The current view of diabetic kidney disease, based on meticulously acquired ultrastructural morphometry and the utility of measuring plasma creatinine and urinary albumin, has been almost entirely focused on the glomerulus. While clearly of great importance, changes in the glomerulus are not the major determinant of renal prognosis in diabetes and may not be the primary event in the development of diabetic kidney disease either. Indeed, advances in biomarker discovery and a greater appreciation of tubulointerstitial histopathology and the role of tubular hypoxia in the pathogenesis of chronic kidney disease have given us pause to reconsider the current “glomerulocentric” paradigm and focus attention on the proximal tubule that by virtue of the high energy requirements and reliance on aerobic metabolism render it particularly susceptible to the derangements of the diabetic state. Such findings raise important issues for therapeutic advances specifically targeting the pathophysiological perturbations that develop in this part of the nephron.
The description of diffuse and nodular glomerulosclerosis by Kimmelstiel and Wilson in 1936 (1) set investigation on a course that has since focused primarily on the glomerulus as a means of understanding the pathogenesis of diabetic kidney disease. Changes in glomerular structure such as mesangial expansion, reduction in capillary surface, and podocyte loss are undoubtedly major features of diabetic kidney disease that help differentiate it from other forms of glomerulonephritis. These findings are, however, juxtaposed with the more recent knowledge that some patients with advanced disease display neither substantial glomerular pathology nor proteinuria and that kidney function declines well before traditional indicators of kidney disease such as microalbuminuria or creatinine-based estimated glomerular filtration rate (eGFR) decline (2). In recognition of these findings, the term diabetic kidney disease rather than diabetic nephropathy is now commonly used. On the background of recent advances in the role of the proximal tubule as a prime mover in diabetic kidney pathology, this review highlights key recent developments. Published mostly in the general scientific and kidney-specific literature, these advances highlight the pivotal role this part of the nephron plays in the initiation, progression, staging, and therapeutic intervention in diabetic kidney disease. From a pathogenetic perspective, as illustrated in Fig. 1 and as elaborated on further in this review, tubular hypoxia as a consequence of increased energy demands and reduced perfusion combine with nonhypoxia-related forces to drive the development of tubular atrophy and interstitial fibrosis in a vicious cycle that promotes disease progression in diabetes. These insights offer new opportunities for therapeutic development.
Normal and Disordered Structure
Anatomically, the proximal tubule refers to that part of the nephron that is directly contiguous with the parietal epithelium of Bowman’s capsule. Measuring approximately 14 mm in length in humans, it consists of three subtly distinct segments. The S1 segment comprises the first two-thirds of the tubule’s early, convoluted component (pars convoluta); the S2 includes the final portion of the pars convoluta along with the initial, cortical part of its straight component (pars recta); and the S3 makes up the remainder of the pars recta as it dives deeply into the cortex and outer medulla (3).
Cells in the S1 segment are characterized by a tall apical brush border, prominent basolateral invaginations, extensive endocytic-lysosomal apparatus, and abundant, often elongated, mitochondria. In the S2, epithelial cells have shorter brush borders, less prominent basolateral invaginations, and smaller mitochondria, while in the S3 basolateral invaginations are absent and mitochondria are fewer (3) (Fig. 2).
The proximal tubule undergoes a range of structural changes in diabetes such as tubular atrophy, interstitial fibrosis, and peritubular capillary rarefaction, each of which correlate closely with declining kidney function (4). Additional dysfunction occurs when, as in cystinosis, atrophy occurs at the critical junction between Bowman’s capsule and the proximal tubule, giving rise to nonfunctioning atubular glomeruli (5). Such changes are commonly observed in patients with type 1 diabetes with overt proteinuria; Najafian et al. (6) noted that in patients with normal to moderately impaired GFR, 17% of glomeruli were atubular and an additional 51% were attached to atrophic tubules (Fig. 3). Similar findings have also been reported in type 2 diabetes, where atubular glomeruli were found in 7% of patients with diabetes, with a further 26% showing glomerulotubular junction abnormalities even in the absence of significant proteinuria so that extent of such abnormalities correlated inversely with creatinine clearance (r = −0.70, P = 0.011) (7).
Primary Tubulopathy Induces Secondary Glomerular Disease
Glomerular pathology in diabetes occurs as a consequence of the interaction between resident glomerular endothelial, mesangial, and epithelial cells with the diabetic milieu. In addition, however, recent studies suggest that the proximal tubule may also contribute to glomerulopathy. In their seminal 2013 study, Hasegawa et al. (8) provide evidence of retrograde trafficking between the proximal tubule and the glomerulus, showing that nicotinamide mononucleotide (NMN) released by proximal tubular epithelial cells diffuses back to the glomerulus to induce podocyte foot process effacement and albuminuria (9) (Fig. 4). Given the importance of podocyte injury not only in the development of proteinuria but also in the progression of glomerulosclerosis and tubuloglomerular junction pathology (10), the triggering of glomerular pathology by the proximal tubule reinforces the primary importance of this region in disease development. We are, however, reminded of the importance of using multiple studies, preferably performed in different laboratories using different animal models, to provide confidence for a new, potentially paradigm-shifting understanding in how diabetic podocytopathy develops.
Cognizant of the absence of significant albuminuria in many patients with declining GFR in diabetic kidney disease, other studies show that proximal tubular injury leads not only to podocytopathy but also to more extensive glomerular injury. Using a mouse model of kidney disease wherein cells of the proximal tubule express the diphtheria toxin receptor, two research groups induced site-selective injury to the proximal tubule. While recovery occurred after a single, low dose administration, repeated dosing of toxin induced all the hallmarks of human diabetic kidney disease with glomerulosclerosis, interstitial fibrosis, capillary rarefaction, tubular atrophy, proteinuria, and elevated serum creatinine (11,12), emphasizing the impact of repeated or continuing injurious stimuli such as those of the diabetic milieu. However, though these studies were not undertaken in diabetic animals, they nevertheless illustrate how intermittent or continuing proximal tubular apoptosis, a common feature of human diabetic kidney disease (13), may lead secondarily to glomerulosclerosis.
The Contiuum of Acute and Chronic Kidney Disease
The proximal tubule is highly susceptible to ischemia and toxin-induced injury that result in acute kidney injury (AKI). Indeed, the term acute tubular necrosis was previously used interchangeably with acute renal failure and AKI. In addition to their propensity to develop chronic kidney disease (CKD), individuals with diabetes are also at much higher risk of AKI (14). While these two disorders were previously viewed as distinct, more recent information indicates that they are closely interrelated, so that patients with CKD are at higher risk of AKI and patients with AKI are at greater risk of progressing to CKD. Indeed, even with apparent full recovery, AKI may be followed by maladaptive tubular repair with fibrosis, inflammation, and microvascular rarefaction that lead to the development of CKD (15–17).
The relationship between episodes of AKI and CKD progression in diabetes is well supported by epidemiological data. Over a 10-year period, Thakar et al. (18) noted that AKI was a common event, occurring in 29% of hospitalized veterans with diabetes. In addition, the study also noted that each episode of AKI conferred a doubling in the risk of progression to CKD stage 4 (eGFR <30 mL/min/1.73 m2), independent of other covariates associated with disease progression (18). Furthermore, the effect of AKI on kidney prognosis was “dose dependent,” worsening incrementally with the number of AKI episodes sustained (Fig. 5). Similar findings have also been reported in the SURvie, DIAbete de type 2 et GENEtique (SURDIAGENE) study of individuals with diabetes. In that study, AKI not only predicted a 2.47-fold increase in the likelihood of doubling serum creatinine or developing end-stage renal disease but was also a major predictor of heart failure hospitalization, myocardial infarction, stroke, and cardiovascular death even after adjusting for eGFR and albuminuria in multivariate analyses (19).
From an intervention perspective, it is noteworthy that sodium–glucose cotransporter 2 (SGLT2) inhibitors with their lowering of systemic blood pressure (20) and increase in afferent arteriolar resistance (21) might, in theory, be expected to increase the likelihood of AKI. However, the reverse relationship was found in the BI 10773 (Empagliflozin) Cardiovascular Outcome Event Trial in Type 2 Diabetes Mellitus Patients (EMPA-REG OUTCOME) trial, wherein empagliflozin reduced AKI and acute renal failure (22). These findings, as discussed later in the section on Na+ transport, suggest that SGLT2 inhibitor–mediated reduced proximal tubular energy requirements may have not only increased the kidneys’ resilience to acute injury (23) but, given their interrelationship with AKI, such findings may also explain the reduction in CKD progression and cardiovascular disease noted in the EMPA-REG OUTCOME trial (22,24).
Biomarkers of Proximal Tubular Injury
The realization that many patients with diabetes and low GFR do not have significant albuminuria and that GFR decline frequently precedes the development of microalbuminuria (25) has led to a vigorous search for alternative or additional biomarkers (26). Among those that appear relatively specific to proximal tubular epithelial cells are kidney injury molecule 1 (KIM-1), liver fatty acid binding protein (L-FABP), and N-acetyl-β-d-glucosaminidase (NAG).
Several studies have examined urinary KIM-1 in diabetes, showing that while its urinary excretion increases commensurately with declining kidney function, it provides little additional information on risk or progression beyond conventional markers in either type 1 or type 2 diabetes (27,28). Serum and plasma concentrations may, on the other hand, be more helpful. In type 1 diabetes, serum KIM-1 concentrations continued to predict eGFR loss and risk of end-stage kidney disease in subjects with type 1 diabetes and proteinuria after adjustment for baseline urinary albumin-to-creatinine ratio, eGFR, and HbA1c (29). In a follow-on study, the same Joslin investigators examined the predictive power of plasma KIM-1 in individuals with type 1 diabetes with normo- and microalbuminuria whose kidney function, as measured by serum creatinine and cystatin C, was normal at baseline (30) (Fig. 6). In a multivariate model, plasma KIM-1 remained strongly associated with the risk of renal function decline regardless of baseline characteristics, reinforcing the view that proximal tubular injury plays a role in the early function decline in diabetes (30).
L-FABP is a 15-kDa protein that, as it names suggests, regulates fatty acid transfer in a range of organs including the proximal tubule (31). An elevated urinary L-FABP concentration provides a rapid and sensitive indicator of AKI risk following cardiac surgery and sepsis that has also been examined as a predictive marker of kidney disease progression in diabetes. In individuals with type 1 diabetes, an increase in urinary L-FABP precedes the development of microalbuminuria and falls with ACE inhibition (32). Further studies attest to the ability of urinary L-FABP to predict kidney disease progression and all-cause mortality in type 1 diabetes, independently of urinary albumin excretion (33), with similar ability to predict cardio-renal end points in Japanese patients with type 2 diabetes without overt proteinuria (34). In another report, however, the FinnDiane (Finnish Diabetic Nephropathy) Study Group found that urinary L-FABP may not improve risk prediction more than albuminuria but still advocated for further studies be done (35).
The urinary excretion of NAG, a lysosomal proximal tubular enzyme that is found predominantly in the proximal tubule, is also increased in diabetes even in the setting of normoalbuminuria and normal eGFR, consistent with the view that proximal tubular dysfunction is a measurable component of early diabetic kidney disease (36). Further increases in urinary NAG are noted in the presence of microalbuminuria and moreover, similar to KIM-1, lower baseline concentrations of NAG were associated with regression of microalbuminuria over a 2-year period in individuals with type 1 diabetes (37).
Tubular Hypoxia: a Driving Force in Diabetic Kidney Disease
Receiving ∼20% of cardiac output, much of the kidney’s high O2 requirements are accounted for by the enormous reabsorptive functions of the proximal tubule that, in turn, render it particularly vulnerable to hypoxia. Indeed, this vulnerability forms the basis for the now well-established “chronic hypoxia theory” of CKD elaborated by Fine et al. in 1998 (38) and further refined in subsequent iterations by other investigators. In experimental diabetes, principally as a result of increased O2 consumption, kidney cortex pO2 is ∼10 mmHg lower than in controls (39). The consequences of hypoxia for proximal tubular epithelial cells is similar to that at other sites, leading not only to apoptosis (40) but also to stimulation of both tubular cells and resident fibroblasts to elaborate increased quantities of extracellular matrix by both transforming growth factor-β (TGF-β)–dependent and –independent mechanisms (38,41–43). The resultant extracellular matrix expansion not only increases the diffusion distance for O2 delivery to the parenchyma but also compresses and disrupts the local architecture, leading to microvascular rarefaction. This then further aggravates the extent of tubulointerstitial hypoxia, setting up a vicious cycle whereby fibrosis begets more fibrosis (Fig. 1).
As elaborated in detail below, the proximal tubule’s propensity to hypoxic injury in diabetes can be attributed to three factors: 1) an increase in metabolic activity as a consequence of the high energy consuming processes of sodium reabsorption and gluconeogenesis, 2) impaired O2 utilization due to altered substrate delivery and mitochondrial dysfunction, and 3) reduced O2 due to microvascular rarefaction (Fig. 1).
The evolutionary move some 365 million years from the sea onto dry land required substantial changes that included not only adaptation to atmospheric O2 but also the ability to avidly reabsorb sodium in this new, comparatively salt-deficient environment. Indeed, 60% of the kidney’s overall energy consumption is devoted to sodium reclamation with the proximal tubule responsible for almost two-thirds, primarily through the activity of the basal Na+/K+ ATPase, quantified as ouabain-sensitive O2 consumption (44).
While sodium–glucose linked transport across the apical membrane of the proximal tubular cell is not of itself an energy-requiring process, its continuing activity is dependent on the maintenance of the electrochemical gradient for Na+, generated by Na+/K+ ATPase activity (Fig. 7). Accordingly, the increase in glucose reabsorptive capacity that develops in diabetes (45) is accompanied by a commensurate demand in Na+/K+ ATPase activity that when measured by ouabain inhibitable O2 consumption increases by about 30% in the experimental setting (46). Here, the SGLT1/2 inhibitor phlorizin was shown to ameliorate the diabetes-induced increase in Na+/K+ ATPase and O2 consumption (46) (Fig. 7). Consistent with these findings, the SGLT2 inhibitor dapagliflozin renders proximal tubular epithelial cells resistant to hypoxia-induced apoptosis, affording protection from ischemia-reperfusion injury (47). Together, these findings raise the intriguing possibility that the reduction in both GFR decline and AKI reported in the EMPA-REG OUTCOME trial (22) may be the result of reduced proximal tubular energy requirements, the ability of the drug to improve glycemia and modulate glomerular hemodynamics notwithstanding (23).
In addition to the SGLTs, apical sodium transport in the proximal tubule is mediated by several other carrier proteins including sodium–hydrogen exchanger 3 (NHE3) that, similar to the SGLTs, require Na+/K+ ATPase to maintain an electrochemical gradient. Increases in NHE3 have been reported in human proximal tubular cells exposed to high glucose and in animals with streptozotocin-induced diabetes (48). A similar increase in Na-lactate cotransporter activity, as a consequence of increased lactate production in the setting of poor glycemic control (49), has also been shown to occur in diabetes (50).
While the proximal tubule reabsorbs glucose, it does not metabolize any of the enormous load that traverses it, relying on lactate, glutamate, and ketones as alternative substrates for energy production (51). Along with hepatocytes, proximal tubular epithelial cells are unique in their ability to undertake gluconeogenesis and export glucose into the circulation, although unlike the liver, the kidney is not responsive to glucagon. From an energy perspective, gluconeogenesis is a demanding process, requiring six energy equivalents (4 ATP, 2 GTP) to synthesize a single molecule of glucose from lactate or pyruvate, contrasting sharply with the 2 molecules of ATP that are generated by glycolysis. Indeed, gluconeogenesis is a major source of the kidney’s ouabain-insensitive O2 usage and energy expenditure, accounting for up to 25% of the energy needed for sodium reabsorption (50).
In the nondiabetic setting, the liver, by a combination of glycogenolysis (50%) and gluconeogenesis (30%), contributes 80% of glucose released into the circulation, with the remaining 20% derived from renal gluconeogenesis (52). In the postprandial state when glycogenolysis and hepatic gluconeogenesis are relatively suppressed, de novo glucose synthesis by the kidney accounts for ∼60% of endogenous glucose release (52). The extent of renal gluconeogenesis is increased in diabetes where in the fasted state gluconeogenic activity increases approximately threefold such that the kidney releases on average 2.21 µmol/kg/min of glucose into the circulation, only marginally lower than the liver’s 2.60 µmol/kg/min (53). Postprandial glucose release by the kidney is similarly increased in subjects with type 2 diabetes when compared with age-, weight-, and sex-matched volunteers without diabetes (54).
Given that the proximal tubule’s metabolic activity is almost entirely oxidative, there is a commensurate reduction in tricarboxylic acid cycle activity and ATP production in the relatively hypoxic diabetic kidney (55,56). Moreover, by requiring more O2 to be consumed for each molecule of ATP generated, the increase in free fatty acids, a component of the dysglycemic state, may exacerbate the extent of ischemia. This relative energy inefficiency can be quantified as the ATP/O2 ratio where glucose has a ratio of 3.17 while palmitate has a ratio of 2.83 (57), and while these differences in oxygen consumption may seem modest, they can have substantial impact when O2 delivery is marginal.
Exacerbating the increased O2 demands in diabetes is the recent realization that mitochondria, the organelles responsible for aerobic energy production, are structurally abnormal and dysfunctional in diabetes (58,59). Indeed, abnormalities of proximal tubular mitochondrial structure and function may be the earliest manifestation of kidney disease. In the rat, for instance, Coughlan et al. (60) found evidence of impaired mitochondrial ATP generation and organelle fragmentation in proximal tubular epithelial cells as early as 4 weeks after the induction of experimental diabetes. That these changes precede increases in urinary albumin excretion, abnormal glomerular morphology, or even elevation of urinary KIM-1 suggests that they may be primary abnormalities. From a therapeutic point of view, such findings also raise the possibility of using strategies that regulate mitochondrial biogenesis such as the silent information regulator 1 activators that are currently undergoing clinical trial in a range of chronic diseases (61).
High glucose and its downstream effector molecules have long been known to increase endothelial cell apoptosis in cell culture (62,63). Importantly, these changes, recognized as capillary rarefaction, are also seen in the in vivo setting as a characteristic feature of diabetic kidney disease that correlates with declining kidney function (64–67) (Fig. 8). Though potentially compensated by endothelial cell regeneration, this reparative process, if anything, is impaired in diabetes (68). As a result, blood supply to the proximal tubule is impaired both by the intrinsic capillary loss within the tubulointerstitium described above and a consequence of glomerular capillary occlusion (64,65).
Given the lack of observable capillary loss in animal models of diabetic kidney disease, much of the experimental work on therapeutic angiogenesis has relied on nondiabetic models such as renal artery stenosis and following subtotal nephrectomy (SNX) that do develop substantial capillary rarefaction. Here, a number of strategies for reconstituting the microcirculation are in development including mesenchymal stem cells (MSCs), so-called endothelial progenitor cells (EPCs), and extracorporeal shockwave therapy. EPCs, for instance, have been shown to exert both proangiogenic and antifibrotic properties, attenuating capillary loss in the tubulointerstitium and glomerulus as well as preserving kidney function in the SNX rat (69). MSCs, on the other hand, have a range of potential beneficial effects that in addition to their angiogenic, immunomodulatory, and anti-inflammatory activities also lower blood glucose in humans with diabetes by as yet unknown mechanisms (70). Most recently, a phase II trial of allogeneic MSCs in patients with advanced diabetic kidney disease has been reported (NCT01843387). This study compared the effects of two doses of MSCs in 30 subjects whose baseline GFR was 20–50 mL/min/1.73 m2, showing a trend in stabilizing measured GFR at 12 weeks when compared with the continuing decline in placebo-treated patients (71).
Perhaps most tantalizing because of its simplicity is the finding that low-energy extracorporeal shockwave therapy (ESWT) induces angiogenesis. In a double-blind, placebo controlled study of patients with severe angina, Kikuchi et al. (72) used this technology to improve blood flow and function in the ischemic heart, reducing pain scores and improving left ventricular ejection fraction. Using a similar strategy in pigs with renal artery stenosis, Zhang et al. (73) reported improvement in microvascular density and tissue oxygenation along with reduced fibrosis and better kidney function after six sessions of ESWT. In exploring the mechanisms that underlay these effects, this group noted elevated expression of vascular endothelial growth factor, mainly in proximal tubular cells along with a reduction in TGF-β expression, providing further substance to the burgeoning exploration of mechanotransduction in disease development and reversal. The effects of ESWT in human diabetic kidney disease, however, remain unknown.
Nonhypoxia-Related Proximal Tubular Mechanisms
In addition to the hypoxia, several other nonhypoxia-related proximal tubule pathways involved in the development of diabetic kidney disease have been the subjects of recent reviews. These include the now well-documented local, predominantly proximal tubule–based renin-angiotensin system (74), the toxic effects of albumin bound fatty acids (75,76), and the activation of epidermal growth factor receptor signaling pathways (77). Still more recently, and not yet the subject of detailed review, is the exploration whereby diabetic kidney disease, like most forms of CKD, once started, continues to progress inexorably. Among the potential contributors to this process is the recent finding that organ stiffness, an inevitable consequence of fibrosis, induces further fibrosis. Briefly, the presence of fibrosis leads to tissue stiffness that can be sensed mechanically by cells. Rather than dampening the fibrogenic process, the presence of a stiff matrix seems to induce a positive feedback cycle to enhance it. The leading contenders in our current understanding of this process are the mechano-transducing transcription cofactors: Yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ). Together, YAP and TAZ perpetuate TGF-β signaling intermediates, Smad2/3, to be retained in the nucleus, thereby perpetuating its fibrogenic activity (78). Consistent with diabetes as a profibrotic state, YAP expression and phosphorylation are increased in diabetic mouse proximal tubular cells so that modulating TAZ-YAP and related pathways has become an important new target for drug development in diabetic kidney disease and other chronic diseases that are characterized by fibrosis (79).
Proximal Tubule–Targeted Therapeutics
The revolution in molecular biology has highlighted the extent of pathophysiological derangements in the diabetic kidney that include a broad range of perturbations in epigenetics, protein–protein interactions, transcriptional changes, and posttranslational modifications. Unfortunately, moving these discoveries into new therapies has been limited in part by the “druggability” of the target, i.e., the likelihood of being able to modulate a target with a small-molecule drug. Indeed, <10% of potential targets are thought to be in this category (80). Antisense technology that mediates specific target–directed degradation of mRNA, microRNA (miRNA), and long noncoding RNA (lncRNA), is, however, far less restricted. Indeed, this technology has already moved to human use with two systemically administered antisense oligonucleotides (ASOs) approved by the U.S. Food and Drug Administration: mipomersen (Kynamro, Genzyme), an inhibitor of apoprotein B synthesis for the management of homozygous familial hypercholesterolemia, and eteplirsen (Exondys 51, Sarepta Therapeutics) for individuals with Duchenne muscular dystrophy that selectively binds to exon 51 of dystrophin premRNA to restore the open reading frame and enable the production of functional dystrophin.
Shorter length (12 vs. the usual 18 or longer) second-generation “shortmer” ASOs have comparatively lower plasma protein binding so that they undergo greater fractional clearance through the glomerulus. As a result of their avid uptake by the brush border of the proximal tubule, these agents are relatively selective for that site. Indeed, this strategy forms the basis for the development of the SGLT2 ASO, ISIS 388626 (Ionis, formerly ISIS), that has completed safety, tolerability, and activity studies in type 2 diabetes (NCT00836225). Such studies highlight the enormous potential for nucleotide-based therapy to pursue novel, nontraditionally druggable targets to ameliorate proximal tubule pathology in diabetic kidney disease.
Although much work needs to be done, substantial data now support the existence of a diabetes-induced proximal tubulopathy as an early disease event that both predicts and contributes to the development of CKD in diabetes. While not ignoring the glomerulus, directing attention to the proximal tubule for biomarker development, therapeutic discovery, and pathophysiological understanding seems prescient.
Acknowledgments. The author regrets that owing to space constraints, much of the excellent work that has been done on this subject could not be included.
Funding. R.E.G. is the Canada Research Chair in Diabetes Complications, and this review was supported in part by Canada Research Chairs Program grant 950-218644.
Duality of Interest. R.E.G. reports having received consulting and lecture fees from Merck, AstraZeneca, Eli Lilly, Boehringer Ingelheim, Mesoblast, and Janssen along with grant funds through his institution from Merck, AstraZeneca, Eli Lilly, and Boehringer Ingelheim. R.E.G. was formerly a shareholder in Fibrotech Therapeutics, which was wholly acquired by Shire Pharmaceuticals in 2014. No other potential conflicts of interest relevant to this article were reported.