Progressive kidney disease is accompanied by a constellation of changes in mineral metabolism, as well as numerous bone and cardiovascular comorbidities that are now collectively referred to as chronic kidney disease-mineral and bone disorder (CKD-MBD).1 Characteristic physiological changes associated with CKD-MBD are hyperphosphatemia, hypocalcemia, 25(OH)D and 1,25(OH)2D deficiency, secondary hyperparathyroidism (SHPT), metabolic bone disease (ranging from adynamic bone disease to osteitis fibrosa), vascular calcification, and increased mortality. Although these abnormalities are invariable features in patients with late-stage kidney disease (ie, CKD stage 5), the mechanism to describe the etiology of these findings has remained elusive.
Recently, a novel phosphaturic hormone, fibroblast growth factor 23 (FGF23), has been identified as a potential contributor to the development of mineral metabolism defects and their associated comorbidities in patients with CKD. To this end, multiple cross-sectional and observational studies in this population suggest that serum FGF23 levels rise early in the course of kidney disease and are independently associated with increased morbidity and mortality.2–5 Despite associations of FGF23 with many components of the CKD-MBD phenotype, there remains a lack of longitudinal investigations evaluating the changes in FGF23 in relation to other mineral metabolism changes, especially in the earliest stages of kidney damage. A lack of reliable markers for the detection of early decrements in renal function has made it difficult to study mineral metabolism changes in humans at these earliest stages of disease.
The creation of genetically altered mouse models that develop chronic kidney damage of predictable onset and progression provides the opportunity to study the interrelationships of mineral metabolism changes at all stages of kidney disease. As one such model, the Col4a3 null mouse, which was created by deletion of the gene encoding the α3 chain of type IV collagen, is a murine model of autosomal-recessive Alport syndrome, a human condition exhibiting hereditary nephritis associated with various ocular and auditory defects. Col4a3 null mice demonstrate severe proteinuria beginning at 5 weeks of age and progressive renal failure over a 12-week period.6, 7 The renal histology of this model has been extensively characterized throughout its life span, making it an attractive model for longitudinal investigations of the complex mineral metabolism changes that occur from the earliest stages of kidney damage all the way to end-stage disease. To explore the suitability of the Col4a3 null mouse as a model for CKD-MBD and to better understand the relation between FGF23 and the formation of the CKD-MBD phenotype, we examined the temporal changes of FGF23 gene expression in bone and alterations in circulating FGF23 levels as they relate to mineral metabolism changes and kidney disease progression in this model.
Materials and Methods
Animal preparation and study protocol
All mice were maintained in accordance with recommendations in the “Guide for Care and Use of Laboratory Animals” from the Institute on Laboratory Animal Resources, National Research Council (National Academy Press, 1996), and all animal protocols were reviewed and approved by the University of Kansas Medical Center Institutional Animal Care and Use Committee before the commencement of this research. Mice containing a targeted deletion of the NC1 domain of the α3(IV) collagen chain7 were originally obtained from The Jackson Laboratory (strain 129-Col4a3tm1Dec, Bar Harbor, ME, USA). Generation of the FGF23-eGFP reporter mouse was performed as previously described.8 Col4a3+/+FGF23+/eGFP and Col4a3−/−FGF23+/eGFP genotypes were generated by first mating Col4a3+/− and FGF23+/eGFP mice, then mating double-heterozygotes. Wild-type littermates of the Col4a3 null mice were used as the control groups for all experiments in this study. Mouse diet was purchased from Harlan-Teklad (Madison, WI, USA) and consisted of 0.6% Ca, 0.54% Pi and 3000 IU vitamin D per kg of diet. Study mice were generated by mating Col4a3+/− males and females to obtain the desired genotypes. Col4a3 null and wild-type mice were enrolled at weaning into one of four cohorts based on predetermined collection time points of 4, 6, 8, and 12 weeks of age. Body weight and survival were recorded weekly for all mice. Because the measured serum parameters and observed phenotype were identical between Col4a3+/− and Col4a3+/+, their values were combined for data analysis. Mice were housed in metabolic cages for urine collection beginning 24 hours before blood and tissue collection. All mice were anesthetized with pentobarbital (50 mg/kg) before euthanasia by exsanguination.
For analysis of eGFP expression, femurs were stored in 10% formalin for 48 hours at 4°C, then transferred to Immunocal decalcifying solution (Decal Chemical, Tallman, NY, USA) for 48 hours at 4°C. Femurs were placed back into 10% formalin at 4°C for 12 hours, transferred to 15% sucrose solution in 1x PBS for 12 hours, then transferred to a 30% sucrose solution for 12 hours. Femurs for eGFP expression were then embedded for frozen section using OCT embedding compound and stored at −20°C. Frozen sections were cut at 10-µm thickness and evaluated under fluorescent light to characterize eGFP expression. DAPI counterstaining of femur sections was performed using mounting medium + DAPI (Vector Labs, Burlingame, CA, USA). Processing of specimens for bone histomorphometry was performed as previously described.9 Bone histomorphometry measurements were performed using the Explora Nova Bone system V4.0 (La Rochelle, France) after Goldner's trichrome (trabecular and osteoid measurements) and tartrate-resistant acidic phosphatase (TRAP; osteoclast measurements) staining of femur sections. Histomorphometry measurements were expressed in standard nomenclature as previously described.10 Kidney specimens for histological analysis were fixed in 4% paraformaldehyde for 24 hours, embedded in paraffin, and cut into 3-µm sections for periodic acid Schiff (PAS) staining.
Serum and urine biochemistries
Serum and urine calcium (Ca) were measured using the Liquicolor kit (Stanbio Lab, Boerne, TX, USA), and phosphorus (Pi) was measured by the phosphomolybdate–ascorbic acid method. Fractional excretion of phosphorus was calculated using the formula: [(urine phosphorus in mg/dL) / (serum phosphorus in mg/dL)] / [(urine creatinine in mg/dL) / (serum creatinine in mg/dL)]. Intact FGF23 was measured by ELISA (Kainos Lab, Tokyo, Japan), serum 1,25(OH)2D by the IDS EIA (Tyne and Wear, UK), PTH by the Mouse Intact PTH ELISA Kit (Alpco Diagnostics, Salem, NH, USA), blood urea nitrogen (BUN) by Quantichrome Urea Assay kit (BioAssay Systems, Hayward, CA, USA), and urine and serum creatinine by the University of Texas Southwestern Mouse Phenotyping Core using a Vitros 250 analyzer (Ortho-Clinical Diagnostics, Rochester, NY, USA).
Quantitative real-time PCR analysis
Tissues for qRT-PCR analysis were snap-frozen in liquid nitrogen and stored at −80°C until processing. Total RNA was extracted after homogenization using TRI-Reagent (Molecular Research Center, Cincinnati, OH, USA) and treated with RNase-free DNase (Qiagen, Valencia, CA, USA). First strand cDNA was synthesized using iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA), with 1 µg of RNA used for each reverse-transcriptase reaction. PCR reactions contained 100 ng cDNA, 300 nM of each primer, and 1X iQ SYBR Green Supermix (Bio-Rad) in 50 µl. The threshold cycle (Ct) of each gene product was normalized to the Ct for GAPDH. Primer sequences used were: FGF23F: 5′-TTTCCCAGGTTCGTCTAGG-3′, FGF23R: 5′-CTCGCAGGTGACTCTCAG-3′; Col4F: 5′-CCTGCTAATATAGGGTTCGAGA-3′, Col4R: 5′-CCAGGCTTAAAGGGAAATCC-3′; NaPi2aF: 5′-ATGCTGGCTTTCCTTTAC-3′, NaPi2aR: 5′-AATCCGAATGAGACTGTG-3′; NaPi-2cF: 5′-CGTGCGACTTTATCAATG-3′, NaPi2cR: 5′-TACTGGGCAGTCAGGTTTCC; KlothoF: 5′-CAACCTCTCGTCTCTTCTGC-3′, KlothoR: 5′-TGGGAACTTCATGTTAGG-3′.
Differences between multiple groups were evaluated by one-way ANOVA, whereas differences between two groups were evaluated by the two-sided t test method. Kaplan-Meier analysis was used to measure survival differences. Computations were performed using Prism 5 software (GraphPad Software, San Diego, CA, USA) and presented as mean ± SEM.
Genotype, body weight, and survival of Col4a3 wild-type and null mice
We examined Col4a3+/+, Col4a3+/−, and Col4a3−/− mice, which were identified by PCR genotyping (Fig. 1A). All genotypes were born with the expected Mendelian frequency. On a standard diet containing 0.6% calcium and 0.54% phosphorus, heterozygous Col4a3 mice were indistinguishable from wild-type (WT) mice within the first 14 weeks of life, whereas Col4a3 null mice demonstrated uremic cachexia (Fig. 1B) and increased mortality beginning around 10 weeks of age (Fig. 1C).
Evidence of progressive kidney disease in Col4a3 null mice
Col4a3 null mice exhibited a time-dependent rise in BUN and creatinine concentrations beginning around 8 weeks of age (Fig. 2A). By 12 weeks, the serum creatinine concentration was significantly elevated in Col4a3 null mice (1.27 mg/dL) compared with WT controls (0.44 mg/dL) (Table 1). BUN levels followed a pattern of rise similar to creatinine in Col4a3 null mice (Fig. 2A). Of note, there was a significant difference in creatinine values between WT and Col4a3 null mice at the 4- and 6-week time points, although a progressive rise in creatinine within the Col4a3 null genotype was not observed until after 8 weeks. Consistent with other reports, the earliest changes in renal histology in null mice occurred at 6 weeks of age in the form of tubular injury and an interstitial infiltration of inflammatory cells (Fig. 2B). After 10 weeks, Col4a3 null mice demonstrated substantial glomerulosclerosis accompanied by massively dilated tubules and interstitial fibrosis (Fig. 2B).
Table 1. Serum Biochemistries in Wild-Type and Col4a3−/− Mice
WT = wild type; KO = knockout.
All measurements presented as mean ± SEM, n ≥ 5 mice per group for each time point.
Statistically different (p < 0.05) compared with 12-week value of same genotype.
Statistically different (p < 0.05) compared with 8-week value of same genotype.
Statistically different (p < 0.05) compared with 6-week value of same genotype.
Statistically different (p < 0.05) compared with wild-type mice at same age.
Temporal changes in serum mineral metabolism parameters
We measured serum markers of phosphate and calcium homeostasis at several time points in Col4a3 null and wild-type mice (Table 1). Col4a3 null mice maintained serum calcium and phosphate concentrations indistinguishable from wild-type mice until the onset of severe renal damage around 12 weeks when frank hyperphosphatemia and hypocalcemia developed. Col4a3 null mice demonstrated a progressive reduction in 1,25(OH)2D levels, from 288 pmol/L at 4 weeks to 39 pmol/L by 12 weeks, and a noticeable separation from wild-type levels by 8 weeks of age. Col4a3 null mice also showed evidence of progressive secondary hyperparathyroidism and increments in FGF23 levels beginning by 6 weeks (Table 1). There was an approximate 40-fold increase in serum FGF23 levels in Col4a3 null mice over the 12-week period, from 133 to 5407 pg/mL. Elevations in FGF23 appeared to occur prior to a progressive increase in BUN and serum creatinine levels within the Col4a3 null mice. In addition, increments in FGF23 tracked with changes in PTH and coincided with increased renal phosphorus excretion (Fig. 3A). Furthermore, a decline in serum 1,25(OH)2D levels mirrored the rise in serum FGF23 and occurred despite increasing PTH levels, which would normally stimulate 1,25(OH)2D production. Consistent with the trends observed in wild-type mice from other studies, wild-type mice in our study demonstrated mild elevations in PTH and FGF23 with age, along with a mild decline in 1,25(OH)2D levels (Table 1).11
Increase in renal phosphorus excretion with kidney disease progression in Col4a3 null mice
Both the fractional excretion of phosphate (Fig. 3A) and total urine phosphate content (Fig. 3B) were increased in Col4a3 null mice compared with wild-type mice beginning at 6 weeks, which corresponded with the earliest changes in renal histology (Fig. 2B) and early increments in circulating PTH and FGF23 levels (Table 1). Consistent with the known effects of PTH and FGF23 to suppress phosphate reabsorption in the proximal tubule, quantitative real-time PCR analysis of kidneys from these mice revealed a progressive decline in sodium-phosphate cotransporter (NaPi-2a & NaPi-2c) gene expression beginning around 6 weeks of age in Col4a3 null mice (Fig. 3C, D), with a marked suppression of gene expression being observed in these mice at 12 weeks. Interestingly, the decline in renal sodium-phosphate cotransporter expression was mirrored by a decrease in renal klotho gene expression (Fig. 3E).
Disconnect between FGF23 serum levels and FGF23 gene transcription in bone in early stages of kidney damage in Col4a3 null mice
Although serum FGF23 levels were elevated by 6 weeks of age in Col4a3 null mice (Fig. 4A), qRT-PCR analysis of calvaria specimens from these mice suggested FGF23 gene transcription in bone to not increase until 12 weeks of age (Fig. 4B), when severe kidney damage was present. To identify the cellular source of FGF23 in bone, we mated our Col4a3 strain to an FGF23 promoter-eGFP reporter mouse line to generate Col4a3 null mice expressing eGFP in bone under the control of the FGF23 promoter. Col4a3 null mice showed increased FGF23 promoter-driven eGFP expression in trabecular osteocytes within the metaphysis of the femur at 12 weeks, whereas osteocytes in cortical bone within the diaphysis appeared to have little eGFP expression (Fig. 4C).
High-turnover bone disease in Col4a3 null mice
Histological analysis by TRAP staining in femurs from 10 week old Col4a3 null mice revealed major increases in osteoclast numbers (N.Oc/BS) (Fig. 5A, B). Col4a3 null mice demonstrated slightly higher bone volume (BV/TV) and trabecular number (Tb.N) compared with wild-type mice, but this difference was not statistically significant. There was no appreciable difference in osteoid thickness (O.Th), osteoid volume (OV/BV), trabecular thickness (Tb.Th), or trabecular separation (Tb.Sp) between Col4a3 null and wild-type mice. Overall, histomorphometric analysis suggested the presence of high-turnover bone disease in Col4a3 null mice.
The discovery of marked elevations of circulating FGF23 levels in patients with CKD has challenged traditional theories regarding the origins of mineral metabolism defects in this setting. Given the complex interactions that exist between PTH, FGF23, and 1,25(OH)2D and a lack of longitudinal data evaluating these hormones at the onset and during the progression of renal failure, it has been difficult to discern the independent contributions of these hormones to the pathogenesis of CKD-MBD. With this investigation, we utilized the Col4a3 null mouse as a model of progressive kidney disease to gain new insight into the temporal relationship that exists between increments in FGF23, the development of mineral metabolism abnormalities, and the associated comorbidities in CKD.
In our initial characterization of this model, Col4a3 null mice were found to develop progressive renal failure with substantial mortality by 12 weeks of age (Figs. 1 and 2). More important, these mice demonstrate mineral metabolism abnormalities that recapitulate the changes reported in cross-sectional observations in humans with CKD-MBD,12, 13 including secondary hyperparathyroidism, a decline in circulating 1,25(OH)2D levels, late hyperphosphatemia and hypocalcemia, and progressive elevations in FGF23 levels (Table 1). In addition, Col4a3 null mice develop late-stage high-turnover bone disease, characterized by osteoclast-mediated bone resorption (Fig. 5), that is consistent with the metabolic bone disease found in humans with elevated PTH levels in the setting of end-stage renal disease (ESRD).14, 15 Interestingly, histological analysis also suggested a nonsignificant increase in bone volume in null mice and no apparent defects in bone mineralization (ie, no evidence of osteomalacia) or growth plate abnormalities. We found no consistent presence of vascular calcification in these mice (data not shown), possibly due to the short survival of these mice or the abbreviated duration of hyperphosphatemia. It is plausible that placing these mice on a diet containing higher levels of phosphate could promote a more reliable presence of vascular calcification by the 12-week time point.
Our next important observation in this model was evidence for an early rise in FGF23 before elevations in the traditional markers of renal function. To this end, increments of FGF23 in Col4a3 null mice were apparent by 6 weeks of age (Table 1), when parenchymal damage was first observed on renal histology (Fig. 2B) and a progressive rise in BUN and serum creatinine levels was not yet observed (Fig. 2A). The significance of this observation could be substantial. Currently, serum creatinine is the most commonly accepted marker for assessing a decline in glomerular function in patients with CKD, yet serum creatinine levels do not increase until a significant proportion of functional renal mass is lost.16, 17 The finding that FGF23 levels rise before creatinine levels suggests this hormone has considerable potential as a predictor for kidney disease progression at an earlier stage of CKD. This seems consistent with cross-sectional data in humans suggesting an association of FGF23 with albuminuria in patients with known cardiovascular disease and only mild renal impairment.18
Our assessment of changes in serum parameters of mineral metabolism in Col4a3 null mice revealed a temporal association between elevations of FGF23 and reductions in serum 1,25(OH)2D levels, consistent with the hypothesis that early elevations in serum FGF23 may contribute to decreased 1,25(OH)2D levels in CKD by suppressing renal 1α-hydroxylase and stimulating 24-hydroxlase activity.19, 20 Similar to patients with CKD, this decline in serum 1,25(OH)2D levels occurred despite a concomitant rise in PTH, which would be expected to stimulate 1,25(OH)2D production by the kidney. This observation implies that the suppressive effects of FGF23 and phosphate retention may override the stimulatory effects of PTH on 1,25(OH)2D production, at least in the setting of CKD. Unexpectedly, we observed unusually low PTH levels in Col4a3 null mice at 4 weeks of age compared with wild-type mice. It is possible that the low PTH levels may represent a transient suppression of PTH secretion by very early increments in FGF23, which would be consistent with the suspected direct action of FGF23 on the parathyroid gland.21, 22 Alternatively, the low PTH levels observed at this early time point may reflect some undefined interaction between early changes in glomerular physiology and pathways regulating PTH production. The recent finding that klotho and FGF-receptor 1 expression are decreased in parathyroid glands in advanced CKD may support the former hypothesis and explain the inability of FGF23 to suppress PTH production as kidney disease progresses in this model.22 Furthermore, early decrements in parathyroid expression of klotho and FGF-receptor 1 may provide a mechanism for the rebound in PTH levels at 6 weeks of age in the face of unaltered serum calcium levels in the Col4a3 null model. Regardless, further studies are needed to determine the etiology and importance of these early changes in serum PTH levels and to identify if these are transient changes that occur universally in the earliest stages of CKD or a unique observation in Col4a3 null mice.
As renal function declined in Col4a3 null mice, we observed an increase in both the fractional excretion of phosphorus and total phosphorus excretion by 6 weeks of age (Fig. 3A, B), that likely resulted from the increase in circulating levels of both PTH and FGF23. Although previous studies in humans suggest that fractional excretion of phosphorus increases with advancing kidney disease,13, 23, 24 an increase in total phosphorus excretion is not a common observation in this setting13 and may be a finding that is unique to either the Col4a3 null model or kidney diseases featuring a more rapid onset and progression. The observed increase in the fractional excretion of phosphorus at 6 weeks correlated with small decrements in sodium-phosphate cotransporter (NaPi-2a and NaPi-2c) and klotho gene expression (Fig. 3C–E), with greater reductions occurring as renal failure progressed and PTH and FGF23 levels became exponentially elevated. Although it has been previously speculated that a decline in renal klotho expression may be the inciting event for a rise in serum FGF23 levels in CKD, decrements in renal klotho expression in Col4a3 null mice occurred simultaneous to the initial rise in serum FGF23 levels, making it difficult to speculate which event occurred first in this model. Interestingly, serum phosphorus levels remained normal in Col4a3 null mice until the presence of severe kidney damage, when reduced renal klotho expression likely resulted in a resistance to the phosphaturic effects of FGF23 and a lack of residual functional nephron mass was present for the excretion of excess total body phosphate.
Similar to previous investigations in uremic animals,25 our evaluation of Col4a3 null mice implicated bone as a source for elevated circulating FGF23 levels in advanced CKD; however, our data suggested that early increments in FGF23 may not result from increased production in bone (Fig. 4A, B). This observation may help explain why circulating FGF23 levels are only modestly elevated in the early to middle stages of CKD (stages I to IV), yet become exponentially elevated in patients with end-stage disease.5 It is plausible that the early rise in FGF23 in the setting of progressive kidney disease results from changes in excretion or post-transcriptional processing of this hormone. Alternatively, a tissue source other than bone may be contributing to the earliest elevations of circulating FGF23. Further studies will be needed to determine whether early increases in FGF23 in CKD are the result of extraskeletal tissue production or some post-transcriptional processing of this hormone. The pattern of FGF23 expression in bone from the Col4a3 null model proved to be unique from previous descriptions of FGF23 expression in inherited and genetic models of FGF23 overexpression.8, 26, 27 In the hereditary hypophosphatemic models caused by mutations of Phex and Dmp1, the primary site of FGF23 production appears to be osteocytes in both cortical and trabecular bone.8, 27 By contrast, Col4a3 null mice demonstrated little expression of FGF23 in cortical osteocytes but widespread expression of FGF23 in trabecular osteocytes (Fig. 4C). The observed difference in cell-type expression patterns for FGF23 production in bone between CKD and hereditary hypophosphatemic disorders may suggest an alternative mechanism for the stimulation of FGF23 production under these unique conditions.
We recognize that there are several limitations to this investigation. First, the relatively short time course for the progression to ESRD in Col4a3 null mice may have prevented the consistent development of several comorbidities that are common in patients with ESRD, such as vascular calcification and osteomalacia. Second, it is plausible that different bone sites may have unique expression patterns of FGF23 in the setting of CKD, making it difficult to definitively preclude an increased production of FGF23 by bone in early CKD. Exhaustive analyses of FGF23 expression patterns in various skeletal locations will be needed to fully understand the contributions of bone to early increments of FGF23 in CKD. Third, despite an observed association between the rise in FGF23 and decline in 1,25(OH)2D levels, we are unable to establish a causal relationship from these observations. However, a recent study investigating the effects of FGF23-deactivating antibodies on vitamin D metabolism suggests a role of FGF23 to promote 1,25(OH)2D deficiency in this setting.28 Finally, given the simultaneous rise of PTH and FGF23 that occurred in Col4a3 null mice, it is clear that an examination of time-dependent changes alone in this model will not be sufficient to decipher the complex interrelationship that exists between PTH and FGF23 in the setting of CKD.
Our data suggest FGF23 to be a novel biomarker for chronic kidney disease-mineral and bone disorder at the early stages of nephron loss. In addition, our results imply that initial increments in circulating FGF23 may result from mechanisms other than increased FGF23 gene transcription in bone. Although an early increase in FGF23 likely prevents hyperphosphatemia in the face of declining functional nephron mass, it appears that this occurs at the expense of promoting 1,25(OH)2D deficiency. Although many controversies remain concerning the intricacies of mineral metabolism regulation in CKD-MBD and the best therapeutic approach to patients with this disorder, our successful characterization of the mineral metabolism phenotype of the Col4a3 null mouse suggests this model to possess considerable value to future hypothesis-driven research that will further advance our knowledge within this field.
JRS was supported by a grant provided by Genzyme Corporation as part of the Genzyme Renal Innovations Program (GRIP). All the other authors state that they have no conflicts of interest.
We thank Dr Timothy Fields and Dr Peter Rowe for their technical assistance with renal and bone histology for this article.
This work was partially supported by National Institutes of Health research grant K08-DK087949 (to JRS) from the National Institute of Diabetes and Digestive and Kidney Diseases.
Authors' roles: Study design: all authors. Data collection: JRS, NH, AI, RG, VD, and YH. Data analysis: all authors. Drafting manuscript: JRS, SL, LDQ. Manuscript revision: JRS, SL, LDQ. All authors approved final content of the manuscript. JRS takes responsibility for the integrity of the data analysis.