Targeted delivery of solutes and oxygen in the renal medulla: role of microvessel architecture

Abstract

Renal medullary function is characterized by corticopapillary concentration gradients of various molecules. One example is the generally decreasing axial gradient in oxygen tension (Po₂). Another example, found in animals in the antidiuretic state, is a generally increasing axial solute gradient, consisting mostly of NaCl and urea. This osmolality gradient, which plays a principal role in the urine concentrating mechanism, is generally considered to involve countercurrent multiplication and countercurrent exchange, although the underlying mechanism is not fully understood. Radial oxygen and solute gradients in the transverse dimension of the medullary parenchyma have been hypothesized to occur, although strong experimental evidence in support of these gradients remains lacking. This review considers anatomic features of the renal medulla that may impact the formation and maintenance of oxygen and solute gradients. A better understanding of medullary architecture is essential for more clearly defining the compartment-to-compartment flows taken by fluid and molecules that are important in producing axial and radial gradients. Preferential interactions between nephron and vascular segments provide clues as to how tubular and interstitial oxygen flows contribute to safeguarding active transport pathways in renal function in health and disease.

Architecture of Medullary Microcirculatory Pathways

pioneering studies by moffat and Fourman (42) and Rollhauser et al. (59) clearly illustrated that the renal medulla is perfused primarily by 1) descending and ascending vasa recta positioned within vascular bundle regions that carry plasma in a papillary or cortical direction, respectively, and by 2) venous capillary networks that carry plasma chiefly in a cortical direction (Fig. 1). Their studies showed that venous capillary networks of the inner stripe of the outer medulla are positioned within interbundle regions, spatially separate from the vascular bundles, and that these networks are continuous, to some degree, with inner medullary capillary networks. Outer medullary capillary networks are associated with thick ascending limbs, descending limbs of long loops, and collecting ducts, whereas inner medullary capillary networks are associated primarily with collecting ducts and thin ascending limbs. Our understanding of vascular anatomy has been further advanced by others (3, 4, 36, 37, 63).

Fig. 1.

Regional distribution of nephrons and blood vessels of the rat inner stripe of the outer medulla and inner medulla. A: blood vessels at outer medullary-inner medullary border of India ink-injected kidney, formalin-fixed, embedded in celloidin, sectioned, cleared, and mounted; modified with permission from Ref. 42. Vascular bundles are denoted by arrowheads (see text for further details). Capillary networks of the outer medullary interbundle regions (*) and inner medullary intracluster regions (**) are shown. B: schematic diagram of nephrons and blood vessels in medullary regions; modified with permission from Ref. 52. Outer medullary interbundle regions are occupied by collecting ducts (CDs), thick ascending limbs, descending thin limbs of long-loop nephrons, and networks of fenestrated interconnecting capillaries (ascending vasa recta) (2, 27). The inner medullary intracluster regions are occupied by collecting ducts, networks of fenestrated interconnecting capillaries (ascending vasa recta), ascending thin limbs, and to a lesser extent aquaporin-1-null descending thin limbs (19, 55, 74). Unbranched ascending vasa recta residing in the medullary interbundle and intracluster regions (27) are not shown in B.

The blood vessels that perfuse the rat renal medulla arise from efferent arterioles of midcortical and juxtamedullary glomeruli. After entering the outer stripe, each efferent arteriole continues as a descending vas rectum, with each vas rectum breaking up into as many as 30 descending vasa recta (DVR) (22). The deeper the originating glomerulus lies below the cortical surface, the deeper its descendant DVR extend into the medulla (14). As the DVR descend through the outer stripe and into the inner stripe of the outer medulla, a number of these DVR assemble to form vascular bundles that include AVR arising from deeper levels. These tightly organized bundles exist in the inner stripe of the outer medulla and the initial third of the inner medulla. At deeper inner medullary levels, these vessels fan out from each other, becoming diffusely distributed throughout the papillary interstitium (58, 74). Vessels within vascular bundles of rat and mouse undergo little or no branching (58, 74).

Within the outer medullary bundles and alongside the DVR and AVR lie descending limbs of short-loop nephrons, positioned either at the bundle periphery (rat) or positioned more centrally within the bundle (mouse; Psammomys) (2). The outer medullary vascular bundles are arranged with the shortest DVR (those breaking up into capillaries in the outer medulla) positioned at the periphery of the bundle (Fig. 1). Long DVR that terminate in the inner medulla lie near to the central core of the bundle (2). This separation of short and long DVR within each bundle indicates that preferential radial flows potentially occur even within vascular bundles; consequently, vascular shunts may lead to targeted delivery of plasma and red blood cells, which could significantly impact medullary oxygenation (see below) (11).

The longest DVR remain within the bundle at the outer medullary-inner medullary border. These long DVR continue to descend through the upper inner medulla within vascular bundles that become more loosely organized with increasing depth. The DVR and adjacent aquaporin-1 (AQP1)-positive descending thin limbs within the bundles are positioned in a way that would facilitate entry of reabsorbed water into the unbranched AVR and at the same time engage in countercurrent exchange of urea (see below). Short terminal portions of inner medullary DVR with continuous endothelia or that express the urea transporter UTB commonly become fenestrated and continue to descend before the vessel terminates in the venous capillary network, or before the vessel begins to ascend, as shown for the rat by ultrastructural studies (22, 63), functional studies (78), and by coexpression of UTB and plasmalemmal vesicle protein 1 (PV1) (55, 74). The overlap of UTB and PV-1 expression suggests that the fenestrations in the terminal portion of the DVR within the upper inner medulla may be insufficient for urea transport across the capillary wall without the presence of UTB (55).

As the DVR descend, they terminate at all levels of the outer and inner medulla, leading to a declining total DVR population at progressively deeper levels below the corticomedullary border. Each DVR typically terminates as a single junction with one fenestrated capillary (or less commonly, as a bilateral junction with several fenestrated capillaries) that continues in an ascending direction and undergoes extensive branching (23, 55, 74), so that DVR, collectively, feed into a complex network of capillaries (Fig. 1) (42, 59). Outer medullary capillary networks arise chiefly from DVR terminating in the outer medulla, and inner medullary capillary networks arise from DVR terminating in the inner medulla. The capillary network of the inner stripe is denser than the capillary networks of the outer stripe and inner medulla (Fig. 1).

In general, inner medullary ascending (venous) capillaries of inner medullary networks ascend some distance, then coalesce at regular intervals along the inner medullary axis to form AVR; however, capillaries of inner medullary intracluster networks lying near the outer medullary-inner medullary border ascend directly into the inner stripe, where they merge with the capillary network of the interbundle region. Consequently, only after passing from the inner medullary intracluster region directly into the outer medullary interbundle region do these capillaries coalesce to form AVR (Fig. 1) (23, 28, 42, 59). These capillaries may represent deeper extensions of interbundle capillary networks previously reported to lie in the inner part of the inner stripe and that link to AVR within the vascular bundles of the innermost outer medulla (27, 72). Capillary networks of the inner part of the inner stripe are thought to be spatially distinct from the capillary networks of the outer part of the inner stripe and the outer stripe, the latter ascending directly among the tubules in the interbundle region and linking to arcuate or interlobular veins (27, 72). The two capillary drainage patterns in the outer medulla suggest two defined regional blood flow patterns may exist (Fig. 1) (27, 72).

Specialized Compartments in the Inner Medulla

In the upper inner medulla of rodent kidneys such as rat and kangaroo rat, about three to five interconnecting fenestrated capillaries abut each collecting duct along its length, forming interstitial channels along the collecting duct axis. Ascending thin limbs combine with these segments to form well-defined interstitial compartments within the intracluster region, the so-called “interstitial nodal spaces” (21, 55, 56). Based on the preferential arrangements of collecting ducts (which reabsorb urea and water) and ascending thin limbs and prebend segments (which reabsorb NaCl but not water), we hypothesized that interstitial nodal spaces (and the intracluster region more generally) might be specialized sites for NaCl and urea mixing in the antidiuretic rat (30, 32, 53, 56).

A mathematical model of the inner medullary urine concentrating mechanism has suggested that NaCl and urea could be sequestered within the intracluster region (30). However, others have argued that solute diffusion between the intracluster and intercluster tubules and vessels may be sufficiently large to prevent establishment of significant lateral NaCl and urea concentration gradients (69). Because the interstitial and luminal O₂ concentrations are orders of magnitude lower than NaCl and urea, and because collecting ducts may have a high O₂ consumption rate, a substantial lateral oxygen tension (Po₂) gradient may be generated, such that Po₂ is significantly higher in the intercluster regions than in the intracluster regions. By sequestering O₂ within the intercluster regions, O₂ delivery to the deep papilla may be preserved.

Countercurrent Exchange Pathways

Examples of countercurrent exchange can be found throughout the medulla: NaCl and urea are generally shunted from AVR to DVR in the process of countercurrent exchange, whereas O₂ is shunted from DVR to AVR. These exchanges delay or prevent loss of solutes from the inner medulla and delay or reduce O₂ delivery to the inner medulla. A number of spatial relationships have long been considered to underlie preferential urea countercurrent flows between tubular structures of the outer and inner medulla based on detailed descriptions in the following references (16, 22, 25, 31, 33, 37, 45, 49, 52, 56, 60, 73). NaCl and O₂ gradients would also lead to spatially dependent NaCl and O₂ countercurrent flows.

Throughout the outer medulla and most of the first half of the inner medulla, all DVR express detectable levels of the urea transporter protein UT-B as determined with immunohistochemistry; however, expression levels are significantly reduced at deeper levels of the inner medulla (23, 24, 54, 74). High transendothelial urea permeability in outer medullary DVR and low transendothelial urea permeability in the papillary DVR of the rat (47, 50, 51) correlate with these UTB expression patterns, providing support for UTB as the predominant pathway for urea flux in DVR. Alternate pathways for urea flux in DVR include the paracellular pathway and/or fenestrations, and these are also likely pathways for NaCl flux (47, 50, 51).

Transendothelial water flux in DVR occurs through the water channel AQP1, which is coexpressed with UT-B within most but not all of these segments (55). Water flux occurs through at least one additional pathway, which has not been identified (48). Water reabsorbed from the outer medullary DVR flows into AVR, thereby providing a shunt pathway for volume efflux, a process that potentially concentrates both plasma solutes and blood cells in DVR and reduces the fluid load that enters the inner medulla. All medullary AVR and capillaries are fenestrated and express PV-1, a protein associated with the fenestral diaphragm (38, 54, 64). Most vessels (both descending and ascending) in the deep papilla (∼2 mm above the papilla tip) are fenestrated (54, 57), with the fenestration fraction increasing with medullary depth (38), although vessels with continuous endothelium do exist in the terminal papilla (63). Fenestrations impart a very high transendothelial permeability to water and small solutes, but not to high-molecular-weight solutes and proteins. In particular in the deep papilla, fenestrations may facilitate the diffusion of oxygen from the DVR plasma into the interstitium.

Throughout the medulla, the DVR and AVR lying within the bundle region (in the outer medulla) or the intercluster region (in the inner medulla) appear to be arranged sufficiently close together to promote effective solute and O₂ countercurrent exchange between them (Fig. 1). UTB-mediated countercurrent exchange of urea between descending and ascending vessels lying near to each other should be considered a feature primarily of the bundle zone in the outer medulla and the intercluster region through approximately the upper ∼50% of the inner medulla (7, 38, 55, 63, 74). Any significant degree of solute recycling between descending and ascending vessels within ∼2 mm above the papilla tip likely involves primarily fenestrated vessels. The disproportionately large fraction of fenestrated descending and ascending vessels in the deep papilla combined with unhindered Na and urea diffusion would account for the nearly equivalent Na and urea transendothelial permeabilities (studied with in vivo microinfusion), reported by Pallone et al. (49, 51) for DVR and AVR (vessels distinguished from each other by the direction of plasma flow) at the papillary tip of the adult female Munich-Wistar rat.

The proximity of AVR to the descending limbs of short-loop nephrons in the outer medullary vascular bundles suggests that countercurrent solute exchange could occur between these segments (58). A similar configuration is found in the upper inner medulla, where the vasa recta and the descending limbs are found in the intercluster region. Despite their proximity, water exchange may be limited: the descending segments of short-loop nephrons of rats, mice, and humans express little or no AQP1 (68, 76); similarly, in most of the terminal inner medullary descending limb segments of the rat and chinchilla, AQP1 protein expression is relatively low (Fig. 1) and transepithelial osmotic water permeability is nearly zero (12, 13, 45, 54).

In rats and mice, the terminal inner stripe segment of short descending limbs, and the initial inner medullary segments of some long descending limbs, express the urea transporter UT-A2 (68). The traditional assumption is that some of the urea ascending from the inner medulla via the AVR is secreted into the UT-A2-positive segment of the short descending limb. If true, deletion of UT-A2 should induce a significant urine concentrating defect. However, that concentrating defect was not observed in UT-A2 knockout mice (67) nor in UT-A2/UT-B double knockout mice (35). What is particularly surprising is that, instead of further aggravating the urine concentrating mechanism, the abolition of UT-A2 permeability in the UT-A2/UT-B double knockout mice restored a nearly normal urine osmolality, which is severely impaired in the UT-B knockout mice (35). An alternative hypothesis is that urea is reabsorbed from, rather than secreted into the UT-A2 segment, a scenario that is likely if urea is actively secreted into the pars recta (73). If so, UT-A2 may play a role in the transient build-up of a urea and osmolality gradient in the inner medulla, rather than the generation of steady-state gradients. This hypothesis, in brief, is based on the observation that for the urea in the short descending limb to return to the inner medulla, the transit time required for the UT-A2-mediated reabsorption and the course through the DVR is shorter than the transit time through the thick ascending limb and collecting duct (for a more complete discussion, see the study by Layton and Bankir) (29).

Total vascular outflow from the rat inner medulla exceeds vascular inflow by ∼30%, the excess outflow representing fluid taken up from descending tubular segments (78). Despite the excess fluid, AVR flow is likely slower than DVR flow, because AVR exceed DVR by about twofold in the rat inner medulla, as determined by functional analyses (5, 39, 78) and by reconstruction of unbranched vessels lying in the bundle region (74). A slower AVR flow would provide a longer time for AVR to equilibrate with local interstitial fluid. As a result, in a concentrating kidney, AVR fluid lags local interstitial fluid osmolality by only a small amount, thereby minimizing any diluting effect and improving countercurrent exchanger efficiency (39). In contrast, there is an AVR/DVR ratio of about four for the anatomic ratio of all fenestrated and nonfenestrated vessels (determined at 2 mm above the papilla tip) in the Munich-Wistar rat (20, 74). The additional fenestrated vessels include interconnecting capillaries in addition to the AVR determined by functional studies. Interconnecting capillaries lie distant from DVR and are unlikely to directly participate in countercurrent exchange with DVR. However, the interconnecting capillaries, which run vertically and laterally, form the interstitial nodal spaces along with collecting ducts and ascending thin limbs and carry absorbate from these segments into the nonbranching AVR in bundle regions. The degree to which interconnecting capillaries join AVR and contribute to their flow has not been quantified for the rat medulla but has been shown to be minimal in the mouse (58).

Medullary Metabolic Activity and Oxygenation

In the mammalian kidney, despite high blood flow and oxygen delivery, Po₂ is relatively low, especially in the renal medulla, which has been measured to be ∼20 and 10 mmHg in the outer and inner medulla, respectively (46). That marked discrepancy between blood supply and oxygenation can be attributed, in large part, to the kidney's high oxygen consumption per tissue weight. The Na-K-ATPase (the sodium pump), the protein complex that drives Na reabsorption, singularly consumes the greatest proportion of energy in the kidney. Na reabsorptive processes expend ∼90% of total renal oxygen; ∼30% of renal Na transport and 30% of renal oxygen consumption are consumed in the outer medulla (41). Renal ATP production occurs chiefly through aerobic metabolism; however, under anaerobic conditions such as has been predicted for the inner medulla, metabolism could consist substantially of anaerobic glycolysis (1, 34). Metabolic activity may be rate limiting for transepithelial solute transport, and this transport itself may, in turn, regulate metabolic activity. The processes by which oxygen is delivered to renal transport epithelia may increase oxygen demand, the direct implication being that renal metabolic autoregulation itself creates challenges with balancing hypoxia and hyperoxia. A host of control mechanisms and signaling pathways are involved in maintaining this balance (44). Due to the complexities involved, systems analyses of blood flow pathways and mathematical models of the multiple functional inputs to these processes are useful for understanding oxygen requirements and metabolism associated with fluid and solute transport in the renal medulla.

Because of the significant distance between the O₂-carrying outer medullary DVR, which lie in the bundle regions, and the thick ascending limbs, which lie in interbundle regions and because of the thick limbs' high metabolic requirements, Po₂ in the thick ascending limbs near the outer-inner medullary boundary is very low (<10 mmHg). This contrasts with Po₂ measured in renal veins (50 mmHg), efferent arterioles (45 mmHg), and proximal and distal tubules (40 mmHg) (71). Hypoxia-induced necrosis of cells lining thick limbs and long-loop descending limbs (but not short-loop descending limbs) gave functional evidence for corticopapillary and transverse Po₂ gradients in the rat outer medulla (6, 26, 62). Thick ascending limbs at the periphery of vascular bundles escape necrosis, and, at least in the mouse kidney, these are predominantly long-loop thick ascending limbs (58). Po₂ gradients along the corticopapillary axis and projecting radially from the vascular bundles are predicted by a mathematical model of oxygen transport that was based on the complex structural organization of the outer medulla (10, 11). The low Po₂ surrounding the thick limbs may be partially alleviated by the Bohr effect, which states that a lower pH reduces the binding affinity of hemoglobin for O₂. Burke et al. (8) observed an increase in pH, from 7.20 to 7.31, after addition of furosemide, a finding that suggests that the interstitial fluid surrounding the thick ascending limbs may have a lower pH than that in the vascular bundles. The presumed greater acidity of blood in the interbundle region where the thick limbs are located, relative to that in the vascular bundles, may facilitate the release of O₂ (9).

Because ∼80% of the DVR do not pass into the inner medulla, ∼80% of the O₂ supplied to the outer medulla does not perfuse the inner medulla. Mathematical models suggest that, owing to the separation of the DVR from the thick ascending limbs, most (>90%) of the O₂ supplied to those deep-reaching DVR at the corticomedullary junction reaches the inner medulla (9, 77). The functional role of the vascular bundles of preserving oxygen delivery to the deep inner medulla is underscored in a recent modeling study (17). The authors conducted a simulation of a hypothetical configuration in which nephrons and vessels are more homogeneously distributed. In particular, the separation between the DVR and thick limbs are much reduced. That simulation predicted a drastic decrease in oxygen delivery to the inner medulla, with the terminal inner medulla almost deprived of oxygen (17).

The vascular structural organization of the upper inner medulla has an analogous impact on oxygenation of that region. While the thin limbs do not have significant active transport (40, 43), functional and modeling studies support a brisk active NaCl reabsorption in the inner medullary collecting duct, (18, 70). A radial oxygen gradient across the inner medullary interstitium has been hypothesized on the basis of permeability and flux values estimated for transport in the inner medullary collecting duct (69). A substantial radial oxygen gradient has been predicted in a mathematical model of medullary oxygen transport (17).

The ratio of Na transport and O₂ consumption is relatively low in the medullary collecting duct compared with other nephron segments. Together with the low medullary Po₂, anaerobic glycolysis may be important in the collecting duct, as suggested by findings in several studies. When aerobic ATP production is directly inhibited, outer medullary collecting duct cells in the mouse maintain ATP levels that are ∼80% of the control (i.e., the case when the cells are not inhibited) (66), and when Na-K-ATPase is directly inhibited in the rabbit, O₂ consumption levels are maintained nearly equal to noninhibited levels of 30–40% (75). In the rat inner medullary collecting duct, cellular ATP levels have been found to reach 40% of controls when oxygen metabolism is inhibited with rotenone (65).

Conclusions and Future Directions

The architecture of the medullary circulatory system is commonly interpreted in terms of how it aids in preserving the corticopapillary solute gradient. This design conflicts with the ability of capillaries to provide oxygen to medullary structures at concentrations that equal those known to exist in the renal cortex. Improved understanding of medullary blood flow will aid in better understanding disorders associated with fluid and solute imbalances. Integrative studies on regional perfusion in vivo are essential for understanding inner medullary blood flow and its regulation, and the impact on various diseases and disorders.

We conclude with an example that illustrates the synergy between model simulations and experimental studies. A recent modeling study by Edwards et al. (15) considers the effect of Na-K-2Cl cotransporter (NKCC2) isoform regulation on NaCl transport in the thick ascending limb and macula densa. Their results suggest that modulation of differential splicing of NKCC2 by a low-salt diet, which induces a shift of NKCC2-A to the B isoform (61) primarily in the thick ascending limb and macula densa cells, significantly enhances NaCl reabsorption along the thick ascending limb. Simulation results also suggest that the isoform shift hyperpolarizes the macula densa basolateral cell membrane, which, taken in isolation, may inhibit the tubuloglomerular feedback signal. However, excessive early distal salt delivery and thus renal salt loss during a low-salt diet may be prevented by an asymmetric tubuloglomerular feedback response, which may be more sensitive to flow increases. This is but one example where mathematical models have provided insights into the mechanism behind disorders and generated ideas for improving outcomes of procedures. A multidisciplinary approach that combines experimental and modeling techniques may have the best chance at attaining a true understanding of renal function and metabolism under physiological and pathophysiological conditions (e.g., hypertension and diabetes).

GRANTS

Research in the authors' laboratories has been supported by the National Science Foundation, IOS-0952885 (T. Pannabecker) and DMS-0340654 (A. Layton), and by National Institute of Diabetes and Digestive and Kidney Diseases Grants DK-08333 (T. Pannabecker) and DK-89066 (A. Layton).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: T.L.P. and A.T.L. prepared figures; T.L.P. and A.T.L. drafted manuscript; T.L.P. and A.T.L. edited and revised manuscript; T.L.P. and A.T.L. approved final version of manuscript.