Quantitative correlation of ENPP1 pathogenic variants with disease phenotype

Abstract

Ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1) codes for a type 2 transmembrane glycoprotein which hydrolyzes extracellular phosphoanhydrides into bio-active molecules that regulate, inter alia, ectopic mineralization, bone formation, vascular endothelial proliferation, and the innate immune response. The clinical phenotypes produced by ENPP1 deficiency are disparate, ranging from life-threatening arterial calcifications to cutaneous hypopigmentation. To investigate associations between disease phenotype and enzyme activity we quantified the enzyme velocities of 29 unique ENPP1 pathogenic variants in 41 patients enrolled in an NIH study along with 33 other variants reported in literature. We correlated the relative enzyme velocities with the presenting clinical diagnoses, performing the catalytic velocity measurements simultaneously in triplicate using a high-throughput assay to reduce experimental variation. We found that ENPP1 variants associated with autosomal dominant phenotypes reduced enzyme velocities by 50% or more, whereas variants associated with insulin resistance had non-significant effects on enzyme velocity. In Cole’s disease the catalytic velocities of ENPP1 variants associated with AD forms trended to lower values than those associated with autosomal recessive forms – 8–32% vs. 33% of WT, respectively. Additionally, ENPP1 variants leading to life-threatening vascular calcifications in GACI patients had widely variable enzyme activities, ranging from no significant differences compared to WT to the complete abolishment of enzyme velocity. Finally, disease severity in GACI did not correlate with the mean enzyme velocity of the variants present in affected compound heterozygotes but did correlate with the more severely damaging variant. In summary, correlation of ENPP1 enzyme velocity with disease phenotypes demonstrate that enzyme velocities below 50% of WT levels are likely to occur in the context of autosomal dominant disease (due to a monoallelic variant), and that disease severity in GACI infants correlates with the more severely damaging ENPP1 variant in compound heterozygotes, not the mean velocity of the pathogenic variants present.

Introduction

Ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1) is a type II transmembrane protein composed of four domains – two somatomedin B-like (SMB) domains, a catalytic domain, and a nuclease domain. The catalytic domain cleaves the high-energy phosphate bonds within nucleotide triphosphate compounds to generate extracellular pyrophosphate (PPi) and nucleotide monophosphates (NMPs) [1], or cyclic nucleotides such as 2’3’-cGAMP to generate 5’-AMP and 5’GMP [2]. The enzymatic products of ENPP1 catalysis generate extracellular signaling molecules that regulate whole organismal physiology – either directly, or as ligands to cell surface or intracellular receptors. For example, ENPP1 regulates soft tissue mineralization through generation of PPi which inhibits extra-skeletal calcification by incorporation into the hydroxyapatite crystal and elimination of crystal contacts necessary for crystal growth [3, 4]. ENPP1 also regulates vascular endothelial proliferation through activity of AMP on purinergic endothelial cell receptors [5], and ENPP1 regulates the innate immune response by modulating inflammatory cytokines and the type I interferon response through hydrolysis of 2’3’-cGAMP – a ligand for STING receptor [6, 7].

The enzyme’s diverse physiologic roles are perhaps reflected by the disparate disease phenotypes present in ENPP1-deficiency, the most dramatic of which are the life-threatening vascular calcifications in biallelic ENPP1-deficient infants afflicted with ‘Generalized Arterial Calcification of Infancy’ (GACI) [8–11]. In this severe disorder, over 50% of affected infants will succumb to cardiovascular collapse and death within the first year of life due to extensive calcifications and stenoses in their large arteries; those that survive will invariably develop a phosphate-wasting rickets driven by increased circulating FGF23, termed ‘Autosomal Recessive Hypophosphatemic Rickets Type 2’ (ARHR2) [12–14]. Moreover, children and young adults affected with ARHR2 often continue to calcify throughout development, manifesting as progressive calcifications of tendons/entheses and fusion of the cervical spine [15]. Sclerosis of the malleus and stapes also frequently occur, advancing progressive hearing loss [16]. The simultaneous co-existence of calcifying and phosphate-wasting phenotypes in the evolution of GACI into ARHR2 demonstrates the opposing site-specific mineralization forces unveiled by ENPP1 deficiency, in which life-threatening tissue calcifications are resisted by adaptive physiologic mechanisms wasting mineralized bone from the skeleton [17, 18].

ENPP1 haploinsufficiency, in contrast, induces early-onset osteoporosis characterized by multiple vertebral fractures and phosphate wasting [19, 20], or conversely ‘Diffuse Idiopathic Skeletal Hyperostosis’ (DISH) characterized by enthesopathies which progress into ossified spinal masses, progressive spinal fusion, immobility, and myelopathy in older adults [21]. Similarly, a polymorphism in ENPP1 has been linked to a rapidly progressive variant of a myelopathy, ‘Ossification of the Posterior Longitudinal Ligament’ (OPLL) [22], in which the posterior longitudinal ligament progressively calcifies. Interestingly, the severity and rate of progression of OPLL directly correlates with circulating FGF23 levels [23, 24] – the same hormone that induces phosphate wasting rickets in ARHR2 – suggesting that similar physiologic mechanisms may be involved in these ENPP1-deficient disorders.

Finally, variants in the SMB domains of ENPP1 can induce either Cole disease [25] or metabolic disorders, depending on their location. The ENPP1 SMB domains are believed to mediate interactions with other proteins on the cell plasma membrane, suggesting that some ENPP1-dependent phenotypes are catalysis-independent. Cole disease induces areas of cutaneous hypopigmentation (typically on the arms and legs) and thickened skin on the palms of the hands and feet (so-called palmoplantar keratoderma), similar to pseudoxanthoma elasticum (PXE) and ARHR2, cutaneous calcifications and tendon enthesopathies. Interestingly, Cole disease has both an autosomal dominant and autosomal recessive inheritance, depending on the specific ENPP1 variant [25, 26]. Finally, large scale GWAS studies have linked a specific polymorphism in the SMB2 domain of ENPP1 – identified as K121Q or K173Q depending on the numbering convention – to be strongly associated with childhood and adult obesity, insulin resistance, type-2 diabetes, diabetic kidney disease, major cardiovascular events in high-risk individuals, and stroke in newborn infants and children with sickle cell anemia [27–31].

To investigate relationships between disease phenotype and ENPP1 catalytic activity and to identify catalysis-independent phenotypes, we quantified the catalytic velocity of ENPP1 variants identified in patients with GACI, ARHR2, Cole disease, OPLL, DISH, early-onset osteoporosis, type 2 diabetes, and stroke associated with sickle cell anemia in 33 patients with ENPP1 pathogenic variants described in the literature and an additional 41 patients from 29 families enrolled in an NIH study, correlating the enzymatic velocity with disease phenotype and outcome. We found that enzyme velocities reduced below 50% of WT levels often induce disease phenotypes as autosomal dominant variants, that disease severity in GACI correlated with the more severely damaging ENPP1 variant in compound heterozygotes, that enzyme velocities associated with GACI ranged from no significant differences compared to WT to the complete abolishment of enzyme velocity, and that a pathogenic variant of ENPP1 long associated with childhood and adult obesity and metabolic syndrome is likely induced by alterations in catalysis-independent ENPP1 protein signaling.

Methods:

Human Subjects:

The human studies were conducted in accordance with the Declaration of Helsinki and were approved by the National Institutes of Health (NIH) Institutional Review Board (IRB). Subjects with pathogenic variants in ENPP1 were enrolled in research protocols 76-HG-0238 (“Diagnosis and treatment of patients with inborn errors of metabolism and other genetic disorders,” identifier: NCT00369421) and/or 18-HG-0064 (“Study of people with generalized arterial calcification of infancy (GACI) or autosomal recessive hypophosphatemic rickets type 2 [ARHR2]),” identifier: NCT03478839).

Identification of Pathogenic ENPP1 Variants:

A panel of ENPP1 pathogenic variants was assembled in combination with clinical data from a cohort of patients enrolled in an IRB-approved study at the NIH. This data set was additionally supplemented with variants reported in the existing literature as of June 2022 in which details of the pathogenic variants and patient clinical data was available. Table 1 summarizes the amino acid variant, domain location, associated disease phenotype, and public reference where available; this table includes patients with pathological ENPP1 variants who were diagnosed with GACI, ARHR2, Cole Disease, DISH, OPLL, early-onset osteoporosis, and type-2 diabetes. In addition, patients presenting with GACI as infants who then progressed to ARHR2 in childhood were also tracked in the clinical data set. Of note, due to heterogeneous inheritance patterns, a given ENPP1 pathogenic variant may be associated with multiple diseases.

Table 1: ENPP1 pathogenic variants and associated phenotypes.

The amino acid residue numbers, resulting protein variants, references, and phenotypes of the 62 ENPP1 pathogenic variants used in this study. ‘GACI -> ARHR2’ in the phenotype column indicates infantile GACI with evolution into ARHR2.‡ Denotes variants that are present in patients with autosomal dominant disease.

Variant Position	Protein variant	Domain	Reference	Associated Phenotype
91	L91P‡	SMB1	[43]	OPLL
92	G92D	SMB1	[13]	ARHR2
120	C120R	SMB1	[26]	Cole Disease, homozygous rec.
126	C126R	SMB1	[44]	GACI
133	C133R‡	SMB1	[45]	Cole Disease, autosomal dom.
149	C149S‡	SMB2	[25]	Cole Disease, autosomal dom.
162	S162G‡	SMB2	NIH Cohort	Hypophos., autosomal dom.
164	C164S‡	SMB2	[25]	Cole Disease, autosomal dom.
173	K173Q	SMB2	[27, 31]	Childhood obesity/diabetes/stroke, autosomal dom.
177	C177S‡	SMB2	[45]	Cole Disease, autosomal dom.
177	C177Y‡	SMB2	[25]	Cole Disease, autosomal dom.
179	N179S‡	SMB2	[21]	DISH, autosomal dom.
186	G186R	SMB2	[46]	GACI
190	W190C	SMB2	NIH Cohort	GACI
195	C195R	Phosphodiesterase	[47]	GACI
216	S216Y	Phosphodiesterase	[44]	GACI
218	D218V	Phosphodiesterase	[48]	GACI
242	G242E	Phosphodiesterase	[44]	GACI
250	P250L	Phosphodiesterase	NIH Cohort [49]	GACI -> ARHR2
252	Y252 Deletion	Phosphodiesterase	[49]	GACI
266	G266R	Phosphodiesterase	NIH Cohort [13]	GACI
268	Y268C	Phosphodiesterase	NIH Cohort [43]	ARHR2
276	D276N	Phosphodiesterase	[44]	GACI
287	S287F	Phosphodiesterase	[43]	OPLL
301	Y301C	Phosphodiesterase	[47]	GACI
305	P305T	Phosphodiesterase	NIH Cohort [49]	GACI -> ARHR2
342	G342V	Phosphodiesterase	NIH Cohort [49]	GACI
349	R349K	Phosphodiesterase	[44]	GACI
356	W356*	Phosphodiesterase	NIH Cohort	GACI
371	Y371F	Phosphodiesterase	[44]	GACI
451	Y451C‡	Phosphodiesterase	[21]	DISH, autosomal dom.
456	R456Q	Phosphodiesterase	[44]	GACI
Variant Position	Effect on protein	Domain	Reference	Associated Phenotype
471	Y471C‡	Phosphodiesterase	NIH Cohort, [19, 47]	GACI -> ARHR2
480	C480R	Phosphodiesterase	NIH Cohort [26]	GACI -> ARHR2 and ARHR2
481	R481W	Phosphodiesterase	NIH Cohort [44]	GACI, ARHR2, and GACI -> ARHR2
481	R481Q	Phosphodiesterase	NIH Cohort	GACI -> ARHR2
500	H500P	Phosphodiesterase	NIH Cohort [44]	GACI and GACI -> ARHR2
504	S504R	Phosphodiesterase	[44]
513	Y513C	Phosphodiesterase	NIH Cohort [44]	GACI
538	D538H	Phosphodiesterase	[47]	GACI
551	Y551C	Phosphodiesterase	NIH Cohort	GACI
570	Y570C	Phosphodiesterase	[44]	GACI
579	L579F	Phosphodiesterase	NIH Cohort [44]	GACI
586	G586R	Linker 1	[50]	PXE
611	L611V	Linker 1	NIH Cohort, [44, 51]	GACI
659	Y659C	Nuclease	NIH Cohort [44]	GACI -> ARHR2
668	E668K	Nuclease	NIH Cohort, [44, 51, 52]	GACI
774	R774C	Nuclease	NIH Cohort, [44, 52]	GACI
777	H777R‡	Nuclease	NIH Cohort, [19, 44]	GACI
792	N792S	Nuclease	[44, 51]	GACI
792	N792K	Nuclease	NIH Cohort
804	D804H	Nuclease	[44]	GACI
805	G805V	Nuclease	NIH Cohort	GACI -> ARHR2 and ARHR2
821	R821H	Nuclease	[44]	GACI
866	E866K	Nuclease	NIH Cohort	ARHR2
886	R886T	Nuclease	[53]	Diabetes
888	R888W	Nuclease	NIH Cohort, [44]	GACI
888	R888Q	Nuclease	NIH Cohort	GACI
901	Y901S	Nuclease	[12]	ARHR2
905	K905fsX15	Nuclease	NIH Cohort, [44]	GACI and GACI -> ARHR2
912	L912S	Nuclease	NIH Cohort	GACI
914	L914*	Nuclease	NIH Cohort	GACI

Design of plasmids encoding ENPP1 pathogenic variants:

To express the various ENPP1 pathogenic variants we started with the secreted form of human ENPP1-Fc (clone 770) described in our previous publications [32]. Using the New England Biolabs Q5 Site-Directed Mutagenesis kit, we eliminating the Fc fusion portion of the ENPP1-Fc by introducing a stop codon at the end of ENPP1 with the following oligos: 5’ gtgagttttgtcagatcaatcttcctggctgaagg 3’ and 5’ ccttcagccaggaagattgatctgacaaaactcac 3’. The resulting plasmid (clone 1881) served as the template for all subsequent Site-Directed Mutagenesis (SDM) reactions necessary to generate the list of pathogenic variants described in Table 1. All oligonucleotides were synthesized by the Keck Oligonucleotide Synthesis Facility. Endotoxin free plasmid DNA was isolated from transformed bacteria using the E.Z.N.A. Endo-free Plasmid DNA Mini Kit from Omega Bio-Tek as per manufacturer’s protocol. All plasmids were sequenced via the Sanger method by the Keck DNA Sequencing Facility to confirm that the desired alterations were correctly introduced into the wild-type human ENPP1 sequence (UniProt ref#P22413).

Recombinant protein expression:

The night before plasmid transfection, Chinese Hamster Ovary (CHO) cells were plated out at approximately 25,000 cells per well into 96-well tissue culture plates and grown at 37C with 5% CO2 in Gibco HAM/F12 media supplemented with 10% FBS and penicillin/streptomycin. Cells were transfected with the plasmids containing ENPP1 pathogenic variants at 100ng/well using a Mirus Bio Transit-CHO Transfection Kit (Madison, WI.) following manufacturer’s protocol. The WT ENPP1 clone 1881 and each pathogenic variant were transfected in triplicate in addition to mock transfected negative control wells.

High-throughput enzyme velocity assay:

After overnight incubation, conditioned media from each well of CHO cells transfected with each ENPP1 variant was transferred into a new 96-well plates. To determine enzyme velocity, 10 uL of conditioned media was diluted into 90 μL of buffer containing 1 M Tris pH 8.0, 50 mM NaCl, 20 μM CaCl₂, 20 μM ZnCl₂, and 1 mM thymidine 5′-monophosphate p-nitrophenyl (TMP-pNP) (Sigma-Aldrich, cat #T4510). The 96 well plates were immediately placed in a Synergy Mx microplate reader with Gen5 software and the velocity of the enzymatically generated chromogenic product, calculated from the linear range of the curve, was reported as the optical density change at 405 nm per unit time (mOD min⁻¹). The assay for each biological triplicate was performed twice to yield six independent measurements of each pathogenic variant. The average velocity across 5 wells of mock transfected CHO cells, generally less than 1% of the enzyme velocity observed in wells transfected with WT ENPP1, was used as background subtraction.

Experimental variability in the high throughput assay was measured with internal controls comparing the enzyme velocities in the conditioned media of two WT ENPP1 clones present on the plates during the assay. These internal controls yielded a 95% confidence interval of ± 28% for the enzyme velocity of these clones, meaning, we are 95% confident that any mutant with an enzyme velocity reduced by more than 56% compared to WT ENPP1 has impaired function. Possible sources of experimental variability include differences in transfection efficiency, DNA delivered, errors in plate reader, and pipetting errors.

Estimating effective organismal ENPP1 exposure from relative enzymatic activities:

To compare the organismal exposure of the ENPP1 pathogenic variants, we plotted the catalytic velocities of each pathogenic variants $\left(V_{var}\right)$ normalized to those of WT ENPP1 $\left(V_{WT}\right)$, or relative catalytic velocity $\left(V_{rel}\right)$, as defined by

$$V_{rel}=\frac{V_{var}}{V_{WT}}$$ #1

in all our figures (y-axis). The $V_{rel}$ of each pathogenic variant allows us to estimate the effective enzymatic output of the ENPP1 variants. The enzymatic velocity of wt ENPP1 and the variants is proportional to their total concentration $\left(C_{var}\right)$ and an enzyme concentration-independent activity $\left(S_{var}\right)$ according to:

$$V_{var}\sim C_{var}{S}_{var}$$ #2.

The ‘Area Under the Curve’ is an accepted measure of organismal exposure to a pharmacologic agent that scales directly with this velocity [32]. Every pathogenic variant will yield differing catalytic velocities due to their differing expression levels $\left(C_{var}\right)$ and concentration-independent activity $\left(S_{var}\right)$. Therefore, the relative activity of a given variant $\left(V_{rel}\right)$ is given by its relative expression level $\left(\frac{C_{var}}{C_{WT}}\right)$ and relative concentration-independent activity $\left(\frac{S_{var}}{S_{WT}}\right)$ according to:

$$V_{rel}=\frac{V_{var}}{V_{WT}}=\frac{C_{var}}{C_{WT}}\frac{S_{var}}{S_{WT}}$$ #3

In silico prediction of the pathogenic variant effects on protein stability:

To predict the change in protein stability (ΔΔG) induced by single amino acid changes in the ENPP1 pathogenic variants we used the program DUET [33] and the cryo-EM structure of human ENPP1 complexed with a biologic inhibitor (PDB 8GHR, [34]). The resolution of this ENPP1 structure did not allow predictions of stability changes induced by SNPs prior to residue 185.

Quantitation of plasma PPi levels:

Blood samples were collected in 4 ml Na-heparin tubes, and plasma was harvested by centrifugation at 2000g at 4 °C for 15 min. Platelet-free plasma was prepared by filtration through a Centrisart I 300 kDa mass cutoff filter. The samples were stored at −80 °C before processing. The PPi assay is semiquantitative and was a modification of methods previously described [35, 36]. In summary, the assay utilized ATP sulfurylase to convert filtered plasma PPi to ATP, which was detected utilizing luciferase/luciferin luminescence to produce light from newly formed ATP. Sensitivity of the assay was determined to be 0.15mM PPi. Inter- and -intra-assay CVs on identical samples were below 20%.

Statistical Methods:

GraphPad Prism 10 was used to statistically analyze all data. Statistical significance between three or more groups was determined using an ANOVA comparison of means using the non-parametric Kruskal-Wallis independent test at a significance level of α= 0.05. Linear regression was performed with a simple linear regression model using Pearson r correlations and P values to determine significance.

Methodology for correlating disease severity with phenotype:

Phenotypic data were associated with each pathogenic variant to investigate a correlation between disease severity and enzyme velocity. The phenotypic data were gathered from published literature reviews in pathogenic variants whose clinical course in affected patients was characterized, and from de-identified patient data available from the NIH cohort. A scoring system was devised to quantitate disease severity in each patient by assigning numerical scores associated with adverse clinical findings, which included vital status, hospitalization, cardiac failure, hypertension, stroke, and other clinically relevant morbidities (Table 2). In this scoring system, the higher the score the patient received, the more severe the disease.

Table 2: Scoring system quantitating disease severity in GACI.

Disease severity corresponds to greater overall score, with maximum score of 23 and minimum score −7.

		Score
Vital status	Survival −1	Unknown 0	Deceased +4
Onset	>90 days −1	Unknown 0	90 days or less + 1
Hospitalization	No −1	Unknown 0	Yes +1
Mechanical ventilation	No −2	Unknown 0	Yes +2
Cardiac failure	No −1	Unknown 0	Yes +1
Hypertension	No −1	Unknown 0	Yes +1
Ventricular hypertrophy	No −1	Unknown 0	Yes +1
Arterial calcification		No/unknown 0	Yes +1
Myocardial infarction		No/unknown 0	Yes +2
Pulmonary edema		No/unknown 0	Yes +2
Pulmonary hypertension		No/unknown 0	Yes +3
Stroke		No/unknown 0	Yes +2
Hearing loss		No/unknown 0	Yes +1
Rickets		No/unknown 0	Yes +1

Data availability:

The data underlying this article will be shared on reasonable request to the corresponding author.

Results:

The location of the 62 ENPP1 pathogenic variants are schematized onto the ENPP1 domain structure in Figure 1, and the enzymatic velocities are displayed in Figure 2. Figure 3 displays the enzymatic velocities of all ENPP1 variants resulting in GACI. Notably, these pathogenic variants occur throughout the primary sequence with effects ranging from the complete abolition of enzymatic velocity to velocities that are without experimental significance when compared to wild type levels. The median velocity of ENPP1 variants below the dashed line in Figure 3 are statistically different than WT ENPP1, and those above the dashed line are not (as defined by a p<0.05, ANOVA Brown-Forsythe and Welch test). Importantly, the statistically significant variants in ENPP1 all reduce the enzymatic velocity of ENPP1 by approximately 50% or greater (Y-axis, Figure 3). Note the 11 ENPP1 pathogenic variants inducing GACI which exhibit enzymatic velocities that are not below WT.

Figure 1: Schematic of the location of ENPP1 pathogenic variants on enzymatic domains.

Figure 2: Enzymatic velocity and disease phenotypes of ENPP1 variants displayed by primary sequence.

The Enzyme velocities for 62 ENPP1 pathogenic variants associated with disease phenotypes were determined simultaneously in triplicate in two separate experiments. The data are displayed as box-plots with whiskers representing the 25–75 quartiles and min-max values, respectively. The individual measurements are displayed as symbols colored by the disease phenotype defined in the legend. Protein domains are boxed at the top of the figure, with boundaries between domains denoted by vertical lines. ENPP1 pathogenic variants in GACI, ARHR2, and PXE exhibit an autosomal recessive (AR) inheritance pattern, while those in Cole disease exhibit an autosomal dominant (AD) inheritance pattern except for the C120R pathogenic variant, which is autosomal recessive. The ENPP1 pathogenic variants in early onset osteoporosis, hypophosphatemia, DISH, OPLL, and childhood obesity/stroke all induce disease in an autosomal dominant manner. While it is recognized that ENPP1 pathogenic variants may induce either GACI or PXE phenotypes, the ENPP1 pathogenic variant G586R has only been described in a PXE patient as a (biallelic, monoallelic, compound heterozygous). Finally, while most, if not all, surviving GACI patients will go on to develop ARHR2 [42], the two ARHR2 pathogenic variants denoted above were only described in ARHR2.

Figure 3: Enzymatic velocities of ENPP1 variants in GACI.

The data are displayed as in Figure 1, for only GACI patients. Measurements below the dotted line are statistically significant relative to WT (p<0.05), and measurement above the dotted line are not (p>0.05, ANOVA multiple comparison with Brown-Forsythe and Welch tests).

Figure 4A shows the enzyme velocities of ENPP1 proteins occurring as heterozygous variants in patients with clinical features, associated with ENPP1-haploinsufficiency. The enzyme velocities of variants associated with Cole’s disease, early-onset osteoporosis, OPLL, and DISH, are all less than half of WT levels. The observation suggests that heterozygous variants resulting in ENPP1 catalytic velocities if they are sufficiently damaging (e.g., with less than 50% of WT levels) are likely to induce a disease phenotype. Figure 3B–C shows variants that induce Cole disease, which occur in the SMB domains, and pathogenic variants that induce disorers other than GACI and Cole disease. Note the three variants associated with spinal hyperostosis (in OPLL and DISH) and early-onset osteoporosis in middle-aged adults as autosomal dominant pathogenic variants – S287F, Y451C, and Y471C [19, 21].

Figure 4: A. Enzyme velocities of ENPP1 variants associated with autosomal dominant disease. B. Enzymatic velocities of ENPP1 variants in Cole Disease. C. Enzymatic velocities of ENPP1 pathologic variants in disorders other than GACI and Cole disease.

Data are displayed as in figure 1. Statistical significance determined by ANOVA Brown-Forsythe and Welch tests, with significance explicitly stated when 0.05>p>0.0001; ****p<0.0001. SSA: sickle cell anemia.

To further characterize the pathogenic variants in GACI patients which were not significantly reduced below WT levels, we plotted the relative catalytic velocities of the corresponding variants in compound heterozygous patients, which in every case we identified possessed a relative catalytic velocity below 50% with the exception of G186R (Figure 5A), which occurred in three siblings with biallelic G186R variants. This observation suggests that the phenotypes of GACI patients with compound heterozygous pathogenic mutations are dominated by the more severely affected allele.

Figure 5: A. Enzymatic velocity of the paired ENPP1 pathogenic variant in compound heterozygous GACI patients where the corresponding allele does not significantly compromise enzymatic velocity.

There were 7 ENPP1 pathogenic variants observed in GACI which did not significantly reduce enzyme velocity below WT levels, 5 in the catalytic domain and 2 in the SMB2 domain. These variants were S162G, G186R, G342V, R349K, R481W, R481Q, and Y570C. The enzymatic velocities of the corresponding alleles in compound heterozygous patients were identified except for patients with R349K and Y570C. In every case the corresponding allele reduced the enzymatic velocity below 50% of WT except patients with G186R, which occurred in siblings with biallelic, homozygous G186R variants. B. Linear regression of the relative catalytic velocity of the most compromised ENPP1 variant with disease severity in compound heterozygous GACI patients. A linear regression analysis found that the GACI disease severity score correlated inversely with the relative catalytic velocity of the most severe ENPP1 variant, demonstrating that disease severity in GACI is dependent on degree to which the catalytic velocity is compromised in the most severely affected allele. F=4.9, DFn=1, DFd=136, R²=0.35, and p=0.029 (simple linear regression analysis). Individual values of disease severity are represented by filled circles, with whiskers representing the standard deviation in each measurement. In most cases the standard deviation does not exceed the symbol width, and so the whiskers are not apparent.

Finally, to investigate a potential relationship of GACI disease outcome to ENPP1 activity, we examined the medical records of 41 homozygous/compound heterozygous ENPP1-deficient patients from 29 families enrolled in an NIH study. We determined a phenotypic composite outcome score for each patient based on clinical parameters described in the methods and listed in Table 2. To investigate whether clinical outcomes in patients with biallelic disease were related to enzyme velocity, we correlated phenotypic scores with the mean of the enzyme velocities of the two variants observed in a compound heterozygote, as well as with the velocity of the more damaging variant. We found that phenotypic severity did not correlate with predicted mean enzyme velocity but did correlate with the velocity of the more severely compromised ENPP1 allele in the compound heterozygotes (R²=0.035, p=0.0288, Figure 5B), supporting the notion that prognosis in GACI compound heterozygotes is governed by the more severely compromising pathogenic variant.

Discussion

Pathogenic variants of the ENPP1 are associated with mineralization disorders (GACI, ARHR2, PXE, DISH, OPLL, phosphate wasting, early onset osteoporosis), metabolic disorders (childhood and adult obesity, type 2 diabetes), cutaneous pigmentation disorders (Cole Disease), and coagulopathies (stroke in children with sickle cell anemia). Certain phenotypes are associated with either monoallelic or biallelic ENPP1 pathogenic variants. To understand whether disease phenotype and severity correlate with enzyme activity, we performed a high-throughput assay to quantify the enzyme velocities of 62 pathogenic variants in a single experiment using a secreted form of the enzyme. To quantify the enzyme velocities of all mutants in triplicate in a single experiment we employed the ENPP colorimetric substrate pNP-TMP, a phosphorylated pyridine whose hydrolysis we and others have established to be proportional to the hydrolysis of native ENPP1 phosphoanhydride substrates such as ATP and cGAMP [37, 38]. The direct measurement of ATP or cGAMP by ENPP1 would require an HPLC assay taking approximately one week to complete, which was impracticable for the high-throughput design of our study. In addition, our assay also reduced the experimental variability by using a soluble form of ENPP1 which eliminated the need to normalize enzyme concentrations with cell membrane protein, as discussed below.

Possible limitations of our study included an assay design which did not directly measure enzyme concentration for each pathogenic variant, and the use of a soluble ENPP1 protein that did not allow us to evaluate pathogenic variants in transmembrane and cytoplasmic domains. Balancing these limitation is the fact that the relative enzyme velocity (Vvar) is an aggregate term accounting for both the enzyme concentration and specific activity, as defined by Eqn. 2. As we have shown that AUC scales directly with Vvar [32], the assessment of Vrel therefore reflects both the differing enzyme concentrations and the organismal exposure of each pathogenic variant, satisfing the goal of our study. Comparisons of our study with previous studies which normalized enzyme concentrations to cell membrane protein yielded good agreement for all pathogenic variants with one exception, further validating our approach (Table 3). The discrepant findings for this one exception (L611V) are due to our evaluation of L611V as a single variant, whereas past studies evaluated L611V in cis with E668K. Supporting the benign nature of the L611V as a single variant is the obseration that this is a common variant in the African American population at 4.9% [39].

Table 3: Comparison of the relative catalytic velocities of ENPP1 variants with prior studies.

The relative catalytic velocities determined herein (Ansh activity) are listed compared with prior studies. There is good concordance with prior measurements with two exceptions: L611V and R481W. See text for discussion.

Pathogenic Variant	Ansh activity	Prior activity 1	Ref. activiy 1	Prior activity 2	Ref. activity 2
C126R	0%	3%	[52]
K173Q	63%	66%	[47]	83%	[52]
C195R	0%	0%	[47]
Y301C	49%	0%	[47]
P305T	4%	0%	[47]
R456Q	2%	11%	[52]
Y471C	33%	43%	[52]
R481W	68%	37%	[52]
S504R	8%	51%	[47]
L579F	0%	0%	[52]
L611V	111%	0%	[52]
Y659C	11%	41%	[47]
E668K	74%	100%	[47]	70%	[52]
R774C	26%	57%	[47]	36%	[52]
H777R	9%	38%	[47]
N792S	26%	17%	[52]
R821H	107%	88%	[47]
R888W	4%	0%	[47]
K905fsX15	0%	0%	[52]

We found that ENPP1 pathogenic variants reducing the relative ENPP1 velocity by 50% or greater were capable of inducing disease phenotypes in an autosomal dominant manner, while those above 50% did not; except for S162G, which was associated with phosphate wasting in a haploinsufficient NIH patient. This notion was supported by a GACI phenotype severity scoring system (Table 2) which correlated disease severity with relative catalytic velocity, revealing that the disease severity correlated with the more compromised ENPP1 allele rather than the mean enzymatic velocity of both alleles in compound heterozygotes, demonstrating that allele co-operativity is a driver of disease phenotype. Other enzyme deficiencies in which allele co-operativity has been identified include biotin deficiency. For example, a common biotinidase variant present in ≈ 4% of the population reduces enzymatic activity by 50%. This variant is asymptomatic in homozygous carriers, but induces biotin deficiency in compound heterozygotes harboring a more severe biotinidase variant.

In summary, our combined findings support: 1) the understanding of ENPP1 as a ‘gene of large effect’ capable of inducing disease phenotypes in an autosomal dominant manner in severe pathogenic variants; 2) the metabolic phenotype induced by ENPP1 deficiency is likely mediated through a catalysis-independent mechanism; 3) ENPP1 pathogenic variants in hypophosphatemic and cutaneous pigmentation disorders commonly occur as autosomal dominant pathogenic variants.

Similarly, seven variants observed in GACI patients with life-threatening vascular calcifications did not reduce enzyme velocities below WT levels, suggesting that the more compromising pathogenic variant may dominate the disease phenotype or other factors than ENPP1 catalytic activity may play an important role in determining plasma [PPi] in individual patients.

To this point, while our studies provide evidence that a catalytic activity threshold exists below which one should expect an ENPP1 disease phenotype, our studies do not explain the variability in disease severity in patients possessing identical pathogenic variants. It is well established that identical ENPP1 pathogenic variants may result in either GACI or ARHR2 in siblings, so clearly ENPP1 catalytic activity cannot entirely account for disease severity, and other factors upstream and downstream of ENPP1 are also at play. These factors include CD73, alkaline phosphatase, and ABCC6, which are all known to modulate the in vivo effects of ENPP1 activity. We also acknowledge that disease-modifying factors not yet identified may play an important, even decisive, role in disease phenotype. To determine how well Vrel predicted plasma [PPi] in GACI patients, we obtained plasma from GACI patients in the NIH cohort, and measured their plasma [PPi] using assay conditions prior to the now-established CLIA PPi assay. As seen in Table 4, some patients show poor correspondence between Vrel and plasma [PPi], demonstrating that factors other than ENPP1 enzymatic activity are likely to play a significant role in extracellular PPi concentrations. CLIA approved assays to reliably quantitate plasma PPi in patients have only recently been established, and the future use of these assays help to clarify the role of upstream and downstream factors in individual patients.

Table 4: Protein variants, calculated Vrel, and Plasma PPi levels in 7 compound heterozygous GACI patients from the NIH cohort.

The variability of plasma [PPi] in patients with identical pathogenic variants (such as patients GACI04 and GACI05), and the non-correspondance of plasma PPi and Vrel (such as in patient GACI17), suggests that upstream and downstream factors such as ABCC6, Alkaline Phosphatase, and CD73 may be important for plasma [PPi] in individual patients.

Patient	Protein Variant (Vrel)	Age at collection	PPi concentration (μM) (normal 1.37–6.93)
GACI69	c.715+1G>C (not determined)	39y	0.53
	p.Asn792Lys (0.063±0.011)
GACI05	p.Cys480Arg (0.345±0.077)	10y	0.63,
	p.Gly805Val) (0.053±0.014)	12y	0.91
GACI04	p.Cys480Arg (0.345±0.077)	7y	0.70,
	p.Gly805Val (0.053±0.014)	9y	0.27
GACI17	p.Tyr551Cys (0.002±0.006)	5y	1.13
	p.Leu579Phe (0.001±0.002)
GACI22	p.Lys905fsX15 (0.001±0.001)	7y	0.93
	p.Arg481Trp (0.681±0.060)
GACI20	p.Tyr471Cys (0.333±0.048)	29y	0.63
	p.Arg481Gln (1.082±0.092)
GACI09	p.Arg888Trp (0.040±0.006)	12y	0.17
	p.Arg774Cys (0.255±0.018)

Finally, our current understanding of ENPP1 function incorporates the presence of both catalysis-dependent and independent mechanisms. This notion is supported by the current study, which demonstrates that some pathogenic variants do not impact catalytic activity, specifically those in the SMB2 domain which are strongly linked to metabolic syndrome and coagulopathy. Until recently the mineralization defects were also presumed to result from catalysis-dependent effects – specifically, decreased plasma PPi. Observations of early-onset osteoporosis in ENPP1-haploinsufficient patients, however, cast doubt on this assumption; PPi inhibits the formation and growth of hydroxyapatite, which constitutes the mineral phase of bone and vascular calcifications. Reduced plasma PPi should therefore increase bone mass, and not cause osteopenia in ENPP1-deficiency [19, 40], and thus, the mechanism by which ENPP1 deficiency induces osteopenia is not apparent from the enzyme’s catalytic activity alone. Recent in vivo and in vitro mechanistic studies have identified catalysis-independent ENPP1 pathways suppressing Wnt through increased Sfrp1 transcription [41], pathways which are activated in ENPP1 deficiency and would provide a rationale understanding for the osteoporosis observed in this setting. Additional investigations into catalysis-independent ENPP1 signaling are likely to provide additional mechanistic insights into the long-recognized effects of ENPP1 pathogenic variants on childhood obesity and metabolic syndrome in adults [27]. Finally, our findings suggest that parents of GACI children possessing severe ENPP1 pathogenic variants will likely exhibit an ENPP1 disease phenotype such as hypophosphatemia, spinal enthesopathy, or early onset osteoporosis. While recognizing the absence in the literature of these phenotypes in GACI parents, the absence of evidence is not evidence of absence, and we will await the results of future studies in this population to confirm our findings.

Highlights:

• ENPP1 pathogenic variants induce disease in AR and AD inheritance patterns.

• Disease phenotypes vary from lethal vascular calcifications to childhood obesity.

• We correlated catalytic velocity with disease phenotype in 62 pathogenic variants.

• ENPP1 enzyme velocities <50% induced disease phenotypes in an AD manner.

• Compound GACI patients possessed pathogenic variants without reduced activity.

• Disease severity in GACI correlated with the more severely damaging ENPP1 allele.

• Variants associated with metabolic syndrome exhibit WT catalytic activities.