Dietary shift leads to venous thrombosis-induced congestive liver failure in CBS-deficient mice

Summary

Cystathionine beta-synthase (CBS) deficiency is an inborn error of metabolism that results in a large increase in plasma total homocysteine (tHcy) and a significant risk of venous thrombosis. Although a mouse model of CBS deficiency (Tg-I278T Cbs^−/−) has several phenotypes in common with human patients, it has not been shown to have elevated thrombosis risk. Here, we describe a novel phenotype in which 40% of Tg-I278T Cbs^−/− mice die of liver failure due to hepatic vein thrombosis shortly after being shifted from a low methionine diet (LMD) to a regular diet (RD). Importantly, no deaths or thromboses occurs if the mice are continuously maintained on RD or LMD for extended periods of time. RNAseq analysis of the livers of Tg-I278T Cbs^−/− mice that were shifted to RD for three days after spending one week on LMD (RD3D), shows significant differences in many transcripts involved in coagulation and fibrinolysis, key processes involved in thrombosis. Interestingly, the liver gene expression profile and serum amino acid profiles of both Tg-I278T Cbs^−/− and Tg-I278T Cbs^+/− mice maintained continuously on RD is also significantly different from RD3D mice. Since the only difference between RD and RD3D mice is their previous exposure to a LMD diet, this shows that the liver transcriptional profile is affected not only by the current diet, but also the animals’ previous dietary history. Overall, our findings indicate that there is a strong gene-diet interaction between the Cbs genotype and dietary methionine, and that this interaction may help explain the thrombosis phenotype in human CBS deficient patients.

Introduction

Human CBS deficiency (also known as classical homocystinuria or HCU) is the most common inborn error of methionine metabolism in humans with an estimated frequency of about 1:100,000 births (Mudd et al 1985). Biochemically, it is characterized by a 10-fold or more elevation in plasma total homocysteine (tHcy), the total of oxidized and reduced forms of homocysteine, and elevated plasma methionine. Clinical phenotypes of CBS deficiency are observed in the vascular, optic, skeletal, and nervous systems. The most common cause of morbidity and death in patients comes from the risk of thrombosis. Blood clots can develop in a variety of vessels, but most presenting patients have venous thrombosis in either peripheral veins (deep vein thrombosis) or in veins in the brain (cerebral venous sinus thrombosis) (Yap et al 2001; Karaca et al 2014; Wen et al 2020; Ochoa-Ferraro et al 2021). The relative risk of thrombosis in CBS deficient patients is 20–1000 times higher than the general population (Kozich et al 2021). Despite much research, the mechanism by which CBS deficiency leads to elevated risk of thrombosis is still unknown (Jacobsen 2000; Undas et al 2005). However, it has been definitively shown that treatments that lower tHcy in CBS deficient patients reduce thrombosis risk (Yap et al 2001). The current treatment strategies to lower tHcy involve using a combination of dietary methionine restriction and supplementation with certain B-vitamins (folic acid and pyridoxine) that stimulate the biochemical reactions that convert homocysteine back to methionine (Morris et al 2017). While this approach can reduce clinical complications of CBS deficiency, dietary restriction is difficult for patients to maintain (Morrison et al 2021).

Homozygous Cbs^−/− mice are born in mendelian ratios but die of liver failure around three weeks of age (Watanabe et al 1995; Gupta et al 2021), making them difficult to study. To circumvent this issue, our lab created the Tg-I278T Cbs^−/− mouse model (Wang et al 2005), which is homozygous for a knockout of the mouse Cbs gene and contains a transgene that expresses a human mutant CBS protein (p.I278T), which is the most common mutation found in patients of European descent (Skovby et al 2010). The transgene is under control of a zinc-inducible mouse Mt1 promoter, allowing levels to be modulated by the presence of ZnSO₄ in drinking water. The inducible transgene allows the mice to survive the neonatal lethality associated with Cbs^−/−, yet still have extremely elevated serum tHcy concentrations (~290 μM vs. ~3 μM for Cbs^+/+).

Tg-I278T Cbs^−/− mice have several phenotypes including a 20% decrease in median lifespan, age-dependent facial alopecia, reduced fat mass, reduced bone mass, and eye changes (Gupta et al 2009) ((Majtan et al 2018). Metabolically they show highly elevated levels of serum tHcy and modest elevations in serum methionine. Surprisingly, Tg-I278T Cbs^−/− maintained on standard mouse chow do not show an increase in thrombosis or abnormalities in blood clotting (Dayal et al 2012), even though this is the major clinical issue in humans. However, the alopecia, fat mass, bone mass, and eye phenotypes of Tg-I278T Cbs^−/− mice can be entirely reversed by feeding the mice a diet that contains 1/8 the methionine found in standard mouse chow indicating that the observed phenotypes in the mice are a result of the elevated tHcy and methionine (Gupta et al 2014; Majtan et al 2018).

Here we describe a serendipitous finding in which adult Tg-I278T Cbs^−/− that had been previously shifted from standard diet to a low methionine diet, will die of congestive liver failure when put back on standard diet. We present evidence that the congestive liver failure is due to thrombosis in the hepatic vein, and that gene expression profiles of shifted mice show large alterations in transcripts involved in hemostasis. Our findings in mice suggest that the thrombosis risk in CBS deficient patients may be influenced by protein dietary fluctuations in combination with elevated plasma homocysteine.

Results

Tg-I278T Cbs^−/− mice die shortly after shift from low-methionine to regular diet.

While testing a novel treatment to lower plasma tHcy in a mouse model of CBS deficiency, adult Tg-I278T Cbs^−/− mice were shifted from a regular diet (RD) to a low methionine diet (LMD) for one week before they were given either drug (CDX-6502) or placebo (Skvorak et al 2023). One day after the treatment, venous blood was collected, and the mice were put back on RD with the intention of performing additional experiments in the future. The surprising finding was that about 40% the Tg-I278T Cbs^−/− mice died 3–6 days after the dietary shift back to regular diet, with mice dying at equal rates in both the treatment and the placebo group (Figure 1A.). Previous studies have shown that unexpected deaths in adult Tg-I278T Cbs^−/− mice under one year of age that have been maintained on either RD or LMD chow rarely occur (Gupta et al 2009; Gupta et al 2014). The observation that both the drug-treated and untreated animals showed similar death rates strongly suggested that the death was not related to drug toxicity.

Figure 1.

Sudden death of Tg-I278T Cbs^−/− after dietary shift to RD from LMD. A. Schematic at top shows experimental design. Survival curve shown below. B. Representative photos of livers of mice on indicated diets. Top row of each genotype shows necropsied livers, second row shows H and E staining at 100x magnification, bottom row shows IHC staining with Caspase 3 antibody. C. Bar charts show quantitation of levels of steatosis, hepatocellular changes, and Caspase 3 staining using a 0–3 scale as assessed blindly by Dr. Cai. Error bars show SE. *P<0.05, ***P<0.001.

Dietary shift in Cbs^−/− mice causes massive liver damage due to thrombosis of the hepatic vein.

To explore the sudden death in more detail, we fed adult Tg-I278T Cbs^−/− and Tg-I278T Cbs^+/− mice one of three different nutritional regimes: (1) RD only, (2) RD followed by LMD for seven days, and (3) RD followed by a shift to LMD for seven days, followed by a return to RD for three days (referred to RD3D). The three-day time point was chosen as this was the earliest time at which we had observed morbidity in the previous study. At the end of the experiment, mice were euthanized, and livers and serum were collected for analysis. Tg-I278T Cbs^−/− mice that were shifted to LMD showed significantly increased amounts of steatosis and exhibited some hepatocellular changes including diffuse vascular swelling, the presence of hyper chromatic nuclei, and an increase in eosinphil infiltration (Figure 1B.). However, RD3D mice had much more dramatic alterations. Most notably, the livers in Tg-I278T Cbs^−/− RD3D mice were much lighter than control livers and had a mottled, punctate appearance, and showed significant amounts of staining for the apoptosis marker Caspase 3. These differences were not observed in control Tg-I278T Cbs^+/− RD3D mice, demonstrating that these effects were due to a combination of the dietary changes and the Cbs genotype.

The pathology observed in the Tg-I278T Cbs^−/− RD3D mice resembles the so-called “nutmeg” liver found in humans and mouse models of hepatic venous congestion (Simonetto et al 2015; Li et al 2018). In humans, this can occur by blockage in the hepatic vein due to thrombus formation (called Budd-Chiari syndrome)(Menon et al 2004). Blockage of the hepatic vein causes an increase in sinusoidal pressure and concomitant liver damage. Consistent with this idea, Tg-I278T Cbs^−/− RD3D mice had large elevations in the serum liver damage markers alanine amino transferase (ALT) and aspartate amino transferase (AST) (Figure 2A and 2B). No difference was observed in serum tHcy or methionine levels in RD vs RD3D Tg-I278T Cbs^−/− mice (Figure 2C and 2D), Interestingly, while the dietary shift in the control Tg-I278T Cbs^+/− mice did not result in death or noticeable liver pathology, we did observe increases in ALT and AST, suggesting that the dietary shift caused some liver damage in mice with CBS activity.

Figure 2.

Serum marker measurements for indicated genotype and diet. A. Alanine amino transferase (ALT). B. Aspartate amino transferase (AST). C. Total homocysteine (tHcy). D. methionine. Statistically different comparisons as determined by ANOVA are shown as indicated.

To further confirm these results and rule out the possibility that these effects might be secondary to changes in the gut microbiome, we repeated the experiment but added Cefoperazone (0.5 μg/ml) to the mouse drinking water for ten days before and during the entire experimental period (Figure 3A). This treatment has been shown to cause a 99.7% decrease in colonic bacteria in C57BL6 mice after 8 days (Bongers et al 2022). Fourteen Tg-I278T Cbs^−/− mice were then shifted to LMD for 7 days, and then shifted back to RD. On days 3 and 4 after shift, one animal died, and four others were judged morbid and euthanized (36% death rate). The other 9 mice survived and were followed until euthanized at day 9, along with the control RD group. Necropsies were performed on all animals, in which we carefully examined where the inferior vena cava merges with the hepatic vein. In four of the five morbid animals we observed visible blood clots in this region (Figure 3B), while none were observed in the regular diet or the survivor mice. Histopathology confirmed the presence of a clot in one of the mice (Figure 3C; other clots were lost during the fixation procedure). In this experiment tHcy, methionine, and ALT was measured at the end of the LMD period, 24 hours after the shift to RD, just before sacrifice in the morbid animals, and at day 9 for the survivors (Figure 3D). Even at the 24-hour time point, we did not see any evidence that shifting from LMD to RD was causing a spike in tHcy, suggesting that the change in the liver is not due to an overshoot of tHcy and methionine when the LM diet was changed back to RD. Thus, these studies support the idea that the observed liver failure in Tg-I278T Cbs^−/− mice is due to blood clot formation in the hepatic vein.

Figure 3.

Dietary shift in presence of anti-biotic. A. Experimental Scheme. Down red arrows show serum collection times. B. Representative necropsy photos of Tg-I278T Cbs^−/− mice on indicated diet. IVC is the inferior vena cava. C. H and E-stained section showing IVC cross section with a large blood clot (green arrow) adjacent to the vascular endothelium (yellow arrow) in a RD3D mouse. D. Serum tHcy, methionine, and ALT.

Serum Amino Acid Analysis.

In addition to tHcy and methionine, we also examined the serum concentrations of 26 other amino acids or amine containing compounds in all six groups (Supplemental Figure 1). One-way ANOVA analysis revealed that all the amino acids were significantly different in at least one of the six groups (Supplemental Data 1). Particularly interesting was the finding that many amino acids were significantly changed between the RD and the RD3D condition, even though the mice were all on RD for at least three days before blood was collected. For example, serum GABA-levels appear greatly reduced in RD3D condition compared to RD in both Tg-I278T Cbs^+/− and Tg-I278T Cbs^−/− mice. Conversely, phenylalanine levels were increased in the same comparison. For some amino acids this “dietary history” effect was dependent on the Cbs genotype. For example, in Cbs^+/− mice serine levels go up dramatically when the mice are on LMD and go back down when put back on RD (RD3D), but in Tg-I278T Cbs^−/− mice serine goes up and stays elevated. Overall, the data suggests that serum amino acid levels reflect not just the current diet, but also the animals “dietary history” and genotype.

To examine this more systematically, we performed 2-way ANOVA to estimate the variance of serum amino acid concentration attributable to genotype, diet, and interaction effect (Supplemental Data 1). The amount of variance that could be explained by these three factors ranged from a high of 88.3% (glycine) to a low of 13% (histidine), with the average being 47.3%. Significant interaction effect between genotype and diet (>10% of the variance) were observed for seven compounds: serine, alanine, asparagine, glutamate, threonine, tyrosine, homocysteine, and ornithine.

With regards to the thrombosis phenotype in the mice, we observed that RD3D mice had a 35% reduction in serum arginine levels. Arginine is converted to nitric oxide (NO) by endothelial nitric oxide synthetase, and this reaction plays an important role in regulating various hemostatic parameters including blood pressure and platelet aggregation (Toda and Toda 2011). However, it is important to note that the difference serum arginine was observed in both Cbs−/− and Cbs+/− mice, so at best it is a contributing factor.

Liver gene expression profiles affected by transient exposure to low methionine diet.

We performed RNA-Seq analysis on livers from RD, LMD and RD3D mice of both genotypes (Tg-I278T Cbs^+/−, Tg-I278T Cbs^−/−) to see how diet and the Cbs genotype affected global gene expression. We initially examined differential gene expression for each pair of dietary treatments (RD vs LMD, LMD vs RD3D, RD vs RD3D) within each genotype. Volcano plot analysis reveals that dietary effects of each of the dietary treatments had a much larger effect on the gene expression profile in Tg-I278T Cbs^−/− mice than on Tg-I278T Cbs^+/− mice (Figure 4A). In control Tg-I278T Cbs^+/− mice, the effects of the dietary shifts were relatively modest, with only 75 transcripts showing a 2-fold difference in expression in at least one of the conditions (Figure 4B, Supplemental Data 2). Over-representation analysis (ORA) and gene set enrichment analysis (GSEA) indicated that transcripts that were altered in response to the dietary shifts included genes involved in fatty acid and cholesterol biosynthesis, PPAR gamma signaling and adipogenesis (Supplemental Figure 2). Interestingly genes that were significantly up or downregulated by LMD, tended not return to the levels found before the animals were shifted (Figure 4C). On average, genes that were decreased only recovered about 33% of their expression difference when shifted back, while genes that were induced were still about 50% elevated. We also observed four transcripts (Pnpla3, Fasn, Mtmr11, Osgin1) that were either up or down regulated after one week on LMD, but in the RD3D condition were even more differentially expressed relative to RD. For the transcripts that were identified as differentially expressed between RD and RD3D, 21 out of 25 had a larger difference between RD and RD3D than between RD and LMD (Figure 4D). These findings support the idea that exposure to LMD for seven days is sufficient to cause alterations in the expression of some genes that can persist or even get more severe for at least three days after the RD diet is restored.

Figure 4.

RNAseq analysis. A. Volcano plots of differences between mice on indicated diets overlayed by Cbs genotype. Black shows differences for Tg-I278T Cbs^+/− and red shows differences for Tg-I278T Cbs^−/− mice. B. Venn diagram showing number of genes that have statistically significant differences in transcript levels between indicated pairs. Criteria for selection was >1.5x difference in expression and p<0.015. C. Examination of behavior of transcripts significantly identified as either up- or down-regulated between RD and LMD in Tg-I278T Cbs^+/− mice. All transcripts level were normalized to one on RD. Median and 95% confidence interval are shown by lines. Transcripts that were more differentially regulated in the RD vs. RD3D condition compared to the RD vs. LMD condition are in green with names on the side. D. Same as C except chosen genes were the most differentially regulated between RD vs RD3D conditions.

Focusing on the Tg-I278T Cbs^−/− mice, we found much larger differences in the expression profiles which resulted in the different dietary treatment groups clustering in three distinct groups based on PCA analysis (Figure 5A). Using a threshold of being at least 2-fold up or down regulated and P<0.01, we identified a total of 1966 transcripts whose level was altered in at least one of the three possible comparison groups (Figure 5B, Supplemental Data 2). We identified 139 transcripts that were significantly changed in RD versus LMD. GSEA and ORA analysis reveals that these transcripts are enriched in genes involved in various aspects of lipid metabolism and drug conjugation (Supplemental Figure 3). Seven of the differentially regulated transcripts from this group (Fasn, Pnpla3, Acss2, Cyp26a1, Cyp2b10, Srebf1, and Slc22a7) were also observed as differentially expressed in Tg-I278T Cbs^+/− mice, but in four of the seven the direction of change was opposite (Fasn, Pnpla3, Acss2, Srebf1, Supplemental Figure 4). Interestingly, these four genes are all key players in lipogenesis. This finding is consistent with our previous studies showing that Tg-I278T Cbs^−/− mice increase fat mass on LMD as opposed to Tg-I278T Cbs^+/− mice which lose weight (Gupta et al 2014).

Figure 5.

Analysis of Tg-I278T Cbs^−/− RNA seq. A. Principle component analysis. Red=RD; green=LMD; blue=RD3D. B. Venn diagram showing number of genes that have statistically significant differences in transcript levels between indicated pairs. Criteria for selection was >2-fold difference in expression and p<0.01. B. GSA and GSEA analysis of RD vs RD3D mice using the KEGG database.

In Tg-I278T Cbs^−/− mice we observed the largest gene expression profile differences in comparing either the RD or LMD group to the RD3D group (Figure 5B). ORA using both the KEGG and WikiPathways databases found enrichment in several different amino acid metabolic pathways, cholesterol metabolism, steroid hormone metabolisms, adipogenesis, drug metabolism, and coagulation pathways (Figure 5C, Supplemental Figure 5). GSEA found several pathways related to inflammation that were upregulated in RD3D mice such as IL-17 signaling, cytokine-cytokine interaction (KEGG), spinal cord injury, and chemokine signaling. Down regulated pathways include tryptophan metabolism, drug metabolism, steroid hormone metabolism, and eicosanoid metabolism.

Dysregulation of coagulation and fibrinolysis pathways in RD3D Cbs^−/− mice.

Given the hepatic vein blockage-phenotype of Tg-128T Cbs^−/− RD3D mice, we were particularly intrigued that the complement and coagulation cascade had been identified as being overrepresented in the ORA analysis. Examination of the 88 genes in the complement and coagulation gene set (mmu04610, KEGG database) reveals that 33 of these genes were significantly (p<0.05) up or down regulated in the RD3D compared to RD mice (Figure 6a). Six genes were significantly up-regulated in RD3D vs RD, while 27 were down regulated. The most up-regulated transcript (30-fold) was the Serpine1 gene, which encodes plasminogen activator-1 (PAI) protein, which is a key regulator of fibrinolysis. This finding was also confirmed at the protein level (Figure 6b). Other up regulated transcripts include the plasminogen activator receptor (Plaur), endothelial protein C receptor (Procr), thrombomodulin (Thbd), and coagulation factor XIII (F13a1). All these upregulated genes are involved in either the formation or breakdown of fibrin in blood clots. Down-regulated genes included the transcripts for clotting factors (Klkb1, Kng1, Kng2, Proc, F2, F7, F9, F10, F12, and F13b), serine protease inhibitors (Serpine1a, 1b, 1c, 1d, c1, g1, and d1), and complement cascade genes (c1ra, c1s1, c8b, c8g, c9, cD59a, cFh, c5ar1, cD59b, cfd). It should be noted that none of these genes were differentially affected by diet in the Tg-I278T Cbs^+/− mice, implying that the dietary-shift has a strong effect on the coagulation/fibrinolysis pathways only in the context of Cbs^−/− genotype.

Figure 6.

RNA seq analysis of complement and coagulation pathway genes. A. Heat map of all 33 genes that were significantly changed in RD vs. RD3D Tg-I278T Cbs^−/− mice. B. Serum PAI-1 (Serpine1) protein levels assessed by ELISA. C. Basic coagulation/fibrinoysis pathway with altered transcripts shown as indicated. Red arrows down show predicted pro-bleeding changes, while blue up arrows show predicted pro-clotting changes.

The overall impression from the transcription changes is relative to the canonical coagulation/fibrolytic pathway (Figure 6c), is that there is a reduction of activity in the steps before thrombin generation and an increase in activity in the steps after.

Discussion

Here we describe the unexpected observation that Tg-I278T Cbs^−/− mice die at a high frequency between three and six days after they are shifted to RD from a LMD. This mortality is caused by congestive liver failure due to the formation of a blood clot at the base of the hepatic vein. This effect is only observed in Tg-I278T Cbs^−/− animals undergoing the dietary shift from LMD to RD and is not observed in unshifted animals maintained either on RD or LMD. In addition, Tg-I278T Cbs^+/− mice that are shifted from LMD to RD also do not show this phenotype, implying that both the genotype and the dietary shift are required for manifestation of the phenotype. Consistent with this idea, is an experiment reported by Majtan and colleagues in which a different mouse model of HCU that was treated with a pegylated-hCBS protein did not show any adverse phenotype when shifted from MRD to RD (Park et al 2020). The finding described here is an unusually powerful example of gene-diet interaction resulting in a rapid onset death phenotype.

Liver RNAseq analysis of RD3D Tg-I278T Cbs^−/− mice shows that the hepatic vein phenotype is associated with dysregulation of many genes that code for factors involved in blood clotting and fibrinogen homeostasis. Nearly 38% of the transcripts identified in the KEGG database complement and coagulation cascade are significantly dysregulated in RD3D Tg-I278T Cbs^−/− mice but are not dysregulated in unshifted or in RD3D Tg-I278T Cbs^+/− mice. Significant downregulation of several transcripts that code for known clotting factors, including Factors XII, IX, X, II, are observed in the livers RD3D Tg-I278T Cbs^−/− mice. Germline mutations in these factors cause increased bleeding, resulting in longer clotting times, larger clots and increased risk of thrombosis (Lippi et al 2012). In humans, Serpina1 (note there are multiple paralogs in mice), Serpinc1, and Serpind1 can all inhibit Factor Xa and Factor IIa and deficiency of each is associated with venous thromboembolism (VTE) and/or deep vein thrombosis (Grover and Mackman 2022). Genetic deficiency of the human ortholog of Serping1 causes hereditary angioedema (HAE), a life-threatening condition of bradykinin induced swelling. A registry study of HAE patients found they had a 4.3-fold increased risk of VTE. In addition, the observed up-regulation of PAI1 (Serpin1e) and F13a would be predicted to stabilize fibrinogen, making clots more stable and increasing the potency of the thrombus. Overall, the data supports a model in which the death of RD3D Tg-I278T Cbs^−/− mice is caused by dysregulation of clotting and fibrinolysis genes, resulting hepatic vein thrombus formation and subsequent congestive liver failure.

A particularly intriguing feature of this study is that Tg-I278T Cbs^−/− mice must first be exposed to low methionine diet and then shifted to RD for the congestive liver failure phenotype to be manifested. However, TgI278T-Cbs^−/− mice maintained on RD do not have the liver phenotype, despite having identical tHcy and methionine levels. Thus, the difference in behavior is not due to a “rebound” effect with regards to tHcy and methionine levels. Rather, the effect must be caused by some physiological change that occurs as a result of the exposure to a low-methionine diet. The genetic pathways that are the most effected by LMD in the liver of Tg-I278T Cbs^−/− mice are lipid biosynthesis and xenobiotic metabolism (glutathione conjugation and cytochrome P450). This may suggest that elevated tHcy may somehow interact with lipids to induce inflammation and produce a prothrombotic environment.

We believe that this mouse phenotype may be related to the thrombosis that is the major cause of morbidity and mortality in human CBS deficiency. Although Tg-I278T Cbs^−/− mice have many phenotypes that are like the human condition (osteoporosis, loss of fat mass, ocular phenotype), the data concerning thrombosis and clotting have been mixed at best. Tg-I278T Cbs^−/− mice on normal diet did not show any difference in prothrombin time (PT) or activated partial thromboplastin time (aPTT) assays (Gupta et al 2009). In addition, no significant difference in in vivo thrombus formation was observed in the carotid artery, mesenteric arteries, or inferior vena cava (Dayal et al 2012). There has been a report of Tg-I278T Cbs^−/− mice having premature death due to intestinal bleeding when put on high methionine diet, but the time required to manifest a phenotype was quite long compared to the one described here (18 weeks vs 3 days) (Park et al 2020) and the serum methionine levels was nearly 20-fold higher. Interestingly, in the high methionine diet experiments, coagulation factor XI was down-regulated, but in the studies reported here, we did not observe a significant change. There has also been a report of decreased tail bleeding times in a different mouse model of CBS deficiency (Maclean et al 2010; Maclean et al 2012), but tail bleed times are complex and influenced by a variety of other factors. In human CBS patients, thrombosis of the hepatic vein and concomitant liver failure has not been reported, but venous thrombosis in both the brain and periphery have been reported. We suspect that the difference in the location of thrombosis is probably a reflection of basic physiological and biomechanical differences between mice and humans, such as the size of the vessel or the relative velocity of the blood.

How might the dietary shifting be relevant to human CBS deficiency? The standard clinical treatment for human CBS deficiency is to try and keep plasma tHcy as low as possible, as this is associated with reduced risk of thrombosis (Morris et al 2017). This generally involves patients having low protein diets. However, these diets are difficult to maintain, and it is not uncommon for patients to “binge”, especially in social situations. We speculate that this sort of binging behavior may be like the shift back to regular diet in our RD3D animals, and that this may increase the risk of thrombosis in patients. In a controlled setting it might be interesting to see if CBS patients that have recently “binged” on protein have any changes in their hemostatic parameters.

Finally, we think it is fascinating that many of the transcripts whose expression is changed by the shift from RD to LMD, are still altered even when the diet is restored to RD. The fact that this occurred in the control animals (Cbs^+/−) suggests that this effect may not simply be a secondary effect of massive liver damage. Rather, it may reflect some sort of “dietary memory”, in which the liver epigenetic gene expression program is somehow altered because of exposure to LMD, and that this memory persists for some period of time which alters the epigenetic readout when the animals are put back on RD. Mechanistically, this would be consistent with findings that have shown that mice shifted to a low methionine diet show bulk changes in the level of histone modifications and alterations in the distribution of the H3K4me3 mark at the promoters of genes (Mentch et al 2015)(Dai et al 2018). It would be interesting to perform a time course study to see how long these gene expression changes persist. Should they persist for a longer period, it may suggest a novel mechanism for how past dietary behavior might influence current metabolism.

Materials and Methods

Mice and diets

Tg-I278T Cbs^−/− mice (mixed sex, 3–8 months old) were generated as described previously (Wang et al 2005). In brief male Tg-I278T Cbs^−/− mice were mated with female Tg-I278T Cbs^+/− mice in cages with water bottles containing 25 mM ZnSO_4. All pups were genotyped between 10–14 days of age. At the time of weaning (around 30 days), mice were put in new cages with non-zinc water. Mice were fed either a Standard Protein Diet (RD; PicoLab Verified-75IF (5V75), 0.6% methionine; LabDiet Advanced Protocol) or a low Methionine Diet (LMD; TD.110591, 0.05% methionine; Envigo). Average age of mice used in the experiments was 150 days old. All animal procedures were approved by the Fox Chase Cancer Center Institutional Animal Care and Use Committee (IACUC).

Dietary shifts experiments

The initial dietary shift experiments for the cohort of animals shown in Figure 1a is described in (Skvorak et al 2023). For the dietary shift experiments shown in Figure 1b, mice were subjected to one of three different dietary treatments. One group of mice were continuously fed RD. A second group of mice were fed RD and then switched to low methionine diet (LMD) for one week. The final treatment group were sequentially fed RD, switched to LMD for a week and then placed back on RD for 3 days (RD3D). At that time mice showed signs of morbidity, such as decreased activity, panting, and hunched posture.

For the experiment shown in Figure 3, Tg-I278T Cbs^−/− mice were given distilled water with dissolved cefoperazone (0.5 g/L, Thermos Scientific Chemicals, 62893–20–3) for 10 days prior to diet changes. In this experiment, mice kept on RD for 2 weeks as a control group. Other mice were shifted to LMD for 7 days and placed back on RD. When mice showed signs of morbidity, mice were euthanized at either day 3 or day 4 (RD3D/4D), while healthy mice were euthanized at day 9 by isoflurane overdose at the end of the experiments. Blood was collected from a submandibular vein and cardiac puncture.

Histopathology

Tissues (liver, inferior vena cava, hepatic veins) were collected and fixed in 10% phosphate-buffered formaldehyde (formalin), dehydrated, and embedded in paraffin. All tissue sections (5 μm) were stained with hematoxylin and eosin (H&E) for morphological evaluation, and liver samples were additionally stained with Caspase-3 antibody for detecting apoptosis (1∶200, # 9661, Cell Signaling Technology). Immunohistochemical staining was performed on a Roche Ventana Discovery XT automated staining instrument (Roche Ventana Medical Systems) using Ventana reagents according to the manufacturer’s instructions. Immune complex was detected using the Ventana discovery anti-rabbit NP-AP kit (760–4817) and developed using the Discovery Red kit (760–228). Slides were counterstained with hematoxylin II (790–2208), followed by Bluing reagent (760–2037). The slides were then dehydrated with ethanol series, cleared in xylene, and mounted. As a negative control, the primary antibody was replaced with normal rabbit IgG. Stained slides were scanned using a Leica Aperio ScanScope CS 5 slide scanner (Leica Aperio) and images were quantitated using a 0–3 scale as assessed blindly by Dr. Qai.

Serum analysis

Serum was reduced with 12 % dithiothreitol at room temperature for 5 minutes and extracted with 10 % sulfosalicylic acid by 1 hour incubation at 4 °C followed by centrifugation at 13,000 rpm at 4 °C for 15 minutes. The supernatant containing amino acids was loaded on an Aracus amino analyzer (MembraPure, Germany). The data acquisition and analysis software Clarity records analogue channels simultaneously (570 nm, 440 nm). Amino acids peaks were identified and quantitated by comparing to known external standards. Alanine aminotransferase (ALT, ab282882, Abcam), aspartate aminotransferase (AST, ab263882, Abcam), and plasminogen activator inhibitor-1 (PAI-1, ab197752, Abcam) levels were quantified in diluted serum (250 ~ 1000x, 125x, and 25x, respectively) following the manufacturer’s instructions.

RNAseq analysis

Total RNAs from liver tissue were extracted using Trizol (15596026, Invitrogen) and cleaned by RNeasy kit (74104, Qiagen) according to the manufacturer’s instructions. RNA from each sample were used to make mRNA-seq library according to the product guide of NEBNext^® Ultra^™ Directional RNA Library Prep Kit for Illumina (E4720L). After processing, libraries were purified using SPRIselect beads, had a quality check on Agilent 2100 bioanalyzer. Sample libraries were subsequently pooled and loaded to the Nextseq 2000 (Illumina). Paired end reads at 65bp were generated. Faseq files were obtained at Illumina base space (https://basespace.illumina.com). For each sample, approximately 60 million PE reads were obtained.

Paired-end FASTQ files were quality checked using FastQC and MultiQC (Ewels et al 2016). FASTQ files were aligned to mouse reference GRCm38, using the STAR aligner (Dobin et al 2013), and resulting BAM files were sorted by read name using samtools sort. Reads aligning to GRCm38 were quantified at the gene-level using htseq-count [HTSeq] (mode = “union”). A PCA plot and related statistics were generated using the DESeq2 [DESeq2] function plotPCA, after applying the rlog transformation to the raw count matrix. Differential expression analysis was carried out with DESeq for the following comparisons of diets: (1) LMD vs. RD, (2) RD3D vs. LMD, and (3) RD3D vs. RD. For each of these comparisons, a list of genes ranked by Wald statistic was produced. Genes with fold-change greater than 2 (up or down) and adjusted p-value less than 0.05 were considered differentially expressed.

ORA and GSEA analysis were performed using Web-Gestalt (Liao et al 2019). Statistical enrichment, all lists were compared to the mouse protein coding genes list. Functional databases analyzed include the KEGG and WikiPathways. For Tg-I278T Cbs^+/− mice, genes used for analysis had >1.5-fold expression differences, P<0.014. For Tg-I278T Cbs^−/− mice, criteria were >2-fold expression, P<0.010.

Statistics

Values in text are mean ± SE. Differences between two groups were analyzed by the Mann–Whitney U test (unpaired and non-parametric test). Significance between more than two groups were determined using one-way ANOVA followed by Dunnett’s multiple comparison tests employing GraphPad Prism 10.0 software. Statistical significance was accepted at the value of P < 0.05.

Supplementary Material

Supplementary Figures

Supplemental Figure 1. Serum amino acid concentrations.

Supplemental Figure 2. Analysis of pathways changed due in RD vs. LMD in Tg-I278T Cbs^+/− mice. A. GSA and GSEA analysis. B. List of all significantly changed genes in RD vs LMD genes categorized by function.

Supplemental Figure 3. ORA and GSEA analysis of Cbs^−/− RD vs. LMD mice

Supplemental Figure 4. Transcripts showing opposite behavior in Cbs^+/− vs. Cbs^−/− livers.

Supplemental Figure 5. ORA and GSEA analysis of RD vs RD3D Cbs−/− mice using Wikipathways.

Synopsis.

Shifting CBS-deficient mice from a low-methionine to a normal methionine diet alters hemostasis resulting in hepatic vein thrombosis and liver failure.

Data and material statement

Processed RNAseq data is available as supplementary data in the article. Additional data used to make the figures in the manuscript is available upon request.

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.