Cellular and organismal function of choline metabolism

Abstract

Choline is an essential micronutrient critical for cellular and organismal homeostasis. As a core component of phospholipids and sphingolipids, it is indispensable for membrane architecture and function. Additionally, choline is a precursor for acetylcholine, a key neurotransmitter, and betaine, a methyl donor important for epigenetic regulation. Consistent with its pleiotropic role in cellular physiology, choline metabolism contributes to numerous developmental and physiological processes in the brain, liver, kidney, lung and immune system, and both choline deficiency and excess are implicated in human disease. Mutations in the genes encoding choline metabolism proteins lead to inborn errors of metabolism, which manifest in diverse clinical pathologies. While the identities of many enzymes involved in choline metabolism were identified decades ago, only recently has the field begun to understand the diverse mechanisms by which choline availability is regulated and fuelled via metabolite transport/recycling and nutrient acquisition. This review provides a comprehensive overview of choline metabolism, emphasizing emerging concepts and their implications for human health and disease.

Choline, a five-carbon quaternary ammonium cation [(CH₃)₃NCH₂-CH₂OH]⁺, was first described in 1862 by Adolf Strecker in his work exploring the chemical composition of living matter¹. Isolated from heated ox and pig bile, Strecker named this nitrogenous compound choline after the Greek word for bile—chole^2,3. Two decades prior, Theodore Gobley, had isolated a related molecule from brain tissue and carp fish eggs that he named ‘lecithin’ after the Greek word for egg yolk—lekithos^2,4. We now know that phosphatidylcholine (PC) is the molecular identity of lecithin. Gobley himself came to understand that choline must be a constituent of lecithin derived from its decomposition^5,6. However, it was not until the pioneering work of Eugene Kennedy in the 1950s that it was understood exactly how mammalian cells utilize free choline to synthesize PC de novo⁷. In the pathway that now bears his name, Kennedy described the biosynthesis of PC from choline, finding that it proceeded through a high-energy cytidine-5′-diphosphate choline (CDP–choline) intermediate⁷.

While mammals have some capacity to synthesize choline de novo, most choline is derived from the diet. In the 1930s, Charles Best found that feeding pancreatectomized dogs with lecithin or choline alone could ameliorate the fatty degeneration of their livers, indicating that dietary choline can influence physiology and disease^8–10. Despite the clear connections between dietary choline availability and animal physiology, choline was not considered an essential nutrient for humans until the late 1970s². Work through the 1980s and 1990s further demonstrated that humans fed choline-deficient diets developed fatty liver disease and muscle and liver damage^11–18. Strikingly, these pathologies were ameliorated when choline was reintroduced to the diet^11,12,14. This body of work led to the recognition of choline as an essential nutrient by the Food and Nutrition Board of the Institute of Medicine in 1998 (ref. 19). Governmental bodies across the globe have followed suit²⁰, with choline adequate intake recommendations currently set at 425 mg per day for women and 550 mg per day for men¹⁹.

Despite the discovery of choline over 160 years ago, our understanding of this essential nutrient and its metabolism is incomplete. Numerous molecular players involved in choline metabolism remain to be discovered, and we still understand little about how choline metabolism is regulated to ensure homeostasis of cellular membranes. Recently, systems biology approaches have identified mechanisms by which free choline and choline-containing metabolites are transported into and throughout mammalian cells; however, organismal choline metabolism, differential demands across tissues and acquisition routes remain poorly understood. Consistent with its essential roles, variants in enzymes involved in choline metabolism result in monogenic diseases, and choline dyshomeostasis is implicated in complex diseases such as cancer, metabolic syndrome and ageing. In this review, we summarize the current state of the field, highlighting the seminal and current literature underpinning our understanding of choline metabolism while also emphasizing the many outstanding questions that remain to be addressed.

Biochemical pathways contributing to choline homeostasis

Within the cell, choline has multiple metabolic fates including acetylcholine, PC and betaine (Fig. 1). In this section, we review the pathways and enzymes that act upon choline to yield these diverse metabolic fates. Additionally, we highlight mechanisms of choline transport and alternative nutrient acquisition strategies used by cells to ensure sufficient choline availability.

Fig. 1 |. Chemistry of choline and its derived metabolites.

Choline is a five-carbon metabolite with a quaternary ammonium cation. The five carbons of choline (highlighted in yellow) are found in nearly all choline-derived metabolites. Through import into the mitochondria and the action of choline dehydrogenase (CHDH) and betaine-aldehyde dehydrogenase (ALDH7A1), betaine is generated from choline. Choline is converted into the neurotransmitter acetylcholine by ChAT. Excess dietary choline is converted to TMA by the gut microbiome. TMA is then oxidized to TMAO by flavin monooxygenases in the liver. In the de novo synthesis of PC through the Kennedy pathway, CHKA/CHKB converts choline to phosphocholine, which is then catalysed to CDP–choline by PCTY1A/PCTY1B. Finally, PC is generated from CDP–choline through the action of CEPT1 or CPT1. The phosphocholine headgroup of PC is transferred to ceramide by SGMS1/SGMS2 to generate SM. PC species can be catabolized by the phospholipases to generate LPC and further deacylated to generate GPC.

Biochemistry of the Kennedy pathway

The most abundant phospholipids in mammalian cells are PC and phosphatidylethanolamine (PE), which represent ~50% and 25% of the total cellular phospholipid pool, respectively²¹. PC is a cylindrically shaped phospholipid that produces flat lipid bilayers and is ubiquitously distributed throughout all cellular membranes²². In mammalian cells, the vast majority of PC and PE pools are generated de novo via parallel biosynthetic pathways (CDP–choline and CDP–ethanolamine) named after their high-energy intermediates and collectively known as the Kennedy pathway. The Kennedy pathway was elucidated by Eugene Kennedy in 1956 (ref. 7) through his seminal work, which revealed that incubation of rat liver homogenate with radioactive P³²-phosphocholine could yield radioactive PC⁷. The complete conversion of radioactive phosphocholine to radioactive PC suggested that PC was the predominant metabolic fate of phosphocholine. In a serendipitous series of experiments, Kennedy, alongside his postdoc, Samuel Weiss, discovered that phosphocholine incorporation into PC proceeded through the high-energy, nucleotide-activated intermediate CDP–choline. Kennedy and Weiss initially believed that the conversion of phosphocholine to PC was ATP dependent, as commercially available liquid ATP stimulated the reaction⁷. However, to their surprise, crystalline ATP failed to similarly stimulate the enzymatic production of radioactive PC, leading Kennedy and Weiss to suspect that the liquid ATP used in their initial experiments contained impurities^7,23. Hypothesizing that the impurity might be another (non-ATP) nucleotide, Kennedy and Weiss tested ITP, UTP, GDP and cytidine-triphosphate (CTP) for their ability to stimulate PC synthesis from phosphocholine, ultimately finding that CTP alone was required for the enzymatic conversion of phosphocholine to PC⁷. Inspired by the biochemical paradigms of nucleotide activation, Kennedy and Weiss hypothesized and eventually demonstrated that CTP served as a coenzyme to activate phosphocholine and phosphoethanolamine for incorporation into PC and PE, respectively, by forming a high-energy, CDP-conjugated intermediate⁷. These discoveries established the broad de novo mechanism by which mammalian cells synthesize PC and PE.

The CDP–choline and CDP–ethanolamine branches of the Kennedy pathway both consist of three enzymatic steps catalysed by multiple enzymatic isoforms, encoded by different genes, at almost every step (Fig. 2). The first step of the CDP–choline branch of the Kennedy pathway is the phosphorylation of choline to phosphocholine. This enzymatic step consumes one molecule of ATP and is catalysed by choline kinase-α (CHKα) and choline kinase-β (CHKβ), encoded by CHKA and CHKB, respectively. Both enzymes are localized in the cytosol and reported to have high micromolar affinity for choline (the K_m values are reported to be around 200 μM and 570 μM for CHKα and CHKβ, respectively²⁴), allowing for the rapid conversion of imported choline to phosphocholine to trap it in the cell²⁵. Consequently, cellular concentrations of phosphocholine are orders of magnitude higher than that of free choline²⁶. Notably, CHKα is the predominant isoform broadly expressed across cell types²⁷ and its loss is incompatible with life, as Chka-knockout mice are embryonically lethal between embryonic day 3.5 (E3.5) and E7.5 (ref. 28).

Fig. 2 |. Choline metabolism in mammalian cells.

Overview of the predominant metabolic pathways responsible for choline consumption and recycling. Extracellular choline is transported intracellularly by high-affinity transporters FLVCR1 and FLVCR2. Other reported choline transporters (SLC44A1/SLC22A2 and SLC22A1/SLC22A2) may also contribute. Via phosphorylation to phosphocholine, intracellular choline is committed to de novo phospholipid synthesis through the Kennedy pathway (enzymes highlighted in yellow). Endogenously synthesized or extracellularly acquired PC can be recycled via the endolysosomal system and phospholipases (transporters and enzymes highlighted in orange) to generate LPC, GPC or free choline. Extracellular LPC can be transported into the cell through MFSD2A to further fuel choline recycling. Intracellular choline can be imported by SLC25A48 into the mitochondria for further metabolism to betaine. De novo synthesis of choline in the liver of mammals occurs via the PEMT pathway, which produces PC from PE. The phosphocholine headgroup of PC is donated to ceramide to generate SM via SMGS1/SMGS2. In cholinergic neurons, choline is transported by SLC5A7 for use in generation of the neurotransmitter acetylcholine. SAH, s-adenosyl homocysteine; PPi, pyrophosphate; CMP, cytidine monophosphate; PA, phosphatidic acid.

The second and rate-limiting step of the CDP–choline pathway generates CDP–choline from phosphocholine and CTP with the release of pyrophosphate. This step is catalysed by CTP:phosphocholine cytidylyltransferases CCTα or CCTβ, encoded by PCYT1A and PCYT1B, respectively. CCTα/PCYT1A is ubiquitously expressed and primarily localized at the nuclear membrane²⁷. The enzymatic activity of CCTα/PCYT1A is stimulated by association with the nuclear membrane, which is favoured when low cellular PC levels cause increased nuclear membrane curvature and mediated via two amphipathic helices²⁹. CCTα/PCYT1A is also localized at the surface of lipid droplets and is required for PC production to circumscribe expanding lipid droplets³⁰. Loss of PCYT1A is incompatible with life, as Pcyt1a-knockout mice display early embryonic lethality with inability to form blastocysts³¹. The second CTP:phoshphocholine cytidylyltransferase enzyme, CCTβ/PCYT1B, shows tissue-specific expression in the brain, retina and testis²⁷ and is localized subcellularly to the endoplasmic reticulum (ER)³².

The third and final step of the CDP–choline pathway generates PC via the transfer of phosphocholine from CDP–choline to a diacylglycerol (DAG) lipid backbone and is catalysed by cholinephospho-transferases. The enzymes capable of catalysing this step are CPT1 and CEPT1, encoded by CHPT1 and CEPT1, respectively^33,34. Both enzymes are ubiquitously expressed²⁷, membrane-bound and localized to the ER and/or golgi apparatus. Recent cryo-electron microscopy structural studies reveal that CEPT1 functions as a homodimer³⁵, and it is suggested that CPT1 functions by a similar mechanism³⁶.

The CDP–ethanolamine branch of the Kennedy pathway mirrors the CDP–choline branch and consists of the same three enzymatic steps, but ethanolamine substitutes choline as the headgroup. The enzymes involved in the CDP–ethanolamine branch of the Kennedy pathway are largely distinct from those in the CDP–choline branch²¹.

Choline transport and uptake

The uptake of choline into cells is the first critical step in its utilization for lipid biosynthesis. At physiological pH, choline carries a net positive charge and cannot readily diffuse across biological membranes. Thus, it must rely on dedicated transporters for transmembrane movement. The concentration of choline in human serum and extracellular fluid is ~10 μM and 3–6 μM, respectively^37–40. High-affinity choline uptake in the brain was appreciated in the 1950s through the study of acetylcholine synthesis and synaptic release^41–43, and subsequently confirmed using radioactive choline tracers^44,45.

The identity of the first high-affinity choline transporter involved in cholinergic signalling was determined in 2000 when Okuda and colleagues identified cho-1 in C. elegans and Cht1 in rats⁴⁶. Shortly thereafter, the mouse and human homologues of CHT1 (also known as SLC5A7) were identified^47,48. CHT1/SLC5A7 has high affinity for choline (K_m 0.5–3 μM)⁴⁹ and the structural basis of choline transport by CHT1/SLC5A7 was recently reported⁵⁰. CHT1/SLC5A7 expression is highly restricted to cholinergic neurons (such as those comprising neuromuscular junctions), suggesting its function is specific to acetylcholine synthesis and activity. Indeed, CHT1/SLC5A7 is localized to synaptic vesicles in cholinergic terminals and, to a lesser extent, presynaptic plasma membranes^51–53. Despite its predominant localization at cholinergic synaptic vesicles, CHT1/SLC5A7 is not functional at these vesicles, as they lack a Na⁺ gradient. Instead, CHT1/SLC5A7 is only active at the presynaptic plasma membrane, which occurs when cholinergic synaptic vesicles fuse to release their acetylcholine cargo. In the synaptic cleft, acetylcholine binds nicotinic and muscarinic cholinergic receptors to induce cholinergic signalling until its activity is quenched by acetylcholinesterase, which hydrolyses acetylcholine to free choline and acetate. Finally, free choline is transported back into presynaptic neurons through CHT1/SLC5A7, where it is recycled to re-form acetylcholine for further neurotransmission. This elegant mechanism couples CHT1/SLC5A7 localization and activity to synaptic acetylcholine release, thereby creating a functional metabolic circuit to facilitate neurotransmission. This critical role of CHT1/SLC5A7 in cholinergic neurotransmission at the neuromuscular junction is reflected in Cht1 (also known as Slc5a7)-deficient mice, which are developmentally normal and born at expected Mendelian ratios, but, strikingly, die within 1 h of birth from difficulty breathing, cyanosis and paralysis⁵⁴.

The first-reported ubiquitous choline transporter is CTL1, also known as SLC44A1 (refs. 55–57). Like that of CHT1/SLC5A7, choline transport by CTL1/SLC44A1 is Na⁺dependent^56,58,59. In humans, complete loss of function of CTL1/SLC44A1 is compatible with live birth, but affected individuals develop early-onset neurodegeneration⁶⁰. CTL1/SLC44A1 has also been implicated in the inflammatory response of macrophages and microglia to Toll-like receptor stimulation by lipopolysaccharide^61,62. Two closely related transporters, CTL2/SLC44A2 and CTL3/SLC44A3, have also been reported to transport choline^49,63–65. Proposed to localize to both the plasma membrane and mitochondria, CTL1/SLC44A1 ad CTL2/SLC44A2 may function to transport choline into the cytosol as well as mitochondrial lumen^66,67. Both transporters are also reported to be carriers for ethanolamine, a metabolite with high structural similarity to choline⁶⁷. Finally, another pair of transporters, OCT1/SLC22A1 and OCT2/SLC22A2, were found to display choline transport activity^68,69. The tissue specificity and precise functional role(s) for these intermediate-affinity and low-affinity choline transporters (K_m 50–500 μM)⁴⁹ are not well known. While it was initially suggested that the collective action of these transporters is sufficient to meet the choline needs of diverse cell types, choline concentrations in the serum and extracellular space in vivo are orders of magnitudes lower than the K_m values of these transporters^37–40, ultimately raising the possibility that a ubiquitously expressed high-affinity choline transporter might exist to ensure adequate choline uptake in all cells throughout the body. It remains possible, however, that these low-affinity and intermediate-affinity choline transporters fuel choline metabolism in specific cell types or under specific physiological conditions.

We recently discovered the identity of a ubiquitously expressed high-affinity choline transporter by leveraging metabolite genome-wide association studies (GWAS) datasets⁷⁰. Our search identified an association between serum levels of the metabolites phosphocholine and PC and FLVCR1 (ref. 70), a plasma membrane transporter previously characterized as a haem exporter^71–73. Using similar computational approaches, the genetic association between FLVCR1 and choline metabolism was made in different patient populations by other groups^74,75. FLVCR1 loss in diverse mammalian cell types resulted in drastic reductions in choline-derived metabolites such as phosphocholine and betaine, and radioactive choline uptake assays directly demonstrated a requirement for FLVCR1 in choline transport⁷⁰. At the same time, the Hamachi laboratory used an innovative click-chemistry-based FACS CRISPR screening strategy and discovered FLVCR1 as a choline transporter⁷⁶.

Surprisingly, cells lacking FLVCR1 remain viable in cell culture, although they become dependent on enzymes involved in phospholipid synthesis (CHKA, PCYT2, PCYT1 and CDIPT) or phospholipid salvage—specifically SPNS1 (ref. 70), a recently discovered lysosomal lysophosphatidylcholine (LPC) exporter discussed in the next section^77–79. Using metabolomics and radioactive uptake assays, we confirmed that FLVCR2, a paralogue of FLVCR1, can also transport choline⁷⁰. While cells in culture can proliferate without FLVCR1, loss of Flvcr1 in mice is embryonically lethal at day E12.5, with Flvcr1-deficient embryos showing severe depletion of choline and downstream metabolites phosphocholine and betaine⁷⁰. Additionally, embryos lacking Flvcr1 display mitochondrial ultrastructural defects, dysfunction and activation of the mitochondrial integrated stress response. Strikingly, perinatal choline supplementation of mice through pregnancy with daily supraphysiological boluses of choline partially rescued the embryonic lethality of Flvcr1-knockout mice, extending developmental lifespan by multiple days. Of note, missense variants in FLVCR1 have been described in rare subsets of patients who display posterior column ataxia and retinitis pigmentosa and/or hereditary sensory autonomic neuropathy^80–84. These clinical genetics associations suggest that impaired choline transport may underlie the pathology of these disorders.

Subsequent to these initial studies^70,76,85, work from the Nguyen laboratory reported similar findings and expanded on these reports^86,87. Mice with constitutive or endothelial cell-specific loss of Flvcr2 are embryonically lethal and phenocopy Fowler’s syndrome, a human condition caused by missense variants in FLVCR2 (ref. 88). Using these mice, the Nguyen group demonstrated that loss of Flvcr2 impacts the levels of choline and choline-derived metabolites in fetal brains⁸⁷. Similarly, endothelial cell-specific loss of Flvcr2 in adult mice reduces choline transport into the brain⁸⁹. Together, these studies suggest that FLVCR2 acts a choline transporter at the blood–brain barrier, thereby implicating choline metabolism in the pathophysiology of Fowler’s syndrome.

In addition to these studies describing the discovery of FLVCR1 and FLVCR2 as the major mammalian choline transporters (K_m ~ 5 μM), three recent cryo-electron microscopy studies provided structural insight into the mechanism of choline transport by FLVCR1 and FLVCR2 (refs. 89–91; Fig. 2). Collectively, these studies demonstrate that FLVCR1 and FLVCR2 adopt the classical topology of a major facilitator superfamily (MFS) transport protein with 12 transmembrane helices arranged into two pseudo-symmetric six-helix bundles connected by a cytoplasmic loop^89–91. Structures of FLVCR1 and FLVCR2 bound to choline revealed that a central coordinating tryptophan residue (FLVCR1 Trp125, FLVCR2 Trp102) is responsible for stabilizing the choline molecule and required for choline transport by FLVCR1 and FLVCR2 (refs. 89–91). Proteoliposome transport assays further demonstrated that FLVCR1 and FLVCR2 are capable of choline transport in cell-free systems^89,90. Two of these studies also reported structures of FLVCR1 and FLVCR2 bound to ethanolamine^90,91. This is consistent with data from the Hamachi laboratory, who showed that increasing doses of ethanolamine could outcompete choline transport by FLVCR1 (ref. 76). Interestingly, the structural basis of ethanolamine transport by FLVCR1 seems to differ from that of choline, as mutation of a single residue (Gln214) abolishes ethanolamine but not choline transport^90,91. Mutation of the analogous residue on FLVCR2 (Gln191), however, abolishes both choline and ethanolamine transport⁹¹. Future work will focus on identifying the endogenous substrates of FLVCR1 and FLVCR2 in vivo and determining their roles in homeostasis and disease.

Alternative strategies for choline acquisition by cells

Mammalian cells display remarkable metabolic plasticity to ensure choline pools are maintained. Critically, the choline head groups of PC, sphingomyelin (SM) and LPC can be recycled from these lipids by the action of various lipases, enabling choline-containing lipids to serve as alternative sources of choline when the availability of the free nutrient is limited.

LPC is an abundant choline-carrying lipid in the serum and extracellular fluid. One mechanism by which extracellular LPC is taken up by cells is via the action of MFSD2A, a cell-surface LPC transporter primarily expressed in the central nervous system (CNS)⁹². Mfsd2a-knockout mice have significantly smaller neonatal brain size and weight, and metabolic profiling of the brains revealed LPC was depleted up to 90%. Mfsd2a-knockout mice also display motor dysfunction after weaning and severe anxiety with deficits in learning and memory, signs reminiscent of omega-3 fatty acid deficiency⁹². MFSD2A was shown to play a similarly critical role in photoreceptor cell development in the eye⁹³. Subsequent work has characterized the structural basis of LPC transport by MFSD2A, which proceeds via a rocker-switch mechanism^94–98. In humans, variants in MFSD2A have been linked to congenital microcephaly and neurodevelopmental syndromes^99–102, and MFSD2A function has been implicated in the pathogenesis of microcephaly caused by congenital Zika virus infection¹⁰³.

In addition to LPC transport, cells can also obtain choline-containing lipids for choline recycling by harnessing the degradative capacity of the endolysosomal system (Fig. 2). Choline-containing glycerophospholipids and sphingolipids bound to serum albumin or lipoprotein particles are taken up by cells and degraded by phospholipases within the lysosome to generate choline-containing catabolites such as LPC, glycerophosphocholine (GPC) and phosphocholine. Work by the Shayman laboratory^104–107 has shown that PLAG2G15 is the major lysosomal phospholipase catalysing the sequential hydrolysis (or transfer) of acyl chains from PC to generate LPC (produced after a single diacylation event) and GPC (produced after two diacylation events)^108–111. Alternatively, PC can be catabolized to LPC and GPC in the ER by the phospholipase PNPLA6/NTE^112–115. LPC and GPC catabolites generated in the lysosomal lumen or the ER lumen can then be recycled via export into the cytosol for incorporation into the glycerophospholipid pool or use in the Land’s cycle, an enzymatic cycle that remodels the fatty acid tails of PC.

The identity of a dedicated lysosomal LPC exporter was unknown until recently. Independent work from both the Silver and Abu-Remaileh groups identified SPNS1, an orphan lysosomal transporter, as capable of transporting LPC and other lysophospholipids with zwitterionic head groups such as lysophosphatidylethanolamine (LPE)^77,78. Whole-cell and lysosomal metabolic profiling of cells with SPNS1-deficient lysosomes revealed an immense accumulation of LPC and LPE compared to control lysosomes⁷⁷. Further, metabolic tracing studies through the endolysosomal pathway demonstrated that SPNS1-exported LPC species are eventually incorporated into PC pools⁷⁷, representing a rapid choline–lipid recycling pathway that supports cell survival under free choline deprivation⁷⁷. Of note, PLA2G15, which deacylates PC to LPC and LPC to GPC, is 1,200 times more active against PC than LPC¹¹⁶, providing a kinetic basis for LPCs’ excess in lysosomes that would enable their transport through SPNS1. Finally, the Ngyuen laboratory also described the role of SPNS1 as a lysosomal LPC exporter, shedding additional light on the physiological function of this pathway during development⁷⁹. Spns1-knockout embryos were significantly smaller than controls and displayed cortical thinning of the brain⁷⁹. Similarly, Spns1 knockdown in livers displayed accumulation of lysosomal LPC and LPE species and increased lysosomal mass concomitant with clinical signs of liver damage⁷⁸. Another choline-containing molecule that is supplied through the lysosome is GPC, which is generated by complete diacylation of PC. The Abu-Remaileh laboratory discovered that the Batten disease gene product CLN3, variants in which cause early-onset neurodegeneration, is required for the lysosomal efflux of glycerophosphodiesters including GPC, whose choline is released in the cytoplasm to be incorporated into choline-containing lipids¹¹⁷. Whether CLN3, a lysosomal transmembrane protein, is directly involved in the transport of GPC or only a factor in the process is yet to be determined.

In addition to these pathways, PC can be generated from PE through phosphatidylethanolamine N-methyltransferase (PEMT), a pathway that is mainly active only in the liver. First described in 1987 (ref. 118), this reaction is catalysed by the ER-resident enzyme PEMT, which uses S-adenosylmethionine (SAM) as the methyl-donating substrate. The role of this pathway is further discussed below.

The molecular fates of choline

Once imported into the cell, the majority of choline is phosphorylated to produce phosphocholine and fed into the de novo synthesis of PC, the most abundant phospholipid in cells, through the CDP–choline branch of the Kennedy pathway. Phosphocholine from PC is also used in the generation of SM through the action of sphingomyelin synthase 1 and 2 (SGMS1 and SGMS2), which catalyses the transfer of phosphocholine from PC to ceramide to generate SM and DAG^119,120. Choline committed to lipid synthesis can be recovered through the catabolism of lipids within the cytosol and lysosome, as previously described. Type A and B phospholipases sequentially cleave the acyl chains from PC to generate 1-acyl-glycerophosphocholine and then GPC. GPC can subsequently be hydrolysed to generate free choline by GPC phosphodiesterases such as GPCPD1 and GDPD5 (refs. 121,122). Free choline can be liberated from PC via the action of type D phospholipases, PLD1 and PLD2 (refs. 123,124). In addition to choline, PC catabolism also generates lipid second messengers such as PA, DAG, LPC and arachidonic acid, with diverse autocrine and paracrine signalling functions¹²⁵. The precise molecular mechanisms by which choline-containing lipids are catabolized across cell types is not well understood; furthermore, whether and how these processes are regulated in physiological homeostasis and disease is nearly unexplored. Finally, the proportional pool sizes of these choline-containing metabolites seem to differ across cell types, although the function remains unclear. For example, GPC is highly abundant in the brain relative to other organs¹²⁶.

A specialized fate of choline in the nervous system is acetylcholine, a critical neurotransmitter controlling many outputs of the nervous system ranging from motor control to memory. In the neuronal cell body, choline acetyltransferase (ChAT), encoded by the CHAT gene, catalyses the addition of an acetyl group from acetyl-CoA to choline to generate acetylcholine. ChAT is thought to exist in both soluble and synaptic vesicle membrane-bound forms, with the soluble version producing the majority of acetylcholine^127,128. The vesicular acetylcholine transporter then loads acetylcholine into synaptic vesicles, which are transported to the nerve terminal. After neuronal stimulation, synaptic vesicles fuse with the presynaptic membrane to release their acetylcholine cargo into the synaptic cleft to engage nicotinic and muscarinic cholinergic receptors for neurotransmission. Acetylcholine is ultimately hydrolysed to acetate and choline by acetylcholinesterase, thereby enabling choline re-uptake into the presynaptic neurons by CHT1/SLC5A7 to be recycled into acetylcholine.

While some molecular fates of choline (that is, acetylcholine, SM and PC) are reversible through deacetylation or lipid catabolism, others are irreversible. One such fate is the oxidation of choline to betaine: once choline is committed to one-carbon (1C) metabolism, it cannot be recovered and recycled for other purposes. In the choline oxidation pathway, cytosolic free choline is transported into the mitochondria, where it is converted to betaine in two enzymatic steps carried out by choline dehydrogenase and betaine-aldehyde dehydrogenase^129,130. Betaine generated de novo in the mitochondria is exported to the cytosol where it can function as an osmolyte. The function of betaine as an osmolyte is physiologically important in the kidney, as decreased betaine levels are seen in patients with chronic kidney disease¹³¹. Additionally, betaine plays critical roles as a methyl donor in 1C metabolism, epigenetic regulation (methylation reactions), nucleotide biosynthesis (purine and thymidine), amino acid homeostasis (glycine, serine and methionine) and redox homeostasis¹³².

Finally, it is worth discussing the fate of unabsorbed dietary choline, betaine and PC that pass through the small intestine and reach the colon. Intestinal bacteria in the microbiome of the colon convert choline and choline-derived metabolites to the metabolite trimethylamine (TMA). TMA enters the circulation via unknown mechanisms and is oxidized to trimethylamine-N-oxide (TMAO) by the host via hepatic flavin monooxygenases^133,134. TMAO has no known metabolic role in humans, although elevated serum TMAO levels are associated with numerous pathologies including cardiovascular disease, obesity and cancer^{133,135–137}.

Subcellular compartmentalization of choline

Organellar distribution of choline metabolism

Most free choline is phosphorylated to phosphocholine in the cytosol. Phosphocholine continues through the CDP–choline arm of the Kennedy pathway to generate CDP–choline and finally PC. The mechanisms by which PC is distributed throughout the cell are incompletely understood. It has been proposed that transport could be mediated through inter-organelle membrane contact sites, vesicular transport through the Golgi network and through the action of specific lipid transfer proteins^138,139. STARD7, a member of the steroidogenic acute regulatory-related lipid transfer domain-containing family, is required for the transport of PC generated by the Kennedy pathway to the mitochondria^76,140. Organellar-membrane specific trafficking of PC generated through the CDP–choline pathway remains to be discovered for most organelles.

Free choline within the cytosol that is not phosphorylated is primarily transported into the mitochondria for commitment to betaine synthesis and eventually 1C metabolism, highlighting the need for a mitochondrial choline transporter. Previous work suggested that CTL1/SLC44A1 and CTL2/SLC44A2 localized to the mitochondria to enable choline transport into the organelle for betaine synthesis^66,67. Recent work from multiple groups, however, has suggested that an alternate solute carrier (SLC) family transporter may be responsible for choline transport into the mitochondria. Using a metabolomics GWAS approach, the Birsoy and Gamazon laboratories identified the orphan mitochondrial transporter SLC25A48 as necessary for choline transport into the mitochondria¹⁴¹. Similar results were also reported by the Köttgen laboratories^142,143. Further, the Kajimura laboratory reported SLC25A48 as a mitochondrial choline transporter that regulates the biosynthesis of purine nucleotides, whose production depends on 1C metabolism¹⁴⁴. Mice with whole-body knockout of Slc25a48 were viable but cold intolerant and displayed defects in brown adipose tissue mitochondrial respiration¹⁴⁴. Taken together, these studies point towards SLC25A48 being the predominant high-affinity mitochondrial choline transporter (Fig. 2).

The metabolism field is only beginning to understand how choline and its derived metabolites are transported across cellular compartments. Innovative approaches to understand this complexity such as organellar immunoprecipitation, CRISPR–Cas9-based genetic screens and metabolic GWAS analyses will hopefully yield new insights into this understudied aspect of choline metabolism.

Organismal choline homeostasis

Dietary requirement for choline consumption

Choline stores within the body are continuously depleted, making it critical that they are continuously replenished to maintain organismal choline homeostasis. While the liver is capable of supplying small quantities of choline via the endogenous PEMT pathway^145,146, the quantity of choline synthesized de novo within the liver is not sufficient to meet physiological demands; thus, choline is an essential nutrient that must be obtained through diet^147,148. Dietary choline is consumed in the form of choline-containing lipids (PC and SM, constituting 50% of total dietary choline) and soluble metabolites (free choline, phosphocholine and GPC, together making up the remaining 50% of total dietary choline)^149–151. Many foods are rich in choline-containing compounds, including meat, poultry, fish, dairy products, eggs, beans and some cruciferous vegetables^{147–150,152}.

Dietary choline is absorbed in both lipid and soluble forms in the small intestine^152,153. Choline-containing lipids are directly packaged into chylomicrons for transport to the liver via the lymphatic vessels¹⁵². A small portion is also degraded by intestinal lipases and hydrolases to release free choline¹⁵². Soluble choline-containing metabolites such as phosphocholine and GPC cannot be directly absorbed and must be broken down to free choline¹⁵². It is still unknown how free choline is absorbed in the small intestine, although it is possible that previously reported choline transporters such as FLVCR1 and CTL1/SLC44A1 (ref. 154) may play a role. Following absorption, free choline enters the portal circulation and traverses the liver, where a portion is taken up and utilized by hepatocytes¹⁵². The remainder passes on to the systemic circulation for uptake by other tissues.

In the United States, it is believed that only 10% of adults meet the National Institutes of Health (NIH) suggested choline intake¹⁴⁸. As such, there is significant interest in better understanding population health consequences of choline under-consumption and, correspondingly, the potential population health benefits of choline supplementation. Despite the existence of generalized recommendations, individual requirements for dietary choline intake may vary dramatically depending on various physiological, genetic and nutritional factors. Pregnancy is a critical modifier of dietary choline requirement, as choline is an essential nutrient for fetal development^147,148. Impressively, the concentration of choline is ten times higher in amniotic fluid compared to maternal serum, a gradient achieved by active transport of choline across the placenta, likely mediated by choline transporters^155,156. Based on studies in guinea pigs, supplying the fetus with adequate choline depletes maternal choline supplies in the liver (but not serum) and can lead to the development of fatty liver in the mother and offspring if dietary choline intake and replenishment is insufficient^157,158. To ensure adequate availability of choline to both the mother and developing child, pregnant and lactating women are recommended to increase their daily choline intake by up to 30%^148,151.

A second factor modifying dietary choline requirement is inherited deficiency in one or more choline-metabolism genes^{147,148,150,151}. The most striking effect is seen with respect to inherited defects in PEMT¹³. In humans, variants in PEMT increase the risk of multi-organ dysfunction from dietary choline insufficiency by 25-fold^159,160. Strikingly, in mice lacking Pemt, a choline-deficient diet leads to end-stage liver failure and death within only 3 days¹⁶¹. Because individuals with PEMT polymorphisms have reduced endogenous synthesis of PC, they depend more heavily on dietary choline sources to meet their physiological requirement. A second gene associated with risk for dietary choline insufficiency is MTHFD1, which catalyses the dehydrogenation of 5,10-methylene tetrahydrofolate to 10-formyl-tetrahydrofolate, thus regulating the availability of folate-derived methyl groups^159,161. Individuals with MTHFD1 variants have increased dietary choline requirements, as choline-derived methyl groups must compensate for the loss of folate-derived methyl groups. Indeed, one study found that premenopausal women who were carriers of a MTHFD1 polymorphism were 15 times more likely to develop organ dysfunction related to choline insufficiency and more likely to have pregnancies affected by neural tube defects¹⁶². While other associations between genes and dietary choline requirement have been suggested, few have been studied in depth, underscoring a need for further investigation in this thread.

Lastly, nutritional factors, particularly dietary availability of the partner methyl donor to choline, folate, can modify the requirement for dietary choline^{148,160,163,164}. Studies in humans demonstrate that choline deficiency, as defined by declines in serum concentrations of choline and PC, occur from moderate dietary choline insufficiency only with low, but not high, folate intake¹⁶³. A similar finding was observed in rats, in which moderate folate deficiency alone was sufficient to deplete liver choline stores¹⁶⁴.

Tissue-specific roles of choline

While choline is a universally essential nutrient in all tissues for its role in membrane phospholipid biosynthesis and methylation reactions, it is also remarkably pleiotropic and plays several tissue-specific roles in the nervous system, liver, lung, kidney and immune system (Fig. 3).

Fig. 3 |. Organismal choline metabolism.

Overview of tissue-specific roles of choline and derived metabolites in mammals. Human diseases associated with perturbations of choline metabolism listed within organ systems.

Nervous system.

Within the nervous system, choline is required for the synthesis of the excitatory neurotransmitter acetylcholine utilized for neurotransmission by cholinergic neurons^165,166. Acetylcholine is synthesized and stored within the axon terminals of cholinergic neurons and released following cellular depolarization. Acetylcholine engages either of two families of cholinergic receptors expressed on postsynaptic neurons: nicotinic cholinergic receptors or muscarinic cholinergic receptors. Nicotinic acetylcholine receptors are choline (and nicotine)-gated ion channels that, following ligand binding, permit the passage of sodium, potassium or calcium ions that locally depolarize the postsynaptic cell and result in an action potential or calcium signalling cascade^165–167. Muscarinic acetylcholine receptors, on the other hand, are choline (and muscarine)-sensitive G-protein-coupled receptors that trigger secondary messenger cascades in the postsynaptic neuron.

Cholinergic neurons are abundant and located throughout regions of both the peripheral nervous system (PNS) and CNS. Within the PNS, cholinergic neurons innervate skeletal muscles to mediate voluntary muscle contractions and project from autonomic ganglia to various target tissues to activate the involuntary autonomic functions of these tissues¹⁶⁶. Within the CNS, cholinergic projection neurons and interneurons are found throughout many regions of the brain and are thought to primarily regulate memory, motivation, arousal and attention¹⁶⁶. Cholinergic transmission in the CNS has also recently been implicated in hippocampal plasticity^166,168,169. Cholinergic dysfunction and its association with Alzheimer’s disease (AD) and Parkinson’s disease (PD) will be discussed further in the next section.

Liver.

The liver stands out as a crucial hub for choline metabolism within the body as the sole tissue with the capability to synthesize choline de novo and the first tissue that gains access to dietary choline absorbed in the small intestine^147,152. Accordingly, choline plays several critical tissue-specific roles in the liver. First, the liver utilizes choline to synthesize PC for the biogenesis of very-low-density lipoproteins (VLDLs), which mediate redistribution of dietary fats and cholesterol to tissues throughout the body^3,170,171. Dietary choline insufficiency in humans and other mammals leads to inadequate VLDL biogenesis and impaired clearance of fat from the liver, culminating in the development of non-alcoholic fatty liver disease (NAFLD)^{145–148,150,152,172}. A second liver-specific function of choline is the production of bile, a secretion composed of bile salts, cholesterol and PC (up to 40% by mass)¹⁷³. Bile is synthesized by hepatocytes in the liver, stored in the gall bladder and released into the small intestine to aid in the solubilization and absorption of dietary fats¹⁷⁴. Of note, choline (in the form of PC) secreted into bile by ABCB4 is one main mechanism by which choline is eliminated from the body^{70–72,173–175}. While Pemt-knockout mice (which lack endogenous choline synthesis) die from a choline-deficient diet in only 3 days, Abcb4-Pemt double-knockout mice on a choline-deficient diet survive to 90 days, presumably because loss of Abcb4 helps to sustain endogenous choline levels^161,176. Finally, the liver is a site for detoxification through chemical conjugation including methylation. Thus, choline is consumed to methylate, detoxify and excrete a plethora of waste compounds in the body¹⁷⁷.

Lung.

The specific function of choline in the lung is in the production of pulmonary surfactant by type II pneumocytes^178,179. Surfactant is a mixture of phospholipids and proteins that functions to lower surface tension within the pulmonary alveoli so that they may easily expand after inhalation. The composition of surfactant is approximately 10% surfactant proteins (SPA-D) and 90% phospholipids, of which almost 70% consists of dipalmitoylphosphatidylcholine [PC(16:0_16:0)]^179,180. Surfactant is tightly packed and stored within specialized organelles known as lamellar bodies, which are exocytosed by type II pneumocytes into the alveoli¹⁷⁸. Continuous synthesis and production of surfactant by type II pneumocytes necessitates a high demand of choline in these cells.

Kidney.

Choline plays a unique role in the kidney as a precursor for an important osmolyte in medullary cells: betaine^181–183. Export of salts by renal tubular cells into the medullary interstitium generates an osmotic gradient that provides the driving force for water reabsorption¹⁸⁴. However, the high salt content of the medullary interstitium increases osmotic pressure on renal medullary cells, which must correspondingly raise their own intracellular osmotic pressure to prevent desiccation. One strategy by which renal medulla cells achieve high intracellular osmolarity is increased uptake of osmolytes such as betaine from the serum/blood, which they achieve through upregulation of the betaine/gamma-aminobutyric acid transporter 1 (BGT1)^181–184.

Immune system.

The immune system deploys choline-containing lipids as potent signalling molecules. LPC is a lipid with proinflammatory properties that is released from activated immune cells such as macrophages in response to stress, infection or inflammation. LPC is produced by hydrolysis of PC by phospholipase A2 and exerts its effects (such as chemotaxis) through binding and activating G-protein-coupled receptors and Toll-like receptors on target cells^185,186. Similarly, platelet-activating factor, an LPC-like molecule derivatized with an additional acetyl group, is released by immune and endothelial cells to activate the PAF receptor on target cells, triggering inflammatory cascades, platelet aggregation, dilation of blood vessels, allergic responses and apoptosis, among other effects¹⁸⁷.

Very recent work by the Moltke and Jay groups has also highlighted a surprising role for acetylcholine in the anti-helminth immune response, at least in mice. Both groups demonstrated that after helminth infection, tuft cells of the small intestinal epithelium release acetylcholine into the lumen of the small intestine that directly induces chloride secretion from luminal epithelial cells. Secreted chloride helps to draw water into the small intestine and clear helminth infection via intestinal ‘flushing’^188,189.

Dysfunction of choline homeostasis in mammals

Consistent with its role as an essential nutrient that both fuels basal cellular metabolism and supports specialized processes in different tissues, choline dysfunction has substantial consequences for human health through all stages of life, from gestation to old age.

Gestation

Choline is a critical nutrient for development during gestation and early life, particularly for the formation and maturation of the nervous system^{148,149,151,152,156,190}. Choline and choline-containing lipids are 2–3 times higher in neonates compared to adults, suggesting choline also plays a critical role in postnatal development^151,191,192. Its vital importance for fetal development is underscored by its tenfold enrichment in amniotic fluid over maternal serum¹⁵⁶. The primary evidence that choline regulates fetal neurological development is the clinical observation that gestational choline insufficiency is strongly associated with neural tube defects, notably mirroring the developmental pathologies observed in folate insufficiency^149,156. Gestational choline insufficiency has additionally been linked to non-neurological developmental abnormalities including cleft lip, hypospadias, heart abnormalities and congenital diaphragmatic hernia^151,193.

Investigations of perinatal choline insufficiency in mice have helped to characterize choline-associated fetal neurodevelopmental abnormalities and their possible mechanisms. Histological examination of mouse pups born to choline-starved mothers revealed hippocampal and midline defects, impaired proliferation, differentiation and migration of neural stem cells, and inhibited cerebral angiogenesis^194,195. These defects were correlated with global DNA hypomethylation within the brain^196,197, prompting the hypothesis that choline insufficiency leads to hypomethylation in neuronal and endothelial fetal progenitor cells, ultimately inducing widespread transcriptional dysregulation and cellular dysfunction in these cells. A second potential mechanism to explain neurodevelopmental defects in choline insufficiency is inadequate synthesis of acetylcholine, an abundant neurotransmitter proposed to serve as a morphogen that regulates the growth, differentiation and plasticity of CNS neurons during development and early life^198,199. Consistently, choline supplementation has been associated with decreased risk for neurodevelopmental abnormalities and increased memory and cognition performance in offspring, observations reproduced in both mice and rats, in which perinatal choline supplementation increased pups’ performance in memory, spatial and explorative tasks^{196,200–207}.

There is additionally a sizeable collection of inborn errors of metabolism caused by deficiency in choline-metabolism genes that present with a broad range of pathologies, from malodour and liver disease to severe muscle dystrophy and neurological dysfunction, which are reviewed thoroughly elsewhere²⁰⁸. Here, we highlight the three inborn errors of choline metabolism associated with the most severe clinical disease and discuss a recently reported inborn error in choline metabolism caused by deficiency in the lysosomal LPC transporter, SPNS1 (refs. 77–79).

The first notable inborn error of choline metabolism is a congenital muscular dystrophy resulting from deficiency of choline kinase beta (CHKB). Homozygous or compound heterozygous variants in CHKB cause a rare form of congenital muscular dystrophy characterized phenotypically by early-onset muscle wasting, childhood-onset speech and motor delays, autism spectrum disorder, cardiomyopathy and seizures²⁰⁹. Histological examination of affected muscle fibres reveals striking enlargement and peripheral distribution of mitochondria and necrotic fibres²⁰⁹. While causing less severe disease than variants in choline kinase alpha (CHKA), which are incompatible with life entirely, that CHKB deficiency is associated with such severe disease supports that it nevertheless plays a significant role in choline metabolism in humans.

A second notable inborn error of choline metabolism is congenital myasthenic syndrome, which results from variants in CHAT (acetylcholine acyltransferase) or CHT1/SLC5A7 (the high-affinity choline transporter), which encode two proteins required for proper acetylcholine production in cholinergic neurons^210,211. The disease is characterized by symptoms attributable to hypoactivity of cholinergic neurotransmission controlling skeletal muscle and autonomic functions, such as poor muscle tone and weakness, poor swallowing, episodic apnoea and vocal cord paralysis^208,210,211.

A third class of inborn errors of choline metabolism is caused by variants in choline-phosphate cytidyltransferase (PCYT1A) and produces three distinct autosomal recessive diseases^21,212,213. PCYT1A is the rate-limiting enzyme in PC biosynthesis that catalyses the formation of CDP–choline, illustrating again the importance of PC homeostasis for proper function of the nervous system. The first disease associated with inborn defects in PCYT1A is spondylometaphyseal dysplasia with rod-cone dystrophy, in which affected individuals suffer from bone abnormalities causing short stature and waddling gate, vision impairment and variable levels of fatty liver and insulin resistance^21,214. The second disease associated with PCYT1A variants is Leber congenital amaurosis, characterized by severe retinal degeneration and blindness before the age of one²¹. Lastly, PCYT1A deficiency is associated with congenital lipodystrophy with severe fatty liver disease, involving abnormal distribution of fat, fatty liver deposits, type II diabetes, dyslipidaemia and short stature²¹. How different variants in PCYT1A cause varying disease phenotypes is not well understood; however, the pathophysiology of each disease is ultimately believed to be driven by PC dyshomeostasis across various cell types⁶⁰.

Recently, a novel inborn error of choline metabolism was described in humans that is due to dysfunction of SPNS1, the lysosomal LPC transporter⁷⁹. Three consanguineous individuals with severe developmental and intellectual disability, neurological impairment and cerebellar hypotrophy were found to harbour a proline-to-leucine mutation at position 295 (p.Pro295Leu) that significantly hindered SPNS1 transport function. Mice completely deficient in Spns1 are not viable, but embryos exhibt marked neurodevelopmental abnormalities, as described earlier. Together, these pathologies in humans and murine models evidence the importance of SPNS1 and ultimately choline–lipid recycling in early neurodevelopment.

Adult life

A plethora of pathologies are associated with choline dysfunction in adults and can result from dietary choline insufficiency or excess to dysfunction of endogenous choline-dependent systems. Before choline had been identified as an essential nutrient and incorporated into total parental nutrition formulations, total parental nutrition-reliant patients were at high risk for choline deficiency. In these susceptible patients, the primary manifestation of choline deficiency was the rapid development of NAFLD associated with a precipitous drop in hepatic PC levels^{14,149,215–217}. Remarkably, experiments in mice demonstrate that choline and PC levels elsewhere in the body, including the brain and serum, exhibit at most modest reductions, suggesting the liver sacrifices its choline stores to recycle and redistribute to vital tissues^176,218. As the liver is depleted of choline-containing metabolites and lipids, it loses the capacity to synthesize PC for VLDL biogenesis and experiences a decreased capacity for methylation. In turn, DNA damage, reactive oxygen species and lipids accumulate within hepatocytes, leading to the development of NAFLD with a high risk of progressing to cirrhosis and eventually hepatocellular carcinoma^216,217,219. Recent research has additionally highlighted the role that choline-derived methyl equivalents play in the pathogenesis of choline-associated NAFLD and hepatocellular carcinoma^197,220 and has suggested that low choline levels induce complex I dysfunction in the mitochondria of hepatocytes, leading to reactive oxygen species and widespread oxidative damage to hepatocytes²²¹.

While it is more common to suffer from choline insufficiency, choline excess can also have clinical repercussions. A surplus of dietary choline acutely presents with symptoms of cholinergic toxicity, such as nausea, vomiting, excessive sweating and fishy malodour, and can lead to hypotension and liver toxicity. Over time, excessive choline intake is linked to cardiac dysfunction and atherosclerosis through increasing levels of TMAO, a metabolite associated with cardiovascular disease. Excessive choline supplementation is associated with changes in the microbiome²²². How organismal choline homeostasis is maintained through dietary absorption remains opaque. The observation that TMAO production is increased with excess choline consumption sugests that when dietary choline uptake is saturated, excess choline is fermented by the microbiome to TMA and eventually converted to TMAO by hepatic flavin monooxygenases. How this balance between choline absorption and TMAO production is reached is unknown. Furthermore, whether free choline or choline-containing metabolites (such as PC and LPC) equally contribute to these fates—dietary choline absorption and TMAO production—is unexplored. The intestinal metabolism of dietary choline in conjunction with the microbiome and the molecular mechanisms involved in the absorption of choline and derived metabolism is deserving of future exploration.

Choline dysfunction can also have aetiologies not related to dietary intake, but to aberrant utilization or production at the cellular and tissue levels. An important disease associated with intrinsic alterations in choline metabolism is cancer²²³. As a structural component of membrane lipids and source of one-carbon units, choline is essential for cellular growth and division. Not surprisingly, many cancers exhibit a phenotype known as the ‘cholinic phenotype’, describing substantially increased uptake and utilization of choline by cancer cells compared to healthy tissues^70,224. This effect is so pronounced that choline-based positron emission tomography tracers are leveraged clinically for identification of both primary and metastatic cancerous lesions^225–227. Additionally, in comparison to healthy tissues, many cancerous tissues have elevated levels of total soluble choline-containing metabolites and lipids and tend to have higher expression of choline-metabolism genes such as CHKA²²⁴. They also exhibit higher uptake and accumulation of synthetic choline–lipid analogues known as alkylphosphocholines, a class of drugs currently in clinical development as novel tumour-targeting theranostics^228,229. Thus, excessive utilization of choline for growth and biosynthetic reactions is critical for supporting the metabolism in rapidly proliferating cancer cells.

Ageing

Choline intake was recently discovered to be a potential modifier of dementia risk in AD progression^230–232. This is not surprising given the body of research that highlights the importance of the cholinergic neurotransmission in memory, learning and, ultimately, in the development of AD pathology^231,232. Namely, brains of individuals with AD are found to have degeneration of deep and early basal forebrain cholinergic neurons²³¹. A number of preclinical studies in mouse models of AD have shown that choline supplementation can improve symptoms and slow progression of disease manifestations^233,234. While the mechanism remains unclear, choline may have neuroprotective effects both through decreasing levels of the neurotoxic metabolite homocysteine and through reducing activation of microglia. Similarly, while less well studied, choline dyshomeostasis is also hypothesized to be involved in the pathogenesis of PD²³⁵.

Future directions and outstanding questions

How the cell senses and regulates phospholipid homeostasis

The precise phospholipid composition of membranes, particularly PC species, is a critical regulator of many cellular processes and is known to be dysregulated in disease. The asymmetric distribution of membrane lipids is a defining feature of cellular plasma membranes and PC species are no exception to this rule. Fully saturated PC species are almost exclusively found in the cytosolic leaflet of the plasma membrane while polyunsaturated PC species are preferentially found in the exoplasmic leaflet²³⁶. How cells are capable of sensing PC levels and the distribution of chain length and saturations—and ultimately integrating this information to regulate PC biosynthesis, remodelling and catabolism—is a poorly understood metabolic phenomenon. Furthermore, how PC abundance is co-regulated with other phospholipid classes is unclear but likely physiologically relevant as the ratio of PCto PE is perturbed in diseases such as NAFLD^237,238.

The most well-studied PC sensor is the protein phosphocholine cytidylyltransferase alpha (CCTα, also known as PCYT1A), which catalyses the final and rate-limiting step in PC synthesis^29,239. PCYT1A exists in both free and membrane-associated forms and, importantly, the membrane-associated form is more active²³⁹. It was shown that PC depletion in membranes favours PCYT1A association through increasing membrane curvature and elasticity; thus, PCYT1A responds to low PC content in membranes by increasing PC biosynthesis^29,239,240. Choline kinase is a second enzyme believed to regulate the flux of PC biosynthesis²⁴¹. As the key enzyme required for trapping choline within cells, it acts as a gateway for the entry of choline into PC biosynthetic pathways (Fig. 4a). While no studies have been conducted to demonstrate this regulatory role in humans, experiments in rats revealed that choline kinase expression increases in response to fatty acid supplementation, ensuring choline is available for the metabolism of fatty acids into PC²⁴². Moreover, experiments in yeast show that two phosphorylation sites regulate choline kinase activity and ultimately PC levels²⁴³. Finally, PC content may also be ‘sensed’ indirectly when depletion within membranes leads to aberrant signalling. Many transmembrane receptors are activated or inhibited by local concentrations of specific lipids and/or have signal transduction mechanisms that rely on the immediate lipidic environment^244–246. Thus, depletion of PC from membranes can alter signalling cascades, and while not intended to sense PC per se, these cascades can ultimately activate transcriptional machineries, modulate metabolic programmes and broadly lead to changes in cellular phenotype²⁴⁷.

Fig. 4 |. Outstanding questions in choline metabolism.

a, Understanding molecular mechanisms involved in sensing and regulating choline metabolism. Phosphorylation of choline by CHKA/CHKB traps choline in the cell and commits it to de novo PC biosynthesis. PCYT1A, the rate-limiting enzyme of PC synthesis, is controlled by subcellular localization with the membrane-bound form being more enzymatically active. Low PC content causes negative membrane curvature, which recruits PCYT1A to the membrane. Membrane-bound PCYT1A shows high enzymatic activity to generate PC and restore phospholipid homeostasis. Microdomains within membranes replete in PC species may contribute to cellular sensing of phospholipid homeostasis by influencing membrane–receptor interactions and subsequent signal transduction. PC species display immense diversity in acyl chain lengths and saturation. The cellular function and mechanisms governing remodelling of PC chains via the Lands cycle is largely unexplored. b, Alternative strategies for choline acquisition and synthesis used by mammalian cells. Choline-containing lipids found in lipoproteins or in conjugation with albumin are endocytosed and delivered through the endolysosomal system to lysosomes. Alternatively, choline-containing lipids can be delivered to the lysosome via autophagy. In the lysosome, choline-containing lipids are broken down into choline, phosphocholine, LPC and GPC. The metabolites can then be incorporated into downstream pathways requiring choline or choline-containing metabolites. In the liver, the PEMT pathway generates PC from PE through three subsequent reactions utilizing SAM as a methyl donor.

Apart from total PC levels, cells are able to sense and adjust the distributions of PC species’ chain lengths and saturations, a process referred to as chain ‘remodelling’²⁴⁸. PC chain remodelling is achieved via the cytosolic Lands cycle and/or the lysosomal SPNS1-dependent salvage pathway^{77,116,249,250}. While these pathways take place in different subcellular compartments, namely the cytosol and lysosome, respectively, their mechanisms follow parallel pathways (Fig. 4a). In the first step of both pathways, one acyl chain is hydrolysed from PC by type A phospholipases (PLA2 in the cytosol²⁴⁹ and PLA2G15 in the lysosome^116,251) to produce the single-chained intermediate, LPC. In the second step, the second acyl chain is added back to LPC by one of four LPC acyltransferase isoforms (LPCAT1–LPCAT4) to regenerate the double-acylated PC species²⁴⁹. Notably, how the localization, function and/or regulation differs between the LPCAT isoforms is still under study. Ultimately, these pathways represent rapid, two-step cycles for PC acyl chain remodelling, and while their biochemical mechanisms are well characterized, how and when cells apply them to alter the PC species landscape remains a mystery.

There is a growing interest in understanding the regulation of phospholipid (and especially PC) synthesis and remodelling given correlations between lipidome alterations and disease states including inflammation/infection, metabolic disease, ageing and cancer^252,253. Characterizing how changes to the phospholipidome occur and why these change drive disease pathophysiology offers an avenue for therapeutic intervention. For instance, understanding the role of lipid saturation in glioma survival enabled the Mischel group to uncover that EGFR-driven gliomas rely on lipidomic drift towards more saturated lipids to support constitutive EGFR signalling and drive rapid tumour growth, thus identifying a druggable pathway in high-grade gliomas²⁴⁴. Ultimately, a deep understanding of how changes in phospholipid acyl chain length and saturation impact signalling pathways, cellular phenotypes and disease progression could open new avenues to treating a diverse array of clinical pathologies.

Cellular and organismal plasticity in obtaining choline

Choline is essential for maintaining basal metabolism in all cells, and as a key structural component of phospholipids, is required in excess to support cell growth and replication. Yet, mammalian cell lines cultured in medium depleted of free choline survive and even continue to replicate for days before succumbing to the nutrient starvation, suggesting they may use multiple mechanisms and/or pathways for choline acquisition⁷⁷.

At the cellular level, one stratagem for acquiring choline is to recycle choline from choline-containing lipids⁷⁷. The endocytic and autophagic pathways capture biomolecular material (including choline-containing lipids) from the exterior and interior of the cell, respectively, for degradation and recycling in the lysosome²⁵⁴. In the case of choline, cells capture extracellular lipid–albumin complexes and lipoprotein particles containing high levels of the choline-containing lipids PC and SM²⁵⁵. PC and SM are degraded by lysosomal acid lipases to release choline-containing catabolites such as LPC, GPC and phosphocholine, which can be recycled into PC biosynthesis or further catabolized to free choline and utilized for methyl equivalents²⁵⁶ (Fig. 4b). The importance of these choline-recycling pathways is highlighted in our recent work, in which we demonstrate that lysosomal LPC salvage and recycling by the lysosomal transporter SPNS1 drastically prolongs survival of pancreatic cancer cells deprived of free choline¹⁷³. This argues that cancer cells, and especially those in choline-limiting environments, critically depend on lysosomal choline–lipid recycling as one mechanism for maintaining PC homeostasis. While most cellular choline demand is for PC biosynthesis, cells require some choline in its free form for conversion to betaine for use in methylation reactions. In this case, endogenous and exogenous choline-containing lipids can be catabolized to liberate free choline. Notably, when cells are low in free choline, PC is cannibalized to produce free choline to drive methylation reactions, particularly in the liver²⁵⁷. Finally, it should be noted that cells can synthesize non-choline phospholipids such as PE to help support overall phospholipid levels and compensate for loss of PC; however, significantly altered ratios of PE:PC will eventually lead to cell dysfunction²³⁷. In liver cells, PE can be directly converted to PC by the enzyme PEMT; however, this pathway is insufficient to meet physiological choline demands, as discussed previously.

Broadening our view, there is impressive plasticity in satisfying choline requirements at the organismal level as well. While a completely choline-free diet eventually results in NAFLD and muscle damage, diets that do not meet the NIH adequate intake are rarely associated with clinical symptoms in otherwise healthy adults, suggesting the human body is highly capable of adapting to reduced choline availability^14,152,172. While choline recycling and redistribution throughout the body are not easily studied in humans, they have been extensively studied in mice, where the liver may serve as a redistribution hub that funnels choline to vital tissues. Understanding if and how choline redistribution to vital tissues occurs in humans is an important area of future research that will contribute to our basic knowledge of choline homeostasis as well as our ability to treat choline-related diseases.