TNAP dephosphorylates phosphocholine and phosphoethanolamine and participates in triglyceride transport from the liver to the bloodstream

Abstract

Tissue-nonspecific alkaline phosphatase (TNAP) is primarily known for its role in skeletal mineralization, through the hydrolysis of inorganic pyrophosphate (PP_i). Here we demonstrate that TNAP-knockout mice exhibit liver steatosis and reduced serum triglyceride levels, mirroring the effects of choline deficiency, which impairs phosphatidylcholine synthesis, an essential component of VLDL. In fasting WT mice, TNAP inhibition via SBI-425 administration decreases choline levels in blood and liver. Incubating mouse or human serum with SBI-425 inhibits the dephosphorylation of phosphocholine and phosphoethanolamine, an alternative substrate for hepatic phosphatidylcholine synthesis. In hepatocytes, TNAP inhibition impedes proliferation when the medium is supplemented with phosphocholine instead of choline. Recombinant TNAP hydrolyzes phosphocholine and phosphoethanolamine with similar efficiency than PP_i. X-ray diffraction and cryo-EM identified the residues in TNAP’s active site interacting with phosphocholine, PP_i and the TNAP inhibitor. In summary, TNAP is the phosphatase enabling cellular choline uptake during fasting, participating in hepatic lipid metabolism.

Introduction

Atherosclerotic plaque calcification, a hallmark of plaque progression and a biomarker of disease severity, has been extensively studied over the last two decades due to its suspected impact on plaque rupture risk and cardiovascular mortality^1,2. Considerable efforts have been made to elucidate the molecular mechanisms underlying plaque calcification, with the goal of identifying therapeutic strategies to prevent plaque rupture. In this context, although the precise molecular triggers of plaque calcification remain largely unclear³, growing evidence suggests a key role for tissue-nonspecific alkaline phosphatase (TNAP)⁴. TNAP, a glycosylphosphatidylinositol (GPI)-anchored protein, is the only enzyme essential for bone mineralization in humans, owing to its ability to hydrolyze inorganic pyrophosphate (PP_i), a potent mineralization inhibitor primarily produced by the liver and released into the bloodstream, and also to a lesser extent, synthesized by other tissues⁵. Inherited TNAP deficiency leads to hypophosphatasia (HPP), a disease that, in its most severe forms, results in perinatal lethality due to a nearly complete lack of skeletal mineralization and consequent respiratory failure⁶. HPP is also associated with seizures due to impaired TNAP-mediated extracellular pyridoxal phosphate (PLP, vitamin B6) dephosphorylation, cellular pyridoxal uptake and intracellular rephosphorylation, and PLP-dependent GABA production⁶. As its name indicates, TNAP is, however, relatively ubiquitous. Beyond its role in physiological mineralization, TNAP is also suspected to induce or exacerbate pathological calcification in soft tissues. TNAP is indeed expressed in calcifying arteries⁷ and in the liver, which releases it in the circulation at levels associated with metabolic syndrome and cardiovascular calcification^8,9,10. Given that defects in hepatic PP_i generation lead to vascular calcification¹¹, TNAP is suspected to contribute to atherosclerotic plaque calcification by hydrolyzing PP_i locally in the vasculature, in the liver, and/or directly in the bloodstream⁴.

In this context, we recently investigated whether TNAP inhibition could prevent plaque calcification in atherosclerotic ApoE-deficient mice⁷. Surprisingly, we found that TNAP inhibition not only affected calcification but also impacted the liver, reducing blood triglyceride (TG) and cholesterol levels and impairing the entire process of atherosclerotic plaque development⁷. Using an NMR metabolomics approach, we identified phosphocholine as a potential TNAP substrate, establishing a possible link between TNAP activity and lipid transport between the liver and the bloodstream⁷. Hypothetically, TNAP would be the enzyme dephosphorylating extracellular phosphocholine, thereby enabling choline uptake by hepatocytes¹². Intracellular choline is a precursor of phosphatidylcholine (PC), the main phospholipid in the membrane of very low-density lipoproteins (VLDLs), necessary for VLDL assembly. VLDLs are then released into the bloodstream to deliver lipids to peripheral tissues. In this framework, TNAP would represent the missing link in the extracellular choline metabolic pathway (Fig. 1). TNAP would not be required for post-prandial choline metabolism, during which choline is released from dietary choline-containing lipids in the intestinal lumen, absorbed into enterocytes, transported through the portal vein, and delivered to hepatocytes^13,14. Instead, TNAP would play a role in post-absorptive metabolism, during which PC associated with lipoproteins undergoes sequential hydrolysis into lyso-PC, glycerophosphocholine (GPC), phosphocholine, and choline¹⁵ (Fig. 1). In this model, TNAP inhibition would recapitulate the effects of choline-deficient diets, which impair PC production and lead to liver steatosis¹⁶. This proposed role for TNAP is supported by the observation that TNAP-deficient Alpl^+/− mice develop liver steatosis¹⁷.

Fig. 1: choline and ethanolamine metabolism from the intestinal lumen to hepatocytes via the bloodstream.

PC is sequentially hydrolyzed in the intestine by sPLA₂-IB, PLB, and GDE5, or in the bloodstream by Lp-PLA₂, lyso-PLA₁, ENPP2, and/or ENPP6. TNAP could be the phosphatase dephosphorylating extracellular phosphocholine and phosphoethanolamine, allowing cellular choline and ethanolamine uptake. In hepatocytes, choline and ethanolamine generate PC and PEA via parallel metabolic pathways, and PEA can be methylated into PC by PEMT. PC is hydrolyzed into Lyso-PC and GPC in the endoplasmic reticulum by the sequential activity of PNPLA8 and PNPLA7, or by PLA₂G15 in lysosomes. Choline serves in hepatocytes as a methyl donor via the production of SAM. BHMT betaine homocysteine methyltransferase, CDP-choline cytidine diphosphate choline, CTP cytidine triphosphate, DH dehydrogenase, ENPP ectonucleotide pyrophosphatase phosphodiesterase, GDE5 glycerophosphodiesterase 5, GPC glycerophosphocholine, Lp-PLA₂ lipoprotein-associated phospholipase A, PC phosphatidylcholine, PLB phospholipase B, PEA phosphatidylethanolamine, PEMT phosphatidylethanolamine N-methyltransferase, PLA₂G15 phospholipase A, group XV, PNPLA patatin-like phospholipase domain containing, SAH S-Adenosyl-homocysteine, SAM S-Adenosyl-methionine, sPLA₂-IB secreted phospholipase A IB.

Additionally, PC can be endogenously produced by the methylation of phosphatidylethanolamine (PEA) (Fig. 1). However, the amount produced through this pathway is generally insufficient to fully compensate for severe choline deficiencies, which makes choline an essential nutrient¹⁸. We suspect that TNAP also plays a role in this alternative route of PC production, as elevated phosphoethanolamine levels in blood and urine are frequently observed in individuals with HPP¹⁹. Akin to phosphocholine, the hydrolysis of phosphoethanolamine into ethanolamine by TNAP would facilitate cellular ethanolamine uptake and subsequent phospholipid synthesis (Fig. 1). However, the potential involvement of TNAP in phosphocholine and phosphoethanolamine dephosphorylation, and its contribution to phosphatidylcholine synthesis and lipid homeostasis, remains speculative.

To address this hypothetical participation of TNAP in choline metabolism, we conducted experiments using TNAP-deficient and wild-type mice, human and mouse blood sera, and cultured human hepatocytes. In addition, we produced recombinant human TNAP in order to characterize its enzymatic activity and three-dimensional structure, and better understand its interaction with phosphocholine, phosphoethanolamine, and TNAP inhibitors. TNAP activity was inhibited in hepatocytes and in vitro using the MLS-0038949 inhibitor, and in sera and mice using SBI-425, a derivative of MLS-0038949 optimized for in vivo studies²⁰. Collectively, our findings demonstrate that TNAP efficiently dephosphorylates phosphocholine both in vitro and in vivo, acting in the bloodstream and at the hepatocyte membrane to enable choline uptake by hepatocytes. Furthermore, our structural analyses indicate that phosphocholine binds within TNAP’s active site in an opposite manner to that of PP_i, and that TNAP inhibitors block the entrance to the catalytic site, preventing the hydrolysis of both substrates.

Results

TNAP-deficient mice develop early-onset liver steatosis

Before thoroughly investigating whether TNAP dephosphorylates phosphocholine to facilitate lipid export from the liver to the circulation, we first aimed to assess whether TNAP-deficient mice exhibit abnormalities in TG homeostasis. Alpl^−/− mice were sacrificed at 11 days of age, shortly before their expected spontaneous death. Impressively, these young mice already displayed increased TG accumulation in the liver and reduced TG levels in plasma (Fig. 2). Consistent with previous reports in adult mice²¹, AP activity in the liver of young Alpl^+/+ mice was weak and primarily localized to the portal arteries within the portal triad (Fig. 2).

Fig. 2: TNAP-deficient mice exhibit liver steatosis and reduced lipidemia.

Oil red O staining of TG accumulation and AP activity in the liver of 11-day-old Alpl^+/+ and Alpl^−/− mice on the left (the bar represents 100 µM), and quantification of TG levels in the liver and plasma on the right. Results are shown as mean ± standard deviation, and statistical analyses were performed as detailed in the “Methods” section. p value for liver TG levels was calculated with a two-sided t-test; p value for plasma TG with a nonparametric Mann–Whitney test. AP alkaline phosphatase, TG triglyceride.

Short-term TNAP inhibition in wild-type mice acutely affects choline and ethanolamine metabolism

Since Alpl^−/− mice do not survive beyond weaning, we used adult wild-type mice and chemically inhibited TNAP to specifically assess its role in both post-absorptive and post-prandial choline metabolisms. Intraperitoneal administration of SBI-425 TNAP inhibitor (TNAPi) led to a marked reduction in serum AP activity, with 96% inhibition in fasted mice and 90% in refed animals (Supplementary Fig. 1B). Notably, after TNAPi administration, the residual AP activity remained slightly but significantly higher in refed animals compared to fasted mice. This difference was likely due to the absorption of intestinal alkaline phosphatase (IAP), which is not inhibited by the TNAPi used in this study ²⁰. Indeed, upon feeding, some IAP is absorbed, contributing between a few percent and up to 20% of total blood AP activity ²². Nonetheless, with more than 90% of blood AP activity inhibited, our experimental conditions effectively allowed us to investigate TNAP’s physiological functions.

Our primary objective was to assess the effects of TNAP inhibition on choline levels in both blood and liver. Analysis and quantification of metabolomics data first indicated that the levels of serum-soluble choline metabolites [choline, phosphocholine and glycerophosphocholine (GPC)] were higher in fasted than in fed animals (Fig. 3A and Supplementary Fig. 2), consistent with established knowledge of choline metabolism (Fig. 1). Indeed, among these molecules, only choline is directly absorbed by intestinal cells, with most of it taken up by hepatocytes from the portal vein before entering systemic circulation¹³. In contrast, during fasting, phosphatidylcholine (PC) in circulating lipoproteins undergoes hydrolysis into lyso-PC, GPC, phosphocholine, and choline, leading to higher levels compared to the post-prandial state¹⁵. As hypothesized (Fig. 1), TNAPi specifically affected choline metabolism during fasting, reducing the levels of choline, phosphocholine, and GPC (Fig. 3A and Supplementary Fig. 2). This suggests that TNAPi disrupted the entire cascade of PC hydrolysis into choline in the bloodstream, rather than solely impairing phosphocholine dephosphorylation. This is further supported by the observed positive correlation between serum choline and phosphocholine levels in both control and TNAPi-treated mice (Fig. 3B).

Fig. 3: Acute TNAP inhibition disrupts choline metabolism in adult wild-type mice.

A Serum levels of choline, phosphocholine, and GPC measured by NMR and B correlation between serum levels of choline and phosphocholine. C Liver content of choline and phosphocholine measured by NMR. D Liver content of PC. E Liver content of GPC. F Correlation between liver content of choline and GPC. G Liver content of phosphatidylethanolamine. H Kidney content of ethanolamine. Results are shown as mean ± standard deviation. Data were tested on normality and on equal variances with the Shapiro–Wilk test and Levene test, respectively. Then, data were analyzed on appropriate parametric t-test or a nonparametric Mann–Whitney test.

In fasted mice, TNAPi not only impacted choline metabolism in the bloodstream, but also in the liver, reducing both choline and phosphocholine levels (Fig. 3C and Supplementary Fig. 3). Despite this decline in choline and phosphocholine levels, total hepatic phosphatidylcholine levels showed a tendency to increase rather than decrease (Fig. 3D), suggesting that the disruption of PC hydrolysis in the bloodstream was associated with a retention of PC in hepatocytes. The PC species that accumulated the most significantly in the liver upon TNAP inhibition [38:4 (p = 0.035), 36:2 (p = 0.048), 36:1 (p = 0.06) (Supplementary Fig. 4)] are in agreement with the fatty acids in PC reported to be primarily sent from the liver to VLDL during fasting (PC with 16:0, 18:2, and 18:0 fatty acids)²³. Part of the PC that was retained in the liver was likely hydrolyzed intracellularly to provide hepatocytes with choline, since the levels of GPC, which increase during PC hydrolysis into choline but not during PC production from choline (Fig. 1), showed a rising trend (p = 0.06, Fig. 3E). Moreover, liver GPC levels were negatively associated with choline levels (Fig. 3F), indicating that the more choline levels were decreased, the more PC was hydrolyzed into choline via GPC. This adaptive PC hydrolysis into choline likely reflects a compensatory mechanism to mitigate choline deficiency and preserve sufficient levels for essential choline-dependent methylation reactions²⁴. Supporting this hypothesis, despite the reduced choline levels in the liver upon TNAPi administration, the concentrations of betaine, dimethylglycine, or sarcosine remained unchanged (Supplementary Fig. 3).

Interestingly, metabolomics data revealed that TNAPi significantly altered other metabolites directly or indirectly associated with choline metabolism. In particular, the levels of dimethylamine (DMA) strongly increased in liver, kidneys, and muscles in response to TNAPi (Supplementary Figs 3, 5 and 6). DMA has two known origins: the first is bacterial degradation of choline in the gut, followed by DMA absorption and urinary excretion without further transformation in the body ²⁵. Since intestinal bacteria and enterocytes compete for choline utilization²⁶, it may be possible that the compromised choline metabolism observed upon TNAP inhibition impacted choline absorption and increased choline availability for bacteria, resulting in increased DMA production. The second source of DMA is the catabolism of methylated arginine released after proteolysis by dimethylarginine dimethylaminohydrolase (DDAH)²⁷. DDAH activity is inhibited by homocysteine²⁸, a key intermediate in methionine metabolism under the dependence of TNAP activity through both choline (Fig. 1) and PLP status. Finally, TNAPi also affected the levels of two additional metabolites associated with methionine metabolism. TNAPi decreased the levels of formate in the liver, kidneys, and muscles and increased the levels of dimethylsulfone in the blood and liver (Supplementary Figs. 3, 5 and 6). Whether TNAPi affects the metabolism of these molecules through the dephosphorylation of phosphocholine, PLP, and/or other substrates remains unclear.

Taken together, these initial findings indicated that a short 3-h TNAP inhibition in fasted mice was sufficient to significantly affect choline metabolism in both the blood and liver. Whether TNAPi also impacted ethanolamine metabolism remained unclear, as ethanolamine levels in the blood and liver were below the detection limit of NMR. Moreover, the total amount of phosphatidylethanolamine in the liver was not affected by TNAPi (Fig. 3G). On the other hand, ethanolamine was detectable in the kidney, where its levels were significantly reduced by TNAP inhibition in fasted mice (Fig. 3H). This result aligns with observations in patients with TNAP deficiency, who often display increased phosphoethanolamine levels in the blood and urine¹⁹. It further supports the hypothesis that TNAP dephosphorylates not only phosphocholine in circulation, but also phosphoethanolamine.

Short-term TNAP inhibition in wild-type mice is sufficient to induce metabolic stress

We then explored using both targeted measurements and untargeted metabolomics analyses whether TNAPi administration induced metabolic changes characteristic of steatohepatitis associated with choline-deficient diets¹⁶. Acute TNAP inhibition for 3 h in fasted mice was likely too brief to significantly increase hepatic TG content or decrease blood TG levels (Fig. 4A). However, when fasted mice were allowed to eat and absorb choline, they exhibited a very significant decrease in hepatic TG content upon TNAP inhibition (p = 0.002), a change not observed in control mice (p = 0.35). While TNAPi had no significant effects on the levels of non-esterified fatty acids (NEFAs) in the serum or liver, it tended to increase serum glucose levels (Fig. 4C), possibly due to enhanced gluconeogenesis from glycerophosphate released during glycerophosphocholine hydrolysis into choline (Fig. 3E)²⁴. More unexpectedly, TNAPi appeared to impact the metabolism of amino acids in muscles. Indeed, TNAPi significantly decreased the levels of the branched-chain amino acids (BCAA) leucine, isoleucine, and valine, which are released by proteolysis during fasting (Supplementary Fig. 6). Whether these changes resulted from dysregulated choline metabolism remains unknown. Interestingly, the transaminase responsible for BCAA metabolism in muscle requires PLP²⁹. It transfers the amino group of BCAA on alpha-ketoglutarate to produce glutamate, whose levels were also decreased by TNAPi (Supplementary Fig. 6). Collectively, these data indicated that TNAPi impacted the metabolic status of fasted mice at multiple levels and in several tissues. In the liver, this metabolic disruption was likely stressful since 3 h of TNAPi administration was sufficient to increase Il1b and Fgf21 transcript levels (Fig. 4D). Hepatic expression of FGF21 is known to increase in mice fed choline-deficient diets³⁰ or when hepatocyte choline metabolism is genetically disrupted²⁴. IL-1β is one of the cytokines produced in the liver in murine models of steatohepatitis induced by choline-deficient diets³¹.

Fig. 4: Acute TNAP inhibition affects liver metabolism in adult wild-type mice.

A Quantification of TG levels in serum and liver. B Quantification of NEFA levels in serum and liver. C Quantification of glucose levels in serum and liver. D RT-qPCR quantification of transcript levels in the liver. Results are shown as mean ± standard deviation. Data were tested on normality and on equal variances with the Shapiro–Wilk test and Levene test, respectively. Then, data were analyzed on appropriate parametric t-test or a nonparametric Mann–Whitney test. Fgf21 fibroblast growth factor21, Il1 interleukin-1, NEFA non-esterified fatty acid, TG triglyceride.

TNAP is the key enzyme responsible for phosphocholine and phosphoethanolamine dephosphorylation in human and mouse blood

The impact of TNAP inhibition on choline metabolism in both blood and liver, despite TNAP’s weak activity in mouse liver (Supplementary Fig. 7 and ref. ²¹), suggested that its primary site of action was the circulation, indirectly influencing liver metabolism. To investigate this, we supplemented human or mouse serum with phosphocholine or phosphoethanolamine and monitored their hydrolysis into choline or ethanolamine over time. Using sera from six adult donors (three men and three women), we observed a progressive conversion of phosphocholine into choline (Fig. 5A–C) and of phosphoethanolamine into ethanolamine (Fig. 5G–I). Notably, the serum with the highest rates of phosphocholine (Fig. 5C) and phosphoethanolamine (Fig. 5I) dephosphorylation also exhibited the highest TNAP activity, as measured with its substrate pNPP (Fig. 5D). TNAP’s direct involvement in the hydrolysis of both phosphocholine and phosphoethanolamine was demonstrated by the complete inhibition of choline and ethanolamine formation upon TNAP inhibition (Fig. 5E, F, J, K, respectively). The same results were obtained when monitoring TNAP’s role in phosphocholine and phosphoethanolamine hydrolysis in mouse serum (Supplementary Fig. 8).

Fig. 5: TNAP is in the bloodstream, the enzyme that dephosphorylates phosphocholine and phosphoethanolamine.

NMR monitoring of phosphocholine (A) or phosphoethanolamine (G) hydrolysis into choline or ethanolamine in human sera over 6 h, after the addition of phosphocholine or phosphoethanolamine at 1 mM. The same experiments were performed in the presence of TNAPi (E, J). For each experiment, an example of NMR spectra is shown on the left (A, E, G, J), and choline or ethanolamine quantification for all donors is shown on the right during time (B, F, H, K), or as formation rate in µM per hour (C, I). The apparition rate of choline and ethanolamine in the presence of TNAPi is shown (F, K). Serum AP activity for all donors is shown (D).

Human hepatocytes use TNAP to dephosphorylate phosphocholine and phosphoethanolamine

Unlike mouse hepatocytes, human hepatocytes express TNAP in vivo. We therefore investigated whether TNAP anchored at the hepatocyte membrane, as in the bloodstream, was involved in phosphocholine and phosphoethanolamine dephosphorylation. To address this, we cultured three human hepatocyte cell lines. Among them, HuH-6 cells were selected because they exhibited the highest AP activity (25 nmol/min/mg), compared to HuH-7 cells (5 nmol/min/mg) and HepG-2 cells (2.5 nmol/min/mg). In HuH-6 cells, both phosphocholine and phosphoethanolamine were clearly hydrolyzed, and this process was inhibited by 60% by TNAPi (Fig. 6A). Given the high selectivity of the TNAPi used here³², and the fact that HuH-6 cells were found by RT-qPCR to express not only TNAP but also IAP transcripts, it is likely that the remaining 40% of hydrolysis was due to IAP-mediated activity.

Fig. 6: TNAP at the hepatocyte membrane dephosphorylates phosphocholine and phosphoethanolamine.

A Quantification of P_i produced by HuH-6 hepatocytes during the hydrolysis of phosphocholine or phosphoethanolamine, in the presence or absence of TNAPi. B DNA quantification in HuH-6 cells cultured or not with choline and phosphocholine, and in the presence or absence of TNAPi. Results are shown as mean ± SEM from 5 independent experiments for substrate dephosphorylation, and from 3 independent experiments for DNA quantification. For the sake of clarity, individual data points are not presented in (A). However, source data are available in the Supplementary data file. Statistical analysis was performed with a paired t-test.

Dephosphorylation of phosphocholine by TNAP is essential for choline uptake in hepatocytes

TNAP, both in the bloodstream and at the hepatocyte membrane, is therefore capable of dephosphorylating phosphocholine and phosphoethanolamine. We next investigated whether TNAP-mediated phosphocholine dephosphorylation is required for choline uptake in hepatocytes. To address this, hepatocytes were cultured for 7 days in a choline-depleted medium, and supplemented or not with 25 μM of choline or 25 μM of phosphocholine, in the presence or absence of TNAPi. Our hypothesis was that cells would be able to grow in a choline-depleted but phosphocholine-supplemented medium only if TNAP was active and able to hydrolyze phosphocholine into choline. As expected, in the absence of both choline and phosphocholine, cell proliferation was completely blocked, as measured by DNA quantification (Fig. 6B). When either choline or phosphocholine was added to the medium, proliferation was restored. As hypothesized, TNAPi had no effect when cells were cultured in the presence of choline, but strongly reduced cell proliferation (by 57%) when cells relied on phosphocholine as their sole source of choline (Fig. 6B). Together, these results strongly suggested that TNAP plays a crucial role in dephosphorylating phosphocholine to facilitate choline uptake in hepatocytes.

Recombinant human and mouse TNAP efficiently dephosphorylate phosphocholine and phosphoethanolamine in vitro

Our findings in mice, sera, and cells strongly suggested that, alongside PP_i and PLP, phosphocholine and phosphoethanolamine are physiological TNAP substrates. To further investigate this, we characterized and compared the ability of recombinant human and mouse TNAP to dephosphorylate phosphocholine and phosphoethanolamine relative to PP_i. We assessed their hydrolysis kinetic parameters in vitro using a Michaelis-Menten model incorporating substrate inhibition. Notably, for both human and mouse TNAP, while the K_m values for phosphocholine and phosphoethanolamine were higher than for PP_i, the catalytic efficiencies were greater, as indicated by the 8-to-10 fold higher k_cat for human TNAP (Fig. 7 for human TNAP and Supplementary Fig. 9 for mouse TNAP). The higher k_cat observed for phosphocholine and phosphoethanolamine compared to PP_i might reflect their opposite charges, as choline and ethanolamine are positively charged and PP_i is negatively charged. These charge differences led us to investigate how TNAP’s active site can accommodate both PP_i, phosphocholine, and phosphoethanolamine.

Fig. 7: TNAP efficiently dephosphorylates phosphocholine and phosphoethanolamine.

A Michaelis-Menten representation, i.e., reaction rate (mM.h⁻¹) of phosphocholine, phosphoethanolamine, and PP_i hydrolysis by human TNAP, depending on the substrate concentration. Experimental data are presented, as well as the fit by a model with substrate inhibition, i.e., following the equation V₀ = V_max[S]/([S]+K_m + [S]²/K_i), where V₀ is the initial reaction rate, V_max the maximal reaction rate, and [S] the substrate concentration. All experiments were repeated at least three times. For the sake of clarity, individual data points are not presented in (A). However, source data are available in the Supplementary data file. B Kinetic parameters (K_m, k_cat, k_cat/K_m, and K_i) are given for phosphocholine, phosphoethanolamine, and PP_i hydrolysis.

Phosphocholine enters into TNAP’s active site through a gate opposite to that used by PP_i

The recombinant human TNAP we produced for enzymatic studies was also used for crystallization experiments. A crystal diffracting at 3.35 Å resolution was obtained, enabling structural determination via molecular replacement, using human TNAP as the template (PDB code 7YIV). The protein crystallized in the space group C222₁, with five subunits in the asymmetric unit (Table 1). As observed in other APs, the conserved dimeric organization was maintained. The same divalent cations were identified, including the three essential cations in the catalytic site (two zinc ions and one magnesium ion), along with a calcium ion at the CA cluster, as seen in the apo structure of the enzyme (PDB codes 7YIV or 7YIW). Crystallization was carried out in the presence of 0.2 M ammonium phosphate, systematically leading to the observation of one inorganic phosphate at the catalytic site. This P_i coordinates with one of the catalytic zinc ions (Zn2) (Fig. 8A) and interacts with residue R184, whose side chain reorients towards the P_i, positioning it at approximately 3.5 Å from the catalytic serine. To investigate how the active site accommodates phosphocholine or phosphoethanolamine, we conducted docking experiments using the apo form of the enzyme. Structure-based virtual “blind” docking was performed on one of the monomers from the structural and functional dimer, due to computational constraints related to the large size of the dimer. Docking analyses of PP_i, phosphoethanolamine, and phosphocholine yielded ΔG values ranging from −4.9 to −3.7 kcal/mol (Supplementary Table 1). These experiments revealed two distinct binding modes for PP_i and phosphocholine. For PP_i, all docking poses displayed an almost identical orientation, with one of its phosphate groups aligning precisely with the P_i observed in the enzyme’s complex structure. The second phosphate extended toward a small positively charged pocket, where it could interact with two arginine residues: R168 and R184 (Fig. 8A). In contrast, phosphocholine exhibited a completely opposite orientation. Five out of ten docking poses displayed very similar binding modes, where the positively charged quaternary amine was nestled into a large negatively charged groove, shaped primarily by two acidic residues, D109 and E452 (Fig. 8A). While these residues do not directly interact with the positive charge, they create an electrostatically favorable environment for its accommodation. Additionally, the positive charge of phosphocholine restricts access to the positively charged pocket, and the presence of its three methyl groups further reinforces this restriction. Notably, the phosphate group of phosphocholine is positioned identically to the phosphate observed in the enzyme’s complex structure. For the remaining docking poses, the quaternary amine of phosphocholine tends to orient toward the solvent, leaving only the phosphate group to interact with the zinc ion. Phosphoethanolamine displayed greater variability in binding modes compared to PP_i and phosphocholine. Some docking poses aligned with phosphocholine’s orientation, while others resembled that of PP_i (Fig. 8C). Unlike the quaternary amine in phosphocholine, which remains consistently positively charged, phosphoethanolamine can exist in different ionization states. Combined with its lack of methyl groups (as compared to phosphocholine), this reduced steric hindrance allows phosphoethanolamine to adopt diverse binding modes. It is conceivable that under varying pH conditions, phosphoethanolamine could favor one of the two orientations observed for PP_i and phosphocholine. Overall, these structural studies confirm that TNAP’s active site can accommodate both the negatively charged PP_i and the positively charged phosphocholine. This is made possible by phosphocholine entering the active site through an opposite pathway compared to PP_i.

Fig. 8: TNAP’s active interaction with phosphocholine, phosphoethanolamine, PP_i, and TNAPi.

Substrate accommodation in TNAP’s active site. The best docking poses for phosphocholine and PP_i obtained with the Apo enzyme are superimposed onto the TNAP structure containing P_i: A Secondary structures and key residues that define the negatively charged zones (D109 and E452) and positively charged zones (R168 and R184). B The electrostatic surface is color-coded to depict the charge distribution, with red representing negatively charged regions (the groove) and blue representing positively charged regions (the pockets). C Two examples demonstrate the orientation of phosphoethanolamine within the active site: one positioned in the positively charged pocket and the other within the negatively charged groove. D Structure of the TNAPi MLS-0038949 (MLS) and global representation of MLS interaction with TNAP’s surface. E MLS forms P_i–cation interactions between the para-di-methoxy-benzene group and Zn²⁺ (Zn2). This interaction is further stabilized by a hydrogen bond between His341 and the sulfonamide group of MLS. Additionally, cation–P_i–stacking interactions are observed between Glu342, His338, and the quinolin-3-yl group. Finally, His338 contributes to the binding through a P_i–stacking interaction with the quinolin-3-yl group of the MLS. F MLS positioning relative to the different substrates of TNAP. The structure in the presence of MLS was superimposed onto the structure containing P_i, as well as the structure used for docking with phosphocholine and PP_i. This comparison highlights the steric hindrance caused by the para-di-methoxy-benzene part of the inhibitor in relation to the phosphate group of the different substrates. G Superposition of the TNAP structure (in green) containing MLS with the AlphaFold model of IAP (in gray), illustrating the interaction specificity between TNAP and the MLS. The three residues interacting with MLS, which differ between TNAP and IAP, are indicated in the respective colors of the two structures.

Table 1 Data collection and refinement statistics (molecular replacement)

TNAPi closes the active site for all substrates

To our knowledge, the most effective chemical scaffolds developed so far to specifically inhibit TNAP rely on a sulfonamide core. The first two sulfonamide-associated TNAP inhibitors were the MLS-003899 inhibitor we used for in vitro and cell culture experiments, and the SBI-425 molecule optimized for in vivo studies, which we used in mice²⁰. The third one is the DS-1211 version developed for human clinical trials³³. We sought to determine how these sulfonamide-based molecules bind and inhibit TNAP, and whether they may differently block the hydrolysis of PP_i, PLP, phosphocholine, and phosphoethanolamine. To this end, we developed a structural approach using single particle analysis by cryo-EM (Supplementary Table 2), to determine how MLS-0038949 (MLS) inhibits human TNAP. The inhibitor was clearly resolved in the density at both active sites (Fig. 8D). Interestingly, it is positioned at the entrance of the catalytic site, with the para-di-methoxy-benzene group positioned directly above Zn²⁺ (site 2) (Fig. 8E–G), forming a cation-π interaction. In this orientation, the aromatic ring is situated 3.8 Å from the catalytic serine. Under this configuration, histidine 341, which is involved in the coordination of zinc number 2, is 3.5 Å from the sulfonamide group, enabling the establishment of a hydrogen bond and contributing to the stabilization of the inhibitor while orienting it. The quinolin-3-yl core seems to be oriented more toward the solvent. However, glutamate 342 and histidine 338, located at 3.4 and 3.3 Å, respectively, might further contribute to stabilization via cation-π interactions. Another residue of importance for the inhibitor’s interaction is histidine 451, whose side chain points toward the nitrogen of the quinolin-3-yl group, at only 3.3 Å from the heterocycle, forming a π-π interaction that helps stabilize MLS. In this configuration, the para-di-methoxy-benzene group prevents the binding of various TNAP substrates, replacing the phosphate found in phosphocholine or phosphoethanolamine (Fig. 8B). MLS has been described as specifically inhibiting TNAP. Our data indeed show that other human APs exhibit a catalytic site organization that could explain this phenomenon. In particular, IAP, instead of histidine 451, contains a serine, which is unable to form a π-π interaction with quinolin-3-yl, potentially causing this group to reorient toward the solvent and destabilize the overall interaction of MLS with the enzyme. Histidine 451 is replaced by a glycyl residue in human germ cell alkaline phosphatase (GCAP) and a phenylalanine residue in placental alkaline phosphatase (PLAP). Additionally, the C119-R138 loop (C120-Q139 in IAP), which participates in the structuring of the catalytic site, has residues E125 and G126 present in TNAP, replaced by residues F126 and Q127 in IAP. Phenylalanine and glutamine are conserved in both GCAP and PLAP. These substitutions create steric hindrance that greatly limits the positioning of MLS, particularly the accommodation of the 2-O-methyl group. These observations confirm that the sulfonamide-associated TNAP inhibitors used in the present study specifically inhibit TNAP, and provide structural evidence of this specificity, and importantly indicate that they probably block the hydrolysis of all TNAP substrates.

Discussion

Our in vivo, in vitro, and in silico findings indicate that TNAP plays a pivotal role in lipid metabolism by dephosphorylating extracellular phosphocholine, thereby facilitating choline uptake into hepatocytes. While previous studies have shown that heterozygous TNAP-deficient mice develop liver steatosis when subjected to a high-fat diet¹⁷, our research reveals that homozygous TNAP-deficient mice exhibit significant liver steatosis and reduced blood TG levels as early as 2 weeks after birth, prior to weaning (Fig. 2). This early onset of hepatic steatosis in the absence of TNAP demonstrates that the enzyme is not merely a secondary participant in lipid metabolism, but likely serves as a key regulatory factor. We hypothesized that TNAP influences TG distribution by participating in choline metabolism, as TG retention in the liver is a hallmark of choline deficiency¹⁶. Our previous observation of dysregulated choline metabolism and reduced TG levels in blood upon TNAP inhibition in mice with atherosclerosis supports this hypothesis⁷. In the present study, acute TNAP inhibition during fasting significantly decreased choline levels in both serum and liver. During fasting, hepatocytes take up NEFAs released by adipocytes; some are beta-oxidized to produce ATP, and some are re-esterified and released into the bloodstream within VLDLs³⁴. As VLDLs deliver their TG content to tissues, they become remnant lipoproteins that return to the liver. Hepatocytes require substantial amounts of choline to synthesize VLDLs, as phosphatidylcholine constitutes up to 70% of VLDL phospholipids³⁵. During fasting, when choline absorption is minimal, it is crucial for hepatocytes to retrieve choline from remnant lipoproteins. While the exact metabolic sequence of lipoprotein PC hydrolysis into choline is still not fully characterized, our findings suggest that TNAP plays a significant role in this process.

In an elegant study, Morita et al. reported that ectonucleotide pyrophosphatase phosphodiesterase (ENPP6), an alkaline phosphatase superfamily member expressed in liver sinusoidal endothelial cells, is crucial for hydrolyzing extracellular glycerophosphocholine into phosphocholine, ultimately facilitating hepatocyte uptake of choline³⁶. Our data demonstrate that TNAP is the phosphatase acting downstream of ENPP6 to generate extracellular choline from phosphocholine. In mice, TNAP-mediated phosphocholine dephosphorylation likely occurs in the bloodstream. We indeed confirm that, in contrast to the human liver, the mouse liver exhibits very low TNAP activity, which is not localized in hepatocytes (Supplementary Fig. 7)²¹. TNAP is, however, active in mouse blood, where it is responsible for phosphocholine dephosphorylation (Supplementary Fig. 8). In humans, where hepatocytes express TNAP, phosphocholine hydrolysis into choline is performed both by circulating TNAP and TNAP anchored at the hepatocyte membrane (Figs. 5 and 6). Moreover, we show that TNAP activity is necessary for hepatocyte proliferation when cells are grown in a choline-depleted medium supplemented with phosphocholine (Fig. 6). Mechanistically, the dephosphorylation of extracellular phosphocholine by TNAP and the subsequent rephosphorylation of intracellular choline by choline kinase are pivotal reactions regulating cellular choline uptake and metabolism. Choline needs to be dephosphorylated to be transported into cells via choline transporters. The initial choline transporters identified were the ubiquitous low-affinity SLC44A1 transporter and its closely related counterparts SLC44A2 and SLC44A3, which may also transport ethanolamine³⁷. More recently, two high-affinity transporter paralogs, FLVCR1 and FLVCR2, were discovered^37,38. FLVCR1 and FLVCR2 have K_m values for choline around 5 µM, similar to the concentrations of choline reported in the blood and extracellular fluids, around 10 µM and 3–6 µM respectively^37,38, and less than those measured in our study in mouse serum (10 µM during fasting and 20 µM after refeeding, Supplementary Fig. 2). Once inside the cells, choline is rapidly rephosphorylated by choline kinase, leading to elevated levels of phosphocholine^37,38. This process ensures that cytosolic concentrations of choline remain sufficiently low to maintain a concentration gradient driving choline uptake from the extracellular space, akin to the mechanism driving glucose entry into cells. TNAP, therefore, likely contributes to this driving force by increasing the extracellular concentration of choline.

In addition to choline, hepatocytes can use ethanolamine to produce phosphatidylethanolamine, which is subsequently methylated to form phosphatidylcholine¹⁸. Although this compensatory pathway cannot fully compensate for severe choline deficiency, it helps sustain PC production under mild choline scarcity. A TNAP deficiency may also disrupt this alternative route of hepatic PC synthesis. HPP, the disease resulting from genetic TNAP deficiency, is often associated with elevated levels of phosphoethanolamine in blood and urine⁶. In our experiments, while ethanolamine and phosphoethanolamine levels in the liver were too low to be quantified, analyses in the kidney revealed strongly decreased ethanolamine levels upon TNAP inhibition, strengthening the notion that TNAP is important for cellular accumulation of ethanolamine in vivo. We further provide evidence that TNAP is the enzyme that dephosphorylates phosphoethanolamine in human blood, as well as at the hepatocyte membrane (Figs. 5 and 6), suggesting a role in liver PC production from blood ethanolamine. Interestingly, phosphoethanolamine levels in the urine of HPP individuals appear to be inversely correlated with the activity of liver-derived TNAP released in the serum, but not with that of circulating bone TNAP³⁹. Liver and bone TNAP are likely both able to hydrolyze phosphoethanolamine, since mouse serum, which contains bone but not liver TNAP²¹, used TNAP to dephosphorylate phosphoethanolamine (Supplementary Fig. 8). Nevertheless, the specific association of phosphoethanolamine levels in HPP patients with the liver but not the bone TNAP isoform may suggest a role of liver TNAP in ethanolamine production in relation to PC synthesis. In conclusion, TNAP appears to participate in physiological lipid distribution through the extracellular dephosphorylation of phosphocholine and phosphoethanolamine, and the subsequent cellular uptake of choline and ethanolamine.

On the other hand, we cannot totally exclude the possibility that TNAP impacts lipid metabolism in part through PLP dephosphorylation and vitamin B6-dependent reactions. Alpl^−/− mice die from epileptic seizures due to impaired extracellular PLP dephosphorylation, cellular PL uptake and rephosphorylation, and PLP-dependent GABA production⁴⁰. Among the 150 or 160 reactions necessitating PLP are both the formation of phosphoethanolamine from sphingosine-1 phosphate and its degradation into acetaldehyde^41,42. However, it is unlikely that liver steatosis that develops in both Alpl^−/− mice (Fig. 2) and Alpl^+/− mice¹⁷ mainly relies on PLP deficiency. Indeed, Alpl^−/− mice but not Alpl^+/− mice develop vitamin B6 deficiency and suffer lethal epileptic seizures⁴⁰, while TNAP^+/− still develop liver steatosis¹⁷. Moreover, while vitamin B6 deficiency affects liver metabolism in mice, it rather tends to increase and not decrease serum TG⁴³. Furthermore, it is also unlikely that our experiments based on a single short SBI-425 administration induced a harmful vitamin B6 deficiency. Indeed, epileptic seizures are among the earliest symptoms of vitamin B6 deficiency⁴¹, and seizures were not detected in a dozen of mouse studies based on much longer SBI-425 treatments^7,44,45. In conclusion, we consider it likely that TNAP participates in the physiological tissue distribution of triglycerides more through the extracellular dephosphorylation of phosphocholine and phosphocholine than through PLP availability.

Mice deficient in TNAP or ENPP6 both exhibit liver steatosis due to the involvement of these enzymes in choline metabolism in the liver. Interestingly, both TNAP- and ENPP6-knockout mice also share other symptoms likely resulting from choline deficiency. For instance, both ENPP6- and TNAP-deficient mice exhibit brain hypomyelination^36,46, likely because choline levels are too low to generate sufficient amounts of sphingomyelin. Again, it is also possible that TNAP deficiency leads to hypomyelination by altering PLP levels^46,47. In addition to decreased myelination, both ENPP6 and TNAP-deficient mice present with hypomineralization^48,49. TNAP may therefore stimulate mineralization not only by hydrolyzing PP_i, but also by allowing osteoblasts to import choline and produce pro-mineralizing, phosphatidylcholine-enriched, matrix vesicles⁵⁰. Given the relatively ubiquitous expression of ENPP6 and TNAP, it is possible that the impairment of extracellular choline generation by their deficiency not only impacts the liver, the brain, and bones, but also other tissues.

Interestingly, patients with HPP often suffer from muscle weakness⁶. TNAP activity in muscles is quite weak as compared to that in the kidneys, but slightly above that in the liver (Supplementary Fig. 1C), suggesting that TNAP activity may play a role in muscle metabolism, not only from the bloodstream, but also directly in muscle cells. We observed that TNAP inhibition significantly decreases the levels of branched-chain amino acids in muscles, which are released by proteolysis during fasting. These BCAAs are crucial molecules for both muscle energy metabolism and functions in the liver, where their catabolic products are sent²⁹. Whether disturbed choline metabolism is responsible for this effect, and whether ENPP6 deficiency also impacts the muscles, are open questions. On the other hand, the enzyme involved in BCAA catabolism in muscles is PLP-dependent, and its deficiency in mice leads to low muscle mass and reduced endurance to exercise²⁹. To conclude, whether the dysregulation of BCAA metabolism in muscles upon TNAP inhibition relies, at least in part, on choline availability or PLP levels, and whether it contributes to muscle weakness in individuals with HPP, is obscure but deserves investigation.

The involvement of TNAP in choline metabolism is important to consider for individuals with HPP. Determining whether human HPP is associated with choline deficiency and related symptoms is challenging. Severe TNAP deficiency is often lethal shortly after birth due to seizures and respiratory complications, making it difficult to identify liver-related symptoms. Treatment with recombinant TNAP now enables children to transition from hypophosphatasia towards hyperphosphatasia⁵¹, which may alleviate liver choline deficiency. That said, about 20% of young patients treated with recombinant TNAP were reported to develop hepatitis^52,53. The origin of these liver symptoms is unknown but deserves investigation. Besides children, who are often treated, adults with moderate TNAP deficiencies are frequently untreated, but the rarity of HPP, combined with its moderate presentation in these cases, makes it difficult to ascertain whether they are more susceptible to developing liver steatosis compared to the general population. Nonetheless, assessing their serum choline levels could reveal subnormal values, in which case choline supplementation would be a straightforward prophylactic measure. We will nevertheless have to take into account that while TNAP deficiency mimics choline deficiency, it differs from it by several features. In particular, choline-deficient diets improve glucose tolerance⁵⁴, while TNAP inhibition tends to increase glycemia (Fig. 4C). These differences likely rely in part on the fact that TNAP deficiency does not generate a true choline deficiency, because it may not compromise choline absorption (Fig. 1). Instead, TNAP deficiency or inhibition likely disturbs choline distribution, on the one hand in different cells and tissues (choline but not phosphocholine can be transported intracellularly), and on the other hand in different choline-containing molecules (choline, phosphocholine, glycerophosphocholine, phosphatidylcholine and others). Moreover, when dietary choline is deficient, phosphatidylethanolamine is methylated by PEMT to help maintain phosphatidylcholine levels (Fig. 1). When TNAP is deficient or inhibited, not only choline metabolism is altered, but also ethanolamine metabolism. It is therefore possible that choline supplementation will not satisfyingly correct choline disturbances induced by TNAP deficiency.

The structural data presented in this article may provide insights into the diversity and variability of symptoms observed in HPP. As already mentioned, HPP is primarily characterized by hypomineralization, seizures, musculoskeletal pain, and muscle weakness. Notably, some HPP patients exhibit only a subset of these symptoms. For instance, an analysis of 151 children with HPP prior to treatment initiation revealed that skeletal manifestations were present in 67.5% of cases, while muscular and neurological symptoms were observed in 48% and 40% of cases, respectively ⁵⁵. These observations may suggest that specific TNAP mutations may differently affect the enzyme’s ability to hydrolyze various substrates, leading to distinct clinical presentations. Notably, elevated levels of phosphoethanolamine are not observed in all HPP patients¹⁹. Importantly, Di Mauro et al. showed that some mutations found in HPP patients differently affect the hydrolysis of PP_i and PLP⁵⁶. Recent studies on another alkaline phosphatase superfamily member, ENPP1, revealed that specific mutations selectively inhibit the hydrolysis of ATP or that of cyclic guanosine monophosphate-AMP without affecting the other⁵⁷. The data we present here suggest that the hydrolysis of PP_i and phosphocholine may also be differently affected by specific mutations. Indeed, we provide evidence that the positively charged phosphocholine enters TNAP’s active site through a pathway opposite to that used by PP_i, interacting with different amino acid residues (Fig. 8). This finding implies that certain mutations could selectively impair the hydrolysis of one substrate while preserving the hydrolysis of another. Our studies indicate that PP_i likely interacts with the residue R184. Variants affecting this residue are associated with mild forms of HPP exhibiting a dominant negative effect, and severe forms when combined with another heterozygous variant⁵⁸. Conversely, phosphocholine appears to interact with the residue E452. Notably, the E452K variant was discovered as a single heterozygous mutation causing a mild adult form of HPP through a dominant negative effect⁵⁹. Investigating whether mutations at R184 and E452 residues differently impact TNAP’s ability to hydrolyze PP_i and phosphocholine, thereby leading to distinct clinical symptoms, warrants further research.

Finally, our results may also be clinically relevant now that patients with pathological calcification are beginning to be treated with TNAP inhibitors. A phase II clinical study (NCT05569252) has been completed in individuals with pseudoxanthoma elasticum, a rare genetic disease due to decreased PP_i production, and associated with ectopic calcification. This trial is likely to be followed by others, in particular, to block cardiovascular calcification in patients with chronic kidney disease and reduce their cardiovascular mortality. Our cryo-EM analyses demonstrate that TNAP inhibitors containing a sulfonamide linker (MLS-0038949, SBI-425, and DS-1211) bind at the entrance of TNAP’s active site (Fig. 8), thereby blocking the hydrolysis of virtually all TNAP substrates. It is therefore likely that patients treated with DS-1211 will develop choline deficiency and subsequently experience all associated metabolic dysregulations. Choline levels should therefore be measured in treated patients, and as for patients with HPP, choline supplementation could be a straightforward measure to compensate for choline deficiency.

In conclusion, the present study indicates that TNAP is the extracellular enzyme dephosphorylating phosphocholine into choline during fasting, allowing hepatocytes to uptake choline. This function is probably not restricted to the liver. Other cell types, in the brain or bones, likely rely on TNAP to obtain choline and produce not only phosphatidylcholine, but also acetylcholine and/or betaine to accomplish their respective functions. Therefore, intense efforts are needed to better understand the pathophysiological consequences of TNAP hydrolysis of phosphocholine, in particular in patients with TNAP deficiency and in patients treated with TNAP inhibitors.

Methods

Mouse experiments

Alpl ^+/+ and Alpl ^+/− mice

Mouse pups born from Alpl^+/− × Alpl^+/− parents (Alpl^tm1Jlm) were genotyped by PCR using toe biopsy and collected at postnatal day 11. Pups were anesthetized with Avertin (0.017 mL/gram body weight), and blood obtained via cardiac puncture was used for the preparation of heparin plasma. Liver tissue and plasma samples were stored in a - 80°C freezer until further analyses. All animal procedures were approved by the AUF committee in Sanford Burnham Prebys Medical Discovery Institute (La Jolla, USA). We have complied with all relevant ethical regulations for animal use. A part of the liver samples was fixed in 4% paraformaldehyde, and optimal cutting temperature (OCT) medium sections were prepared as previously described⁶⁰. Localization of lipids was visualized with Oil Red O stain by the standard protocol. Alkaline phosphatase (AP) activity was stained with the Azo Dye methods⁶¹. TG levels in plasma and liver were measured using a kit from Cayman Chemical Co. (Ann Arbor, MI) by following the manufacturer’s protocol. Briefly, the liver tissue was homogenized in the supplied dilution buffer supplemented with 5 mM EDTA, and the supernatant was used for the assay. The protein concentration of each sample was determined by using BCA reagent (Pierce, Rockford, IL).

Wild-type mice

Eight-week-old male mice (Janvier Labs, France) were housed at the animal facility of INSA (Lyon, France; Cetil n◦C2EA-102), and kept in conventional cages with four mice per cage. All animal experiments were performed with the approval of our institutional animal care committee and of the ethics committee of the French Ministry of Education, Research and Innovation (APAFIS#15387-2018060708052279). We have complied with all relevant ethical regulations for animal use. A total of 40 mice were utilized and divided into four groups (n = 10) with varying treatments (Supplementary Fig. 1A). All mice underwent a 10-h fasting period and half of them were then subjected to an intraperitoneal injection of 10 mg/kg of SBI-425, which was dissolved at 1 mg/mL in a solvent composed of 10% DMSO and 10% Tween 80. SBI-425 was synthesized at the Prebys Center for Drug Discovery, Sanford Burnham Prebys Medical Discovery Institute (La Jolla, USA)²⁰. The remaining control mice were injected with the solvent alone. One hour after the injection, half of the mice were provided with food for an additional 2 h, while the other half continued to fast for the same duration (Supplementary Fig. 1A). Then, mice were euthanized by cervical dislocation after anesthesia with isoflurane and intracardiac blood puncture, and organs and tissues were collected, immediately frozen in liquid nitrogen and stored at −80 °C.

Biochemical analyses

AP activity assay in the serum was adapted from a previously published protocol and performed in a 96-well plate, with pNPP as substrate⁶². TG (MAK266 SIGMA), NEFAs (Ab65341 Abcam), and glucose (EIAGLUC Invitrogen) levels were assayed in serum and liver using colorimetric kits and following the manufacturer’s protocols. Staining of AP activity in liver sections was performed on an 11-week-old mouse. Briefly, the liver was rinsed with sterile phosphate-buffered saline (PBS), pH 7.4, frozen in OCT (Thermo Scientific), and stored at −80 °C before serial cryo-sectioning. Cryostat sections (10-μm thick) were cut at −21 °C using a Leica CM3050 S cryostat (Leica), mounted on glass slides (SuperFrost Plus Gold glass slides, Thermo Scientific). Sections were stained for AP activity using nitroblue tetrazolium/5-Bromo-4-chloro-3-indolyl phosphate (NBT/BCIP) solution⁷.

RNA extraction, reverse transcription, and quantitative polymerase chain reaction (RT-qPCR)

Mouse organs were ground in a liquid nitrogen-cooled mortar. Ground organs were lysed and purified with the NucleoSpin® RNA II kit (Macherey-Nagel) following the manufacturer’s protocol. One microgram of each RNA sample was retro-transcribed with Superscript II reverse transcriptase (Invitrogen). Quantitative PCR was performed using iTaq Universal SYBR green supermix on the CFX96 Touch Real-Time PCR Detection System (Biorad). All primers were obtained from Sigma. The primer sequences and PCR conditions are given in Supplementary Table 3. Relative quantification analyses were calculated using the 2^−ΔCt method by the CFX Manager software (Biorad). Gapdh (glyceraldehyde-3-phosphate dehydrogenase), Rpl13a (Ribosomal Protein L13a), and Hprt (Hypoxanthine-guanine phosphoribosyltransferase) were used as references. The target gene expression values were calculated by normalization with the geometric mean of the reference values.

Metabolomics and lipidomics analyses

Mouse frozen tissues were weighed, and extracts were prepared following the Beckonert protocol⁶³, which involves a chloroform:methanol:water extraction [2:2:1.425 (v/v/v)]. Organ and serum NMR spectra were obtained on an 800 MHz Ascend Avance III Bruker spectrometer with an Avance Neo console and a cryoplateform, equipped with a 3 mm BBI probe. All spectra were acquired at 30 °C. Before spectra acquisition, samples were maintained at 4 °C and handled by a Bruker SampleJet automated sample changer. ¹H 1D NMR spectra were recorded for each sample, using zgpr for liver, kidneys, and muscles, and PROJECT (quantitative cpmg)⁶⁴ for serum. Both sequences were acquired with 128 scans, a relaxation delay of 6.6 s,and an acquisition time of 3.4 s. 2D NMR experiments were conducted with standard parameters to facilitate peak assignment in 1D spectra. ¹H–¹H TOCSY, ¹H–¹³C HSQC, and J-Resolved were recorded on a liver sample, and ¹H–¹H TOCSY and J-Resolved were recorded on a serum sample. Non-Uniform Sampling was used for the TOCSY and the HSQC up to 25%. Prior to the Fourier transform, NMR free induction decays underwent multiplication by an exponential function corresponding to a 0.3 Hz line-broadening factor. The sodium 2,2,3,3-tetradeutero-3- trimethylsilylpropionate (TSP) singlet at −0.016 ppm was used to calibrate liver, kidneys, and muscles spectra, and TSP was also used as concentration reference. Serum NMR spectra were calibrated using the center of the glucose doublet at 5.230 ppm via Topspin 3.6 (Bruker GmbH, Rheinstetten, Germany). A commercial standard lactate solution (1 g/L, Fisher) served as a concentration reference. The ERETIC2 utility from TopSpin incorporated a synthetic peak into all spectra, facilitating cross-referencing⁶⁵. Metabolite identification and quantification were obtained using ChenomX NMR Suite 8.0 Software (Edmonton, Canada) for all organs and serum. Lipidomics analysis of liver extracts was performed by the MetaToul-Lipidomique Core Facility (I2MC, Inserm 1297, Toulouse, France) from MetaToul (Toulouse metabolomics and fluxomics facilities, www.metatoul.fr), which is part of the French National Infrastructure for Metabolomics and Fluxomics MetaboHUB-ANR-11-INBS-0010.

NMR analysis of phosphocholine and phosphoethanolamine hydrolysis in sera

Six human serum samples [three females (41, 53, and 65 years old) and three males (50, 62, and 65 years old)] were obtained from the EFS (Etablissement français du sang, Auvergne Rhône-Alpes, France). NMR analyses were performed at the Lyon 1 University NMR Facility (CCRMN). Duplicate measurements were realized in 5 mm tubes, containing 66.6% (v/v) serum, 10% (v/v) D₂O in buffer (10 mM Tris-HCl, pH 7.4). Sera from three 3-month-old C57Bl/6 mice were used for NMR analysis. Analysis was realized in 3 mm NMR tubes, containing 16.6% (v/v) serum, 10% (v/v) of D₂O in Tris-HCl buffer at 10 mM at pH 7.4. After the substrates addition at a final concentration of 1 mM, samples were kept at 37 °C and NMR spectra were recorded every 90 min to follow their hydrolysis. SBI-425 TNAP inhibitor was added at a final concentration of 1 mM. Serum NMR spectra were obtained on a 500 MHz Ascend Avance III Bruker spectrometer, equipped with a 5 mm and 3 mm BBI probe. Cpmgpr1d pulse sequence was used, and spectra were acquired at 37 °C. Before the start of the experiment and the addition of substrate, samples were maintained at 4 °C. Spectrum processing and quantification using the ERETIC2 utility were performed as described before.

Hepatocyte cultures

HuH-6 human hepatocytes (Hölzel Diagnostica) were routinely cultured in complete media (Dulbecco’s Modified Eagle Medium, DMEM) containing 4.5 g/L glucose and supplemented with heat-inactivated 10% fetal bovine serum (FBS), 100 U/mL penicillin, 100 μg/mL streptomycin, 20 mM 4-(2-hydroxyethyl)-1-piperazine ethane sulfonic acid (HEPES) and 2 mM L-glutamine at 37 °C in a humidified atmosphere containing 5% CO₂. For P_i quantification, TNAP activity and expression, cells were seeded at a density of 10,000 cells per cm². Cells were collected 7 days after seeding. For DNA quantification, HuH-6 cells were seeded at a density of 8000 cells per cm². The day after, the medium was replaced by the treatment medium without choline (custom DMEM from Cliniscience) and without FBS, containing insulin/transferrin/selenium (ITS, Sigma-Aldrich) and recombinant human epidermal growth factor (rhEGF, Immunotools) at 0.5 ng/mL. After 48 h, the medium was replaced again by treatment medium without choline and FBS, containing ITS and EGF 0.5 ng/mL, with or without choline at 25 μM, with or without choline phosphocholine at 25 μM, and with or without SBI-425 at 25 μM. Medium was replaced twice a week, and cells were collected in PBS after 7 days and centrifuged. The pellet was resuspended in lysis buffer (10 mM Tris-HCl, pH 8.0, containing 0.1 mM EDTA, 0.5 M NaCl, and 0.05% Triton X-100), and cells were disrupted by sonication. Hoechst 33258 solution at 0.2 μg/mL in the lysis buffer without TX-100 was added to the mixture, and fluorescence light emission was measured at 460 nm (excitation 365 nm) with a microplate spectrophotometer reader (TECAN Infinite®200 PRO).

Determination of alkaline phosphatase activity

For the determination of AP activity based on the amount of p-nitrophenol (pNP) released after dephosphorylation of p-nitrophenyl phosphate (pNPP), organs were ground in a liquid nitrogen-cooled mortar and homogenized in 0.2% NP-40 detergent. Cells were collected in 0.2% NP-40 and disrupted by sonication. Organs and cell lysates were centrifuged at 3000 × g for 5 min at 4 °C, and the resulting supernatants were collected. The sample was mixed with reagent assay (0.56 M 2-Amino-2-methyl-1-propanol buffer, pH 10.5, 10 mM p-nitrophenyl phosphate (pNPP), 1 mM MgCl₂). Absorbance was acquired for 10 min at 405 nm by a TECAN Infinite®200 PRO. The slope was calculated to determine AP activity levels in each sample. For serum, AP enzymatic activity was expressed as nmol of p-nitrophenol formed/min/mL of serum. For cells and organs, the slope was normalized to protein amounts measured with the BCA protein assay (Pierce). The enzymatic specific activity was expressed as nmol of p-nitrophenol formed/min/mg.

Human recombinant TNAP expression and purification

Human TNAP was expressed using the Expi293 Expression System Kit (A14635, Thermo Fisher Scientific), according to the protocol described recently⁶⁶. The human ALPL gene was synthesized from Genscript, where the amino acid sequence 18–500 was cloned into a pcDNA3.4 plasmid, with a TEV-protease recognition sequence followed by a FLAG tag and an 8x polyhistidine tag on the C-terminal extremity, as well as an HA signaling peptide sequence on the N-terminal. Expi293F cells were cultured in Expi293 medium at 37 °C, in a humidified atmosphere with 8% CO₂, on a 20 mm diameter orbital shaker set to 125 rpm. When cell density reached 2.5 × 10⁶ cells/mL, the cells were transiently transfected with the pcDNA3.4 plasmid using Expifectamine293. These cells were then harvested after a 96-h incubation period. For the purification, transfected cells were pelleted using centrifugation, then the supernatant was collected and subjected to affinity chromatography using a HisTrap FF crude 5 mL Nickel-binding column (Cytiva). The elution was performed using a 0–300 mM imidazole gradient in HEPES buffer (20 mM HEPES, pH 7.5, 150 mM NaCl). The purity of eluates was evaluated by SDS-PAGE. After concentration of the pure eluates, the sample was further purified by size exclusion chromatography (SEC), using a Superdex 200 Increase 10/300 GL column (GE HealthCare Life Sciences) in equilibration buffer (20 mM HEPES, pH 7.5, 150 mM NaCl). The purity of the eluates was examined by SDS-PAGE, and polydispersity and oligomeric state were determined using Dynamic Light Scattering with a NanoSizer (Malvern).

Human TNAP crystallization, X-ray diffraction, and docking

A crystallization screening was performed on 96-well plates using the MOSQUITO Liquid Handler (TTP Labtech). After 51 days, nucleation was observed in a condition containing 0.2 M ammonium phosphate, 20% (w/v) PEG3350. Crystals were cryoprotected by soaking for 2–5 s in a crystallization solution supplemented with 15% (v/v) ethylene glycol before flash-freezing in liquid nitrogen. Diffraction data were collected from a single crystal at the ESRF (beamline ID30A-3) using a wavelength of 0.9677 Å. The crystal was maintained at 100 K under a nitrogen gas stream. The initial model was solved by molecular replacement using Phaser embedded in Phenix (version 1.21.2-5419) using the human TNAP structure (PDB ID 7YIV) as a starting model. Iterative refinement cycles were carried out using Phenix and Coot (version 0.9.6). Following refinement, the Ramachandran plot indicated that 95.71% of residues were in favored regions, 3.83% in allowed regions, with only 0.46% classified as outliers. The crystallographic parameters and data collection statistics are given in Table 1. Docking of PP_i, phosphoethanolamine, and phosphocholine into TNAP’s structure was performed using AutoDock Vina. Briefly, the protein structure was prepared for docking by removing unwanted ligand or water molecules and by adding polar hydrogen atoms using Discovery Studio Visualizer v19.1.0.18287 (BIOVIA, Dassault Systèmes, San Diego, 2018). The same program was used to build the substrate molecules as PDB files and for their subsequent energy minimization. PyRx 0.8 was then used for converting all molecules to AutoDock Ligand format (PDBQT)⁶⁷. The three-dimensional grid box for molecular docking simulation was obtained using Autodock tools embedded in PyRx. The Grid box was centered to cover the active site and all essential residues, while an exhaustiveness value of 500 was used, considering the relatively wide search area (18 × 54 × 25 ų). The docking experiments were then carried out with the program AutoDock Vina v1.1.2⁶⁸. In all calculations, the small-molecule compounds were treated as being flexible, whereas the target protein was treated as being rigid. The docking results were analyzed by comparing the binding interactions and binding energies between substrate molecules and the enzyme. A visual examination of the best poses was performed using PyMOL (Schrödinger, LLC - https://pymol.org).

Human TNAP characterization by cryo-electron microscopy (cryo-EM)

To determine how the MLS-0038949 inhibitor may inhibit the hydrolysis of phosphocholine or phosphoethanolamine, we performed a structural approach using single particle analysis by cryo-EM. A TNAP sample was purified from Expi cells by affinity chromatography and gel filtration. To determine the concentration of MLS-0038949 to use with TNAP, we first carried out in vitro enzyme inhibition assays and demonstrated that a concentration of 0.6 µM was sufficient to inhibit 100% of the enzyme (concentrated at 75 nM). To have an excess of inhibitor in our cryo-EM conditions, 20 µM of MLS-0038949 was first solubilized in the TNAP buffer and incubated with 2 µM of purified TNAP for 10 min on ice. Quantifoil R2/1 grids coated with 2 nm of continuous carbon were first glow-discharged for 1 min at −3 mA on an ELMO glow discharge system and then loaded on a TFS Vitrobot Mark IV cryo-plunger. The sample chamber was saturated at 95% humidity and maintained at 22 °C for grid preparation. Four microliters of the mixture TNAP-MLS-0038949 were then loaded onto the grid inside the cryo-plunger chamber. The blotting conditions used were 10 s of pre-blot, 5 s of blot at blot force 0, and 0 s of post-blot. After blotting, the grid was quickly plunged in liquid ethane maintained at −182 °C for sample vitrification. Around 5000 movies were collected on a TFS Krios G4i microscope operating at 300 kV at the Polaris beamline at SOLEIL synchrotron. A total dose of 50 e⁻/Ų was used for imaging conditions at 165,000× magnification. The dataset was completely processed with the suite cryoSPARC v4.5.1. After data curation, Patch Motion correction, and CTF estimation, the coordinates of the particles were found using the blob picker function (Supplementary Table 2). Around 860k particles were then extracted using a 360-pixel window size. After several rounds of 2D classification and selection of 2D class averages showing high-resolution details, ~466k particles were selected. A 3D ab initio reconstruction followed by a 3D non-uniform refinement with C2 symmetry allowed us to obtain a 2.3 Å resolution cryo-EM map (estimated with gold-standard FSC). The model of the TNAP could be easily fitted, and side chains could be adjusted in the density to get a better interpretation of the map. Extra density in the active site could be clearly fitted with the MLS model, as well as 3 sugars on each known post-translational modified site (N140, N230, N271, N303, N430).

Substrate hydrolysis kinetics in vitro and in cellulo

Human and mouse recombinant TNAP activity was measured as follows. Kinetics were obtained using the Malachite Green Phosphate Assay kit (Sigma-Aldrich). For enzymatic parameters determination in vitro, the reaction mixture was prepared in a reaction buffer (25 mM Tris-HCl, pH 9.0, 150 mM NaCl) with substrates at different concentrations and 125 nM of human recombinant TNAP or 24 nM of commercial mouse TNAP (Biolegend). For each substrate concentration, the reaction rate was calculated, and the kinetic parameters were determined using a generalized reduced gradient nonlinear solving method, and experimental data were fitted according to the Michaelis-Menten model: v = k_cat[E] [S]/(K_m + [S] + [S]²K_i). Prior to substrate hydrolysis in cellulo, cells were washed twice with a reaction buffer (25 mM Tris-HCl, pH 7.4 or 9.0, 150 mM NaCl, 1 mM MgCl₂, and 1 mM CaCl₂). Then, cells were incubated in the same buffer with 25 μM substrate at 37 °C. The results were expressed as μM of released P_i, and initial reaction rates were determined. TNAP inhibition was achieved with 25 μM of MLS-0038949 (Merck)⁶⁹.

Statistics and reproducibility

The number of experimental points and of experiments is indicated in the corresponding “Methods” section and/or in the figure legends. Data are expressed as the mean ± SEM or SD, as indicated in the Figure legends. Statistical analyses were performed with Past 3.20 software. Data were tested on normality and on equal variances with the Shapiro–Wilk test and Levene test, respectively. Then, data were analyzed on appropriate parametric or nonparametric tests.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.