Mapping early serum proteome signatures of liver regeneration in living donor liver transplant cases

The liver is the only solid organ capable of regenerating itself to regain 100% of its mass and function after liver injury and/or partial hepatectomy (PH). This exceptional property represents a therapeutic opportunity for severe liver disease patients. However, liver regeneration (LR) might fail due to poorly understood causes. Here, we have investigated the regulation of liver proteome and phosphoproteome at a short time after PH (9 h), to depict a detailed mechanistic background of the early LR phase. Furthermore, we analyzed the dynamic changes of the serum proteome and metabolome of healthy living donor liver transplant (LDLT) donors at different time points after surgery. The molecular profiles from both analyses were then correlated. Insulin and FXR‐FGF15/19 signaling were stimulated in mouse liver after PH, leading to the activation of the main intermediary kinases (AKT and ERK). Besides, inhibition of the hippo pathway led to an increased expression of its target genes and of one of its intermediary proteins (14‐3‐3 protein), contributing to cell proliferation. In association with these processes, metabolic reprogramming coupled to enhanced mitochondrial activity cope for the energy and biosynthetic requirements of LR. In human serum of LDLT donors, we identified 56 proteins and 13 metabolites statistically differential which recapitulate some of the main cellular processes orchestrating LR in its early phase. These results provide mechanisms and protein mediators of LR that might prove useful for the follow‐up of the regenerative process in the liver after PH as well as preventing the occurrence of complications associated with liver resection.


| INTRODUCTION
The liver is the only solid organ capable of regenerating itself to recover 100% of its mass and function. 1 Liver regeneration (LR) is a complex process that involves a finely tuned network of molecular events and interaction of different cell types to ensure efficient proliferation until complete restoration of liver physiology. 2 The process of LR is usually divided into three phases known as the priming/initiation phase (early), the progression/maintenance phase, and the termination phase (late), which are regulated by the coordinated action of growth factors and cytokines secreted close to the site of injury or transferred to the liver from blood. 3 There is an intricate mechanistic overlapping across these three phases, which eventually share a common sequence of events making it difficult to determine which pathway specifically regulates which phase and to predict an efficient LR. 4 Liver transplantation (LT) continues to be the only available curative treatment for patients with acute liver failure and end-stage liver disease. 5 Although LT improves the quality of life and long-term results in patients, the incidence of postoperative complications remains high due to drastic pathophysiological changes in the liver. In general, risk factors can be categorized as related to patient comorbidities, primary liver function, and perioperative events. Hepatic regenerative capacity can be adversely affected by certain chemical and drug-induced injuries, viral infections, or chronic inflammation leading to fulminant hepatic failure. Moreover, complications may arise from the clamp of the portal vein during surgery and the concomitant hypo perfusion of organs, being the kidney one of the most affected. Postoperative acute kidney injury (AKI), even when transient, has been associated with poor long-term graft and reduced patient survival. 6 Associated with the surgery, it may also happen that LR is insufficient or not initiated at all in some patients due to excessive liver resection. 7 Small-for-size syndrome usually occurs when the relative graft volume is <30%-40% of the standard liver volume, leading to slow/no recovery of liver function and ultimately graft failure. 8 The identification of factors associated with posthepatectomy liver failure is critical for the selection of appropriate patients to undergo liver resection as well as to elaborate strategies to prevent mortality. 9 Thus, an enhanced understanding of LR is needed for an efficient follow-up and to maximize the rates of clinical success.
Emergent analytical approaches grant for the discovery of new biomolecular markers that may allow for monitoring LR and to detect early adverse events. Most studies have usually focused on factors and pathways occurring within the liver, neglecting the circulating signals in serum that can be measured by noninvasive protocols. Investigating the serum proteome emerges as a powerful resource to define protein signatures in an unbiased manner that can be associated to a specific physiological or pathological condition. 10 While recent studies describe the dynamic changes of plasma metabolites in the progression of LR and search for potential biomarkers, 11 changes in circulating proteins that might reflect the events occurring by the regenerating liver remain to be revealed.
In this study, we have defined early changes at the level of liver proteome and phosphoproteome in a wellestablished murine model of LR after partial hepatectomy (PH). The regulated proteins pointed to a coordinated reprograming of hepatocyte homeostasis associated to the recovery of the liver parenchyma. In parallel, the dynamics of the serum proteome and metabolome were investigated in living donor liver transplant (LDLT) donors at different time points after surgery. Integration of both analyses resulted in the identification of protein and metabolite drivers of LR that may prove useful to predict LR outcomes.

| MATERIALS AND METHODS
An extended version of the methods is provided as Supplementary methods S1.

| Human serum samples
Human samples from healthy liver donors (n = 7) were collected at baseline (preoperative stage, T0) and several time points after liver resection (1H, 12H, 24H, 48H, 72H, 5D, 30D, and 8M). Then, were stored by the Biobank of the Universidad de Navarra, which is integrated in the Spanish National Biobanks Network and they were processed following standard operating procedures with the appropriate approval of the Ethics and Scientific Committees (064/2010 and 2017.146). This study was approved by the human research review committee of the University of Navarra, and was conducted in compliance with the ethical standards formulated in the Helsinki Declaration of 1996 (revised in 2000), upon obtention of the informed consent from all patients. The surgical technique has been previously described. 12 2.2 | PH model LR was induced in mice by 2/3 PH as previously described 13 . Control mice underwent sham operation (SH) that did not involve liver resection. At 9 h after hepatectomy, control mice (SH, n = 5) and hepatectomized mice (PH, n = 5) were sacrificed and liver tissue samples were snap frozen or formalin-fixed and paraffin embedded. All mouse experiments were approved by the Comunidad de Madrid (PROEX 845/2019) and the CNB Ethics Committees in strict accordance with the Spanish and European Union laws and regulations concerning the care and use of laboratory animals.

| RT-qPCR
Total RNA was extracted from mouse liver tissue using TRI-Reagent (Sigma Aldrich). cDNA strands were synthesized from 2 μg total RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), employing random primers. mRNA levels were quantified by qPCR in a Quant-Studio 5 Real-Time PCR (RT qPCR) System (Thermo Fisher) using the SYBR EvaGreen-based reaction mix (5Â PyroTaq EvaGreen qPCR Mix Plus ROX; Cmb-Bioline) and specific primers (Table S1); β-actin or gapdh were used as housekeeping genes for normalization purposes. Relative quantities (i.e., relative to the sample with the lowest expression or the control sample) were calculated using the 2 ÀΔΔCt method.

| SureQuant-targeted quantitation
Total protein amount from serum samples was measured using the Pierce 660 nm protein assay (Thermo Fisher Scientific). Protein digestion using S-Trap filter (Protifi, Huntington, New York) was performed following the protocol described in. 14 Peptide quantification was done with Qubit™ Fluorometric Quantitation (Thermo Fisher Scientific) using spiked in PQ500™ Reference Peptide kit (Biognosys) as reference. Liquid chromatography-tandem mass spectrometry (LC-MS/ MS) analysis was done by reversed-phase chromatographic separations and SureQuant method as indicated in Supplementary methods S1. Data from Orbitrap Exploris 480 MS analysis were processed with the Skyline-daily program (version 20.1.1.196). Total area ratio endogenous/Internal standard was extracted. After no. N/A values filtering out, the statistical analysis was performed by linear modeling. False discovery rate control was used for multiple testing correction similarly as described elsewhere. 15

| Label-free LC-MS/MS shotgun proteomics
Liver tissue was homogenized in urea lysis buffer. Proteins were precipitated with methanol/chloroform/water (4:1:3 vol.), resuspended, and reduced in 7 M urea, 2 M Thiourea, 100 mM TEAB, 5 mM TCEP, pH 8, for 45 min at 37 C. Cysteine thiols were alkylated with 10 mM methyl methanethiosulfonate for 10 min at room temperature. Subsequently, samples were fourfold diluted with TEAB (final concentration: 25 mM). Proteins were then digested with trypsin from porcine pancreas (Sigma-Aldrich) at a final trypsin: protein ratio of 1:10 and digested overnight at 37 C. Tryptic peptides were dried by vacuum centrifugation. For phosphoproteome analysis, the phosphopeptide fraction was obtained from the total protein digest using MOAC affinity chromatography with TiO 2 as described in. 16 One μg of the total peptide digest was analyzed by 1D-nano LC (Eksigent Technologies nanoLC Ultra 1D plus, SCIEX, Foster City, California) coupled to ESI-MS/MS (high-speed Triple TOF 5600 mass spectrometer, SCIEX, Foster City, California) with a Nanospray III source. One μg of the phosphopeptide sample was subjected to 1D-nano LC ESI-MSMS analysis using a nano liquid chromatography system (Ultimate 3000) coupled to Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific).
Bioinformatics analysis was performed with Perseus 1.6.5.0.2 software. 18 For functional enrichment analysis, the list of differentially expressed genes was subjected to functional analysis using ingenuity pathway analysis (IPA) software v.6875226 (Ingenuity ® Systems, www. ingenuity.com).

| Targeted proteomics: Parallel reaction monitoring
To study the one-carbon metabolism (OCM) pathway in mouse liver, the abundance of 12 participating enzymes was determined by parallel reaction monitoring (PRM) as previously described. 19,20 We also measured the levels of PHB1 in the same samples and followed the same protocol. The selection of proteotypic peptides for inclusion in the PRM development combined in-house shotgun proteomics-based experimental data with public information from the PeptideAtlas (https://www.peptideatlas.org; Table S2). Data files were analyzed with Skyline (v21.2.0.568).

| Analysis of metabolites
Serum and tissue samples were processed as indicated in the Supplementary methods S1. The analysis was carried out by randomizing the samples for each platform run. Samples were measured using an Agilent 1290 Infinity II UHPLC system coupled to an Agilent 6460 QqQ-MS (Agilent Technologies) detector in negative ESI mode. For CE-MS analysis, samples were analyzed by a capillary electrophoresis system (7100 Agilent Technologies) coupled to a time-of-flight mass spectrometer (6224 Agilent Technologies) with online preconcentration; the coupling was equipped with an ESI sprayer G1607 from Agilent Technologies.
The data obtained were integrated using MassHunter Quantitative Analysis B.10.0 software (Agilent Technologies). Statistical analysis was done using R, version 4.1.3. Negative and zero values in the data set were imputed using the "missForest" R package. missForest 21 is a library based on machine learning algorithm random forest. 22 Unnormalized data set was used to compute Kruskal-Wallis test. After that, Dunn's post hoc test was calculated to evaluate specific statistical differences between groups. All p-values were adjusted through the Benjamini-Hochberg method. 23 To cluster bile acid (BA) time patterns "mfuzz" 24 package was used applying the elbow method. 25

| Statistical analysis
Data are shown as means ± SEM unless otherwise indicated. The number of replicates was as indicated in the figure legends. Statistical differences were examined using the two-tailed Student's t-test for comparison of two independent groups. Significance was set at p < 0.05, unless otherwise indicated. Calculations were performed using GraphPad Prism 9.4.1. software.

| Data deposition
Proteomics data are available via ProteomeXchange with identifier PXD038316.

| Shotgun proteomic and phosphoproteomic analysis at 9 h upon PH in mice
Aiming to identify protein indicators of LR progression, we have dissected in depth the early events occurring 9 h after PH in mice. To this end, we have done a systematic analysis of liver proteome and phosphoproteome using a label-free shotgun proteomics approach ( Figure S1). We identified 13,160 peptides corresponding to 1443 protein groups, of which 166 were differentially regulated (95 downregulated and 70 upregulated in PH livers; Figure 1A and Table S3). After phosphopeptide enrichment, 3322 phosphopeptides were identified, which represented 1551 phospho-protein. A total of 118 proteins were differentially phosphorylated (44 decreased and 72 increased their phosphorylation in PH mice; Figure 1A and Table S4). The heat map ( Figure 1B) and PCA ( Figure S2) 26 showed a good experimental reproducibility, and a sample clustering according to their corresponding biological condition (SH and PH). Phosphorylation variations were not associated to changes in protein abundance but we found a wide overlapping between the biological processes in which the F I G U R E 1 Shotgun proteomics and phosphoproteomics analysis 9 h after partial hepatectomy. (A) Volcano plot (Àlog10 [p-value] and log2 [fold-change]) of the proteins and phosphoproteins found in liver from partial hepatectomy (PH) mice compared with sham operation (SH) control animals. Proteins and phospho-proteins are represented as gray dots (p-value > 0.05), black dots ( p-value < 0.05), red dots ( pvalue < 0.05; log 2 fold change < À1) and green dots (p-value < 0.05; log 2 fold change > 1). (B) Heat map of differential proteins and phosphopeptides. (C) Venn diagram of both differential molecules and canonical pathways from proteomics and phosphoproteomics data set. (D) Ingenuity pathway analysis of the differentially represented proteins between control (SH) and PH mice identifying the top enriched categories of canonical pathways. X-axis indicates the significance (Àlog p-value) of the functional association that is dependent on the number of genes in a class as well as on the biological relevance. differential proteins and phosphoproteins are involved as would be expected for two complementary regulatory levels of LR ( Figure 1C).
To investigate upstream checkpoints that would explain the differential protein profile in PH livers, we searched for master regulators. The top candidates were peroxisome proliferator-activated receptor alpha (PPARα), p53, and insulin ( Figure S3), which would coordinate the regulation of the cellular processes occurring during early LR, as suggested by IPA enrichment analysis: (1) Cell proliferation (hippo signaling), (2) Metabolic reprogramming (glucose and lipid metabolism, mitochondrial function), and (3) Cell differentiation (intercellular junctions, OCM; Figure 1D).

| Induction of cell proliferation signals 9 h upon PH in mice
Activation of PPARα promotes hepatocyte proliferation and controls BAs synthesis through activation of ERK1/2 and c-Jun. 27 We found a significant increase of ERK1/2 ( Figure 2A) and c-Jun ( Figure 2B) phosphorylation in PH mice suggesting the activation of proliferation pathways. We next analyzed the mRNA levels of the mitogenic fibroblast growth factor (FGF15) target genes, as it mediates the activation of ERK1/2 and c-Jun 28 and regulates BAs production. We observed a significant decrease of Cyp7a1, Cytc, and Idh3a and a decreasing trend of Cyp8b and Pgc1α in PH mice ( Figure 2C). These results indicated a stimulation of FGF15 signaling, whose synthesis in the ileum is activated by BAs, which are essential in the beginning of the regeneration process, as they have promitotic properties. 29 According with the stimulation of FGF15 signaling, we observed a significant increase in the concentration of most BAs in PH mice (Table S7). This is further supported by activation of the hippo-YAP signaling pathway as evidenced by the significant upregulation of all hippo target genes (Areg, Id1, Fgf1, Ccnd1, Ctgf, and Myc) in PH mice ( Figure 2D). Among these genes, Areg can be considered as an early trigger of LR, and could be a promising candidate for supporting therapy in small-for-size transplantation to stimulate liver growth. 30 Besides, 14-3-3 protein zeta/delta (1433Z) levels increased in PH mice ( Figure 2E), which also participates in the regulation of the hippo pathway. 31 However, the strongest effects on the regulation of this pathway result from remodeling of the cytoskeleton, by the action of cell-cell and cell-matrix junction components. 32 In this direction, results from phosphoproteomics data revealed a regulation of intercellular junction pathways such as tight junction, actin cytoskeleton signaling, and adherents junction, through a significant decrease in the phosphorylation of catenin alpha-1 (CTNA1), catenin beta-1 (CTNB1), CAP-Gly domain-containing linker protein 1 (CLIP1), and myosin light chain kinase (MYLK) in PH mice, which participate in tight and adherens junctions ( Figure 2F). These results point to the remodeling of the hepatocyte cytoskeleton, which is crucial for epithelial-mesenchymal transition and proliferation, triggering the entry of quiescent cells into the cell cycle. 33

| Energy metabolism reprogramming at 9 h upon PH in mice
Proliferating cells require metabolic activity to generate energy and molecular intermediates to produce new cells and tissue mass. 34 Proteomics data reflected a higher abundance of glycolytic enzymes such as glucose-6-phosphate isomerase (GPI), aldolase B (ALDOB), phosphoglycerate mutase 1 (PGAM1) in PH mice suggesting the activation of glycolysis ( Figure 3A). The reduction of l-lactate dehydrogenase A chain (LDHA) may indicate that pyruvate is not being converted into lactate, favoring aerobic energy production in the mitochondria instead of the anaerobic lactic acid cytosolic fermentation, which is observed in cancer (Warburg effect). Supporting glucose catabolism activation, we found higher levels of the three liver hexokinase isoforms, although glucokinase ( gk), the hepatocyte-specific isoform, 35 was the only increase statistically significant in PH mice ( Figure 3B). The low hk2 expression might be considered a good indicator of controlled growth as its accumulation has been associated with tumoral phenotypes. 36 The decrease of fructose-1,6-bisphosphatase 1 (FBP) and phosphoenolpyruvate carboxykinase 1 (PCK1 and pck1; Figure 3C,D) suggest a negative regulation of gluconeogenesis, which might explain the drop of blood glucose after PH. 37 Insulin is one of the master regulators in the early phases of LR through AKT-FoxO1 axis activation 9 h after PH, as AKT was significantly phosphorylated in PH livers ( Figure 3E), leading to the repression of FoxO1 target genes ( Figure 3D). Furthermore, we observed a significant increase of insulin receptor substrate-1 (IRS-1) phosphorylation in PH compared with SH mice (Figure 3F), further supporting the activation of this pathway for cell survival and proliferation. Insulin signaling also plays an important role in the regulation of fatty acid (FA) metabolism, underscoring the close coordination between lipid and glucose metabolism in hepatocytes during LR. 38 Fasn (FA synthase) was significantly increased in PH, suggesting stimulation of de novo synthesis of hepatic FAs ( Figure S4A), which are mainly directed to the mitochondrial β-oxidation, as indicated by the increase trifunctional enzyme subunit alpha (HADHA) and trifunctional enzyme subunit beta (HADHB) in the regenerating liver from PH mice ( Figure S4B). Furthermore, isocitrate dehydrogenase (IDH1) and aconitate hydratase 1 (ACO1) were significantly decreased after PH ( Figure S5), which suggest the accumulation of cytosolic citrate that would be available for FAs synthesis via acetyl-CoA and for TCA cycle.
Serine plays an important role in cellular metabolic processes that link glycolysis with lipid synthesis to stimulate cell proliferation. 39 We found upregulation of 3-phosphoglycerate dehydrogenase (PHGDH), phosphoserine aminotransferase 1 (PSAT1), and phosphoserine phosphatase (PSPH) in PH mice ( Figure S6), suggesting an activation of de novo serine biosynthesis. The increase in serine synthesis could be associated to the requirement of proteins of a regenerating liver and is consistent with the significant accumulation (eIF4A1) and phosphorylation (eIF3B, eIF4G1, and eIF4G3) of translation initiation factors in PH mice ( Figure 4B,C), which suggest activation of protein synthesis.
The metabolic alterations in liver proteome suggested that mitochondrial function is activated in early LR ( Figure 1D). This inference is based on a particularly strong upregulation of the five protein complexes of electron transport chain in PH compared with SH control mice ( Figure 5A). However, no significant changes of these complexes were observed at the mRNA levels ( Figure 5B), suggesting that stabilization of their protein components would explain their increase, more than a transcriptional regulatory event. Interestingly, we observed a significant increase of PHB1 (Prohibitin 1) and a tendency to increase of PHB2 (Prohibitin 2) in regenerating liver 9H after PH ( Figure 5C). Prohibitins (PHB1 and PHB2) are proteins that form a ring-shaped complex that is anchored to the inner membrane of the mitochondrion, inhibit m-AAA protease and physically interacts with and stabilizes the respiratory chain complexes promoting mitochondrial super-complex assembly. 40 Remarkably the observed increment on PHB1 was further confirmed by targeted PRM assay ( Figure S9A). The changes in PHB1 would stabilize respiratory complexes contributing to the enhanced mitochondrial activity. 41 Overall, these observations agree with previous studies 42 suggesting that after PH, the remnant liver tissue is highly glycolytic and biosynthetic as demanded by proliferating cells.

| Inhibition of OCM at 9 h upon PH in mice
Shotgun proteomics analysis indicated a significant decrease of methionine metabolism enzymes (MAT1A, GNMT, and AHCY) and trans-sulfuration pathway enzyme (CBS) after PH ( Figure 6B). Since OCM is essential in maintaining quiescent and differentiated hepatocytes, [43][44][45][46] and its regulation during LR is poorly understood, we have performed a systematic analysis of 12 enzymes involved in this pathway. To this end we used a standardized PRM assay targeting enzymes catalyzing reactions of the methionine cycle (MAT1A, MAT2A, MAT2B, GNMT, AHCY, BHMT, and MTAP), the folate cycle (SHMT1 and 2, DHFR), and the transsulfuration pathways (CTH and CBS). 19 A significant F I G U R E 4 Increase in abundance and phosphorylation of translation initiation complex proteins 9 h after partial hepatectomy. (A) Interaction network of principal components of the eIF4F complex using STRING software. (B) Normalized labelfree quantification intensity of IF4A1 in liver tissue from sham operation (SH) and partial hepatectomy (PH) mice (n = 3). (C) Normalized label-free quantification intensity of phospho-eIF3B, phospho-eIF4G1, and phospho-eIF4G3 in liver tissue from SH and PH mice (n = 3). Data shown as the mean ± SEM from at least three independent experiments. *p < 0.05; two-tailed Student's t-test. reduction of MAT1A, GNMT, CBS, and SHMT1 ( Figure 6C) and a tendency to be downregulated for the other enzymes was observed in PH mice. Downregulation of MAT1A would facilitate hepatocytes proliferation, whereas there was no accumulation of MAT2A and MAT2B in the regenerating liver ( Figure 6C), contrary to what it is observed in hepatocellular carcinoma (HCC) where MAT2A and B accumulation are considered of bad prognosis. 47 In parallel, the RNA levels of these enzymes were also measured to investigate their transcriptional regulation and showed a similar trend with one exception ( Figure 6D). GNMT downregulation would contribute to regulate OCM flux within hepatocytes and therefore, their proliferation, differentiation, and response to druginduced injury. 48 It is worth noting that mRNA levels of Gnmt remain unaltered in PH livers ( Figure 6D), which suggests that the regulation of GNMT 9 h after PH occurs posttranslationally. Interestingly, it has been reported that Mir-873-5p downregulates GNMT, which contribute to fibrosis progression in the cholestatic liver. 49 Regarding the trans-sulfuration pathway, CBS was downregulated in PH mice. Its function in LR is not known, but its regulation may be related to the levels of homocysteine, which are predicted to be diminished due to downregulation of AHCY. Further, the decrease of SHMT suggests inactivation of the folate cycle. Finally, the measurement of two central metabolites in this cycle (S-Adenosyl-Lhomocysteine, SAH and S-Adenosyl-L-methionine, SAM) confirmed the downregulation of OCM during LR, as we observed a marked decrease in both SAH and SAM in PH mice (Table S7). Thus, we show here data supporting OCM reprogramming, which might contribute to switch hepatocytes from a quiescent to a controlled proliferative configuration and support the value of the systematic measurement of OCM to assess the physiological status of the liver and would allow for discrimination among a controlled regenerative process and a tumor progression.

| Serum proteomic and metabolomic profile of LDLT donors recapitulate the main cellular processes observed in PH murine model
Finally, we wanted to investigate whether the cellular reprogramming underlying the early phase of LR F I G U R E 5 Increase of mitochondrial proteins 9 h after partial hepatectomy. (A) Normalized label-free quantification intensity of mitochondrial proteins in liver tissue from sham operation (SH) and partial hepatectomy (PH) mice (n = 3). (B) Relative quantification of mRNA levels of some mitochondrial respiratory complexes, normalized to those of GAPDH, in liver tissue from SH and PH mice (n = 3). (C-E) Normalized label-free quantification intensity of prohibitins (C), VDAC (D), and PRDX5 (E) in samples in (B). Data shown as the mean ± SEM from at least three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001; two-tailed Student's t-test. observed in mice after PH could be reflected in circulating serum proteins in humans. To address this question, the dynamic changes of the serum proteome from LDLT healthy liver donors (n = 7) at baseline (preoperative stage, t = 0) and several time points (1H, 12H, 24H, 48H, 72H, 5D, 30D, and 8 M) after liver resection were analyzed using the SureQuant™ ( Figure 7A). Overall, 540 peptides were detected, corresponding to 353 identified and quantified proteins, of which 56 proteins were regulated along the regeneration process (Table S5). Most proteins were liver-specific as assessed with Metascape software (Figure 7B), suggesting their association with LR. Moreover, they were grouped according to their dynamic regulation into four main clusters ( Figures 7C  and S7) which provided a sequential organization of the enriched functional processes in which they are involved: innate immune response, glucose/lipid metabolism, coagulation system, cell growth/proliferation, and mitochondrial function ( Figure 7D). It is worth noting that the functional outcomes observed previously in the PH murine model match with the F I G U R E 6 Reprogramming of one-carbon metabolism (OCM) 9 h after partial hepatectomy. (A) Scheme of the OCM and transsulfuration pathway; the reactions catalyzed by the enzymes monitored are indicated. (B) Normalized label-free quantification intensity of 1CM enzymes in liver tissue from sham operation (SH) and partial hepatectomy (PH) mice (n = 3). (C) Quantification of OCM enzymes by the multiple-reaction monitoring method in liver tissue from SH and PH mice (n = 3). All proteins were detected across the analyzed samples, and significant differences were observed as indicated. (D) Relative quantification of mRNA levels of OCM enzymes, normalized to those of β-Actin, in liver tissue from SH and PH mice (n = 3) Data shown as the mean ± SEM from at least three independent experiments. *p < 0.05, **p < 0.01, ****p < 0.0001; two-tailed Student's t-test. human serum data, suggesting that serum profiling provides a reliable readout of liver function during regeneration.
The first cluster consisted of proteins that showed significant upregulation 1 h after liver hepatectomy ( Figure S7A) and included enzymes processing glycolytic intermediates steps (phosphoglycerate mutase 1, PGAM1; alpha-enolase, ENOA; all-trans-retinol dehydrogenase, ADH4; l-lactate dehydrogenase A chain, LDHA) suggesting the activation of glycolysis as the main energy source to support proliferation. Also included the increase of stress proteins, such as protein disulfide-isomerase (PDIA1) and heat shock cognate 71 kDa protein (HSP7C), might serve as QC to prevent misfolding and aggregation of newly synthetized proteins. As a readout of early inflammation, the raise of moesin (MOES) suggests the activation of circulating T cells. 50 The increase of MYG that is cleared by the kidney agrees with previous observations, 51 and points to this protein as a marker to increase the accuracy of predicting the occurrence of post-LT acute kidney injury. 6 The second cluster includes proteins that peak 12 h after PH ( Figure S7B) and participates in acute phase response (Scavenger receptor cysteine-rich type 1 protein M130, C163A), proliferation (14-3-3 protein epsilon, 1433E; 14-3-3 protein zeta/delta, 1433Z; and catalase-A, CATA), extracellular matrix remodeling (matrix metalloproteinase-9, MMP9; osteopontin, OSTP), hepatic inflammation and liver injury (SPARC-like protein 1 [SPRL1]). 52 According to the high-energy demands of the regenerating liver, there was an increase in creatine kinase M-type (KCRM), and in fructose-bisphosphate aldolase B (ALDOB) that play a central role in energy transduction in tissues with large energy demands.
The third cluster integrates proteins peaking at 24-48 h ( Figure S7C) and their main biological functions are mostly related to those regulated during the first hour of LR, such as innate immune response and inflammation (macrophage colony-stimulating factor 1, CSF1; lipopolysaccharide-binding protein, LBP; c-reactive protein, CRP; serum amyloid A-1 protein, SAA1; serum amyloid A-2 protein, SAA2), cytoskeletal remodeling (calpain-1 catalytic subunit, CAN1), and glycolysis (fructose-bisphosphate aldolase ALDOA; isocitrate dehydrogenase [NADP] cytoplasmic, IDHC). ALDOA has the lowest K m for F1,6BP catalysis compared with ALDOB and ALDOC, so it may represent a major push to glycolytic flux and energy production in regenerating hepatocytes. 53 The last cluster of proteins increased 72 h after the initiation of the LR process ( Figure S7D). They participate in stimulation of plasma coagulation through platelet aggregation (platelet factor 4, PLF4; and von Willebrand factor) and, conversely, preventing coagulation (vitamin K-dependent protein S, PROS). This balance between procoagulant and anticoagulant proteins might mediate a wound-healing response after PH, while promoting cell proliferation and liver parenchyma growth. 54 In line with this idea, the increase of PROS in serum after hepatectomy, indicates the stimulation of cell proliferation through TAM receptors signaling. 55 At this time point we also observed an increase of misfolded protein repair (transitional endoplasmic reticulum ATPase, TERA); matrix extracellular remodeling (tenascin, TENA; bone marrow proteoglycan, PRG2; cellular redox signaling [thioredoxin, THIO] and mitochondrial function [prohibitin, PHB]). Since changes in PHB1 levels in serum could have translational relevance and might be a good reporter for mitochondrial activation and for efficient starting of LR, we measured its levels by enzyme-linked immunosorbent assay (ELISA) to confirm the result from SureQuant analysis. A significant increase of PHB1 was observed in LDLT donors at 72 h after liver donation ( Figure S9B).
The decrease of IGFALS, which binds and stabilizes insulin-like growth factors (IGFs) in serum, may increase the bioavailability of IGFs, as described elsewhere, thereby modulating their metabolic activities. 56 There is ample evidence supporting that insulin is necessary for LR in the early period exerting strong effects on hepatocyte growth and metabolism. 57 Moreover, continuous intraportal insulin perfusion after LDLT may promote early (1 week) LR and functional recovery after transplant. 58 Here, we propose the measurement of IGFALS levels as a readout of insulin signaling.
Finally, we also analyzed serum metabolites such as amino acids (glycine and serine) and BAs including nonconjugated, Gly-conjugated, and Tau-conjugated BAs along LR up to 30 days postresection. Most significant changes occurred at 72 h (Table S6). The increase of both serine and glycine, could be used as efficient LR hallmarks, since it has been related to improved survival rates, stimulated cell proliferation, and reduced reperfusion injury after partial liver donation. 39,59 Interestingly, after 30 days the initial levels of both amino acids are recovered, as would be expected when the regenerative process has been completed ( Figure S8A). Regarding BAs, their metabolism is directly involved in the process of LR, as previously described for our PH mice model, in agreement with other studies. 60 All monitored BAs were grouped into two main clusters according with their dynamics ( Figure S8B). It is especially important to note that only the BAs from Cluster no. 1 showed statistically significant changes. This cluster is composed mainly by cholic acid and their Gly-conjugated and Tau-conjugated forms, whose concentration was significantly increased 72 h postresection and returned to preoperative levels by Day 30. This correlates with previous work where most BAs returned to initial values 10 days after PH. 61 The combination of coherent proteomics and metabolomics data, which align with the cellular events driving LR in hepatocytes, pave the way to develop noninvasive follow-up strategies for individuals involved in liverresection treatments.

| CONCLUSIONS
In conclusion, we have defined a protein and metabolite profile in the serum of LDLT donors that resumes many of the cellular processes undergone during LR after PH, pointing to biomolecules that open new perspectives for the clinical management of the patients. A metabolic reprogramming resulting from insulin signaling ensures energy production and the synthesis of biomolecules to support hepatocyte proliferation. Activation of mitochondrial respiration mediated by prohibitin activity, plays a central role in these processes. The rise of PHB1 in human serum after PH might be a promising sensor of mitochondrial activation and efficient LR in LDLT cases. Downregulation of OCM enzymes in the early phase of regenerating liver might facilitate hepatocyte growth. The lack of activation of MAT2A and B, which are bad prognostic factors for HCC, indicates regulated cell proliferation during LR and not uncontrolled cell growth. In this work, we have been F I G U R E 8 (A) Workflow for human serum samples (n = 7) and mouse liver tissue (n = 5) analysis using LC-MS/MS. Bioinformatics analysis was performed using Skyline-daily, MaxQuant, and Perseus. Statistics were performed by Proteobotics SL and GraphPad Prism. Functional integration was done by ingenuity pathway analysis (IPA) and STRING softwares. (B) Schematic representation of proteins and biological processes altered in the early phase of liver regeneration. BAs, bile acids; FAs, formic acids. defined a panel of serum proteins (IGFALS, PHB1, and 1433Z) and metabolites (BAs, serine, and glycine) that resumes many of the cellular processes undergone in the liver to allow regeneration after PH (Figure 8). Noninvasive monitoring of these molecules in the serum of patients that require liver resection could allow the follow-up of the regenerative process in the liver as well as preventing the occurrence of complications associated with liver resection and open new avenues to develop alternative or complementary therapeutic strategies.