Visualization of sodium dynamics in the kidney by magnetic resonance imaging in a multi-site study

Sodium magnetic resonance imaging (MRI) is a powerful, non-invasive technique to assess sodium distribution within the kidney. Here we undertook pre-clinical and clinical studies to quantify the corticomedullary sodium gradient in healthy individuals and in a porcine model of diuresis. The results demonstrated that sodium MRI could detect spatial differences in sodium biodistribution across the kidney. The sodium gradient of the kidney changed significantly after diuresis in the pig model and was independent of blood electrolyte measurements. Thus, rapid sodium MRI can be used to dynamically quantify sodium biodistribution in the porcine and human kidney.

Sodium imaging was performed at site B with two Helmholtz loop coils placed anteriorly and posteriorly over the kidneys, protocol as above.
Three pigs were imaged on the above HDx platform and a further three on an MR750 at site B. Pigs were anaesthetised with propofol (0.4 mg/kg/h) and fentanyl (8 µg/kg/h), and ventilated. Arterial blood was sampled from pigs before the baseline scan, and subsequently at each time point.
Pigs were placed supine in the scanner. T1 weighted 1 H imaging was performed using an 8channel cardiac array (FOV = 380mm, reconstruction matrix = 288x288x192, TR = 43 ms, TE = 1 ms, FA = 12 degrees). Subsequent sodium imaging was performed with a Helmholtz loop coil pair as described above, including respiratory gating, with 5 NEX per volume and imaging acquired in 5 min blocks. Phantoms used were as described above in the site A protocol. IV furosemide (0.5 mgkg -1 ) was administered after baseline imaging and sodium imaging was performed with 5 min resolution for 30 minutes.
Arterial blood gas was sampled between each dynamic scan, with serum electrolytes (sodium, potassium, and chlorine) quantified.
Ex vivo renal sodium T1 weighted imaging was performed on four separate porcine kidneys post excision which were placed into a saline (154 mmolL -1 ) filled phantom at 37°C using a Helmholtz loop pair, with the coils placed above and below the phantom in the magnet. The imaging protocol consisted of 1 H scout images (Fast Imaging Employing Steady-state Acquisition, TR = 4 ms, TE = 2 ms, inversion time = 210 ms, flip angle = 65 degrees, slice thickness = 20 mm, slice spacing = 50 mm, averages = 16, FOV = 320mm, matrix = 256x256) and the sodium sequence described above, with the following modifications: TR = 250 ms, inversion preparation (adiabatic hyperbolic secant), inversion times = 20, 22, 26, 30, 34, 42 ms. T1 maps were calculated fitting a single exponential to data with non-linear least squares fitting in Matlab, fitting for T1 and the sodium density signal (M0). In total, four kidneys were imaged. Sodium concentration maps were formed from the M0 image after T1 fitting using normal saline (154 mmolL -1 ), and noise (i.e. 0 mmolL -1 ) as reference standards.

Image post-processing and analysis
Regions of interest (ROIs) were drawn on the coronal T2 images by a researcher, from three central slices in both kidneys, and were used to segment the renal cortex, medulla, and the whole kidney. Image processing and statistical analysis was performed in Matlab Where the ConcentrationCoefficient and ConcentrationOffset are derived from a linear fit between sodium phantom signal values and a region of noise.
1)2+& is the signal in a given voxel in the imaging slice.
The corticomedullary sodium gradient was defined by segmenting the kidney into concentric layers of equal thickness as previously described 21 . Twelve and seven layers were used for human and porcine segmentation, respectively. Voxels in each layer were averaged across all subjects at each site and combined to form a site-averaged corticomedullary sodium gradient.
After site data pooling the porcine sodium gradient was assessed as above. Dynamic alterations in the gradient were assessed (mM/mm) at each time point.
The % difference between each serum electrolyte at a given time point and baseline was calculated according to equation 3: Where 0+-64+&+,-)&7,+ is the concentration of a given metabolite (mmolL -1 ), and 0+-64+&+*,-)&7,+ is the concentration after administration of furosemide, at a given time point. Data were averaged over the subject group after % change was calculated.
Ex vivo renal analysis was performed by drawing regions of interest around each kidney, and the average T1 was calculated for all kidneys. Further regions were drawn segmenting the cortex and medulla for each kidney, with the mean sodium concentration in each compartment calculated over all the ROIs.
Further ex vivo analysis was performed to validate imaging derived sodium concentration measurements with a sodium assay based on the requirement of sodium ion as a cofactor for the enzymatic activity of b-Galactosidase (Sodium Assay Kit, MAK247, Merck). Twelve biopsy samples were acquired from ex vivo kidneys from the medulla (N = 6) and cortex (N = 6) and tissue were processed according to the manufacturer's instructions. Briefly, kidneys were excised and washed in saline (154 mmolL -1 ) to remove blood from the renal vasculature. Biopsy samples were acquired, and flash frozen in liquid nitrogen, and stored at -80°C for later analysis. After defrosting, samples were homogenised using 10 mmolL -1 Dithiothreitol, after which samples were spun for 10 minutes and supernatants taken for spectrophotometric analysis. A 23 Na standard curve was set up as per the assay instructions, and sample absorbance at 405 nm recorded to estimate sodium concentration.
Ex vivo chemical results were correlated with ex vivo imaging data using a linear fit in Matlab.

Figures
Supplementary figure 1 -Ex vivo sodium assay results correlate with ex vivo imaging.