Cystathionine γ-lyase is a major regulator of cognitive function through neurotrophin signaling and neurogenesis

Significance

Once considered to function predominantly in the peripheral systems, cystathionine γ-lyase (CSE) is emerging as a key player in neuroprotection. Prior studies had considered cystathionine β-synthase (CBS) to be the principal enzyme governing H₂S signaling in the brain. In this study, through an integrated approach combining genetic, proteomic, biochemical, and behavioral studies, we demonstrate that CSE is crucial for maintaining brain homeostasis and that loss of CSE is sufficient to trigger cognitive deficits. This work resolves the longstanding question regarding CSE’s role in brain function and elevates CSE from a peripheral player to a key therapeutic target for brain health.

Abstract

Cystathionine γ-lyase (CSE), the enzyme responsible for neuronal cysteine and hydrogen sulfide production, is dysregulated in aging and neurodegenerative diseases including Alzheimer’s disease and Huntington’s disease, both marked by cognitive decline in addition to motor deficits. To determine whether CSE loss directly causes cognitive decline, we genetically ablated CSE in mice. This loss was sufficient to induce oxidative damage, compromise blood–brain barrier integrity, impair neurogenesis and neurotrophin signaling, and elicit cognitive deficits. Global proteomic analysis further revealed molecular alterations that contribute to impaired neurogenesis. Our findings establish CSE as an essential guardian of homeostatic brain health and identify it as a potential therapeutic target for neurodegenerative disorders.

Cystathionine γ-lyase (CSE) is a key enzyme in the reverse transsulfuration pathway (Fig. 1A), the sole mammalian biosynthetic enzyme for the semiessential amino acid cysteine and one of three enzymes that produces the gasotransmitter hydrogen sulfide (H₂S) in the brain (1). Two other enzymes, cystathionine β-synthase (CBS) and 3-mercaptopyruvate sulfurtransferase (3-MST), also generate H₂S in the brain, with CSE predominantly neuronal, 3-MST present in both neurons and astrocytes, and CBS localized to glial cells (2, 3). CSE utilizes either cysteine or homocysteine as a substrate to generate the gasotransmitter, which signals via the posttranslational modification termed S-sulfhydration/persulfidation (4–7). S-sulfhydration is widespread throughout the body and regulates diverse physiological processes, including responses to stress stimuli, mitochondrial function, and neuronal signaling (4). At the molecular level, S-sulfhydration modulates the activity of numerous proteins, including key metabolic proteins and neuroprotective enzymes such as glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and the E3-ubiquitin ligase parkin (8–10). It influences organellar stress responses and cellular signaling (11–16), and also protects cysteine residues from irreversible oxidation (17–19).

Fig. 1.

Cognitive deficits in Cth^−/− mice. (A) Schematic representation of the reverse transsulfuration pathway. CSE utilizes cystathionine to generate cysteine, which in turn is used by CSE or CBS to produce hydrogen sulfide (H₂S). 3-MST utilizes 3-mercaptopyruvate, which in turn is generated from cysteine by cysteine aminotransferase (CAT), to produce H₂S. Both CSE and CBS also utilize homocysteine to produce H₂S. (B and C) Barnes maze test reveals impaired spatial memory in Cth^−/− mice, compared to wild-type mice (n = 10, mean ± SEM, two-way repeated measures ANOVA with Tukey’s multiple comparisons test, **P < 0.01). (D) Schematic representation of the open field test. (E) Open field test reveals equivalent locomotion of wild-type and Cth^−/− mice (n = 12 to 18, mean ± SEM, two-tailed unpaired t test, ns). (F) Locomotor activity of Cth^−/− mice across different bins or phases of the open field test remains unaltered (n = 12 to 18, mean ± SEM, two-way repeated measures ANOVA with Tukey’s multiple comparisons test, ns). (G) Body weight of 6 to 7-mo-old Cth^−/− mice were comparable to WT controls (n = 10, mean ± SEM, two-tailed unpaired t test, ns). (H and I) Sensorimotor function in Cth^−/− mice. (H) Time required to retrieve buried cookie remains unaltered between genotypes at 6 to 7 mo of age, suggesting no changes in olfaction. Each point represents an individual mouse (n = 10, mean ± SEM, two-tailed unpaired t test, ns). (I) 6 to 7-mo-old Cth^−/− mice needed similar time for tape removal as that of WT mice, indicating no sensorimotor deficits. Each point represents an individual mouse (n = 10, mean ± SEM, two-tailed unpaired t test, ns). (J) Barnes maze test reveals normal spatial memory in Cth^−/− mice, compared to WT mice at 2 mo of age (n = 10, mean ± SEM, two-way repeated measures ANOVA with Tukey’s multiple comparisons test, ns). (K) Cysteine levels are equivalent in 6- to 7- mo- old wild-type and Cth^−/− mice. Each point represents an individual mouse [n = 4, mean ± SEM, two-tailed unpaired t test, not significant (ns)]. (L) Total glutathione levels (GSH) are equivalent in 6- to 7-mo-old wild-type and Cth^−/− mice. Each point represents an individual mouse (n = 3, mean ± SEM, two-tailed unpaired t test, ns). (M) Plasma homocysteine levels in 6- to 7-mo- Cth^−/− mice were significantly elevated (n = 4, mean ± SEM, two-tailed unpaired t test, ****P < 0.0001). (N) Levels of CBS and 3-MST are equivalent in wild-type and Cth^−/− hippocampus as revealed by western blot analysis using β-actin as a loading control. Note the absence of CSE protein in western blot analysis. (O) Quantitation of CBS. Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, ns). (P) Quantitation of 3-MST. Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, ns).

Across species, aging is associated with decreased S-sulfhydration, increased cysteine oxidation, and reduced CSE levels (17). Elevated CSE expression and H₂S signaling have also been linked to longevity strategies, including dietary restriction and mTOR inhibition (20–22), and CSE has been identified as a prolongevity gene in several screens (23, 24). We and others have previously shown that H₂S donors are neuroprotective in preclinical models of Alzheimer’s disease (AD) driven by human genetic mutations known to cause heritable disease and also that brain CSE and S-sulfhydration decline in preclinical models of aging and AD (10, 17, 25, 26). Dysregulation of the transsulfuration pathway and CSE was also observed in several other neurodegenerative states including Huntington’s disease (HD), Parkinson’s disease (PD), and ataxia (9, 27, 28). While insufficient H₂S production can impair neural function, excessive H₂S levels are equally harmful, as demonstrated in Down syndrome (29, 30).

Historically, CBS and 3-MST have been considered the principal sources of H₂S in the brain, while CSE was considered more prominent in peripheral tissues (31, 32). In this study, we challenged this view by testing whether depletion of CSE alone, without alterations in other H₂S-synthesizing enzymes, could produce neuropathological changes and cognitive deficits characteristic of human neurodegenerative disease.

Results

CSE Depletion Causes Premature Cognitive Impairment.

To determine whether CSE depletion causes cognitive impairment, we assessed learning and memory in 6-mo-old Cth^−/− mice using the Barnes maze, a test of spatial learning and memory (33). In this test, mice were placed on an elevated round platform with 20 holes equally spaced around its perimeter, with one hole containing an escape chamber (Fig. 1B). Spatial cues are placed outside the maze and a bright light is focused on the platform. Mice are then placed in the center of the maze and the time taken to reach the escape box is recorded (primary latency) for a maximum of 3 min. Mice normally avoid bright light and quickly learn to locate the escape box with greater efficiency over 6 d. Cth^−/− mice, however, were significantly impaired as compared to WT littermates (Fig. 1C), indicating impaired spatial memory.

Importantly, this deficit was not attributable to altered locomotor activity, as open field testing revealed no significant differences in movement parameters between genotypes (Fig. 1 D–F). There were no significant differences in the body weights of the 6-mo-old Cth^−/− mice or their wild-type controls (Fig. 1G). Furthermore, the mice lacking CSE did not display any sensorimotor impairments as revealed by the hidden cookie and the tape removal tests (Fig. 1 H and I). Notably, the cognitive impairment was not present from birth, as 2-mo-old Cth^−/− mice performed normally in the Barnes maze test (Fig. 1J), indicating a progressive, age-related onset.

We further evaluated potential metabolic and compensatory mechanisms. Brain cysteine and GSH levels were similar between genotypes (Fig. 1 K and L), however plasma homocysteine levels were increased in the 6-mo-old Cth^−/− mice (Fig. 1M). Expression of the other H₂S-producing enzymes, CBS and 3-MST, showed no compensatory changes (Fig. 1 N–P). Together, these data indicate that loss of CSE alone is sufficient to induce premature, aging-related cognitive deficits, independent of changes in cysteine/GSH or compensation by CBS or 3-MST.

Global Proteomics of the Cth^−/− Hippocampus.

To assess the impact of CSE loss on the brain, we performed global proteomic profiling of the hippocampi from 6-mo-old Cth^−/− mice and their WT littermates (Fig. 2A and Dataset S1). Using a direct data-independent acquisition (DIA) default workflow, we identified 7,203 proteins, some of which showed differential abundance as a function of genotype (Fig. 2 B and C). Intensity histograms showed no global shift in total protein abundance in the Cth^−/− samples (SI Appendix, Fig. S1A) and biological replicates were highly reproducible (Pearson r = 0.96 to 0.99) (SI Appendix, Fig. S1B). Approximately 800 to 900 proteins were significantly altered (P < 0.05) in Cth^−/− mice relative to WT controls. Gene ontology (GO) analysis focusing on biological processes showed that transport proteins were most impacted by loss of CSE(Fig. 2D), while a GO analysis focusing on cellular components revealed that synaptic proteins were most impacted (Fig. 2E). Last, we assessed molecular functions (MF), indicating specific activities of gene products, such as “nitric oxide synthase binding” and “glutathione transferase activity” (Fig. 2F). Further pathway enrichment analysis using the Kyoto encyclopedia of genes and genomes (KEGG) highlighted perturbations in pathways linked to neurodegeneration and reactive oxygen species (ROS) in Cth^−/− mice (Fig. 2G).

Fig. 2.

Unbiased proteomics of mice lacking CSE reveals neurodegenerative changes (A) Brains from 6-mo-old WT and Cth^−/− mice were harvested, hippocampi were microdissected and snap-frozen. Samples were prepared for proteomics analysis using the S-Trap method and subjected to LC–MS analysis. (B) Volcano plot showing hippocampal proteome of Cth^−/− mice, compared with wild-type controls (n = 3 mice/genotype). Proteins above P value of 0.05 were considered as significantly changed. The proteins marked in red, blue, and gray indicate upregulation, downregulation, or no change, respectively. (C) Clusterogram demonstrating total hippocampal proteome consisting of 7,203 proteins from WT and Cth^−/− mice. (D) GO analysis showing significantly affected pathways for biological processes in the hippocampus of Cth^−/− mice analyzed using David database. (E) GO analysis showing significantly affected pathways for cellular components in the hippocampus of Cth^−/− mice analyzed using David database. (F) GO analysis showing significantly affected pathways for molecular functions (MF) in the hippocampus of Cth^−/− mice, compared to wild-type controls, analyzed using David database. (G) KEGG analysis in Cth^−/− hippocampus. (H–S) Abundance of key proteins in proteomics. Proteomics analysis revealed enhanced abundance of Hebp1, a heme-binding protein (H), Ftl1, ferritin light chain (I) and Tfrc, transferrin receptor (J) suggesting dysregulated iron homeostasis. Increased histone H2AX suggesting DNA damage (K), superoxide dismutase 1, SOD1 (L), Nicotinamide nucleotide transhydrogenase, Nnt (M), decreased vesicular glutamate transporter 1 Slc17a7 or Vglut1 (N), decreased Synaptogyrin-1, Syngr1 (O), indicating compromised synaptic functioning, decreased Semaphorin (P) indicative of impaired axonal guidance, decreased disabled-1 (Dab 1), involved in neuronal positioning during development (Q), reduced abundance of CD34 (R) suggesting altered brain vasculature, and increased levels of Ighg (S)indicating compromised blood–brain barrier (BBB). Each point represents an individual mouse (n = 3, mean ± SEM, two-tailed unpaired t test, ****P < 0.0001, **P < 0.01, *P < 0.05). (T) Pathway terms that overlap between the upregulated or downregulated proteins in Cth^−/− (fold change 1.2, P-value < 0.05).

Notable individual changes in Cth^−/− mice included alterations in iron homeostasis proteins [heme binding protein 1 (Hebp1), ferritin light chain (Ftl1), and transferrin receptor (Tfrc)] (Fig. 2 H–J). Markers of DNA damage and redox imbalance were also elevated, including H2AX (Fig. 2K), superoxide dismutase 1 (SOD1) (Fig. 2L), and nicotinamide nucleotide transhydrogenase (Nnt) (Fig. 2M) (34, 35).

We further observed dysregulation of glutamate homeostasis and synaptic proteins consistent with changes reported in human AD, as evidenced by reduced levels of vesicular glutamate transporter 1 (VGLUT1, also known as Slc17a7) (Fig. 2N) and synaptogyrin-1 (Syngr1) (Fig. 2O), respectively. Levels of axon guidance and neuronal positioning proteins [semaphorin 3e (Sema3e) (Fig. 2P) (36) and disabled-1 (Dab1) (Fig. 2Q), respectively] (37, 38) were significantly decreased as well. Markers indicative of blood–brain barrier (BBB) dysfunction, a feature of neurodegenerative disease (39–44), were also present in Cth^−/− mice, such as reduced CD34 (Fig. 2R), a transmembrane phosphoglycoprotein found in endothelial progenitor cells of the blood–brain barrier (BBB). Additionally, immunoglobulin gamma heavy chain polypeptide (Ighg) was markedly increased in brain parenchyma (Fig. 2S). Collectively, these changes overlap with proteomic features of neurodegeneration in AD.

We further filtered proteins by P-value of less than or equal to 0.05 and a fold difference of at least 1.2 fold, and then compared the up- and downregulated signatures with a public human postmortem hippocampal AD gene expression dataset (GSE36980) and performed GSEA. Pathways were retained with nominal P-value< 0.05 and FDR < 0.10. Genes downregulated in Cth^−/− hippocampus were also downregulated in human AD and PD brains and were enriched for pathways critical for neurodegeneration and nervous system development (Fig. 2T, SI Appendix, Fig. S1 C–H, and Dataset S2).

Global Proteomics of the Cth^−/− Plasma.

We further performed global proteomic profiling of the plasma from 6-mo-old Cth^−/− mice and their WT littermates (SI Appendix, Fig. S2 and Dataset S3). Approximately 83 proteins were significantly altered (P < 0.05) in Cth^−/− mice relative to WT controls (SI Appendix, Fig. S2A). GO analysis focusing on biological processes showed that proteins involved in immune response, glucose, lipid, and energy metabolism were significantly impacted by loss of CSE (SI Appendix, Fig. S2B), while a GO analysis focusing on cellular components revealed that immunoglobulin complexes, extracellular proteins, high-density lipoprotein particles, and hemoglobin–haptoglobin complex were most impacted (SI Appendix, Fig. S2C). Analysis of MF revealed changes in antigen binding, immunoglobulin receptor binding, haptoglobin binding, oxygen binding, and changes in insulin receptor activity (SI Appendix, Fig. S2D). Further pathway enrichment analysis using KEGG highlighted perturbations in pathways linked to complement and coagulation cascades, amino acid biosynthesis pathways, carbon metabolism, and cytoskeletal function in muscles in Cth^−/− mice (SI Appendix, Fig. S2E).

CSE Deficiency Triggers Oxidative Stress in the Brain.

Guided by our proteomic results and the known role of S-sulfhydration in limiting oxidative damage (17, 45), we tested whether oxidative stress was elevated in Cth^−/− brain. Lipid peroxidation, which involves free radical-mediated oxidation of polyunsaturated fatty acids, yields the neurotoxic product 4-hydroxy-2-nonenal (4-HNE), which forms Michael adducts with cysteine residues and disrupts protein function (46). Immunofluorescent staining revealed significantly elevated 4-HNE in hippocampal subfields [CA1, dentate gyrus (DG), and CA3] in 6 to 7 mo old Cth^−/− mice, indicating elevated lipid peroxidation (Fig. 3 A and B). We further quantified levels of lipid peroxidation by measuring malondialdehyde (MDA), which is generated by peroxidation of polyunsaturated fatty acids. MDA levels were significantly higher in Cth^−/− hippocampus than in WT littermates (Fig. 3C).

Fig. 3.

CSE depletion elevates oxidative stress. (A) Elevated 4-hydroxy-2-nonenal (4-HNE) staining (red) in the hippocampal subfields, CA1, dentate gyrus (DG), and CA3 in the Cth^−/− mice. NeuN (green) staining was used for staining neurons. n = 5, (Scale bar, 50 μm.) (B) Quantitation of 4-HNE staining in the hippocampal subfields, CA1, dentate gyrus (DG), and CA3 in the Cth^−/− mice. Each point represents an individual mouse. (n = 4, mean ± SEM. Student’s t test. ****P < 0.0001). (C) Increased basal levels of malondialdehyde (MDA), a product of lipid peroxidation in Cth^−/− mice. (n = 5). Each point represents an individual mouse. (n = 4 to 5, mean ± SEM. Student’s t test. ****P < 0.0001). (D) Increased protein carbonylation in the Cth^−/− mice. Hippocampal lysates were treated with 2,4-dinitrophenylhydrazine (DNPH), which reacts with protein carbonyl groups generated by oxidation. The DNPH-derivatized samples were electrophoresed and subjected to western blotting using anti-DNPH antibodies and quantitation (n = 5). (E) Quantitation of D. Each point represents an individual mouse. (n = 5, mean ± SEM. Student’s t test. *P < 0.05). (F) Increased basal levels of catalase-1 and superoxide dismutase 1 (Sod1) in Cth^−/− mice reflecting increased oxidative stress as revealed by western blot analysis (n = 5). (G) Quantitation of Catalase-I. (n = 5, mean ± SEM. Student’s t test. *P < 0.05). (H) Quantitation of superoxide dismutase 1 (SOD). (n = 5, mean ± SEM. Student’s t test. *P < 0.05). (I) Increased DNA damage as assessed by 8-hydroxy-2-deoxyguanosine (8-OHdG, in green) staining in the hippocampi of Cth^−/− mice. (Scale bar, 10 μm, Zoom = 5 μm.) (J) Quantitation of 8-OHdG in K. Each point represents an individual mouse. (n = 5, mean ± SEM. Student’s t test. ****P < 0.0001). (K) Accumulation of the DNA-damage marker γ-H2AX (red) in the hippocampus of Cth^−/− mice. DAPI (white) was used to visualize the nuclei. n = 4, (Scale bar, 20 μm.) (L) Quantitation of γ-H2AX in K. Each point represents an individual mouse. (n = 4, mean ± SEM. Student’s t test. Student’s t test. ***P < 0.001).

We next assessed protein carbonylation, another indicator of oxidative stress, using 2,4-dinitrophenylhydrazine (DNPH), which reacts with protein carbonyls produced by oxidants. This was increased in Cth^−/− mice compared to WT littermates (Fig. 3 D and E). Concordant with a perturbed redox state and our proteomic data, levels of the antioxidant enzymes catalase 1 and SOD1 were elevated in Cth^−/− mice (Fig. 3 F–H), consistent with a compensatory response to heightened oxidative stress.

CSE Depletion Elevates DNA Damage in the Brain.

Because oxidative stress induces DNA lesions and DNA damage is linked to neurodegeneration (47, 48), we evaluated DNA damage in Cth^−/− hippocampus. Immunostaining showed significantly elevated 8-hydroxy-2-deoxyguanosine (8-OHdG), an oxidized guanosine derivative, in 6- to 7-mo-old Cth^−/− mice compared to WT littermates (Fig. 3 I and J). Levels of γ-H2AX, the phosphorylated form of histone H2AX, a marker of DNA damage that correlates with the number of double-stranded DNA breaks and increases with aging (49, 50)^, were also substantially higher in Cth^−/− hippocampi, relative to WT littermates, confirming increased DNA damage (Fig. 3 K and L).

Loss of CSE Disrupts Iron Homeostasis in the Brain.

While low levels of iron are essential for various physiologic processes in the brain, excessive iron accumulation is neurotoxic (51). Notably, H₂S signaling and its metabolites are intricately linked to iron homeostasis. For example, iron nonenzymatically regulates H₂S production, while H₂S regulates iron uptake, transport, and accumulation (52, 53). Additionally, increased iron accumulation is associated with ferroptosis, an iron-dependent form of programmed cell death that occurs in aging-related neurodegenerative diseases (54). H₂S signaling may protect against this form of cell death (55, 56). Because proteomics indicated altered iron-related proteins (Fig. 2 H–J), we evaluated iron homeostasis in Cth^−/− hippocampus.

The iron homeostasis protein, Hebp1, was significantly increased in the hippocampi of CSE knockout mice in our proteomics study (57). Elevated levels of Hebp1 contribute to heme-induced cytotoxicity through apoptosis and have been suggested as both as a driver of neuronal cell death and an early biomarker of neurodegenerative disease (57). Our proteomics analysis indicated increased Hebp1 levels (Fig. 2 B and H), which we further validated by western blot to be an approximately fivefold increase in Cth^−/− brain, relative to WT littermates (SI Appendix, Fig. S3 A and B). We next assessed changes in iron transport proteins in Cth^−/− mice, as tissue iron levels are regulated by both cellular storage and transport. Ferritin, the primary intracellular protein that binds ferric iron in the brain, (51) exhibited higher levels in the brains of Cth^−/− mice, relative to WT littermates, as reflected by immunohistochemical staining of Ftl (SI Appendix, Fig. S3 C and D). Cth^−/− mice also displayed reduced levels of the transferrin receptor protein (Tfrc), which mediates iron transport into cells (SI Appendix, Fig. S3 E and F). These findings suggest activation of compensatory mechanisms to limit iron intake in Cth^−/− mice, an expected response to excessively high iron levels. Last, Perls staining confirmed increased iron deposition in Cth^−/− mice (SI Appendix, Fig. S3G). Together, these findings indicate disrupted iron handling and accumulation following CSE loss.

CSE Depletion Impairs Blood–Brain Barrier Integrity.

BBB breakdown is an early biomarker of cognitive decline, including in the initial stages of AD (42, 58–60), and H₂S donors preserve BBB integrity under various pathological conditions and injuries (45). The specific role of CSE in BBB maintenance, however, has not been defined. Given the observed decrease in CD34 levels (Fig. 2R) and increase in Ighg within the brain parenchyma (Fig. 2S) in CSE-deficient mice, we employed transmission electron microscopy (TEM) to obtain a more definitive structural assessment. Using TEM, we observed frequent disruptions of hippocampal capillary endothelium, breaches of the BBB, and damage to astrocytic end-feet in Cth^−/− mice, features that were absent in WT littermates (Fig. 4 A and B). In addition to the hippocampus, Cth^−/− mice exhibited increased astrocytic end-feet swelling in the cortex as compared to WT littermates (SI Appendix, Fig. S4 A, B). Consistent with structural disruption, western blot analysis (Fig. 4 C and D) and immunohistochemistry (Fig. 4 E–H) showed increased parenchymal infiltration of peripheral immunoglobulin G (IgG) infiltration in Cth^−/− brain, indicating compromised BBB function.

Fig. 4.

Loss of CSE triggers breakdown of the BBB. (A) Transmission electron microscopy of the hippocampus shows BBB capillary endothelium breaks (black arrow) in the Cth^−/− mice, but not in wild-type mice. (Scale bar, 1 μm). (B) Quantitation of BBB and endfeet reveals significantly increased damage in the Cth^−/− mice as compared to the wild-type mice. Each point represents an individual mouse (BBB: n = 5, mean ± SEM. Student’s t test. ****P < 0.0001; Endfeet: n = 5, mean ± SEM. Student’s t test. ****P < 0.0001). (C) Increased extravasation of IgG in the hippocampus of 6-mo-old Cth^−/− mice as compared to wild-type controls as revealed by western blot analysis (n = 4). (D) Quantitation of C. (n = 4, mean ± SEM. Student’s t test. ****P < 0.0001). (E) Representative image showing IgG extravasation in the hippocampus parenchyma of 6-mo-old Cth^−/− mice as revealed by DAB staining. (F) Quantitation of E. Each point represents an individual mouse (n = 3, mean ± SEM. Student’s t test. **P < 0.01). (G) Representative immunofluorescence image showing IgG extravasation in the hippocampus parenchyma in the hippocampus of 6-mo-old Cth^−/− mice as revealed by IgG staining. (H) Quantitation of G. (n = 3, mean ± SEM. Student’s t test. **P < 0.01). (I) Reduced pericyte coverage (CD 13, red) of blood vessel endothelial cells (CD 31, green) in the hippocampus of 6-mo-old Cth^−/− mice, relative to wild-type controls. (J) Quantitation of I. Each point represents an individual mouse (n = 4 to 5, mean ± SEM. Student’s t test. *P < 0.05). (Scale bar, 50 μm.)

We further investigated pericyte coverage of BBB endothelial cells, which is reduced in both AD and chronic injury associated with traumatic brain injury (44, 61). Pericytes modulate several neurovascular functions, including BBB maintenance, vascular stability, capillary blood flow, and clearance of toxic molecules from the brain (62), and we observed that pericyte coverage of BBB endothelial cells was significantly reduced in Cth^−/− mice, relative to WT littermates (Fig. 4 I and J). Thus, loss of CSE compromises BBB integrity.

CSE Deficiency Impairs Postnatal Hippocampal Neurogenesis.

Because Cth^−/− mice display cognitive deficits, we evaluated hippocampal neurogenesis, a process critical for hippocampal plasticity that drives learning and memory. Neurogenesis declines with aging and in neurodegenerative diseases, including models of AD (42, 61), and H₂S donors have been shown to promote neurogenesis (63, 64).

Markers of immature neurons were markedly reduced in Cth^−/− hippocampus. Expression of doublecortin (DCX), a microtubule associated protein expressed in migrating and differentiating neurons (60), was profoundly diminished in Cth^−/− mice compared to WT littermates (Fig. 5 A and F). Polysialylated form of the neural cell adhesion molecule (PSA-NCAM), which similarly marks immature neurons (65), was also reduced (Fig. 5 B and G) and the number of DCX⁺ PSA-NCAM⁺ cells was significantly lower in Cth^−/− mice (Fig. 5 B and H).

Fig. 5.

Impaired postnatal hippocampal neurogenesis in Cth^−/− mice. (A) Representative immunofluorescence image of the dentate gyrus shows decreased expression of doublecortin (DCX; red) in Cth^−/− mice, compared to wild-type mice (Scale bars, 200 and 20 μm.) DAPI (blue) was used to stain the nuclei. (B) Representative immunofluorescence image of the dentate gyrus shows colocalization of DCX (red) and polysialylated neural cell adhesion molecule (PSA-NCAM, green). Expression of DCX⁺PSA-NCAM⁺ cells were profoundly diminished in Cth^−/− mice, compared to wild-type mice (Scale bars, 100 and 20 μm.) DAPI (blue) was used to stain the nuclei. (C) Representative immunofluorescence image of the dentate gyrus shows decreased expression of Ki67 (cyan) in Cth^−/− mice, compared to wild-type mice (Scale bars, 10 μm.) DAPI (blue) was used to stain the nuclei. (D) Representative immunofluorescence image of the hippocampus shows decreased expression of SOX2 (green staining) in Cth^−/− mice, compared to wild-type mice (Scale bars, 20 μm.) DAPI (blue) was used to stain the nuclei (E) Adult hippocampal neurogenesis as assessed by BrdU staining. Mice were injected intraperitoneally with BrdU at 150 mg/kg and analyzed 30 d later for BrdU⁺ cells, which were significantly decreased in Cth^−/− mice, compared to wild-type mice (Scale bars, 20 μm.) (F) Quantitation of A (DCX⁺ cells). Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, ***P < 0.001). (G) Quantitation of B (PSA-NCAM⁺ cells). Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, ****P < 0.0001). (H) Quantitation of B (DCX⁺PSA-NCAM⁺ cells). Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, ***P < 0.001). (I) Quantitation of C (Ki67⁺ cells). Each point represents an individual mouse (n = 4, mean ± SEM, two-tailed unpaired t test, ****P < 0.0001). (J) Quantitation of D (SOX2⁺ cells). Each point represents an individual mouse (n = 4 to 5, mean ± SEM, two-tailed unpaired t test, ****P < 0.0001). (K) Quantitation of BrdU⁺ cells. Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, ***P < 0.001).

We further examined the number of neuronal progenitor cells in Cth^−/− mice. During the process of adult hippocampal neurogenesis (AHN), Ki67- and sex-determining region Y (SRY)-box 2 (SOX2)-expressing neural stem cells in the subgranular zone of the hippocampus divide to generate cells characterized by a high proliferative capacity (66, 67). Ki-67 is present in actively dividing cells (68) and SOX2 is a transcription factor that regulates the multipotency of neural stem cells (67, 69). Notably, SOX2 deficiency impairs adult hippocampal neurogenesis, and reduced SOX2 is a feature of mouse and human AD, with the degree of depletion correlating with disease severity (69, 70). Expression of both Ki67 and SOX2 was significantly decreased in the hippocampus of 6-mo-old Cth^−/− mice, compared to WT littermates (Fig. 5 C, D, I, and J), indicating reduced proliferation of neural stem cells.

Consistent with these findings, survival of newborn hippocampal neurons was impaired, as assessed by established methods with 5-bromo-2’-deoxyuridine (BrdU) labeling (71, 72), (Fig. 5 E and K), reflecting impaired adult hippocampal neurogenesis.

CSE Deficiency Dysregulates Neurogenic Signaling.

Consistent with reduced neurogenesis, decreased levels of multiple neurotrophins and neurogenesis-regulating genes were observed in Cth^−/− brain as assessed using a pathway focused PCR array (Fig. 6A). These included neurogenin-2 (Neurog2), achaete-scute homolog 1 (Ascl1), brain-derived neurotrophic factor (Bdnf), cAMP response element-binding protein (Creb), and neurotrophin-3 (Ntf3) (Fig. 6 A–F), paralleling patterns seen in aging and AD (73). Conversely, expression of paired box gene 3 (Pax3), a transcription factor that negatively regulates neurogenesis, and C-X-C chemokine ligand (Cxcl1), which is elevated in human AD, were increased in Cth^−/− mice (Fig. 6 G and H), reflecting dysregulated neurogenesis (74, 75).

Fig. 6.

Impaired neurotrophin signaling in Cth^−/− mice. (A) PCR-microarray shows decreased levels of neurogenesis-related transcripts in Cth^−/− brain, relative to wild-type controls. (B–H) Quantitation of expression of genes involved in neurogenesis (NeuroG2, Ascl1, Bdnf, Creb, Pax3, and Cxcl1) by quantitative RT-PCR reveals a decreased levels of genes promoting neurogenesis and neurotrophin encoding genes, and increased expression of genes involved in negative regulation of neurogenesis in Cth^−/− hippocampus, compared to wild-type controls. We observed a trend of decrease in Ntf3 expression in Cth^−/− hippocampus. Each point represents an individual mouse (n = 3 to 4, mean ± SEM, two-tailed unpaired t test, ****P < 0.0001, ***P < 0.001, **P < 0.01). (I) Western blot showing decreased cAMP response element binding protein (CREB) activation in the Cth^−/− hippocampus, compared to wild-type controls. (J) Decreased levels of phosphorylated CREB (p-CREBS133) in Cth^−/− hippocampus, compared to wild-type controls (n = 5, mean ± SEM, two-tailed unpaired t test, *P < 0.05). (K) Levels of total hippocampal CREB were equivalent in Cth^−/−mice and wild-type controls (n = 5, mean ± SEM, two-tailed unpaired t test, ns). (L) The p-CREBS133/total CREB ratio was diminished in the hippocampus of 6- to 7-mo-old Cth^−/− mice, compared to wild-type controls Each point represents an individual mouse (n = 5, mean ± SEM, two-tailed unpaired t test, *P < 0.05). (M–O) Altered BDNF levels in the Cth^−/− hippocampus. (N) Levels of pro-BDNF in the Cth^−/− hippocampus, compared to wild-type controls as assessed by western blot analysis (n = 5, mean ± SEM, two-tailed unpaired t test, ns). (O) Levels of mBDNF in the Cth^−/− hippocampus, compared to wild-type controls (n = 5, mean ± SEM, two-tailed unpaired t test, *P < 0.05). (P) Model for processes contributing to learning and memory deficits. cAMP-response element binding protein (CREB), a key transcription factor involved in learning and memory formation may be activated through multiple pathways. Activation of adenylyl cyclase results in increased conversion of ATP to cAMP. cAMP stimulates activity of protein kinase A (PKA), which phosphorylates CREB at Ser133. This residue may also be phosphorylated by other kinases. Phosphorylated CREB activates transcription from promoters harboring the cAMP-response element (CRE)leading to increase of proteins involved in regulation of neurogenesis and cognitive functions, including brain-derived neurotrophic factor (BDNF), which is processed to its mature form (mBDNF). mBDNF acts on its receptor, TrkB. Signaling by BDNF modulates the activity of N-methyl D aspartate (NMDA) receptor and AMPA receptors, which results in influx of Ca²⁺ ions, which mediates a signaling cascade through phosphorylation of proteins including CAM kinases. H₂S produced by CSE may augment these processes at multiple levels, including activation of NMDA receptors, increasing cAMP levels, stimulation of CREB phosphorylation and modulating BDNF levels.

To further assess the mechanism by which loss of CSE diminishes neurogenesis and cognitive functions, we examined key signaling proteins involved in the process. The transcription factor CREB plays central roles in learning, memory, and synaptic plasticity, which are dependent on transcriptional activity of CREB (76–79). CREB activity is regulated through phosphorylation, a process impaired in AD (80). Key kinases involved in CREB activation include cAMP-dependent protein kinase (PKA) and Ca²⁺/calmodulin-dependent protein kinase (CaMK), both of which phosphorylate CREB at serine 133 (Ser133) (79, 81, 82), which facilitates its interaction with the transcriptional coactivators CBP/p300 to promote transcriptional activation of target genes (83). Analysis of CREB in hippocampal lysates from mice lacking CSE revealed a decrease in Ser133 phosphorylation of CREB (Fig. 6 I–L). Additionally, the PI3K/Akt signaling cascade also regulates CREB phosphorylation, which influences hippocampal progenitor proliferation and differentiation (84) and role of activated Akt in the process is another possibility.

Activation of CREB stimulates the transcription of proteins involved in synaptic plasticity, most notably BDNF. As a pivotal regulator of synaptic plasticity and memory formation, BDNF plays a critical role in maintaining neuronal integrity. Reduced BDNF levels occur in AD, which may cause the progressive neuronal atrophy observed in affected brain regions (85, 86). Altered levels of BDNF or a disruption of BDNF-TrkB signal pathway, where TrkB is the receptor for mature BDNF (mBDNF) have been linked to synapse loss and cognitive dysfunction (87, 88). We analyzed levels of BDNF and observed that levels of mBDNF were dysregulated in the hippocampus of Cth^−/− mice (Fig. 6 M–O). Together, these changes indicate multilevel disruption of memory-related signaling associated with reduced neurogenesis (Fig. 6P).

To assess whether neurogenesis impairment occurred earlier in the lifespan of Cth^−/− mice, we evaluated neurogenesis in younger 2-mo-old Cth^−/− mice and their WT littermates. Immunohistochemical analysis for DCX and PSA-NCAM showed normal neurogenesis levels in these younger Cth^−/− mice, ruling out a developmental cause for the postnatal neurogenesis deficiency in 6-mo-old mice (SI Appendix, Fig. S5 A–D). Thus, Cth^−/− mice prematurely exhibit an age-dependent decline in postnatal neurogenesis.

Discussion

CSE is one of the three principal enzymes responsible for H₂S production in mammals, alongside CBS and 3-MST. Historically, CBS was considered the dominant source of H₂S in the brain, while CSE’s contribution was thought to be minimal. However, emerging evidence has revealed that CSE plays a significant signaling role in the central nervous system, with its expression and enzymatic activity notably reduced in AD. Despite these findings, the specific involvement of CSE in cognitive decline remained uncharacterized.

Using Cth^−/− mice, we combined unbiased proteomics, biochemical assays, histology, and behavioral testing to define the consequences of CSE depletion in the brain. Loss of CSE produced a constellation of pathological changes linked to neurodegeneration, including markedly increased oxidative damage (elevated levels of protein carbonylation and lipid peroxidation), elevated DNA damage, and iron deposition. Proteomic and biochemical data indicated perturbation of redox homeostasis and iron-handling pathways, consistent with ferroptosis-relevant processes.

One of the most striking consequences of CSE deficiency is the deterioration of the blood–brain barrier (BBB). TEM revealed endothelial breaches and astrocytic end-feet damage in hippocampal capillaries of Cth^−/− mice. These ultrastructural abnormalities were accompanied by reduced pericyte coverage and abnormal parenchymal infiltration of IgG, hallmarks of compromised BBB integrity. Given that BBB breakdown is an early feature of cognitive decline, these results position CSE as a critical regulator of neurovascular unit stability and a protector against peripheral insults. Our results demonstrate that BBB disruption in CSE-deficient mice is not confined to the hippocampus but is also evident in the cortex, thereby broadening the scope of vascular vulnerability associated with CSE deficiency. Mechanistically, CSE plays a central role in the transsulfuration pathway, which includes the metabolism of homocysteine. Its genetic ablation therefore causes elevated plasma homocysteine levels, as previously reported (89), which was confirmed in the 6-mo-old CSE-deficient mice used in this study. Hyperhomocysteinemia is a recognized independent risk factor for cardiovascular and neurodegenerative disorders such as dementia, Alzheimer’s disease (AD), and Parkinson’s disease (PD) (90–92). Elevated homocysteine has also been implicated in BBB disruption (93). Taken together, these findings support the hypothesis that CSE ablation-induced redox imbalance, partly mediated by homocysteine accumulation, may contribute to BBB dysfunction. This is likely compounded by additional abnormalities arising from impaired H₂S signaling and reduced protein sulfhydration.

CSE deficiency impairs postnatal hippocampal neurogenesis, which is critical for learning and memory, processes disrupted in AD (94, 95). Cth^−/− mice exhibited reduced expression of neurogenesis markers (DCX, PSA-NCAM), fewer proliferating neural stem cells (Ki67+, SOX2+), and diminished survival of newborn neurons (BrdU+). These deficits were age-dependent, emerging in 6-mo-old mice but absent in younger cohorts, indicating a premature decline in neurogenic capacity. Given the hippocampus’s central role in memory formation, impaired neurogenesis likely contributes directly to the cognitive deficits observed in these mice. At the molecular level, CSE deficiency disrupted key pathways involved in neurogenesis and synaptic plasticity. Expression of neurotrophins and transcription factors such as Neurog2, Ascl1, BDNF, CREB, and Ntf3 was significantly reduced, while negative regulators like Pax3 and Cxcl1 were elevated. CREB phosphorylation, a critical step for memory-related gene transcription, was impaired, along with decrease in production of BDNF. These changes compromise the transcriptional machinery required for neuronal survival, differentiation, and synaptic function.

The cumulative impact of BBB breakdown, impaired neurogenesis, and disrupted signaling cascades manifests as cognitive deficits in Cth^−/− mice. These deficits are likely driven by reduced hippocampal plasticity, diminished synaptic integrity, and compromised neurovascular support. The dysregulation of BDNF and its receptor TrkB, both essential for memory consolidation, further underscores the link between CSE depletion and cognitive dysfunction. Our findings are consistent with reports in literature, where H₂S donors were shown to promote neurogenesis (63, 64, 96); however, the role of CSE in neurogenesis has been demonstrated in this study.

While depletion of CSE may impact neuronal H₂S production, H₂S levels were not directly measured in our study, nor was H₂S replacement therapy employed to evaluate the reversibility of the phenotypes observed in CSE knockout mice. Our previous work has demonstrated reduced protein sulfhydration in CSE-deficient brains (17), which may disrupt signaling pathways critical for cognitive function and neuroprotection. Moreover, we have shown that H₂S donors can restore sulfhydration and mitigate neurodegeneration in Alzheimer’s disease models where CSE activity is compromised (25). Sulfhydration is not a direct consequence of H₂S itself, and H₂S cannot directly modify cysteine residues. Importantly, H₂S levels and polysulfide concentrations do not necessarily correlate under physiological or pathological conditions (97). This distinction is particularly relevant in disease states, where altered redox environments or enzymatic activities may differentially affect the formation and stability of reactive sulfur species. Further studies are needed to define the downstream chemical pathways altered by CSE loss or dysfunction. There exists a possibility that some of the observed effects may stem from non-H₂S-related functions of CSE and it remains possible that some of the observed effects may arise from non-H₂S-related functions of CSE. Although measurements of cysteine and GSH levels in the hippocampus of CSE-deficient mice revealed no significant alterations, we cannot exclude the possibility that subtle changes in these endogenous molecules at critical subcellular or regional sites may exert profound biological effects.

The CSE/H₂S axis may further influence gene regulatory processes that may be impacted with the loss of CSE. It has been shown that H₂S donors induce transcriptomic changes, throughout the animal and plant kingdoms as well as in bacteria (15). While gene promoters may harbor response elements for various transcription factors, currently there is no evidence of a response element on DNA that directly binds H₂S or its derivatives. However, H₂S may influence gene regulation through several mechanisms. Sulfhydration of transcription factors and their coregulators may alter their activity to modulate gene expression. For example, H₂S-mediated sulfhydration regulates NF-κB (98), Keap1 (99, 100), and SIRT1 (101). Additionally, H₂S may influence chromatin structure and accessibility, or influence epigenetic changes, further contributing to transcriptional regulation (15, 102). In this study, we observed extensive proteomic alterations in the mice lacking CSE, which may stem from changes in transcription factor abundance or activity. For instance, transcript levels of Pax3, a transcription factor known to negatively regulate neurogenesis, was upregulated in Cth^−/− mice. Additionally, signaling mediated by CREB, a positive regulator of genes involved in learning and memory was impaired. Although total CREB protein levels remained unchanged, its phosphorylated (active) form was markedly decreased, suggesting impaired transcriptional activation of genes regulated by CREB. These findings demonstrate that CSE deficiency disrupts the transcriptional landscape, particularly affecting factors involved in neuronal differentiation. Additional transcription factors may also be impacted, either directly through altered redox signaling or indirectly via downstream pathways. This represents a promising avenue for future investigation.

In summary, CSE loss initiates a multilevel cascade of oxidative stress, DNA and iron-related damage, BBB breakdown, impaired neurogenesis, and disrupted CREB/BDNF signaling that converges on reduced hippocampal plasticity and cognitive dysfunction. These findings establish CSE as a central regulator of brain homeostasis and suggest that restoring CSE activity or downstream H₂S signaling may be a therapeutic strategy to preserve neurovascular integrity, support neurogenesis, and mitigate cognitive decline in aging and neurodegenerative disease. Future research will explore the interplay between CSE-mediated sulfhydration and other posttranslational modifications, such as nitrosylation and cyanylation, to further elucidate the roles of atypical gaseous neurotransmitters including H₂S, NO, carbon monoxide (CO), and hydrogen cyanide (HCN) and to identify therapeutic opportunities (103–109).

Materials and Methods

Animals.

The study protocol (MO21M457) was approved by the Institutional Animal Care and Use Committee of Johns Hopkins University. All animal procedures were performed in accordance with the NIH guidelines (Guide for the Care and Use of Laboratory Animals). The generation and breeding scheme of Cth^−/− mice were described previously (27, 89). These mice were backcrossed for nine generations to the C57BL/6J mice. Mice were grouped-housed in individually ventilated cages with five mice per cage. Corn cob was used as bedding and nestlets were provided as nesting material. All the mice were maintained in a controlled environment at a temperature of 22 °C ± 1 °C, with a light–dark cycle of 12 h (lights on at 07:00) and provided with a normal chow diet (containing 13% fat, 23.3% protein, and 63.7% carbohydrate; LabDiet) and water ad libitum.

Behavioral Analysis.

Both cognitive and locomotor tests were conducted.

Cognitive test.

The Barnes Maze test was performed as described previously, with minor modifications (25). In brief, the acquisition was performed in a block of four trials for six consecutive days. Escape latencies were recorded and the learning curve was plotted by averaging four trials.

Locomotor test.

The open field was performed in an activity chamber (Photobeam Activity System; San Diego Instruments, San Diego, CA, USA). The mice were placed in the center of the activity chamber installed with infrared beams, and activity was monitored for 45 min. The cumulative and session-wise beam breaks in the x and y directions were recorded and plotted as total ambulatory activity.

Sensorimotor tests.

We assessed sensorimotor deficits using the tape removal test. The test was performed as described previously (110). In addition, a buried food test (Hidden cookie test) was conducted as described previously (111) to check any gross malfunctioning of the olfactory system.

Immunoblotting.

Tissues were microdissected and lysed in cold IP buffer containing 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% Triton X-100, protease inhibitors (cOmplete, EDTA-free Protease Inhibitor Cocktail from Sigma), and protein phosphatase inhibitor (PhosSTOP™ phosphatase inhibitor, Sigma). Lysates were centrifuged at 14,000 g for 20 min, and the supernatant was collected. The total protein content was analyzed using Bradford assay (Bio-Rad), and samples were normalized for equal load. Protein samples were prepared by adding 1× final concentration of NuPAGE LDS Sample Buffer (Invitrogen) with DTT followed by incubation at 95 °C for 5 min. Subsequently, protein samples were loaded on a NuPAGE 4 to 12% Bis-Tris gel (Thermo Fisher, Scientific USA), electrophoresed in 1× NuPAGE MES SDS running buffer (Thermo Fisher, Scientific USA), and transferred to Immobilon-FL (Millipore). Membranes were blocked for 1 h at room temperature with 5% BSA or 5% milk in TBS-tween followed by immunoblotting with the indicated antibodies. The primary antibodies and dilutions used were as follows: anti-CSE (1:1,000), anti-CBS (1:1,000), anti-3-MST (1:1,000), anti-pCREB (1:1,000), anti-CREB (1:1,000), anti-DNP (1:150), anti-SOD (1:1,000), anti-Hebp1 (1:1,000), anti-IgG (1:500). Horseradish Peroxidase (HRP) conjugated secondary antibodies (1:10,000) were used for detection with SuperSignal West Pico chemiluminescence reagent (Thermo Fisher, Scientific USA). Membranes were imaged on Bio-Rad ChemiDoc Touch Imaging System (Bio-Rad Laboratories). Band intensities were quantified Fiji Image J (NIH, Bethesda, MD) and normalized to respective loading controls.

Additional details of reagents and methods are available in the SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Data, Materials, and Software Availability

Study data are included in the article and/or supporting information.