To investigate the possibility of using combined blood oxygen level-dependent (BOLD) imaging and diffusion-weighted imaging (DWI) to detect pathological and physiological changes in renal tissue damage of the kidney induced by chronic renal hyperfiltration.
Materials and Methods
The apparent diffusion coefficient (ADC) and the T2* value within the inner compartments of the kidneys of 17 rats with diabetes mellitus were compared with the results obtained from a control group (N = 16). The influence of dynamic changes of the renal function on the blood-oxygen saturation was evaluated by comparing the T2* values before and after the active reduction of tubular transport by furosemide injection.
All compartments of the diabetic kidney showed significantly (P < 0.05) lower T2*-values compared to the control group. In particular, the very low values in the outer stripe (OS) of the outer medulla (OM) (T2*-normal: 69.4 ± 10.9 msec; T2*-diabetic: 51.4 ± 13.9 msec) indicated either hypoxia due to hyperfiltration, or renal blood volume changes. Diffusion imaging of the same area showed significantly lower ADC values (ADC-normal: 1.45 ± 0.26; ADC-edema: 1.19 ± 0.25 [10–9m2/s]) that correlated with pathological findings on histopathology. The injection of furosemide significantly (P < 0.05) increased T2* in all compartments of both populations while the ADC remained unchanged.
ALTHOUGH PHYSIOLOGICAL PARAMETRIC imaging techniques, such as diffusion imaging (1) and blood oxygen level-dependent (BOLD) imaging (2), have been widely applied to assess the functional state of the brain, their clinical use in other organs, such as the kidney, has so far been limited.
Renal involvement in diabetes mellitus is the main cause of end-stage renal failure in developed countries (3) and a leading cause of morbidity and mortality in diabetic patients (4). The unavoidable price of efficient renal function appears to be a delicate balance between sufficient oxygenation and medullary hypoxia (5). An increased tubular load, such as that caused by diabetes or decreased medullary perfusion, can easily shift the balance toward hypoxia. To evaluate the potential of BOLD and diffusion MRI techniques for the diagnosis of early stages of the renal damage associated with diabetes, we applied both techniques to an experimental model of diabetic nephropathy in 17 rats.
BOLD sequences can noninvasively demonstrate the level of intrarenal oxygen tension, as shown by Prasad et al (6, 7). Those authors used the transverse relaxation rate R2* to establish a link between the low level of oxygenation in the medulla and the active reabsorption along the medullary thick ascending limb of the loop of Henle.
Diffusion imaging, as shown by recent studies in the brain (8–10), is sensitive to both cellular edema and cellular atrophy, and hence to the tissue damage typically induced by acute or chronic hypoxia. Nevertheless, the clinical application of diffusion-weighted imaging (DWI) of the kidney has so far been limited to a few studies (11–16) due to the extreme motion sensitivity of the method. Furthermore, the complicated internal structure of the kidney leads to very anisotropic diffusion characteristics and thus necessitates advanced diffusion methods, such as diffusion tensor imaging (DTI) (17) and isotropic diffusion preparations (18–20), in order to obtain reproducible results.
According to the classification of Mogensen et al (21), the evolution of diabetic lesions within the kidney is associated with an initial renal hypertrophy and glomerular hyperfiltration. This phase is difficult to detect since it is, in general, asymptomatic. Subsequently, the hyperfiltration can lead to progressive glomerular sclerosis with micro-albuminuria. In the penultimate stage, both the glomerular filtration rate and the renal plasma flow progressively decrease, leading to proteinuria and chronic renal failure.
Since the resulting hyperfiltration increases the reabsortive tubular load, the aims of this study were to detect the resulting medullary hypoxia directly with BOLD imaging, and to investigate the possible influence of resulting pathological changes of the tubular cells on the ADC.
MATERIALS AND METHODS
The animal experiments were performed in accordance with the Declaration of Helsinki principles for animal research. Diabetes mellitus was induced in 17 rats by an intraperitoneal injection of 50 mg/kg of streptozotocine (Sigma-Aldrich). This agent destroys the islets of Langherhans within 24–48 hours, and thus type I diabetes subsequently develops with an hyperglycemia of >16 mmol/liter 2–3 weeks after injection. At this stage, the renal status is characterized by a nephromegaly, which appears on the first day of hyperglycemia, whereas the hyperfiltration appears 5 days later. The rats in this study weighed 250–300 g and were killed after the MRI session. A group of 16 rats, with the same body-weight range, was used as a control.
For imaging, a general anesthesia was induced by intraperitoneal injection of 0.5 mL/kg of chloral hydrate (Sigma-Aldrich), 8% diluted with serum. A 24 G catheter was positioned in the tail vein. To avoid movements of the kidneys during breathing, and magnetic susceptibility artifacts related to the adjoining bowel, an in-house-made plastic holder was used to block the movements of the kidneys and isolate them from bowel loops while avoiding compression of the parenchyma and renal vessels, as shown in Figure 1.
Pharmacological Test and Induction of Hypoxia
An injection of furosemide (Lasilix; Hoechst-Marion-Roussel Laboratories) was used to evaluate the influence of both the decrease of cellular workload due to active tubular reabsorption on intrarenal T2* measurements and the increase of tubular flow on ADC values. After the initial application of the imaging protocol, 10 diabetic rats and 11 normal rats received an intravenous injection of 10 mg/kg of Lasilix. The imaging protocol was repeated 10 minutes after the injection.
Since BOLD imaging is sensitive not only to the blood-oxygen saturation, but also to the blood volume, the possible influence of the anesthetics and the furosemide on the latter factor could lead to ambiguous results. In order to identify areas of the kidney where the blood-oxygen saturation is not the predominant contribution to T2* variations, an additional control experiment was performed. A short exposure (60 seconds) to a pure nitrogen atmosphere changes the blood-oxygen saturation without introducing other major changes in the hemodynamic parameters. No physiologic monitoring was performed during this experiment because it was previously demonstrated (22) on isolated turtle hearts, perfused with 95% nitrogen, that cardiac output and ventricular pressure development were unaffected by 2 or 3 hours of hypoxia. Therefore, we believed there was no reason for any significant change in the renal hemodynamics after only 1 minute of exposure. Nitrogen was administered by applying a dedicated mask on the head of five healthy rats and changing the composition of the inspired gas from normal air (21% oxygen concentration) to 100% nitrogen for 1 minute. The measurement of the T2* values was performed prior to and during nitrogen breathing. This allowed us to depict the compartments of the kidney wherein blood-oxygen variations correlate well with the T2* evolution.
MRI was performed at day 5. All images were obtained on a 1.5-T clinical MRI system (Philips Medical Systems, Best, The Netherlands) equipped with actively shielded magnetic field gradients with a maximal amplitude of 23 mT/m. Radiofrequency (RF) excitation was performed with the integrated body-coil, while for signal detection a surface coil of 5-cm diameter was used. To maintain the comparability between the different imaging sequences, all images were acquired with the same geometry. The image matrix of 256 × 102 and the FOV of 150 mm × 60 mm resulted in a nominal pixel size of 0.59 mm × 0.59 mm. Three slices with a thickness of 3 mm were acquired per experiment.
For BOLD-contrast imaging a gradient-echo sequence with nine echoes in the phase-encoded acquisition train was employed. The resulting echo time (TE) was 5.4 msec for the first echo and 3.6 msec (average TE = 19.8 msec) interecho interval between the subsequent echoes. The repetition time (TR) was 184 msec, with an excitation pulse flip angle of 30°, resulting in an imaging time of 5.5 min for 16 signal averages (including the scan preparation). For the experiments in which the animals were exposed to 60 seconds of pure nitrogen breathing, the BOLD sequence was slightly modified in order to obtain the images in less than 1 minute. Only four signal averages and a reduced imaging matrix of 256 × 80 were used.
The measurement of the ADC employed a segmented DW echo-planar imaging (EPI) sequence with 19 echoes per excitation with TEimaging = 24 msec. Diffusion preparation (b = 0, 130, 260, and 390 seconds/mm2) was achieved by using an isotropic diffusion preparation with a diffusion preparation time (TEdiff) of 64 msec, which is based on the isotropic preparation sequence proposed by Wong et al (19) and described for this application in detail by Ries et al (20). This results in a complete TE of 88 msec. The typical imaging time with a TR of 1.2 seconds was 7.83 minutes. The gradient waveforms and the amplitudes of the diffusion gradients were numerically calculated off-line in order to achieve a maximal isotropic diffusion weighting without giving rise to gradient cross-terms. Although the animals were immobilized in an adjustable frame, a navigator echo for each excitation pulse was acquired after the diffusion preparation to suppress residual motion artifacts (23).
In addition to the functional images, a standard T2-weighted image (TE = 100 msec) with the same geometry was acquired before each set of parametric images. This image was used both as an anatomical reference image and to monitor any changes in T2 induced by the pharmacological test. Note that since T2* changes are much more sensitive to blood oxygenation than are T2 changes, and quantitative T2 measurements are rather time-consuming, no high-precision T2 evaluation was performed.
The DW and BOLD images were analyzed off-line on a LINUX workstation using in-house software written in IDL (Interactive Data Language) and C++. The decay constant for each image series was calculated by applying a Levenberg-Marquard fitting algorithm to a monoexponential model on a pixel-by-pixel basis. The signal intensity of the T2-weighted images, and the ADC and T2* values were obtained in regions of interest (ROIs) chosen by a clinical radiologist from the T2-weighted images containing 15–20 pixels placed in the cortex (CO), the outer (OS) and inner (IS) stripes of the outer medulla (OM), and the inner medulla (IM) of each kidney. The signal intensity of the T2-weighted images was normalized to the latissimus dorsi muscle in order to account for signal changes arising from signal drifts and changes in receiver gain. The relative signal changes in the images obtained before and after the pharmacological test were then calculated.
Biological and Pathological Examination
After the rats were killed, 600 μl of blood were sampled by direct puncture of the inferior vena cava for evaluation of plasma creatinine (by Jaffé's method) and blood glucose (glucose oxydase method) levels.
For the microscopic analysis, both kidneys of all diabetic rats were removed, sectioned longitudinally, and fixed in a Dubosc-Brasil solution (600 mL of 80° alcohol, 240 mL of 30% formol, and 100 mL of saturated picric acid). Microsections (4–5 μm thick) were obtained, stained with hematoxylin-eosin-saffron (HES), and analyzed by the renal pathologist (C.D.) of our institution with light microscopy and a magnification of 15–100×.
The statistical significance of the differences of the obtained values was evaluated with an unpaired t-test with equal variance for comparisons of the different animal groups, and a paired t-test with equal variance for tests within the same animal group. The acceptance criterion was in both cases a P-value of <0.05.
Biological and Pathological Results
The mean glucose and creatinine levels of the diabetic rats at the date of the examination were 47.8 ± 14.5 mmol/liter and 69.5 ± 14.22 μmol/liter, respectively, compared to 11.8 ± 2.05 mmol/liter and 42.22 ± 8.1 μmol/liter, respectively, in the normal control group.
It was not possible to perform the pathology exam on three of the diabetic rats. Under microscopic examination, the cellular lesions were sparse, mainly affecting tubular cells distant from medullary vessels, and were predominantly located in the OS of the OM in all cases. As shown in Figure 2, the pathological examination showed three different kinds of lesions: edematous cells with clarification and swelling, atrophic cells, and foci of cellular necrosis. According to the type of cellular damage, diabetic rats were reclassified into the following three groups: 1) diabetic kidneys with predominantly edematous cells (N = 8), 2) diabetic kidneys with predominantly atrophic cells (N = 3), and 3) diabetic kidneys with dominating necrosis foci (N = 3).
BOLD Imaging/T2 Imaging
With the T2- and T2*-weighted sequences, four intrarenal compartments were identified in the normal and diabetic animals. These compartments were identical in all animals in (from outside to inside) the CO, OS, and IS of the OM and IM, as shown in Figures 3 and 4.
Table 1 shows the T2* values obtained before and after respiration of pure nitrogen in the five healthy rats. T2* in the CO, OS, and IS decreased by 30–40% in all animals, with low interanimal variations. Although the T2* values of the IM reflect, in principle, the same behavior, the interanimal variations were much higher and thus the changes induced by the nitrogen were statistically not significant.
Table 1. Average T2*-Values (N = 5) in Kidney Compartments Before and After Exposure to Pure Nitrogen Atmosphere for One Minute
T2* before nitrogen [10−3 seconds]
T2* after nitrogen [10−3seconds]
All differences are statistically significant (P < 0.05) except the values for the inner medulla.
76.0 ± 9.7
43.8 ± 4.4
65.9 ± 12.2
38.1 ± 6.1
87.9 ± 15.7
53.2 ± 8.8
157.0 ± 74.5
102.3 ± 40.0
As shown in Table 2, all compartments of the diabetic kidneys (N = 17) showed statistically significant lower T2* values than those of normal kidneys (N = 16). In both groups, T2* was significantly lower within the OS than in all other compartments of the kidney. In all animals, the lowest value was always observed in the OS. However, in each compartment, no correlation was found between the T2* value and glycemia or creatinine levels (–0.3 < r < 0.1 and –0.1 < r < 0.36, respectively). A reclassification of the T2* findings of the diabetic population according to the pathological results into edematous, atrophic, and necrotic groups did not show any statistically significant differences.
Table 2. Regional Mean T2* Values [10−3 seconds] in Normal and Diabetic Rats Before and 10 Minutes After Injection of Furosemide
Normal rats (N = 11)
Diabetic rats (N = 10)
T2* before furosemide
T2* after furosemide
T2* before furosemide
T2* after furosemide
Except for the differences in the inner medulla, all changes showed statistical significance (P < 0.05).
83.9 ± 11.3
87.0 ± 15.0
67.4 ± 10.4
75.3 ± 9.6
70.7 ± 11.0
75.0 ± 11.48
49.2 ± 12.15
61.5 ± 17.78
91.9 ± 18.2
94.3 ± 23.1
64.7 ± 20.1
75.4 ± 22.9
145.3 ± 31.8
162.2 ± 38.8
101.1 ± 29.1
120.0 ± 44.4
Effect of Furosemide on T2*
The infusion of furosemide induced, 10 minutes after injection, a significant increase in T2* in all compartments of the kidneys of the normal (N = 11) and diabetic (N = 10) groups (Fig. 5). These changes are summarized in Table 2. The normal group showed a relative T2* change of +6.1% within the OS, and somewhat smaller positive changes were seen in the CO and the IS (+3%). As shown in Table 2, the relative changes for both groups were largest in the OS, and the diabetic group showed the largest modifications (+25.0 %).
The mean relative changes of the signal intensity of the T2-weighted images obtained before and after the pharmacological test were found to be <12.5% for both populations, as shown in Table 3, and their offset from their initial value was within half a standard deviation.
Table 3. Relative Signal Changes (%) on T2-Weighted Images Before and 10 Minutes After Injection of Furosemide
11.7 ± 24.1
12.5 ± 22.9
0.7 ± 15
1.1 ± 8.0
6.5 ± 11.8
−3.5 ± 9.2
The comparison of the diffusion measurements in the kidneys of the diabetic (N = 17) and normal (N = 14) rat groups is presented in Table 4. Two normal animals were excluded due to poor immobilization of the kidney. In the four renal compartments, the ADC values did not show statistically significant differences. However, when we considered the type of cellular damage, the ADC in the OS did show, for all animals with edematous cellular damage, statistically significantly lower values compared to the normal control group (Table 5 and Figure 6). Unfortunately, the resulting groups were too small for us to perform a multifactorial analysis to investigate the correlation between the groups. The injection of furosemide did not produce a significant variation in the ADC in any of the four intrarenal compartments of normal (N = 11) or diabetic (N = 10) kidneys, as shown in Table 6.
Table 4. Comparison of Regional Mean ADC Values [10−9 m2/seconds] Between Normal and Diabetic Rat Kidneys
ADC normal (N = 14)
ADC diabetic (N = 17)
1.44 ± 0.26
1.53 ± 0.31
1.45 ± 0.26
1.38 ± 0.32
1.09 ± 0.23
1.28 ± 0.29
1.04 ± 0.21
1.20 ± 0.38
Table 5. Mean ADC Values in Outer Stripe of Outer Medulla in Normal and Diabetic Rats According to Pathologic Results
Table 6. Regional Mean ADC Values [10−9 m2/seconds] in Normal and Diabetic Rats Before and 10 Minutes After Furosemide
Normal rats (N = 11)
Diabetic rats (N = 10)
All differences are statistically not significant (P > 0.05).
1.43 ± 0.27
1.45 ± 0.23
1.53 ± 0.32
1.70 ± 0.19
1.45 ± 0.26
1.36 ± 0.22
1.56 ± 0.37
1.59 ± 0.34
1.08 ± 0.24
1.11 ± 0.25
1.35 ± 0.31
1.35 ± 0.28
1.03 ± 0.21
1.14 ± 0.26
1.25 ± 0.44
1.38 ± 0.36
The activity of the Na+-K+-ATPase pump is highest in the proximal tubule and in the thick ascending limb of the loop of Henle, located in the OS of the OM, indicating the importance of active Na+ reabsorption at these sites (24). This pump has two major roles: it actively transports reabsorbed Na+ out of the cell and back into the systemic circulation via the peritubular capillaries, and maintains a low intracellular Na+ concentration that allows luminal Na+ to continue to enter into the cell via a concentration gradient (25). Although this requires almost half of the total oxygen utilization of the kidney (26), the medulla is relatively poorly oxygenated, causing the cells from the OS of the OM to operate close to ischemic conditions, and hence to be particularly susceptible to injury (27). All the conditions that result in an increased filtration rate expose the kidney to an increase in oxygen consumption, especially in the OS of the OM, because of an increase of Na+-K+-ATPase pump activity (28). The resulting high oxygen demand may exceed the arterial supply, possibly resulting in ischemic injury in this area. In diabetic nephropathy in particular, it has been shown that, initially, kidney growth causes an increase in proximal reabsorption and subsequently an increase in glomerular filtration, due to a tubuloglomerular feedback mechanism (29). This mechanism confirms the importance of modifications of the tubular workload and their consequences in diabetic nephropathy.
In our study, we were able to differentiate the CO and the three medullary compartments. The lowest T2* was found within the OS of the OM, which corresponds adequately with renal physiology. Prasad et al (6, 7) were the first to report lower T2* values within the medulla as compared with the CO, but without differentiation of various medullary compartments. Baseline T2* values in all compartments of diabetic kidneys were significantly lower than the values obtained in the normal control group. These differences were observed in the CO (around 20% shorter in diabetic kidneys) and the medulla (around 30% shorter in the three compartments of diabetic kidneys). The hypothesis that oxygen consumption increases in all compartments of the kidney, predominating in the medulla, due to an increased reabsorption work in this model is likely. Consequently, we suspect the OS has a much lower energy reserve than the other parts of the kidney. This could explain the increased risk of ischemia in this area caused by hyperfiltration, as evidenced by the predominance of cellular lesions in this compartment of diabetic kidneys.
The link between the injection of furosemide and an augmented medullary blood-oxygenation has been established (30). Our findings are comparable with previous results (7) showing an increase of T2* within the medulla and, to a lesser degree, the CO. In our study, the largest relative increase in T2* was observed in the OS for both groups, which also corresponds adequately with known renal physiology: these increased T2* values (and slightly increased T2 values) (31, 32) are compatible with an improved blood-oxygen level caused by a decreased workload. The cortical increase, whereas furosemide is a loop diuretic, could be explained by the fact that in rats the upper part of the ascending limb is located within medullary rays of the renal CO. The increase in the IS could be explained by the deeper location of ascending limbs from juxta-medullary nephrons. This increase was four times larger in the diabetic population than in the healthy control group, indicating that the blood-oxygen saturation was much lower in the diabetic group.
The presence of microscopic alterations of tubular cells in this early phase of diabetic nephropathy was an interesting finding. The absence of correlation between the T2* values and the type of cellular changes (edema vs. atrophy vs. necrosis) strengthens the argument that the observed MR changes are not influenced by “anatomical or pathological” changes, but by functional changes only. However, definitive conclusions could not be drawn because the rat population was relatively small. On the other hand, since these T2* changes are related to the specific phase of hyperfiltration of diabetic nephropathy, it is conceivable that this effect could be only transient, because with chronic renal failure the oxygen consumption should be decreased and T2* unchanged. This should be taken into account in future human trials.
However, BOLD MRI cannot distinguish between the changes in oxygenation caused by an increased supply of oxygen—for example, those caused by increased blood flow or reduced energy consumption. Furthermore, variations in the pH value modify the hemoglobin-O2 dissociation curve, leading to a pH-dependent BOLD effect.
The preparation experiment we performed in our study employed a pure nitrogen atmosphere in order to achieve a strong variation of the blood-oxygen saturation while changing neither blood flow nor blood volume. Consequently, a well defined shortening of T2* can be expected in those areas where T2* depends purely on blood-oxygen saturation. The CO, OS, and IS show a very distinct decrease of T2*; the IM followed the same trend, but with large interindividual variations (Table 1). The high variations of the values obtained in the IM led to the conclusion that other factors, such as the dominant presence of urinary water or pH differences between the IM and the peripheral blood, severely biased the measurements. Since the signal changes cannot be purely attributed to blood-oxygen variations, the IM was excluded from the rest of the discussion.
Therefore, we must take into consideration the three main contributing factors of BOLD signal: flow, blood-volume, and blood-oxygen concentration.
A glomerular cortical hyperfiltration in diabetic rats is accompanied by an increased blood flow in the CO and the medulla (33–38), which should lead to larger T2* values in all compartments of the diabetic kidney. As the observed T2* values in the diabetic kidneys show the opposite behavior, we can conclude that the magnitude of effect due to increased workload dominates over any possible blood-flow effect.
Blood volume in the CO is expected to increase due to the vasodilatation of the afferent arteries (34–38) secondary to hyperfiltration, whereas it has been demonstrated that the efferent arteries (seating downstream of the glomerulus) remained unaltered (39). Therefore, since the medullary vasa recta originate from the efferent arteries, the medullary blood volume can be expected to be much less affected by diabetes.
This idea is further supported by the T2* findings after furosemide injection. Furosemide reduces the blood flow within the medulla (30), but shows vasoactive effects in this compartment only when used in high doses (>10 mg/kg) (31). Therefore, after injection of furosemide, decreased T2* values would be expected if the reduction of blood flow were the dominating factor for the reduction of tubular workload, and a major change of the signal intensity of the T2-weighted images would be expected if vasoactive effects dominate.
However, these findings have to be carefully interpreted considering that the induced change of the renal reabsorption might change the urine/blood volume fraction within the OM, which might also lead to increased T2 (and thus T2*) values. However, since T2 within the IS does not seem to be affected by furosemide, this factor does not appear to be the dominating effect.
Therefore, we can consider that a higher oxygen consumption, related to increased active water reabsorption secondary to hyperfiltration, can explain a medullary decrease of T2* in early diabetic nephropathy. Alternative explanations, such as alterations of glucose metabolism, could also be discussed; however, in our opinion, they are too speculative.
The complementary information from the ADC values of both populations corresponds with the above findings. Although the ADCs of the CO, IS, and IM did not show any statistically significant differences among them, the reclassification of the diabetic population according to the observed histopathology appeared to indicate a link between the observed ischemic cellular damage and the observed ADC. The majority of the cases did show a reduced ADC in the diabetic rats, probably due to cellular edema (Table 5). However, since a multifactorial analysis between the groups was not performed, these small differences must be considered with caution. Nevertheless, these findings correspond well with the results of Yamashita et al (14) and Vexler et al (40), who reported decreased ADC values in cases of chronic renal failure and renal ischemia, respectively.
Müller et al (12) attributed high ADC values in the medulla to tubular flow-dependent effects caused by water transport mechanisms in the renal tubules. However, if ADC measurements were sensitive to osmotic transport processes within the kidney, higher ADC values (due to hyperfiltration in the diabetic kidneys) should have been observed. Since this was not the case, it appears that the ADC values do not depend on tubular flow effects. Furthermore, our experiments showed no significant ADC changes in the OM following the injection of furosemide, although the active reabsorption of water after the injection was practically halted and tubular flow was increased.
The combination of BOLD-contrast and DWI may enable a comprehensive assessment of the functional status of the kidney. The high-energy metabolism of the kidney causes large differences in the blood-oxygen saturation within the inner compartments, which can be assessed using BOLD-contrast imaging and the time course of the changes caused by pharmacological tests. BOLD-contrast imaging apparently can depict tissue at risk from ischemia by providing information related to the tubular workload of the kidney. The ADC measurements are not sensitive to the current energy metabolism or the tubular flow or reabsorption, but rather seem to reflect histopathological changes within the tissue. However, the physiological and pathological significance of changes in the diffusion coefficient values still have to be assessed.