Protein prediction models support widespread post-transcriptional regulation of protein abundance by interacting partners

Solid stress compression enhances breast cancer cell migration through the upregulation of Interleukin-6

Apart from biochemical signals, tumour cells respond to biophysical and mechanical cues from their environment. The mechanical forces from the tumour microenvironment could be in the form of shear stress, tension, or solid stress compression. In this study, we explore the effects of solid stress compression on tumour cells. Solid stress compression, a prevalent biomechanical stimulus accumulated during tumour growth, has been shown to enhance invasive and metastatic phenotypes in cancer cells. However, the underlying molecular mechanism that elicits this aggressive metastatic phenotype, especially in breast cancer, is not extensively studied. Using an established 2D in vitro setup to apply incremental solid stress compression, we found that migratory and invasive capacities of aggressive breast cancer cells were enhanced in a biphasic manner. We also found that the transcript and protein levels of Interleukin-6 (IL-6) and SNAI1 were upregulated in response to solid stress. The resultant increased secretion of IL-6 could in turn lead to autocrine activation of downstream signalling pathways and impact on cancer cell migration and invasion.

Multi-omics analysis reveals discordant proteome and transcriptome responses in larval guts of Frankliniella occidentalis infected with an orthotospovirus

The western flower thrips, Frankliniella occidentalis, is the principal thrips vector of Orthotospovirus tomatomaculae (order Bunyavirales, family Tospoviridae), a devastating plant-pathogenic virus commonly referred to as tomato spotted wilt virus (TSWV). The larval gut is the gateway for virus transmission by F. occidentalis adults to plants. In a previous report, gut expression at the transcriptome level was subtle but significant in response to TSWV in L1s. Since it has been well documented that the relationship between the expression of mRNA and associated protein products in eukaryotic cells is often discordant, we performed identical, replicated experiments to identify and quantify virus-responsive larval gut proteins to expand our understanding of insect host response to TSWV. While we documented statistically significant, positive correlations between the abundance of proteins (4189 identified) and their cognate mRNAs expressed in first and second instar guts, there was virtually no alignment of individual genes identified to be differentially modulated by virus infection at the transcriptome and proteome levels. Predicted protein-protein interaction networks associated with clusters of co-expressed proteins revealed wide variation in correlation strength between protein and cognate transcript abundance, which appeared to be associated with the type of cellular processes, cellular compartments and network connectivity represented by the proteins. In total, our findings indicate distinct and dynamic regulatory mechanisms of transcript and protein abundance (expression, modifications and/or turnover) in virus-infected gut tissues. This study provides molecular candidates for future functional analysis of thrips vector competence and underscores the necessity of examining complex virus-vector interactions at a systems level.

Inferring protein from transcript abundances using convolutional neural networks

BACKGROUND

Although transcript abundance is often used as a proxy for protein abundance, it is an unreliable predictor. As proteins execute biological functions and their expression levels influence phenotypic outcomes, we developed a convolutional neural network (CNN) to predict protein abundances from mRNA abundances, protein sequence, and mRNA sequence in Homo sapiens (H. sapiens) and the reference plant Arabidopsis thaliana (A. thaliana).

RESULTS

After hyperparameter optimization and initial data exploration, we implemented distinct training modules for value-based and sequence-based data. By analyzing the learned weights, we revealed common and organism-specific sequence features that influence protein-to-mRNA ratios (PTRs), including known and putative sequence motifs. Adding condition-specific protein interaction information identified genes correlated with many PTRs but did not improve predictions, likely due to insufficient data. The integrated model predicted protein abundance on unseen genes with a coefficient of determination (r2) of 0.30 in H. sapiens and 0.32 in A. thaliana.

CONCLUSIONS

For H. sapiens, our model improves prediction performance by nearly 50% compared to previous sequence-based approaches, and for A. thaliana it represents the first model of its kind. The model's learned motifs recapitulate known regulatory elements, supporting its utility in systems-level and hypothesis-driven research approaches related to protein regulation.

Single Cell Proteomics Reveals Novel Cell Phenotypes in Marfan Mouse Aneurysm

Background

Single-cell omics technology is a powerful tool in biomedical research. However, single cell proteomics has lagged due to an inability to amplify peptides in a similar fashion to nucleotide strings. Single cell proteomics is important because proteins are the main functional unit in cells, and they often poorly correlate with mRNA quantities. In this paper we describe the first single cell proteomic analysis of complex tissue, comparing aneurysmal and normal mouse aorta from males and females. We also compare and integrate our single cell proteomic profiles with a matching single cell transcriptomics dataset.

Methods

We compared single cell proteomes between male and female, wild-type and Fbn1 C1041G/+ Marfan mice (N=3 per group). Individual cells from mouse aortic root single cell suspensions were deposited in 384 well plates and subjected to ultra-sensitive nanoflow liquid chromatography-ion mobility-time of flight-mass spectrometry. The data were analyzed with leiden clustering to identify cell types. Statistical analyses were performed to detect differential proteins within cell types and multi-omics analysis integrated single cell proteomics with published single cell RNA-seq.

Results

We identified all major aortic cell types including 7 distinct smooth muscle cell subtypes. The proportion of these cells varied based on sex and the Fbn1 C1041G/+ genotype. Differentially expressed proteins between male and female in addition to wild-type and Marfan samples uncovered enhanced endothelial to mesenchymal transition patterns in endothelial cells from male Marfan mice. Comparisons between single cell RNA and single cell proteomic profiles showed similarities in major subtypes but not smooth muscle cell subtypes. Multi-omics analysis of these two single cell platforms demonstrated a potential novel role for smooth muscle cell derived angiotensin signaling in the Marfan phenotype.

Conclusions

Single cell proteomics identified new subpopulations of vascular smooth muscles cells and novel cell type specific protein signatures related to sex differences and aneurysm formation.

Abbreviations

Next generation sequencing (NGS), Mass spectrometer (MS), Single cell proteomics by Mass Spectrometry (ScOPE-MS), Marfan's syndrome (MFS), Fibrillin 1 (FBN1), Transforming growth factor β (TGFβ), Smooth muscle cell (SMC), Single cell proteomic (scProteomic), Differentially expressed proteins (DEPs), Wild-type (WT), Hanks' balanced salt solution (HBSS), Fetal bovine serum (FBS), Dulbecco's Modified Eagle Medium (DMEM), Data-independent acquisition parallel accumulation-serial fragmentation (DIA-PASEF), Magnetic assisted cell sorted (MACS), Single Cell Analysis in Python (Scanpy), Kyoto Encyclopedia of Genes and Genomes (KEGG), Principal component analysis (PCA), Uniform manifold projection (UMAP), Single cell transcriptomic (scTranscriptomic), Smoothelin (Smtn), Transgelin (Tagln), Myosin heavy chain 11 (Myh11), Platelet endothelial cell adhesion molecule 1 (Pecam1), Dipeptidase 1 (Dpep1), Uncoupling protein 1 (Ucp1), Low-density lipoprotein receptor-related protein (Lrp1), DNA ligase 3 (Lig3), Capsaicin channel transient receptor potential vanilloid 1 (Trpv1), Endothelial to mesenchymal transition (endMT), Intercellular adhesion molecule 1 (Icam1), Intercellular adhesion molecule 2 (Icam2), Endothelial cell-selective adhesion molecule (Esam), Calponin 1 (Cnn1), Vimentin (Vim), Zinc finger E-box-binding homeobox 1 (Zeb1), Snail family transcriptional repressor 1 (Snai1), Tropomyosin alpha-4 chain (Tpm4), Angiotensin converting enzyme (Ace).

Combined statistical-biophysical modeling links ion channel genes to physiology of cortical neuron types

Neural cell types have classically been characterized by their anatomy and electrophysiology. More recently, single-cell transcriptomics has enabled an increasingly fine genetically defined taxonomy of cortical cell types, but the link between the gene expression of individual cell types and their physiological and anatomical properties remains poorly understood. Here, we develop a hybrid modeling approach to bridge this gap. Our approach combines statistical and mechanistic models to predict cells' electrophysiological activity from their gene expression pattern. To this end, we fit biophysical Hodgkin-Huxley-based models for a wide variety of cortical cell types using simulation-based inference, while overcoming the challenge posed by the mismatch between the mathematical model and the data. Using multimodal Patch-seq data, we link the estimated model parameters to gene expression using an interpretable sparse linear regression model. Our approach recovers specific ion channel gene expressions as predictive of biophysical model parameters including ion channel densities, directly implicating their mechanistic role in determining neural firing.

Proteome-wide copy-number estimation from transcriptomics

Protein copy numbers constrain systems-level properties of regulatory networks, but proportional proteomic data remain scarce compared to RNA-seq. We related mRNA to protein statistically using best-available data from quantitative proteomics and transcriptomics for 4366 genes in 369 cell lines. The approach starts with a protein's median copy number and hierarchically appends mRNA-protein and mRNA-mRNA dependencies to define an optimal gene-specific model linking mRNAs to protein. For dozens of cell lines and primary samples, these protein inferences from mRNA outmatch stringent null models, a count-based protein-abundance repository, empirical mRNA-to-protein ratios, and a proteogenomic DREAM challenge winner. The optimal mRNA-to-protein relationships capture biological processes along with hundreds of known protein-protein complexes, suggesting mechanistic relationships. We use the method to identify a viral-receptor abundance threshold for coxsackievirus B3 susceptibility from 1489 systems-biology infection models parameterized by protein inference. When applied to 796 RNA-seq profiles of breast cancer, inferred copy-number estimates collectively re-classify 26-29% of luminal tumors. By adopting a gene-centered perspective of mRNA-protein covariation across different biological contexts, we achieve accuracies comparable to the technical reproducibility of contemporary proteomics.

An Extensive Atlas of Proteome and Phosphoproteome Turnover Across Mouse Tissues and Brain Regions

Understanding how proteins in different mammalian tissues are regulated is central to biology. Protein abundance, turnover, and post-translational modifications like phosphorylation, are key factors that determine tissue-specific proteome properties. However, these properties are challenging to study across tissues and remain poorly understood. Here, we present Turnover-PPT, a comprehensive resource mapping the abundance and lifetime of 11,000 proteins and 40,000 phosphosites across eight mouse tissues and various brain regions, using advanced proteomics and stable isotope labeling. We revealed tissue-specific short- and long-lived proteins, strong correlations between interacting protein lifetimes, and distinct impacts of phosphorylation on protein turnover. Notably, we discovered that peroxisomes are regulated by protein turnover across tissues, and that phosphorylation regulates the stability of neurodegeneration-related proteins, such as Tau and α-synuclein. Thus, Turnover-PPT provides new fundamental insights into protein stability, tissue dynamic proteotypes, and the role of protein phosphorylation, and is accessible via an interactive web-based portal at https://yslproteomics.shinyapps.io/tissuePPT.

Protein turnover regulation is critical for influenza A virus infection

The abundance of a protein is defined by its continuous synthesis and degradation, a process known as protein turnover. Here, we systematically profiled the turnover of proteins in influenza A virus (IAV)-infected cells using a pulse-chase stable isotope labeling by amino acids in cell culture (SILAC)-based approach combined with downstream statistical modeling. We identified 1,798 virus-affected proteins with turnover changes (tVAPs) out of 7,739 detected proteins (data available at pulsechase.innatelab.org). In particular, the affected proteins were involved in RNA transcription, splicing and nuclear transport, protein translation and stability, and energy metabolism. Many tVAPs appeared to be known IAV-interacting proteins that regulate virus propagation, such as KPNA6, PPP6C, and POLR2A. Notably, our analysis identified additional IAV host and restriction factors, such as the splicing factor GPKOW, that exhibit significant turnover rate changes while their total abundance is minimally affected. Overall, we show that protein turnover is a critical factor both for virus replication and antiviral defense.

Interferons dominate damage and activity in juvenile scleroderma

OBJECTIVES

Juvenile scleroderma is a heterogeneous group of diseases associated with sclerotic skin lesions, grouped as juvenile systemic sclerosis and juvenile localized scleroderma. This study aims to measure the cytokine and chemokine levels involved in interferon (IFN) signalling in patients with juvenile scleroderma and determine their correlation with disease severity.

METHODS

Twenty-nine juvenile localized scleroderma, five juvenile systemic sclerosis, and nine healthy controls were included in the study. Cytokines and chemokines involved in IFN gene signalling (IL-1, IL-6, IL-8, IP-10, MCP1, TNF-α, CXCL-11, IFN-α, IFN-β, IFN-γ) and IFN-stimulated genes (ISGs), including IFI27, IFI44, ISIG15, IFIT1, OAS1, RSAD2, were measured by ELISA and RT-PCR method, respectively.

RESULTS

A significant increase in IFN-α, IFN-β, IFN-γ, TNF-α, IL-1, IL-6 IL-8, IP-10, and MCP1 levels was observed in patients with juvenile systemic sclerosis compared with the healthy control group. Furthermore, IFN-α and IP-10 were elevated in both juvenile localized scleroderma and juvenile systemic sclerosis compared to the healthy control group. IFN-γ and IFN-α positively correlated with LoSAI and LoSDI levels, respectively. According to PGA-A analysis, IFN-β, IFN-γ, TNF-α, IL-8, IP10, MCP1, and CXCL11 were significantly higher in active disease than in the inactive state in both groups.

CONCLUSION

The results suggest that IFN signalling may be impaired in patients with juvenile scleroderma. Significant changes were observed in cytokines and genes related to IFN signalling, which may have a crucial role in monitoring disease activity. In addition, we have gained important insights into the possibility of using IFN-α and IFN-γ as biomarkers for monitoring juvenile scleroderma activity and damage.

Proteome-wide association study using cis and trans variants and applied to blood cell and lipid-related traits in the Women's Health Initiative study

In most Proteome-Wide Association Studies (PWAS), variants near the protein-coding gene (±1 Mb), also known as cis single nucleotide polymorphisms (SNPs), are used to predict protein levels, which are then tested for association with phenotypes. However, proteins can be regulated through variants outside of the cis region. An intermediate GWAS step to identify protein quantitative trait loci (pQTL) allows for the inclusion of trans SNPs outside the cis region in protein-level prediction models. Here, we assess the prediction of 540 proteins in 1002 individuals from the Women's Health Initiative (WHI), split equally into a GWAS set, an elastic net training set, and a testing set. We compared the testing r2 between measured and predicted protein levels using this proposed approach, to the testing r2 using only cis SNPs. The two methods usually resulted in similar testing r2, but some proteins showed a significant increase in testing r2 with our method. For example, for cartilage acidic protein 1, the testing r2 increased from 0.101 to 0.351. We also demonstrate reproducible findings for predicted protein association with lipid and blood cell traits in WHI participants without proteomics data and in UK Biobank utilizing our PWAS weights.

Multi-dataset Integration and Residual Connections Improve Proteome Prediction from Transcriptomes using Deep Learning

Proteomes are well known to poorly correlate with transcriptomes measured from the same sample. While connected, the complex processes that impact the relationships between transcript and protein quantities remains an open research topic. Many studies have attempted to predict proteomes from transcriptomes with limited success. Here we use publicly available data from the Clinical Proteomics Tumor Analysis Consortium to show that deep learning models designed by neural architecture search (NAS) achieve improved prediction accuracy of proteome quantities from transcriptomics. We find that this benefit is largely due to including a residual connection in the architecture that allows input information to be remembered near the end of the network. Finally, we explore which groups of transcripts are functionally important for protein prediction using model interpretation with SHAP.

Gene expression response under thermal stress in two Hawaiian corals is dominated by ploidy and genotype

Transcriptome data are frequently used to investigate coral bleaching; however, the factors controlling gene expression in natural populations of these species are poorly understood. We studied two corals, Montipora capitata and Pocillopora acuta, that inhabit the sheltered Kāne'ohe Bay, Hawai'i. M. capitata colonies in the bay are outbreeding diploids, whereas P. acuta is a mixture of clonal diploids and triploids. Populations were sampled from six reefs and subjected to either control (no stress), thermal stress, pH stress, or combined pH and thermal stress treatments. RNA-seq data were generated to test two competing hypotheses: (1) gene expression is largely independent of genotype, reflecting a shared treatment-driven response (TDE) or, (2) genotype dominates gene expression, regardless of treatment (GDE). Our results strongly support the GDE model, even under severe stress. We suggest that post-transcriptional processes (e.g., control of translation, protein turnover) modify the signal from the transcriptome, and may underlie the observed differences in coral bleaching sensitivity via the downstream proteome and metabolome.

A Ratiometric Catalog of Protein Isoform Shifts in the Cardiac Fetal Gene Program

The fetal genetic program orchestrates cardiac development and the re-expression of fetal genes is thought to underlie cardiac disease and adaptation. Here, a proteomics ratio test using mass spectrometry is applied to find protein isoforms with statistically significant usage differences in the fetal vs. postnatal mouse heart. Changes in isoform usage ratios are pervasive at the protein level, with 104 significant events observed, including 88 paralog-derived isoform switching events and 16 splicing-derived isoform switching events between fetal and postnatal hearts. The ratiometric proteomic comparisons rediscovered hallmark fetal gene signatures including a postnatal switch from fetal β (MYH7) toward ɑ (MYH6) myosin heavy chains and from slow skeletal muscle (TNNI1) toward cardiac (TNNI3) troponin I. Altered usages in metabolic proteins are prominent, including a platelet to muscle phosphofructokinase (PFKP - PFKM), enolase 1 to 3 (ENO1 - ENO3), and alternative splicing of pyruvate kinase M2 toward M1 (PKM2 - PKM1) isoforms in glycolysis. The data also revealed a parallel change in mitochondrial proteins in cardiac development, suggesting the shift toward aerobic respiration involves also a remodeling of the mitochondrial protein isoform proportion. Finally, a number of glycolytic protein isoforms revert toward their fetal forms in adult hearts under pathological cardiac hypertrophy, suggesting their functional roles in adaptive or maladaptive response, but this reversal is partial. In summary, this work presents a catalog of ratiometric protein markers of the fetal genetic program of the mouse heart, including previously unreported splice isoform markers.

Proteomics applications in next generation induced pluripotent stem cell models

INTRODUCTION

Induced pluripotent stem (iPS) cell technology has transformed biomedical research. New opportunities now exist to create new organoids, microtissues, and body-on-a-chip systems for basic biology investigations and clinical translations.

AREAS COVERED

We discuss the utility of proteomics for attaining an unbiased view into protein expression changes during iPS cell differentiation, cell maturation, and tissue generation. The ability to discover cell-type specific protein markers during the differentiation and maturation of iPS-derived cells has led to new strategies to improve cell production yield and fidelity. In parallel, proteomic characterization of iPS-derived organoids is helping to realize the goal of bridging in vitro and in vivo systems.

EXPERT OPINIONS

We discuss some current challenges of proteomics in iPS cell research and future directions, including the integration of proteomic and transcriptomic data for systems-level analysis.

Proteome-wide copy-number estimation from transcriptomics

Protein copy numbers constrain systems-level properties of regulatory networks, but absolute proteomic data remain scarce compared to transcriptomics obtained by RNA sequencing. We addressed this persistent gap by relating mRNA to protein statistically using best-available data from quantitative proteomics-transcriptomics for 4366 genes in 369 cell lines. The approach starts with a central estimate of protein copy number and hierarchically appends mRNA-protein and mRNA-mRNA dependencies to define an optimal gene-specific model that links mRNAs to protein. For dozens of independent cell lines and primary prostate samples, these protein inferences from mRNA outmatch stringent null models, a count-based protein-abundance repository, and empirical protein-to-mRNA ratios. The optimal mRNA-to-protein relationships capture biological processes along with hundreds of known protein-protein interaction complexes, suggesting mechanistic relationships are embedded. We use the method to estimate viral-receptor abundances of CD55-CXADR from human heart transcriptomes and build 1489 systems-biology models of coxsackievirus B3 infection susceptibility. When applied to 796 RNA sequencing profiles of breast cancer from The Cancer Genome Atlas, inferred copy-number estimates collectively reclassify 26% of Luminal A and 29% of Luminal B tumors. Protein-based reassignments strongly involve a pharmacologic target for luminal breast cancer (CDK4) and an α-catenin that is often undetectable at the mRNA level (CTTNA2). Thus, by adopting a gene-centered perspective of mRNA-protein covariation across different biological contexts, we achieve accuracies comparable to the technical reproducibility limits of contemporary proteomics. The collection of gene-specific models is assembled as a web tool for users seeking mRNA-guided predictions of absolute protein abundance (http://janeslab.shinyapps.io/Pinferna).