Frese AN et al. (2024), Quantitative proteome dynamics across embryogen...

XB-ART-60629

iScience 2024 Apr 19;274:109355. doi: 10.1016/j.isci.2024.109355.

Show Gene links Show Anatomy links

Quantitative proteome dynamics across embryogenesis in a model chordate.

Frese AN , Mariossi A , Levine MS , Wühr M .

???displayArticle.abstract???
The evolution of gene expression programs underlying the development of vertebrates remains poorly characterized. Here, we present a comprehensive proteome atlas of the model chordate Ciona, covering eight developmental stages and ∼7,000 translated genes, accompanied by a multi-omics analysis of co-evolution with the vertebrate Xenopus. Quantitative proteome comparisons argue against the widely held hourglass model, based solely on transcriptomic profiles, whereby peak conservation is observed during mid-developmental stages. Our analysis reveals maximal divergence at these stages, particularly gastrulation and neurulation. Together, our work provides a valuable resource for evaluating conservation and divergence of multi-omics profiles underlying the diversification of vertebrates.

???displayArticle.pubmedLink??? 38510129
???displayArticle.pmcLink??? PMC10951915
???displayArticle.link??? iScience
???displayArticle.grants??? [+]

Species referenced: Xenopus tropicalis Xenopus laevis
Genes referenced: nodal rho sms smyd1
GO keywords: embryo development

???attribute.lit??? ???displayArticles.show???

	Graphical abstract
	Figure 1. Absolute proteomics of the Ciona egg (A) Schematic of label-free proteomics utilized to determine absolute protein concentrations. Unfertilized Ciona eggs were lysed, and human proteins of known concentrations (UPS2) were added to the lysate as a reference standard. Following normalization as outlined in the materials and methods, we detect ∼195,000 peptides and estimate protein concentrations for ∼6,000 proteins. (B) Table of selected proteins in the unfertilized egg including the top 5 most abundant and some transcription factors important to embryonic development. (C) Histogram of all quantified proteins in the Ciona egg (gray) with superimposed kernel density estimates (KDE) of transcription factors (TFs - red) and signaling molecules (SMs - blue). Both TFs and SMs follow a distribution similar to the global egg proteome (black) but with a lower median concentration. The complete data is provided in Table S1. (D) Stoichiometries of protein complexes. Concentrations of subunits from a shared protein complex display comparable values and show typically a statistically different distribution than the entire proteome (∗p < 0.01, two-way ANOVA with Tukey’s multiple-comparisons test).
	Figure 2. Proteome and RNA analyses during Ciona embryogenesis (A) Overview of the transcriptome and proteome time-course experiments. Staged embryos were collected at eight developmental stages, beginning with unfertilized egg (unfE), fertilized egg (fertE), 16-cell stage (cell-16), initial gastrula (iniG), late neurula (latN), middle tailbud II (midTII), late tailbud II (latTII), and hatching tadpole (larva). Each stage is represented by a unique color code, and abbreviation; both are kept consistent throughout the figures. Time indicates hours postfertilization (hpf). (B) Number and overlap of identified protein-coding genes in the transcriptome and proteome datasets. (C) Donut plot with the percentage of protein evidence categories from UniProt that are identified at the proteome level (9,419 entries). Evidence level: (1) protein evidence; (2) transcript evidence; (3) homology; (4) predicted. (D) Histogram of Pearson correlations between RNA and corresponding protein dynamics throughout Ciona development (gray). The lines represent kernel density estimates (KDE) for all genes (black), transcription factors (red), and signaling molecules (blue). Notably, mRNA dynamics correlate poorly with protein dynamics. n = 7021 pairs. (E) Example of high Pearson correlation between RNA and protein dynamics for the transcription factor Hox10. (F) K-means clustering used to classify RNA (left) and protein (right) dynamics for each gene during Ciona development. The thickness of the lines scales with the number represented in each cluster, as indicated in the legend. (G) GO term analysis used to discern the functional relevance of each of the clusters (indicated by matching colors) identified in F.
	Figure 3. Comparison of development between chordate and vertebrate (A) Experimental design of the inter-species comparative developmental transcriptome and proteome time courses. Full circles highlight stages of development sampled for RNA-seq and proteomics. Mya, million years ago. (B) K-means co-clustering of the dynamics of orthologs (3,325) between Ciona and Xenopus development. The thickness of the line scales with the number of proteins represented in each cluster. The number of proteins in each cluster are quantified in the legend. Xenopus proteome time series from Sonnett et al.106 (C) GO term analysis identifying the functional significance of each of the clusters from B. The color of the clusters in B is kept consistent. (D) The log2 fold change (FC) protein correlation between Ciona and Xenopus TFs. Here, FC is defined as the ratio of relative protein abundance in the larva stage compared to the egg. Most TFs show similar behavior with the notable exception of Ybx. (E) Relative protein dynamics of TFs Ybx, Smyd1, Tfap2-r.b, Arid3, and E2f4/5. Each exhibit large fold changes in both organisms. Colors are preserved in these five proteins from the plotting in D. These TFs are canonically important for organism development by regulating transcriptional activation during the cell cycle, early muscle development, ectoderm development, gene activation through chromatin remodeling, and Nodal signaling respectively. Ybx exhibits signs of being maternally deposited in Ciona, but not in Xenopus, suggesting functional evolutionary divergence of this ortholog from chordate to vertebrate. Xenopus illustrations © Natalya Zahn (2022).
	Figure 4. The protein anti-hourglass model (A) Similarity heatmaps showing Pearson similarity between the two species for each investigated time point. Developmental stages are color-coded as defined in Figure 3A. The black line follows the highest correlation of the Xenopus time-point for each Ciona stage (n = 3,350, Xenopus transcriptome from Hu et al.,53 and Session et al.,115. Xenopus proteome from Sonnett et al.106). (B) Temporal divergence of gene (blue) and protein (red) expression from Xenopus embryogenesis to each Ciona stage. Maximal similarity is represented by the smallest distance from the center line, revealing a nested hourglass model in which the proteome exhibits more evident bottlenecks at early and later stages. Gray boxes outline these periods of minimal divergence. Regardless of stage, proteins show higher similarity between the two species' developmental mapping than RNA-seq, suggesting that protein dynamics are evolutionarily more conserved than mRNA dynamics (n = 3,350, Xenopus transcriptome from Hu et al.,53 and Session et al.,115. Xenopus proteome from Sonnett et al.106).
	Figure S1 LF-MS quality control and and genome-free protein reference database, related to Figure 1 A, Coomassie gel of Ciona egg lysate. A strong band corresponding to the 100 kDa apolipoprotein-B-like yolk protein (Vitellogenin) is visible 80. B, Bioinformatics pipeline for MS reference dataset construction. 1,222,451,669 RNA-seq reads were aggregated from five different studies and translated into a protein reference database as described previously 115 to obtain 163,907 95% non-redundant proteins. C, The genome-free reference database, and several Ciona proteomes, were used to analyze the same TMTproC MS dataset. The bar heights correspond to the total number of peptides identified in the MS dataset for each annotation. The red line represents the size of the proteome database. Our reference database outperforms the KY21 proteome slightly in terms of peptide identification, but at the expense of a much larger size and manual annotation of proteins. D, Comparison of peptides identification using genome-free reference database and KY21 proteome reference. The Venn diagram illustrates the shared and unique peptides identified using each reference. E, Examples where genome-free reference can help improve current KY21 assembly from SNPs detection (blue insert), correction of mis-annotated coding sequences, and accurate annotation of selenoproteins. F, Density plot of the distribution of the amino-acid based sequence coverage of the 6,219 protein detected before collapsing isoforms. The mean sequence coverage is indicated by the black dotted line.
	Figure S2 Characterization of the identified Ciona proteome, related to Figure 2 A, Linear discriminant analysis (LDA) is used to discriminate target and decoy peptides. Example of MS fraction with separation between the decoy (red) and target groups (green). 0.5 % peptide-level false discovery rate (FDR) is then estimated to discriminate between correct (dark green) and incorrect (light green) peptide−spectrum matches (PSMs). Rug plots at the bottom showing all sampling data. B,Histogram of peptide frequencies identified per protein in the dataset, with the protein count annotated on top of each bar. Dashed line indicates median values for the dataset. Insert: number of times a peptide occurs in the top proteins. C, Temporal expression of 35 protein isoforms represented by 2 to 4 unique splice variants.
	Figure S3 Quality control and sample correlation of stage-specific RNA-seq samples, related to Figure 2 A, Pairwise scatterplots of gene expression levels in transcripts per million (TPM) between biological replicates R: Pearson correlation coefficient. Stages are colouredcoded in all plots, as shown in Figure 2A. B, Principal component analysis (PCA) of the developmental RNA-seq time course showing the first two principal components (PCs), which together explain ~85% of the variance in the data. Samples from the same stage cluster together and there is a smooth progression through developmental time. C, Correlation matrices between RNA-seq experiments, calculated using Spearman (ρ). High agreement between biological replicates is observed. D, Boxplot of the distributions of numbers of genes detected at each stage (≥ 2TPM). E, Ridgeline plots of the distribution of genes by gene expression levels used to experimentally define a cut-off value of TPM ≥ 2 to deem a gene expressed.
	Figure S4 Protein expression dynamics during embryogenesis, related to Figure 2 A, Summary of the temporal pattern of protein expression during embryogenesis. Individual proteins (n = 7,095) are depicted in gray, while the data representing the median for each cluster is superimposed and color-coded. B, Tissue-level marker protein levels. Metaplots display the temporal dynamics for selected proteins. ‘SV' denotes sensory vesicle. Capital letters represent human orthologs, while lowercase letters indicate Ciona gene names. C, Percentage of all annotated transcription factors (TF), signaling molecules (SM), phosphatases (Pho), and kinases (Kin) detected at the protein level. Numbers near the y-axis denote the number of genes for these protein classes. D, Principal Component Analysis (PCA) of the expression levels of proteins in the embryo time series. Points on the graph represent individual proteins (7,095), with color coding indicative of different protein classes as in C. A 'salt and pepper' pattern for TFs and SMs is observable. E, Stage-specific protein differences at fertilization (from unfE to ferE), maternal-zygotic transition (from fertE to cell16), and in preparation for swimming tadpoles (from latTII to larva). Stages are color-coded as in Figure 2A.
	Figure S5 Proteogenomics and dynamic range of transcript and protein expression, related to Figure 2 A, Genes abundance spans a broader range of orders of magnitude compared to protein abundance. In both cases, approximately 90 % of the transcriptome or proteome is concentrated within about three orders of magnitude around the median value. B, Dynamic range of gene abundance with the proportion of transcripts detected exclusively by RNA-seq (grey) and those also identified at the protein level (yellow). Proteins associated with lower abundance genes are less detected. C, Cumulative abundance plots of transcriptome (represented by circles) and proteome (represented by squares) ranked by abundance (x-axis), with their respective contributions to the total transcriptome and proteome (y-axis), in the unfertilized egg. The seven most abundant genes and proteins are listed in descending order, these are not the same. Note the protein line rises more quickly than the gene line and the more uniform distribution of transcription factors (TF), signaling molecules, and transcription regulators (kinases, phosphatases, and zinc finger (ZF) genes), within the transcriptome, while these elements are more concentrated at higher ranks in the proteome.
	Figure S6 Different RNA and protein dynamics during development for TFs with known involvement in early Ciona development, related to Figure 2 A, i-iii Dlx.b, Tbx2/3, and Foxp have similar RNA and protein dynamics, largely matching in relative expression across development. B, i-iii Elk, Rar, and Tfap2-r.b have similar trends in their respective RNA and protein dynamics, with RNA being expressed earlier than protein and degrading while protein expression remains high. C, Foxh.b, Hnf4, and Sp8/9 do not have strong trends in their RNA and protein dynamics. RNA and protein expression seem more sporadic, with RNA coming in distinct waves that are not necessarily followed by protein.
	Figure S7 Extended transcriptome analysis, related to Figure 4 A, Comparison of the orthologues temporal associations between Ciona and Xenopus. These shared genes are more active in the earlier stages. B, Heatmaps of comparisons of all single copy one-to-one orthologs in an extended transcriptome using Spearman correlation (n=7,636; based on Xenopus and Ciona data from Hu et al., 2017) C, Heatmaps of comparisons of all single copy one-to-one orthologs using Cosine similarity (n=7,636; Xenopus data from Session et al. 2016 and Ciona data from this paper).
	Figure S8 Extended proteome analysis, related to Figure 4 A,B,C, Heatmaps representing Pearson (A, B) and Spearman (C) correlations from pairwise comparisons of one-to-one orthologs data between Ciona and Xenopus. Irrespective of the reference frog protein dataset used for comparison against the Ciona proteome, a consistent pattern emerges, showing the highest similarity at early and late stages of embryogenesis (compare A to B). The two frog independent time series align at three specific timepoints: st.1, st.12, and st.30. The time series from Sonnet et al., 2018 provides more stages in early development, while that from Itallie et al., 2021, covers more stages in late development, specifically st.41 and st.48. Regardless of the metric used (Pearson or Spearman) to quantify proteome similarity across species, consistent results are observed (compare B with C) (n protein = 3,350 from Sonnet et al., 2018; n protein = 5,376 from Itallie et al., 2021).
	Figure S9 Chordates share minimal proteome similarity during mid-developmental stages, related to Figure 4 A, Spearman's rho proteome similarity, normalized using the minimum and maximum values from all one-to-one ortholog sets, from Xenopus to each Ciona stage. Regardless of the frog proteome time series (top and bottom), a consistent pattern emerges in this proteome time series, showing minimal similarity at the stage of neurulation (in between vertical grey lines). B, Normalized Spearman similarity from Ciona to two independent Xenopus time series (top and bottom), revealing maximal similarity at the onset of embryogenesis and during the tadpole stages. Red points indicate identical frog timepoints in the two time series. Vertical grey lines highlight the developmental window with the highest divergence.

References [+] :

Abdul-Wajid, Exploiting the extraordinary genetic polymorphism of ciona for developmental genetics with whole genome sequencing. 2014, Pubmed