Abstract

   Non-alcoholic fatty liver disease is associated with an irregular
   serine metabolism. Serine hydroxymethyltransferase 2 (SHMT2) is a liver
   enzyme that breaks down serine into glycine and one-carbon (1C) units
   critical for liver methylation reactions and overall health. However,
   the contribution of SHMT2 to hepatic 1C homeostasis and biological
   functions has yet to be defined in genetically modified animal models.
   We created a mouse strain with targeted SHMT2 knockout in hepatocytes
   to investigate this. The absence of SHMT2 increased serine and glycine
   levels in circulation, decreased liver methylation potential, and
   increased susceptibility to fatty liver disease. Interestingly,
   SHMT2-deficient mice developed simultaneous fatty liver, but when fed a
   diet high in fat, fructose, and cholesterol, they had significantly
   less inflammation and fibrosis. This study highlights the critical role
   of SHMT2 in maintaining hepatic 1C homeostasis and its stage-specific
   functions in the pathogenesis of NAFLD.

   Subject terms: Metabolic syndrome, Molecular medicine
     __________________________________________________________________

   Mouse genetics indicates that serine hydroxymethyltransferase 2
   maintains hepatic methylation potential and plays stage-specific roles
   in the pathogenesis of non-alcoholic fatty liver disease.

Introduction

   Non-alcoholic fatty liver disease (NAFLD) is a significant health issue
   that affects around a quarter of the global population. This condition
   is progressive and includes hepatic steatosis, steatohepatitis, liver
   fibrosis/cirrhosis, and hepatocellular carcinoma, which can be
   life-threatening. Despite the high prevalence of NAFLD, no approved
   treatments are available, highlighting the urgency for more research
   and development in this field^[38]1,[39]2.

   NAFLD can arise from impaired one-carbon (1C) metabolism, a metabolic
   process that involves the transfer of bioactive methyl groups (1C
   units) via pathways such as the folate and methionine
   cycles^[40]3,[41]4. These methyl groups play a vital role in producing
   essential metabolic intermediates, including nucleotides and
   phospholipids, and participate in gene regulation^[42]5. Recent
   research on patients and mice has demonstrated that maintaining proper
   hepatic methylation potential, as determined by the concentration of
   the main 1C metabolite S-adenosylmethionine (SAM), is crucial for
   preventing NAFLD. SAM-dependent phosphatidylcholine (PC) synthesis is
   pivotal for upholding the integrity of the phospholipid metabolome.
   This, in turn, either supports the assembly of very low-density
   lipoprotein (VLDL), which promotes hepatic triglyceride (TG) disposal,
   or prevents endoplasmic reticulum (ER) stress, thereby reducing the
   risk of hepatocyte de novo lipogenesis (DNL) hyperactivation.
   Ultimately, this process helps reduce the risk of hepatic fat
   accumulation and NAFLD progression^[43]6–[44]9.

   Serine is the primary supplier of 1C units in the human body, and
   isotopic tracing has shown that almost all methyl groups used to
   replenish the SAM 1C pool are derived from this amino acid^[45]10.
   Studies indicate that individuals with NAFLD and metabolic syndrome
   (MetS) tend to have lower levels of serine and glycine in
   circulation^[46]11,[47]12. These two amino acids have been suggested as
   promising biomarkers for diagnosing NAFLD and guiding nutritional
   interventions^[48]12–[49]15. The enzyme serine hydroxymethyltransferase
   (SHMT) breaks down serine into glycine, producing 1C for methylating
   folate in the folate cycle^[50]16. The methylated folate species
   supports the methionine cycle, which produces SAM. Systems biology
   modeling suggests that NAFLD may be caused by a deficiency in serine
   and reduced SHMT activity^[51]17.

   Mammals have two SHMT paralogs: cytosolic SHMT1 and mitochondrial
   SHMT2^[52]18. While the global SHMT1 knockout mouse model has been
   observed to be healthy, fertile, and viable with elevated hepatic SAM
   levels, indicating that SHMT1 may not be necessary to maintain the 1C
   pool^[53]19, the whole-body SHMT2 deletion leads to embryonic
   lethality, indicating its crucial role^[54]20,[55]21. However, global
   SHMT2 knockout is unsuitable for simulating metabolic diseases that
   usually impact adults. It is yet to be defined how the critical
   serine-catabolic enzyme SHMT2 regulates 1C homeostasis and contributes
   to metabolic health in adult animals.

   We have modeled hepatic SHMT2 deletion in mice, which survives
   embryonic development. Our research has revealed the crucial role of
   SHMT2 in maintaining hepatic methylation potential in adult animals. We
   have further discovered that SHMT2 has a nonlinear function in NAFLD:
   inhibiting hepatic steatosis but supporting liver fibrosis development.
   These results have established the pivotal role of SHMT2 in regulating
   hepatic 1C levels while uncovering its potential therapeutic values
   against NAFLD at various stages.

Results

SHMT2 is essential for maintaining hepatic methylation potential

   In mice, whole-body deletion of SHMT2 causes embryonic lethality due to
   faulty fetal hematopoiesis in embryonic liver^[56]22, making it
   impossible to assess adult phenotypes. To address this, we created a
   floxed SHMT2 mouse allele (SHMT2^fl/fl) with two loxP sites spanning
   exons 3–12, allowing for conditional targeting by lox/Cre recombination
   (Fig. [57]S1a). Next, we crossbred this allele with the Albumin-Cre
   allele^[58]23 to specifically target the hepatocytes in the liver.
   Successful recombination at the SHMT2 locus was confirmed by PCR-based
   mouse genotyping (Fig. [59]S1b). The liver is a vital metabolic organ
   that expresses a significant amount of SHMT2^[60]24. The resulting
   mouse strain has been named SHMT2^HKO. Mice possessing homozygous
   SHMT2^HKO genes exhibited no discernible abnormalities at birth and
   were capable of reproduction. Compared to the SHMT2^fl/fl control,
   their liver SHMT2 protein levels decreased by approximately 50% at
   birth, continued to decrease to about 30% at weaning, and were almost
   undetectable in adulthood (Fig. [61]1a). Throughout the process of
   liver development, there is a notable increase in the number of
   hepatocytes while the number of non-parenchymal cells (NPCs) gradually
   decreases after birth^[62]25. This observation suggests that targeting
   hepatocyte SHMT2 in SHMT2^HKO mice was successful, without affecting
   SHMT2 expression in NPCs.

Fig. 1. The deletion of SHMT2 in the liver significantly elevates the levels
of glycine in circulation.

   [63]Fig. 1
   [64]Open in a new tab

   a SHMT2 protein levels were tested in the SHMT2-deficient (HKO) and
   control (fl/fl) mice liver at specific times using Western blotting. b
   The heatmap shows amino acid expression in control and SHMT2-deficient
   mice. c–e Bar graphs illustrate the variation in circulating amino acid
   levels between SHMT2-deficient and control mice. (Mean ± SD; *p < 0.05,
   **p < 0.01, ****p < 0.0001; t-test).

   We examined the role of the SHMT2 enzyme in the breakdown of the serine
   amino acid, which results in the production of glycine. To do this, we
   fed SHMT2^HKO mice a chow diet until they were 22 weeks old and then
   analyzed their circulating amino acid profiles. Our findings showed
   significant changes in the levels of serine, glycine, and aspartate in
   the SHMT2^HKO mice (Fig. [65]1b–e). We found that the absence of SHMT2
   increased the levels of serine in the bloodstream, indicating a
   decrease in hepatic serine catabolic activity. Interestingly, the
   SHMT2^HKO mice had three times more serum glycine levels than the
   control mice, which cannot solely be attributed to the expected
   decrease in glycine production. Additionally, deleting hepatic SHMT2
   led to lowered serine levels and elevated glycine levels in the liver
   (Fig. [66]2a, b). These findings suggest that deleting hepatic SHMT2
   has metabolic impacts beyond the breakdown of serine and the production
   of glycine in intact animals.

Fig. 2. SHMT2 deficiency reduces liver methylation potential.

   [67]Fig. 2
   [68]Open in a new tab

   a, b Bar graphs illustrate serine and glycine levels in the livers of
   SHMT2-deficient (HKO) and control (fl/fl) mice. (Mean ± SD;
   ***p < 0.001, ****p < 0.0001; t-test). c Comparison of metabolite
   levels in mouse livers (HKO vs. fl/fl) shown on a volcano plot. d
   Bubble plot showing enriched pathways of differentially expressed
   metabolites in mouse livers (HKO vs. fl/fl). e A cartoon diagram
   showing the mapping of 1C DEMs in the folate, methionine cycles, and
   choline degradation pathways. Ser serine, Gly glycine, Sar sarcosine,
   DMG dimethylglycine, THF tetrahydrofolate, CH[2]-THF methylene
   tetrahydrofolate, CH[3]-THF methyl tetrahydrofolate, Met methionine,
   Hcy homocysteine, SAM S-adenosylmethionine, SAH S-adenosylhomocysteine,
   NAM nicotinamide, MNA 1-methylnicotinamide, GAA guanidinoacetate, SHMT2
   serine hydroxymethyltransferase 2, SARDH sarcosine dehydrogenase, MS
   methionine synthase, BHMT betaine homocysteine methyltransferase, MAT1A
   methionine adenosyltransferase 1A, GNMT glycine N-methyltransferase. f
   The heatmap shows the DEM levels of the methionine, serine and glycine,
   PE and PC pathways in mouse livers. g–i Liver SAM and SAH levels were
   analyzed to assess hepatic methylation potential in HKO and fl/fl mice.
   (Mean ± SD; **p < 0.01; t-test).

   We conducted targeted metabolomics analysis of liver metabolites to
   evaluate the metabolic reprogramming that occurs with the deletion of
   hepatic SHMT2. We identified 48 metabolites that were differentially
   expressed (DEM) out of 188 metabolites analyzed, with a cutoff of
   p < 0.05 and FDR < 0.20 (Fig. [69]2c, Supplementary Data [70]1, [71]2).
   Pathway enrichment analysis showed that the DEMs were mainly localized
   within the top pathways of the methionine metabolism, serine and
   glycine metabolism, phosphatidylethanolamine (PE) biosynthesis, and
   phosphatidylcholine (PC) biosynthesis, which are crucial components of
   the 1C metabolic process and the related biogenesis of phospholipids
   (Fig. [72]2d). In the liver of SHMT2^HKO mice, we observed a distinct
   pattern of metabolite expression, where the levels of 1C-donating
   metabolites, including serine, choline, betaine, and 1-methyl
   nicotinamide (1-MNA), decreased, and the levels of 1C-recipient
   metabolites, including glycine, guanidinoacetate (GAA), and putrescine,
   increased (Fig. [73]2e, f). These findings suggest that the deletion of
   SHMT2 has led to a hepatic 1C deficiency. We also measured the levels
   of SAM and S-adenosyl homocysteine (SAH), which indicates methylation
   potential^[74]26. We found that SHMT2^HKO mice had decreased SAM levels
   and increased SAH levels in the liver, resulting in a significant
   decline in the SAM/SAH ratio (Fig. [75]2g–i). These results highlight
   the critical role of SHMT2 in maintaining the liver’s 1C pool and
   methylation potential, even if other pathways can also supply 1C units
   in hepatocytes (Refer to pathway diagram in Fig. [76]2e).

   Glycine plays a crucial role in the formation of glutathione, which is
   the primary antioxidant of the body. We observed that SHMT2^HKO mice
   showed a significant increase in glycine levels in their liver and
   bloodstream. This led us to investigate whether it affected the
   glutathione concentration in the liver. However, we found no
   significant change in hepatic glutathione abundance due to SHMT2
   deletion (Fig. [77]S2).

The absence of SHMT2 leads to the development of simultaneous fatty liver

   As reduced liver methylation capacity increases susceptibility to
   NAFLD, we examined if lack of SHMT2 impacts liver pathogenesis. We
   observed SHMT2^HKO mice under chow-diet-feeding conditions for up to 22
   weeks. After analyzing their growth kinetics, we found that the absence
   of SHMT2 did not cause significant changes in body weight or body
   composition (Fig. [78]3a, b). SHMT2 knockout animals displayed no
   differences in glycemia, glucose tolerance, insulin sensitivity,
   circulating triglyceride (TG), or total cholesterol (TC) levels
   compared to the control group (Fig. [79]3c–e). On the other hand, the
   deletion of SHMT2 significantly increased hepatic levels of TG and TC
   and circulating alanine aminotransaminase (ALT) levels (Fig. [80]3f,
   g). Histological analysis and Oil Red O staining consistently showed
   that the abundance and size of liver lipid droplets significantly
   increased with SHMT2 deletion (Fig. [81]3h). Our findings suggest that
   SHMT2 prevents hepatic steatosis and protects the liver from steatosis
   damage in mice under chow-fed conditions.

Fig. 3. The development of hepatic steatosis occurs concomitantly with the
deficiency of SHMT2.

   [82]Fig. 3
   [83]Open in a new tab

   a Growth kinetics of SHMT2-deficient (HKO) and control (fl/fl) mice. b
   Body composition of 22-week-old mice was measured using Eco-MRI during
   harvesting. c The glucose intolerance test was conducted when the
   subject reached 18 weeks of age (n = 7). d The insulin sensitivity test
   was conducted when the subject reached 20 weeks of age (n = 6). e
   Measuring the levels of total triacylglycerol (TG) and total
   cholesterol (TC) in circulation. f Levels of TG and TC in the liver.
   (Mean ± SD; *p < 0.05; t-test). g Levels of ALT in circulation
   (Mean ± SD; *p < 0.05; t-test). h The mouse liver sections were stained
   using H&E (top) and Oil Red O (bottom). Scale bar: 100 μm.

The absence of SHMT2 worsens hepatic steatosis caused by an over-nutritious
diet, but it lessens inflammation and fibrosis in the liver

   Overnutrition is known to exacerbate NAFLD. To investigate the effects
   of NAFLD progression in rodents, the AMLN diet, which is high in fat,
   fructose, and cholesterol, is commonly utilized^[84]27,[85]28. In this
   study, SHMT2^HKO mice were switched from their regular diet to the AMLN
   diet at four weeks of age and were harvested at 22 weeks of age,
   following 18 weeks on the AMLN diet. The SHMT2^HKO mice on the AMLN
   diet demonstrated increased fat mass, resulting in higher body weight
   compared to the control group (Fig. [86]4a). Biochemical analyses
   revealed that these mice had elevated levels of circulating and hepatic
   TG and TC (Fig. [87]4b, c). Additionally, they exhibited increased
   fasting glycemia and more severe glucose disposal impairment and
   insulin insensitivity after 14 and 16 weeks of exposure to the AMLN
   diet (Fig. [88]4d, e). Liver histological analysis indicated that
   SHMT2^HKO mice had increased lipid droplet deposition in hepatocytes,
   mainly exhibiting excessive microvesicular steatosis in the pericentral
   zone, resulting in higher steatosis grading compared to the control
   (Fig. [89]4f, g). However, the SHMT2^HKO mice demonstrated
   significantly reduced lobular inflammation and pericellular fibrosis,
   leading to a significant decrease in inflammation and fibrosis
   pathological grading compared to the control group (Fig. [90]4f, g).
   The SHMT2^HKO mice exhibited a notable increase in the levels of
   circulating ALT (Fig. [91]4h). We concluded that SHMT2 deletion in
   hepatocytes worsens liver steatosis and injury but reduces inflammation
   and fibrosis in the AMLN diet treatment.

Fig. 4. SHMT2 deficiency worsens fatty liver but improves AMLN diet response
by reducing inflammation and fibrosis.

   [92]Fig. 4
   [93]Open in a new tab

   a Eco-MRI measured HKO and fl/fl mice’s body composition after 18 weeks
   on AMLN diet at 22 weeks old (Mean ± SD; *p < 0.05, **p < 0.01;
   t-test). b The levels of total triacylglycerol (TG) and total
   cholesterol (TC) in mouse circulation after 18 weeks on AMLN diet.
   (Mean ± SD; *p < 0.05; t-test). c The levels of TG and TC in mouse
   livers after 18 weeks on AMLN diet. (Mean ± SD; *p < 0.05; t-test). d
   Glucose tolerance test conducted on mice after 14 weeks on AMLN diet
   (Mean ± SD; n = 6; ***p < 0.001; t-test). e Insulin sensitivity test
   conducted on mice after 16 weeks on AMLN diet. (Mean ± SD; n = 6;
   *p < 0.5; t-test). f H&E (left), Oil Red O (middle), and Sirius Red
   (Right) staining of mouse liver sections after 18 weeks on AMLN diet.
   Scale bar: 100 μm. g Pathological grading of liver steatosis (left),
   lobular inflammation (middle), and fibrosis (right). (Mean ± SD;
   *p < 0.05, **p < 0.01; t-test). h Levels of ALT in circulation
   (Mean ± SD; **p < 0.01; t-test).

The deletion of SHMT2 does not have a discernible impact on the expression of
hepatic electron transport chain proteins

   The proper oxidation of mitochondria is crucial in disposing of fatty
   acids and preventing NAFLD. Previous research has indicated that SHMT2
   is critical in maintaining electron transport chain (ETC) protein
   expression and mitochondrial respiration in murine embryonic
   fibroblasts and human HEK293 cells^[94]22,[95]29–[96]31. However, upon
   studying SHMT2^HKO mice livers under chow-fed conditions, Western
   blotting analysis revealed that the protein markers for each
   respiratory complex were expressed at similar levels in comparison to
   control mice (Fig. [97]5a). We isolated primary hepatocytes from mice.
   Western blot analysis confirmed the complete absence of SHMT2 protein
   in hepatocytes from SHMT2^HKO mice (Fig. [98]5b, top). Furthermore,
   Seahorse respirometry analysis conducted on the primary mouse
   hepatocytes demonstrated that the SHMT2 knockout had minimal impact on
   basal and maximal cellular oxygen consumption (Fig. [99]5b, bottom).
   After the SHMT2 gene was deleted, the AML12 mouse hepatocyte cell line
   produced similar results (Fig. [100]S3). Therefore, it can be concluded
   that SHMT2 is not essential for the mitochondrial respiration of mouse
   hepatocytes.

Fig. 5. The deletion of SHMT2 does not impact the expression of ETC proteins
in hepatocytes.

   [101]Fig. 5
   [102]Open in a new tab

   a Western blotting measures each ETC complex’s protein levels in the
   livers of SHMT2-deficient (HKO) and control (fl/fl) mice. The
   individual complex markers were NDUFS1 (complex I), SDHB (complex II),
   UQCRC2 (complex III), COX II (complex IV), and ATP5A (complex V). b Top
   panel: Western blotting confirms SHMT2 knockout in primary hepatocytes.
   Bottom panel: respirometry analysis for the SHMT2 knockout (KO) and
   wild-type (WT) primary hepatocytes (n = 11).

Transcriptomics profiling shows that deleting SHMT2 promotes the expression
of the DNL pathway while reducing inflammation and fibrosis pathways in the
diet-induced NAFLD model

   Using RNAseq, we analyzed the liver transcriptome to explore how SHMT2
   affects the progress of diet-induced NAFLD in mice. We discovered a
   list of genes with varying expression levels between SHMT2^HKO and
   control mice when fed an AMLN diet (Supplementary Data [103]3, [104]4).
   Gene set enrichment analysis (GSEA) showed that upregulated genes were
   highly enriched in lipogenesis pathways, while downregulated genes were
   highly enriched in the inflammatory response and extracellular matrix
   remodeling pathways (Fig. [105]6a–c). Our qRT-PCR analyses revealed
   that the AMLN diet decreased the hepatic expression of de novo
   lipogenesis (DNL) genes in SHMT2^fl/fl control mice, except for Scd1.
   However, this decrease was not present in SHMT2^HKO mice. Furthermore,
   SHMT2^fl/fl mice exhibited elevated hepatic Scd1 expression responsive
   to the AMLN diet, which was even more pronounced in SHMT2^HKO mice
   (Fig. [106]6d). Western blotting confirmed the upregulated hepatic
   protein expression of DNL genes in SHMT2^HKO mice under AMLN diet
   feeding (Fig. [107]6e). Conversely, the AMLN diet significantly
   increased the expression of hepatic inflammatory and fibrotic genes in
   SHMT2^fl/fl mice, but this was significantly reduced by hepatic SHMT2
   deletion (Fig. [108]6f, g). These findings suggest that SHMT2 plays a
   crucial role in suppressing genes involved in de novo lipogenesis,
   while promoting inflammation and fibrosis pathways in a diet-induced
   model of NAFLD.

Fig. 6. Transcriptomics profiling reveals that SHMT2 deficiency increases DNL
gene expression while reducing inflammatory and fibrotic gene expression in
mouse livers.

   [109]Fig. 6
   [110]Open in a new tab

   a RNA-seq analysis showed pathway enrichment of differentially
   expressed genes (DEGs) in the livers of both SHMT2-deficient and
   control mice after being on the AMLN diet for 18 weeks. b GSEA plots
   show DEG enrichment for fatty acid biosynthesis and inflammatory
   response pathways in the liver of HKO and fl/fl mice fed the AMLN diet.
   c RNA-seq detected DEGs of fatty acid biosynthesis (left) and
   inflammatory response (right) the AMLN-fed mouse livers shown in heat
   maps. d The gene expression of the DNL pathway was measured in the
   livers of specified mice using qRT-PCR (Mean ± SD; *p < 0.05,
   **p < 0.01, ***p < 0.001, ****p < 0.0001; t-test). e The protein
   expression of DNL enzymes in mouse livers fed with the AMLN diet was
   measured by Western blotting. The gene expression of the inflammatory
   (f) and fibrotic (g) pathways was measured in the livers of specified
   mice using qRT-PCR. (Mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001,
   ****p < 0.0001; t-test).

The relationship between SHMT2 genetic variations and expression with
metabolic diseases in humans

   We conducted research on functional genomic data using the Common
   Metabolic Disease Knowledge Portal (CMDKP)^[111]32 to investigate
   potential correlations between single nucleotide polymorphisms (SNP) in
   the SHMT2 gene and metabolic phenotypes contributing to metabolic
   diseases. Our analysis was focused on specific SNPs that met the
   following criteria: they are located between the start and stop sites
   of the SHMT2 gene’s transcription, present in the 1000 Genomes
   reference database, and part of the “bottom-line” meta-analyzed genetic
   associations for the relevant phenotype in CMDKP database. These
   variations are particularly suitable for identifying the linkage
   disequilibrium of the SHMT2 gene, which plays a role in developing
   complex metabolic diseases. Our results indicate that SNPs in SHMT2 are
   significantly associated with changes in metabolic parameters such as
   circulating lipids and lipoproteins, body mass indexes, and diabetic
   indicators (Fig. [112]7a and Table [113]S1). These findings suggest
   that genetic variations in SHMT2 may influence SHMT2 gene expression
   and activity, ultimately impacting the development and outcome of
   metabolic syndrome in humans. Upon conducting a more in-depth analysis
   of SHMT1, which encodes the cytosolic counterpart of SHMT2, it was
   observed that the SHMT1 gene SNPs are also significantly associated
   with human metabolic phenotypes that define metabolic syndrome
   (Fig. [114]S4 and Table [115]S2). These results indicate that both
   SHMT1 and 2 are associated with susceptibility to metabolic syndrome in
   human populations.

Fig. 7. Metabolic syndrome and non-alcoholic fatty liver disease (NAFLD) in
humans are linked to genetic and expression level variations of SHMT2.

   [116]Fig. 7
   [117]Open in a new tab

   a SHMT2 polymorphism affects metabolic syndrome parameters in humans,
   as shown in the bubble plot. b SHMT2 expression is upregulated in NASH
   patients’ liver samples, per the meta-analysis shown in the forest
   plot. The Y axis displays GEO accession numbers while the X axis
   represents the standard mean difference (SMD) of SHMT2 gene expression
   between NASH and healthy control (Ctrl) groups.

   Furthermore, we analyzed GSE datasets to determine whether SHMT2
   expression is altered in liver specimens of NAFLD patients. Our
   meta-analysis demonstrated that SHMT2 mRNA levels are significantly
   higher in NASH patients compared to healthy controls in two data sets
   (Fig. [118]7b). This indicates that SHMT2 is a metabolic gene that
   significantly associates with the development and outcome of metabolic
   syndrome and liver disease in human populations.

Discussion

   The liver plays a crucial role in metabolic processes that require a
   significant amount of 1C units, responsible for over 80% of the body’s
   methylation reactions^[119]33. To acquire the necessary 1C pool, two
   major pathways are utilized: the liver-specific choline/betaine
   degradation pathway and the universal serine catabolic pathway. A vital
   question arises on whether these pathways are interchangeable. Past
   research has indicated that disrupting the critical 1C-donating enzyme
   betaine homocysteine methyltransferase (BHMT) in the choline/betaine
   degradation pathway reduces hepatic methylation potential in
   mice^[120]34. Our latest findings indicate that disrupting serine
   catabolic enzyme SHMT2 in the liver reduces SAM while increasing SAH,
   resulting in a diminished SAM/SAH ratio and indicating reduced
   methylation potential. Moreover, SHMT2 loss significantly impacts
   methionine metabolism and the methyl metabolome, leading to a 1C
   deficiency in the liver. The methyl donors or carriers, such as serine,
   choline, betaine, sarcosine, and 1-MNA, decrease, while the methyl
   recipients, such as glycine and GAA, increase. Our results suggest that
   both serine catabolism and choline degradation are essential to fulfill
   the liver’s transmethylation demands for 1C units.

   An unexpected discovery was made when the enzyme SHMT2, responsible for
   producing glycine, was removed from hepatocytes. Surprisingly, the
   SHMT2 removal resulted in an increase in glycine levels both in the
   liver and bloodstream. These findings may reflect the critical role of
   glycine in regulating the 1C balance in the liver through participating
   in a regulatory process called “methyl sink.” When there is an excess
   of 1C, glycine is converted into sarcosine by GNMT, which consumes SAM.
   Conversely, when there is a shortage of 1C, sarcosine dehydrogenase
   (SARDH) breaks down sarcosine into glycine, producing a unit of 1C with
   the help of folate (Refer to pathway diagram in Fig. [121]2e)^[122]35.
   SAM regulates the activation of GNMT through allosteric regulation, and
   it is known that GNMT is the most abundant liver methyltransferase,
   accounting for 1% of liver soluble protein^[123]36. The increase in
   glycine levels in mice without hepatic SHMT2 could be due to SAM
   depletion, leading to the deactivation of GNMT and activation of SARDH.
   This, in turn, inhibits glycine methylation and mobilizes sarcosine
   storage to balance the 1 C pool deficiency, producing a significant
   amount of glycine in the liver that can affect circulation. Our
   research suggests that glycine may be a central component of hepatic
   regulatory “methyl sink”, playing a vital role in balancing the 1C pool
   of the liver.

   Deleting SHMT2 in whole mice arrested erythroblast development in the
   fetal liver, ultimately leading to embryonic death at E13.5^[124]22.
   This highlights the critical role of serine-derived 1C units in
   facilitating the folate cycle and nucleotide synthesis, which are
   crucial for cell proliferation. Interestingly, mice with SHMT2 deletion
   specific to their hepatocytes were able to survive without any
   developmental abnormalities. This suggests that hepatocytes have an
   alternative 1C source for the folate cycle, such as sarcosine, which is
   derived from the degradation of choline and betaine. It is possible
   that the liver has a unique pathway for breaking down
   choline/betaine/sarcosine, which can compensate for the loss of SHMT2
   in hepatocytes. Our research sheds light on how hepatocytes can adapt
   to acquire 1C units to support the folate cycle and cell proliferation,
   ultimately promoting survival.

   This study has uncovered a novel function of SHMT2 in inhibiting fatty
   liver development. While previous studies had connected changes in
   serine metabolism to NAFLD^[125]12,[126]17, our findings offer mouse
   genetic evidence demonstrating the significance of SHMT2-mediated
   serine catabolism in the development of NAFLD. Previous research has
   shown that SHMT2 plays a crucial role in maintaining the electron
   transport chain (ETC) and mitochondrial respiration in mouse embryonic
   fibroblasts and HEK293 cells^[127]22,[128]29–[129]31, which could
   affect lipid disposal through mitochondrial fatty acid oxidation.
   Interestingly, disrupting SHMT2 did not impact cellular oxygen
   consumption or ETC protein expression in liver cells, indicating that
   liver cells have a unique mechanism to safeguard mitochondria, akin to
   serine catabolism. It is possible that the 1C units produced by the
   liver-specific choline degradation pathway may be enough to replenish
   the mitochondrial 1C-folate pool, but not enough to compensate for SAM
   deficiency in SHMT2-deficient hepatocytes. SAM dysregulation can, in
   turn, lead to an accumulation of fat in the liver.

   Previous studies revealed that the restriction of glycine availability
   had a detrimental impact on glutathione biosynthesis and fatty acid
   oxidation, leading to the exacerbation of hepatic steatosis and
   fibrosis in mice^[130]14. Despite the significant increase in
   circulating and hepatic glycine levels, SHMT2^HKO mice did not exhibit
   notable changes in hepatic glutathione abundance or hepatocyte
   respiratory capacity. These findings suggest that the disruption of 1C
   homeostasis, rather than glycine availability, may be the underlying
   cause of hepatic pathogenesis in SHMT2^HKO mice.

   Depleting BHMT or methionine adenosyltransferase 1 A (MAT1A) in mice
   lowers methylation potential, resulting in NAFLD^[131]6,[132]34. This
   is due to the impaired SAM-dependent synthesis of phosphatidylcholine
   (PC), which causes defective very-low-density lipoprotein (VLDL)
   assembly in the liver and, thus, fat buildup. As deleting SHMT2 reduces
   SAM levels and disrupts liver methylation potential, further research
   is necessary to confirm whether SHMT2-mediated 1 C synthesis is
   necessary to maintain liver PC synthesis, VLDL assembly, and fat
   disposal.

   Our research on the diet-induced development of non-alcoholic
   steatohepatitis (NASH) revealed that SHMT2 disruption can have both
   positive and negative effects on liver health. While it worsened
   steatosis, it could also aid in reducing liver inflammation and
   fibrosis in cases with an inflammatory and fibrogenic diet. SHMT2
   disruption increased the hepatic DNL pathway due to the derepression of
   DNL genes when there was an overload of lipids. Sterol regulatory
   element binding proteins (SREBPs) recognize nutrient signals and mature
   in the ER to activate lipid synthesis genes^[133]37. It is critical to
   maintain the integrity of the ER membrane, which requires adequate PC
   content to prevent excessive SREBP activation^[134]38. Thus, the
   increase in liver DNL genes may be due to lower SAM-dependent PC
   synthesis, decreased ER PC content, and subsequent SREBP derepression
   in SHMT2 deficient mice. Further research is necessary to determine
   whether SHMT2 disruption may worsen fatty liver under AMLN diet
   treatment by decreasing VLDL-mediated hepatic lipid disposal and
   increasing SREBP-mediated hepatic lipid synthesis.

   A significant discovery of this study indicates that SHMT2 is necessary
   for developing liver inflammation and fibrosis. Despite having a
   significant increase in liver steatosis and injury, mice lacking SHMT2,
   when given the AMLN diet, had reduced collagen deposition and immune
   cell infiltration in their livers, as revealed in the histological
   analyses, as well as lower levels of inflammation and fibrosis-related
   genes. The SHMT2^HKO mice showed increased lipogenesis that contributed
   directly to liver steatosis and injury. However, this increase was not
   compatible with reduced inflammation and fibrogenesis, which are
   typically stimulated by liver steatosis and injury. It is possible that
   additional SHMT2 activity is required to activate the liver’s immune
   and fibrogenic response to steatosis and injury. Liver injury triggers
   inflammation and fibrosis by instigating interaction between
   hepatocytes, immune cells, and fibrogenic cells^[135]39. The animal
   model in the current study specifically targeted SHMT2 in the
   hepatocytes of the mouse liver, highlighting the importance of
   communication between these cells. This communication may involve
   serine catabolic activity derived from hepatocytes, which supports the
   activation and expansion of immune and fibrogenic cells in the liver,
   likely by generating methyl folate species and SAM. These metabolites
   are crucial for nucleotide synthesis and transmethylation reactions,
   which support cell proliferation and epigenetic gene regulation.
   Further research is necessary to determine if the serine-derived 1 C
   activities from hepatocytes regulate the metabolism or signaling of
   immune and fibrogenic cells in supporting their proliferation and
   functional expansion.

   A significant association was found between the SHMT2 genetic
   variation, its expression, and increased susceptibility to NAFLD and
   MetS in human populations, as well as SHMT2’s stage-specific regulation
   of the onset and progression of NAFLD in the model animals. These
   findings suggest that precision medicine strategies are necessary when
   considering therapeutic interventions for serine catabolism in NAFLD at
   different stages, despite its significant value as a therapeutic target
   for the disease. It is also important to be cautious about interfering
   with SHMT2 during pregnancy because it is essential for embryonic
   development.

Methods

Animal studies

   All studies conducted were authorized by the Institutional Animal Care
   and Use Committee of Wayne State University. We have complied with all
   relevant ethical regulations for animal use. Mice were maintained under
   a 12-hour day-night cycle and fed ad libitum with a control chow diet
   (#D12450K, Research Diets) or an AMLN diet (#D17010103, Research
   Diets). The SHMT2^fl/fl strain was generated in the C57BL/6 J
   background by flanking exon 3–12 of the SHMT2 gene with loxP sites
   using CRISPR/Cas9-mediated genome editing (Cyagen). The Albumin-Cre
   strain in the C57BL/6 J background was obtained from the Jackson
   Laboratory (Strain# 003574). The SHMT2^HKO strain was generated by
   crossbreeding the SHMT2^fl/fl with the Albumin-Cre strain.

   In the standard experiment protocol, male mice were weaned at three
   weeks of age and began consuming a chow diet. The mice were either kept
   on a chow diet or switched to the AMLN diet at 4 weeks of age. They
   were then exposed to the AMLN diet for 18 more weeks until they were 22
   weeks old. GTT and ITT analyses were performed at 18 and 20 weeks of
   age, respectively. Body composition analysis was performed at 22 weeks
   of age on an EchoMRI instrument. For tissue harvesting, mice were
   fasted for 5 h. Blood samples were drawn from the inferior vena cava of
   the anesthetized animals for biochemical analysis. Dissected liver
   tissues were either flash-frozen in liquid nitrogen for biochemical
   analysis or fixed in 10% neutral buffered formalin for histological
   analysis.

Mouse primary hepatocyte isolation and cell culture

   To obtain primary mouse hepatocytes, a two-step collagenase perfusion
   method was used. The procedure involved opening the abdominal cavity
   and cannulating the inferior vena cava (IVC) with a catheter. The
   portal vein was incised to allow sufficient outflow during perfusion.
   The liver was perfused with buffer A (calcium-free HBSS with 25 mM
   HEPES and 0.5 mM EDTA) for 3 min, followed by buffer B (low-glucose
   DMEM with 0.4 mg/mL collagenase IV) for 5 min at a rate of 8–10 ml/min.
   After perfusion, the liver was transferred to a sterile Petri dish and
   gently shaken to disperse the hepatocytes. The isolated hepatocytes
   were filtered using a 70 μm strainer and purified by Percoll density
   gradient centrifugation. The purified hepatocytes were plated on
   collagen-coated dishes, incubated in DMEM with 10% FBS for 4 h, and
   then maintained in DMEM without FBS until treatment.

   Murine AML12 hepatocytes (ATCC) were maintained in a 1:1 mixture of
   DMEM and Ham’s F12 medium, supplemented with 10% FBS, 1:100 ITS-G
   (Invitrogen), 100 U of penicillin/ml and 0.1 ng of streptomycin/ml. The
   AML12 cell line with a SHMT2-knockout was described previously^[136]29.
   The oxygen consumption rate was measured on a Seahorse XF analyzer
   (Agilent).

Serum amino acid and blood biochemistry tests

   The serum was prepared by centrifugation of the clotted blood at
   10,000 g, 4 °C for 10 min. The serum amino acid levels were measured on
   an LC-MS/MS instrument (Waters) equipped with Cortec UPLC C18 column
   and Xevo TQ-S mass spectrometer using Kairos Amino Acid 100 Kit
   (Waters) as reference standards. Serum triglycerides and total
   cholesterol levels were measured using commercial kits from Wako and
   Pointe Scientific. Serum ALT was measured using the commercial kit from
   BioAssay Systems.

Liver metabolite extraction and targeted metabolomics

   For the isolation of hydrophilic metabolites, liver tissues were
   homogenized in 80% aqueous methanol by sonification. After
   centrifugation clarification, the supernatants were lyophilized on a
   SpeedVac vacuum concentrator. The metabolite extract pellets were
   stored at −80 °C until use. The metabolite levels were measured on an
   AB SCIEX LC-MS/MS system equipped with a SHIMADZU Nexera ultra
   high-performance chromatography system and a hybrid triple quadrupole
   and linear ion trap mass spectrometer, as described previously^[137]40.

   The metabolite abundance data were subjected to median normalization,
   log transformation, mean-center scaling, fold change analysis and
   two-sample t-tests using the MetaboAnalystR package available on the
   MetaboAnalyst 5.0 metabolomics analysis platform
   ([138]https://www.metaboanalyst.ca/MetaboAnalyst/). The differentially
   expressed metabolites (DEMs) were identified using the significance
   cutoff of p < 0.05 and FDR < 0.20. The DEM pathway enrichment analysis
   was performed on MetaboAnalyst 5.0 using the Mus musculus SMPDB
   database. The results were visualized and plotted with GraphPad Prism
   9.

Liver lipid extraction and measurement

   For lipid extraction, 50 mg of liver tissues were homogenized in 1 ml
   methanol by sonication on ice using a Branson Digital Sonifier. The
   homogenates were mixed with an additional 330 ul methanol and 2.66 ml
   chloroform by vertexing and incubated at 4 °C for 4 h, and then mixed
   with 1.6 ml 0.9% NaCl by vertexing followed by centrifugation at
   4500 g, 4 C for 15 min. The organic phase was collected and evaporated
   under nitrogen flow. The lipid extracts were resuspended in isopropanol
   with 2%Triton. Liver triglycerides and total cholesterol levels were
   assessed using commercial kits obtained from Wako and Pointe
   Scientific, respectively.

Glucose tolerance test

   After a period of overnight fasting, mice were given an injection of
   glucose (2 g/kg body weight) into their peritoneal cavity. Blood
   samples were then taken from their tail veins at 0, 15, 30, 60, 90, and
   120 min after the injection. These samples were tested for glucose
   concentration using a glucometer (Contour Next).

Insulin tolerance test

   After a 5-h period of fasting, insulin was administered to mice via an
   intraperitoneal injection at a dosage of 0.65U/kg of body weight. Blood
   samples were obtained from the tail vein at intervals of 0, 15, 30, 60,
   and 90 min post-insulin injection. The blood glucose levels were
   measured using a glucometer (Contour Next).

Histological analysis

   For H&E and Sirius Red staining, liver tissues were fixed in 10%
   neutral buffered formalin overnight, dehydrated into 70% ethanol, and
   then paraffin-embedded and sectioned. Tissue section slides were either
   stained with hematoxylin and eosin (H&E, ThermoFisher Scientific); or
   stained with Picro-Sirius Red solution containing saturated aqueous
   picric acid with 0.1% Sirius red F3B (Sigma) for 1 h, followed by
   rinses with 0.5% acetic acid.

   For Oil red O staining, formalin-fixed liver tissues were stepwise
   cryoprotected in 15% and 30% sucrose for 4 h and overnight,
   respectively, then embedded with OCT compound (Tissue-Tek) and
   flash-frozen in liquid nitrogen and stored at −80° C. The cryosections
   were fixed with 10% neutral buffered formalin for 30 min, rinsed with
   ddH[2]O and 60% isopropanol, and subsequently stained with 0.35% Oil
   Red O dye (Rowley Biochemical) for 15 min, followed by rinse in 60%
   isopropanol and counterstaining for nuclei with alum Hematoxylin.

   The tissue section images were captured on a BZ-X800 microscope
   (Keyance) and scored using the following NAFLD histological scoring
   system: steatosis grade (0, <5%; 1, 5–33%; 2, >33–66%; 3, >66%),
   lobular inflammation (0, No foci; 1, <2 foci per 200x field; 2, 2–4
   foci per 200x field; >4 foci per 200x field), and fibrosis stage (0,
   None; 1, perisinusoidal or periportal; 2, perisinusoidal and
   portal/periportal; 3, bridging fibrosis)^[139]41.

Western blotting

   The tissues or cultured cell pellets were homogenized in RIPA lysis
   buffer (Tris-HCl pH7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium
   deoxycholate, 0.1% SDS) by sonication on a Branson Digital Sonifier.
   Protein concentrations were determined with the BCA reagent (Pierce).
   Protein lysates (40 ug) were resolved by sodium dodecyl
   sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and proteins
   were transferred onto nitrocellulose filters. The blots were saturated
   with 5% milk and probed with antibodies against SHMT2 (#HPA020543,
   Sigma, 1:1000), NDUFS1 (sc-271510, Santa Cruz, 1:1000), Total OXPHOS
   Antibody Cocktail (ab110411, Abcam, 1:1000), ACLY (#15421-1-AP,
   Proteintech, 1:1000), ACC1 (#21923-1AP, Proreintech, 1:1000), FASN
   (#10624, Proteintech, 1:1000) or β-actin (A2066, Sigma, 1:1000).
   Following a wash with PBST (PBS containing 0.1% Tween 20), the blots
   were incubated with peroxidase-coupled goat anti-rabbit immunoglobulin
   G (Sigma, 1:5000). The immunolabeled protein bands were detected by
   enhanced chemiluminescence (ECL) method (Perkin Elmer) on an Azure C600
   imager.

RNA-seq, Bioinformatics, and qRT-PCR analysis

   For RNA-seq transcriptomics profiling, total RNAs were isolated from
   liver tissues using RNeasy kit (Qiagen). The library construction and
   high throughput sequencing were conducted at Azenta Life Sciences. The
   cDNA libraries were constructed with the NEBNext Ultra RNA Library Prep
   Kit for Illumina (New England Biolabs), quantified with a Qubit 2.0
   Fluorometer (Life Technologies) and a TapeStation (Agilent), and
   sequenced on a HiSeq 4000 sequencer (Illumina) using 2x150bp paired-end
   method. Base calling and FRSTQ file generation were achieved with the
   HiSeq Control and bcl2fasq 2.17 software. Sequence reads were trimmed
   to remove adapter sequences and poor-quality nucleotides using
   Trimmomatic v.0.36 and then mapped to the Mus musculus GRCm38 ERCC
   reference genome using the STAR aligner v.2.5.2b, resulting in the BAM
   files with >25 million total mapped reads per sample. The unique gene
   hits were extracted using EnsDb.Mmusculus.v79:Ensembl based annotation
   from R package, further normalized using edgeR and subjected to
   limma-voom-transformation and t-statistics calculation using the eBayes
   function of the limma R package to identify the differentially
   expressed genes (DEGs) with the significance cutoff of p < 0.05, adj
   p < 0.25. Gene set enrichment analysis (GSEA) were performed with the
   setting of m2.cp.wikipathways.v2023.1.Mm.symbols.gmt, 1000
   gene_set-based permutations, chip
   platform = Mouse_Gene_Symbol_Remapping_MSigDB.v2023.1.Mm.chip, weighted
   enrichment statistic, and metric for ranking genes = tTest, and the
   significance cutoff of p,0.05 or FDR < 0.25.

   For qRT-PCR analysis, total RNAs were isolated from liver tissues with
   the TRI Reagent (Sigma). cDNA libraries were constructed by random
   priming using the High-Capacity cDNA Reverse Transcription Kit (Applied
   Biosystems) and then used as a template for qPCR amplification with
   Luna Universal qPCR Master Mixes (New England Biolabs) on an AriaMx
   cycler (Agilent). The mRNA levels were determined as the delta-delta
   threshold cycle (ΔΔC[T]) and normalized to the PPIA mRNA level. The PCR
   primers used are listed in Table [140]S3.

Human data analysis

   The associations between SHMT2 common genetic variants and metabolic
   phenotypes in European Ancestry descendants were retrieved from the
   Common Metabolic Diseases Knowledge Portal (CMDKP,
   [141]https://hugeamp.org/). The statistical analysis of the gene-level
   phenotypic associations was performed using the Multi-marker Analysis
   of GenoMic Annotation (MAGMA) method, available on the Portal platform,
   to calculate the bottom-line genetic associations.

   For meta-analysis of SHMT2 gene expression, the mRNA profiling for
   liver samples of healthy human donors (n = 38) and NASH patients
   (N = 35) were retrieved from the NCBI Gene Expression Omnibus under the
   accession of [142]GSE89632^[143]42 and [144]GSE126848^[145]43. The
   unique gene counts were filtered, and subjected to TMM normalization
   and logCPM transformation with edgeR. The resulting SHMT2 expression
   values (log-transformed) were used to calculate the standardized mean
   difference (SMD, Hedge’s g) between the groups of NASH and healthy
   control and presented as a forest plot using the random-effects model
   of the metafor R package.

Statistics and reproducibility

   All figures display individual data points representing single readings
   for each sample. Sample sizes and error management are specified in the
   figure legends.

Reporting summary

   Further information on research design is available in the [146]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [147]Peer Review File^ (4.7MB, pdf)
   [148]Supplementary Information^ (1.8MB, pdf)
   [149]Description of Supplementary Materials^ (15KB, docx)
   [150]Supplementary Data 1^ (33.8KB, xlsx)
   [151]Supplementary Data 2^ (22.7KB, xlsx)
   [152]Supplementary Data 3^ (1.6MB, xlsx)
   [153]Supplementary Data 4^ (472.6KB, xlsx)
   [154]Supplementary Data 5^ (53.7KB, xlsx)
   [155]reporting summary^ (1.5MB, pdf)

Acknowledgements