Impaired Growth Plate Maturation in XLH Is due to Both Excess FGF23 and Decreased 1,25-Dihydroxyvitamin D Signaling

Abstract

X-linked hypophosphatemia (XLH) is the most common form of hereditary hypophosphatemic rickets. The genetic basis for XLH is loss of function mutations in the phosphate-regulating endopeptidase X-linked (PHEX), which leads to increased circulating fibroblast growth factor 23 (FGF23). This increase in FGF23 impairs activation of vitamin D and attenuates renal phosphate reabsorption, leading to rickets. Previous studies have demonstrated that ablating FGF23 in the Hyp mouse model of XLH leads to hyperphosphatemia, high levels of 1,25-dihydroxyvitamin D, and is not associated with the development of rickets. Studies were undertaken to define a role for the increase in 1,25-dihydroxyvitamin D levels in the prevention of rickets in Hyp mice lacking FGF23. These mice were mated to mice lacking Cyp27b1, the enzyme responsible for activating vitamin D metabolites, to generate Hyp mice lacking both FGF23 and 1,25-dihydroxyvitamin D (FCH mice). Mice were fed a special diet to maintain normal mineral ion homeostasis. Despite normal mineral ions, Hyp mice lacking both FGF23 and Cyp27b1 developed rickets, characterized by an interrupted, expanded hypertrophic chondrocyte layer and impaired hypertrophic chondrocyte apoptosis. This phenotype was prevented when mice were treated with 1,25-dihydroxyvitamin D from day 2 until sacrifice on day 30. Interestingly, mice lacking FGF23 and Cyp27b1 without the PHEX mutation did not exhibit rickets. These findings define an essential PHEX-dependent, FGF23-independent role for 1,25-dihydroxyvitamin D in XLH and have important therapeutic implications for the treatment of this genetic disorder.

X-linked hypophosphatemia (XLH; OMIM# 307800) is the leading cause of inherited hypophosphatemic rickets, with an incidence rate of 1 in 20 000 (1). Individuals with XLH exhibit low serum phosphate levels because of increased circulating fibroblast growth factor 23 (FGF23). FGF23 acts as a phosphaturic factor by decreasing the expression of the type IIa sodium-phosphate co-transporter on the apical membrane of proximal renal tubular cells. This impairs renal phosphate reabsorption leading to hypophosphatemia (2). Rickets in Hyp mice and other hypophosphatemic animal models is secondary to expansion of the hypertrophic chondrocyte (HC) layer of the growth plate resulting from impaired phosphate-induced HC apoptosis (3, 4). In addition to impairing renal phosphate reabsorption, increased circulating FGF23 leads to decreased circulating 1,25-dihydroxyvitamin D (1,25D) levels by decreasing 1-alpha hydroxylase (Cyp27b1) (5) the enzyme that converts vitamin D into its active form 1,25D, as well as by increasing Cyp24a1 (5), the enzyme that inactivates vitamin D metabolites.

The genetic basis of XLH is inactivating mutations in the phosphate-regulating endopeptidase X-linked (PHEX) gene, which is primarily expressed in osteoblasts and osteocytes. How this leads to increases in FGF23 is not known, nor is it known whether this mutation has FGF23-independent effects on the growth plate. The naturally occurring PHEX mutation in the Hyp mice phenocopies human XLH: Hyp mice have similar skeletal, mineral ion, and hormonal abnormalities as humans with XLH, including increased FGF23, hypophosphatemia, and impaired activation of 1,25D (6). Blocking FGF23 activity, using either pan-FGFR inhibitors (7), neutralizing antibodies (8-10), or by knocking out FGFR1 (11), FGF3/4 (12), or the FGF23 co-receptor α-Klotho (13), increases serum phosphate and circulating 1,25D levels, and improves the Hyp skeletal phenotype. Mice lacking type IIa sodium-phosphate co-transporter exhibit a similar degree of hypophosphatemia as seen in other hypophosphatemic murine models, but increased endogenous 1,25D signaling rescues rickets in these mice despite the persistent hypophosphatemia (14). Consistent with this, skeletal defects in Hyp mice are significantly attenuated by treatment with 1,25D (15) or 1,25D analogs (16). To dissect the relative roles of 1,25D and FGF23 in the Hyp mouse growth plate phenotype, Cyp27b1 was deleted from Hyp mice lacking FGF23 to generate Hyp mice lacking both FGF23 and 1,25D (FCH mice).

Materials and Methods

Animal Studies

Animal studies were approved by the Institutional Animal Care and Use Committee. All mice were on a C57BL/6J genetic background, housed in a virus and parasite-free barrier facility with a 12-hour light/dark cycle. The following mouse lines were used for these studies (Fig. 1A): wild-type (WT), Cyp27b1 null mice (CYP) (17), FGF23 null mice (MMRRC, 036748-JAX), and Hyp mice (JAX, 000528). These mice were mated to obtain CYP-HYP (CH), FGF-HYP (FH), FGF CYP (FC), and FGF-CYP-HYP (FCH) mice.

Figure 1.

Effects of 1,25D on the body weight, femur length, and serum ions in FCH mice. Mouse breeding scheme (A) WT: Wild type, FGF: FGF23KO, Cyp: Cyp27b1 KO, Hyp: Hyp, CH: Cyp; Hyp, FH: FGF; Hyp, FC: FGF; Cyp and FCH: FGF; Cyp; Hyp. Body weight (BW) (B), femur length (C), and serum calcium (D) and phosphate (E) levels were evaluated D30 from WT, Cyp27b1 (CYP) null mice, fibroblast growth factor 23 (FGF-23) null mice, Hyp mice (HYP), CYP-HYP double knockout mice (CH), FGF-HYP double knockout mice (FH), FGF-CYP double knockout mice (FC), FGF-CYP-HYP triple knockout mice (FCH), and 1,25D-treated FCH mice (FCH + D). The squares indicate males, and the dots indicate female data points in the graphs. Data represent the mean ± SEM. The asterisk above the bars represents a P value < .005.

All mice lacking Cyp27b1 were weaned on day 18 onto a rescue diet enriched with calcium, phosphorus, and lactose to maintain normal mineral ion levels (10, 18). Other mice were maintained on house chow. To evaluate the effect of restoring 1,25D signaling in FCH mice, mice were injected with 1,25D subcutaneously (44 pg/g of body weight) every 36 hours (Calcitriol, Akron NDC 17478-931-01) or an equal volume of vehicle (saline) from day 2 until the time of sacrifice on day 30. This dose and dosing interval was used to prevent 1,25D-related toxicity in mice lacking FGF23 because they have decreased inactivation of vitamin D metabolites resulting from low expression of Cyp24a1.

Serum Mineral Ions

At the time of sacrifice on day 30, serum was collected to evaluate mineral ion levels. The Calcium (CPC) Liqicolor detection kit (STANBIO Texas, USA; Cat# 0150-250), and Phosphate Assay Kit (Abcam, Cambridge MA, USA; ab65622) were used, following the manufacturers’ protocols.

Histological Analyses

Tibiae were fixed overnight in neutral-buffered formalin at 4°, decalcified using 20% EDTA, and processed for paraffin sectioning. Hematoxylin and eosin staining (H&E) and immunohistochemistry (IHC) were performed on 5-µm-thick paraffin sections. To evaluate the expression of Collagen type-X (Coll-X), 2 mg/mL hyaluronidase (Sigma H-3506, St Louis, MO)-based antigen retrieval was performed at 37 °C for 30 minutes, before incubation with anti-Coll-X (ABclonal, Woburn MA; Cat# A6889, 1:300, RRID:AB_2767448) at 4 °C overnight (19). Bone Sialo Protein IHC was performed after antigen retrieval using 2 mg/mL hyaluronidase at 37 °C for 30 minutes, before overnight incubation anti-BSP (Invitrogen, Cat# AB_2746540, 1:300, RRID:AB_2746540). For osteopontin IHC, antigen retrieval was performed using 0.05% trypsin solution at 37 °C for 30 minutes before overnight incubation with anti-Opn (Proteintech, Cat#229521-AP, 1:200, RRID:AB_2783651).

Signals were detected with anti-Rabbit-HRP (Cell Signaling, Danvers MA; Cat# 7074, 1:500, RRID:AB_2099233) and DAB-HRP Kit (Vector Laboratories, Burlingame, CA; Cat# 4100). For safranin-O and fast green staining, mouse tibial paraffin sections were stained with 0.02% fast green solution and 0.1% safranin-O following a standard protocol.

The terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was used to evaluate the number of mature HCs undergoing apoptosis in the growth plates. Paraffin sections were permeabilized with 20 μg/mL proteinase K for 30 minutes at 37 °C. Apoptotic cells were detected using the In Situ Cell Death Detection Kit (Roche, Millipore Sigma, St Louis MO; Cat# 11684795910), and sections were mounted using Fluoroshield mounting media with DAPI (Millipore Sigma, USA; Cat. #F6057). TUNEL and DAPI double-positive HCs were counted in the distal 2 rows of the HC layer.

Statistical Analyses

Data represent the mean ± standard error of the mean. Statistical significance was determined by ANOVA (using Prism version 9.2.0, GraphPad Software, San Diego, CA, USA). P values < .05 were considered statistically significant.

Results

Mouse Phenotypes

Previous investigations have demonstrated that impairing FGF23 action in Hyp mice prevents rickets (7, 8, 11-13, 20). However, this decrease in FGF23 signaling is associated with a significant increase in 1,25D, thus raising the question as to the relative contribution of impaired FGF23 action vs increased 1,25D signaling in the prevention of rickets in Hyp mice. To address this, Hyp mice were mated to mice lacking the 1,25D 1 α hydroxylase to generate Cyp27b1 knockout (KO)/Hyp mice (CH) and to mice heterozygous for FGF23 ablation to generate FGF23 KO/Hyp mice (FH). These mice were mated to generate Hyp mice lacking both Cyp27b1 and FGF23 (FCH) (Fig. 1A). On sacrifice day 30, body weight, femoral length, and serum calcium and phosphate were evaluated. As shown in Fig. 1B-E, Hyp mice are smaller than WT controls with a reduced femoral length and exhibit normocalcemia and hypophosphatemia. When maintained on a lactose-enriched high-calcium/high-phosphate diet to circumvent abnormalities in intestinal calcium absorption, mice lacking Cyp27b1 exhibit normal body weight, femur length, and mineral ion levels, as previously reported ((21, 22) and Fig. 1B-1E). Consistent with other studies (23, 24), FGF23 KO mice exhibited a lower body weight and shorter femoral length, associated with hypercalcemia and hyperphosphatemia (Fig. 1B-1E).

Genetic deletion of Cyp27b1 on the Hyp background (CH) normalizes body weight but does not prevent the shorter femur length and hypophosphatemia seen in Hyp mice (Fig. 1B-1E, CH), whereas deletion of FGF23 in the Hyp background (FH) phenocopies the FGF23 KO phenotype including decreased body weight, shorter femur length, hypercalcemia, and hyperphosphatemia ((24, 25) and Fig. 1B-1E). However, abolishing 1,25D signaling in FGF23 KO mice by ablating the vitamin D receptor (26) or Cyp27b1 ((27) and Fig. 1B-1E, FC) normalizes their phenotype demonstrating that impaired growth and mineral ion homeostasis in FGF23 KO mice are associated with enhanced 1,25 signaling. In contrast, ablation of Cyp27b1 and FGF23 in Hyp mice (FCH) does not rescue the lower body weight or reduced femur length despite normalizing mineral ion levels (Fig. 1B-1E, FCH), suggesting that, in Hyp mice, 1,25D has effects on growth that are independent of FGF23 and hypophosphatemia.

Growth Plate Phenotype

As previously reported, Hyp mice exhibit rickets, characterized by a disorganized and expanded HC layer, and the growth plates of the FGF23 KO and FH mice, which exhibit hypercalcemia and hyperphosphatemia, did not exhibit rickets (Fig. 2 and Table 1).

Figure 2.

Histological analyses of P30 tibial growth plates. (A) Hematoxylin and eosin (H&E) staining. (B) Immunohistochemistry (IHC) for Collagen type X (Coll-X), a marker of hypertrophic chondrocytes. Brackets on H&E and Coll-X IHC images indicate the hypertrophic chondrocyte layer. Black arrows on H&E, and Coll-X IHC indicate the interrupted hypertrophic zone of the growth plate of FCH mice. Wild-type (WT), Cyp27b1 (CYP) Cyp27b1 null mice, fibroblast growth factor 23 (FGF-23) knockout mice, Hyp mice (HYP), CYP-HYP double knockout mice (CH), FGF-HYP double knockout mice (FH), FGF-CYP double knockout mice (FC), FGF-CYP-HYP triple knockout mice (FCH), and 1,25D-treated FCH mice (FCH + D) were analyzed at day 30. Data are representative of that from of at least 4 mice per genotype/treatment group. The scale bar represents 100 μM.

Table 1.

Quantification of the number of HCs per column and of apoptotic HCs

	WT	CYP	FGF23 KO	HYP	CH	FH	FC	FCH	FCH + D
# HC/column	4.04 ± .52	4.62 ± .69	2.75 ± .45b	13.34 ± 2.15a	13.26 ± 1.15a	3.66 ± .50	4.46 ± 1.26	9.18 ± 1.53a	3.41 ± .33
# Apoptotic HCs	13.37 ± 1.22	12.83 ± 1.46	20.82 ± 1.88a	3 ± 1.41a	3.14 ± 1.36a	12.4 ± .49	7 ± 1.04	3.2 ± 1.63a	11 ± .46

Data represent the mean and SD that obtained from at least 5 mice per genotype/treatment group.

Abbreviations: WT, wild-type; CYP, Cyp27b1 knockout mice; FGF-23, fibroblast growth factor 23 knockout mice; HYP, Hyp mice; CH, CYP-HYP double knockout mice; FH, FGF-HYP double knockout mice; FC, FGF-CYP double knockout mice; FCH, FGF-CYP-HYP triple knockout mice; FCH + D, 1,25D-treated FCH mice.

^a P < .005 vs WT.

^b P = .466 vs WT.

Cyp27b1 knockout mice fed a high calcium, lactose-enriched diet have a normal growth plate (Fig. 2). However, this diet is not able to prevent rickets in Hyp mice lacking Cyp27b1 (CH) (Fig. 2 and Table 1). To address if this was due solely to persistent hypophosphatemia associated with the increased circulating FGF23 observed in Hyp mice, the growth plate of CH mice lacking FGF23 was examined (FCH). Consistent with the decrease in femoral length observed in these FCH mice, their growth plates exhibited a rachitic growth plate characterized by expansion of the HC layer despite normal mineral ion homeostasis. Of note, the HC layer of the FCH growth plate was interrupted by cells that did not express Collagen type X (black arrow in Fig. 2B, FCH). Mice lacking both FGF23 and Cyp27b1 (FC), who did not have the PHEX mutation, exhibit a normal growth plate phenotype.

To further analyze the growth plate defect in FCH mice, Safranin O staining was performed, which showed absence of cartilage proteoglycans in the interrupted region of the growth plate where Collagen type X expression was absent (Fig. 3). IHC was performed to evaluate the expression of the SIBLING proteins bone sialo protein (BSP) and osteopontin (Opn). Reduced expression of both BSP and Opn was observed in FCH growth plates, demonstrating that overexpression of these proteins is not responsible for the phenotype observed.

Figure 3.

Histological analyses of P30 tibial growth plates. Safranin-O and fast green staining for proteoglycans (A). Immunohistochemistry (IHC) for bone sialo protein (BSP) and osteopontin (OPN), in day 30 growth plates of WT, FCH, and FCH + D mice. Black arrows, on safranin-O and IHC images, indicate the interrupted growth plate of FCH mice. Wild-type (WT), FGF-CYP-HYP triple knockout mice (FCH), and 1,25D-treated FCH mice (FCH + D). The scale bar represents 100 μM. Data are representative of 3 mice per genotype/treatment group.

Evaluation of Hypertrophic Chondrocyte Apoptosis

Previous studies have demonstrated that impaired phosphate-mediated hypertrophic chondrocyte apoptosis leads to the expansion of the HC layer in hypophosphatemic mice (3, 4). Consistent with this, the hypophosphatemic Hyp and CH mice exhibited a significant reduction in TUNEL-positive HCs in their growth plates (Fig. 4 and Table 1). However, the normophosphatemic FCH mice also exhibited a reduction in TUNEL-positive HCs (Fig. 4), demonstrating that 1,25 signaling is required for the normalization of HC apoptosis in the FH mice.

Figure 4.

Hypertrophic chondrocytes apoptosis. White arrows indicate the TUNEL-labeled late HC in the growth plates from day 30 mice. Nuclei are stained blue with DAPI. Wild-type (WT), Cyp27b1 (CYP) null mice, fibroblast growth factor 23 (FGF-23) knockout mice, Hyp mice (HYP), CYP-HYP double knockout mice (CH), FGF-HYP double knockout mice (FH), FGF-CYP double knockout mice (FC), FGF-CYP-HYP triple knockout mice (FCH), and 1,25D-treated FCH mice (FCH + D). Data are representative of that from at least 4 mice per genotype/treatment group. The scale bar represents 100 μM.

Restoring 1,25D Signaling Prevents Rickets in Hyp Mice Lacking FGF23 and Cyp27b1

Based on the observation that preventing increases in 1,25 signaling in Hyp mice lacking FGF23 (FH) leads to rickets (FCH), studies were performed to address the hypothesis that 1,25D has FGF23-independent effects on growth plate maturation in Hyp mice. To this end, FCH mice were treated with 1,25D (44 pg/g every 36 hours) from day 2 until sacrifice day 30. This normalized their body weight and femur length and did not alter mineral ion levels relative to untreated FCH mice (Fig. 1, FCH + D). Histological analysis, IHC for Coll-X and Safranin-O staining demonstrated normalization of the growth plate abnormalities in FCH mice treated with 1,25D (Fig. 2A-B and Fig. 3, FCH and FCH + D). This treatment led to a decrease in BSP expression and an increase in Opn expression in the FCH growth plate (Fig. 3), consistent with the known effects of 1,25D on suppression of BSP gene expression (28) and induction of Opn expression (29). Treatment of FCH mice with 1,25D also restored hypertrophic chondrocyte apoptosis (Fig. 4 and Table 1, FCH and FCH + D).

Discussion

The growth plate defects observed in Hyp mice are thought to be secondary to FGF23-induced hypophosphatemia, which leads to impaired HC apoptosis and rickets. Inhibition of FGF23 activity in Hyp mice (FH) rescues the rachitic phenotype (7, 8, 11-13, 20). However, deletion of FGF23 in Hyp (FH) mice leads to hyperphosphatemia, hypercalcemia, and increased circulating 1,25D levels, raising the question as to the relative roles of 1,25D signaling, increases in mineral ions, and lack of FGF23 in preventing the rachitic phenotype in FH mice. The current studies establish a role for impaired 1,25D signaling, independent of phosphate, in the pathogenesis of rickets in Hyp mice because, despite having normal serum Ca and Pi levels, FCH mice exhibit a rachitic phenotype with an interrupted HC layer, a phenotype that is normalized by 1,25D treatment. Impairing 1,25D signaling in VDR or Cyp27b1 knockout mice does not lead to rickets in the setting of normal mineral ion levels (21, 30). Similarly, deletion of Cyp27b1 in mice lacking FGF23 (FC) does not impair growth plate maturation unless they are in the Hyp background (FCH). These data suggest that 1,25D exerts critical, FGF23-independent effects to maintain growth plate homeostasis in the setting of a PHEX mutation.

Although the major effect of 1,25 on growth plate maturation in WT mice is thought to be the maintenance of normal mineral ion homeostasis (30), 1,25D also has been shown to have direct effects on mediators of HC apoptosis in vitro. Studies in primary cultured HCs demonstrate that 1,25D treatment increases both basal and phosphate-induced mitochondrial ERK1/2 phosphorylation, which is essential for the activation of Caspase-9 mediated apoptosis in HCs (10). Recent investigations have demonstrated that phosphate activates this mitochondrial apoptotic pathway in HCs by activating vascular endothelial growth factor (VEGF) receptor 2 signaling, and that matrix metalloprotease (MMP)-dependent release of VEGF from the matrix is required for this effect (15). These findings are of particular interest in that 1,25D has been shown to increase the expression of VEGF and of MMPs in chondrocytes (31-33).

Although these studies help to elucidate how 1,25D preserves normal growth plate maturation in the setting of low serum phosphate, they do not explain why mice lacking both FGF23 and Cyp27b1 develop rickets only if they are in the Hyp background. This novel observation suggests that 1,25D attenuates the consequences of PHEX mutation that are distinct from its effects on FGF23. There has been significant interest in the pathogenic effect of the high levels of acidic serine and aspartate-rich motif (ASARM) peptides derived from Small Integrin Binding Ligand N-linked Glycoproteins proteins (MEPE, dentin matrix protein-1, BSP, dentin sialophosphoprotein, and osteopontin) observed in XLH. These peptides have been implicated in both the hypophosphatemic and skeletal phenotypes of XLH. Phosphorylation of ASARMs by FAM20C has been shown to be important for their actions (34). Because 1,25D treatment of Hyp mice decreases FAM20C (10), it is unlikely that 1,25D exerts its beneficial effects in XLH by inhibiting ASARM phosphorylation. However, the levels of the MEPE-derived ASARM peptide in dentin extracts were shown to be lower in patients with XLH who were treated with 1,25D and phosphate during growth (35), suggesting that 1,25D may inhibit the expression of the parent SIBLING proteins. Notable in this respect, 1,25D has been shown to decrease the expression of dentin matrix protein-1 (36) and BSP (28); however, BSP levels were not found to be increased in the FCH growth plate (Fig. 3). 1,25D increases the expression of FGF23 (10), dentin sialophosphoprotein, and osteopontin (29). Consistent with this, although a decrease in osteopontin was observed in the FCH growth plates, 1,25D treatment led to increased expression along with normalization of the growth plate phenotype. Thus, misexpression of parent SIBLING proteins is unlikely to underly the abnormal growth plate in FCH mice; however, these studies are unable to directly assess levels of cleaved ASARM peptides. Although 1,25D inhibits PHEX expression (37), it is possible that it induces the expression of other phosphatases that can compensate for impaired PHEX activity, thereby attenuating the FGF23-independent XLH phenotype.

A major strength of these studies is the use of a germline deletion approach, which enables a complete knockout of FGF23 and Cyp27b1 in the mouse model of XLH. This allows investigations to define a role for these 2 hormones in growth plate maturation in XLH. In addition, restoration of 1,25 D signaling in this background allows confirmation of phenotype(s) that are directly attributable to the absence of 1,25D. However, because these studies use global knockouts, they do not permit the identification of the precise tissue in which 1,25D signaling exerts its effects. Consistent with findings in humans and mice with PHEX mutations and its X-linked dominant mode of inheritance, no sexual dimorphism was observed despite the fact that the affected females express a normal PHEX allele in addition to the mutant allele.

These studies define an essential, FGF23-independent, and phosphate-independent effect of 1,25D signaling in the Hyp growth plate. The novel finding that, despite the absence of FGF23 and normal mineral ion levels, Hyp mice lacking Cyp27b1 (FCH) develop rickets, suggests that 1,25D attenuates effects of the PHEX mutation that are unrelated to FGF23 overexpression. These findings have important clinical implications and support the optimization of 1, 25D signaling to achieve maximum therapeutic benefit in XLH. Although current treatment approaches focus on blocking FGF23 signaling with an antibody, this treatment improves but does not cure rickets (9). Similarly, the ability of this treatment to increase 1,25D levels wanes over time. Thus, our studies suggest that complementary strategies directed at inhibiting FGF23 action, and optimizing 1,25D signaling would be expected to lead to improved outcomes in growing children with XLH.

Abbreviations

1,25D

1,25-dihydroxyvitamin D

ASARM

acidic serine and aspartate-rich motif

BSP

bone sialo protein

CH

Cyp27b1 knockout/Hyp

Cyp27b1

1-alpha hydroxylase

FC

lacking both FGF23 and Cyp27b1

FCH

Hyp mice lacking both FGF23 and Cyp27b1

FGF23

fibroblast growth factor 23

FH

FGF23 knockout/Hyp

H&E

hematoxylin and eosin

HC

hypertrophic chondrocyte

IHC

immunohistochemistry

KO

knockout

Opn

osteopontin

PHEX

phosphate-regulating endopeptidase X-linked

TUNEL

terminal deoxynucleotidyl transferase dUTP nick end labeling

VEGF

vascular endothelial growth factor

WT

wild-type

XLH

X-linked hypophosphatemia

Funding

This work was funded by National Institutes of Health (NIH) (R01-AR072650 and P30-AR075042).

Author Contributions

P.S.Y. designed and conducted studies, interpreted data, and wrote the manuscript. M.M.K. conducted studies. J.S.M. designed and conducted studies and interpreted data. T.T. conducted studies. E.S.L. designed and conducted studies, interpreted data, and edited the manuscript. M.B.D. designed the studies, interpreted the data, and participated in writing the manuscript.

Disclosures

The authors have nothing to declare.

Data Availability

Original data generated and analyzed during this study are included.