Novel methods for microCT-based analyses of vasculature in the renal cortex reveal a loss of perfusable arterioles and glomeruli in eNOS-/- mice

Prior to vascular perfusion and sacrifice, the WT and eNOS-/- mice were in generally good health with no obvious problems. Upon visual inspection of the microCT images, it was evident that the selected resolution (a combination of voxel size and signal-to-noise ratio) was not sufficient to accurately segment individual capillaries, but could easily detect glomeruli as their interwoven capillary networks provided greater signal and reduced volume averaging. This result was expected since individual capillaries are only 2–3 times larger than the 5 μm isotropic voxel size used for these studies. Hence, very few voxels encapsulate only capillary volume, while most voxels contain an average of contrast-filled capillary space and non-contrasted tissue around capillaries. While the glomeruli have an apparent thickness that is substantially greater than the afferent and efferent arterioles, their thickness is similar to other, slightly larger vessels in the renal cortex. The overlapping thickness ranges of these structures made it impossible to isolate glomeruli exclusively on the basis of thickness. In contrast, precapillary vessels on the order of >20 μm thick were visible and easily segmented and analyzed on a consistent basis as determined by thickness labeling of the segmented vasculature and the corresponding histograms (Fig. 2).

Decreased perfusion of contrast into eNOS-/- kidneys compared to those from WT mice was clear upon visual inspection of the 3D reconstructions, which was confirmed by quantification of vascular volume and structure of the whole kidney (Figs. 3a-d). Total kidney volume and total vascular volume were not significantly different between the genotypes (Figs. 3e and f). However, vascular volume calculated as a fraction of the total kidney volume (vascular volume/kidney volume; VV/KV) was significantly lower in eNOS-/- mice compared to WT (Fig. 3g), as was vascular density, represented as the number of vessels/mm (Vessel Number; V.N) (Fig. 3h). To determine which types of vessels were lost and/or obstructed in the eNOS-/- kidneys, vessel thickness in the whole kidney was analyzed by plotting the distribution of vessel abundance at each measured thickness and examining differences between the genotypes at individual thicknesses (Fig. 3i). eNOS-/- kidneys had significantly fewer perfusable vessels that those from WT mice primarily in the range of 20–40 μm in thickness, although significant differences were also found with a few higher thickness measurements (Fig. 3i).

As it relates to vasculature and perfusion, kidney function is primarily determined by factors in the renal cortex where resistance vessels, arterioles, and glomeruli control filtration and reabsorption. Hence, we sought to determine whether the renal vascular phenotype of eNOS-/- mice was more pronounced when the analysis was restricted to the renal cortex. Through visual inspection and measurement of mid-coronal microCT images from multiple kidneys, we determined that the boundary between the renal cortex and medulla was best approximated as 600 μm deep from the kidney capsule and created a volume of interest for each kidney comprising only this volume (Fig. 1).

Similar to the whole kidney analysis, three-dimensional renderings of vessels in the renal cortex show a clear decrease in the number and total volume of perfused cortical blood vessels in eNOS-/- kidneys (Fig. 4a-d). Total cortical volume was not significantly different between genotypes (Fig. 4e). However, the cortices of eNOS-/- kidneys had significantly less vascular volume, vascular volume/cortical volume, and vascular number (all p < 0.05), with a corresponding increase in the distance between cortical vessels (V.Sp; p < 0.05) (Figs. 4f-i). Vessel thickness was again examined as a distribution of the total volume of vessels at each measured thickness. The total volume of vessels in the range of 20–30 μm was significantly lower in the renal cortices of eNOS-/- mice compared to WT (Fig. 4j).

Isolation of glomeruli from other vascular structures in the cortex produced a number of perfused glomeruli in wild type mice (5290 ± 1758) that was approximately half of the total glomeruli in WT mice obtained using a stereology method (12,283 ± 663) [17]. Comparison between the genotypes revealed a striking difference in the number of perfused glomeruli between WT and eNOS-/- mice (Fig. 5a and b). Quantification confirmed that there were significantly fewer perfused glomeruli in the cortex of the eNOS-/- mice (Fig. 5c).

Precise analysis of renal vascular structure and perfusion is critical to understand the mechanisms of and test potential interventions for numerous renal diseases in animal models. While contrast enhanced microCT imaging and analysis of the renal vasculature has been published previously [1‐4], the current data presents an improved analysis capable of detecting changes in arteriole, prearteriole, and glomerular perfusion in the renal cortex. The ability of this approach to focus the analysis on cortical vessels within a specified range(s) of thickness appears to greatly improve the sensitivity of the technique to detect the phenotype of kidneys from eNOS-/- mice. The exclusion of the renal medulla from the volume of interest prevents artifacts created by incomplete filling of the veins which may influence or skew the analysis. Furthermore, this novel method for extracting the perfused glomeruli from the cortical vasculature provides a relatively rapid and cost effective means to detect phenotypic or treatment induced changes in total glomerular number that is not subject to sampling errors in histological assessments.

Applying the vascular thickness distribution to the unique vascular structure of the renal cortex may also provide additional information about the nature of perfusion or structural changes. In the renal cortex, collecting veins are the thickest structures (>100 μm), arterioles are 15–30 μm thick, prearterioles and glomeruli (which appear as solid spheres here) have overlapping thickness in the range of ~30–90 μm, while isolated capillaries and the fine venous structures surrounding collecting ducts are <15 μm. As applied to imaging, the Nyquist theorem states that the smallest structure that can be accurately resolved is in a noise-free image must be at least twice the thickness of the nominal resolution (voxels size). Therefore, at the resolution used here (5 μm voxels; ~12.5 μm resolution) isolated capillaries and fine veins are below the limit of reproducible detection. By excluding structures <15 μm or >100 μm thick, then examining differences in the remaining cortical vascular volume or abundance within a given thickness range, one can surmise whether the observed differences occurred primarily at the arteriolar, prearteriolar, or glomerular levels. Hence, the data presented here demonstrating decreased cortical vascular volume in vessels with thickness 20–40 μm (Fig. 4) suggests a decrease in perfusable renal arterioles in eNOS-/- mice.

This analysis also measured vessel separation in the cortex, a measurement of the mimimum distance between vessels and found a significant increase in the cortex of eNOS-/- mice compared to WT. While this measurement is most commonly applied to the analysis of trabecular bone structure in the literature [18], here it provides insight into the amount of tissue supplied per vessel. As vessel separation increases, the amount of tissue that relies on each vessel increases. Hence, V.Sp may prove useful in detecting areas of potential hypoxia or nutrient depravation in some disease models.

These data clearly demonstrated that microfil-perfusable arterioles and glomeruli are significantly decreased in eNOS -/- mice. Further histological investigation is required to determine if this is due to a reduction in glomerular number or because of vascular restrictions that limit microfil filling of the vessels. While such investigations are beyond the scope of this methodological paper, the fact that eNOS -/- mice at this age show normal serum creatinine levels, kidney weight, and urine albumin excretion [14], suggest that the observed microCT changes may be due to poor perfusion caused by the renal lesions, including focal glomerulosclerosis and renal scaring reported in these mice [13, 14].

Contrast-enhanced microCT analysis of renal vasculature also has a number of limitations which should be carefully considered and controlled, when possible. Since this technique is dependent on reproducible perfusion, extensive practice and confirmation of reproducibility is required before using this to compare phenotypes. Our experience suggests that adequate precision of vascular measurements in wild-type animals can be achieved, although the accuracy of the three-dimensional reconstructions is incomplete, as demonstrated by “breaks” often seen in smaller vessels in the microCT images (data not shown). The acquisition settings and 5 μm voxels used here were sufficient to consistently segment perfused vessels as small as 20 μm in thickness, but the efficiency of detection decreased beginning at about 50 μm thickness (Figs. 3 and 4).

The mean glomerular number in WT mice using the methods presented here are approximately one-half of what is reported with stereological methods. This underdetection is likely due to a combination of incomplete filling and artifacts of the methods used to select the glomeruli including the threshold, noise filter, and volume filtering steps. However, it is important to recognize that these methods are not intended to provide a substitute for stereological quantification of the exact number of glomeruli, but rather a less laborious method to screen for between-group changes in the number of glomeruli that can be perfused with microfil or any other radiopaque contrast agent. Futures studies will benefit from the development and use of contrast agents with lower, more physiological, viscosity.

While new analytical methods are presented here, others have previously employed contrast enhanced microCT and MRI for renal vascular characterization. Beeman and colleagues recently described the use of cationized ferritin as a contrast agent for MRI based quantification of glomerular volume and number in excised human kidneys [19]. This technique was capable of quantifying glomeruli with an approximate 6 % false positive rate. However, the technique required >10 hours of images acquisition time per kidney in a 7 T MRI to achieve the 117 μm isotropic voxel size needed to accurately identify the glomeruli. This technique was previously developed in rats using 9.4 T field and reduced FOV, which still required almost 5 h of acquisition time [20]. Hence, while this approach holds promise for future development, it is currently not feasible for either human or animal studies.

The use of ex vivo contrast enhanced microCT to image renal vasculature has been reported previously [1‐3, 5‐7, 21]. This technique was first described by Garcia-Sanz in a paper that compared the use of synchrotron and conventional x-ray sources to quantify vascular volume in the renal cortex and regions of the medulla [2]. Their paper reported that while synchrotron-based images were superior in quality microCT images produced from the conventional, polychromatic, x-ray source were of sufficient quality to measure vascular volume in the various anatomic regions and were consistent with histological measurements reported elsewhere [2]. At that time, only the synchrotron microCT was capable of generating images with voxels as small as 6 μm and hence the polychromatic microCT images could not accurately visualize glomeruli. Toyota et al, was the first to report quantification of glomerular volume in 2004. Again, they used synchrotron-based microCT to generate images with 5.6 or 9.6 μm voxels and manually selected 400 glomeruli per kidney to analyze the mean glomerular volume in the 9.6 μm images [1]. While this was a substantial advance, the use of a synchrotron radiation source and time required to manually identify a subset of glomeruli likely prevented additional use of the technique. Another laboratory has reported the use of polychromatic microCT in multiple papers [5‐7], however, their analyses were performed in only 12 equidistant two-dimensional slices of the microCT scans rather than in the 3-dimensional renderings. While this was likely due to limited computing power, it nevertheless, reduces the accuracy and precision of the approach. Hence, while laying the groundwork for the current study, previous studies employing contrast enhanced microCT to analyze renal vasculature structures have been limited in scope due to availability of affordable high resolution polychromatic systems, computing power, and analytical software capable of handling these extremely large data sets.

The application of these methods to in vivo images presents an opportunity for future refinement and application that could potentially capture changes in renal blood flow in response to acute stimuli. However, this technique requires the acquisition of images with a nominal isotropic voxel size of approximately 10 μm or less. Given the time, and more importantly, the radiation dose required to achieve this resolution in vivo, repeated in vivo measurements would be challenging with currently available technology. However, since current small animal scanners are marginally capable of performing such scans, this will likely be possible with the next generation of in vivo microCT technology.

Thanks to advancements in engineering and computer technology, the work presented here represents, to our knowledge, the most rigorous and widely accessible method for complete volumetric analysis of murine renal vasculature published to date. This approach holds the potential to substantially improve renal phenotyping and disease characterization in murine models which may improve understanding of numerous disease models, biological mechanisms, and treatments. The wide availability of affordable polychromatic microCT systems comparable to that used here and powerful parallel processing computer clusters should enable widespread utilization of these techniques.