Substrate supply, compartmentation, and utilization in hepatic de novo lipogenesis

Abstract

Hepatic de novo lipogenesis (DNL) is a fundamental process that supports energy storage, membrane biogenesis, and lipid signaling. However, chronically elevated hepatic DNL is a risk factor for insulin resistance and liver fat accumulation. If sustained, this can drive inflammation and fibrosis, with progression to cirrhosis and sometimes hepatocellular carcinoma, as well as numerous cardiometabolic comorbidities. Therefore, new discoveries on the basic mechanisms that control hepatic DNL may enable therapeutic modulation that improves major health outcomes. This review synthesizes recent advances in how the liver channels lipogenic substrates through mitochondrial and cytosolic pathways into DNL. It also highlights how these substrates regulate lipogenic flux by supplying cytosolic acetyl-CoA and NADPH.

Hepatic de novo lipogenesis in physiology and metabolic liver disease

Lipids are essential biomolecules that support organismal homeostasis by storing energy, forming cellular membranes, and serving as signaling mediators. Cells obtain lipids from the diet or de novo lipogenesis (DNL) (see Glossary), the synthesis of fatty acids from non-lipid precursors. Adipose tissue is the major site of DNL and lipid storage. The liver also performs DNL to store and direct lipids to the circulation and peripheral tissues, especially during acute caloric surplus. DNL requires both carbon and reducing power in the form of NADPH to build a fatty acyl chain (Figure 1). Synthesized fatty acids can then be incorporated into triglycerides and other complex lipids.

Figure 1: Overview of hepatic de novo lipogenesis (DNL).

Hepatic DNL uses acetyl-CoA as the carbon substrate and NADPH as reducing power to synthesize fatty acids. Acetyl-CoA carboxylase (ACC) converts acetyl-CoA to malonyl-CoA, and fatty acid synthase (FASN) extends the fatty acyl chain by adding 2-carbon units from malonyl-CoA. FASN consumes 2 NADPH per 2-carbon addition cycle. FASN continues to add 2-carbon units until it is allosterically inhibited by palmitoyl-CoA, which has a fatty acyl chain of 16 carbons (C₁₆). Then, the fatty acid is released from the acyl carrier protein (ACP) as palmitate and may be esterified to glycerol, which can hold up to three acyl chains to form triglycerides. Excess hepatic triglyceride accumulation causes steatosis, which can progress to steatohepatitis, cirrhosis, and sometimes hepatocellular carcinoma.

In healthy individuals, hepatic DNL rates are low during fasting and increase after feeding, driven in part by postprandial insulin signaling that stimulates DNL [1]. However, individuals with whole-body insulin resistance, with or without type 2 diabetes (T2D), typically have higher hepatic DNL than insulin-sensitive individuals [2–4]. Hepatic steatosis is defined by lipids accumulating to >5% of liver weight and is the hallmark of metabolic dysfunction-associated steatotic liver disease (MASLD), formerly non-alcoholic fatty liver disease (NAFLD). Patients with MASLD also commonly have elevated rates of DNL [4–7]. Continued lipid accumulation that is exacerbated by elevated DNL instigates inflammation and fibrosis characteristic of metabolic dysfunction-associated steatohepatitis (MASH), formerly non-alcoholic steatohepatitis (NASH) [8]. Continued inflammation and fibrosis can cause cirrhosis, sometimes requiring liver transplantation, and may progress to hepatocellular carcinoma. Moreover, excessive DNL not only harms the liver, but it is also linked to extrahepatic complications, including tumor growth in mice and coronary artery disease in humans [9, 10]. Thus, prolonged elevation of hepatic DNL may lead to serious adverse health outcomes.

Given these hepatic and systemic consequences, there is a critical need to better understand the mechanisms that control DNL. This includes delineating the complex metabolic network that supplies carbon and reducing equivalents to support and regulate DNL. Accordingly, DNL has been intensely investigated, resulting in therapeutics in clinical trials, yet effective therapeutic targeting has remained elusive [11, 12]. This review highlights recent findings on metabolic sourcing for DNL and related regulatory mechanisms.

Carbon substrates for hepatic de novo lipogenesis

Carbon sources that fuel hepatic DNL

Cytosolic acetyl-CoA is the central building block for DNL. Carbohydrates, especially fructose, serve as major carbon substrates for production of lipogenic acetyl-CoA, which resides in the cytosol. Foundational research identified enzymatic pathways that convert sugars into cytosolic acetyl-CoA and subsequently into fatty acids, enabling development of initially promising therapeutics [11, 12]. However, these compounds target obligate DNL enzymes such as acetyl-CoA carboxylase (ACC) and fatty acid synthase (FASN). Despite potential benefits, these inhibitors impair physiologically essential hepatic DNL, with secondary adverse effects on non-hepatic tissues. For example, the ACC inhibitor MK-4074 induces hypertriglyceridemia [13]. Together, these findings illustrate a need to better understand how DNL routes carbon from a diverse suite of precursor molecules into lipogenic acetyl-CoA. Recent work shows that extrahepatic glucose and fructose metabolism, together with hepatic metabolism of fructose, amino acids, and ketone bodies, contributes to DNL and highlights mitochondrial pathways as key control points (Figure 2).

Figure 2: Cytosolic acetyl-CoA production for hepatic DNL.

Abbreviations: ACSS2, acetyl-CoA synthetase, Acetate + ATP + CoA -> Acetyl-CoA + AMP + PPi; AACS, acetoacetyl-CoA synthetase, Acetoacetate + ATP + CoA -> Acetoacetyl-CoA + AMP + PPi; LDH, lactate dehydrogenase, Lactate + NAD⁺ -> Pyruvate + NADH + H⁺; ALT2, alanine transaminase 2, Alanine + α-Ketoglutarate -> Pyruvate + Glutamate; GLS2, glutaminase 2, Glutamine + H₂O -> Glutamate + NH₄⁺; MPC, mitochondrial pyruvate carrier, Pyruvate (cytosol) + H⁺ (cytosol) -> Pyruvate (mitochondria) + H⁺ (mitochondria); OGDH, oxoglutarate dehydrogenase, α-Ketoglutarate + CoA + NAD⁺ -> Succinyl-CoA + NADH + CO₂; CiC, mitochondrial citrate carrier, Malate (cytosol) + Citrate (mitochondria) -> Malate (mitochondria) + Citrate (cytosol); ACLY, ATP-citrate lyase, Citrate + ATP + CoA -> Oxaloacetate + Acetyl-CoA + ADP + Phosphate; OGC, oxoglutarate carrier, Malate (cytosol) + α-Ketoglutarate (mitochondria) -> Malate (mitochondria) + α-Ketoglutarate (cytosol); IDH1, isocitrate dehydrogenase 1, α-Ketoglutarate + CO₂ + NADPH -> Isocitrate + NADP⁺; ACC, acetylCoA carboxylase, Acetyl-CoA + HCO₃^- + ATP -> Malonyl-CoA + H⁺ + ADP + Phosphate; FASN, fatty acid synthase, Acetyl-CoA + (n) Malonyl-CoA + (2n) NADPH + (2n) H⁺ -> (n+1) CoA + (n) CO₂ + (2n) NADP⁺ + (n-1) H₂O + C_2n+2 Fatty Acid; OAA, oxaloacetate; α-KG, α-ketoglutarate.

Dietary sugar and differences in first-pass processing of fructose versus glucose

The carbohydrates fructose and glucose are structurally similar yet markedly differ in hepatic lipogenic potential. In human NAFLD (MASLD) patients, oral fructose stimulates hepatic DNL, whereas glucose does not [7]. Likewise, dietary supplementation with fructose or sucrose (a glucose-fructose disaccharide), but not glucose, over 7 weeks increases hepatic DNL [14]. Together, these studies illustrate the higher hepatic lipogenic potential of fructose than glucose in humans.

Recent studies have investigated how fructose and glucose contribute to DNL. Several works traced dietary fructose handling in the small intestine, a major nutrient absorption site [15, 16]. ¹³C-tracing showed that the small intestine metabolizes fructose in a ketohexokinase-dependent manner. When supplied at low doses, ¹³C-labeled fructose appeared in circulation as glucose, TCA cycle intermediates, and amino acids derived from intestinal metabolism. However, at higher fructose loads or during intestinal ketohexokinase disruption, the small intestine fails to fully process fructose. A portion is absorbed intact and metabolized by the liver, consistent with rapid hepatic fructolysis by ketohexokinase after fructose ingestion [15–17]. This is followed by conversion to pyruvate and mitochondrial metabolism to generate citrate, which ATP-citrate lyase (ACLY) converts into cytosolic acetyl-CoA for DNL. In parallel, excess fructose is metabolized by the colonic microbiota to TCA cycle metabolites, amino acids, or short-chain fatty acids [15]. These metabolites are either excreted into feces or absorbed into portal circulation. Notably, the liver metabolizes microbiota-derived acetate to acetyl-CoA using acetyl-CoA synthetase (ACSS2), supporting DNL through a microbiota-mediated circuit [18]. This route is also active during ACLY disruption. Thus, dietary fructose can supply hepatic DNL through multiple routes.

Compared to fructose, intestinal glucose is minimally metabolized and mainly taken up into circulation. A fraction is taken up by the liver, whereas most is utilized by skeletal muscle [15]. This is physiologically rational, because during times of high glucose availability, skeletal muscle can replenish glycogen stores without compromising brain glucose supply. In addition, unlike fructose, intestinal glucose absorption correlates linearly with glucose dosage over a broad range and is not readily saturated [15]. Thus, glucose is less likely than fructose to reach the colonic microbiota and to be converted to metabolites such as acetate for ACSS2-dependent DNL. Overall, these mechanisms begin to explain the difference between the lipogenic potential of glucose and fructose.

Extrahepatic conversion of glucose into hepatic lipogenic precursors

These data raise the question of how glucose ultimately contributes to hepatic DNL. This is important because hyperglycemia secondary to systemic and hepatic insulin resistance is common in MASLD patients. By mass action, hyperglycemia could increase hepatic glucose uptake and conversion to fatty acids within the constraints of limited hepatocyte glycolytic capacity. Excess fructose may amplify this effect, consistent with the observation that aldolase-B–deficient mice that accumulate fructose-1-phosphate exhibit increased glucose-supplied hepatic DNL [19]. This phenotype depends upon glucokinase, suggesting that glucose flux through hepatic glycolysis supplies the lipogenic carbon. Indeed, glucose can trace into lipids in both mice and primary murine hepatocytes [20–22]. However, because insulin resistance can decrease net hepatic glucose uptake, the accompanying hyperglycemia, as an isolated factor, may not be a major driver of elevated hepatic DNL [23].

Given limited first-pass metabolism of glucose, extrahepatic glucose metabolism plays the primary role in channeling dietary glucose into hepatic DNL. A recent investigation traced ¹³C substrates (glucose, lactate, acetate, glutamine, and alanine) into palmitate in mouse primary hepatocytes [24]. In this system, other substrates contributed more to DNL than glucose. In vivo, dietary ¹³C-glucose labeled plasma and hepatic lactate, pyruvate, and alanine in addition to fatty acids in the liver. The authors concluded that extrahepatic glucose supplied DNL via these metabolites. Consistent with this, directly administered lactate/pyruvate is a potent hepatic DNL substrate in vivo [25, 26].

Contribution of amino acids and ketone bodies to DNL

Amino acids as lipogenic carbon sources:

In obese patients, dietary protein intake is associated with increased risk of MASLD and MASH, raising the possibility that amino acids contribute to elevated DNL [24, 27]. Based on known pathways, many amino acids can feed DNL through TCA cycle metabolism and flux to cytosolic acetyl-CoA. Given their high plasma abundance, alanine and glutamine are reasonable major substrates [28]. Consistent with this, ¹³C-glutamine and ¹³C-alanine trace into palmitate in mouse primary hepatocytes [24]. In vivo, ¹³C-glutamine contributed more to DNL than ¹³C-glucose, whereas alanine was not directly tested. In a proof-of-principle physiological experiment, disruption of amino acid catabolism or a low protein diet decreased hepatic triglycerides in the ob/ob mouse model of T2D [24]. Thus, amino acids can meaningfully contribute to hepatic DNL in vivo and may uniquely drive MASLD progression.

Ketone bodies as lipogenic carbon sources:

As reviewed recently, ketone bodies may also supply DNL [29]. Hepatic ketogenesis provides extrahepatic tissues with noncarbohydrate energy during starvation. Within hepatocytes, acetyl-CoA condensation produces the ketone bodies acetoacetate and β-hydroxybutyrate. However, these ketones can also be routed into fatty acid synthesis [25, 30–33]. While seemingly paradoxical, ketones could maintain essential DNL during fasting and be stored again as lipids during refeeding. Mechanistically, acetoacetyl-CoA synthetase (AACS) and mitochondrial β-hydroxybutyrate dehydrogenase (BDH1) contribute to ketone-supplied DNL, but neither appears essential, suggesting additional mechanisms await discovery [31]. Moreover, ketone-supplied DNL can be upregulated when other routes are disrupted. In mice, DNL from acetoacetate increases during disruption of mitochondrial citrate-export mediated DNL [25]. Overall, these studies show ketone bodies are indeed lipogenic but that much more remains to be learned about basic mechanisms and regulation.

Mitochondrial compartmentation and transport as control points for lipogenic flux

Many DNL substrates require mitochondrial metabolism for trafficking into the cytosol and the lipogenic acetyl-CoA pool. However, because these substrates cannot passively diffuse across the mitochondrial inner membrane, conductance through membrane transporters (carriers) is required. Recent work highlights substrates and transporters contributing to hepatic DNL.

Circulating lactate/pyruvate requires mitochondrial routing to supply hepatic DNL. Lactate and pyruvate are imported into the liver, where lactate is oxidized to pyruvate, and pyruvate enters mitochondria via the mitochondrial pyruvate carrier (MPC). Accordingly, hepatic MPC deletion markedly decreases DNL from lactate/pyruvate [25]. Notably, MPC deletion also decreases protein abundance of the lipogenic enzymes ACLY, ACC, and FASN [25]. Consistent with this, hepatic MPC loss also decreases DNL from acetate, which does not require carbon routing through mitochondria. Furthermore, hepatic MPC disruption in normal chow-fed mice decreases total DNL from all substrates as measured by deuterated water tracing [34]. These observations in mice may translate to the clinic, where in NASH (MASH) patients, MSDC-0602K, a compound that inhibits the MPC, improved hepatic steatosis [35]. Overall, these studies show the MPC controls hepatic DNL and may also mediate MASLD development and progression.

For continued routing into DNL, after pyruvate is imported by the MPC into the mitochondrial matrix, it is oxidized to acetyl-CoA. Mitochondrial acetyl-CoA must then be delivered to the cytosol. However, there are no known mitochondrial acetyl-CoA transporters. The major mechanism of mitochondrial acetyl-CoA export is first, conversion to citrate, then export by the mitochondrial citrate carrier (CiC). ACLY then cleaves cytosolic citrate to acetyl-CoA. Consistent with this, hepatic CiC deletion decreases DNL from lactate/pyruvate but not acetate [25]. Furthermore, hepatic CiC deletion decreases hepatic steatosis in mice with diet-induced obesity [36]. These genetic data confirm early biochemical studies suggesting that the CiC exports mitochondrial carbon for DNL [37–41].

Like pyruvate, alanine requires mitochondrial metabolism to supply DNL. Alanine-fed DNL is routed through deamination to pyruvate, by either cytosolic or mitochondrial alanine transaminase (ALT1/ALT2), with expected flux through citrate and the CiC. Yet, understanding of the alanine DNL pathway remains incomplete. Although the mitochondrial serine transporter sideroflexin 1 (SFXN1) has been reported to transport alanine, the primary mitochondrial alanine transporter remains unidentified [42].

Conversely, the most direct path for glutamine-supplied DNL bypasses pyruvate and the CiC. As with alanine, the identity of the major mitochondrial glutamine transporter is unknown. Nonetheless, once glutamine enters mitochondria, hepatic mitochondrial glutaminase (GLS2) deamidates it to glutamate, followed by deamination and oxidation to α-ketoglutarate. α-ketoglutarate may be exported to the cytosol and, depending on cellular redox state, be reductively carboxylated by isocitrate dehydrogenase (IDH1) to isocitrate. Cytosolic aconitase then readily converts isocitrate to citrate that is available for DNL. This supplies net carbon for DNL whereas forward flux of α-ketoglutarate through the TCA cycle would only provide oxaloacetate as an acetyl-CoA carrier. Consistent with this model, genetic disruption of either GLS2 or IDH1 in ob/ob mice decreases hepatic triglycerides, further supporting a role for glutamine as an in vivo lipogenic substrate [24].

Identifying extrahepatic carbon sources and how substrate compartmentation regulates DNL programming

As highlighted above, a strong body of evidence demonstrates that carbohydrates, amino acids, and ketone bodies supply carbon substrate for DNL. However, outstanding questions remain. Since hyperglycemia is characteristic of insulin resistance, a key systemic question is how glucose feeds DNL in insulin-sensitive versus -resistant states. Experiments could measure hepatic glucose uptake and incorporation into lipids in each state. Moreover, they could identify sites of extrahepatic glucose degradation and the resulting substrates that feed hepatic DNL. Potential tissues include skeletal muscle and erythrocytes. Both consume glucose and release lactate [43–45]. Experiments could also locate where amino acids are produced. Skeletal muscle produces alanine and glutamine, and the brain releases glutamine [43, 46]. Such studies may delineate the distinct, organ-specific pathways by which glucose and amino acids supply DNL.

Mitochondrial metabolism regulates DNL. Loss of the MPC, but not CiC decreases lipogenic protein expression [25]. This could reflect epigenetic or posttranslational modifications by intermediates such as lactate. Additional studies could explore how metabolic compartmentation between the cytosol and mitochondria dictates substrate preference for DNL. Furthermore, work is necessary to define how plasma membrane transporters regulate hepatic uptake of DNL precursors. Acetate transporters are an ideal starting point because acetate does not require mitochondrial metabolism for DNL. Plasma membrane transporters for lactate, pyruvate, glutamine, and alanine, and whether specific isoforms differentially channel these substrates into intracellular and lipogenic metabolism, merit focused investigation. This is paramount given evidence that extrahepatic release of these substrates supplies the liver with lipogenic carbon. Overall, these efforts will better define the metabolic pathways of DNL.

Nutritional regulation of hepatic DNL

Carbohydrates and amino acids distinctly regulate hepatic DNL

To transform carbon substrates into lipids through DNL, the liver must express the appropriate enzymes and metabolite transporters. As a mechanism of physiological coherence, some substrates activate and facilitate their own channeling into DNL. For example, glucose and fructose activate the transcription factor carbohydrate response element-binding protein (ChREBP), which increases expression of enzymes necessary to convert carbohydrates to lipids. Fructose-mediated ChREBP activation also stimulates DNL from acetate, consistent with fructose-ChREBP signaling upregulating a broader DNL program [18, 19].

Fructose and glucose also differentially regulate DNL by modulating its reciprocal, fatty acid oxidation (FAO). Decreased FAO during DNL prevents futile cycling. In high-fat diet-fed mice, fructose supplementation decreased FAO more than glucose [47]. Moreover, fructose decreased activity of the rate-limiting enzyme of FAO, carnitine palmitoyl transferase 1A (CPT1A) more than glucose. In the same study, fructose increased abundance of malonyl-CoA, an allosteric CPT1A inhibitor, and CPT1A acetylation, consistent with CPT1A inhibition by both allosteric and post-translational mechanisms. Thus, fructose-mediated FAO suppression may reinforce its greater lipogenic potential over glucose. Lastly, fructose may prime hepatocytes for DNL through modulating mitochondrial dynamics by stimulating mitochondrial fission (Box 1).

Box 1: Fructose regulates mitochondrial dynamics to support DNL.

Fructose may prime the liver for DNL by altering mitochondrial morphology and dynamics. Compared to glucose, addition of fructose to a high fat diet increases mitochondrial fission, decreases mitochondrial fusion, and suppresses mitophagy [47, 88]. Genetic ablation of mitophagy decreases lipogenic protein expression and hepatic triglycerides [89]. Inhibiting mitochondrial fission decreases lipogenic gene expression and lipid droplets in cultured epithelial cells [90]. Conversely, mitochondrial fusion or inhibiting mitophagy promotes FAO [89, 91, 92]. Together, these studies indicate fructose may regulate DNL by stimulating mitochondrial fission and blunting mitophagy.

Like sugars, amino acids, as previously discussed for glutamine and alanine, can serve as DNL substrates. Yet, the mechanisms by which the liver senses their availability for and regulates their shunting into DNL remain less understood. Amino acid sensors may monitor total energy status rather than specifically regulate DNL. The amino acid sensor GCN2 inhibits protein synthesis during amino acid deprivation. Genetic GCN2 disruption decreases lipogenic gene expression and hepatic triglycerides [48, 49]. Similarly, deleting the amino acid sensor TAS1R1 decreases hepatic triglycerides [50]. On the other hand, greater amino acid abundance increases DNL. Supplementing primary hepatocytes with all 20 amino acids increases lipogenic gene expression [51]. Notably, removing proline negates this effect, and additional proline supplementation amplifies it. Interestingly, proline hydroxylation is necessary for full ChREBP activity and may partially explain this finding when free proline is limiting [52]. These studies link amino acid sensing to DNL, highlight a distinct role for proline, and illustrate a need to more deeply delineate regulatory mechanisms. Future work should define how glutamine and alanine are sensed and partitioned into DNL and identify enzymes and transporters that control their routing.

CoA sequestration as a potential mechanism of DNL substrate-level control

α-ketoglutarate, a metabolite generated from both sugar and amino acid metabolism, may inhibit DNL through a mechanism distinct from the major sugar- and amino acid–driven regulatory pathways. Supplementary dietary α-ketoglutarate decreases hepatic triglycerides and ACC and FASN protein abundance in hyperlipidemic mice [53]. Mechanistically, high levels of α-ketoglutarate oxidation to succinyl-CoA by oxoglutarate dehydrogenase (OGDH) may sequester coenzyme-A (CoA). CoA sequestration could limit synthesis of lipogenic cytosolic acetyl-CoA, either by mass action or regulatory feedback mechanisms. Consistent with this model, hepatic ChREBP knockdown induces OGDH expression and decreases CoA levels [54]. Intriguingly, in the case of α-ketoglutarate, this could blunt excessive DNL during periods of high sugar and amino acid supply. As another potential path to the same outcome, the branched chain amino acids (BCAAs) are metabolized to mitochondrial acetyl- and succinyl-CoA. Like α-ketoglutarate, BCAA supplementation can decrease hepatic steatosis [55–57]. Yet, the BCAA leucine is also utilized as a hepatic DNL substrate by routing through ketogenesis and the AACS pathway [25]. Thus, depending on metabolic context, BCAAs may serve as a DNL substrate or constrain hepatic DNL through mitochondrial CoA retention.

Future experiments to resolve substrate-level control will require modulating nutritional state and measuring DNL fluxes alongside substrate availability, signaling, and compartmentalized metabolite pools. Isotope tracing paired with targeted perturbations of ChREBP, amino acid sensors, and key plasma membrane and mitochondrial transporters can define when each substrate drives DNL. Parallel profiling of CoA species and acetyl-CoA pools can test whether CoA sequestration constrains DNL across diverse nutrient states. Such integrative approaches may reveal control nodes that decrease pathological DNL without suppressing DNL essential for healthy hepatic and whole-body function.

Reducing equivalents for hepatic DNL

Cytosolic NADPH supplies reducing power for DNL

Hepatic DNL requires not only carbon substrate but also reducing power, specifically cytosolic NADPH. FASN utilizes 2 NADPH molecules per two-carbon unit added to a nascent fatty acyl chain. However, lipogenic NADPH sources remain unclear, given the liver expresses 352 known or predicted NAD(P)H metabolizing enzymes [58]. Recent work tested candidate sources of cytosolic NADPH for DNL, including serine catabolism, the pentose phosphate pathway (PPP), malic enzyme 1 (ME1), and IDH1 (Figure 3) [26].

Figure 3: Cytosolic NADPH production for hepatic DNL.

Abbreviations: G6PD, glucose-6-phosphate dehydrogenase, Glucose-6-Phosphate + NADP⁺ -> 6-Phosphogluconolactone + NADPH + H⁺; 6PGD, 6-phosphogluconate dehydrogenase, 6-Phosphogluconate + NADP⁺ -> Ribulose-5-Phosphate + CO₂ + NADPH; SHMT1, serine hydroxymethyltransferase 1, Serine + Tetrahydrofolate -> Glycine + (5,10)-Methylene-Tetrahydrofolate; ALDH1L1, aldehyde dehydrogenase 1 family member L1, 10-Formyltetrahydrofolate + NADP⁺ + H₂O -> Tetrahydrofolate + CO₂ + NADPH + H⁺; MTHFD1, methylenetetrahydrofolate dehydrogenase, cyclohydrolase, and formyltetrahydrofolate synthetase 1, (5,10)-Methylene-Tetrahydrofolate + NADP⁺ + H₂O -> 10-Formyltetrahydrofolate + NADPH + H⁺; ME1, malic enzyme 1, Malate + NADP⁺ -> Pyruvate + CO₂ + NADPH; IDH1, isocitrate dehydrogenase 1, Isocitrate + NADP⁺ -> α-Ketoglutarate + CO2 + NADPH; G6P, glucose-6-phosphate; 6PGL, 6phosphogluconolactone; 6-PG, 6-phosphogluconate; R5P, ribose-5-phosphate; 3PG, 3-phosphoglycerate; THF, tetrahydrofolate; 5,10-MeTHF, 5,10-methylene-tetrahydrofolate; 10-fTHF, 10-formyltetrahydrofolate; OAA, oxaloacetate.

Oxidative glucose metabolism through the PPP produces NADPH. Fructose feeding increases activity of the PPP enzyme glucose-6-phosphate dehydrogenase (G6PD) in mice [59]. However, a recent study found that the PPP supplies lipogenic NADPH in brown adipose tissue, but not in primary hepatocytes or the liver [26]. Rather, the authors found that serine-supported folate cycling produces NADPH in both primary hepatocytes and the liver. Cytosolic serine hydroxymethyltransferase (SHMT1) transfers a methyl unit (one-carbon) from serine to tetrahydrofolate (THF) to form 5,10-methylene-THF. This metabolite undergoes two oxidation reactions in sequence that reduce NADP+ to NADPH and regenerate THF. SHMT1 deletion or inhibition suppresses both NADPH production and serine-derived hydride incorporation into hepatic palmitate. These findings align with human studies. In patients with hepatic steatosis, serum serine negatively correlates with liver fat fraction and serum triglycerides, with additional investigation necessary to evaluate causality [60]. Overall, these data suggest serine metabolism is a key source of lipogenic NADPH.

In addition to serine, ME1 and IDH1 may furnish NADPH for DNL. Fructose feeding increases hepatic ME1 activity [59]. An ME1 null mutation decreases hepatic steatosis in high-fat diet fed female mice [61]. As described earlier, IDH1 disruption decreases hepatic triglycerides in ob/ob mice [24]. These studies are consistent with ME1 and IDH1 regulating hepatic lipid metabolism, with more work being required to directly evaluate their roles in production of lipogenic NADPH.

Coordinating NADPH supply with lipogenic demand

Linking substrate supply to utilization is necessary for efficient DNL. The liver gauges carbon substrate abundance and appropriately upregulates DNL. Since this requires NADPH, many DNL carbon sources are also capable of supporting NADPH synthesis. For example, pathways include pyruvate through malate and ME1, pyruvate through isocitrate and IDH1, and glucose through the PPP. In addition, DNL substrates may regulate NADPH producing enzymes by transcription or post-translational modifications.

As examples, glucose and fructose stimulate ChREBP to increase expression of DNL enzymes, raising the question of whether ChREBP concomitantly upregulates NADPH production. Indeed, ChREBP disruption decreases ME1 and G6PD mRNA abundance in vivo and in vitro [62, 63]. However, there appear to be no reports that identify ChREBP regulation of hepatic serine metabolism or IDH1. Chromatin immunoprecipitation studies in mice identified IDH1 as a ChREBP target in white adipose tissue, but not the liver [64]. This suggests alternative regulatory mechanisms for these pathways.

Furthermore, DNL substrates may directly modulate hepatic NADPH production. Supplementing primary rat hepatocytes with lactate or pyruvate increases G6PD and malic enzyme activity [65]. Adding the transcription inhibitor actinomycin D or translation blocker cycloheximide abolished pyruvate-induced malic enzyme activity (lactate and G6PD were not tested). This suggests lactate and pyruvate may increase enzyme activity by increasing gene expression and protein abundance. This is important given these metabolites are DNL substrates and NADPH is necessary to utilize them for DNL.

Carbon substrates may regulate NADPH producing enzymes by post-translational modifications. IDH1 acetylation at K93 is necessary for maximal IDH1 activity [66]. Mutating this site decreased NADPH production in AML12 hepatocytes and a mouse model of acute liver failure. Acetylation is an attractive mode of DNL regulation. An overabundance of acetyl-CoA could increase activity of NADPH producing enzymes. This would couple acetyl-CoA supply to NADPH biosynthesis for DNL. Thus, further investigation of protein acetylation may be especially fruitful.

NADPH partitioning between DNL and oxidative stress defense

NADPH metabolism is multifaceted. It supplies reducing power for biosynthesis and counteracts oxidative stress by regenerating reduced glutathione (GSH). Thus, the liver must partition NADPH between DNL and oxidant defense. The transcription factor nuclear factor erythroid 2–related factor 2 (NRF2) coordinates the transcriptional response to oxidative stress. NRF2 upregulates NADPH producing enzymes such as G6PD, 6-phosphogluconate dehydrogenase (6-PGD), and ME1 but also represses lipogenic gene expression [67, 68]. Thus, NRF2 may differentially regulate NADPH availability for oxidant defense versus DNL. In accord, disrupting hepatic glutathione synthesis induces NRF2 and represses lipogenic gene and protein expression [69].

DNL substrates may also generate NADPH for oxidant defense. For example, hepatic pyruvate carboxylase (PC) disruption decreases hepatic NADPH likely by limiting pyruvate cycling for NADPH production by cytosolic malic enzyme [70]. This is associated with decreased GSH, increased markers of oxidative stress and inflammation, and NRF2 induction. Notably, high-fat diet fed hepatic PC knockout mice have decreased hepatic macrovesicular steatosis but preserved microvesicular steatosis. Microvesicular steatosis is associated with increased odds of steatohepatitis [71]. This suggests there are complex trade-offs and interactions between NADPH production, DNL, hepatic lipid load, and oxidative stress.

Key open questions in hepatic NADPH supply

With many hepatic NADPH producing reactions, it will be important to determine how NADPH sourcing for DNL changes across physiologic and disease states. Conflicting data on the PPP warrant further study of its role in producing lipogenic NADPH. Decreased hepatic glucose uptake may limit PPP flux, and oxidant defense may divert NADPH away from DNL. Given this, the regulation of reactions producing carbon substrate for DNL may be especially impactful to examine. For example, malate oxidation by ME1 generates pyruvate in addition to NADPH. It also remains to be resolved how the liver commits NADPH to DNL. Because the liver must simultaneously sustain numerous NADPH-dependent processes, spatiotemporal regulation may contribute at both the portal triad and subcellular levels. Physical association of NADPH producing enzymes with FASN may channel NADPH directly into fatty acid synthesis. Overall, answering these questions is critical to understand how the liver synthesizes NADPH for DNL.

Concluding remarks and future perspectives

This review highlights how carbohydrates, amino acids, and ketone bodies supply carbon for DNL, and mechanisms of cytosolic NADPH production. However, there remain many questions to address (see Outstanding Questions).

Outstanding Questions:

1. How is dietary glucose routed through peripheral metabolism to supply hepatic DNL, and what mechanisms regulate these fluxes?

2. What regulatory framework controls utilization of amino acids for DNL?

3. Under what conditions do ketone bodies serve as major DNL substrates?

4. What physiologic states and signals determine which enzymes generate cytosolic NADPH for DNL? Are there major contributors beyond SHMT, MTHFD, ME1, and IDH1? Does the hepatic PPP contribute NADPH to DNL in vivo?

5. How does the liver partition NADPH toward DNL?

Carbohydrates are well-established substrates for DNL. However, recent evidence demonstrates roles for amino acids and ketone bodies. Further work is necessary to delineate when these substrates are used for DNL and how the liver regulates their channeling into DNL. Investigation can begin with physiologic and pathologic states associated with elevated circulating amino acids and ketone bodies. For example, glucocorticoids increase protein catabolism and associate with increased DNL [72]. Glucocorticoid production increases during stress, such as sepsis and starvation, and they are often prescribed to treat inflammatory diseases. Future experiments can test whether glucocorticoid-associated increases in hepatic DNL depend upon amino acid supplied DNL.

Ketone bodies rise during fasting and during conditions such as diabetic ketoacidosis (DKA). Refeeding-induced insulin release, or direct insulin administration, decreases ketone bodies and stimulates DNL. Consistent with this, insulin dose correlates positively with hepatic steatosis in patients with type 1 diabetes, who are prone to DKA during insulin deficiency [73]. Future studies can investigate the regulatory role of insulin by tracing β-hydroxybutyrate and/or acetoacetate into fatty acids under a range of insulin doses.

Like carbon supply, there are many possible NADPH sources for DNL. However, the physiologic circumstances and stimuli directing NADPH synthesis remain poorly characterized. When cytosolic NADPH is limited, mitochondrial NADPH equivalents could be exported for DNL. A recent study used deuterated glucose to study compartmentalized NADPH production in cancer cells [74]. Although the authors did not observe NADPH shuttling between the cytosol and mitochondria, for completeness these experiments could be repeated in hepatocytes. However, the technique utilized relies on glycolytic and PPP flux, which may be low in the liver as previously described. To circumvent these limitations, total NADPH production can be measured noninvasively using magnetic resonance spectroscopy (MRS) [75, 76]. This can be combined with acute administration of metabolites (glucose, malate, citrate) that directly feed NADPH synthesis. DNL can be simultaneously measured using MRS, stable isotope tracers, or positron emission tomography [77]. These approaches can identify substrates for DNL suitable for further mechanistic investigation with potential clinical applications (Box 2).

Box 2: Glucose-6-phosphate dehydrogenase deficiency may limit NADPH availability for oxidative defense and DNL.

Clinically, noninvasive measurements of DNL and NADPH production could benefit patients with genetic perturbations in these pathways. G6PD deficiency is a disorder that impairs NADPH production. Children with G6PD deficiency have decreased plasma cholesterol during the initial stages of hemolytic crisis [93]. This is a period of high NADPH demand for mitigating oxidative stress and supporting cholesterol synthesis for erythrocyte production. Accordingly, plasma levels of malondialdehyde, a marker of oxidative stress, are also increased in G6PD deficient patients [94]. However, in high-fat diet fed mice lacking hematopoietic G6PD, neither liver triglycerides nor serum cholesterol are altered [95]. Because hepatic G6PD remains intact in this model, these results are difficult to interpret in the context of hepatic DNL. These findings suggest potential differences between mice and humans lacking G6PD, and they may help explain the limited contribution of the PPP to hepatic NADPH production in mouse models. Noninvasive approaches could define how NADPH is partitioned between oxidative defense and DNL and inform patient care, particularly G6PD-deficient populations.

Future, studies of especially high impact may further delineate mechanisms regulating carbon- and NADPH-sourcing for DNL. We propose four major areas for future study. First, as described previously, the transcription factor ChREBP regulates hepatic DNL. Like ChREBP, sterol response element binding proteins (SREBPs) are transcription factors that augment hepatic DNL. Defining how ChREBP- and SREBP-based regulation interact will help clarify transcriptional regulation of DNL. Second, allostery and post-translational modifications regulate DNL enzyme activity. Rapidly advancing technologies now enable mechanistic tests of how specific carbon substrates and NADPH sources regulate DNL [78, 79]. Third, future studies can test how DNL substrates modify the surrounding inter- and intrahepatic microenvironment and subcellular hepatocyte organization to promote DNL. Macrophages can be either pro- or anti-inflammatory, with DNL found to support both programs [80, 81]. However, these studies did not test liver-derived macrophages, which may limit their relevance. Stellate cells instigate fibrosis, causing progression of liver disease, but are sensitive to DNL inhibition [82]. Within hepatocytes, fructose alters mitochondrial structure and promotes DNL (Box 1), but the impact of other DNL substrates on hepatocyte subcellular organization is unknown. Fourth, neural-hormonal mechanisms modulating hepatic DNL have recently been demonstrated, with much to learn about the underlying biochemical-molecular mechanisms [83].

Finally, a few topics beyond the scope of this review should be acknowledged. First, it is necessary to understand how lipogenic carbon and reducing equivalent availability, utilization, and regulation change during the progression from steatosis to MASH to cirrhosis. Identifying differences between disease stages may improve classification and facilitate development of specific therapeutics. Also, this review does not comprehensively address hormonal regulation of DNL. Hormones such as insulin almost certainly regulate how carbon and reducing power are sourced for hepatic DNL, an important topic to address in future studies.

In conclusion, DNL contributes to hepatic lipid accumulation during insulin resistance, provoking inflammation and fibrosis. Given that this is a risk factor for insulin resistance and can progress to cirrhosis, liver failure, and cancer, it is critical to understand how carbon substrates and reducing equivalents are supplied and how they regulate DNL. Understanding basic mechanisms of hepatic DNL may improve understanding and treatment of metabolic liver disease, the larger metabolic syndrome, and numerous comorbidities.

Highlights:

Dietary glucose is stored as hepatic glycogen or metabolized by peripheral tissues. Peripheral glucose metabolism produces lactate and pyruvate, which are released into circulation, taken up by the liver, and channeled into DNL.

Dietary fructose metabolism by the small intestine, liver, and colonic microbiota supplies substrate for hepatic DNL.

Amino acids and ketone bodies can also contribute carbon to DNL.

Cytosolic acetyl-CoA synthesis uses both mitochondrial-dependent and -independent pathways.

One-carbon serine metabolism through folate cycling supplies NADPH for hepatic DNL, alongside less defined contributions of malic enzyme 1 and isocitrate dehydrogenase 1. The role of the pentose phosphate pathway in supplying NADPH for hepatic DNL remains unclear.

Oxidative stress can suppress DNL to conserve NADPH for regenerating reduced glutathione.

Glossary

¹³C Tracing

technique utilizing the isotope carbon-13 (¹³C) to measure incorporation of specific substrates into various metabolic intermediates.¹³C-tracing involves administering a ¹³C-labeled substrate, sampling cells or tissue of interest, and using mass spectrometry or nuclear magnetic resonance to differentiate between naturally abundant ¹²C metabolites and ¹³C-enriched metabolites using the mass difference between ¹²C and ¹³C [84]

Carbohydrate response element binding protein (ChREBP)

transcription factor that stimulates expression of genes involved in glycolysis and DNL. ChREBP is thought to be allosterically activated by sugar phospho-metabolites

Carbon substrate

molecules that donate carbon atoms for DNL. This term refers to macromolecules (e.g. carbohydrates) or metabolic intermediates (e.g. lactate, acetate)

Cirrhosis

liver injury resulting from chronic inflammation and fibrosis, causing potential liver failure

De novo lipogenesis (DNL)

fatty acid synthesis from acetyl-CoA. This process uses NADPH for reducing power and occurs primarily in the cytosol, distinct from the mitochondrial fatty acid synthesis pathway

Fatty acid (beta) oxidation (FAO)

catabolizes fatty acids to mitochondrial acetyl-CoA, supporting TCA cycle oxidation and, in liver, ketogenesis (acetoacetate and β-hydroxybutyrate) to fuel extrahepatic tissues

Glutathione

tripeptide of glutamate, cysteine, and glycine that cycles between reduced (GSH) or oxidized (dimer with disulfide bond, GSSG). GSH detoxifies reactive oxygen species through glutathione peroxidases and is regenerated by NADPH-dependent glutathione reductase

Hepatocellular carcinoma

neoplastic transformation of hepatocytes, often preceded by cirrhosis

Ketone Bodies

refer to acetoacetate, β-hydroxybutyrate, and acetone. The redox pair acetoacetate (oxidized) and β-hydroxybutyrate (reduced) are the most common ketone bodies. They are produced by hepatic fatty acid oxidation during fasting to supply extrahepatic tissues with energy

Metabolic dysfunction-associated steatotic liver disease (MASLD)

is defined as hepatic steatosis with one cardiometabolic criterion (BMI, fasting blood glucose, blood pressure, plasma triglycerides, plasma HDL) and excluding other primary causes of steatosis. This term was introduced in 2023 to replace non-alcoholic fatty liver disease (NAFLD) and ranges from isolated steatosis to cirrhosis [85–87]

Metabolic dysfunction-associated steatohepatitis (MASH)

is defined as MASLD with inflammation and hepatocyte injury on biopsy. This term was introduced in 2023 to replace non-alcoholic steatohepatitis (NASH) [85–87]

Pentose phosphate pathway (PPP)

utilizes glucose to produce NADPH via glucose-6-phosphate dehydrogenase (G6PD) and 6-phosphogluconate dehydrogenase (6PGD)

Reducing power

electron-donating capacity provided by reduced cofactors that reduce metabolic intermediates to a lower oxidation state by hydride transfer. The enzyme fatty acid synthase uses NADPH to reduce β-ketoacyl intermediates to β-hydroxyacyl intermediates, producing NADP⁺ as a byproduct

Steatohepatitis

hepatic steatosis with hepatocyte ballooning and inflammation on biopsy

Steatosis

hepatic fat content equal to or greater than 5% of total liver weight diagnosed by imaging or biopsy

Triglycerides

lipid storage molecules consisting of glycerol esterified to three fatty acid chains