectation maximization (OSEM2D) algorithm with 3 × 30 s, 15 × 10 s, 11 × 60 s and 9 × 300 s time frames for 60 min. MRI and PET datasets were aligned using Siemens Inveon Research Workplace (IRW, version 4.2.0.8) software employing a standard transformation matrix.

HOROS (www.horosproject.org) was used for MRI region-of-interest (ROI) drawing and output. On the central slice of the tumour, one ROI was positioned within tumour tissue taking care to avoid the susceptibility effect at the edge of the brain. The high resolution T₁ and T₂ image sets were used for placement of the tumour ROIs. The tumours with low enhancement on the T₁ images (Figures 1C–E) were more easily visualised on the T₂ TSE images. The ROI placements were then checked on the TE = 6 ms images, both before and after Gd injection, to ensure the ROI regions of the images were not distorted due to susceptibility effects. A second ROI was positioned in the thalamus to represent normal brain tissue (Figure 1). The ROIs were then propagated to all time point images, exported as csv files and imported into excel for calculations.

Calculation of the gadolinium concentration derived independently from T₂* and T₁ time changes were determined from equations modified as previously described [8], using TE1 = 3 ms and TE2 = 6 ms, TR = 75 ms, pulse angle = 45°, r2* = 6.35 s⁻¹ mM⁻¹ and r1 = 6.1 s⁻¹ mM⁻¹ (Gadovist solutions at 7.0 T), A = mean ROI intensity (TE = 3 ms) and B = mean ROI intensity (TE = 6 ms), M_osinα = 2,700.

The T₂* was calculated using: T 2 ∗ =   T E 1 L n ( A B )

The T₂* derived concentration was then calculated using: T 2 ∗       c o n c e n t r a t i o n = ( 1 T 2 ∗   −   1 T 2 b a s e l i n e ∗   )   ÷ r 2 ∗  

The baseline values were determined by averaging data points 3–12 during the baseline period before the Gd injection.

The T₁ was calculated for TE = 3 and 6 ms image pairs using the following equations: D   = 1 −   A × ( A B ) M o     ×  sin ∝   T 1   = 1 ÷ ( T R   ÷   ( D D × c o s ∝ ) )

The T₁ derived concentration was then calculated using: T 1   c o n c e n t r a t i o n = ( 1 T 1   −   1 T 1 b a s e l i n e   )   ÷ r 1

The Appendix contains the equations expressed in a form that can be pasted into excel to perform the calculations from TE = 3 ms and TE = 6 ms image pairs. The values were then expressed as change from the baseline and 3 point floating averaging was applied.

Prism 8 software (GraphPad Software, United States) was used for plotting the data.

Spontaneous brain tumours 5 months post induction displayed variable size and T₁ gadolinium enhancement (Figure 1). Similar to tumour growth in humans, the model was extremely heterogenous in growth rates and not all animals produced a tumour. The tumours produced included a large tumour with strong Gd enhancement (Figure 1A), a large tumour with reduced Gd enhancement (Figure 1B), and three slow growing tumours with weak Gd enhancement (Figures 1C–E). The T₂ weighted images provided easier visualisation of the tumours with weak enhancement. This allowed the potential of the dual echo method to measure changes in physiology during tumour growth to be assessed.

Firstly, we compared Gd and PET data from the thalamus region as control tissue in all mice. A rapid initial increase and subsequent decrease in the T₂*-derived Gadovist concentration (Figure 2A) is consistent with vascular Gd bolus passing through the tissue and then equilibrating throughout the total blood volume. The slower uptake and gradual decrease observed in the T₁-derived Gadovist concentration curve (Figure 2B) is consistent with Gadovist crossing the BBB and entering the brain tissue [5]. [¹⁸F]DPA-714 thalamus time activity curve (TAC) displayed a rapid increase followed by a decrease (Figure 2C) in activity. Plots were consistent for mice 1, 3, 4 and 5 with lower T₂* and TAC curves for mouse 2.

The T₂* concentration curve for mouse 2 was below the other four animals (Figures 2A, 3E, green data). Increased resistance during intravenous dose injection was noted, consistent with poor cannulation and the solution entering the tail tissue instead of the tail vein. This illustrates the power of the T₂* derived Gd concentration to ensure cannulation consistency in group studies where animals with decreased agent delivery can be excluded from further analysis. The decreased Gd delivery to the normal tissue was also reflected in the decreased activity of the PET tracer activity curve (Figure 2C), illustrating the parallel changes in both Gd and PET dynamic tracer injection. The PET dynamic curve of mouse 2 was flat compared to the decrease in intensity post bolus of the other mice consistent with slow release of the tracer injected into the tail tissue, rather than the vein. The T₁ derived Gd concentrations curves (Figure 2B) reflect tissue uptake of Gd and were grouped together. This is consistent with the intact blood brain barrier restricting transport of the Gadovist from the blood into the brain tissue. The reduced vascular delivery in mouse 2 did not appear to alter the normal tissue uptake of Gadovist. The decreased intensity of the Gd and PET tracer curves in the normal tissue of mouse 2, was not exhibited in the tumour tissue of mouse 2, as discussed below.

The tumour tissue in mice 1 and 2 both displayed dramatically enhanced T₂* (Figure 3A) and T₁ (Figure 3B) derived concentration curves relative to corresponding thalamus curves (Figures 2A,B) and compared to the tumour tissue in mice 3, 4, and 5. Mice 3, 4 and 5 showed modest increases in T₂* and T₁ derived concentrations in the tumours (Figure 3C) compared to the normal thalamus tissue (Figure 3E).

The extent of the dramatic increase of delivery and uptake of Gadovist in the two large tumours was not evident in the time activity curves representing the tumour uptake of [¹⁸F]DPA-714 (Figure 3F). For example, at 10 min the T2* curve for the tumour in mouse 2 is approximately 8 times the other curves (Figure 3A), yet the TAC curve is only about double the other curves (Figure 3F). The TAC for the tumour in mouse 2 was similar in intensity to the other three tumours (Figure 3F), despite a higher bolus (Figure 3A) and the TAC amplitude being of less than half the intensity in the thalamus compared to the other mice.