Naturally occurring stable calcium isotope ratios are a novel biomarker of bone calcium balance in chronic kidney disease

Dysregulated calcium homeostasis is common in chronic kidney disease and causally associated with disorders of bone mineralization. However, radiological measures and biomarkers do not allow accurate evaluation of bone calcium balance. Non-radioactive calcium isotopes, 42 Ca and 44 Ca, are present in our diet and sequestered into body compartments following principles of kinetic isotope fractionation. Isotopically light 42 Ca is preferentially incorporated into bone, while heavier 44 Ca is excreted. The ratio ( 44/42 Ca serum ) increases when bone formation exceeds resorption and vice versa, re ﬂ ecting bone calcium balance. We measured these calcium isotopes by inductively coupled plasma mass-spectrometry in blood, urine and feces of 42 children with chronic kidney disease and 92 receiving dialysis therapy. We compared the isotope ratios with bone biomarkers and determined total bone mineral content by dual-energy x-ray absorptiometry and peripheral quantitative CT expressed as age-adjusted z-scores. The 44/42 Ca serum ratio positively correlated with serum calcium, 25-hydroxyvitamin D and alkaline phosphatases and inversely with serum parathyroid hormone and other bone resorption markers. The 44/42 Ca serum ratio positively correlated with age-adjusted z-scores of tibial trabecular bone mineral density and total bone mineral content measured by peripheral quantitative CT, and hip bone mineral density measured by dual-energy X-ray absorptiometry. Signi ﬁ cant and independent predictors of total bone mineral content, measured by, were the 44/42 Ca serum ratio and parathyroid hormone. The 44/42 Ca serum ratio, repeated after four weeks, highly correlated with baseline values. When adjusted for calcium-containing medications and kidney impairment, the 44/42 Ca serum ratio in patients receiving dialysis was 157% lower than that of age-matched children and 29% lower than levels in elderly women with osteoporosis, implying signi ﬁ cantly lower bone mineral content. Thus, calcium isotope ratios may provide a novel, sensitive and noninvasive method of assessing bone calcium balance in chronic kidney disease.

Dysregulated calcium homeostasis is common in chronic kidney disease and causally associated with disorders of bone mineralization. However, radiological measures and biomarkers do not allow accurate evaluation of bone calcium balance. Non-radioactive calcium isotopes, 42 Ca and 44 Ca, are present in our diet and sequestered into body compartments following principles of kinetic isotope fractionation. Isotopically light 42 Ca is preferentially incorporated into bone, while heavier 44 Ca is excreted. The ratio ( 44/42 Ca serum ) increases when bone formation exceeds resorption and vice versa, reflecting bone calcium balance. We measured these calcium isotopes by inductively coupled plasma mass-spectrometry in blood, urine and feces of 42 children with chronic kidney disease and 92 receiving dialysis therapy. We compared the isotope ratios with bone biomarkers and determined total bone mineral content by dual-energy x-ray absorptiometry and peripheral quantitative CT expressed as age-adjusted z-scores. The 44/42 Ca serum ratio positively correlated with serum calcium, 25-hydroxyvitamin D and alkaline phosphatases and inversely with serum parathyroid hormone and other bone resorption markers. The 44/42 Ca serum ratio positively correlated with age-adjusted z-scores of tibial trabecular bone mineral density and total bone mineral content measured by peripheral quantitative CT, and hip bone mineral density measured by dual-energy X-ray absorptiometry. Significant and independent predictors of total bone mineral content, measured by, were the 44/42 Ca serum ratio and parathyroid hormone. The 44/42 Ca serum ratio, repeated after four weeks, highly correlated with baseline values. When adjusted for calciumcontaining medications and kidney impairment, the 44/42 Ca serum ratio in patients receiving dialysis was 157% lower than that of age-matched children and 29% lower than levels in elderly women with osteoporosis, implying significantly lower bone mineral content. Thus, calcium isotope ratios may provide a novel, sensitive and noninvasive method of assessing bone calcium balance in chronic kidney disease. I n patients with chronic kidney disease (CKD), mineral bone disorder is common and associated with significant morbidity. 1 Dysregulated calcium (Ca) and phosphate homeostasis leads to abnormalities in bone formation and resorption, and is causally associated with reduced bone mineral density (BMD) 2 and an increased risk of fractures. 3 The growing skeleton of children is uniquely vulnerable: 29% in CKD stage 2 and >90% on dialysis have deficient mineralization, 4 manifesting as bone deformities, and a 2 to 3 fold higher fracture risk compared with healthy peers. 3 The assessment of bone health is a major challenge for physicians. 5,6 Although bone biopsy with histomorphometry is considered "gold standard" for assessing bone formation, it is invasive and cannot be easily repeated in children. Bone biomarkers reflect either the bone formation or the bone resorption process but not the net state of bone turnover, nor the bone calcium balance (BCaB). 7 Radiological changes may take months to years to manifest changes in BMD. 7 Dual-energy X-ray absorptiometry (DXA), although widely used, is not a reliable predictor of fracture risk, and is not recommended as a screening tool in patients with CKD. 8 Traditional Ca balance studies administering oral and i.v. Ca isotopes have been performed but are mainly limited to Ca absorption studies. 1,[9][10][11][12] Radioactive isotopes of Ca have been used in mass balance experiments, but these studies are highly laborious, requiring a "metabolic ward" setting, limiting their routine clinical use. Ca balance has been extensively studied in healthy individuals, but only 2 Ca balance studies have been performed in patients with CKD. 13,14 A biomarker of BCaB that is easily measurable, sensitive, and specific and reflects "realtime" changes in bone mineralization and demineralization is required.
Natural Ca isotopes have been extensively studied in marine geological systems (changes in marine calcification reflect global warming) 15,16 and space travel (to study BCaB in lowgravity environments). 17,18 These naturally occurring stable (i.e., nonradioactive) Ca isotopes are present in our diet and incorporated into different body compartments, and they may provide a novel, noninvasive method of assessing BCaB. Naturally occurring nonradioactive Ca isotopes ( 40 Ca, 42 Ca, 43 Ca, 44 Ca, 46 Ca, and 48 Ca) [19][20][21] are present in our food and water 22 and sequestered in different body compartments following distinct rules of kinetic isotope fractionation, depending on their atomic mass. 15 Isotope fractionation is the physicochemical mechanism for separation of heavier and lighter isotopes in at least 2 body compartments, 21 such that isotopically light Ca is preferentially enriched in the course of chemical transport reactions (e.g., from the diet to the skeleton), whereas the heavy isotope is preferentially excreted in urine and feces due to a Rayleigh distillation behavior. [22][23][24][25][26] As a major consequence, each body compartment is characterized by a unique and individual Ca isotope ratio, which reflects age, sex, and health status, as shown in our previous work. 23,24,27 The ratio of Ca isotopes (e.g., when studying 42 Ca and 44 Ca, the ratio would be expressed as d 44/42 Ca) compared with a known standard is described mathematically in a compartment model [22][23][24]26,28  Recently, we studied the naturally occurring Ca isotope ratios in the serum, urine, and feces of healthy individuals and found significantly higher d 44/42 Ca serum values in children compared with young adults. 27 The d 44/42 Ca serum values correlated with height and pubertal stage, reflecting an avid uptake of Ca by the growing skeleton. 27 The d 44/42 Ca serum value correlated positively with biomarkers of bone formation, and inversely with bone resorption markers, and was a significant and independent predictor of tibial cortical BMD measured on peripheral quantitative computed tomography (pQCT) scan, 27 a reliable measure of bone mineral content. 2,29 Our group and others have also studied the sensitivity of natural Ca isotope measures in animal models 15,20,30,31 and healthy adults, 22,24,25,28,[32][33][34][35][36][37] where d 44/42 Ca in serum and urine closely reflected the interventions that alter bone homeostasis, such as complete bed rest, 25 vitamin D supplementation, 35 bisphosphonate treatment, 22 and myelomainduced bone disease. 37 In a rodent model of 6 animals with CKD, the d 44/42 Ca serum values were lower compared with control animals and correlated with BMD. 31 Although disorders of BCaB are common in CKD, 4,8 the Ca isotope fractionation technique has not been applied in this population so far.
We hypothesize that natural Ca isotope fractionation is a sensitive measure of changes in BCaB in children with CKD and on dialysis. We correlated d 44

METHODS
This is a prospective, multicenter study across 5 European pediatric nephrology centers. Children aged <18 years with CKD stages 4-5 (CKD4-5) and those receiving chronic dialysis (hemodialysis [HD] or peritoneal dialysis [PD]) were included. Children with preexisting bone disease (inherited or acquired), those with fractures in the preceding 6 months, those with any acute illness in the preceding 2 weeks, those on glucocorticoid therapy in the preceding year, or those with a lifetime cumulative steroid exposure of >6 months, or those who had received bisphosphonates at any time, were excluded. A total of 134 children (42 in CKD4-5 and 92 on dialysis) were included in the analysis. All participants and/or their caregivers provided informed consent or assent as appropriate. The study is registered on clinicaltrials.gov (NCT03285854) and approved by all local research ethics committees.

Study measures
Participants were asked to complete a 3-day food diary and bring a 24-hour urine sample and a single fecal sample on the day of the study visit. The study visit included anthropometric measures, fasting blood sampling, DXA, and pQCT. In HD patients, all measurements were taken before a midweek dialysis session. Details of anthropometry, biomarker assays, and radiological measures have been described 27 (full details in Supplementary Methods).

Repeatability study
A longitudinal study was conducted in a subset of patients to examine subject variability and test the validity of the Ca isotope technique. In 22 children (7 in CKD5 and 15 on dialysis), we repeated fasting serum and urine tests after a median interval of 24 (interquartile range, [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] days. Patients were selected to ensure that there was no change in the CKD status or urine output (24-hour measures recorded) and no change to their Ca-containing medications or vitamin D prescriptions between the 2 assessments. The intrasubject variability in Ca intake was determined from the 3-day food records and was AE6.3% (interquartile range, 2.1%-9.6%). All biomarkers were measured, but radiological measures were not repeated given the short interval from baseline measures.

Comparison with healthy controls
Ca isotope values in CKD and dialysis patients were compared with our recently published data from age-matched healthy children and young adults 27 as well as with data from a study in elderly women with and without osteoporosis. 24 In both studies, healthy participants underwent similar biochemical tests and radiological investigations, including DXA and pQCT.

Ca isotope measurements
After Ca extraction and chemical purification (details in Supplementary Methods), Ca isotope measurements were performed on a multicollector inductively coupled plasma mass spectrometer (Neptune plus; Thermo Fisher Scientific) at the mass spectrometer facilities of the GEOMAR Helmholtz Centre for Ocean Research, Kiel, Germany. The mass spectrometer was equipped with 9 Faraday cups, of which 8 are moveable. The mass spectrometer was set up to measure masses 42, 43, 43.5, and 44 simultaneously. To suppress interfering Ca and Ar hydrides (e.g., 40Ar1H2 on 42Ca), an APEX IR (ESI) sample introduction system was used. All measurements were performed in medium resolution (m/Dm w 4000) on the interference-free plateau of the low mass side of the peaks. This was achieved by choosing an appropriate center cup mass of 43.687 AE 0.001 atomic mass unit (amu) and verified on a daily basis, as previously described. 24 Instrumental fractionation (mass bias) was corrected by applying the standard-sample-bracketing approach. The measurement of a sample was bracketed by measurements of a w5 mg/ml Ca solution prepared from a 10,000 mg/g Ca inductively coupled plasma reference solution (Ca inductively coupled plasma). Every sample was measured at least 4 times during a session, and the mean value was used for further calculations. The Ca balance model described above is based on the assumptions that the diet contains Ca from natural biological sources (dairy, vegetables, and meat) 23,39 and that kidney function is normal. To compare the d 44/42 Ca serum values of CKD and dialysis patients with healthy controls, adjustments to the measured values were performed (full details in Supplementary Methods and Supplementary Tables S1 and S2).

Baseline characteristics
The median age of the study population was 11.3 (interquartile range, 6.1-15.7) years. There was no difference in sex, race, or underlying kidney disease between groups, but children on dialysis were older and had lower anthropometric measures (Table 1). Children on dialysis had a significantly higher Ca intake from medications and total Ca intake from diet and medications compared with those in CKD4-5 (P ¼ 0.04 and P ¼ 0.02, respectively; Table 1). A total of 91% of children on dialysis received a vitamin D analog compared with 64% with CKD (P ¼ 0.002), but the median vitamin D dosage was comparable between groups. There was a positive association between the cholecalciferol and alfacalcidol dose and d 44/42 Ca serum in CKD and dialysis patients (Supplementary Figure S1). Unlike age-matched healthy children in whom there was a strong correlation between  no correlation was seen in CKD4-5 and dialysis patients. A total of 37 (40%) of the dialysis cohort were anuric.

Serum Ca isotope fractions and biomarkers of mineral bone disease
Children with CKD had a lower serum ionized Ca, phosphate, magnesium, and parathyroid hormone (PTH) compared with those on dialysis (Supplementary Table S3

Ca isotope fractions and radiological measures
There was no significant difference in DXA or pQCT measures between children with CKD4-5 and those on dialysis (Supplementary Table S3). In dialysis patients, both the DXA hip and lumbar spine z-scores correlated positively with Figure 3). In dialysis patients, both the tibial trabecular volumetric BMD z-score and the total bone mineral content z-score (calculated from the trabecular and cortical BMD measured by pQCT) correlated positively with On univariable analysis, significant correlations with the total bone mineral content z-score (in the dialysis cohort) were d 44/42 Ca serum , ionized Ca, ALP, PTH, cross-linked    Figure S2A). There was no difference in d 44/ 42 Ca serum values between dialysis patients who were anuric compared with those with a urine output (P ¼ 0.11; Supplementary Figure S2B Ca serum values at repeated testing correlated with serum ionized Ca (R 2 ¼ 0.34, P ¼ 0.03) and serum ALP (R 2 ¼ 0.29, P ¼ 0.04), and inversely with PTH (R 2 ¼ 0.29, P ¼ 0.01) and tartrate-resistant acid phosphatase 5b (R 2 ¼ 0.43, P ¼ 0.01; in dialysis patients only).

Comparing Ca isotope values of CKD and dialysis patients with age-matched healthy controls
After adjustment for the isotopic composition of Ca-containing medications and kidney impairment, the d 44/42 Ca serum levels were significantly higher in healthy

DISCUSSION
In this proof-of-concept study, we have shown that the ratio of naturally occurring Ca isotopes in serum (d 44/ 42 Ca serum ) is a significant and independent predictor of BCaB and may be a novel biomarker of BCaB in patients on dialysis. The d 44/42 Ca serum values correlated with a wide range of currently available measures of bone mineral status, including bone densitometry and biomarkers of bone formation and resorption used in routine clinical and research practice, but were the strongest predictor of total bone mineral content. Children with CKD and on dialysis had d 44/42 Ca serum values that were 157% (95% confidence interval, 108%-248%) lower than age-matched children 27 (Figures 4 and 5b) and 29% (95% confidence interval, 20%-35%) lower than levels reported even in elderly osteoporotic women, implying a significant loss of bone mineral content.
Mineral bone disorder of CKD causes abnormalities in BCaB, particularly in the growing skeleton. 40 The d 44/42 Ca serum values of our CKD and dialysis patients indicate significant reduction in BCaB compared with healthy age-matched children ( Figure 5b). As shown in our earlier work, the growing skeleton of healthy children avidly absorbs Ca with the highest d 44/42 Ca serum values and an F Bone_Loss /F Bone_Gain ratio of w0.4 (Figure 6), implying that 2.5 times more Ca is accumulated in the bones than lost by bone resorption. 27 Bone mineral accretion continues into the 30s, when peak bone mass (the maximum amount of bone acquired at the end of skeletal development) is reached, 41 and the F Bone_Loss / F Bone_Gain ratio is w1. 27 With increasing age, there is progressive bone resorption leading to a net loss of Ca. 24 The CKD cohort with an average age of 10 years showed a bone resorption/formation ratio comparable to that of healthy 30year-old adults (F Bone_Loss /F Bone_Gain w1; Figure 6). Children on dialysis had an F Bone_Loss /F Bone_Gain ratio below w10,    for total bone mineral content measured by pQCT. We used pQCT as the best available measure of bone mineral content to compare with isotope data. Unlike DXA, pQCT, clearly distinguishing trabecular and cortical bone, has been used to show progressive decrease in bone mineral content with disease severity in CKD patients 29 and predicts fracture risk. 2 Importantly, several limitations of pQCT must be recognized that limit it application (it provides an image of a single bone compared with the whole skeleton, is not widely available, is highly operator dependent, and lacks robust reference data), 6 thereby stressing the need for a more robust marker of bone Ca content. Second, traditional biomarkers only qualitatively reflect either bone formation or bone resorption, but do not indicate the changes in bone Ca content. 5,7 In current clinical practice, DXA is our only tool for estimating BCaB. However, serum Ca accounts for <0.1% of total body Ca, and due to tight negative feedback control, cannot reflect the total body Ca. 1 Although bone histology is considered the "gold standard," this is difficult to perform and cannot be easily repeated. The sensitivity of DXA in predicting changes in BMD is poor: radiology tests capture only a single area of the skeleton and can take months to years to manifest a change. 6,42 As change in BCaB is a dynamic process, BMD measurement by radiography or histology cannot provide information on short-term changes in Ca status, such as with pharmacologic interventions. Thus, Ca isotope measures provide an urgent unmet need in the diagnosis and management of mineral bone disorder of CKD. In future, the predictive value of bone turnover markers coupled with Ca isotope diagnostics to determine the BCaB marker  35 In patients with multiple myeloma, a malignancy that causes an activation of osteoclasts, resulting  Figure S3). Only 0.5 g (i.e., <0.1% of total body Ca) is present in the circulation, and this is exchanged rapidly with the considerably larger Ca reservoir of 1200 to 1500 g in the skeleton, such that Ca has a short residence time in the blood in the order of only about 1 hour (details in Supplementary Methods: the concept of Ca residence time in human blood). Thus, deviations in Ca concentration as well as Ca isotope composition following a dialysis session are rapidly and completely returned to normal within a few hours. The baseline Ca isotope levels (measured at 12 or 48 hours after the previous dialysis session in PD and HD patients, respectively) were comparable across HD and PD cohorts (Supplementary Figure S3). Thus, d 44/42 Ca serum is determined mainly by the interaction with bone Ca fluxes, reflecting the ratio of bone Ca absorption and resorption.
This proof-of-concept study is the first to investigate the use stable Ca isotopes in adults or children with kidney dysfunction, but limitations of our work as well as the outline of future studies to promote our understanding of the utility of isotopes in medicine must be noted. We were unable to evaluate the effect of growth on changes in Ca isotope ratios as this was a cross-sectional study design. The CKD cohort was small and heterogeneous, perhaps explaining the lack of correlations between d 44/42 Ca serum and some biomarkers and radiological measures, prompting the need for further study with larger patient numbers. Although controlled diets or weighed food records are the most rigorous analysis of dietary Ca intake, in routine clinical practice, analysis of 3-day food diaries 43 offers the most accurate assessment. The overall uncertainty of the reported data may be larger because of systematic uncertainties introduced by the corrections for different diets and kidney dysfunction. Ca absorption could not be studied; nevertheless, individual patient data of d 44/ 42 Ca serum already account for their absorbed dietary Ca. We measured DXA hip and lumbar spine rather than "total body less head" DXA scans, in keeping with standard clinical practice and latest guidelines, 44 but acknowledge that variations due to skeletal heterogeneity should be considered. We measured total ALP and not bone-specific ALP, but bonespecific ALP represents 80% to 90% of total circulating ALP in growing children with normal liver function, 45 and guidelines suggest that total ALP can be routinely used, except in patients with liver disease. 5 A crucial question in dialysis patients is to determine the fate of the absorbed Ca: whether it is in the bones or ectopically deposited in extraosseous tissues, like blood vessels; further studies to measure the Ca isotopic composition in bone and arterial biopsy samples collected from the same individual are under way. Unfortunately, phosphorus has only 1 stable isotope, and so the isotope fractionation technique that requires at least 2 isotopes cannot be applied to study phosphorus levels.
In summary, this proof-of-concept study suggests that the naturally occurring stable Ca isotope ratio in serum is a significant and independent predictor of bone Ca content in children with CKD and on dialysis and is more sensitive and accurate than routinely used measures of BMD, such as DXA or bone biomarkers. The clinical utility of Ca isotopes as a biomarker of BCaB may also be applicable in other patient groups who have diseases that affect bone health (such as adults with CKD or on dialysis, chronic childhood diseases, and inherited bone diseases) or receive medications, like steroids, chemotherapy, and antiresorptive treatments, that affect the bone. Future studies that perform serial measurements of Ca isotope ratios may aid prognostication of fracture risk in patients with CKD as well as other disease cohorts.
DISCLOSURE AE and AK are cofounders of Osteolabs GmbH. All the other authors declared no competing interests. Research, Kiel, Germany, is gratefully acknowledged for providing laboratory space and mass spectrometer capacities.   Table S3. Bone biomarkers, calcium (Ca) isotope ratios, and radiological evaluation. Table S4. Multivariable regression analysis for predictors of the total bone mineral content z-score.     Figure S5. A schematic diagram adapted from the calcium (Ca) balance model in Figure 1 for chronic kidney disease (CKD) and dialysis patients, showing the complex processes involved when Ca homeostasis is affected in renal impairment. For CKD patients, the compartment model for healthy individuals (Figure 1) can still be applied, with levels closely correlating with the estimated glomerular filtration rate. However, for dialysis patients, including those who are anuric, the kidney compartment is replaced by a dialysis compartment. Extraosseous calcification is shown as a potential compartment that can account for Ca isotope fractionation. Supplementary References.