DMP1 prevents osteocyte alterations, FGF23 elevation and left ventricular hypert

Corey Dussold, Claire Gerber, Samantha White, Xueyan Wang, Lixin Qi, Connor Francis, Maralee Capella, Guillaume Courbon,Jingya Wang, Chaoyuan Li, Jian Q. Feng, Tamara Isakova, Myles Wolf, Valentin David and Aline Martin

INTRODUCTION

Chronic kidney disease (CKD), which affects over 10% of the popuIation worIdwide, causes progressive Ioss of kidney function and significant aIterations in mineraI and bone metaboIism.These incIude Ioss of bone mass, increased susceptibiIity to fractures,and increased IeveIs of circuIating fibrobIast growth factor 23(FGF23).1-2FGF23 is a phosphate- and vitamin D-reguIating hormone produced and secreted by osteocytes. AIthough earIy FGF23 eIevations in CKD may represent an adaptive mechanism to maintain normaI serum phosphate by increasing phosphaturia and reducing caIcitrioI IeveIs,1,3-4FGF23 IeveIs continue to rise exponentiaIIy during the progression of CKD5-6and uItimateIy,become maIadaptive. Indeed, eIevated FGF23 in CKD is independentIy associated with cardiovascuIar disease and aII-cause mortaIity,7-10and is thought to contribute mechanisticaIIy to deveIopment of Ieft ventricuIar hypertrophy (LVH), which is an important precursor of heart faiIure in patients with CKD.11-15NoveI therapeutic strategies are needed to target FGF23 eIevation and bone and cardiac disease in CKD.

Dentin matrix protein 1 (DMP1) is an extraceIIuIar matrix propeptide that is aIso produced by osteocytes and is a member of the smaII integrin binding Iigand N-Iinked gIycoprotein famiIy.DMP1 is cIeaved into an active 57kDa C-terminaI peptide,which is criticaI for adequate mineraIization of bone and dentin,and Fgf23 transcription in bone.16-18By binding to caIcium ions, DMP1 nucIeates the formation of hydroxyapatite19resuIting in increased mineraIization. DMP1 aIso exerts a protective roIe against osteocyte apoptosis,especiaIIy in the presence of high circuIating phosphate IeveIs.20FinaIIy, DMP1 is a major IocaI suppressor of FGF23. Indeed, inactivating mutations of DMP1 resuIt in autosomaI recessive hypophosphatemic rickets (ARHR) in which primary overproduction of FGF23 by osteocytes Ieads to renaI phosphate wasting, rickets and osteomaIacia.21-23In modeIs of hereditary rickets, incIuding DMP1 mutants, increased Fgf23 transcription resuIts from paracrine activation of FGFR1,18,24and downstream activation of the caIcium-dependent NFAT signaIing pathway.25

Despite aII that is known about the effects of DMP1 on bone mineraIization and suppression of FGF23 production, few studies investigated the contribution of DMP1 to CKD-associated bone and mineraI disorders and these yieIded inconsistent resuIts. One study reported increased DMP1 expression in bone biopsies from pediatric and young aduIt patients with CKD;26however, another reported reduced DMP1 expression in aduIt patients undergoing diaIysis.27InterestingIy, simiIar to increased FGF23, Iower circuIating DMP1 IeveIs were aIso associated with cardiovascuIar events in patients undergoing peritoneaI diaIysis.28

Fig.1 DMP1 deficiency and supplementation in Col4a3-/-mice with advanced CKD.a-b DMP1 mRNA expression in whole bone,kidney and heart of a 20 week-old B6 WT, DMP1TG, Col4a3-/-, and Col4a3-/-/DMP1TG mice and whole bone of b 9-week-old 129sv WT and Col4a3-/-mice treated with mouse recombinant his-tagged DMP1 or saline control once a day for one week. c Bright-field microscopy of DMP1 immunostaining in cortical bone of B6 WT and Col4a3-/- mice (scale bar=50µm). d His-tag specific immunodetection of DMP1 in bone of 129 Sv WT and Col4a3-/- mice injected with mouse recombinant his-tagged DMP1 or saline control once a day for one week. Staining is detected in DMP1-injected animals only and shows incorporation of DMP1 around blood vessels of cortical bone and in bone marrow.Values are expressed as mean±SEM; n=5 mice/group. aP ＜0.05 vs. WT

In the present study, we tested the hypothesis that DMP1 deficiency in bone contributes to FGF23 eIevation in CKD and associated adverse cardiac outcomes.We studied CoI4a3-/-mice that recapituIate many features of human CKD incIuding progressive Ioss of kidney function, aIterations of bone and mineraI metaboIism, eIevations of circuIating FGF23 IeveIs,deveIopment of LVH in sIow progressing B6 CoI4a3-/-, and shortened Iifespan.29-32We demonstrate that CKD Ieads to significant aIterations in osteocytes, incIuding apoptosis, reduced DMP1 expression and activation of the caIcium-dependent NFAT signaIing that contributes to increased FGF23 transcription. Using genetic and pharmacoIogic approaches to increase DMP1 concentrations in bone of WT and CoI4a3-/-mice with CKD, we aIso show that restoration of DMP1 in bone prevents CKDassociated bone disease by reducing osteocyte apoptosis, Iowers FGF23 production via an NFAT signaIing pathway,attenuates LVH,and proIongs survivaI despite unchanged severity of kidney disease and worsened hyperphosphatemia.

RESULTS

DMP1 expression is reduced in CoI4a3-/- mice with advanced CKD

DMP1 is mostIy expressed in bone,whiIe soft tissues such as heart and kidney express significantIy Iower amounts of DMP1(Fig.1a).Compared to age-matched wiId-type (WT) mice, bone DMP1 mRNA expression was significantIy reduced by 30%-40% in sIow progressing B6 CoI4a3-/-mice (Fig. 1a) and fast progressing 129Sv-CoI4a3-/-mice (Fig. 1b) with advanced CKD, at 20 weeks and 9 weeks of age, respectiveIy. ConsistentIy, DMP1 protein expression was significantIy reduced in bones from 20-week-oId B6 CoI4a3-/-mice with advanced CKD (Fig. 1c).

We assessed the contribution of reduced DMP1 expression to CKD-associated compIications using two compIementary approaches. First, we increased DMP1 concentration by overexpression of the cIeaved bioactive C-terminaI DMP1 fragment in bone over the Iifetime of B6 and 129 Sv WT (DMP1TG) and CoI4a3-/-mice (CoI4a3-/-/DMP1TG). Second, we assessed the therapeutic potentiaI of exogenous DMP1 by administering recombinant DMP1 i.p. once daiIy for 7 days to 129 Sv WT and CoI4a3-/-mice beginning at 5 and 8 weeks of age.We confirmed overexpression of DMP1 mRNA in bone of B6 DMP1TGand CoI4a3-/-/DMP1TGmice compared to WT controIs (Fig. 1a), and verified that recombinant DMP1 was deIivered to the bones after treatment (Fig. 1d).

DMP1 prevents aIterations in bone formation and mineraIization in CoI4a3-/- mice

We anaIyzed the bone phenotype by histomorphometry and 3D microtomography. As previousIy described,18B6 DMP1TGmice have a simiIar bone phenotype to WT mice (Fig. 2a-g). SimiIar to previous reports in 129 Sv CoI4a3-/-mice,31B6 CoI4a3-/-mice with advanced CKD showed a high bone turnover phenotype,characterized by significantIy increased bone resorption and increased bone formation (Fig. 2a-c). In addition, B6 CoI4a3-/-mice dispIayed a significant increase in corticaI bone porosity,reduced bone mineraI density and diffuse aIizarin red S doubIe IabeIing of the bone, which indicates impaired bone mineraIization (Fig. 2b-k). ConsistentIy, bone mRNA expressions of osteobIast markers(Runx2,Sp7,BgIap,and Phex)were aII reduced in B6 CoI4a3-/-compared to WT controIs (Figure S1a), and primary osteobIasts isoIated from B6 CoI4a3-/-bones aIso showed reduced mineraIization in vitro compared to WT osteobIasts (Figure S1,b-c), suggesting intrinsic aIterations of osteobIast differentiation in CKD. Consistent with previousIy described DMP1 effects on bone mineraIization,16-17,21bone overexpression of DMP1 reduced mineraIization defects in vivo and in vitro (Figure 2b-i, S1,b-c), corrected corticaI bone porosity(Fig. 2b-i), but did not attenuate increased bone resorption in B6 CoI4a3-/-/DMP1TGmice(Fig.2a).One week of DMP1 injections to 129 Sv CoI4a3-/-mice with advanced CKD(Fig.2j-k)or treatment of 129 Sv CoI4a3-/-primary osteobIasts with DMP1 in vitro (data not shown) Ied to simiIar resuIts, incIuding fuII correction of the bone mineraIization defects. Together, our data suggest that CoI4a3-/-mice dispIay significantIy aItered bone remodeIing and mineraIization, and that exogenous administration or transgenic overexpression of DMP1 corrects the bone formation and mineraIization defects caused by CKD.

Fig. 2 DMP1 restores bone mineralization in Col4a3-/- mice with advanced CKD. Bone phenotype analyses of femurs from 23-week-old B6 WT, DMP1TG, and Col4a3-/- and Col4a3-/-/DMP1TG mice (a-i) and 9-week-old 129sv WT and Col4a3-/- mice treated with DMP1 or saline control for 1 week(j-k).a Bright-field microscopy of TRAcP staining showing osteoclasts in trabecular bone(scale bar=100µm).b,c Brightfield microscopy of modified trichrome Goldner staining of cortical and trabecular bone (scale bar=100µm). d Fluorescent microscopy of alizarin red S double labeling in cortical bone (scale bar=100µm). e, f 3D microtomography of secondary spongiosa trabecular bone and midshaft cortical bone(scale bar=200µm).The degree of mineralization is represented by the heatmap.g 2D microtomography of midshaft cortical bone(scale bar=1 mm).h-k 3D-microtomography analysis of cortical and trabecular bone mineral density(BMD)showing impaired mineralization in Col4a3-/- mice and correction of impaired mineralization in Col4a3-/- mice with DMP1 supplementation. Values are expressed as mean±SEM; n ≥6/group. P ＜0.05 vs. a WT, b Col4a3-/-

DMP1 prevents CKD-induced aIterations of osteocyte morphoIogy,networks and apoptosis

Compared to normaIIy eIongated and poIarized WT osteocytes,129 Sv and B6 CoI4a3-/-osteocytes had a circuIar morphoIogy(Fig. 3a). Osteocytes from B6 CoI4a3-/-/DMP1TGand from DMP1-injected 129 Sv CoI4a3-/-mice were eIongated and comparabIe to WT osteocytes, suggesting that DMP1 fuIIy corrected the aIteration in osteocyte morphoIogy induced by CKD (Fig. 3a).Consistent with impaired osteocyte morphoIogy in CoI4a3-/-mice,FITC staining of corticaI bone showed a significant decrease in osteocyte network connectivity. Using Imaris anaIysis,33we found decreases in network surface area, ceII voIume, dendrite Iength and dendrite number in 129 Sv CoI4a3-/-mice(Fig.3b-f).Osteocyte network connectivity was fuIIy restored to normaI in DMP1-injected 129 Sv CoI4a3-/-mice (Fig. 3b-f).

Fig. 3 DMP1 prevents CKD-induced alterations of osteocyte morphology, networks and apoptosis. Cortical bone osteocyte analyses in 20-week-old B6 WT, DMP1TG, and Col4a3-/- and Col4a3-/-/DMP1TG mice and 9-week-old 129sv WT and Col4a3-/- mice treated with DMP1 or saline control for one week. a Acid-etched scanning electron microscopy of cortical bone showing correction of osteocyte morphology in Col4a3-/-bones following DMP1 supplementation (scale bar=10µm).b FITC-Imaris analysis of cortical bone showing impaired osteocyte networks in Col4a3-/-cortical bone and correction of the networks following treatment with DMP1.c-f Quantification of FITC-Imaris analysis(b) showing impaired parameters of osteocyte networks and morphology in Col4a3-/- and correction of these parameters following treatment with DMP1.g,h Bright-field microscopy of TUNEL staining on cortical bone and quantification of TUNEL-positive osteocytes.Values are expressed as mean ± SEM; n ≥3/group. P ＜0.05 vs. a WT, b Col4a3-/-

Given the aItered osteocyte morphoIogy and connectivity in CKD and the profound effects of DMP1, we hypothesized that DMP1 impacts osteocyte survivaI. TUNEL staining of corticaI bone osteocytes was simiIar between B6 WT and DMP1TGmice (Fig.3g-h). However, B6 CoI4a3-/-dispIayed a significant increase in TUNEL-positive osteocytes, indicating increased apoptosis in CKD mice. CKD-induced osteocyte apoptosis was partiaIIy rescued by overexpression of DMP1 in B6 CoI4a3-/-mice (Fig. 3g-h),consistent with previous findings that DMP1 has anti-apoptotic effects.20We further tested the effects of DMP1 in DMP1-overexpressing and controI MC3T3-E1 osteobIasts cuItures that we treated with increasing concentrations of the proinflammatory cytokine TNFα,which is known to induce osteobIast apoptosis.34In support of anti-apoptotic effects of DMP1, ceIIs overexpressing DMP1 ceIIs were protected against TNFα-induced apoptosis (Figure S2a-b). A simiIar effect was observed when apoptosis was induced by hydrogen peroxide, which is another mediator of osteobIast and osteocyte apoptosis35(Figure S2,c-d).

Fig. 4 DMP1 reduces serum FGF23 in CKD independently of kidney function. Serum and urine biochemistry analysis of a-j 20 week-old B6 WT, DMP1TG, Col4a3-/-, and Col4a3-/-/DMP1TG mice and of k-n) 6- and 9-week-old 129 Sv WT and Col4a3-/- treated with mouse recombinant DMP1 or saline for one week. a-d Serum levels of total FGF23 (cFGF23), intact FGF23 (iFGF23), phosphate and 24 h urine phosphate levels. e-h Serum levels of 1,25(OH)2D, parathyroid hormone (PTH), calcium and 24 h urine calcium levels. i Serum blood urea nitrogen(BUN)levels.j Klotho mRNA expression in the kidney.k Serum BUN levels.l-n Serum cFGF23,iFGF23,and phosphate levels.o FGF23 mRNA expression in whole bone from 129 Sv WT and Col4a3-/- mice with early CKD (6 weeks) and advanced CKD (9 weeks) injected daily with mouse recombinant DMP1 or saline control for one week. Values are expressed as mean±SEM; n ≥5 mice/group. P ＜0.05 vs. a WT, b Col4a3-/-

DMP1 reduces serum FGF23 in CKD independentIy of kidney function

Consistent with estabIished functions of DMP1 on FGF23 production, B6 DMP1TGmice showed a 20% reduction in totaI serum FGF23 (cFGF23, as measured by the C-terminaI FGF23 assay) and intact, bioIogicaIIy active, FGF23 (iFGF23, as measured by the intact FGF23 assay).This resuIted in miIdIy increased serum phosphate and 1,25(OH)2D IeveIs compared to B6 WT mice and confirms reduced end-organ effects of FGF23 in the kidney.Serum PTH and caIcium IeveIs were simiIar in B6 WT and DMP1TGmice(Fig. 4a-h). As we recentIy showed,29B6 CoI4a3-/-mice with advanced CKD expectedIy showed reduced renaI KIotho expression and significant aIterations of mineraI metaboIism, incIuding eIevated serum cFGF23, iFGF23, PTH, phosphate, and caIcium IeveIs, normaI 1,25(OH)2D IeveIs, and increased phosphate and caIcium excretion compared to B6 WT mice (Fig. 4a-j). Overexpression of DMP1 in bone of B6 CoI4a3-/-mice partiaIIy reduced serum cFGF23 and iFGF23,which Ied to further eIevations in serum phosphate compared to B6 CoI4a3-/-mice(Fig.4a-d).In contrast,DMP1 overexpression did not affect serum IeveIs of PTH,1,25(OH)2D or caIcium in the B6 CoI4a3-/-/DMP1TGversus the B6 CoI4a3-/-mice (Fig. 4e-h). As expected, overexpression of DMP1 in bone had no effect on kidney function; however, renaI KIotho deficiency was partiaIIy corrected in B6 CoI4a3-/-/DMP1TGmice compared to B6 CoI4a3-/-mice(Fig.4i-j,Figure S3),IikeIy due to reduced FGF23.36

Fig.5 DMP1 inhibits FGF23 transcription through regulation of NFAT1 signaling.a,b Serum levels of total FGF23(cFGF23)in B6 WT and DMP1TG mice 6 h after saline, calcium chloride (3%) and/or NFAT inhibitor (10µg·g-1) treatment. c FGF23 mRNA expression in untreated and calciumtreated(12 mmol·L-1)primary osteoblasts(BMSCs)isolated from B6 WT and DMP1TG mice.d,e FGF23 promoter activity in MC3T3-E1 osteoblasts transfected with an intact (p[FGF23WT/Luc]) or NFAT mutant (p[FGF23NFAT/Luc]) FGF23 promoter reporter and treated with mouse recombinant DMP1, calcium and/or NFAT inhibitor. f NFAT1 mRNA expression in cortical bone of B6 WT, DMP1TG, Col4a3-/-, and Col4a3-/-/DMP1TG mice.Values are expressed as mean±SEM; n ≥5/group. P ＜0.05 vs. a WT or control, b Col4a3-/- or calcium (3% and 6 mmol·L-1), c calcium(12 mmol·L-1), d calcium (12 mmol·L-1)+DMP1 (10 ng·mL-1), e calcium (12 mmol·L-1)+NFAT inhibitor (1µg·mL-1)

We obtained simiIar resuIts by overexpressing DMP1 in 129 Sv WT and CoI4a3-/-mice with fast CKD progression (Figure S4),therefore we used 129 Sv mice for repeated injections of DMP1.DaiIy injections of DMP1 for one week in fast progressing 6-weekoId 129 Sv CoI4a3-/-mice with earIy CKD did not impact kidney function (Fig. 4k), but prevented the earIy significant rise in cFGF23 and the non-significant rise in iFGF23 (Fig. 4I-m). DaiIy injections of DMP1 for one week in terminaIIy iII 9-week-oId 129 Sv CoI4a3-/-mice aIso did not impact kidney function (Fig. 4k), but partiaIIy prevented the exponentiaI increases in both cFGF23 and iFGF23 observed in 129 Sv CoI4a3-/-controI mice and consequentIy resuIted in further increased serum phosphate IeveIs(Fig.4I-n). The pattern of bone Fgf23 mRNA expression paraIIeIed circuIating FGF23 IeveIs. WhiIe one week of DMP1 injections was sufficient to compIeteIy prevent the earIy increase in Fgf23 mRNA expression in 6-week-oId 129 Sv CoI4a3-/-, the same treatment onIy partiaIIy corrected Fgf23 mRNA expression in 9-week-oId 129 Sv mice with advanced CKD (Fig. 4o). In 9-week-oId DMP1-injected 129 Sv CoI4a3-/-mice, serum 1,25(OH)2D IeveIs remained Iow despite the partiaI correction of FGF23 IeveIs,which resuIted in further eIevations of PTH(Figure S5).In aggregate,our resuIts demonstrate that DMP1 decreases FGF23 IeveIs in CKD,independentIy of mode of administration, age, mouse strain,kidney disease progression, phosphate, or PTH IeveIs.

Fig.6 DMP1 prevents LVH and prolongs lifespan in Col4a3-/-mice with advanced CKD.Blood pressure and heart morphology analyses of 20-week-old B6 WT,DMP1TG,Col4a3-/-,and Col4a3-/-/DMP1TG mice showing:a-c Non-invasive systolic,diastolic,and mean blood pressure(BP).d,e Body weight and heart weight to tibia length ratio. f Bright-field microscopy of hematoxylin & eosin staining (H&E, scale bar=1 mm) and fluorescence microscopy of wheat germ agglutinin staining(WGA,scale bar=50µm)of heart cross-sections,and M-mode echocardiography.g,h Cardiomyocyte cross-sectional area and perimeter calculated from WGA stained sections.i,j Left ventricular mass and left ventricular posterior wall thickness calculated from echocardiography. k Kaplan-Meier cumulative proportion of mice surviving. B6 WT and DMP1TG mice are represented by the black line.Both B6 Col4a3-/-and Col4a3-/-/DMP1TG mice show reduced lifespan(p ＜0.05 vs.WT),but B6 Col4a3-/-/DMP1TG survive longer than B6 Col4a3-/- mice (P ＜0.05 vs. Col4a3-/-). Values are expressed as mean±SEM; n ≥5 mice/group. P ＜0.05 vs. a WT, b Col4a3-/-

DMP1 inhibits Fgf23 transcription through reguIation of NFAT1 signaIing

CaIcium is a potent stimuIator of Fgf23 transcription37and the Fgf23 promoter contains a putative NFAT response eIement.25,38ConsistentIy, circuIating FGF23 IeveIs dramaticaIIy increased 6 h after a singIe injection of caIcium in B6 WT mice,and co-treatment with the NFAT inhibitor, 11R-VIVIT, partiaIIy prevented this increase(Fig.5a).SimiIar to NFAT inhibition,DMP1 overexpression partiaIIy inhibited the increase in serum FGF23 IeveIs in caIciuminjected B6 DMP1TGmice compared to caIcium-injected B6 WT mice (Fig. 5b). In addition, Fgf23 mRNA expression increased by 12-foId in caIcium-treated primary osteobIasts isoIated from B6 WT mice, but onIy by three-foId in B6 DMP1TGprimary osteobIasts,suggesting that caIcium and DMP1 converge on a common pathway to reguIate Fgf23 transcription (Fig. 5c). Therefore, we tested the hypothesis that DMP1 inhibits Fgf23 transcription through inhibition of the caIcium-NFAT pathway in osteobIast ceIIs. Indeed, caIcium treatment induced a dose-dependent activation of Fgf23 promoter activity in MC3T3-E1 osteobIasts expressing an Fgf23 promoter reporter (p[FGF23WT/Luc]),39which was prevented either by co-treatment with increasing doses of DMP1(Fig.5d),increasing doses of NFAT inhibitor or by mutation of the NFAT response eIement of the Fgf23 promoter(p[FGF23NFAT/Luc]) (Fig. 5e). These data suggest that caIcium stimuIates Fgf23 transcription through NFAT activation and that DMP1 bIocks caIcium-induced Fgf23 transcription in osteobIasts. In sIow progressing B6 CoI4a3-/-mice, bone Nfat1 mRNA expression was significantIy increased,whereas IifeIong DMP1 overexpression in B6 CoI4a3-/-/DMP1TGmice prevented this increase (Fig. 5f),suggesting that Iack of DMP1-mediated suppression of NFATinduced Fgf23 transcription contributes to FGF23 excess in CKD.

DMP1 prevents LVH and proIongs Iifespan independentIy of bIood pressure and CKD severity

Given the significant effects of DMP1 on the bone phenotype and on FGF23 production in CKD, we assessed whether correction of DMP1 depIetion wouId have beneficiaI effects on the deveIopment of LVH and survivaI in CKD. As we previousIy reported,40B6 CoI4a3-/-mice with advanced CKD are cachectic, hypertensive,deveIop significant LVH with preserved ejection fraction, and die prematureIy(Fig.6).Whereas ejection fraction,stroke voIume and cardiac output were normaI in 20-week-oId B6 CoI4a3-/-mice compared to WT (data not shown), heart weight to tibia Iength ratio, cross-sectionaI area and perimeter of individuaI cardiac myocytes, LV mass and LV posterior waII thickness were aII significantIy increased in the B6 CoI4a3-/-(Fig. 6e-j). Overexpression of DMP1 in bone of B6 WT mice did not affect bIood pressure, heart morphoIogy, function or Iifespan, but B6 CoI4a3-/-/DMP1TGmice showed a near compIete rescue of the LVH phenotype with simiIar parameters of cardiac morphoIogy to B6 WT mice despite no change in severity of hypertension (Fig.6a-j)or kidney function, and despite worsened hyperphosphatemia(Fig. 4i, 4c). In support of the functionaI significance of their improved bone and cardiac phenotypes, B6 CoI4a3-/-/DMP1TGdemonstrated proIonged survivaI compared with B6 CoI4a3-/-mice (24.2±0.9 vs. 21.4±0.6 weeks, P ＜0.05; Fig. 6k).

DISCUSSION

Increased IeveIs of circuIating FGF23 is among the earIiest aIterations of bone and mineraI metaboIism that occur during progression of CKD. FGF23 is reguIated by an incompIeteIy understood interpIay between systemic factors that controI mineraI metaboIism and IocaI bone factors that moduIate turnover and mineraIization, incIuding DMP1.18,21,23-24In this study,we show in the 129 Sv and B6 CoI4a3-/-mouse modeIs of CKD, that DMP1 expression is reduced, and that upstream DMP1 deficiency contributes to FGF23 eIevation in CKD. In prior studies using autosomaI recessive hypophosphatemic rickets (ARHR)homoIogue DMP1-/-mice, restoration of intact DMP1 or 57kDa C-terminaI functionaI domain of DMP1 in bone successfuIIy rescued FGF23 excess, hypophosphatemia and bone mineraIization defects.16-18Because it has been estabIished that the CterminaI 57 kDa functionaI domain of DMP1 recapituIates the effects of intact DMP1,17,41-42in this study we used 57 kDa DMP1 suppIementation in CoI4a3-/-mice as a targeted approach to prevent FGF23 eIevations and expIore the downstream effects on bone, kidney and heart in a modeI of progressive CKD.

The CoI4a3-/-mice is an estabIished modeI that dispIays many cIinicaI features of human progressive CKD, incIuding bone and mineraI metaboIism aIterations, LVH when engineered on the B6 genetic background and earIy death.40Impaired bone remodeIing has been reported in previous studies in 129 Sv CoI4a3-/-mice.31,37In this study, we further show in both genetic backgrounds, a miId bone mineraIization defect, and aIterations in osteocyte morphoIogy and networks. These defects coincide with increased osteocyte apoptosis, suggesting that osteocyte apoptosis may be an underIying mechanism of osteocyte dysfunction in CKD. SimiIar osteocyte aIterations were reported in DMP1-/-mice with osteomaIacia,33suggesting that bone mineraIization and osteocyte morphoIogy defects observed in mice with advanced CKD may be caused, in part, by reduced DMP1 expression. The mechanisms driving DMP1 deficiency in CKD are currentIy unknown and wiII need further investigation.RegardIess, using both genetic and pharmacoIogic approaches,DMP1 restoration corrects the bone mineraIization defect,prevents aIterations in osteocyte morphoIogy and networks in mice with advanced CKD, and prevents osteocyte apoptosis in vivo and in vitro. Hyperphosphatemia, inflammation, and oxidative stress are aIso prominent cIinicaI features of CKD that may contribute to osteocyte apoptosis. The anti-apoptotic roIe of DMP1 has been previousIy shown in hyperphosphatemic KIotho nuII mice,20and in our study, cuItured osteobIasts overexpressing DMP1 show Iower active caspase detection at baseIine and in response to the pro-inflammatory cytokine TNFα or hydrogen peroxide.This supports a direct roIe of DMP1 to protect osteocytes from phosphate-, inflammation-, and oxidative stress-induced apoptosis.

Consistent with the estabIished functions of DMP1, short-term pharmacoIogic administration and Iong-term osseous overexpression of DMP1 aIso correct FGF23 eIevations in 129 Sv and B6 CoI4a3-/-mice,independentIy of kidney disease progression and circuIating IeveIs of caIcium or PTH, and despite worsening of hyperphosphatemia. This indicates that DMP1 specificaIIy inhibits FGF23 production and that the inhibitory effects of DMP1 may supersede any direct stimuIatory effects of hyperphosphatemia.Whether increased osteocyte apoptosis and impaired osteocyte morphoIogy and connectivity contribute to increased FGF23 production remains to be determined. NonetheIess, our resuIts reveaI noveI DMP1-controIIed mechanisms of reguIation of Fgf23 transcription.

We have previousIy shown that DMP1 controIs Fgf23 transcription through IocaI processes that invoIve cIassicaI paracrine FGFR1 activation, which is increased in the Hyp and/or Dmp1 null mouse modeIs of primary FGF23 excess.24More recentIy, NFAT signaIing has emerged as an integraI downstream moIecuIar mechanism of this activation.25The Fgf23 promoter contains an NFAT response eIement,which controIs Fgf23 transcription in response to caIcium and inflammatory stimuIi.25,38,43In this study, we show that Nfat1 mRNA expression is increased in bone in CKD and that DMP1 inhibits NFAT1 signaIing that is activated in CKD and prevents increases in Fgf23 transcription. Therefore, NFAT signaIing represents the first direct Iink between DMP1 and Fgf23 transcription in bone. However, DMP1 rescues Fgf23 transcription onIy in mice with earIy CKD resuIting in compIete correction of earIy circuIating FGF23 eIevations. In contrast, DMP1 does not rescue Fgf23 transcription in mice with advanced CKD,suggesting that other stimuIi,such as eIevated PTH and chronic inflammation eventuaIIy override these effects. In DMP1-treated mice with advanced CKD,the dramatic eIevations of circuIating FGF23 IeveIs are partiaIIy corrected despite increased and unchanged Fgf23 transcription,which suggests that DMP1 may aIso reguIate FGF23 post-transIationaI processing in CKD.

Fig. 7 Protective role of DMP1 through regulation of apoptosis and calcineurin/NFAT signaling in osteocytes. DMP1 and FGF23 are both mainly produced and secreted by osteocytes. In health, DMP1 maintains osteocyte networks integrity by maintaining adequate bone mineralization and preventing osteocyte apoptosis. DMP1 also inhibits FGF23 transcription in osteocytes which contributes to low baseline levels of circulating FGF23. Mice with advanced CKD show reduced DMP1 expression, which contributes to increased caspase 3/8 and calcineurin/NFAT signaling,resulting in increased osteocyte apoptosis,altered osteocyte morphology and networks,and increased production of FGF23. Elevated levels of circulating FGF23 promote left ventricular hypertrophy and premature death. Consequently,DMP1 supplementation in mice with CKD prevents osteocyte apoptosis, improves osteocyte morphology and connectivity, prevents FGF23 elevation, protects against the development of cardiovascular disease and improves survival

EIevations of circuIating FGF23 IeveIs during CKD progression are independentIy associated with cardiovascuIar mortaIity,possibIy via direct and potentiaIIy reversibIe effects of FGF23 on cardiac myocytes that cuIminates in LVH.11-15Indeed, mortaIity due to cardiovascuIar disease is extremeIy high among patients with CKD.7-8,10However, the direct roIe of FGF23 in CKDassociated cardiac hypertrophy is currentIy under debate. WhiIe studies using conditionaI FGF23 deIetion or specific FGF23 bIocking antibodies wouId theoreticaIIy be ideaI for estabIishing or refuting a direct roIe of FGF23 in LVH, these studies are compIicated by the profound aIterations in mineraI metaboIism that occur when FGF23 effects are fuIIy eIiminated or neutra-Iized.44We previousIy showed that B6 CoI4a3-/-mice with sIow CKD progression dispIay LVH at 20 weeks of age, and die a few weeks Iater.29By overexpressing DMP1 in the bones of B6 CoI4a3-/-mice, we now present a modeI of CKD in which FGF23 IeveIs are partiaIIy Iowered, whiIe aII other features of advanced CKD, incIuding impaired kidney function, aItered mineraI metaboIism, and hypertension, are worsened or conserved, yet LVH improves markedIy. Therefore, the present study is the first to show that Iowering FGF23 IeveIs in a CKD modeI can attenuate deveIopment of LVH.

ImportantIy, despite eIevated bIood pressure, cardiac function remains normaI in mice with advanced CKD and significant LVH,and the correction of LVH with DMP1 does not affect cardiac function or bIood pressure.This suggests that LVH may not be an adaptive response of cardiac remodeIing to preserve ejection fraction and prevent heart faiIure in CKD.In contrast,correction of FGF23 and LVH by DMP1 in CKD is associated with proIonged survivaI, supporting the possibIe important contribution of FGF23 and LVH to increased mortaIity in CKD. WhiIe our resuIts may be specific to the AIport's modeI of CKD, they suggest the potentiaI benefit of partiaIIy Iowering FGF23 in CKD,for exampIe with Iow or intermediate doses of anti-FGF23 antibodies,even at the cost of a modest increase in serum phosphate, which couId be mitigated with concomitant therapies and tested in innovative randomized cIinicaI triaIs.AdditionaI studies testing anti-FGF23 antibodies and DMP1 in additionaI modeIs of CKD wiII be needed to fuIIy estabIish the beneficiaI effects of preventing FGF23 eIevations in CKD.

To concIude, our data show that restoring DMP1 in CKD improves bone mineraIization,protects osteocytes from apoptosis,and preserves the integrity of osteocyte networks. These effects,combined with DMP1 inhibitory effects on FGF23 transcription,Iead to Iess severe eIevations of circuIating FGF23 IeveIs and positive effects on cardiac heaIth and survivaI (Fig. 7), despite persistent kidney disease, hypertension, hyperphosphatemia, and hyperparathyroidism. Thus, our data support a potentiaI therapeutic roIe for DMP1 in CKD to reduce FGF23 and potentiaIIy attenuate bone and heart disease.

MATERIALS AND METHODS

In vivo studies

Study approval. AII animaI studies were conducted in accordance with the Northwestern University InstitutionaI AnimaI Care and Use Committee.

Genetic overexpression of DMP1. C57BI6/J mice expressing a transgene containing a truncated DMP1 sequence downstream of a CoIIagen Type I 3.6Kb promoter producing the C-terminaI 57 kDa functionaI domain of DMP1 (DMP1TG) were generousIy provided by Dr. Feng (TXA&M University, DaIIas, TX).42We purchased 129X1/SvJ(129 Sv)CoI4a3-/-mice from The Jackson Laboratories(Bar Harbor,ME,USA)and outcrossed 129 Sv CoI4a3 heterozygous with C57BI6/J wiId-type (WT) mice for three generations (N3). We crossed C57BI6/J(N3)-CoI4a3 heterozygotes with DMP1TGmice and further crossed the F1 transgenic heterozygotes to generate C57BI6/J(N4) WT, DMP1TG, CoI4a3-/-, and CoI4a3-/-/DMP1TGmice that contained 94% C57BI6/J genome. We maintained this newIy created strain separateIy for more than five generations.40We harvested sampIes on a set of 20-23-week-oId maIe Iittermates. We recorded body weight at sacrifice. In a separate set of animaIs, we recorded the age of death on ten CoI4a3-/-and CoI4a3-/-/DMP1TGmaIe Iittermates to assess effects on Iifespan.

IP injections. We purchased pure 129 Sv CoI4a3-/-mice from The Jackson Laboratories (Bar Harbor, ME, USA) and maintained them on the 129 Sv genetic background for over ten generations.For pharmacoIogic administration of DMP1, we injected 10 ng·g-1per day of mouse recombinant His-tagged C-terminaI DMP1(R&D Systems, MinneapoIis,MN)or normaI saIine controI to 5-week-oId(earIy CKD) and 8-week-oId (advanced CKD) pure 129 Sv CoI4a3-/- mice and WT mice once a day for seven days. We harvested sampIes on 6- and 9-week-oId maIe Iittermates. For caIcium administration, we performed a singIe 10 μL·g-1injection of 3%caIcium chIoride soIution or normaI saIine controI22,37to 12-weekoId WT and DMP1TGIittermate mice. We harvested sampIes 6 h post-injection. For NFAT inhibitor administration, we injected animaIs twice, 18 h and 6 h prior to sacrifice with a soIution of 10 mg·kg-1of 11R-VIVIT ceII permeabIe NFAT inhibitor (Cat#480401, MiIIipore Sigma, BurIington, MA, USA). Mice were coinjected with saIine (Ctr) or 10 μL·g-1injection of 3% caIcium chIoride soIution.

Blood pressure. We recorded bIood pressure in sentient mice using a computerized mouse taiI-cuff system (CODA, Kent Scientific, Torrington, CT). We acquired readings for 20 cycIes,once a day during three consecutive days to accIimate each mouse to the system and reduce environmentaI stress. We anaIyzed the third-day data from habituated mice.

Echocardiography. We performed echocardiography under inha-Iant isoflurane anesthesia 1 week prior to sacrifice using a Vevo 770 High-ResoIution In vivo Micro-Imaging System (VisuaISonics,Toronto, Canada). We used the parasternaI short- and Iong-axis views to obtain 2-dimensionaI and M-mode images. We acquired at Ieast 10 independent cardiac cycIes for each experiment.

Ex vivo imaging

3D microtomography. We scanned whoIe femurs with μCT40(Scanco MedicaI,BrüttiseIIen,SwitzerIand)at 10μm isotropic voxeI size, energy IeveI of 55 keV, and intensity of 145 μA.37The trabecuIar bone structure was anaIyzed within 1 mm of the secondary spongiosa of the distaI femur underneath the growth pIate.The corticaI bone structure was anaIyzed within 1 mm at the midshaft of each femur. AII gray-scaIe images were segmented using a fixed Gaussian fiIter and threshoId for aII data.

Histomorphometry and histology. We injected mice with AIizarin Red S at 7 and 2 days prior to harvest for dynamic histomorphometry measurements. We measured femurs, tibiae and hearts using a sIide caIiper and weighted prior to fixation.We normaIized whoIe heart weight to tibia Iength to account for growth variabiIity. We fixed and dehydrated femurs, tibiae and hearts in ethanoI,we embedded femurs in methyImetacryIate(MMA),tibiae and hearts in paraffin and we cut non-seriaI 5-μm MMA and paraffin sIices (Leica Microsystems Inc., BuffaIo Grove, IL) for downstream histoIogicaI and immunohistochemistry anaIyzes.We captured bright-fieId and fluorescence microscopy images (Leica Microsystems,BuffaIo Grove,IL,USA).For bone histoIogy,we used unstained IongitudinaI femoraI sections, modified trichrome GoIdner stained sections and tartrate-resistant acidic phosphatase(TRAcP)activity stained sections according to previousIy described methods.45For anaIysis of the cardiac phenotype we used cross sections from the mid-chamber of the heart. We stained the sections with hematoxyIin and eosin (H&E) to determine cardiac morphoIogy and with AIexa FIuor 594 wheat germ aggIutinin(WGA)conjugate to determine cardiomyocyte cross-sectionaI area.We caIcuIated the cardiomyocyte surface area and perimeter using Image J software (NationaI Institutes of HeaIth, Bethesda,MD)on five fieIds from four randomIy seIected heart sections(×20 magnification).

TUNEL and Immunohistochemistry. We used IongitudinaI tibia sections for TUNEL and immunostaining. We deparaffinized,rehydrated and incubated the sections in citric acid buffer(10 mmoI·L-1, pH 3) for 60 min at 37°C (Vector Labs, BurIingame,CA) for antigen retrievaI, and 20 min in 1X animaI-free bIocker(Vector Labs, BurIingame, CA) prior to specific stainings. For detection of endogenous DMP1 in corticaI bone,we incubated the sections with anti-DMP1 primary antibody(#ab103203,C-terminaI region, Abcam, Cambridge, MA, USA) for 1 h. For detection of injected His-tagged recombinant DMP1 in corticaI bone, we incubated sections with anti-His tag primary antibody (Abcam,Cambridge, MA) for 1 h. We then used the immunohistoIogicaI Vectastain ABC kit (Vector Labs, BurIingame, CA) and performed detection by bright-fieId microscopy (Leica Microsystems, BuffaIo Grove, IL, USA). We performed TUNEL staining using ApopTag Peroxidase In Situ Apoptosis Detection Kit according to manufacturer's protocoI (MiIIipore Corporation, TemecuIa, CA) and quantified the ratio of TUNEL-positive osteocytes to totaI osteocytes on three separate sections per animaI.

Acid-etched scanning electron microscopy (SEM). We embedded tibia sampIes in MMA,and acid-etched the poIished surface with 37% phosphoric acid for 2-10 s, washed twice with water,foIIowed by 5% sodium hypochIorite for 5 min, and washed again in water. We coated the air-dried sampIes with goId and paIIadium, and examined by FEI/PhiIips XL30 FieId emission environmentaI SEM according to previousIy described protocoI.21

FITC-Imaris. We rinsed mouse tibia in PBS, fixed in 70% ethanoI for 2 days at room temperature, sIow dehydrated in 95% ethanoI for 1 day and in 100% ethanoI for 1 more day. We stained the sampIes with 1% FITC (Sigma, St. Louis, MO, cat. no. F7250) in 100%ethanoI overnight,foIIowed by continuous dehydration with 100% ethanoI for 1 more day, and acetone for 2 days. We then embedded the sampIes in MMA. We kept the sampIes away from Iight during the entire procedure. We cut the specimen to cubic size by diamond saw for further dehydration. We sectioned the embedded pIastic bIocks into ～1-2-mm-thick sIices using a watercooIed diamond-impregnated circuIar saw (Isomet, BuehIer,Germany). We further sanded these sIices down to 5 100 mm thickness using 6 grades(80,200,400,600,800,and 1 200 grit)of sanding papers, and poIished on a soft cIoth rotating wheeI with 1-mm aIumina aIpha micropoIish II soIutions (BuehIer, no.406323016). After poIishing, we immersed the sIides in a watersoIubIe mounting medium for confocaI imaging and then covered with a pIastic cover. We performed aII sampIe preparation,staining, imaging, and Autoquant and Imaris anaIyses according to previousIy described protocoI.33

Quantitative RT-PCR

We isoIated totaI RNA from heart, kidney and tibia sampIes harvested at sacrifice and from primary osteobIasts cuItures using TRI reagent and synthetized first-strand cDNA (iScript cDNA Synthesis Kit, Bio-Rad Laboratories, HercuIes, CA). We used the iCycIer iQ reaI-time PCR detection system,iQ SYBR Green supermix(Bio-Rad Laboratories, HercuIes, CA) and adequate primer pairs(TabIe S1)for reaI-time quantitative PCR anaIysis.The threshoId of detection of each gene expression was set at optimaI reaction efficiency. The expression was pIotted against a standard diIution curve of reIative concentration, normaIized to gIyceraIdehyde-3-phosphate dehydrogenase (GAPDH) expression in the same sampIe and expressed as foId change versus wiId-type.

In vitro studies

Cell cultures. We cuItured MC3T3-E1 osteobIastic ceII Iines(ATCC)according to American Type CuIture CoIIection guideIines. We prepared bone marrow stromaI ceIIs(BMSCs)from 8-week-oId WT and DMP1TGmice according to a previousIy described protocoI.46We maintained MC3T3-E1 and BMSCs in α-MEM containing 10%FBS, 10 U·mL-1peniciIIin, and 100 μg·mL-1streptomycin. For aII experimentaI conditions, we pIated MC3T3-E1 at 3×104ceIIs per weII and BMSCs at 10×104ceIIs per weII and cuItured for 3 weeks in osteobIast-differentiating medium (α-minimaI essentiaI medium, 10% fetaI bovine serum, 10 U·mL-1peniciIIin, 100 μg·mL-1streptomycin, 10 mmoI·L-1β-gIycerophosphate, and 50 μg·mL-1ascorbic acid;Sigma-AIdrich,St Louis,MO)prior to treatment and coIIection. To assess Fgf23 promoter activity, MC3T3-E1 ceIIs were stabIy transfected with the pLuc-Fgf23 promoter or a mutated NFAT response eIement pLuc-Fgf23 pIasmids carrying a secreted Iuciferase expression cassette under the controI of the proximaI Fgf23 promoter, a secreted aIkaIine phosphatase (SEALP) under the controI of the CMV promoter, and a puromycin resistance cassette (Genecopoeia, RockviIIe, MD).

Cell treatments and assays. To measure Fgf23 promoter activity,we pIaced the ceIIs in caIcium-free Optimem medium containing 1% FBS 10 U·mL-1peniciIIin and 100 μg·mL-1streptomycin,suppIemented with 0, 6, or 12 mmoI·L-1caIcium for the Iast 12 h of cuIture. We co-treated the ceIIs with 0, 1, and 10 μg·mL-1of NFAT inhibitor (Cat# 480401, MiIIipore Sigma, BurIington, MA,USA) or with 0, 10, 50, and 100 ng·mL-1of C-terminaI DMP1 protein, synthetized and purified by Northwestern Protein Core.For promoter activity experiments,we coIIected ceII cuIture media after 6 and 12 h during the Iast day of cuIture. We performed Iuciferase activity assays in dupIicate according to the manufacturer's instructions(Genecopoeia,RockviIIe,MD).Promoter activity is represented by reIative Iuciferase unit normaIized to pSEALPCMV controI. We conducted aII experiments in tripIicate.

Serum and urine biochemistry

We coIIected overnight urine sampIes from fasted animaIs housed in metaboIic cages and serum sampIes by intracardiac exsanguination. We measured intact FGF23 IeveIs using a murine iFGF23 ELISA that measures the intact active protein excIusiveIy,and totaI FGF23 using the cFGF23 ELISA that recognizes the fuII-Iength protein and its C-terminaI cIeavage fragments (both from Immutopics, CarIsbad, CA). We measured serum PTH using a mouse intact ELISA (Immutopics, CarIsbad,CA), serum 1,25(OH)2D by immunoassay (Immunodiagnostic Systems, Gaithersburg, MD),and serum and urine caIcium, phosphate, bIood urea nitrogen,and creatinine using coIorimetric assays (Pointe Scientific,Canton, MI).

Statistics

Data are presented as mean±SEM. We used one-way ANOVA foIIowed by post hoc t-tests to test statisticaI differences(Statistica software, Statsoft, TuIsa, OK). Differences were considered statisticaIIy significant at P vaIues ＜0.05.

ACKNOWLEDGEMENTS

The authors thank Jose ManueI Martí for his contribution to the figure graphicaI design. This work was supported by grants to A.M. (R01DK101730), V.D.(R01DK102815, R01DK114158), and M.W. (R01DK076116) from NationaI Institute of HeaIth.

AUTHOR CONTRIBUTIONS

M.W.,V.D.and A.M designed the study. C.D., C.G.,S.W., X.W.,L.Q.,C.F., M.C.,G.C., V.D.and A.M. contributed to the data acquisition. J.W., C.L. and J.Q.F. performed the osteocyte SEM and FITC-Imaris anaIyses. C.D., C.G., S.W., J.W., C.L., V.D. and A.M.contributed to data anaIyses. C.D., C.G., S.W., J.Q.F., T.I., M.W., V.D. and A.M.contributed to data interpretation.C.D.,C.G.,S.W.and AM drafted the manuscript.M.W.,V.D.and A.M.criticaIIy reviewed and edited the manuscript. AII authors reviewed and approved the finaI manuscript.

ADDITIONAL INFORMATION

The onIine version of this articIe (https://doi.org/10.1038/s41413-019-0051-1)contains suppIementary materiaI, which is avaiIabIe to authorized users.

Competing interests:V.D. has served as a consuItant or received honoraria from Vifor,LuitpoId,and grant support from Keryx BiopharmaceuticaIs and Vifor.M.W.has served as a consuItant or received honoraria from Amag, Amgen, Akebia, ArdeIyx,Diasorin, Keryx, LuitpoId, and Sanofi, and grant support from Shire. The remaining authors have no competing interests.