General study design and experimental conditions
All experimental procedures were approved by the Local Committee for Animal Research Ethics (permit number M285-11) and performed according to the guidelines of the Swedish Animal Welfare Agency and in agreement with international guidelines. Female Wistar adult rats (Charles River Laboratories, Germany) with a weight of 220–250 g were used. Animals were housed on a 12-h light/dark cycle with ad libitum access to food and water. All experiments, including surgery and MRI, were conducted while the animals were anaesthetised. The sizes n1 and n2 of the control group (index 1) and the study group (index 2) were calculated using the formulas n1 = [(\( {\sigma}_1^2 \) + \( {\sigma}_2^2 \)/K)(1.96 + 0.84)2]/[|m1-m2|2] (for power = 0.8 and α = 0.05) and n2 = K·n1, where K is the enrolment ratio [24]. For the CBF endpoint the following input data were used: K =1.3, mean m1 = CBFmean~ 160 mL/100 g/min [25], expected mean m2 ~ 136 mL/100 g/min, \( {\sigma}_1 \) = \( {\sigma}_2 \) = SDCBF ~ 20 mL/100 g/min [25]. Although sample size calculations are based on the primary endpoint, a supplementary calculation based on CTT was performed for corroboration, using K = 1.3, m1 = CTTmean ~ 1.43 s [18], expected m2 = 1.57 s, \( {\sigma}_1 \) = \( {\sigma}_2 \) = SDCTT~ 0.12 s [18]. In accordance with the conditions of the ethical approval, all animals were euthanised immediately after the MRI experiments using a lethal dose of pentobarbital by intravenous or intraperitoneal injection.
Animal anaesthetics, surgery and blood preparation
Animal anaesthetics
Anaesthetics were administered by inhalation through a locally manufactured mask, and anaesthesia was induced by 4% isoflurane and then maintained at 2.2% in a 1,000 mL/min oxygen mixture. Nitrous oxide (N2O), commonly used in an anaesthetic mixture, was avoided in this case because it is known to produce a significant increase in CBF in rats by up to 100% [26]. During the surgical procedure (normally taking between 20 and 30 min), an animal body temperature of 37 °C was maintained by an electrical heating pad.
Animal preparation
Animals were placed on the surgery table, and a 21G butterfly needle with catheter PE 50, filled up with heparin saline, was inserted into the lateral tail vein for transfusion of RBC suspension. Fresh blood was obtained from healthy anaesthetised donor rats (of the same provenience and characteristics as the rats in the CG/SG) by cardiac puncture. The blood from one donor rat (7–8 mL) was used for the preparation of the RBC suspensions for two transfusions.
Preparation of the red blood cells
The RBCs were separated from plasma by centrifugation at a relative centrifugal force of 1500G during 30 min. The RBCs were divided into two groups, i.e., (A) suspension with normal RBCs, injected into the animals in the control group (CG) and (B) suspension with RBCs with reduced deformability, injected into the animals in study group (SG). To reduce the deformability of the cell membrane, the RBCs of group B were subjected to minimal hardening by incubation in 0.025% glutaraldehyde in phosphate-buffered saline at pH 7.4 during 30 min at room temperature. Following the incubation, the RBCs were washed in phosphate-buffered saline three times to completely remove the glutaraldehyde. The RBCs of group A were handled in the same way, but without adding the glutaraldehyde solution. The washed erythrocytes were diluted in the phosphate-buffered saline. Haematocrit was measured and adjusted by dilution to a level equal to that of the experimental rats [20].
MRI experiments
For the MRI investigation, all animals were rapidly repositioned from the surgical table into an MRI animal holder, where the head of the rodent was secured using ear bars, bite bar and a nose cone to minimise motion during image acquisitions. The neck of each rat was also fixated with adhesive tape to reduce respiratory movement. The animal positioning, relative to the radiofrequency and gradient coils, was carefully considered to minimise animal-to-animal variability. Throughout the MRI experiments, breathing rate and oxygen saturation level were controlled, and the body core temperature was monitored by use of a rectal thermometer and maintained at 37 °C using warm air flowing into the magnet bore. The breathing rate was monitored with a MRI-compatible system (Model 1025, SAII, Stony Brook, NY, USA) using a respiratory air-filled pressure sensor placed in contact with the abdomen, and the breathing rate was controlled by adjusting the depth of anaesthesia to maintain a stable respiration rate around 60 min−1.
The MRI experiments were performed using a 9.4 T horizontal bore animal scanner (Agilent Inc., Palo Alto, CA, USA) with the 205/120 HD (High Duty Cycle) gradient coil. A 72-mm inner diameter volume coil was used for radiofrequency transmission, and the signal was received using a 4-channel array head coil. T2-weighted fast spin echo anatomical images were used for slice localisation. To standardise the scan procedures, the imaging slice of interest was positioned just frontal to the bregma, which is an easily recognisable structure in MRI images, located at the most forward crossing fibre of the anterior commissure [27] (Fig. 1a). A single-slice flow-sensitive alternating inversion recovery ASL sequence with a three-shot segmented spin-echo echo-planar imaging readout was implemented with the following parameters: field of view 32 mm2; matrix 642; slice thickness 1 mm; repetition time 5.1 s; echo time 10 ms; spectral width 250 kHz; echo spacing 0.32 ms; tag region width 5 mm (to ensure that there was no mismatch between the effect of the slice selective and the global inversion pulses on the imaging slice [28]). A hyperbolic secant adiabatic inversion pulse with a bandwidth of 20 kHz was used for both the slice selective and global inversions. To minimise the signal from the static tissue, two pre-saturation pulses were applied in the slice of interest.
For each dataset, ASL images at ten different time points (corresponding to different inversion times) were acquired for model fitting according to Eq. 2, starting at inversion time (TI) 300 ms followed by a 300 ms increment between time points. Before the injection of RBCs, one ASL dataset was acquired, in both CG and SG rats, to be used as baseline, and after injection totally five ASL datasets were acquired for each animal in both groups. Hence, in the analysis (described below), six datasets for each animal were evaluated separately, i.e., an initial pre-injection baseline measurement (assigned time t = 0) and five postinjection measurements. Considering that the biological half-life of the damaged RBCs was assumed to be approximately 13 min [20], the first postinjection data point was acquired 10 min after injection and subsequent postinjection data points were acquired in 11 min and 30 s intervals.
Tissue T1, as required for the modelling according to Eq. 2, was estimated separately on a region of interest (ROI) basis in a location corresponding to the area used in the ASL measurements, using the inversion recovery fast spin-echo imaging method. The parameters used for T1 estimation were as follows: repetition time 5.1 s; TI 0.022, 1, 2, 3, 4, 5 s; effective echo time 28 ms; echo-spacing time 14 ms; echo train length 4; number of excitations 4. The first echo was assigned to the centre of k-space.
Theory and postprocessing procedure
Labelled and control images were pair-wisely subtracted to attain perfusion-weighted ASL-MRI images, and these difference maps were denoised by wavelet-domain filtering [29] using an in-house software program written in IDL 7.1 (ITT Visual Information Solutions, Boulder, CO, USA).
The CBF values were calculated using a general kinetic model [2] according to the Buxton approach (Eq. 1):
$$ \Delta M(t)=\left\{\begin{array}{cc}0& 0<t\le BAT\\ {}\frac{2{M}_{0B}\bullet CBF\bullet \alpha }{\Delta R}{e}^{-\frac{(BAT)}{T_{1b}}}\bullet \left(1-{e}^{-\Delta R\left(t- BAT\right)}\right)\bullet {e}^{\frac{-\left(t- BAT\right)}{T_{1 app}}}& BAT<t\le BAT+ BL\\ {}\frac{2{M}_{0B}\bullet CBF\bullet \alpha }{\Delta R}{e}^{-\frac{(BAT)}{T_{1b}}}\bullet \left(1-{e}^{-\Delta R\bullet BL}\right)\bullet {e}^{\frac{-\left(t- BAT\right)}{T_{1 app}}}& BAT+ BL<t\end{array}\right. $$
(1)
where ∆M is the magnetisation difference (obtained from difference maps), T1b [= 2.2 s at 7 T [30]] is the spin-lattice (longitudinal) relaxation time of blood, M0B is the equilibrium magnetisation of arterial blood, α is the inversion efficiency (assumed to be 1), BAT is the bolus arrival time, BL is the bolus length, ΔR is R1b- R1app, where R1b is the relaxation rate of blood (R1b=1/T1b = 1/2.2 s), and R1app= 1/T1app is the apparent relaxation rate measured under the inflow of fresh blood magnetisation. Values of M0B, BAT, BL, and T1app were all estimated by fitting the Buxton model to the dynamic, multi-TI, experimental ASL-MRI data.
The relative arterial transit time (rATT) and the mean capillary transfer time (CTT) were extracted by fitting the model in Eq. 2, which was adapted to allow for the use of an experimental arterial input function, according to Appendix.
$$ \Delta M(t)=\frac{rTTT\bullet S}{\sqrt{4\pi \bullet CTT}}{\int}_0^t\Delta {M}_a\left(\tau \right)\frac{e^{-\frac{t-\tau }{T1}}}{{\left(t-\tau \right)}^{3/2}}\mathit{\exp}{\left(-\frac{rTTT-\left(t-\tau \right)}{\sqrt{4\bullet CTT\bullet \left(t-\tau \right)}}\right)}^2 d\tau $$
(2)
where rTTT is the relative total transit time (rTTT = rATT + CTT), ΔMa is the measured arterial concentration (i.e., the arterial input function), registered in a large artery in the same slice as the tissue concentrations curves. S is a scaling factor which accounts for proportional errors in the arterial concentration estimation, owing to issues such as spatial variations in coil sensitivity, differences in equilibrium magnetisation, differences in T2 and arterial partial volume effects. To obtain a reasonable estimate of S, preliminary fittings were conducted with S as a free parameter. In the final analysis, to prevent overfitting, S was taken as the mean value over all pre- and post-injection time points for each animal, obtained from the preliminary fitting procedure. T1 is the longitudinal relaxation time of the tissue, calculated by least-squares fitting of Eq. 3 to the experimental inversion recovery MRI signal from each ROI:
$$ {M}_Z(TI)={M}_0\bullet \left(1-2\alpha \bullet {e}^{\frac{- TI}{T_1}}+{e}^{\frac{- TR}{T_1}}\right) $$
(3)
where MZ (TI) is the magnitude MR signal collected at each TI, α is the inversion efficiency and M0 is the fully relaxed magnetisation. M0, α and T1 are fitted parameters.
The underlying fitting process was performed in MATLAB (R2018a, The MathWorks, Inc., Natick, MA, USA) using FMINUIT (version 2.3.1, Giuseppe Allodi, Dipartimento di Fisica, Universita di Parma, Italy), which is based on MINUIT (CERN, Geneva, Switzerland) [31].
ROI-based analyses
Regional CBF and CTT values were calculated, on a ROI basis, for six predefined cerebral ROIs, corresponding to the left cortex, right cortex, left putamen, right putamen, and white matter. A ROI including the total brain tissue in the chosen slice was assumed to represent global brain levels and is referred to as the whole brain (WB) estimate below (see Fig. 1). ROIs were placed by an experienced preclinical MRI scientist (A.B.). Results were reported as absolute baseline levels in combination with relative changes after injection of damaged RBCs, as this facilitates a direct comparison of the effects (in terms of magnitude and variability) on CBF with the corresponding effects on CTT.
Statistical analysis
Statistical analysis was accomplished using a repeated measures ANOVA test (α = 0.05) to test for effect of group (CG versus SG) and effect of time after injection (XLSTAT, Addinsoft, Paris, France). A limited number of planned observations were conducted, and no correction for multiple comparisons was required [32]; p values are reported, which enables readers to apply an alternative approach for estimating a different level of inference.