Determination of the brain–blood partition coefficient for water in mice using MRI

Abstract

Cerebral blood flow (CBF) quantification is a valuable tool in stroke research. Mice are of special interest because of the potential of genetic engineering. Magnetic resonance imaging (MRI) provides repetitive, noninvasive CBF quantification. Many MRI techniques require the knowledge of the brain–blood partition coefficient (BBPC) for water. Adopting an MRI protocol described by Roberts et al (1996) in humans, we determined the BBPC for water in 129S6/SvEv mice from proton density measurements of brain and blood, calibrated with deuterium oxide/water phantoms. The average BBPC for water was 0.89±0.03 mL/g, with little regional variation within the mouse brain.

Introduction

Cerebral perfusion imaging techniques are often based on injection or creation of a tracer substance in the blood pool. Depending on the technique used, quantification of cerebral blood flow (CBF) requires subsequent determination of arterial and brain concentrations of the used tracer and its partition coefficient between brain tissue and blood. The partition coefficient is specific to a substance, species, and brain region, and has to be determined experimentally. Arterial spin labeling magnetic resonance imaging (MRI) techniques such as flow sensitive alternating inversion recovery MRI use magnetically labeled water as tracer. These techniques are increasingly being used in mice (Prass et al, 2007; Muir et al, 2008; Royl et al, 2009; Farr and Wegener 2010). The brain–blood partition coefficient (BBPC) for water has been determined for humans from measurements of brain and blood water content (Davis et al, 1953; Dittmer, 1961; Herscovitch and Raichle, 1985). A value of 0.9 mL/g was found as an average over brain regions with considerable variation between white matter (0.82 mL/g) and gray matter (0.99 mL/g; Herscovitch and Raichle, 1985). This has been closely reproduced in a human MRI study (Roberts et al, 1996). Regional variation of the BBPC has also been demonstrated in monkeys, with values ranging from 0.69 to 0.90 mL/g (Kudomi et al, 2005). Therefore, studies in humans or monkeys should use regional-specific BBPCs when quantifying CBF with positron emission tomography or MRI. In mouse MRI studies, an average value of 0.9 mL/g is usually adopted from human studies (Herscovitch and Raichle, 1985). However, distinction between gray and white matter is less clear in mice, and it remains unclear whether the average value obtained from humans applies to mice. Depending on the method used, the BBPC is a linear factor in the calculation of CBF for arterial spin labeling MRI (van Dorsten et al, 1999; Foley et al, 2005; Leithner et al, 2008). Therefore, the erroneous assumption of a BBPC may lead to a proportional error in the absolute quantification of CBF. To our knowledge, the BBPC for water has not been determined for mice.

In this study, we used an MRI protocol similar to the protocol described by Roberts et al (1996) for humans to determine the BBPC in mice. To this end, relative proton densities were measured for brain tissue of mice in vivo and in vials of fresh, anticoagulated mouse blood. Determination of proton density was validated using deuterium oxide/water calibration phantoms.

Materials and methods

We performed all animal experiments in strict accordance with the national and international guidelines. All animal experiments described herein were approved by the local official committee (Landesamt für Arbeitsschutz, Gesundheitsschutz und technische Sicherheit, Berlin, Germany). Male or female 129S6/SvEv mice (BfR, Berlin, Germany), 8 to 16 weeks old, were used for all experiments.

Eleven mice were anesthetized with 1.2% to 1.8% isoflurane in oxygen. Body temperature was continuously monitored and kept within physiological limits using a heated water jacket. Respiratory rate was also continuously monitored and kept within physiological limits by adjusting the isoflurane concentration.

In a first experimental group (n=6), fresh mouse blood was obtained shortly before MRI measurement from animals killed during other experiments. The blood was anticoagulated with ethylenediaminetetraacetate (EDTA), sealed in a plastic tube (2 mm diameter) and placed adjacent to the mouse head (Figure 1C). Five additional plastic tubes (2 mm diameter) were filled with a deuterium oxide/water mixture, with water content ranging from 20% to 100% in steps of 20% (mL water per mL total volume). The phantoms were placed adjacent to the head of the mouse (Figure 1C).

Figure 1.

(A) CuSO₄ calibration phantom showing minor B0 field inhomogeneities. These data were used to correct proton density measurements. (B) Calibration of proton density measurements with deuterium oxide/water phantoms, measured simultaneously with mice. In each of the six experiments, five tubes containing 20% to 100% water were measured (see images in C). The regression line through the mean measured proton densities shows excellent recovery of true water content. (C) Relative proton density images of one axial slice through the mouse brain with one vial of anticoagulated mouse blood and five calibration phantom tubes, imaged at echo time (TE)=1.2 and 3.2 milliseconds.

In a second experimental group (n=5), MRI was performed without blood or water/deuterium oxide phantoms. After killing the animal, blood and brain water content was determined by weighing blood and brain before and after desiccation at 180°C for 18 hours.

The MRI was performed on a Bruker 7T PharmaScan 70/16, with a Bruker 98/38 mm radio frequency (RF) Coil, operating on Paravision software platform (Bruker, Karlsruhe, Germany). In both experimental groups, proton density measurement was performed by a series of nine RF phase-spoiled gradient echo images (fast low angle shot (FLASH) method), with different repetition times (TR=0.1, 0.5, 1, 2, 3, 4, 5, 6, and 8 seconds, echo time (TE)=3.2 milliseconds, flip angle 90°C, slice thickness 1 mm, FOV 2.8 × 2.8 cm², 256 × 256 matrix, scan time ∼130 minutes). The signal intensity was monoexponentially fitted to

(SI is the signal intensity, T_1app is the apparent T₁, and M₀ is the equilibrium magnetization). For a three-dimensional assessment of B0 field inhomogeneities of the coil, a CuSO₄ phantom that filled the sample volume of the animal measurements was imaged with the same sequence (Figure 1A). Relative proton densities were then calculated by correcting M₀ for B0 field inhomogeneities as described by Roberts et al (1996).

In the first experimental group (n=6), one axial slice was imaged together with five deuterium oxide/water phantoms and a mouse blood sample (see above). In addition, the imaging sequence was repeated with a shorter TE of 1.2 milliseconds and a 64 × 64 matrix (total scan time ∼37 minutes) to evaluate T2^* effects on proton density calculation.

The second experimental group (n=5) was added to allow for a detailed region of interest (ROI) analysis and to perform desiccation experiments (see above). In this group, proton density measurement was performed in 10 1-mm thick slices covering the whole mouse brain.

From the measured proton densities of the deuterium oxide/water phantoms, a regression equation was obtained that related relative proton density measurements to water content (Figure 1B). This equation was used to calculate the water content of mouse brain and mouse blood. Hence, the water content of brain and blood was obtained as mL/mL. However, for historical reasons, CBF is usually quantified in mL/(g × min). This requires the partition coefficient to be given in mL/g (not mL/mL). Therefore, we corrected for the density of brain tissue in our BBPC calculation. A density of mouse brain of 1.04 g/mL was adopted from Bothe et al (1984), and the BBPC was calculated using Equation (2):

(C is the water content in mL/mL as obtained from calibrated proton density measurements, ρ_brain is the density of brain tissue=1.04 g/mL).

Results

The CuSO₄ phantom revealed minor regional B0 field inhomogeneities (Figure 1A), which were three-dimensionally used in the calculation of relative proton densities as described by Roberts et al (1996). Figure 1C shows two proton density images (TE=3.2 and 1.2 milliseconds) of one axial mouse brain slice along with five deuterium oxide/water phantoms and one vial of fresh, anticoagulated mouse blood. Within the brain slice, regional differences in proton densities are minor. Relative proton density measurements of the deuterium oxide/water phantoms yielded an excellent recovery of true water content (Table 1; Figure 1B; regression equation: measured water content=0.9995 × (true water content)+0.0007). This underlines that water content can be reliably determined using the chosen MRI approach.

Table 1. Water contents determined with MRI proton density measurements.

	BBPC (mL/g)	Brain (mLH2O/mL)	Blood (mLH2O/mL)	100% H2O	80% H2O	60% H2O	40% H2O	20% H2O
Mean	0.89	77	83	100	80	59	40	20
s.d.	0.03	2	5	1	2	2	1	1

BBPC, brain–blood partition coefficient; MRI, magnetic resonance imaging.

Mean and s.d. from n=6 measurements of mouse brain, mouse blood, and D₂O/H₂O calibration phantoms are shown. BBPC is given in mL/g assuming a density of 1.04 g/mL for mouse brain (Bothe et al, 1984).

For each of the six animals, the water content of the brain was calculated using the mean relative proton density of one axial brain slice, and the regression equation was determined from the five deuterium oxide/water phantoms that were measured together with this animal. The mean water content of mouse blood was calculated in the same way using the proton density measurement of the slice through a plastic tube filled with mouse blood. The results are shown in Table 1. A mean water content of blood of 0.83±0.05 (mL water)/(mL blood) (mean±s.d.) and a mean water content of brain of 0.77±0.02 (mL water)/(mL brain) (mean±s.d.) were determined. This translates to a BBPC of 0.89 mL/g assuming a brain density of 1.04 g/mL (Bothe et al, 1984). Measurements at TE=1.2 and 3.2 milliseconds lead to the same BBPC when averaging across all six animals, excluding significant T2^* effects on our BBPC determination.

Additional measurements of regional proton density on five mice yielded little regional variations of BBPC. Regional BBPC were 103%±1% of average whole brain BBPC for cortex and striatum, 101%±2% for thalamus, 100%±2% for hippocampus, 97%±1% for brainstem, and 95%±3% for cerebellum. The average whole brain BBPC determined from desiccation experiments on five mice was 0.93 mL/g, therefore, slightly higher than the BBPC determined with MRI.

Discussion

Using the data on brain and blood water content with Equation (2), we determined an average BBPC for water of 0.89±0.03 mL/g (mean±s.d.) for mice from MR measurements. This value can be used in the quantification of CBF in mice with MRI methods. It is close to the average value determined for humans: 0.9 mL/g or 0.95 mL/mL (Herscovitch and Raichle, 1985; Roberts et al, 1996).

Desiccation experiments yielded a slightly higher whole brain average BBPC of 0.93±0.01 mL/g. Brain and blood water contents determined with MRI were slightly lower than those determined by desiccation. This has been reported previously and might be related to small water fractions invisible to MR (Lin et al, 1997; Venkatesan et al, 2000). It remains unclear whether the whole amount of water detected by desiccation participates in free water exchange between brain and blood during the time scale of a perfusion measurement with MRI.

In contrast to desiccation experiments, MRI allowed for ROI analyses. Regional variations of BBPC were small in regions commonly delineated in mouse stroke studies (<3% for cortex, striatum, hippocampus, and thalamus). The BBPC was slightly lower for cerebellum and brainstem, in accordance with regional brain water content measurements (Stonestreet et al, 2003).

Further to determine the BBPC, our study indicates that MRI determination of brain water content is feasible and could potentially replace brain water content measurements (with limitations, see Lin et al, 1997; Venkatesan et al, 2000) by desiccation techniques, as are commonly used in experimental animal studies to quantify brain edema. In contrast to desiccation, MRI would allow for longitudinal as well as regional water content determination.

Our study has several limitations. Changes in hematocrit affect the BBPC for water, as water content of red blood cells differs from plasma water content. Mouse blood was obtained after decapitation from large vessels in our study and hematocrit might be different from hematocrit in smaller vessels in the brain. However, substantial variations in hematocrit will only lead to minor changes in BBPC (changes from 25% to 55% have been estimated to lead to BBPC changes from 0.86 to 0.93 mL/g; Herscovitch and Raichle, 1985). The BBPC might vary with age or strain. We measured young 129S6/SvEv mice (8 to 16 weeks old), which are typically used in animal stroke experiments. Therefore, our BBPC measurements might not extend to older animals or other strains. Variations of BBPC with age might be around 10%, as water content of the brain decreases roughly by that amount during aging. Variations of brain water content with gender have not been found in C57/Bl6 mice rendering large variations of BBPC with gender unlikely (Liu et al, 2008). Proton density measurements could in principle be used to correct for regional differences of the BBPC. However, in the mouse brain, regional differences in proton density and hence BBPC were low in regions routinely delineated in mouse stroke studies. Finally, CBF quantification with arterial spin labeling MRI techniques might resemble more closely the microspheres technique than a diffusible tracer experiment (Buxton, 2005), because the time scale of the experiment is on the order of a second rather than of a minute as in positron emission tomography experiments. Therefore, the calculation methods that do not incorporate the BBPC have been developed. However, many recent studies continue to use calculation algorithms including the BBPC. When applying these studies on mice, the obtained BBPC of 0.89 mL/g should be used.