Live nephron imaging by MRI

Abstract

The local sensitivity of MRI can be improved with small MR detectors placed close to regions of interest. However, to maintain such sensitivity advantage, local detectors normally need to communicate with the external amplifier through cable connections, which prevent the use of local detectors as implantable devices. Recently, an integrated wireless amplifier was developed that can efficiently amplify and broadcast locally detected signals, so that the local sensitivity was enhanced without the need for cable connections. This integrated detector enabled the live imaging of individual glomeruli using negative contrast introduced by cationized ferritin, and the live imaging of renal tubules using positive contrast introduced by gadopentetate dimeglumine. Here, we utilized the high blood flow to image individual glomeruli as hyperintense regions without any contrast agent. These hyperintense regions were identified for pixels with signal intensities higher than the local average. Addition of Mn²⁺ allowed the simultaneous detection of both glomeruli and renal tubules: Mn²⁺ was primarily reabsorbed by renal tubules, which would be distinguished from glomeruli due to higher enhancement in T₁-weighted MRI. Dynamic studies of Mn²⁺ absorption confirmed the differential absorption affinity of glomeruli and renal tubules, potentially enabling the in vivo observation of nephron function.

there has been considerable interest in observing individual nephrons and studying their function (6, 13, 14). Current techniques for nephron counting involve histology or acid maceration, thus are impractical for longitudinal preclinical studies or eventual clinical use. Micropuncture can study glomerular filtration of individual nephrons (7), but this destructive technique is applicable only to nephrons that are very close to the cortical surface. On the other hand, MRI is a nondestructive technique that has proven invaluable for structural and functional nephrology (4, 8, 9, 19–22). Contrast enhancement, based on the delivery of cationized ferritin, has enabled nondestructive observation of individual glomeruli (1, 2, 3, 5, 10). However, these high-resolution studies were mostly performed on ex vivo kidneys. In vivo observation of glomeruli is more challenging (5, 18), because it requires sensitive detection of remote signals emitted from deep inside the body. Whereas smaller MRI detectors placed close to the tissue of interest can improve local sensitivity, the need for cables to transfer the locally detected signals is cumbersome and introduces a risk of infection. Recently, a wireless amplified NMR detector (WAND) (15) was developed to actively amplify emitted signals from small, implantable MR detectors (11, 17). Such enhanced sensitivity enabled the in vivo observation of individual glomeruli using transverse dephasing (T₂^*) contrast introduced by cationized ferritin, and renal tubules enhanced by gadopentetate dimeglumine (Gd-DTPA) in longitudinal relaxation (T₁) weighted images (16). In this work, we extended the contrast available for kidney anatomy and function. Glomeruli were observable in native kidney as hyperintense regions due to the high blood flow inside glomerulus arterioles. After MnCl₂ was infused intravenously, the majority of Mn²⁺ accumulated in renal tubules and enhanced tubular signals to a greater extent, while glomeruli appeared as less enhanced regions compared with their surrounding tissues. Dynamic studies of MnCl₂ absorption confirmed the differential affinity of glomeruli and renal tubules to Mn²⁺, paving way for in vivo studies of nephron function. The WAND technique can be used to noninvasively monitor nephron physiology in vivo and could potentially be used to study structure-function relationships and macromolecular filtration of individual nephrons. The ability to image glomeruli and renal tubules nondestructively may eventually enable the clinical use of the WAND as a chronic monitoring device for transplanted kidneys.

METHODS

The WAND is a double-frequency resonator (Fig. 1A) tuned to the first and second harmonics of Larmor frequency, with dimensions of 7 × 3.5 × 3.5 mm³ (Fig. 1B). A pumping signal applied near twice the Larmor frequency amplifies the MR signal through frequency mixing. For 1 mW of pumping power, the WAND has a gain of 23 dB with a 500-kHz bandwidth. This gain level is sufficient to maintain the local sensitivity of the WAND for a distance separation that is 10 times its own diameter. Compared with a local detector of similar dimension with direct cable connection, the WAND retains 70% sensitivity, which is much higher than the sensitivity achievable by an external coil or an internal detector with passive coupling (16).

Fig. 1.

A: circuit diagram of a wireless amplified NMR detector (WAND). This circuit is a double-frequency resonator with 2 inductors (L₁ and L₂) and two capacitors (C₁ and C₂). The lower resonance mode is tuned to receive weak MR signals, and the higher resonance mode is to receive the strong pumping signal at approximately twice the Larmor frequency. The capacitor C₂ is a nonlinear component whose capacitance is dependent on the applied voltage. The weak MR signal can exchange energy with the strong pumping signal through nonlinear frequency mixing. As a result, signal amplification can occur. B: photograph of the polydimethylsiloxane-coated resonator whose constituting components are labeled by symbols defined in A.

Animal experiments were approved by the National Institutes of Health Animal Care and Use Committee. Nine male Sprague-Dawley rats (300–350 g) were used. During surgery, rats were anesthetized with 2% isoflurane, with their kidneys exposed through an abdominal incision. The WAND was fixed onto the medial renal surface using 10% (wt/vol) type A gelatin and 1% (wt/vol) glutaraldehyde (Sigma-Aldrich). The kidney was returned to the abdominal cavity before the incision was closed. Antibiotic ointment was applied. Ketoprofen (5 mg/kg) was administered once a day during the 7-day recovery period.

During MRI sessions, rats were ventilated with 84% oxygen, 14% nitrogen, and 2% isoflurane. Animals were secured in the supine position with an abdominal restraint belt. A 22-mm-diameter surface coil placed externally beneath the left kidney was used for signal acquisition, and a coaxially placed 22-mm pumping coil pumped the internal WAND. To locate the WAND's position, a low-resolution multislice Fast Low Angle Shot (FLASH) image was acquired without pumping, using echo time (TE) = 6 ms, repetition time (TR) = 387.5 ms, flip angle (FA) = 30°, number of scans (NS) = 1, field of view (FOV) = 4 × 4 cm², 1-mm slice thickness, and 156 × 156-μm² in-plane resolution. High-resolution images were subsequently acquired with pumping, using TE = 3.5 ms, TR = 65 ms, FA = 35°, NS = 20, FOV = 0.9 × 0.9 cm², 0.2-mm slice thickness, and 70 × 70-μm² in-plane resolution. To saturate signals of inflowing blood, flow saturation was performed by a 90° saturation pulse applied on a 30-mm slice separated from the imaging plane by 30 mm on the proximal end. Two rats received 10 mg/ml cationized ferritin (Sigma-Aldrich) at 5 mg·kg⁻¹·min⁻¹ for 15 min and scanned with transverse dephasing (T₂^*)-weighted gradient refocused echo (GRE), using TE = 10 ms, TR = 65 ms, FA = 20°, NS = 20, FOV = 0.9 × 0.9 cm², 0.2-mm slice thickness, and 70 × 70-μm² in-plane resolution. Seven rats received 25 mM MnCl₂ at 1.25 μmol·kg⁻¹·min⁻¹ over 40 min and scanned with longitudinal relaxation (T₁)-weighted GRE, using TE = 3.5 ms, TR = 65 ms, FA = 75°, and other parameters remained the same as for T₂^*-weighted images.

All images were processed using Matlab (Mathworks). The local pixel average was evaluated within a contiguous region of 7 × 7 pixels, whose length was slightly larger than the average distance separation between individual glomeruli. The relative intensity was set as the ratio between the signal intensity and the local average. To segment contrast regions with 87% confidence, a local threshold for each pixel was chosen as 1.5 times the SD of relative intensities within this contiguous region, so that the threshold curve approximately passed through the half height positions of major peaks in relative intensity profiles. Significantly (or less) enhanced regions were identified for pixels with relative intensities greater (or smaller) than unity by their local thresholds, respectively. To evaluate the similarity between two images, a cross-correlation coefficient was calculated according to

$$r=\frac{{\displaystyle \sum _{m,n}}\left|\left(X_{m,n}-\overline{X}\right)\left(Y_{m,n}-\overline{Y}\right)\right|}{\sqrt{{\displaystyle \sum _{m,n}{\left(X_{m,n}-\overline{X}\right)}^2}}\sqrt{{\displaystyle \sum _{m,n}{\left(Y_{m,n}-\overline{Y}\right)}^2}}},$$

where X_m,n and Y_m,n are the pixel intensities for the (m, n) pixel in these two images, respectively, while X̄ and Ȳ are average intensities over the entire region of interest. Signal intensities in post-contrast images were normalized against the precontrast image before being averaged over the bright and less enhanced regions.

RESULTS

To locate the position of the WAND inside the body, a low-resolution FLASH image was acquired by the external coil without pumping (Fig. 2A). A black rectangle corresponds to the polydimethylsiloxane-coated WAND, and the higher signal intensity nearby is due to the detector's passive coupling to the external coil. When the gain was adjusted to 23 dB, the imaging sensitivity was enhanced by 21 dB in regions proximal to the detector. As a result, it was possible to obtain high-resolution images within a 9 × 9-mm² region defined by a dashed square. As shown in Fig. 2B, bright regions appear in a native kidney, which are particularly obvious in the magnified view (Fig. 2C1) within a 3.2 × 1.4-mm² rectangle defined by a green dashed box. The one-dimensional (1D) intensity profile along a red dashed line is plotted as a red curve in Fig. 2C2, with multiple peaks above the local average (blue curve). By dividing the red curve with the blue curve, the relative intensity profile is plotted as the red dashed curve in Fig. 2C3. The threshold curve (grey) for peak identification is above unity by 1.5 times the local SD of relative intensity values. According to the relative intensity profile in Fig. 2C3, these bright regions have an average distance separation of 0.35 ± 0.05 mm and average diameter of 0.12 ± 0.03 mm. The image in Fig. 2D1 was acquired in the presence of flow saturation. The disappearance of hyperintense regions in flow-saturated images indicates that bright regions in T₁-weighted images are due to blood flow into the detection slice.

Fig. 2.

Assignment of flow-enhanced regions in a native kidney as glomeruli. A: low-resolution Fast Low Angle Shot (FLASH) image acquired without wireless amplification when the pumping power is turned off. The acquisition parameters are echo time (TE) = 6 ms, repetition time (TR) = 387.5 ms, flip angle (FA) = 30°, NS = 1, field of view (FOV) = 4 × 4 cm², 1-mm slice thickness, and 156 × 156-μm² in-plane resolution. B: high-resolution image acquired in the presence of wireless amplification, with FOV defined by the white dashed box in A. The acquisition parameters are TE = 3.5 ms, TR = 65 ms, FA = 35°, number of scans (NS) = 20, FOV = 0.9 × 0.9 cm², 0.2-mm slice thickness, and 70 × 70-μm² in-plane resolution. C1: 3.2 × 1.4-mm² region of interest (ROI) defined by the green dashed box in B. C2: 1-dimensional (1D) intensity profile along the red dashed line in C1. The 1D intensity profile is shown in red, and the average intensity profile is shown in blue. The average intensity for each pixel was obtained by 2-dimensional averaging performed over a surrounding region with 7 × 7 pixels. C3: relative intensity profile (red dash) obtained by dividing the absolute intensity with the average intensity, as well as the threshold curve (gray) for positive contrast identification. This curve is above unity by 1.5 times the local SD of relative intensity values within the surrounding region. Peaks above the threshold curve are labeled by black dashed lines in C1. D1: same region of interest in another image acquired with the same parameters as in C1 but in the presence of flow saturation. Most bright dots disappear in D1, as confirmed by the 1D intensity profile in D2. E1: T₂^*-weighted image acquired after a bolus injection of cationized ferritin at 75 mg/kg, with TE = 10 ms, TR = 65 ms, FA = 20°, and other parameters the same as in C1. E2: 1D intensity profile (red) and the average intensity profile (blue). E3: relative intensity profile (red dash) and the threshold curve (gray) for negative contrast identification. This threshold curve is below unity by 1.5 times the local SD of relative intensity values within its surrounding region. F1, G1, and H1: images obtained from another rat, where F1 has the same parameters as in C1, G1 has the same parameters as in D1, and H1 has the same parameters as in E1. The 1D-intensity profiles in F2, G2, and H2 and the relative intensity profiles in F3, G3, and H3 were obtained in the same manner as their counterparts in C2, D2, and E2 and C3, D3, and E3, respectively.

To positively identify these bright regions as glomeruli, cationized ferritin was infused to detect individual glomeruli, as previously reported (16). Figure 2E1 shows the T₂^*-weighted image of the same region after administration of cationized ferritin. Due to the ferritin iron content, it will decrease glomerular signal intensity when selectively bound to glomerular basement membranes that are negatively charged (1, 2, 5, 10, 19). Figure 2E1 has multiple dark regions whose intensities fall below the average intensity curve in Fig. 2E2. In Fig. 2E3, these dark regions can be identified by dips in the relative intensity profile (red dash) that are below the local threshold (gray). The dark regions in Fig. 2E1 have a similar pattern as the bright regions in Fig. 2C1, with a cross-correlation coefficient of 0.93 ± 0.01. The strong colocalization of flow contrast in a native kidney and T₂^* contrast after ferritin injection is reproduced in Fig. 2, F1 and H1, in a different rat. These results indicate that individual glomeruli are observable as focal enhanced regions in T₁-weighted images due to blood flow.

In addition to glomerular blood flow and cationized ferritin, MnCl₂ was also used for contrast enhancement. To track MnCl₂ absorption, multiple T₁-weighted images with flow saturation were acquired at different time points. As shown in Fig. 3A1, the native kidney had little contrast when signals from inflowing blood were saturated. Halfway into the MnCl₂ infusion, bright ribbons became visible (Fig. 3A2). At the end of the 40-min infusion, these bright ribbons were most obvious, with an average width of 0.18 ± 0.03 mm and an average distance separation of 0.33 ± 0.05 mm (Fig. 3A3). Subsequently, lower intensity regions became visible (Fig. 3A4) and remained (Fig. 3, A5 and A6). Images in Fig. 3, A4–A6 were segmented for bright and lower intensity regions based on local thresholds described in methods. Figure 3A7 is the average segmentation mask for Fig. 3, A4–A6, where the green color corresponds to bright regions and the pink color corresponds to lower intensity regions. Signal intensities in post-contrast images (Fig. 3, A2–A6) were normalized against the pre-contrast image (Fig. 3A1). These normalized intensities were averaged over the green and pink masks before being plotted over the time course. As shown in Fig. 3B, during MnCl₂ infusion, signals in bright regions increased more quickly (green curve), while signals in lower intensity regions increased more slowly (pink curve). Both regions had maximum intensities immediately after infusion finished, followed by a slow decrease over time.

Fig. 3.

High-resolution Mn-enhanced MRI. The T₁-weighted images in A1–A6 were acquired with flow saturation before infusion, and at 21, 41, 56, 70, and 94 min after the start of infusion. The infusion was completed within 40 min. Detailed acquisition parameters are TE = 3.5 ms, TR = 65 ms, FA = 75°, NS = 20, FOV = 0.9 × 0.9 cm², 0.2-mm slice thickness, 70 × 70-μm² in-plane resolution, and RO = 3.2 × 1.4 mm². Images in A4–A6 are segmented for bright and less enhanced regions according to the threshold criteria described in methods. A7: average masks for images in A4–A6, where the green mask corresponds to bright regions and the pink mask corresponds to less enhanced regions. B: time series plot of the averaged normalized intensities in regions defined by the green and pink masks, where the error bars correspond to differences among 5 different rats.

The bright regions in Mn-enhanced images can be assigned to tubular mass (the sum of all tubules in the voxel), because Mn²⁺ is known to be effectively reabsorbed by epithelial cells of renal tubules (12). To assign the lower intensity regions, coregistration experiments were performed. The image in Fig. 4A1 was acquired without a contrast agent. It has multiple bright regions as a result of glomerular blood flow. According to the 1D-intensity profile (red curve) in Fig. 4A2, these bright regions have signal intensities that are significantly above their local average (blue curve). The gray curve in Fig. 4A3 is the threshold for peak identification, which is above unity by 1.5 times the local SD of the relative intensity values. After administration of MnCl₂, a T₁-weighted image was acquired with flow saturation (Fig. 4B1). The 1D-intensity profile along the red dashed line has multiple lower intensity valleys as well as higher intensity peaks (Fig. 4B2). These lower intensity regions along the 1D profile are identified by the gray threshold curve in Fig. 4B3. The lower intensity regions in Fig. 4B1 have a similar pattern as the bright regions in Fig. 4A1, with a correlation coefficient between images of 0.92 ± 0.01. This correlated pattern supports the assignment of lower intensity regions in Fig. 4B1 to glomeruli. Figure 4, D1 and E1, show results from another rat kidney. To confirm that the differential intensity of glomeruli in T₁-weighted images were not a result of transverse dephasing, T₂^*-weighted images were acquired (Fig. 4, C1 and F1). The only contrast detectable was due to dark blood vessels in Fig. 4C1 (labeled by green triangle). No contrast for glomeruli was detected, indicating that Mn contrast was due to T₁ weighting.

Fig. 4.

Correlated pattern between bright regions in a native kidney and less enhanced regions after MnCl₂ injection. A1: 3.2 × 1.4-mm² ROI in a high-resolution image acquired with the following parameters: TE = 3.5 ms, TR = 65 ms, FA = 35°, NS = 20, FOV = 0.9 × 0.9 cm², 0.2-mm slice thickness, and 70 × 70-μm² in-plane resolution. B1: same ROI in a T₁-weighted image acquired after a bolus injection of MnCl₂ at 50 μmol/kg body wt. This image was acquired in the presence of flow saturation, with FA = 75° and other parameters remain the same as in A1. C1: same region of interest in a T₂^*-weighted image acquired with TE = 10 ms, TR = 65 ms, FA = 20°, and other parameters the same as in A1. A2 and B2: 1D-intensity profiles (red) and the average intensity profiles (blue) along the dashed lines in A1 and B1, respectively. The average intensity value for each pixel was obtained by 2D averaging performed over a contiguous region with 7 × 7 pixels. A3 and B3: relative intensity profiles obtained by dividing the red curve with the blue curve in A2 and B2, respectively. In A3, the threshold curve (gray) for peak identification is above unity by 1.5 times the local SD of the relative intensity values within a surrounding region of 7 × 7 pixels; in B3, the threshold curve (gray) for dip identification is below unity by 1.5 times the local standard deviation defined in the same manner. The T₂^*-weighted image in C1 has few dark regions, except the blood vessels that also appear in B1. These vessels marked by green triangles in C2 are not labeled by black dashed lines in B1 as exogenous contrast induced by MnCl₂. D1, E1, and F1: images obtained from another rat, where D1 has the same parameters as in A1, E1 has the same parameters as in B1, and F1 has the same parameters as in C1. The 1D intensity profiles in D2, E2, and F2 and the relative intensity profiles in D3, E3, F3 were obtained in a similar manner as their counterparts in A2, B2, and C2 and A3, B3, and C3, respectively.

DISCUSSION

In this work, the WAND was used to detect individual glomeruli and tubular mass in vivo with MRI. This millimeter scale detector was surgically implanted on the medial surface of the kidney to observe the renal cortex deep lying inside the body. This is a difficult region to observe with traditional MR detectors. Image analysis was performed on highly sensitive signals within a 3.2 × 1.4-mm² region that is ∼0.4 mm away from the detector's surface. In non-contrast images, individual glomeruli were observable as positive contrast due to blood flow. After Mn²⁺ administration, glomeruli appeared as lower intensity regions compared with their surroundings. Based on previous histological studies with the radio isotope ⁵²Mn, manganese is expected to accumulate in the epithelial cells of renal tubules rather than in glomeruli (12). In MR images, the higher enhancement of regions is most likely due to epithelial increase in Mn²⁺. Because the 200-μm slice thickness is much larger than the distance separation between individual renal tubules, bright ribbons in MR images likely represent several overlying tubules that are collectively enhanced.

Both Mn²⁺ and Gd-DTPA are used clinically as T₁-enhancement contrast agents. In a previous study, Gd-DTPA was used in combination with the WAND to image rat kidneys at high resolution (16). Because Gd-DTPA could have strong T₂^* effects at increased concentrations, it was only used at 5 μmol/kg body wt to avoid significant signal loss after water reabsorption in renal tubules. Signal enhancement in tubular lumens was also observed when Gd-DTPA concentrated in lumens and enhanced luminal water signals. However, individual glomeruli were not detected in this previous study, probably because the Gd-DTPA concentration was too low to create sufficient contrast. In the current study, we used a higher dose of Mn²⁺ at 50 μmol/kg body wt and were able to see both tubules and glomeruli. Taken together, the contrast available from blood flow and Mn²⁺ can complement the information provided by Gd-DTPA alone.

In conclusion, the in vivo observation of nephrons has been made possible by the superior sensitivity of the WAND. Individual glomeruli can be positively visualized by blood flow and negatively visualized by ferritin and Mn²⁺. Renal tubules can be visualized by Mn²⁺ epithelial reabsorption, which is a mechanism different from Gd-DTPA filtration. Future work to quantify the time course of Mn²⁺ and Gd-DTPA enhancement should add a semiquantitative picture of glomerular filtration and tubular reabsorption. In addition to a 3D view of nephron morphology, this should enable a better understanding of changes in nephron structure and function with renal diseases. Although the WAND is an invasive device with limited field of view, its usage as an implantable device is justified when surgical treatment is required, and its limited field of view can be enlarged by multiple implanted detectors. It is expected that the WAND can potentially be used as a chronic monitoring device for the transplanted kidney in the clinic.

GRANTS

This study was supported in part by the intramural research program of the National Institute of Neurological Disorder and Stroke.

DISCLOSURES

K. Bennett owns Nanodiagnostics, LLC.

AUTHOR CONTRIBUTIONS

Author contributions: C.Q. and A.K. contributed conception and design of research; C.Q., S.D., and N.B. performed experiments; C.Q. analyzed data; C.Q., X.Y., and N.P. interpreted results of experiments; C.Q. prepared figures; C.Q. drafted manuscript; C.Q., R.A.S., K.M.B., and A.K. edited and revised manuscript; C.Q. and A.K. approved final version of manuscript.