Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 May 31;195(2):1561-1585.
doi: 10.1093/plphys/kiae059.

Gene expression and metabolite levels converge in the thermogenic spadix of skunk cabbage

Haruka Tanimoto  1 , Yui Umekawa  2 , Hideyuki Takahashi  3 , Kota Goto  4 , Kikukatsu Ito  1   4
Affiliations

Gene expression and metabolite levels converge in the thermogenic spadix of skunk cabbage

Haruka Tanimoto et al. Plant Physiol. .

Abstract

The inflorescence (spadix) of skunk cabbage (Symplocarpus renifolius) is strongly thermogenic and can regulate its temperature at around 23 °C even when the ambient temperature drops below freezing. To elucidate the mechanisms underlying developmentally controlled thermogenesis and thermoregulation in skunk cabbage, we conducted a comprehensive transcriptome and metabolome analysis across 3 developmental stages of spadix development. Our RNA-seq analysis revealed distinct groups of expressed genes, with selenium-binding protein 1/methanethiol oxidase (SBP1/MTO) exhibiting the highest levels in thermogenic florets. Notably, the expression of alternative oxidase (AOX) was consistently high from the prethermogenic stage through the thermogenic stage in the florets. Metabolome analysis showed that alterations in nucleotide levels correspond with the developmentally controlled and tissue-specific thermogenesis of skunk cabbage, evident by a substantial increase in AMP levels in thermogenic florets. Our study also reveals that hydrogen sulfide, a product of SBP1/MTO, inhibits cytochrome c oxidase (COX)-mediated mitochondrial respiration, while AOX-mediated respiration remains relatively unaffected. Specifically, at lower temperatures, the inhibitory effect of hydrogen sulfide on COX-mediated respiration increases, promoting a shift toward the dominance of AOX-mediated respiration. Finally, despite the differential regulation of genes and metabolites throughout spadix development, we observed a convergence of gene expression and metabolite accumulation patterns during thermogenesis. This synchrony may play a key role in developmentally regulated thermogenesis. Moreover, such convergence during the thermogenic stage in the spadix may provide a solid molecular basis for thermoregulation in skunk cabbage.

PubMed Disclaimer

Conflict of interest statement

Conflict of interest statement. None declared.

Figures

Figure 1.
Figure 1.
Thermogenesis of S. renifolius. A) Images showing the thermogenic stage of intact S. renifolius (upper left), a thermal image capturing the surface temperature (lower left), and a longitudinal section indicating the positions of florets and pith (right). B) Comparative data for air temperature (Ta), spadix temperature (Ts), and their difference (Ts−Ta). The stages of thermogenesis are denoted as “Pre” (prethermogenic), “Hot” (thermogenic), and “Post” (postthermogenic). The mean values and standard deviations from 4 independent measurements are presented in bar graphs with error bars, respectively. Statistically significant differences are indicated by different letters above the bars (Tukey–Kramer test: P < 0.05). C) Details of the analyzed samples, indicated by labels corresponding to the thermogenic stages of the spadix in S. renifolius. Stages are labeled as “Pre_x” for prethermogenic, “Hot_x” for thermogenic, and “Post_x” for postthermogenic samples, where “x” denotes the specific sample number. Each label also includes information on the spadix temperature (Ts), the sampling site [either Fujine (F) or Kanegasaki (K)], and the analysis method used (RNA-seq or CE-MS). Empty boxes in “Pre_1” and “Pre_4” denote samples that were not subjected to CE-MS or RNA-seq analysis due to insufficient tissue quantity resulting from the small size of the pith tissues at the prethermogenic spadices. CE-MS, capillary electrophoresis-mass spectrometry.
Figure 2.
Figure 2.
Differentially expressed genes (DEGs) in the thermogenic spadices of S. renifolius.A) Volcano plots highlighting the DEGs between the florets and pith at various thermogenic stages. The threshold for DEG detection [P-value < 0.05 and |fold change (FC)| > 2] is shown with dashed lines. B) Hierarchical clustering and heatmap analysis of DEGs. Samples are designated by stage and tissue type (Fig. 1C). Normalized expression values are presented as z-scores based on TPM values. C) GO enrichment analysis for the DEGs, segregated by tissue type (florets or pith) and thermogenic stage in the spadices (Pre, Hot, or Post). Each panel corresponds to 1 of the 3 GO categories: biological process (top), cellular component (middle), and molecular function (bottom). The top 5 enriched GO terms for each tissue type and thermogenic stage were identified, and overlapping terms were excluded from the combined list. Circle size corresponds to the number of genes associated with each term, and the intensity reflects the −log10 (P-value). The stages of thermogenesis in the spadices are denoted as “Pre” (prethermogenic), “Hot” (thermogenic), and “Post” (postthermogenic). TPM, transcripts per million.
Figure 3.
Figure 3.
Dynamic gene expression patterns during spadix development in S. renifolius. A) Nonhierarchical k-means clustering delineates 25 distinct expression profiles for identified differentially expressed genes (DEGs). Individual gene expression trajectories within each profile are depicted in thin lines, with the average trajectory for each cluster highlighted by a bold line. Thermogenic stages and sampled tissues are denoted by color-coded bars along the horizontal axis. B, C) Gene expression changes in florets (B) and pith (C) throughout spadix development. DEGs are organized into 6 distinct groups based on k-means clustering. Homologous gene expression patterns between the 2 tissues are grouped and marked with identical numbers, with “F” denoting florets and “P” indicating pith in both panels. The scatter plots in both panels show TPM values, with each dot representing the calculated average expression level of a gene across analyzed samples. The dots are color-coded to indicate different thermogenic stages, and the corresponding k-means clustering numbers are provided. D, E) Representative genes and their expression levels in florets and pith from the groups shown in panels (B) and (C). The top 5 genes by TPM values at each thermogenic stage for each tissue were selected, with overlaps excluded from the combined list. TPM values are normalized to z-scores, with expression levels visualized on a heatmap. Gene identifiers, BLAST descriptions, k-mean clustering numbers, and group identifications are shown. The stages of thermogenesis in the spadices are denoted as “Pre” (prethermogenic), “Hot” (thermogenic), and “Post” (postthermogenic). TPM, transcripts per million.
Figure 4.
Figure 4.
Analysis of the top 10 differentially expressed genes (DEGs) in the florets and pith during the thermogenic stage of spadices in S. renifolius. Comparisons of TPM values across prethermogenic (Pre), thermogenic (Hot), and postthermogenic (Post) stages of spadices for A) florets and B) pith are shown. Genes are listed according to their expression levels, and corresponding values for the prethermogenic and postthermogenic stages are also shown for comparison. The bar graphs represent the average TPM values of multiple independent samples (n = 4 except for prethermogenic pith: n = 3) ± standard deviation. Statistically significant differences, as determined by the Tukey–Kramer test (P < 0.05), are denoted by different letters beside the bar plots. The gene categorization based on group and k-means cluster identification (Fig. 3B and C), is indicated to the left of each panel. Gene identifiers and putative functions are also given. TPM, transcripts per million.
Figure 5.
Figure 5.
Metabolite profiling of florets and pith across various thermogenic stages in the spadices of S. renifolius. A) Principal component analysis (PCA) plot of the analyzed metabolites. PC1 (31.6%) contributes to the separation of tissues with thermogenic status, while PC2 (17.7%) reflects the transition of thermogenic stages. Colored areas for each stage and tissue represent 95% confidence intervals (MetaboAnalyst). Arrows indicate the developmental progression from the prethermogenic (Pre), through thermogenic (Hot), to postthermogenic (Post) stages of spadices in both florets and pith. B) Hierarchical clustering diagrams of metabolites, grouped according to a dissimilarity scale, are presented on the top (Clusters A and B) and on the left (Clusters I, II, and III). Metabolites, categorized as sugars and sugar phosphates, organic acids, amino acids, nucleotides, polyamines, and others, are denoted by different color labels. Each sample is identified by its thermogenic stage (Pre, Hot, or Post) and tissue type (F for florets, P for pith), as represented by respective identifiers (e.g. Hot_P4 refers to the 4th pith sample collected from the thermogenic stage of spadices). Normalized data of metabolite concentrations are represented by the z-score. C) PC loading values of metabolites in PC1 and PC2 from panel (A). Sugars and sugar phosphates, organic acids, amino acids, nucleotides, polyamines, and others are differentiated by colors as shown in the panel. D) Accumulation of metabolites expressed as fold changes in prethermogenic (Pre), thermogenic (Hot), and postthermogenic (Post) stages of spadices. Metabolites categorized as sugars and sugar phosphates, organic acids, amino acids, nucleotides, polyamines, and others are shown. Upregulated metabolites either in florets or pith are depicted with their fold changes in different colors. The log2 (|fold change (FC)|) is represented by the size of the circles, as indicated in the panel. E) A visualization of the changes in metabolite accumulation from prethermogenic (Pre), through thermogenic (Hot), to postthermogenic (Post) stages. Metabolites are organized into 9 distinct groups using k-means clustering. F) A representative display of metabolite accumulation profiles (Groups A to E) corresponding to the diverse stages of thermogenesis in the florets, alongside their respective k-means cluster numbers. Each stage of thermogenesis includes 4 biological replicates, each represented as a separate dot based on the z-score. G) Metabolites within each group are categorized into sugars and sugar phosphates, organic acids, amino acids, nucleotides, polyamines, and others, each of which is represented by a unique color.
Figure 6.
Figure 6.
Changes in the accumulation of adenylates and genes for AMP production in the spadices of S. renifolius. A) A comparison of adenylate nucleotide levels across prethermogenic (Pre), thermogenic (Hot), and postthermogenic (Post) stages of spadices in florets and pith. Data for AMP, ADP, ATP levels, AMP/ATP, ADP/ATP ratios, and energy charge are presented. Energy charge is defined by the following formula: (ATP + ½ ADP)/(ATP + ADP + AMP). The mean values and standard deviations from independent measurements (n = 4 except for Pre_Pith n = 3) are presented in bar graphs with error bars, respectively. Statistically significant differences are denoted by different letters above the bars (Tukey–Kramer test: P < 0.05). Specifically, lowercase letters denote comparisons within the “florets” group, while uppercase letters are used for comparisons within the “pith” group. Significant differences between the florets and the pith are indicated by the number of asterisks: 1 asterisk (*) denotes P < 0.05, 2 asterisks (**) denote P < 0.01, and 3 asterisks (***) denote P < 0.005, as determined by Student's t-test. Each thermogenic stage (Pre, Hot, or Post) and tissue type (florets or pith) are differentiated by distinct colors. B) List of genes potentially contributing to AMP production in the thermogenic florets, ranked by their expression levels. Genes that are upregulated in florets, with fold changes indicated by the size of the circles, are displayed. Averaged gene expression levels (n = 4), denoted by (log2 (TPM) + 1), are shown. Circles of different sizes represent the log2 fold change (FC) values, ranging from 2 to 8. Filled circles indicate upregulation in the florets. Gene identifiers and their putative functional names are also provided. DW, dry weight; TPM, transcripts per million.
Figure 7.
Figure 7.
Integrated overview of transcriptome and metabolome analyses in the thermogenic florets of S. renifolius. A) A comprehensive map of the central carbon metabolic pathways and mitochondrial metabolism. B) Interconversion of adenylate nucleotides and the activation of associated pathways. C) An overview of sulfur and methionine metabolism. The diagrams in panels (A) to (C) incorporate information on gene expression and metabolite accumulation in the florets and pith at the thermogenic stage of spadices. Color gradients in the lower panel (C) show variations in metabolite accumulation (log10 (nmol/mg DW)) and gene expression levels (log2 (TPM) + 1). Circles of different sizes represent the log2 fold change (FC) values, ranging from 2 to 8. Filled red circles indicate upregulation in the florets, blue filled circles represent upregulation in the pith, and gray filled circles denote nonsignificant differences. Arrows denote upregulation trends in either the florets (red) or the pith (blue), with bold arrows indicating a P-value < 0.05 and FC > 2 or FC < −2. Narrow arrows show a P-value < 0.05 and FC ≤ 2 or FC ≥ −2, and gray arrows point to nonsignificant results (P-value ≥ 0.05). DW, dry weight; TPM, transcripts per million. 2OG-DH, 2-oxoglutarate dehydrogenase; ACO, aconitate hydratase; AOX, alternative oxidase; ATPBM, ATP synthase subunit beta; ATPO, ATP synthase subunit O; CS, citrate synthase; CX5C3, cytochrome c oxidase subunit 5C-3; COX6A, cytochrome c oxidase subunit 6a; Cyt c, cytochrome c; dicarboxylate carrier (UCP4), dicarboxylate carrier (mitochondrial uncoupling protein 4); ENO, enolase; FBA, fructose-bisphosphate aldolase; FBPase, fructose-1,6-bisphosphatase; FRK, fructokinase; Fumarase, fumarate hydratase; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; HEX, hexokinase; ICDH, isocitrate dehydrogenase; INV, invertase; ME, malic enzyme; MST, monosaccharide transporter; NDA, internal alternative NAD(P)H:ubiquinone oxidoreductase; NDB, external alternative NAD(P)H:ubiquinone oxidoreductase; NDBAB, NADH dehydrogenase [ubiquinone] 1 beta subcomplex subunit 10-B; PDH, pyruvate dehydrogenase; PEPase, phosphoenolpyruvate phosphatase (bifunctional purple acid phosphatase); PEPC, phosphoenolpyruvate carboxylase; PEPCK, phosphoenolpyruvate carboxykinase; PFK, ATP-dependent 6-phosphofructokinase; PFK2/F2,6BPase, phosphofructokinase-2/fructose-2,6-bisphosphatase; PFP, pyrophosphate-fructose 6-phosphate 1-phosphotransferase; PGI, glucose-6-phosphate isomerase; PGK, phosphoglycerate kinase; PGM, phosphoglycerate mutase; PK, pyruvate kinase; PLT, polyol transporter; Pyr carrier, mitochondrial pyruvate carrier; STP, sugar transport protein; Suc-CoA ligase, succinate-CoA ligase; SUS, sucrose synthase; SWEET, sugars will eventually be exported transporter; UCP, mitochondrial uncoupling protein; USP, UDP-sugar pyrophosphorylase; TPI, triosephosphate isomerase; 1,3-BPG, 1,3-bisphosphoglycerate; 2OG, 2-oxoglutarate; 2PGA, 2-phosphoglycerate; 3PGA, 3-phosphoglycerate; AcCOA, acetyl-CoA; Cit, citrate; DHAP, dihydroxyacetone phosphate; F2,6BP, fructose 2,6-bisphosphate; F6P, fructose 6-phosphate; FBP, fructose 1,6-bisphosphate; Fum, fumarate; G1P, glucose-1-phosphate; G6P, glucose 6-phosphate; GA3P, glyceraldehyde 3-phosphate; Iso-cit, isocitrate; Mal, malate; OAA, oxaloacetate; PEP, phosphoenolpyruvate; Pyr, pyruvate; Suc, succinate; Suc-CoA, succinyl-CoA; UDP-Glu, uridine diphosphate glucose; ADK, adenylate kinase; ADSL, adenylosuccinate lyase; ADSS, adenylosuccinate synthetase; AMPD, AMP deaminase; APY, apyrase; Asn synthase, asparagine synthetase; APT, adenine phosphoribosyltransferase; adoK, adenosine kinase; NDK, nucleoside diphosphate kinase; NSH, ribonucleoside hydrolase; PNP, purine nucleoside phosphorylase; SurE, 5′-nucleotidase SurE; ADP, adenosine diphosphate; AMP, adenosine monophosphate; ATP, adenosine triphosphate; CDP, cytidine diphosphate; CMP, cytidine monophosphate; CTP, cytidine triphosphate; GDP, guanosine diphosphate; GMP, guanosine monophosphate; GTP, guanosine triphosphate; IMP, inosine monophosphate; S-AMP, adenylosuccinate; UDP, uridine 5′-diphosphate; UMP, uridine 5′-monophosphate; UTP, uridine triphosphate. AAT, aspartate aminotransferase; ACCO, 1-aminocyclopropane-1-carboxylate oxidase; ACCS, 1-aminocyclopropane-1-carboxylate synthase; APS, ATP sulfurylase; ARD, 1,2-dihydroxy-3-keto-5-methylthiopentene dioxygenase; CBL, cystathionine beta-lyase; CGS, cystathionine gamma-synthase; CIMS, 5-methyltetrahydropteroyltriglutamate-homocysteine methyltransferase; CYSC, bifunctional L-3-cyanoalanine synthase/cysteine synthase; DCD, D-cysteine desulfhydrase; DEP, bifunctional methylthioribulose-1-phosphate dehydratase/enolase-phosphatase E1; GGT, glutathione hydrolase; GSH1, glutamate-cysteine ligase; GSH2, glutathione synthetase; GTOMC, tocopherol O-methyltransferase; HMT, homocysteine S-methyltransferase; JMT, jasmonic acid carboxyl methyltransferase; LAP, leucine aminopeptidase; LCD, L-cysteine desulfhydrase; MGL, methionine gamma-lyase; MMT, methionine S-methyltransferase; MTAN, 5′-methylthioadenosine/S-adenosylhomocysteine nucleosidase; MTK, methylthioribose kinase; MTI, methylthioribose-1-phosphate isomerase; PRH, 5′-adenylylsulfate reductase; ROMT, trans-resveratrol di-O-methyltransferase; SAHH, adenosylhomocysteinase; SAMDC, S-adenosylmethionine decarboxylase; SAMS, S-adenosylmethionine synthase; SBP1/MTO, selenium-binding protein 1/methanethiol oxidase; SIR, sulfite reductase 1 [ferredoxin]; SPSY, spermine synthase; SULTR, sulfate transporter; MST, thiosulfate/3-mercaptopyruvate sulfurtransferase; 3-MP, 3-mercaptopyruvate; ACC, 1-aminocyclopropane-1-carboxylate; APS, adenosine 5′-phosphosulfate; C2H2, ethylene; CH3SH, methanethiol; Cys, L-cysteine; CYT, cystathionine; D-Cys, D-cysteine; DHKMP, 1,2-dihydroxy-3-keto-5-methylthiopentene; DMDS, dimethyl disulfide; DMTS, dimethyl trisulfide; GSH, glutathione; H2S, hydrogen sulfide; Hcy, homocysteine; KMTB, α-ketomethylthiobutyrate; Met, L-methionine; MTA, 5-methylthioadenosine; MTR, 5-methylthioribose; MTR-P, 5-methylthioribose 1-phosphate; MTRu-P, 5-methylthioribulose-1-phosphate; OAS, O-acetylserine; SAHcy, S-adenosyl-L-homocysteine; SAM, S-adenosylmethionine; Ser, L-serine; SMM, S-methylmethionine; SO32−, sulfite; SO42−, sulfate; dSAM, S-adenosylmethioninamine (decarboxy-AdoMet).
Figure 8.
Figure 8.
Impact of hydrogen sulfide on the respiration capacities of terminal oxidases in the mitochondria from thermogenic florets of S. renifolius. A) Endogenous hydrogen sulfide content in the thermogenic stage of the spadix. The analysis was performed on the florets and pith from an S. renifolius thermogenic spadix. Results are presented from 3 biological replicates. The error bars show the mean ± standard deviation. Statistically significant differences are indicated by 2 asterisks based on Welch’s t-test (P = 0.00434). B) Western blot analysis using total proteins extracted from thermogenic florets. Two primary antibodies targeting different epitopes, one for the N-terminal region (SBP1N) and the other for the C-terminal region (SBP1C) of the SBP1/MTO protein, were used. The protein loading was standardized to 20 μg per lane. Coomassie Brilliant Blue (CBB) staining is shown for reference. C) Effects of hydrogen sulfide on AOX- and COX-mediated respiration capacities under various temperatures. AOX and COX capacities at indicated hydrogen sulfide concentrations are shown with fitted curves as indicated in the panel. Arrows indicate the points of intersection between the curves of the 2 oxidases. Assays were performed in triplicate and error bars display the mean ± standard deviation. D) Percentage of AOX capacities in total oxygen consumption by different terminal oxidases. The ratio of AOX respiration was calculated as the AOX respiration rate (V(AOX)) divided by the sum of the AOX and COX respiration rates (V(AOX) + V(COX)). Assays were performed in triplicate and error bars display the mean ± standard deviation. In panels (C) and (D), quadratic curve fitting has been applied to the measurements. FW, fresh weight; SBP1/MTO, selenium-binding protein 1/methanethiol oxidase; AOX, alternative oxidase; COX, cytochrome c oxidase.
Figure 9.
Figure 9.
Convergence and unification of gene expression and metabolite accumulation patterns in the florets of the thermogenic spadices of S. renifolius. A) Representative images of S. renifolius at prethermogenic (Pre), thermogenic (Hot), and postthermogenic (Post) stages. In the prestage, the spathe is removed to reveal the spadix. B) Patterns of gene expression and metabolite accumulation in the florets across different stages. The symbols in the panel represent distinct gene expression/metabolite accumulation patterns in each group. The groups presented correspond to those for transcriptomic analysis (Fig. 3B) and for metabolomic analysis (Fig. 5F). C) Schematic overview of changes in gene expression and metabolic pathways during the development of the florets. Note that distinct patterns of gene expression during development largely merge maximally at the thermogenic stage, as shown by different symbols represented in the panel (B). Key shifts during the thermogenic stage include the upregulation of central carbon metabolism and mitochondrial respiratory chain components, including AOX. The SWEETs stimulate the glycolytic pathway. Increased ATP consumption is mediated by the NDK-driven nucleotide interconversion and the expression of asparagine synthetase. The elevation of AMP stimulates catabolic processes, thereby triggering a surge in metabolic heat-production. SBP1/MTO produces hydrogen sulfide by decomposing methanethiol, which represses COX activity and redirects metabolic flux toward the AOX respiratory pathway, contributing to metabolic heat-production. A shift toward methanethiol degradation and a reduction in the production of malodorous compounds (DMDS, DMTS) substantially diminishes the odor release from an intact spadix of S. renifolius. SBP1/MTO, selenium-binding protein 1/methanethiol oxidase; AOX, alternative oxidase; COX, cytochrome c oxidase; NDA, internal alternative NAD(P)H:ubiquinone oxidoreductase; NDB, external alternative NAD(P)H:ubiquinone oxidoreductase; NDK, nucleoside diphosphate kinase; SWEET, sugars will eventually be exported transporter; DMDS, dimethyl disulfide; DMTS, dimethyl trisulfide; AcCOA, acetyl-CoA; ADP, adenosine diphosphate; AMP, adenosine monophosphate; Asn, asparagine; Asp, aspartate; ATP, adenosine triphosphate; Cit, citrate; dNDP, deoxynucleoside diphosphate; dNTP, deoxynucleoside triphosphate; DHAP, dihydroxyacetone phosphate; F6P, fructose 6-phosphate; FBP, fructose 1,6-bisphosphate; G6P, glucose 6-phosphate; Mal, malate; NDP, nucleoside diphosphate; NTP, nucleoside triphosphate; OAA, oxaloacetate; PEP, phosphoenolpyruvate; PRPP, 5-phosphoribosyl 1-pyrophosphate; SAM, S-adenosylmethionine; Suc, succinate; 2OG, 2-oxoglutarate.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Albre J, Quilichini A, Gibernau M. Pollination ecology of Arum italicum (Araceae). Bot J Linn Soc. 2003:141(2):205–214. 10.1046/j.1095-8339.2003.00139.x - DOI
    1. Angeles-Núñez JG, Tiessen A. Arabidopsis sucrose synthase 2 and 3 modulate metabolic homeostasis and direct carbon towards starch synthesis in developing seeds. Planta. 2010:232(3):701–718. 10.1007/s00425-010-1207-9 - DOI - PubMed
    1. Baena-González E, Rolland F, Thevelein JM, Sheen J. A central integrator of transcription networks in plant stress and energy signalling. Nature. 2007:448(7156):938–942. 10.1038/nature06069 - DOI - PubMed
    1. Baena-González E, Sheen J. Convergent energy and stress signaling. Trends Plant Sci. 2008:13(9):474–482. 10.1016/j.tplants.2008.06.006 - DOI - PMC - PubMed
    1. Bansal MP, Oborn CJ, Danielson KG, Medina D. Evidence for two selenium-binding proteins distinct from glutathione peroxidase in mouse liver. Carcinogenesis. 1989:10(3):541–546. 10.1093/carcin/10.3.541 - DOI - PubMed
-