Tradeoff-in-the-Nephron: A Theory to Explain the Primacy of Phosphate in the Pathogenesis of Secondary Hyperparathyroidism

Chronic kidney disease (CKD) causes secondary hyperparathyroidism (SHPT). The cardinal features of SHPT are persistence of normocalcemia as CKD progresses and dependence of the parathyroid hormone concentration ([PTH]) on phosphate influx (IP). The tradeoff-in-the-nephron hypothesis integrates these features. It states that as the glomerular filtration rate (GFR) falls, the phosphate concentration ([P]CDN) rises in the cortical distal nephron, the calcium concentration ([Ca]CDN) in that segment falls, and [PTH] rises to maintain normal calcium reabsorption per volume of filtrate (TRCa/GFR). In a clinical study, we set GFR equal to creatinine clearance (Ccr) and IP equal to the urinary excretion rate of phosphorus (EP). We employed EP/Ccr as a surrogate for [P]CDN. We showed that TRCa/Ccr was high in patients with primary hyperparathyroidism (PHPT) and normal in those with SHPT despite comparably increased [PTH] in each group. In subjects with SHPT, we examined regressions of [PTH] on EP/Ccr before and after treatment with sevelamer carbonate or a placebo. All regressions were significant, and ∆[PTH] correlated with ∆EP/Ccr in each treatment cohort. We concluded that [P]CDN determines [PTH] in CKD. This inference explains the cardinal features of SHPT, much of the evidence on which other pathogenic theories are based, and many ancillary observations.


Introduction
Chronic kidney disease (CKD) causes the parathyroid hormone concentration ([PTH]) to rise to abnormally high values. This phenomenon, secondary hyperparathyroidism (SHPT), begins early in the course of CKD and increases in prevalence and severity as the glomerular filtration rate (GFR) falls [1][2][3][4][5]. A secondary skeletal lesion, osteitis fibrosa, evolves with SHPT and presumably contributes to the increased fracture risk of patients with CKD [6,7]. Excessive PTH may also play a role in extraskeletal manifestations of uremia [8,9].
SHPT exhibits two reproducible characteristics: the ionized calcium concentration ([Ca] i ) is consistently physiologic until GFR is severely reduced [1,3], and [PTH] varies directly and substantially with phosphate influx (I P ). In experimental CKD, [PTH] is elevated at customary I P but falls to normal if I P is reduced in proportion to GFR [10][11][12][13][14]. We have not found a reported exception to this rule.
The pathogenesis of SHPT is unresolved. In this paper we present a hypothesis, tradeoff-in-the-nephron, that integrates the primacy of I P with the paradox of normal [Ca] i and high [PTH]. The hypothesis is compatible with evidence on which other pathogenic theories are based, and it illuminates many ancillary observations. We suggest that resistance to the calcemic action of PTH arises in the cortical distal nephron (CDN), where PTH regulates calcium reabsorption [15]. An increased

The Original Tradeoff Hypothesis
Bricker proposed the following sequence of events to explain the role of phosphate in SHPT [42]: intake and gastrointestinal absorption of phosphate continue unabated as nephrons are lost; a temporary rise in plasma phosphate ( [36,38], and oral phosphate raised [PTH] in such patients even though [P] s fell simultaneously [32]. In patients with hypophosphatemia due to impaired phosphate reabsorption, high I P raised [PTH] without correcting [P] s [39]. In vitro, modest increments in [P] p did not reduce [Ca] i [43].

Skeletal Resistance to PTH
As Slatopolsky, Kaplan, and their colleagues were linking SHPT to I P , others focused on the paradox of high [PTH] and normal [Ca] i . A source of calcium seemed resistant to PTH, and the skeleton was assumed to be that source. We have found no evidence that the CDN was considered.
Massry and colleagues measured effects of infused parathyroid extract (PTE) on the serum calcium concentration ([Ca] s ) in humans. PTE raised [Ca] s by more than 1.0 mg/dL in subjects with normal GFR and by approximately 0.5 mg/dL in patients with mild, advanced, or end-stage renal disease [44]. Llach and colleagues examined responses to endogenous PTH by infusing the chelating agent ethylenediaminetetraacetic acid (EDTA); in comparison to control subjects, patients with mild CKD responded to EDTA with more severe hypocalcemia, much higher [PTH], and a more delayed recovery of [Ca] s [45].
Three hypotheses were offered to explain the blunted calcemic response in CKD: a deficiency of 1,25-dihydroxyvitamin D (1,25D) undermined the effect of PTH on osteolysis; circulating phosphate mediated skeletal resistance by an unknown mechanism; and chronically increased [PTH] down-regulated PTH receptors in bone. In dogs made uremic by ureteral ligation or nephrectomy, preliminary administration of 1,25D improved but did not normalize the calcemic response to PTE [46]. Somerville and Kaye found that 1,25D ameliorated PTH resistance in chronic but not acute renal failure [47]; in contrast, phosphate was the agent of resistance when uremia was created by intravenous infusion of urine from intact kidneys [48]. In an isolated rat-tail preparation, the same investigators demonstrated that phosphate could inhibit calcium release from bone [49].
In 5/6 nephrectomized dogs, Kaplan and colleagues observed that neither 1,25D nor phosphate restriction could normalize the calcemic response to PTH even though each intervention restored it partially [50]. Rodriguez and colleagues also achieved partial improvement with these interventions but found that parathyroidectomy restored the calcemic response completely [41,51]. It should be noted that parathyroidectomized animals were maintained with a high-calcium diet post-operatively and a low-phosphate diet during the PTH infusion [51].
We are reluctant to attribute SHPT to skeletal resistance to PTH. If kidneys are functional, and if GFR is assumed to equal creatinine clearance (C cr ), then the flux of calcium into plasma (I Ca ) equals the urinary excretion rate (E Ca ), and the impact of I Ca on [Ca] i is measurable as calcium excreted per volume of filtrate (E Ca /C cr ) [16]. If the skeletal resistance theory is correct, we should not see normal [PTH] when E Ca /C cr is low, or high [PTH] when E Ca /C cr is high. However, we found normal [PTH] despite minimal E Ca /C cr in some control subjects, and high [PTH] despite robust E Ca /C cr in some patients with CKD [16]. Low I Ca did not provoke SHPT at normal GFR, and high I Ca did not prevent it at reduced GFR. We doubt that the skeleton is the principal site of PTH resistance in SHPT.

Deficiency of 1,25-Dihyroxyvitamin D
The active metabolite of vitamin D, 1,25-dihydroxyvitamin D (1,25D), is synthesized throughout the nephron [52]. Its concentration falls as nephrons are lost, and SHPT is widely attributed to this phenomenon [1,2,4,5]. In theory, a reduction in [1,25D] could necessitate a rise in [PTH] by compromising intestinal absorption and tubular reabsorption of calcium [15,35,53], but the preferred explanation for SHPT at present is loss of the suppressive effect of 1,25D on PTH gene transcription [54,55]. This attribute of the metabolite is the basis for treatment of SHPT with vitamin D receptor activators (VDRAs) [56].

Direct Stimulation of PTH Secretion by Circulating Phosphate
In 1996, two groups showed that parathyroid tissue from normal rats secreted PTH in proportion to the phosphate concentration ([P]) in culture medium [57,58]. Two years later, the observation was repeated with hyperplastic tissue from patients with SHPT [59]. Whereas changes in [Ca] i altered [PTH] within one hour [58], changes in [P] did so over 3-5 h [57,58]. [P] did not affect PTH gene transcription [57,59]; observations by Moallem and colleagues suggested indirectly that cytosolic proteins stabilized PTH mRNA in response to high [P] [60].
Evidence of a direct relationship between [P] s and [PTH] was also found in vivo. Takahashi, Slatopolsky, and their colleagues demonstrated strong linear correlations between [PTH] and [P] s in rodents subjected to 5/6 nephrectomy [28,57]. Kates and colleagues confirmed a similar relationship in humans with CKD, but it was demonstrable only in subjects with serum creatinine ([cr] s ) ≤3.0 mg/dL [61]. On some occasions our group also found significant linear regressions of [PTH] on [P] s [18].
We do not doubt that hyperphosphatemia increases PTH synthesis in CKD. However, when kidneys are functional, correlations between [P] s and [PTH] may reflect dependence of both concentrations on a third variable. If E P and TR P are rates of excretion and tubular reabsorption of phosphorus, [P] s equals the sum of E P /C cr and TR P /C cr [19]. E P /C cr quantifies the contribution of I P to [P] s , but it also serves as a mathematical surrogate for [P] CDN , which we believe to be the principal determinant of [PTH] in CKD [17,18]. In patients with Stage 3 and 4 CKD, we found that [PTH] varied directly with E P /C cr and [P] s before administration of sevelamer or a placebo, but with E P /C cr alone after treatment [17]. We therefore attributed the correlation of [PTH] with [P] s to a dependence of both concentrations on E P /C cr [18]. In our study and that of Kates and colleagues, most values of [P] s were in the normal range and were in fact lower than many fasting values of control subjects without SHPT [17,18,61,62]. Consequently, we suspect that [P] CDN , as represented by E P /C cr , determined [PTH] in both studies. When kidneys are functional, the putative effect of [P] s on [PTH] cannot be separated from that of [P] CDN .

Impaired Suppression of the PTH Gene by Fibroblast Growth Factor 23 (FGF23)
FGF23 is a hormone made predominantly but not exclusively by osteocytes [63,64]. In CKD, its concentration is already increased when [PTH] begins to rise [32,65]. Its effects on parathyroid glands and renal tubules are initiated by simultaneous binding to a cognate receptor, FGFR1c, and a co-receptor, the membrane form of klotho [66]. When GFR is normal, FGF23 suppresses transcription of the PTH gene [67], but this action dissipates as GFR falls because FGFR1c and klotho recede in parathyroid tissue [68,69].
PTH and FGF23 reduce proximal tubular phosphate reabsorption by promoting removal of sodium-phosphate co-transporters from the brush border membrane [66], and both hormones increase calcium reabsorption in the distal convoluted tubule [70]. The actions of the two hormones are thought to be integrated at both sites, and both may be required to maintain normal [P] s and [Ca] i in CKD [70][71][72]. In theory, it is possible that the loss of the genomic effect of FGF23 in parathyroid tissue facilitates synthesis of PTH in CKD. It is also possible that the calcium-reabsorbing action of FGF23 promotes reversal of SHPT when I P is reduced in proportion to GFR [12][13][14]29].

Deficiency of 25-Hydroxyvitamin D (25D)
Although definitions of vitamin D insufficiency and deficiency are debated, [25D] ≥30 ng/mL (74.9 nmol/L) is generally accepted as evidence of full repletion [73][74][75][76] A more protracted trial yielded qualitatively similar results [81]. Although ample doses of 25D induce partial reversal of SHPT, vitamin D insufficiency is not the primary cause of SHPT in CKD.

Tradeoff-in-the-Nephron
The ultrafilterable fraction of plasma calcium (Ca uf ) consists of Ca i and a small amount bound to organic anions in complexes [82]. In normal health, [Ca] uf is maintained by influx from the gastrointestinal tract and by tubular reabsorption of filtered calcium. I Ca determines and equals E Ca [16].
The filtration rate of calcium, (GFR)[Ca] uf , is the sum of its excretion and reabsorption rates: (  [16,82]. In our experience, mean [Ca] i of 5.0 mg/dL (1.25 mmol/L) was accompanied by mean [Ca] uf of 5.4 mg/dL. Since I Ca and E Ca fell in tandem with GFR, E Ca /C cr and TR Ca /C cr approximated 0.1 mg/dL and 5.3 mg/dL at any GFR [16].
We used Equation (4) to examine TR Ca /C cr as a function of [PTH] in seven patients with primary hyperparathyroidism (PHPT), 29 patients with CKD (mean MDRD estimated GFR of 29.5 mL/min/1.73 m 2 , range 14-49), and 28 controls with normocalcemia and estimated GFR >60 mL/min/1.73 m 2 [16]. Because of wide dispersion around mean values, [PTH] was not significantly different in PHPT and SHPT even though the 11 highest values in the study were seen in the latter, but concentrations were significantly higher in both of these groups than in controls. Fasting E Ca /C cr , the measurable consequence of calcium influx, was comparable in all three groups. This finding led to the conclusion that increased TR Ca /C cr , not increased I Ca , had caused hypercalcemia in PHPT [16]. Simultaneously, the results showed that [PTH] sufficient to increase TR Ca /C cr in PHPT had maintained normal TR Ca /C cr in SHPT ( Figure 1). We therefore inferred that the CDN is partially resistant to the calcemic effect of PTH in CKD [16].
We reasoned that under conditions of reduced GFR and normal I P (measurable as E P ), the concentration of phosphate in the CDN ([P] CDN ) would be greater than normal, as Bank and colleagues had demonstrated by micropuncture [83]. We hypothesized that high [P] CDN would reduce the availability of Ca for reabsorption through the formation of soluble complexes or crystals, and would, thereby, necessitate increased [PTH] to maintain normal TR Ca /C cr , [Ca] uf , and [Ca] i . We believed that this hypothesis would elucidate the role of phosphate influx in the pathogenesis of SHPT and would explain the persistence of normocalcemia despite high [PTH] in CKD.
Supporting evidence for the hypothesis was available. Tiselius and colleagues had argued that distal tubular filtrate is normally supersaturated with calcium-phosphate compounds, and had shown with in vitro simulations that calcium-phosphate crystals would be the first to form in the CDN after addition of calcium [84,85]. In rats subjected to 3/4 nephrectomy, Haut and colleagues had found that a high-phosphate diet promoted calcium deposition in lumens and cells of cortical nephrons, and had shown that kidney calcium content rose on this diet even if [P] s remained normal [86]. Biopsies had also revealed calcium deposition within CDNs of patients with phosphate-induced acute kidney injury [87]. Most importantly, treatment of SHPT with the calcimimetic agent cinacalcet had reduced [PTH], [Ca] s , and calcium reabsorption, but had not reduced E Ca (or by inference, I Ca ) [71]. Supporting evidence for the hypothesis was available. Tiselius and colleagues had argued that distal tubular filtrate is normally supersaturated with calcium-phosphate compounds, and had shown with in vitro simulations that calcium-phosphate crystals would be the first to form in the CDN after addition of calcium [84,85]. In rats subjected to ¾ nephrectomy, Haut and colleagues had found that a high-phosphate diet promoted calcium deposition in lumens and cells of cortical nephrons, and had shown that kidney calcium content rose on this diet even if [P]s remained normal [86]. Biopsies had also revealed calcium deposition within CDNs of patients with phosphate-induced acute kidney injury [87]. Most importantly, treatment of SHPT with the capable of causing high TR Ca /C cr in patients with PHPT maintained normal TR Ca /C cr in patients with CKD. Reproduced from [16] with permission of the publisher (Dustri-Verlag). E Ca , Urinary excretion rate of calcium, mass/time; C cr , Creatinine clearance (volume/time); TR Ca , Rate of tubular reabsorption of calcium, mass/time; PTH, Parathyroid hormone; CKD, Chronic kidney disease.
We published evidence for the tradeoff-in-the-nephron hypothesis in 2014. Our underlying assumptions were that glomerular filtration of phosphate is virtually complete [88]; I P determines and equals E P at any GFR [19][20][21]25,35]; [P] CDN rises at customary I P as GFR falls [83]; and increased [P] CDN promotes complexation of Ca as described above [84][85][86][87]. For simplicity, we also assumed that delivery of filtered phosphate to the CDN equals E P even though phosphate may be secreted into the distal nephron in CKD [83,89].
Twenty-nine patients with eGFR of 14-49 mL/min/1.73 m 2 participated in a study designed to examine the tradeoff-in-the-nephron hypothesis [17]. They were seen in a research clinic on five occasions, each separated by four weeks. Informed consent was obtained at the first visit, and patients who were taking intestinal phosphate-binding agents discontinued them at that time. A course of cholecalciferol was prescribed at the second visit to minimize any possible contribution of vitamin D deficiency to SHPT. Patients were instructed in a phosphate-restricted diet at the third visit and were asked to continue the diet through the end of the study. At the fourth visit, subjects were randomly assigned to a course of sevelamer carbonate or placebo with meals. Metabolic studies obtained at this visit revealed that the dietary instruction had been ineffective. Results of the therapeutic trial were ascertained at the fifth visit.
We argued algebraically that E P /C cr is proportional to [P] CDN and hypothesized that [PTH] would therefore vary directly with E P /C cr [17]. The purpose of sevelamer carbonate administration was to reduce this ratio. ∆E P /C cr was negative in all sevelamer recipients, and the mean change was −0.5 ± 0.1 mg/dL. In placebo recipients, ∆E P /C cr was evenly distributed over a range of positive and negative values, and the mean change was 0.04 ± 0.12 mg/dL. We interpreted dispersion around this mean as evidence of random variation in phosphate intake.
In both groups, we found significant linear regressions of [PTH] on E P /C cr and of ∆[PTH] on ∆E P /C cr after treatment (Figure 2). Sevelamer recipients in whom ∆[PTH] did not vary with ∆E P /C cr tended to have extremely low E Ca /C cr . The results supported the hypothesis that high [P] CDN necessitates high [PTH] to achieve normal TR Ca /C cr , and also suggested that sufficient [Ca] CDN is essential to the salutary effect of reduced I P on [PTH] [17]. where "∆" = change during treatment. All regressions are statistically significant. Adapted from [17] with permission of the publisher (Dustri-Verlag). EP, urinary excretion rate of phosphorus, mass/time; Ccr, creatinine clearance, volume/time.

Compatibility of Tradeoff-in-the-Nephron with Existing Data
Tradeoff-in-the-nephron is a straightforward hypothesis.  on E P /C cr before and after administration of sevelamer carbonate for four weeks. Graphs (b) and (d) show the same regressions before and after administration of a placebo for four weeks. Graphs (e) and (f) show regressions of ∆[PTH] on ∆E P /C cr in the sevelamer and placebo groups, respectively, where "∆" = change during treatment. All regressions are statistically significant. Adapted from [17] with permission of the publisher (Dustri-Verlag). E P , urinary excretion rate of phosphorus, mass/time; C cr , creatinine clearance, volume/time.

Compatibility of Tradeoff-in-the-Nephron with Existing Data
Tradeoff-in-the-nephron is a straightforward hypothesis. It states that high [P] CDN reduces [Ca] CDN by complexation and thus necessitates high [PTH] to maintain normal calcium reabsorption.
In theory, calcium, 1,25D, or phosphate could affect the synthesis and release of PTH in CKD. Of these, only calcium regulates immediate secretion of stored hormone through its interaction with the membrane calcium receptor [90]. If I P affects [PTH] by determining calcium availability for reabsorption, then changes in I P should alter [PTH] quickly. In vivo and in vitro studies have confirmed this expectation [24,30,31,58].
Tradeoff-in-the-nephron explains why [PTH] was high as long as E P /C cr was high [23,26] and low as long as E P /C cr was low [34]. The hypothesis explains why [PTH] fell with I P while hyperphosphatemia persisted [27,40]. It accounts for the chronicity of SHPT in CKD, in which [P] CDN is continuously increased at normal I P [83]. The hypothesis explains why [PTH] correlated with E P but not [P] s in early CKD [91], and with E P /C cr but not [P] s after administration of sevelamer or placebo [17]. It accounts for increased calciuria despite high [PTH] after an oral bolus of phosphate [32]. It explains why [PTH] was elevated in patients with mild CKD, normal I P and low-normal [P] s [35,36], and why [PTH] rose after a bolus of phosphate even though [P] s fell simultaneously [32]. The hypothesis provides a mechanism for high [PTH] in response to high I P despite persistent hypophosphatemia [39]. Most importantly, it predicts normalization of [P] CDN , E P /C cr , and [PTH] when I P is reduced in proportion to GFR [10][11][12][13][14]17,24,[28][29][30]50].
The principal alternatives to tradeoff-in-the-nephron involve skeletal resistance to the calcemic action of PTH, the effect of 1,25D to suppress transcription of the PTH gene, and direct stimulation of PTH synthesis and secretion by circulating phosphate. Much of the evidence for these theories is compatible with our hypothesis. In subjects with functioning kidneys, 1,25D could have enhanced the calcemic response to PTH through its independent effect on calcium reabsorption in the CDN [15]. In addition to limiting calcium egress from bone [41,49], phosphate could have introduced resistance to PTH in the CDN by the mechanism implied in our hypothesis. Instead of making bone more sensitive to PTH, parathyroidectomy could have necessitated a diet that ensured maximal calcium reabsorption from the CDN in response to the hormone [51].
Recurrent themes emerge from studies of the calcemic response to PTH. Typically, the magnitude of the response was less at reduced than at normal GFR, and preparatory phosphate restriction or 1,25D administration mitigated but did not eliminate this difference [38,41,[44][45][46][47]50,51]. A notable exception occurred when I P was brought to zero in a model of uremia that left kidneys intact; in that instance, the calcemic response was restored completely [48]. These observations make sense if PTH acted on the CDN as well as the skeleton to raise [Ca] s . When filtrate contained no phosphate, a full complement of nephrons permitted a normal response to PTH even though experimental animals were uremic [48]. In other studies, a deficit of nephrons imposed a limit on the response to PTH that neither phosphate restriction nor 1,25D could overcome [38,41,[44][45][46][47]50,51].
The premise that the CDN is the site of PTH resistance is also supported by effects of the calcimimetic agent cinacalcet. In patients with Stage 3 and 4 CKD, the drug reduced [PTH] by 43.1%, but simultaneously kept mean [Ca] s between 8.5 and 9.0 mg/dL even though E Ca rose or remained unchanged [71]. Since I Ca determined E Ca , and since I Ca and TR Ca maintain [Ca] uf at a given GFR [16], it follows that reduction of [PTH] with cinacalcet led to reduction of TR Ca /GFR. High [PTH] was apparently required for reabsorption sufficient to maintain normocalcemia [71].
The capacity of VDRAs to suppress PTH gene transcription can be exploited before ESRD is reached [77,92], but efficacy of the intervention does not confirm reversal of pathogenesis. If deficiency of 1,25D were the cause of SHPT, then normal [PTH] would be incompatible with low [1,25D] in CKD. Numerous investigators have documented this combination after sufficient reduction of I P [13,14,33,34,40,41], and tradeoff-in-the-nephron explains why the combination is possible.
E P /C cr is a determinant of [P] s , and [P] s is a linear function of E P /C cr in CKD [18,20,28,57,61]. At the same time, E P /C cr is approximately proportional to [P] CDN [17,18]. If [PTH] varies directly with [P] s in vivo, the reason may be that [PTH] also varies directly with E P /C cr . We suggest that this confounding association is responsible for correlations between [PTH] and [P] s in Stage 3 and 4 CKD [18,61].

Therapeutic Implications of Tradeoff-in-the-Nephron
Tradeoff-in-the-nephron implies that [PTH] is normal if [P] CDN is normal. E P /C cr is our surrogate for [P] CDN . Since I P determines E P , a reduction of I P in proportion to GFR yields normal E P /C cr . Proportional reduction of I P was precisely the intervention that prevented and reversed SHPT in animal models of CKD [10][11][12][13][14]50]. It follows that normalization of E P /C cr should do the same for patients with SHPT.
In the 1980s and 1990s, European investigators employed severe dietary phosphate restriction to reduce [PTH] in patients with CKD [33,34,40]. Today, in the United States, a similar result requires a drastic revision of eating habits, including avoidance of phosphate preservatives [93,94]. This effort is necessary because to date, the most successful human studies of intestinal phosphate binders have reduced E P by 25%-50% and [PTH] by 13%-35% [37,[95][96][97][98]. Our theory and many animal studies suggest that E P /C cr must be reduced to normal to reverse SHPT completely; if GFR has been reduced by 80%, E P must be reduced by 80%. In addition to diet and binders, blockade of sodium-hydrogen exchanger 3 (NHE3) and inhibition of the intestinal sodium-phosphate 2b co-transporter may ultimately be required to lower I P sufficiently [99,100]. Our experience suggests that normal E Ca /C cr must also be established [17,91]. Attainment of [25D] >30 ng/mL may reduce [PTH] modestly, but we endorse it for other reasons [74]. We presume that normalization of [PTH] is desirable, but concede that the point is debatable [101].

Conclusions
Discordant empiric observations undermine each of the major theories concerning the pathogenesis of SHPT. We have sought a unifying explanation for the two most consistent features of the syndrome, which are dependence of [PTH] on I P and persistence of normal [Ca] i until CKD is far advanced. Tradeoff-in-the-nephron accounts for these features. The hypothesis also provides alternate explanations for much of the evidence on which other theories are based, and it sheds light on numerous ancillary observations. It traces SHPT to high [P] CDN and predicts normal [PTH] at normal E P /C cr . An abundance of evidence is consistent with this prediction. The veracity of tradeoff-in-the-nephron is testable in patients by rigorous but feasible interventions.