Glutamine/glutamate metabolism rewiring in reprogrammed human hepatocyte-like cells

Human dermal fibroblasts can be reprogrammed into hepatocyte-like (HEP-L) cells by the expression of a set of transcription factors. Yet, the metabolic rewiring suffered by reprogrammed fibroblasts remains largely unknown. Here we report, using stable isotope-resolved metabolic analysis in combination with metabolomic-lipidomic approaches that HEP-L cells mirrors glutamine/glutamate metabolism in primary cultured human hepatocytes that is very different from parental human fibroblasts. HEP-L cells diverge glutamine from multiple metabolic pathways into deamidation and glutamate secretion, just like periportal hepatocytes do. Exceptionally, glutamine contribution to lipogenic acetyl-CoA through reductive carboxylation is increased in HEP-L cells, recapitulating that of primary cultured human hepatocytes. These changes can be explained by transcriptomic rearrangements of genes involved in glutamine/glutamate metabolism. Although metabolic changes in HEP-L cells are in line with reprogramming towards the hepatocyte lineage, our conclusions are limited by the fact that HEP-L cells generated do not display a complete mature phenotype. Nevertheless, our findings are the first to characterize metabolic adaptation in HEP-L cells that could ultimately be targeted to improve fibroblasts direct reprogramming to HEP-L cells.

metabolites was performed using HMDB database (https://www.hmdb.ca) with an accuracy of 10ppm. The identity of glutamic acid (exact mass: 147.0531) was unequivocally confirmed using a chemical standard. Differential glutamate concentration in cell media was validated in a new set of experiments (data not shown). In parallel, we estimated the rates of glutamine and glutamate uptake/secretion by HDF and HEP-L cells normalizing per mg of protein since HEP-L cells are 3.8-fold bigger and contain 4.1-fold more protein per cell ( Supplementary Fig. S2). Glutamine consumption was 2-fold higher in HEP-L cells reaching a value comparable to primary cultured human hepatocytes (PHH) (Fig. 2B). In parallel, glutamate utilization changed dramatically. While HDF were net consumers of glutamate from cell media (HMM media contains glutamate), HEP-L cells, as well as PHH, were net producers (Fig. 2C). These changes run in parallel to profound transcriptomic remodeling of genes involved in glutamine/glutamate metabolism and transport, such as those involved in deamidation  Table S6. N = 7. **p < 0.005; ***p < 0.001.
Glutamate flux reversal in HEP-L cells. Net production of glutamate in cell media from HEP-L cells can only be explained by de novo synthesis and secretion. Glutamine is the main cellular source of glutamate. The fate of the carbon and nitrogen moiety of glutamine present in the media was established by incubation with [U-13C, 15 N] glutamine ( 13 C 5 15 N 2 -Gln). 13 C 5 15 N 2 -Gln is transported into cells, where it is deamidated to 13 C 5 15 N 1 -glutamate ( 13 C 5 15 N-Glu; M6) by glutaminases (Fig. 3A). Non-proteinogenic glutamate is largely converted to α-ketoglutarate via either glutamate dehydrogenase or transaminases. We found accumulation of 13 C 5 15 N-Glu, 13 C 5 -Glu (M5; originated from transamination of 13 C 5 -α-ketoglutarate (α-KG) with unlabeled nitrogen) and 15 N-Glu (M1; originated from transamination of unlabeled α-KG with glutamine-derived 15 N) in the cell media from HEP-L cells and primary cultured human hepatocytes, but not from HDF ( Fig. 3B; Table S2). Accumulation of glutamate in cell media was corroborated by 13 C-NMR of replicate experiments ( Supplementary Fig. S3). As expected, such difference cannot be attributed to a lack of glutaminase activity in HDF ( Supplementary Fig. S4), although, de novo expression of phosphate-activated glutaminase (GLS2) at levels comparable to PHH could explain this change (Fig. 3C).
Interestingly, unlabeled glutamate present in 13 C 5 15 N 2 -Gln tracing media was consumed by HDF, but secreted by HEP-L cells (Fig. 3D), indicating the novo generation of glutamate from non-glutamine sources. Glutamate can also be synthesized from glucose through glycolysis coupled to the TCA cycle (Fig. 3E). We found a significant accumulation of 13 C 2 -Glu and 13 C 3 -Glu in cell media from HEP-L cells, but not from HDF incubated with 13 C 6 -glucose ( 13 C 6 -Glc; Fig. 3F). By combining the data collected from incubations with 13 C 5 15 N 2 -Gln and 13 C 6 -Glc, we estimated that de novo synthesized and secreted glutamate originates from glutamine (71%) and glucose (15%). Contribution from non-glutamine non-glucose sources may account for another 15% approximately (Fig. 3G). We observed a low capacity to produce glutamine by glutamate-ammonia ligation catalyzed by glutamine synthetase (GLUL) in HEP-L cells (Fig. 3H). A metabolic feature shared with PHH. In parallel, GLUL mRNA levels in HEP-L cells were reduced 3-fold, similar to PHH levels (Fig. 3I).
Glutamine utilization is reduced in HEP-L cells. Glutamine metabolic pathways are summarized in Fig. 4A. To assess glutamine´s contribution to cellular protein, we quantified 13 C isotopic enrichment in methanol precipitated cellular protein using combustion/isotope-ratio mass spectrometry (C/IRMS). HDF incorporated 70% more carbon-13 when incubated for 24 hours with 13 C 5 15 N 2 -Gln (Fig. 4B). Since, HDF and HEP-L cells are quiescent ( Supplementary Fig. S5), this result cannot be attributed to higher proliferation rates. This result was not totally unexpected, since 58% of glutamine uptaken by HEP-L cells is secreted back to the media as glutamate. Thus, only 42% of glutamine (1.37 micromol/mg of protein*day compared to 1.79 micromol/mg of protein*day in HDF) is actually used in HEP-L cells for cellular metabolic processes apart from glutamine to glutamate deamidation and secretion.
De novo synthesis of proline, a major aminoacid in extracellular matrix proteins, was 14-fold higher in HDF (Fig. 4C,D). Upregulation of the catabolic enzymes ALDH4A1 (Δ 1 -pyrroline-5-carboxylate dehydrogenase; P5CDH) and PRODH (proline dehydrogenase) in HEP-L cells can explain this metabolic difference (Fig. 4E). This transcriptional change also explains the low production of labelled ornithine in HEP-L cells, given that both enzymes are shared in proline and ornithine synthesis pathway (Fig. 4F). Enhanced proline and ornithine catabolism in HEP-L cells are probably accounting for the presence of 15% non-glutamine, non-glucose derived glutamate in HEP-L cells supernatant 14 , since HMM media contains unlabeled ornithine and proline (100 and 30 mg/L respectively).

Reductive glutamine metabolism is increased in HEP-L cells.
Glutamine can contribute carbon to lipogenic acetyl-CoA through glutaminolysis and reductive carboxylation. In glutaminolysis, 13 C 5 15 N 2 -Gln-derived 13 C 5 -α-ketoglutarate generates 13 C 4 -succinate, 13 C 4 -fumarate and 13 C 4 -malate isotopologues by forward cycling in TCA cycle (Fig. 5A). Decarboxylation of 13 C 4 -malate via malic enzyme produces 13 C 3 -pyruvate, which may give rise to 13 C 3 -lactate by LDH or 13 C 2 -acetyl-CoA by PDH. Using 13 C 3 -lactate levels as a readout, we show that glutaminolysis pathway is of minor relevance in our experimental conditions (Fig. 5B), something expected for terminally differentiated cells 15 as well as non-diving cells 16 . Moreover, condensation of 13 C 2 -acetyl-CoA with 13 C 4 -oxalacetate would produce 13 C 6 -citrate, which is hardly detected (Fig. 5C). Alternatively, citrate can also be produced by reductive carboxylation of α-KG, which will ultimately render lipogenic acetyl-CoA by cytoplasmic ATP-citrate lyase (ACL) (Fig. 5A). We found higher contribution of 13 C 5 -citrate and 13 C 3 -malate, both suggestive of greater reductive carboxylation in HEP-L cells. In fact, α-KG to citrate ratio, a readout of reductive glutamine metabolism 17 suggests that reductive carboxylation is approximately 2-3-fold and 10-fold more active in HEP-L cells and PHH respectively (Fig. 5D).
To definitively assess the contribution of glutamine to lipogenic acetyl-CoA through reductive or oxidative glutamine metabolism, we used 13 C 5 15 N 2 -Gln and 5-13 C-Gln tracers and performed isotopomer spectral analysis 18,19 . Incubation with 5-13 C-Gln allows to determine the contribution of glutamine to lipids by reducing carboxylation (Fig. 5E). Both cell types generated labelled myristic acid (C14:0), although HEP-L cells rate was 53% higher (Fig. 5F). When the same analysis was performed using 13 C 5 15 N 2 -Gln as a tracer to estimate the total contribution of glutamine to fatty acid synthesis, no difference was found, indicating that reductive carboxylation is the primary route for glutamine incorporation into fatty acids. Finally, we quantified the retention of isotope label in citrate and malate pools from cells incubated with 1-13 C-Gln (Fig. 5G). HEP-L cells accumulated more 1-13 C isotopologues. However, we found differences between label retention in citrate and malate and also between HEP-L cells and primary cultured human hepatocytes (Fig. 5H).

Discussion
Direct conversion of human fibroblasts to HEP-L cells can be achieved by expression of a specific set of transcription factors 2 . HEP-L cells have been characterized by gene expression profiling and limited functional assays, such as CYP450 activity, glycogen storage (PAS staining) or ac-LDL intake. However, metabolic reprogramming of HEP-L cells (switch between fibroblastic and hepatic metabolic features) has not been covered in these studies. In this paper, we have addressed changes in glutamine/glutamate metabolic pathways in reprogrammed cells and the transcriptomic changes underneath. HEP-L cells acquire a high capacity to deamidate glutamine into glutamate and secrete it into the media, diverting glutamate from other metabolic pathways; moreover glutamate-ammonia www.nature.com/scientificreports www.nature.com/scientificreports/ ligation to glutamine is reduced. Incorporation of glutamine/glutamate carbon moiety to cell protein is reduced, as well as the synthesis of proline from glutamate. However, HEP-L cells significantly increase glutamine anaplerosis into fatty acids by reductive carboxylation. All this characteristics are in agreement with the metabolism of primary cultured human hepatocytes as shown in this work.
Glutamine/glutamate metabolism is heterogeneous in the liver. This is accomplished by drastic differences in periportal to perivenous expression of glutamine/glutamate transporters, phosphate-activated glutaminase (GLS2) and glutamine synthase (GLUL) 9,13 . Periportal hepatocytes uptake glutamine from the portal vein, deamidate it and secrete large amounts of glutamate while keeping glutamine synthesis very low 11 . HEP-L cells also perform this glutamate-secreting pathway by de novo expression of liver-specific glutaminase (GLS2), not inhibited  13 C enrichment of cellular protein from HDF and HEP-L cells. Cell precipitates from methanol treatment were washed and 13 C/ 12 C ratio estimated by C/IRMS. 13 C enrichment is defined as the difference between the percentage of 13 C in sample and the natural abundance estimated in cells incubated without tracer (1.0799 ± 0.0023%; n = 8). Data is individually plotted (n = 8 per group); ***p < 0.001. (C) Diagram summarizing proline/ornithine biosynthetic pathway and the main isotopic metabolic transitions of 13  N 2 -Gln enters the TCA cycle after deamination to 13 C 5 -α-ketoglutarate. 13 C 5 -α-ketoglutarate can be further metabolized via forward cycling to 13 C 4 -succinate, 13 C 4 -malate and so on (red). 13 C 5 -α-ketoglutarate can also generate 13 C 5 -citrate by reductive carboxylation initiated by reverse cycling catalysis of isocitrate dehydrogenase (IDH; green). As an external TCA pathway, 13 C 4 -malate can be decarboxylated to 13 C 3 -pyruvate by malic enzyme (ME; golden) that will ultimately produce 13 C 3 -lactate or www.nature.com/scientificreports www.nature.com/scientificreports/ by glutamate concentration 8,9 , at levels comparable to human liver (90%), remodeling of glutamate transport expression and low levels of glutamine synthase (GLUL) a classical perivenous hepatocyte marker 13 . A shift in the intracellular-extracellular concentration gradient of glutamate might also be critical for glutamate flux reversal ( Supplementary Fig. S6).
Glutamine metabolic pathways participate in proteins, lipids and sugar metabolism. We found that HEP-L cells reduce the utilization of glutamine for intracellular metabolism when compared to HDF. Among the different pathways, we detected reduced incorporation of glutamine carbon moiety to cellular protein and reduction of proline synthesis. Surprisingly, glutamine anaplerosis into fatty acids by reductive carboxylation (Gln→aKG→Cit→Acetyl-CoA lipo ) was higher in HEP-L cells.
According to our analysis, HDF reprogramming to HEP-L cells involves a metabolic rewiring that leads to the loss of metabolic pathways characteristic of HDF and the acquisition of metabolic pathways present in hepatocytes in concurrence with changes in the transcription of metabolic-related genes. Fibroblast is one of the most common collagen-producing cell in the body. Collagen is composed by 28% of proline/hydroxyproline 6 . Proline is synthesized in the cell from glutamate 20 . HEP-L cells hardly accumulated labeled proline despite containing 3-fold higher glutamate concentration (Supplementary Fig. S6) and we found that expression of genes involved in degradation of proline to glutamate were the most probable cause of this change. Yet, metabolic differences exist between HEP-L cells and PHH. Contribution of glutaminase activity to glutamate pool is several fold greater in HEP-L cells. Also, the importance of reductive carboxylation in glutamine anaplerosis is of higher relevance in PHH based on several parameters, i.e. the levels of citrate M5 and malate M3 in cells incubated with 13 C 5 15 N 2 -Gln, the ratio αKG/citrate in cells incubated with unlabeled media and the levels of citrate M1 in cells incubated with 1-13 C-Gln. It is interesting to note that we found important differences between malate and citrate M1 abundance in the latter incubations.
Reductive carboxylation of glutamine-glutamate derived α-ketoglutarate was initially described as an alternate pathway in rat liver 21 . In their tracing experiments, 5-14C-glutamate incorporated to fatty acids (10%). Ulterior papers confirmed the original results in liver and elaborated a sophisticated model based on the reversibility of isocitrate dehydrogenase 12,22-24 . Briefly, using 13 C glutamate tracers, a striking 10.8% M5 citrate when M5 glutamate was perfused in rat liver. In fact, M5 enrichment of citrate amounts to 45% of the M5 enrichment of α-ketoglutarate. The authors concluded that the formation of M5 citrate can only be explained by the reversal of the isocitrate dehydrogenase reaction. Later studies concluded that reversion of isocitrate dehydrogenase reaction would provide cytosolic acetyl-CoA from mitochondrial α-ketoglutarate for fatty acid synthesis. Thus, the reversal of the isocitrate dehydrogenase reaction should be included in isotope labeling models of the citric acid cycle. Reductive carboxylation was later observed in heart and brown adipocytes 25 . The role of reductive carboxylation in liver, heart or adipose tissue, is to maintain lipogenesis under hypoxia, when acetyl-CoA supply from glucose-derived oxidation might be limited 26 . We interpret the increased capacity of HEP-L cells to divert glutamine via reductive carboxylation to FA synthesis compared to HDF as an important metabolic reprogramming into the hepatocyte cell lineage supporting a metabolic rewiring of glutamine-glutamate metabolism.
In summary, we found major changes in glutamine/glutamate metabolism in reprogrammed HEP-L cells compatible with hepatocyte function and validated using freshly isolated primary cultured human hepatocytes. In particular, some of these changes are suggestive of periportal hepatocyte function i.e. intense deamidation of glutamine, secretion of large amounts of glutamate 27 , low glutamine synthesis, high expression of GLS2 mRNA and down-regulation of GLUL mRNA. We also found the loss of active proline synthesis characteristic of HDF. All these changes are in agreement with reprogramming into HEP-L cells. Nevertheless, metabolite pathways differences exist between HEP-L cells and mature hepatocytes that are probably indicating uncomplete reprogramming. In fact, our conclusions are limited by the fact that HEP-L cells cells generated in this study do not display a complete mature phenotype. Expression of genes such as CYP2C9, CYP2C19, CYP2D6, ARG1, OTC, CPS1, PXR, CAR, NURR1 or ESR1 28 are hardly expressed compared to hepatocytes (data not shown). Implementation of recently improved protocols 29 , will probably bring metabolic function in HEP-L cells closer to PHH. In this sense, future metabolic profiling of cells in the process of reprogramming, may also help identify metabolic barriers that hamper HEP-L cells reprogramming and design improved cell media to facilitate a complete reprogramming into hepatocytes.

Methods
All methods were carried out in accordance with relevant guidelines and regulations. All experimental protocols were approved by the Instituto de Investigación Sanitaria la Fe.
Human liver samples were obtained in agreement with the rules of the hospital's Ethics Committee. Informed consent was obtained from all donors. Patients had no known liver pathology nor did they receive medication during the weeks prior to surgery. None of the patients were habitual consumers of alcohol or other drugs. Human hepatocytes were isolated from 2 liver biopsies (<5 g), using a two-step collagenase perfusion technique 32 . Hepatocytes were seeded on fibronectin/collagen-type I-coated (Sigma-Aldrich, Madrid, Spain) dishes and cultured with Ham's F-12/Williams (1:1) medium (Gibco BRL, Paisley, Scotland). Seeded hepatocytes were allowed to settle for 6 hours before experiments were initiated (Table S3).
qRT-PCR, immunofluorescence, PAS staining, Indocyanine green transport and ELISA. qRT-PCR, immunofluorescence, PAS staining and Indocyanine green transport were performed as previously described 33 . Control Liver RNA sample was obtained by combining human liver total RNA from five donors ( Table S4). The antibodies used in this study are described in Table S5. Fluorescence images were taken in Olympus FV1000 confocal mounted on an IX81 inverted microscope.
To determine the presence of human albumin in cell media and mice sera, we used a human Albumin ELISA Quantitation Set (Bethyl Laboratory; Montgomery, TX, USA) according to the manufacturer´s instructions. Reference value (primary cultured human hepatocytes) were obtained from previous data from our group 34 . In vivo transplantation. Transplantation of HEP-L cells in male CB17/Icr-Prkdc scid/Crl mice was done as previously described 35 . Animals were acquired from Charles River Laboratories (Willmington, MA, USA) and housed at the animal facilities of the Instituto de Investigación Sanitaria La Fe. All experiments were performed in accordance with relevant guidelines and regulations and approved by the Institutional Animal Ethics Committee of the Instituto de Investigación Sanitaria La Fe and Generalitat Valenciana (reference number IP.RBM.#6A-3-2015). Briefly, three hours after the injection of 300 mg/kg of APAP, mice were anaesthetized with a sevoflurane/O 2 mixture and the lower pole of the spleen was exposed. Animals received an intrasplenic injection of 10 6 HEP-L cells LT cells in 200 μl of PBS within seconds. The control mice, which had also received APAP treatment, received an intrasplenic injection of PBS. Thirty days after infusion, mice were sacrificed under anaesthesia (sevoflurane/O 2 mixture). Blood was collected and serum aliquots were protected from light and stored at −80 °C until analysis. Isotopic labelling and sample processing. HDF cultured on collagen coated 12w plates were infected with lentiviral vectors expressing HNF4, HNF1A and FOXA3. After 11 days, glutamine or glucose in HMM media was substituted with isotope-labelled compounds at the same final concentration. Cell media was collected after 24 hours and immediately frozen in liquid N 2 , and kept at −80 °C until analysis. Before analysis, samples were diluted 1/100 with water containing D4-Succinic acid, D5-Glutamic and D5-Phenylalanine as internal standards (IS; final concentration 2ppm) and filtered through a modified PES 3 K molecular exclusion filter (VWR; Radnor, PA, USA). Cell monolayers were washed three times with 1 mL of cold PBS and metabolites extracted immediately by scraping the cells with 600 µL of a methanol/water (4:1) solution at −20 °C. The cell extract was transferred to a clean tube and immediately frozen in liquid N 2 , and kept at −80 °C for at least 24 hours. When ready for analysis, samples were centrifuged at 12000 rpm for 15 min at 4 °C. The supernatant was transferred to a clean tube, IS added and dried under vacuum; the dry residue was stored at −80 °C until analysis. Before analysis, residues were resuspended in 50 µL of water. Pellet was resuspended in 600 μL NaOH 0.5 M and protein quantified by Lowry in a small aliquote. The rest of the pellet was washed twice with cold methanol/ water (4:1) solution to decrease sodium concentration and 13 C enrichment determined by C/IRMS (Combustion/ Isotope Ratio Mass Spectrometry) using a stable isotope ratio mass spectrometer (Thermo Finnigan TM MAT253) coupled to an elemental analyzer (Carlo Erba EA1108) in Servizos de Apoio a Investigacion (SAI, Universidade da Coruña).

Liquid chromatography high-resolution mass spectrometry (LC-HRMS) analysis.
Chromatographic analysis was performed on an Agilent 1290 Infinity II (Agilent Technologies, Santa Clara, CA, USA) HPLC system equipped with a quaternary pump, vacuum degasser and an autosampler with a temperature controller. Chromatographic separation of metabolites was achieved on a 150 mm × 2.1 mm, 4 µm particle size Synergi-Hydro C18 column (Phenomenex Inc, Torrance, CA, USA) with the following separation conditions: www.nature.com/scientificreports www.nature.com/scientificreports/ solvent A, water/FA (99.8:0.2); solvent B, ACN; separation gradient, initially 1% B, held for 2 min and then linear 1-80% B in 8 min, washing with 98% B for 2 min and column equilibration with 1% B for 7 min; flow rate, 0.25 mL/min; injection volume range, 0.2-4.5 µl. Autosampler and column temperatures were set at 6 °C and 23 °C, respectively.
Mass spectrometry analysis was carried out by an Agilent 6550 Q-ToF (Agilent Technologies, Santa Clara, CA, USA) detector equipped with an electrospray (ESI) source with Jet Stream Technology. Column flow was conducted into the mass analyzer in the time range of 0.7-12 min diverting the rest of run time to waste. Isotopologue study of samples was separately performed in positive and negative ionization modes under full MS scan mode. Positive assignments were achieved through autoMS/MS analysis with exclusion and inclusion lists in the respective ionization mode. Shared MS conditions of analysis were: gas temp, 130 °C; drying gas, 14 L/min; nebulizer, 30 psig; sheath gas, 10 L/min; capillary voltage, 3500 V and 3000 V for positive and negative ionization modes, repectively; fragmentor, 380 V; octapole 1 RF, 400 V; isolation width, narrow (1.3 m/z); Nozzle voltage, 500 V funnel exit DC, funnel RF HP and funnel exit RF LP, 50, 150 and 60 V, respectively; lock masses, 121.0509/922.0098 and 119.0363/980.0164 for positive and negative, respectively; considered m/z range, 40-750; data acquisition, centroid mode.
Particular settings for full MS isotopologue analysis: acquisition rate, 1.5 scans/sec (cycle time, 0.666 sec). The autoMS/MS analyses merged full MS with MS/MS experiments with settings: acquisition rate for both scan events, 4 scans/sec (cycle time, 1.1 sec); collision induced dissociation (CID) energies, 10 and 20; Max precursors per MS/MS cycle, 3; Active MS/MS exclusion, after 2 scans during 0.3 min; isotope mode, off; Precursor threshold, 3000 counts; exclusion list, included lock masses and those from a blank injection; inclusion list; included m/ zs from low-abundance metabolites in samples.
Before sample analysis, the MS device was tuned and calibrated in low mass range and high resolution mode (4 GHz). Considered mass tolerance for full MS and MS/MS analyses for data processing was 10 ppm.
LC-MS data processing. Untargeted cell media metabolic profiling was performed by analyzing processed media samples from cells incubated with unlabeled nutrients with the Molecular Feature Extraction wizard of the Profinder B.08.00 program. Molecular features (characterized by exact mass and retention time) with MFE score above 95% and present in all samples of at least one group (HDF or HEP-L cells) were selected for statistical analysis.
Positive assignments were manually achieved by the comparison of the MS/MS breakdown profiles of metabolites from samples with those from either commercial standards and detailed in the Metlin database (https:// metlin.scripps.edu). Two in-house libraries listing the identified compounds were built to carry out automated data processing. Automated peak integration of unlabeled and labeled metabolites from samples was carried out by the Targeted Feature Extraction wizard of the Profinder B.08.00 program. In this case, there was loaded a customized library detailing the retention time and neutral unlabeled/labeled masses of identified metabolites considering all the isotopologue combinations (from [M0] to [Mn]), while setting isotope abundance and spacing scores to 0%. Mass and retention time tolerances used by Profinder analysis were 10 ppm and 0.2 min, respectively. Isotopologues abundance were corrected for the presence of naturally occurring isotopes using FLUXFIX software 36 .
Absolute quantification of glutamic acid and glutamine isotopologues was carried out through their respective relative response factors using D5-glutamic and D4-succinic acids as IS, respectively. The rest of metabolites considered in this research were normalized and relatively quantified through the integrated peak area ratios of analytes and their respective IS. Absolute intracellular metabolite concentration was determined dividing the amount of metabolite in the cell extract/plate expressed as milimoles by total cellular volume determined in parallel plates (average cell volume x plate cell number). Alternatively, LC/MS data was normalized by mg of protein in the extracted cell pellet.
Fatty acid analysis. Lipidome was extracted and total lipids were saponified as described previously 37 .
Chromatographic analysis was performed on an an Acquity Ultra Performance LC system (Waters, Milford, MA, USA) equipped with an Acquity UPLC BEH C18 column (100 × 2.1 mm; 1.7 µm) (Waters, Milford, MA, USA) as described 38 . Mass spectrometry analysis was carried out by a Synapt G2-Si Q-TOF (Waters, Milford, MA, USA).a, CA, USA) detector equipped with an electrospray (ESI) source with Jet Stream Technology. Data analysis was performed using LipidMS package (https://CRAN.R-project.org/package=LipidMS). Labelling of fatty acids for isotopomer spectral analysis (ISA) was conducted over 2 days of culture with medium change after the first day to prevent tracer/nutrient depletion. ISA was performed using the simple network described in 18 . Uncorrected 13 C mass isotopomer distributions were fitted to the model equations and the tracer enrichment in lipogenic acetyl-CoA (D value) and percentage of newly synthesized lipids and de novo lipogenesis, g(t) were estimated by GRG nonlinear evolutionary method using Solver add-in in Excel 39 . NMR analysis. Samples were placed on ice and allowed to thaw for 20 min. 500 µl of medium was added to 100 µl of phosphate buffer (100 mM Na2HPO4 pH 7.4, in 100% D2O) containing 15 mM 3-(trimethylsilyl) propionic −2,2,3,3-d4 acid sodium salt (TSP), as internal standard. The samples were vortexed and transferred into a 5 mm NMR tube. Samples were analyzed on a Bruker Ultrashield Plus 600 MHz spectrometer equipped with a 5 mm TCI cold probe. 13 C experiments were acquired with a 29761 Hz spectral width, 32768 data points, acquisition time of 0.55 s, relaxation delay of 60 s and 256 scans at 27 °C. Spectra were processed with exponential line broadening to 3 Hz and zero filling to 65536 points. Multiplicity Heteronuclear Single Quantum Correlations (HSQC) were performed for representative samples for signal assignment. For these experiments, 256 t1 increments were used and 32 transients were collected. The relaxation delays were set to 1.5 s and the experiments were acquired in the phase-sensitive mode. Following Fourier transformation, 1D spectra were manually phased, www.nature.com/scientificreports www.nature.com/scientificreports/ baseline corrected and referenced to the TSP peak (0.00 ppm) using MestReNova 8.1. Metabolite identities were assigned by comparison to reference values for chemical shift and multiplicity. Quantification was carried out with the Eretic Signal (Bruker Biospin). Metabolite signals were integrated with MestreNova 8.1 using the GSD deconvolution option. Statistical analysis. The number of experiments (n) throughout the paper refers to the number of independent biological replicates (cell plates, animals…). The infection was performed using lentivirus generated in three time-separated batches. Sample size was estimated using Monte Carlo Simulations. Based on mean values and standard deviations from preliminary experiments a sample size of 7 per group was needed to achieve a statistical power of 80% at the standard significance level of alpha = 0.05. When needed the sample size was increased to 15. Primary cultured human hepatocytes used for metabolic studies were derived from two donors (a total of 7 replicates distributed in 3 and 4 cell plates from each donor). Descriptive statistics (mean ± s.d.) are used for experiments with n ≥ 10. When n < 10, individual points are plotted. Statistical differences were calculated using two-tailed Student's t-test for unpaired samples (also known as Welch's t-test). Normal distribution was confirmed using the Shappiro-Wilk test. Volcano plot was generated using Metaboanalyst software (www.metaboanalyst.ca). Data related to every metabolite-related determination, was acquired by an investigator blinded to the group allocation since all samples were number-coded. Data analysis was performed by another investigator.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.