Visualization of kidney fibrosis in diabetic nephropathy by long diffusion tensor imaging MRI with spin-echo sequence

Renal fibrosis (RF) is an indicator for progression of chronic kidney disease (CKD). Although diabetic nephropathy (DN) is the leading cause of CKD and end-stage renal disease in Western populations, the ability of MRI to evaluate RF in DN patients has not been determined. As a first step to identify possible MRI methods for RF evaluation, we examined the use of diffusion tensor imaging (DTI) MRI to evaluate RF in a rat model of DN (SHR/NDmcr-cp(cp/cp): SHR/ND). The signal-to-noise ratio in DTI MRI was enhanced using a spin-echo sequence, and a special kidney attachment was developed for long-term stabilization. The changes in renal temperature and blood flow during measurement were minimal, suggesting the feasibility of this method. At 38 weeks of age, RF had aggressively accumulated in the outer stripe (OS) of the outer medulla. FA maps showed that this method was successful in visualizing and evaluating fibrosis in the OS of the SHR/ND rat kidney (r = 0.7697, P = 0.0126). Interestingly, in the FA color maps, the directions of water molecule diffusion in RF were random, but distinct from conventional water diffusion in brain neuron fibers. These findings indicate that DTI MRI may be able to evaluate RF in CKD by DN.

DWI technique, is used to visualize the brain fiber structure by enhanced signals in a fractional anisotropy (FA) map 13,14 . This is because DTI can differentiate water molecule diffusivities both along and against neuronal fibers. Consequently, it is easy to speculate that this MRI technique may also be able to detect RF, by analogy to its detection of neuronal fibers in the brain. To date, DTI has been trialed for the evaluation of kidney tissue damage in CKD patients 15 , including glomerulonephritis 16 , diabetic nephropathy 17 , and allograft kidney injury 18 . However, in all of these studies, the kidney tissue FA values were decreased, rather than being enhanced as observed in the brain, and the correlation between signal intensity reduction and RF was obscure. Furthermore, the FA value in the inner medulla (IM) of rats with DN plus unilateral nephrectomy was found to be negatively correlated with interstitial fibrosis 19 . This body of evidence suggests that it is very difficult to visualize and evaluate RF using DTI in a similar manner to neuronal fibers, particularly in DN tissue where intracellular or interstitial edema affects the manner of water molecule diffusion. Previously, we developed a long DTI MRI method with spin-echo sequence to evaluate RF in UUO model rats 20 . Here we report successful visualization of RF in diabetic rats with enhanced fibrosis (SHR/NDmcr-cp(cp/cp): SHR/ND) using our DTI MRI method with spin-echo sequence and long duration measurement enabled by a special kidney attachment.

Results
Characterization of SHR/ND rats. The SHR/ND rat is a diabetic rat model with concentrated and enhanced RF 21 . As a pilot study, we examined the disease phenotypes of SHR/ND rats. SHR/ND rats showed extreme obesity and hypertension, with the kidney tissue of older rats showing massive fibrotic changes concentrated in the outer stripe (OS) of the outer medulla ( Supplementary Fig. S1a,b). Oral administration of the antihypertensive drug telmisartan at 5 mg/kg/day from 8-38 weeks of age to improve the hypertension and RF 22 dramatically reduced these fibrotic changes ( Supplementary Fig. S1a,b).
Confirmation of feasibility of long DTI MRI measurement with a special attachment. We speculated that our previously developed long DTI MRI method with spin-echo sequence 20 would detect the concentrated RF in 38-week-old SHR/ND rats, based on the results of our pilot study. However, the rats were so obese that it was very difficult to hold and fix the kidney from outside of the body using the previous attachment. Therefore, we developed a semi-invasive method to fold and retain the kidney in a humid environment in these obese rats ( Supplementary Fig. S2a). To confirm the feasibility of this DTI MRI system, we examined renal perfusion and temperature in kidneys held by the attachment for 3 h. During the 3 h, the renal temperature gradually reduced from 34.68 ± 0.40 °C to 32.70 ± 0.97 °C (Fig. 1a), but the change was not significant (n = 5; Wilcoxon matched test, p = 0.062). The relative renal blood flow at 3 h based on 0 h was 97.8 ± 7.8% (n = 5) measured using an Super Paramagnetic Iron Oxide (SPIO) (Fig. 1b,c). The relative renal blood flow data between 0 h and 3 h also showed no significant difference (one-sample t-test, p = 0.56).
DTI MRI evaluation of RF in DN model rats at 8 weeks of age. The FA value of the IM on DTI MRI was found to be the only parameter associated with total RF, with the FA values in rats with DN plus unilateral nephrectomy being decreased in the cortex (CO) and OS of the medulla 19 . These findings suggested that MRI would be sufficiently sensitive to detect RF, as shown by the total decrease in the FA value in rats with DN plus unilateral nephrectomy.
At 8 weeks of age, SHR/ND rats had a higher body weight, more profound hypertension, lower renal function, more elevated serum glucose, and more urine protein than control WKY/ism (WKY) rats (Table 1). However, the degree of RF was similar in the kidney tissues of 8-week-old SHR/ND and WKY rats (Figs 2a,b and S3a-d). Examination of these kidneys by DTI MRI showed that the calculated FA values in the CO, OS, and inner stripe (IS) of the outer medulla were lower in SHR/ND rats compared with WKY rats, while the FA values in the IM did not differ significantly.
DTI MR evaluation of RF in DN model rats at 38 weeks of age. SHR/ND rats showed more profound metabolic syndrome phenotypes at 38 weeks of age than at 8 weeks of age (Table 2), with the kidney tissue of older rats showing massive fibrotic changes concentrated in the OS (Figs 3e,f and S4b). Oral administration of the antihypertensive drug telmisartan at 5 mg/kg/day from 8-38 weeks of age to improve RF 22 dramatically reduced these fibrotic changes (Figs 3i,j and 4c and S4d-j). Consistent with the concentrated fibrotic changes in untreated SHR/ND rats, the FA signals were sufficiently high to be visible in the OS of the kidneys in the same animals ( Fig. 3e-g), but were not observed in WKY rats or SHR/ND rats treated with telmisartan ( Fig. 3c,k). Furthermore, the directions of FA in fibrotic areas of the OS in untreated SHR/ND rats appeared to be random (Fig. 3m), being distinct from the FA color map images of nerve fibers in the brain 23 . The FA values of the OS were higher in SHR/ ND rats than in other rats, and were higher in the OS than in other portions of the same kidneys (Figs 4a and S4k-m). The FA values of the CO were reduced in SHR/ND rats with or without telmisartan, consistent with the results at 8 weeks of age and with previous findings 19,24 . The mean diffusivity (MD) looked similar in all groups of rats ( Fig. 3d,h,i), but the quantified MD values were lower in the CO, OS, and IS of SHR/ND rats with or without telmisartan than in WKY rats (Fig. 4b). Based on the differences in FA values between the kidneys of WKY and SHR/ND rats, we analyzed the data from SHR/ND rats with and without telmisartan treatment only. We observed a significant correlation between the fibrotic area measured by Sirius Red staining and the FA value only in the OS (r = 0.7697, P = 0.0126) (Fig. 4d). Even if we included the data of WKY rats in addition to Fig. 4d, we still observed a significant correlation in OS (r = 0.7812, p = 0.0004) with the tendencies of other compartments not changed.

Discussion
This study identified an MRI method that can visualize RF, even in the presence of DN. ADC values calculated from diffusion MRI data were reported to accurately indicate RF in patients with CKD 11 and in a mouse UUO model 10 . As the kidneys handle considerable volumes of water to maintain body fluid volumes, their ADC values are likely to be heavily affected by circulation and urine currents 12 . Indeed, the ADC values in human kidneys were reported to be highly variable 25 . FA values are calculated from DTI MRI data, and performed in six different directions. This study used a spin-echo sequence, rather than the echo-planar imaging method conventionally used in DTI MRI, to enhance the signal-to-noise ratio 26 . Therefore, the MRI required 3 h for completion. To keep the kidney still during this time period without tissue compression, we devised a special procedure and device to isolate and fix the kidney, even in an extremely obese rat. To confirm the feasibility of this DTI MRI system, we examined the kidney temperature and blood flow during 3 h under the same settings used for the DTI MRI. We found that the changes in kidney temperature and blood flow during 3 h were not significant. In addition, we also selected an approximately-unbiased and automated method for measuring RF ( Supplementary Fig. S6) for reliability for data sampling.
Although a previous study attempted to evaluate histopathological changes by DTI MRI in a rat model of DN induced by streptozotocin and uninephrectomy, the FA values in the CO, OS, and IM of DN rats following uninephrectomy were not linearly correlated with tissue fibrosis 19 . We suspected that the increase in FA was undetectable because these rats experienced only modest fibrotic changes. Moreover, at the low b-value (300 s/mm 2 ) used, MR images are easily influenced by perfusion and flow effects. The present study utilized a rat DN model with more aggressive RF and a higher b-value (601 s/mm 2 ), resulting in the detection of increased FA in the fibrotic kidney tissues and visualization of RF in rats with DN. The striking difference between the FA values in the CO and OS of the kidneys in SHR/ND rats may make the FA signal increases more visible. Similarly, use of a higher b-value (1000 s/mm 2 ) in a previous study resulted in FA being significantly higher in the presence of massive liver fibrosis compared with modest liver fibrosis (4 vs. 2 weeks after CCl 4 insult) 27 . Although the mechanisms underlying the decreased diffusion anisotropy in DN tissues remain to be determined 17,19 , tissue edema may be a possible cause. Interestingly, the brain tissue in fulminant hepatic failure exhibits massive edema, which was reported to be a cause of the reduction in its FA value 28 . Our DTI MRI method could successfully detect enhanced signals in the aggressive RF area, suggesting that fibrotic tissue changes may increase FA values even in DN with edema. In contrast, the MD values in this study differed from those in a previous report 19 . Although the reasons for the discrepancies were not determined, they may arise through differences in the model animals or b-values. In the future, it will be interesting to measure multiple ADC components 29 in DN rats with RF.
In rat experiments, our DTI MRI method requires a very long time and is partially invasive owing to oscillation of the kidney and its surrounding fat. For studies in humans, the influence of oscillation by breathing is preventable by respiratory/pulse triggering measurement. However, we need to devise a more sensitive sequence to shorten the duration of measurement.

Experimental disease model and design.
Seven-week-old WKY and SHR/ND rats were allowed to acclimate for 7 days. Seven WKY rats and five SHR/ND rats were treated p.o. with vehicle (0.5% methyl cellulose; Wako Pure Chemical Industries Ltd., Osaka, Japan), and five SHR/ND rats were treated with 5 mg/kg/day telmisartan (Tokyo Chemical Industry Co. Ltd., Tokyo, Japan) suspended in 0.5% methyl cellulose from 8-38 weeks of age. Body weight, blood pressure, heart rate, urine volume, and urine protein excretion were measured weekly. After each examination, rats were euthanized and kidney samples were harvested for measurements of fibrotic markers. As a pilot study, we conducted almost the same experimental procedure without MRI measurements using WKY rats (n = 9), SHR/ND rats treated with vehicle (n = 9), and SHR/ND rats treated with telmisartan (n = 9). Fig. S2a) was used in hugely obese SHR/ND rats, in which the kidneys were surrounded by large amounts of visceral fat. The skin around the left kidney was cut and the kidney was carefully isolated from the surrounding fat tissue, taking care not to induce additional tissue injury or bleeding.

MRI of the kidneys of DN rats.
Before MRI measurements, the rats remained drug-free for about 4 days to wash out antihypertensive drug or vehicle. MR images were obtained using a Unity INOVA MR spectrometer (Varian Associates Inc., Palo Alto, CA, USA) with a JASTEC Horizontal Magnet 4.7 T (JMTB-4.7/310/SS; Japan Superconductor Technology Inc., Hyogo, Japan). In SHR/ND rats, a rabbit volume coil (Takashima Seisakusho Co. Ltd., Tokyo, Japan) was used for transmission, along with an in-house-made surface coil ( Supplementary  Fig. S2b). All measurements were performed in a 20 °C room. The rats were sedated by inhalation of mixed gas (O 2 0.35 L/min, N 2 O 0.15 L/min, 1.5% isoflurane; Mylan Inc., Canonsburg, PA, USA). All images were acquired without respiratory or pulse triggering. To compare different imaging sequences, images were acquired in the same geometry for the four portions, namely the CO, OS of outer medulla, IS of outer medulla, and IM.  MR image analysis. ADC, MD, and FA maps were obtained using FSL imaging software (FSL.5.0.6) 30 . To accurately measure MRI signal intensity, ADC, MD, and FA in four portions (CO, OS, IS, and IM) were determined using Image J software and its Restore Selection function as described 9 . Use of this function resulted in successful selection of the exact same portions of the kidney anatomically referenced in standard T 2 -weighted images for measuring MRI signal intensity, ADC, MD, and FA.

RF analyses.
Conventional methods for measuring RF use Masson trichrome staining of randomly obtained magnified fields, followed by image analysis with Image J software. However, because of possible biases in selecting fields, a Biorevo BZ-9000 microscope (Keyence, Osaka, Japan) and BZ-X analyzer software (Keyence) were used. After automated capture of 40x magnified images over the entire tissue sample, a single high-resolution image of the whole kidney was created seamlessly ( Supplementary Fig. S6a,b). Each whole kidney image was dissected into four portions (CO, OS, IS, and IM), and the renal fibrosis in each portion was quantified by Micro Cell Count software (Keyence) using common parameters ( Supplementary Fig. S6c-f). Sirius Red staining was used instead of Masson trichrome staining, because the former is more specific for collagen types I and III 35 . DAB based immunohistochemistry (collagen type I or α-SMA) was not used, because the software cannot be applied to the heterogeneous background staining observed in DAB based immunohistochemistry.
Statistical analysis. Unpaired and nonparametric t-tests (Mann-Whitney tests) were used to compare two different kidney tissues, while one-way analysis of variance and Dunnett's multiple comparison test were used to compare three or more samples or kidney tissues. All statistical analyses were performed using GraphPad Prism 6 software (GraphPad Software Inc., San Diego, CA, USA).