Regenerating liver uses ammonia to support de novo pyrimidine synthesis and cell proliferation

The liver exhibits a remarkable capacity for complete regeneration following partial hepatectomy (PHx). In murine models, routinely used for investigating liver regeneration¹, restoration of the tissue after ~ 35% PHx is achieved within a timeframe of fewer than seven days. This rapid proliferation surpasses even the robust growth observed in fast-proliferating tissues such as tumors. However, unlike tumors, the liver ceases to grow upon reaching its original mass, determined by cell number^2,3,4. Liver regeneration involves in its initial phase a stage referred to as compensatory cellular hypertrophy (CCH), which results in rapid liver expansion after PHx due to increased hepatocyte size, followed by the phase of robust proliferation^5,6,7. Not surprisingly, regeneration of liver tissue after PHx involves altered metabolism to support this highly dynamic process by engaging in particular anabolic pathways^8,9,10,11. Both cell growth to a larger size and proliferation are dependent on mitochondrial function^12,13.

The de novo pyrimidine synthesis pathway is important for cell proliferation to provide nucleic acid precursors. Not much is known about its role in liver regeneration. We have recently found that cancer cells with dysfunctional mitochondria import these organelles with their DNA payload via horizontal mitochondrial transport^14,15, in order to restore mitochondrial respiration that is critically linked to de novo pyrimidine synthesis, enabling transition through the S-phase of the cell cycle¹⁵. It can be estimated that approximately half a billion cells need to form in the mouse liver within the span of < 7 days following ~35% PHx. Consequently, we postulated that such robust cell growth necessitates the engagement of de novo pyrimidine synthesis for rapid transition through the cell cycle.

We show here the critical role of de novo pyrimidine synthesis in liver regeneration linked to rapid metabolic remodeling, whereby detoxification of ammonia in the liver is suppressed in favor of the use of this toxic product to support anabolic pathways epitomized by the de novo pyrimidine pathway.

Results

Inhibition of dihydroorotate dehydrogenase suppresses liver regeneration

Our recent research showed that de novo pyrimidine synthesis (see scheme in Fig. 1a) is of importance for tumor formation that cancer cells devoid of mitochondrial DNA (mtDNA) import mitochondria with DNA from the surrounding stroma to restore respiration needed to drive conversion of dihydroorotate (DHO) to orotate. This is catalyzed by dihydroorotate dehydrogenase (DHODH), an enzyme coupled to the mitochondrial electron redox chain¹⁶. Since it takes 5–7 days to fully regenerate liver after removal of the left lateral lobe (Fig. 1b), which equates to about ~ 35% PHx¹⁷, we anticipated that generation of the high number of cells requires a switch from homeostasis to metabolic pathways that support rapid cell proliferation required for fast transition via the cell cycle, supported by de novo pyrimidine synthesis. It has been shown that inhibition of DHODH, using the specific inhibitor BAY-2402234 that suppresses activity of the enzyme with an IC₅₀ of 3–5 nM¹⁸, results in suppression of tumor growth¹⁹. We therefore tested the effect of the inhibitor on liver regeneration following ~ 35% PHx (Fig. 1a, b). Figure 1c shows that the loss in body weight after PHx of about 1 g (assuming mouse weight of ~ 20 g) was followed by recovery to near that of control mice within 5 days. On the other hand, BAY-2402234 given to mice post-PHx by daily gavage (2 or 5 mg/kg) resulted in a complete halt in body weight increase. This is further corroborated by the liver to body weight ratio (LBWR), where the full restoration of liver in mice at 7 days is blocked by the DHODH inhibitor (Fig. 1d). Of note, treated mice exhibited similar liver status parameters compared to untreated mice 3–14 days after ~ 35% PHx (Supplementary Fig. 1) and had a 100% survival rate (Supplementary Fig. 2), suggesting that treatment with BAY-2402234 following 35% PHx is well tolerated and does not cause toxicity¹⁸.

Nat Commun. 2025 Nov 4;16(1):9664: Fig. 1: Liver regeneration and hepatocyte proliferation are suppressed by a DHODH inhibitor.

a Scheme of the de novo pyrimidine pathway illustrating the precursor substrates and respective enzymes to make the final product uridine-5-monophosphate (UMP), which can be inhibited by BAY-2402234 (BAY) specifically targeting DHODH in the mitochondria. b Mouse liver subjected to partial hepatectomy (PHx) showing removal of the left lateral lobe (~35% of the whole liver) with full regeneration of the remaining liver occurring within 7 days. Relative body weight (c) and liver/bodyweight ratio (LBWR, %) (d) of control mice and mice subjected to ~ 35% PHx with or without daily gavage of BAY. Assessment of proliferation using western blot (WB) of MCM2, PCNA, and pHH3 (e), and using the EdU assay (f) in control and ~35% PHx mice with or without daily gavage of BAY at the indicated timepoints. g A section of regenerating liver was imaged for E-cadherin (red) to demarcate the periportal region surrounding the portal vein (PV), GS (green) to stain the pericentral cells surrounding the central vein (CV), and EdU-positive cells (white) showing spatial localization of proliferating cells encompassing periportal to pericentral distribution within the liver (inset). For panel (c), multiple t tests using the Holm-Sidak method were used (control n = 4, PHx n = 8, PHx + 2 mg/kg BAY n = 4, PHx + 5 mg/kg BAY n = 4; p-values for control vs. PHx at days 1, 2, and 3 post-PHx are 0.0003, 0.0025, and 0.0122 respectively; p-values for PHx vs. PHx+2 mg/kg BAY at days 2, 3, 4, 5, 6 and 7 post-PHx are 0.0362, 0.0075, 0.0057, 0.0009, 0.0003, and 0.0061, respectively; p-values for PHx vs. PHx + 5 mg/kg BAY at days 2, 3, 4, 5, 6 and 7 post-PHx are 0.0370, 0.0370, 0.0085, 0.0264, 0.0040, and 0.0018, respectively). For panels (d and f), unpaired t test was used (d: control n = 15, day 0 post-PHx n = 4, day 3 post-PHx n = 15, day 3 post-PHx + 2 mg/kg BAY n = 10, day 7 post-PHx n = 15, day 7 post-PHx + 2 mg/kg BAY n = 4; p-values for control vs. post-PHx at days 0, 3, 3 + 2 mg/kg BAY, and 7 + 2 mg/kg BAY are < 0.0001, 0.0067, < 0.0001, and 0.0002, respectively; p-values for post-PHx day 3 vs. day 3 + 2 mg/kg BAY and post-PHx day 7 vs. day 7 + 2 mg/kg BAY are 0.0036 and 0.0035, respectively; (f) control n = 11, day 3 post-PHx n = 9, day 3 post-PHx + 2 mg/kg BAY n = 10, day 7 post-PHx n = 11, day 7 post-PHx + 2 mg/kg BAY n = 4; p-values for control vs. post-PHx days 3, 3 + 2 mg/kg BAY, 7, 7 + 2 mg/kg BAY are < 0.0001, 0.0007, < 0.0001, and < 0.0001, respectively; p-values for post-PHx day 3 vs. post-PHx day 3 + 2 mg/kg BAY, day 7, day 7 + 2 mg/kg BAY are < 0.0001; p-values for post-PHx day 3 + 2 mg/kg BAY vs. post-PHx day 7 and day 7 + 2 mg/kg BAY are 0.0017 and 0.0038, respectively). Images presented are representative of at least three independent biological experiments. Data are expressed as mean ± S.D. The symbols ‘*/+’, ‘**/++’, ‘***/+++’, and ‘****/++++’ denote statistical significance at p < 0 .05, p < 0.01, p < 0.001, and p < 0.0001, respectively; ns, not significant. Source data are provided as a Source Data File.

De novo pyrimidine pathway is required for liver regeneration

We next investigated whether BAY-2402234 suppresses de novo pyrimidine synthesis (see scheme in Fig. 1a) in the context of liver regeneration. We first assessed the effect of the inhibitor on DHODH-dependent respiration (Fig. 2a) and DHODH activity (Fig. 2b), and found that suppression of both parameters considerably lowered the orotate-to-DHO ratio (Fig. 2c).

Nat Commun. 2025 Nov 4;16(1):9664: Fig. 2: Liver regeneration requires de novo pyrimidine synthesis.

Assessment of DHODH-dependent respiration (a), DHODH activity (b), and orotate-to-DHO ratio (c) in control mice and mice subjected to ~ 35% PHx with or without daily gavage of 2 mg/kg BAY-2402234 (BAY). LC-MS data showing levels of de novo pyrimidine metabolites glutamine (Gln), carbamoyl aspartate (CA), dihydroorotate (DHO), uridine monophosphate (UMP), their MALDI-TOF representative images with intensity quantification (d), and their MALDI-TOF spatial localization including glutamate (Glu) and aspartate (Asp) (e) demarcated by pericentral (C) GS staining (green), and periportal (P) taurocholic acid (TCA) accumulation in liver of control mice and mice subjected to ~ 35% PHx with or without daily gavage of 2 mg/kg BAY. MALDI-TOF images were acquired with the 9AA matrix and measured in the negative mode. Images are displayed with TIC normalization, 99% quantile hotspot removal, linear interpolation and weak denoising level. Peak intensities are rescaled to the full color map. For panels (a–d), unpaired t test was used, and data are expressed as mean ± S.D. (a: all groups have n = 4; control vs. day 3 post-PHx + 2 mg/kg BAY p < 0.0001; control vs. day 7 post-PHx + 2 mg/kg BAY p = 0.0003; day 3 post-PHx vs. day 3 post-PHx + 2 mg/kg BAY p < 0.0001; day 7 post-PHx vs. day 7 post-PHx + 2 mg/kg BAY p = 0.0028; b: all groups have n = 4; control vs. day 3 post-PHx + 2 mg/kg BAY p < 0.0001; control vs. day 7 post-PHx + 2 mg/kg BAY p < 0.0001; day 3 post-PHx vs. day 3 post-PHx + 2 mg/kg BAY p < 0.0001; day 7 post-PHx vs. day 7 post-PHx + 2 mg/kg BAY p = 0.0003; c: control, post-PHx days 3, 7 and 14 have n = 4; day 3 post-PHx + 2 mg/kg BAY has n = 5; day 3 post-PHx + 2 mg/kg BAY vs. control p = 0.0042; day 3 post-PHx + 2 mg/kg BAY vs. day 3 post-PHx p = 0.0034; day 3 post-PHx + 2 mg/kg BAY vs. day 7 post-PHx p = 0.0418; day 3 post-PHx + 2 mg/kg BAY vs. day 14 post-PHx p = 0.0319; d, Gln: n = 5 for all groups; control vs. day 3 post-PHx p = 0.0004; control vs. day 7 post-PHx p = 0.0476; d, CA: n = 4 for all groups; control vs. day 7 post-PHx p = 0.0004; control vs. day 3 post-PHx + 2 mg/kg BAY p = 0.0044; d, DHO: n = 4 for all groups; control vs. day 7 post-PHx p = 0.013; control vs. day 3 post-PHx + 2 mg/kg BAY p = 0.0011; d, UMP: n = 4 for all groups; control vs. day 3 post-PHx p = 0.0035; boxplot in d: midline depicts median value; the lower and upper hinges correspond to the first and third quartiles, representing the 25^th and 75^th percentiles respectively; the upper whisker extends from the hinge to the largest value no further than 1.5 * IQR from the hinge; the lower whisker extends from the hinge to the smallest value at most 1.5 * IQR of the hinge). The symbols ‘*/+’, ‘**/++’, ‘***/+++’, and ‘****/++++’ denote statistical significance at p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively. Source data are provided as a Source Data File.

Methods

MALDI imaging measurement

MALDI-TOF spectra of 3 biological replicates for a control (Ctrl) group, 3 days post ~35% PHx (D3-PHx) group, and 3 days post ~ 35% PHx with BAY-2402234 treatment (D3-PHx + B) group (9 in total) were acquired in sets of 3 (representing an experimental cohort where each group is represented per acquisition) using the rapifleX MALDI-TOF/TOF mass spectrometer (Bruker Daltonics) operated by the fleXcontrol v4.2 software (Bruker Daltonics) in the reflector negative-ion mode with the 355 nm SmartBeam™ 3D laser (spatial resolution of 50 × 50 µm, m/z range of 20–1000) at the constant laser fluence of 88% and laser frequency of 5 kHz. 200 shots were summed up from each position. The instruments were set up accordingly: ion source 1, 19.973 kV; PIE, 2.664 kV; lens, 11.353 kV; reflector 1, 20.810 kV; reflector 2, 1.034 kV; reflector 3, 8.577 kV. The pulsed ion extraction time was set to 90 ns, detector gain was 2302 V. Data were acquired with a digitizer speed at 2.5 GS/s. Calibration was done externally using red phosphorus to achieve precision up to 2 ppm⁴³. Spatial navigation for imaging data acquisition was done in the flexImaging v6.0 software (Bruker Daltonics).

MALDI-FTICR imaging experiments on a representative biological sample per experimental group (3 in total) from mice treated with ¹⁵NH₄Cl were conducted in the negative ion mode using the MALDI-solariX 12T-2ω Fourier transform ion cyclotron resonance (FTICR) mass spectrometer (Bruker Daltonics) equipped with the Smartbeam II 2 kHz laser, and operated by the ftmsControl v2.3.0 software. The laser focus was selected to achieve a lateral resolution of 35 × 35 µm by summing up 100 laser shots at the 1 kHz laser frequency and constant laser power. The instrument parameters were adjusted to maximize ion intensities of compounds of interest, considering the sample preparation, including the MALDI matrix. First, for samples overlaid with the DAN matrix, data were collected in the mass range of 40–650 m/z with one acquired spectrum per pixel. The data were collected with 2 M data points in the transient (0.21 s), which provided the estimated resolving power of 48,000 at m/z of 400. The ion optic parameters were optimized to maximize ion transmission within the defined m/z range, including the funnel RF amplitude (80 Vpp), collision cell (collision voltage: 1.5 V, DC bias: -1.0 V), time-of-flight delay (0.4 ms), and transfer optics (6 MHz, Q1 m/z 100). Second, the MALDI MSI method used for the samples prepared with the 9-AA matrix was adjusted as follows. MS data were acquired in the mass range of 100–650 m/z and continuous accumulation of selected ions mode, allowing for a selection of multiple mass windows, including one isolation window of 70 Da with the Q1 mass of 130 m/z, and the second isolation window of 300 Da with the Q1 mass of 380 m/z. The parameters of ion transfer optics were kept the same as used in the above-mentioned MALDI-FTICR MSI method except for the Q1 mass. Data were acquired with 2 M data points in the transient (0.56 s), providing an estimated resolving power of 130,000 at m/z 400. Both MSI methods were externally calibrated on clusters of red phosphorus prior to data acquisition, achieving mass accuracy better than 1 ppm⁴³. Spatial navigation for imaging data acquisition was set in the flexImaging v5.0 software (Bruker).

LC-MS and LC-MS data processing

A Q-Exactive orbitrap mass spectrometer with an Ion Max source and HESI II probe attached to the Vanquish Horizon UHLPC system was used to evaluate polar metabolites. The liquid chromatography – mass spectrometry (LC-MS) system underwent weekly cleaning and calibration with positive and negative Pierce ESI Ion Calibration Calmix (Thermo Scientific). To prepare samples for polar metabolite extraction, snap-frozen liver samples from 3 to 5 biological replicates after ~35% PHx were pulverized, and ~ 10 mg of tissue was vortexed in 60% cold methanol (− 20 °C) for 10 min. 500 μl of cold chloroform (− 20 °C) was added to the mixture and vortexed for 10 min. The solution was centrifuged at maximum speed for 10 min at 4 °C, and the top layer (polar fraction) was collected for polar metabolite analysis. 2 μl of the sample adjusted to 2 μg protein was run through a SeQuant ZIC-pHILIC 5 μm 150 × 2.1 mm analytical column (Sigma) with a 2.1 × 20 mm guard column (Sigma) attached to the front end. The column oven was set to 25 °C, and the autosampler was set to 4 °C. Buffer A comprised 20 mM ammonium carbonate (Sigma) and 0.1% ammonium hydroxide (Sigma) in HPLC-grade water (Sigma), and buffer B compromised 100% acetonitrile (Sigma). The liquid chromatography was set to the flow rate of 0.15 ml/min. The metabolites were separated using the linear gradient from 80% buffer B to 20% buffer B over the course of 20 min, followed by an increasing gradient from 20% buffer B to 80% buffer B for 0.5 min, and by isocratic flow at 80% buffer B for 7.5 min. The mass spectrometer was set to full scan, polarity switching mode, with the spray voltage set to 4.0 kV, heated capillary to 350 °C, and the HESI probe to 30 °C. The MS data were acquired in the mass range of 70–1000 m/z. The sheath gas flow was set to 10 units, auxiliary gas to 1 unit, and sweep gas flow to 1 unit. Resolution of scans was set to 70,000, AGC target to 10⁶, and maximum injection time to 20 ms. An additional scan between 220 and 700 m/z was used to enhance nucleotide detection in the negative mode, together with the maximum injection time set to 80 ms.