Cystathionine β-synthase deficiency: different changes in proteomes of thrombosis-resistant Cbs−/− mice and thrombosis-prone CBS−/− humans

Cystathionine β-synthase (CBS)-deficient patients are prone to vascular thrombosis. In contrast, Cbs−/− mice show no abnormalities in blood coagulation. To identify molecular basis underlying these disparately different thrombotic phenotypes, we analyzed plasma proteomes of Cbs−/− vs. Cbs+/+ mice (8-month-old, 12/group, sex-matched) and CBS−/− vs. CBS+/+ humans (37 ± 7-year-old, 10–14/group, sex-matched) using label-free mass spectrometry. We identified 117 and 41 differentiating plasma proteins in Cbs−/− mice and CBS−/− humans, respectively. Twenty-one proteins were shared between CBS−/− humans and Cbs−/− mice, with sixteen changed in the opposite direction. Proteins involved in blood coagulation and complement/coagulation cascades represented a greater fraction of the differentiating proteins in CBS−/− patients (51%) than in Cbs−/− mice (21%). Top canonical pathways, identified by Ingenuity Pathways Analysis, such as LXR/RXR, FXR/RXR activation (− log[P-value] = 30–31) and atherosclerosis signaling (− log[P-value] = 10–11) were similarly affected in Cbs−/− mice and CBS−/− humans. The Coagulation System was affected stronger in CBS−/− humans than in Cbs−/− mice (− log[P-value] = 15 vs. 10, respectively) while acute phase response and complement system were affected stronger in Cbs−/− mice (− log[P-value] = 33 and 22, respectively) than in humans (− log[P-value] = 22 and 6, respectively). Other pathways, including IL-7 signaling and B cell development were affected only in Cbs−/− mice. Taken together, our findings suggest that differences in these processes, in particular in the Coagulation System, could account for the thrombotic phenotype in CBS−/− patients and the absence of thrombosis in Cbs−/− mice. Overall, our findings suggest that Cbs−/− mice have a better adaptive response to protect from prothrombotic effects of hyperhomocysteinemia than CBS−/− humans.

www.nature.com/scientificreports/ enhancing the synthesis of Hcy-thiolactone 2 . The dysregulation of Hcy metabolism in human CBS deficiency is accompanied by pathological changes in many organs, including the cardiovascular system. CBS −/− patients have a blood clotting defect 1 underlying their high susceptibility to vascular thromboembolism (50% chance by the age of 30), which is the major cause of morbidity and mortality 1 . Although the mechanisms leading to the onset of thromboembolism in CBS −/− patients are not fully understood 4 , 32% of those patients suffer thromboembolic stroke and 4% have myocardial infarction 1 . Other vascular thromboembolic incidents occur in the peripheral veins (51%) and arteries (11%) of CBS −/− patients 1 .
To study mechanisms underlying the pathology associated with elevated Hcy, the Cbs gene has been deleted in mice. The Cbs −/− homozygous mutants showed a neonatal phenotype due to liver dysfunction and only a few percent of them survived to adulthood 5 . The neonatal lethality was circumvented by providing betaine in the drinking water 6 or by creation of transgenic Cbs −/− mice harboring wild type or mutant versions of the human CBS gene under control of the zinc-inducible metallothionein promoter, Tg-I278T Cbs −/− mouse 7 . The Cbs −/− mice recapitulate most, but not all, of the pathological changes observed in human CBS −/− patients 6,7 including the reduced life span 7 .
Surprisingly, Tg-I278T Cbs −/− mice that have severe HHcy (plasma total Hcy 250 μM) do not have the thrombotic phenotype 7,8 found in CBS −/− patients 1,4,9,10 . This was established by finding no differences between Tg-I278T Cbs −/− vs. Tg-I278T Cbs +/+ mice in the prothrombin time, activated partial thromboplastin time 7 , bleeding time, and the occlusion time in the chemical FeCl 3 vascular injury model 8 that was used to demonstrate impaired fibrinolysis caused by the formation of S-Hcy-annexin 2 in wild type mice with plasma total Hcy elevated (to 70 μM) by dietary intervention 11 . Tg-I278T Cbs −/− mice also do not show elevated arterial or venous thrombosis in other assays, including photochemical injury of the carotid artery, injury of the carotid artery and mesenteric arterioles, and ligation of the inferior vena cava 8 . Hemostatic and hemodynamic parameters are also not affected in Tg-I278T Cbs −/− mice.
Different susceptibilities of CBS −/− patients and Cbs −/− mice to thrombosis might be reflected in their proteomes. Previous work has identified CBS deficiency-responsive proteins in human plasma 12 . In the present work we have used label-free mass spectrometry and the Ingenuity Pathway Analysis resources to interrogate plasma proteomes of thrombosis-resistant Cbs −/− mice and compared them with plasma proteomes of thrombosis-prone CBS −/− humans.

Results
Total Hcy levels in CBS −/− patients and Cbs −/− mice were severely elevated compared to corresponding unaffected controls (71.2 ± 55.5 vs. 9.7 ± 5.9 μM in human plasma 12 ; 236 ± 80 vs. 10.2 ± 5.3 μM in mouse plasma), as expected 13,14 . In each group of subjects, label-free nanoLC-MS/MS mass spectrometry identified 196-198 proteins in human and mouse plasma with a minimum of two peptides and 1% false discovery rate (FDR). The overlap between duplicate injections was > 90% at the protein level as determined by the Proteome Discoverer (PD) analysis.
To identify variations between samples in terms of global proteomic profile, principal component analysis (PCA) was carried out. There was a clear difference in PCA profiles between Cbs −/− mice and their Cbs +/+ siblings (Fig. 1). The PCA also differentiated between males (lower clusters) and females (upper clusters) within each Cbs genotype. However, there was an overlap in PCA profiles between CBS −/− patients and CBS +/+ controls ( Fig. 1) and between males and females within each CBS genotype.
We identified 117 proteins differentiating between Cbs −/− vs. Cbs +/+ in mice in the present study while 41 proteins differentiating between CBS −/− vs. CBS +/+ were identified in humans 12 (Fig. 2). Of these, ninety-six proteins (82%) were affected by CBS deficiency only in mice, while twenty (49%) were affected only in humans. There were twenty-one proteins (accounting for 17% and 51% of the differentiating proteins in mice and humans, respectively) that were affected by CBS deficiency in both mice and humans (Fig. 2). The differentiating proteins are listed in Supplemental Table S1.
Overlap between proteins affected by CBS genotype in mice and humans. We identified twentyone proteins whose levels were affected both by the human CBS −/− genotype and the mouse Cbs −/− genotype (Fig. 2, Table 1). The twenty-one common proteins represented 52% and 18% of the total number of proteins affected in CBS −/− patients and Cbs −/− mice, respectively.
Organ specificity of Cbs −/− genotype effects on protein levels. To establish whether Cbs −/− genotype-affected proteins identified in mouse plasma are organ specific and to identify their possible sources, we carried out label-free mass spectrometry analyses of proteomes of mouse organs. We identified fifteen proteins that were affected by Cbs deficiency both in plasma and in at least one organ; seven of those proteins were upregulated and six downregulated in the plasma and at least one organ. For the majority of proteins (13 out  www.nature.com/scientificreports/ of 15), Cbs −/− genotype had similar effects on protein expression in the plasma and in liver, kidney, spleen, and heart (Table 2). Specifically, SerpinA1d, a component of the complement/coagulation cascades, was downregulated in plasma, liver, kidney, spleen, and heart. Carboxylesterase Ces1c, a lipid metabolizing enzyme, was downregulated in plasma, liver, and heart. Immunoglobulins Ighg2c, Ighg2b, Ighm, and Igkc were downregulated in plasma and spleen, while Igkc was similarly downregulated in kidney and heart ( Table 2).
The following proteins were upregulated: apoliprotein ApoE in plasma, liver, kidney, spleen, and heart; lysosomal antigen-processing enzyme, cathepsin CtsB in plasma, liver, and kidney; clusterin Clu in plasma and kidney; glutathione peroxidase Gpx3 in plasma, kidney, and spleen; haptoglobin Hp in plasma, kidney, and heart; Mup20 in plasma, liver, and kidney; lumican Lum in plasma and kidney ( Table 2).
Levels of two proteins were changed in opposite directions in plasma and organs. An acute phase response protein, hemopexin Hpx, which was upregulated in the plasma, was downregulated in heart and spleen. An iron transport protein, serotransferrin Tf, which as upregulated in plasma, was downregulated in the brain (Table 2).
Validation of label-free mass spectrometry analysis. To validate label-free mass spectrometry experiments, we quantified five mouse and two human proteins using ELISA or Western blot assays (Table 3). We found that the acute phase response protein anti-thrombin III (SERPINC1), and the oxidative stress response protein GPX3, downregulated (0.84-fold) and upregulated (1.41-fold, P = 0.005), respectively, in the label-free experiments in CBS −/− patients (Supplemental Table S1), were similarly affected by the CBS genotype in the ELISA assays (0.62-fold and 1.19-fold, respectively).

Bioinformatics analysis.
To identify biological pathways linked to proteins affected by Cbs deficiency in mice and CBS deficiency in humans, we performed bioinformatics analyses using IPA resources. Analyses for the human CBS deficiency have been described before 12 . We have found that in the mouse Cbs deficiency, proteins from 19 pathways were overrepresented among differentially expressed proteins, while proteins from 17 pathways were enriched in the human.      (Fig. 4B).
Cbs-deficient mice. IPA identified five biological networks associated with mouse Cbs deficiency ( Table 4).
The two top-scoring networks were Protein Synthesis, Cardiovascular Disease, Hematological Disease and Developmental Disorder, Hereditary Disorder, Immunological Disease. The Protein Synthesis, Cardiovascular Disease, Hematological Disease network, with a score of 50, contains 35 proteins: 25 were quantified by the labelfree mass spectrometry and 10 were identified by IPA to interact with the quantified proteins (Fig. 5A). This network involves proteins participating in lipid metabolism and the complement/coagulation cascades that interact with Erk1 and Crp. The Developmental Disorder, Hereditary Disorder, Immunological Disease network, with a score of 40, contains 35 proteins: 21 quantified by the label-free mass spectrometry and 14 identified by IPA to interact with the quantified proteins (Fig. 5B). This network contains proteins of the complement/coagula-   CBS-deficient patients. IPA identified two biological networks associated with human CBS deficiency 12 (Table 4). The Humoral Immune Response, Inflammatory Response network, with a score of 60, contains 34 proteins: 24 quantified by label-free mass spectrometry and 10 identified by IPA to interact with the quantified proteins (Fig. 5C)

Discussion
In this comparative proteomic study, we found that thrombosis-susceptible CBS −/− patients and thrombosisresistant Cbs −/− mice exhibit different expression patterns of proteins involved in the blood coagulation, complement/coagulation cascades, and acute activation and immune responses. Notably, proteins involved in the blood coagulation and complement/coagulation cascades represented a greater fraction of proteins affected in CBS −/− patients (51%) than in Cbs −/− mice (21%). Further, although some proteins were affected in both CBS −/− patients and Cbs −/− mice, these common proteins we affected in opposite directions in the two species. Our findings suggest that species-specific differences in the effects of CBS deficiency on these processes could account for the pro-thrombotic phenotype of CBS −/− patients and the thrombosis-resistant phenotype of Cbs −/− mice. Comparative proteomics of CBS deficiency have not been studied before, and, to our best knowledge, this is the first proteomic study of disparately different thrombotic phenotypes in humans and mice. Overall, our findings suggest that Cbs −/− mice have a better adaptive response to protect from prothrombotic effects of hyperhomocysteinemia than humans. Discussion of specific changes in the proteomes of human CBS −/− patients and Cbs −/− mice identified in the present work and their roles in blood coagulation follows below. Our findings show that KNG1 protein levels are reduced in CBS −/− patients, but unaffected in Cbs −/− mice. KNG1 is an inhibitor of thiol proteases that inhibits the thrombin-and plasmin-induced aggregation of thrombocytes (https ://www.nextp rot.org/entry /NX_P0104 2/). Variants in the KNG1 gene are associated with venous  Table 4. Drawing shown in panel (C) is from Sikora et al. 12  www.nature.com/scientificreports/ thrombosis 15 . Thus, reduced KNG1 levels could contribute to the pro-thrombotic phenotype observed in CBS −/− patients 1,4,9,10 . The anti-thrombin SERPINC1, genetic variants of which are associated with venous thrombosis 15 , was downregulated in CBS −/− humans and upregulated in Cbs −/− mice. The downregulation of the anti-thrombin SER-PINC1 would increase thrombin activity (and thus increase blood clotting) in CBS −/− patients. In contrast, the upregulation of the anti-thrombin Serpinc1 would reduce thrombin activity (and thus reduce blood clotting) in Cbs −/− mice. These findings suggest that the changes in levels of components regulating thrombin activity contribute to the thrombotic phenotype in CBS −/− patients and the absence of thrombosis in Cbs −/− mice 7,8 .
We found that SERPIND1, heparin cofactor 2, was downregulated in CBS −/− patients and upregulated in Cbs −/− mice (Table 1, Supplemental Table S1). Because SERPIND1 deficiency can lead to increased thrombin generation and a hypercoagulable state 16 , these findings could account for the thrombotic phenotype of CBS −/− patients and thrombosis-resistance of Cbs −/− mice.
APOH is a component of an important anticoagulation pathway. It inhibits the contact-activated intrinsic blood coagulation pathway by binding phospholipid at the surface of damaged cells 17 , von Willebrand factordependent platelet adhesion and aggregation 18 , and the pro-coagulation activities of thrombin 19 . Polymorphisms in the APOH gene that affect these functions, are risk factors for venous thrombosis in a Chinese population 20 . The present findings that APOH is downregulated in CBS −/− patients and upregulated in Cbs −/− mice ( Table 1,  Supplementary Table S1), could contribute to the thrombotic phenotype of CBS −/− patients and the absence of thrombosis in Cbs −/− mice.
The present study found that the acute phase protein Orm2 was one of the most highly (4.95-fold) upregulated proteins in Cbs −/− mice, while ORM2 was significantly downregulated (0.8-fold) in CBS −/− patients ( Table 1, Supplemental Table S1). Because of its effects on inflammation, ORM2 has been suggested to have a protective effect on stroke in humans. In mice Orm2 has been show to protect from ischemic stroke: levels of Orm2 increase in a mouse model of ischemic stroke and administration of the Orm2 protein significantly decreases infarct size and neurological deficit score 21 . Thus, our present findings, in conjunction with the mouse Orm2 study, suggest that the high upregulation of Orm2 would protect Cbs −/− mice from ischemic stroke, and that this protective response is not induced in CBS −/− patients.
Although previous findings show that Cbs −/− mice do not have a thrombotic phenotype 7,8 found in CBS −/− patients 1,4,9,10 , severe stroke phenotype can be induced in homozygous Cbs −/− mice by the middle cerebral artery ischemia/reperfusion injury 22 . A similar stroke phenotype can also be induced in heterozygous Cbs +/− mice, which have only slightly elevated plasma Hcy levels, 6.1 vs. 3.1 μM. Surprisingly, treatment with the N-methyl-D-aspartate receptor antagonist memantine protects Cbs +/− but not Cbs −/− mice from experimental stroke mice. This has been suggested to be due to an increased expression of Hcy-insensitive N-methyl-D-aspartate receptor subunit NR2B in Cbs −/− mice. However, it is unclear whether activation of N-methyl-D-aspartate receptors contributes to the experimental stroke phenotype of Cbs −/− mice 22 .
We also found that several proteins of the Coagulation System, such as Serpina1b, Serpina10, and the anticoagulant protein Pros1, which are known to inhibit thrombin activity or its generation, were upregulated in Cbs −/− mice, but not in CBS −/− patients (Table 1, Supplementary Table S1). Pros1 inhibits thrombin generation by binding and inhibiting FXa and FVa in the prothrombinase complex 23 , Serpina1b inhibits plasmin and thrombin, while Serpina10, a negative regulator of blood coagulation, inhibits the activated coagulation factors X and XI 24 . These findings suggest reduced propensity to clotting in Cbs −/− mice, consistent with absence of thrombosis in these animals 7,8 .
In addition to the Coagulation System, we found that proteins affected by CBS deficiency both in humans and mice were significantly enriched in other molecular pathways, many of which directly or indirectly affect blood clotting (Fig. 4). For example, one of the CBS deficiency-responsive pathways-LXR/RXR Activation-involves RXR, which has a negative regulatory role in platelet function/aggregation and thrombus formation 25 . Notably, we identified the LXR/RXR pathway proteins that were downregulated in CBS −/− patients but upregulated in Cbs −/− mice (AHSG, APOA1, APOH, APOM, CLU, GC, HPX, SERPINF2) ( Table 1, Supplementary Table S1), indicating opposite regulation of the expression/function of those proteins in Cbs −/− mice and CBS −/− humans. One of those proteins, AHSG, a potent anti-inflammatory cytokine 26 , is known to be involved in platelet degranulation (https ://plate letwe b.bioap ps.bioze ntrum .uni-wuerz burg.de/), while SERPINF2 is a serine protease inhibitor responsible for inactivating plasmin 27 . As treatments with RXR ligands have cardioprotective effects 25 , our findings may offer a similar strategy for the prevention of thrombosis in CBS −/− patients.
The interconnections between the blood coagulation and complement cascades have been observed in various clinical and experimental settings 28 . Here, using the label-free mass spectrometry we have identified such interconnections in CBS −/− patients and Cbs −/− mice and shown that CBS deficiency is characterized by complement-associated inflammatory responses. We have also identified specific changes in the levels of protein www.nature.com/scientificreports/ components of the blood coagulation system that could provide an explanation for the thrombotic phenotype in CBS −/− patients and the absence of thrombosis in Cbs −/− mice. These findings also suggest that ORM2, F10, SOD3, HP, SERPING1, SERPINA10, SERPINA3N, SERPINA7, and/or CES2E proteins might be targeted for potential anti-thrombotic interventions in humans.

Materials and methods
CBS-deficient patients. The study included plasma samples from the Danish 29 (n = 4) and Polish 30,31 (n = 6) CBS-deficient patients (mean age 37.3 ± 7.3 years, 5 males, 5 females) and gender-and age-matched healthy control individuals (n = 14). Characteristics of the CBS-deficient patients and controls have been described 12 .
Briefly, two of the Polish CBS-deficient patients had a stroke before diagnosis in childhood at age 3 to 15. Danish CBS-deficient patients were diagnosed at age 7, 21, 22, and 36; two of them had episodic transient ischemic attacks at age 22 and 36. Patients were on a Hcy-lowering vitamin B 6 treatment, except for the six vitamin B 6 non-responsive patients who were treated with betaine. All experimental protocols conformed to the Ethical Guidelines of the World Medical Association Declaration of Helsinki and was approved by the Bioethics Committee of the University of Medical Sciences, Poznań, Poland (decision no. 661/16). Written informed consent was obtained from all participants or legal guardians. The procedures followed were in accordance with institutional guidelines and regulations.
Human plasma samples. Blood was drawn by venous puncture during routine health examination visits at doctor's office as previously described 12 . Blood samples were collected into EDTA tubes, plasma was separated by centrifugation (2,000 g, 15 min, 4 °C), collected, and stored at − 80 °C. Blood and tissue collection and processing. Blood was collected from the cheek vein into Eppendorf tubes containing 1% (v/v) 0.5 M EDTA. After centrifugation (2,000 g, 15 min, 4 °C), separated plasma and cells were frozen at − 80 °C. Mice were euthanized using CO 2 , organs collected and frozen on dry ice. Liver, kidneys, spleen, and brain were powdered with dry ice using mortar and pestle, and stored at − 80 °C. Proteins were extracted from powdered tissues using triethylammonium bicarbonate pH 8.5 buffer, 0.1% SDS (Millipore-Sigma) with sonication (Bandelin SONOPLUS HD 2070) on wet ice (five 30-s strokes with 1 min cooling interval after each stroke). Extracts were clarified by centrifugation (15,000 g, 5 min, 4 °C), protein was precipitated with acetone (40 min, − 20 °C), the protein pellet collected by centrifugation (15,000 g, 10 min, 4 °C), and dissolved by sonication in 50 mM ammonium bicarbonate. Protein concentration was quantified using a 2-D Quant kit (GE Healthcare).

Label-free mass spectrometry. Dionex UltiMate 3,000 RSLC nanoLC System attached to Q Exactive
Orbitrap mass spectrometer (Thermo Fisher Scientific) was used for analyses as previously described 12 . Tryptic peptides were separated on a C18 column Acclaim PepMap RSLC nanoViper (75 µm × 25 cm, 2 µm) using 4-60% acetonitrile gradient, 0.1% formic acid (300nL/min, 30 °C, 230 min). Mass spectra were acquired on the Q Exactive in a data-dependent mode using top 10 data-dependent MS/MS scans. Target value for the full scan MS spectra was 1 × 10 6 with a maximum injection time of 100 ms and a resolution of 70,000 at m/z 400. The MS scan range was 300-2000 m/z. Ten most intense ions charged two or more were selected with a 2-Da isolation window and fragmented by higher energy collisional dissociation with NCE 25. The ion target value for MS/MS was 5 × 10 4 with a maximum injection time of 100 ms and a resolution of 17,500 at m/z 400. Data analysis. Proteins were identified using UniProt human database with a precision tolerance 10 ppm for peptide masses and 0.08 Da for fragment ion masses as previously described 12 . Two missed trypsin cleavages were allowed. S-carbamidomethyl-Cys was a fixed modification and Met-sulfoxide was allowed as a variable modification. For protein identification/quantification, each dataset was imported into MaxQuant 1.5.3.30 version (MPIB, Martinsried, Germany). Protein was positively identified if at least two peptides were found by the Andromeda search engine, and a peptide score had a false rate of discovery of 0.01. The analyses were based on label-free quantification (LFQ) of peptide intensities. The MaxQuant data were filtered for reverse identifications (false positives), contaminants, and proteins "only identified by site". The mean LFQ intensities + /− standard Statistics. The data were analyzed using Perseus ver. 1.5.3.2 software (part of MaxQuant package) as previously described 12 . Numeric data were log-transformed and samples were annotated with their group affiliation. Data were filtered-proteins with valid values in 70% of samples in at least one group were kept in the table. For multiple comparisons, one-way analysis of variance (ANOVA) with a Bonferroni correction for multiple testing was performed. Two-sample t-test was used for comparisons between two groups with P values < 0.05 used for truncation. The resulting lists of differentiating proteins were normalized using Z-score algorithm for hierarchical clustering of data. Multivariate analyses were carried out by untargeted principal component analysis (PCA) 12 .
Bioinformatics analysis. Proteins that were quantified as unique and non-redundant were used in analyses. Proteins were considered to be differentially expressed if the FC difference was statistically significant (P < 0.05). The proteins were considered as differentiating if it was quantified by at least two peptides with > 99% confidence. Uncharacterized proteins were excluded from the analysis. Bioinformatic analysis to discover biological pathways and networks for proteins affected by the Cbs −/− genotype were performed out using the Ingenuity Pathway Analysis software (IPA, Ingenuity Systems, Mountain View, CA). The data sets containing differentially expressed proteins were uploaded into the IPA Knowledge database to map proteins and metabolites to global molecular networks and identify their interactions with other proteins in the database.

ELISA and western blot assays.
To validate differentially expressed proteins, we quantified human SER-PINC1 and GPX3, mouse Gpx3, Pon1, and Serpinc1 using commercial ELISA kits (FineBiotech and AssayPro). Triplicate assays were performed following the manufacturers' protocols. A 450 was read using microplate reader (Infinite M200 Pro, Tecan, Switzerland). Mouse Mup20 and IgG were quantified by Western blotting using commercial anti-Mup and anti-IgG antibodies.
Homocysteine assays. Mouse and human plasma total Hcy was quantified by an HPLC-based method as previously described 14 .