Loss of function variants in PCYT1A causing spondylometaphyseal dysplasia with cone/rod dystrophy have broad consequences on lipid metabolism, chondrocyte differentiation, and lipid droplet formation

Abstract Spondylometaphyseal dysplasia with cone-rod dystrophy (SMD-CRD) is a rare autosomal recessive disorder of the skeleton and the retina caused by biallelic variants in PCYT1A, encoding the nuclear enzyme CTP:phosphocholine cytidylyltransferase α (CCTα), which catalyzes the rate-limiting step in phosphatidylcholine (PC) biosynthesis by the Kennedy pathway. As a first step in understanding the consequences of PCYT1A variants on SMD-CRD pathophysiology, we generated and characterized a series of cellular models for SMD-CRD, including CRISPR-edited PCYT1A-null HEK293 and ATDC5 cell lines. Immunoblot and PC synthesis assays of cultured skin fibroblasts from SMD-CRD patient cell lines revealed patient genotype-specific reductions in CCTα steady state levels (10-75% of wild-type) and choline incorporation into PC (22-54% of wild-type). While PCYT1A-null HEK293 cells exhibited fewer and larger lipid droplets in response to oleate loading than their wild-type counterparts, SMD-CRD patient fibroblasts (p.Ser323Argfs*38 homozygotes) failed to show significant differences in lipid droplet numbers or sizes as compared to controls. Lipid droplet phenotypes in PCYT1A-null HEK293 cells were rescued by transfection with wild-type, p.Ala99Val, and p.Tyr240His human PCYT1A cDNAs. While both edited cellular models had normal morphology and proliferation rates compared to unedited controls, Pcyt1a-null ATDC5 cells demonstrated accelerated rates of chondrocyte differentiation as compared to their wild-type counterparts. Lipidomics revealed changes in 75-200 lipid levels in PCYT1A-null HEK293 and ATDC5 cells or in SMD-CRD patient fibroblasts as compared to wild-type controls. The specific lipids altered and extent of change varied by cell type. Importantly, both PCYT1A-null HEK293 cells and SMD-CRD patient fibroblast cell lines had decreased phosphatidylcholine:phosphatidylethanolamine (PC:PE) ratios and decreased levels of several lysophosphatidylcholine (LPC) species as compared to wild-type controls, suggesting compensatory PC production through increased LPC remodeling by LPCAT or decreased conversion of PC to LPC by phospholipase A2. Our results show that all tested PCYT1A alleles associated with SMD-CRD are hypomorphic and suggest involvement of PCYT1A in chondrocyte differentiation, PC:PE ratio maintenance and LPC metabolism, and lipid droplet formation. Author Summary Rare genetic disorders can reveal the function of genes on an organismal scale. When normal gene activity is lost, patients can experience a range of symptoms, often dependent on the residual activity of the encoded protein. Rare variants in the gene PCYT1A can cause multiple inherited disorders, including a disorder of the skeleton and the retina characterized by short stature, bone abnormalities, and blindness. PCYT1A is required for normal cellular function, particularly lipid metabolism, but the role of this gene in human disease is still poorly understood. To determine consequences of genetic variants in patients with this disorder, we made and studied a series of cellular models, including cells cultured from patients and CRISPR-edited cell lines lacking normal copies of PCYT1A. Here we show that patient variants lead to reduced PCYT1A expression and/or function and have adverse consequences on cell biology and lipid metabolism that are often cell-type specific. This work advances understanding of the role of lipid metabolism in skeletal and eye development.

consequences: homozygosity for a Pcyt1a knockout allele results in embryonic lethality by day 115 E3.5 and failed implantation [9], and MT-58 CHO cells with a temperature-sensitive Pcyt1a point 116 mutation resulting in undetectable CCT protein at 40 o C undergo apoptosis at this temperature 117 [11,12]. Based on this evidence, some degree of PC production by the Kennedy pathway appears 118 to be essential for life, at least in these model systems. Fly models with eye-specific RNAi 119 knockdown or ocular mosaic homozygous knockout of CCT1 (the fly ortholog of PCYT1A) 120 showed reduced response to light on ERG, decreased rhodopsin expression, and failed rhabdomere 121 formation [13]. These phenotypes were rescued with wild-type and nuclear localization signal 122 mutant CCT1. In at least some cells, however, PCYT1A is expressed at levels exceeding that 123 required for basal function. For example, MT-58 CHO cells at the permissive temperature have 124 normal growth rates with only ~5% CCT expression compared to control CHO cells [12]. 125 Although most PC is produced by the CCTα-dependent Kennedy pathway in mammalian 126 cells, there are alternative pathways for PC production [6,14] (Fig 1). PCYT1B, a paralog of 127 PCYT1A, encodes the enzyme CCTβ, which catalyzes the same Kennedy pathway reaction as 128 CCTα but has expression limited predominantly to the brain and reproductive tissues of mice and  Phosphatidylethanolamine can also be interconverted to lysophosphatidylethanolamine by 151 phospholipase-A 2 (PLA 2 ) and LPE acyltransferase (LPEAT) enzymes.

153
CCTα functions as a homodimer comprised of 367-residue monomers with several 154 functional domains (Fig 1)  translocation of CCTα to nuclear membranes in wild-type fibroblasts, but did induce nuclear  Proliferation studies in HEK293 cells 265 Given the high concentration of PC in plasma membranes of cultured cells, we 266 hypothesized that CCTα deficiency would limit their proliferation. Interestingly, we found that 267 PCYT1A-null HEK293 cells proliferated at normal rates as compared to wild-type cells cultured 268 in standard medium supplemented with 10% FBS (Fig 5A). To determine if this normal growth 269 rate was dependent on uptake of exogenous lipid from the medium, we also tested cells cultured Pcyt1a-null ATDC5 cells proliferated at normal rates as compared to their wild-type counterparts 291 ( Fig 5B). Surprisingly, Pcyt1a-null ATDC5 cells differentiated into chondrocytes more quickly 292 than their wild-type counterparts ( Fig 6B). In total, the levels of ~75-200 lipids were significantly changed in the PCYT1A-null or SMD-CRD 313 cells, and the specific lipids altered were largely variable depending on the cell type tested (S2 314   Table, S3 Table, S4 Table). However, there were two significant changes in the lipidome shared 315 by both SMD-CRD cells and PCYT1A-null cells: reduced LPC content and reduced 316 phosphatidylcholine: phosphatidylethanolamine (PC:PE) ratios (Fig 7).

317
The levels of several LPC species (Fig 7A)  HEK293 cells and in SMD-CRD patient fibroblast cell lines as compared to controls (Fig 7A).

322
The reduction over all LPC classes attributed to loss of PCYT1A or the p.Ser323Argfs*38 mutation 323 was 3.3-fold for PCYT1A-null ATDC5 cells, 3.9-fold for PCYT1A-null HEK293 cells, and 1.7-324 fold for SMD-CRD fibroblasts. We also observed significant decreases in the 325 lysophosphatidylethanolamines (1.5-fold in null ATDC5 cells) and increases in 326 lysophosphatidylserines (2-fold in null HEK293 cells) (S2 Table, S4 Table).  Table), in line with a reduced rate of PC 346 synthesis. The PCYT1A-null HEK293 lipidome showed significantly increased 347 galactosylceramide and lactosylceramide (1.3-fold) (S4 Table).  The only lipid whose overall cellular content was decreased in two cell types was LPC, LPCAT or to decreased conversion of PC into LPC by PLA 2 (Fig 1). The former mechanism is 473 supported by recent work by Lee   Accuprime Taq polymerase (Invitrogen, Cat#12339-016) and standard thermal cycling conditions 531 using custom primer sequences (S1 Table). 12574-018, primer sequences in S1 Table), and cloned into the pcDNA3 mammalian expression  Table). All plasmids were sequence-verified prior 543 to use. 607 (S1 Table), modified into oligonucleotides for cloning purposes, and cloned according to the  Next, we treated cells with puromycin-containing medium at either 2 μg/mL (HEK293 cells) or 4 617 μg/mL (ATDC5 cells) 24 hours post-transfection. Medium was replaced daily for 2-3 days, until 618 untransfected control cells died. Single cells were then isolated by serial dilution in puromycin-619 free medium, expanded into isogenic clones, subjected to DNA extraction, and screened for both 620 wild-type and deletion-containing (S1 Table)  We quantified lipid droplet sizes and numbers as described previously [24]. Using ImageJ 662 software (v1.47), we converted images to 8-bit, adjusted image threshold, and used the "analyze 663 particle" command with a particle size distribution of 20 to infinity pixels^2 set to exclude on 664 edges.

672
Lipid standards for mass spectrometry 673 We obtained the following lipid standards from Avanti Polar Lipids (Alabaster, AL): 1,2- approximately 80% confluency, cells were scraped in ultrapure water, pelleted at 1000 x g, 695 resuspended in ultrapure water, and subjected to total protein concentration measurements using 696 the BCA protein assay kit (Cat#23225, Pierce). We stored cell extracts at -80°C until further 697 processing. 698 We obtained a total lipid extract from cell pellets using a modified Bligh-Dyer procedure 699 [50]. In brief, we diluted each cell suspension (200 µg protein) to 200 µl with ddH 2 O and sonicated 700 ten times using short bursts with a sonic homogenizer (Fisher Scientific, Waltham, MA) for 30 701 seconds followed by 30 seconds on ice. We transferred homogenates to glass tubes and gently 702 mixed with 800 µL of ddH 2 O, followed by 2.9 mL methanol/dichloromethane (2:0.9, v/v) 703 containing internal standards for 12 lipid classes to each sample. We then added 1 mL of ddH 2 O 704 and 0.9 mL dichloromethane to obtain a biphasic mixture, which was incubated at 4 o C for 30 min 705 and centrifuged at 4 o C for 10 min at 3000 x g to separate organic and aqueous phases. 1 mL of the 706 organic phase was transferred to a 2 mL glass vial and stored at -20 o C until use. We finally dried fragment peaks were present in 7 of the 8 pooled runs, with a coefficient of variation (CV) for peak 731 identifications less than 20%. Peak identifications meeting these criteria were then used to develop 732 a targeted method in LipidView. The targeted method was used to identify these pre-validated