Agenda

Cellular interactions within the tissue micro-environment form the basis of health and disease for all organisms. Exposure to nutrients, toxins, and neighboring cells trigger coordinated molecular responses that impact cellular function and metabolism in a beneficial, adaptive, or detrimental manner. Acquiring molecular information at cellular resolution is thus crucial for developing a comprehensive understanding of the biology in an organism. Matrix-assisted laser desorption/ionization (MALD) imaging mass spectrometry (IMS) addresses this need by combining the spatial fidelity of classical microscopy with the molecular specificity of a mass spectrometer. This presentation will highlight our work developing new, high-performance technologies for improving the spatial resolution, sensitivity, and specificity of MALDI IMS for metabolite, lipid, and protein mapping. This will include the utilization of high mass resolution FT-ICR MS and high spatial resolution Q-TOF platforms to address the molecular complexity associated with direct tissue analysis. Further, I will describe recent advances in integrated, multimodal methods that enable molecular signals to be correlated to specific biological tissue features. These technologies will be demonstrated through applications that include understanding the molecular drivers of host-pathogen interactions and the construction of comprehensive molecular tissue atlases.

As the Worlds population is getting older the prevalence of age associated diseases such as Alzheimer’s disease (AD) is steadily increasing accordingly, currently affecting 12% over the age of 65.  As the underlying mechanism remain unclear there is still no curative treatment. It is therefore critical to understand how key pathological factors of AD including beta-amyloid (Aβ) plaque formation are interconnected and implicated in nerve cell death, clinical symptoms and disease progression.

The main goal of our research is to elucidate the biochemical processes underlying early Aβ plaque formation in brain tissue.

We use advanced chemical imaging modalities including hyperspectral confocal microscopy, electron imaging and mass spectrometry imaging to delineate in vivo Aβ build up and deposition at cellular length scales. Together with stable isotope labelling, these tools allow us to visualize Aβ aggregation dynamics within single plaques across different brain regions and to follow the fate of aggregating Aβ species from before and throughout the earliest events of precipitating plaque pathology.

We further integrate these experiments with functional amyloid microscopy using structure sensitive fluorescent probes that allow delineate the chain of events underlying amyloid aggregation, deposition and maturation. 

These data, provide a detailed picture of the earliest events of precipitating amyloid pathology at scales not previously possible. The results from these studies bring considerable novel information about the deposition mechanism of Aβ and its toxic interactions with the surrounding. This will open up for developing tailored strategies to affect AD pathology prior to any neurodegenerative mechanisms as well as to develop new biomarkers for AD.

Over the past decades, large-scale proteomics studies have allowed us to begin to better understand human phenotype at a protein level. However, most proteomics measurements only identify peptides and do not precisely determine alternatively spliced sequences or posttranslational modifications of the proteins. The Blood Proteoform Atlas (BPA) presents the primary structures of ~30,000 unique proteoforms, nearly ten times more than in previous studies, expressed from 1,690 human genes across 21 cell types and plasma from human blood and bone marrow. The results indicate that proteoforms better describe protein-level cell biology and are more specific indicators of differentiation than their corresponding proteins (genes), which are more broadly expressed across cell types. Finally, we demonstrate the potential for clinical application by interrogating the BPA in the context of liver transplantation and identifying cell and proteoform signatures that distinguish normal graft function from acute rejection and other causes of graft dysfunction.

Top-down mass spectrometry (MS)-based proteomics is arguably the most powerful technology to comprehensively characterize proteoforms that arise from genetic variations, alternative splicing, and post-translational modifications (PTMs), but myriad challenges still remain. We have been developing novel technologies to address the challenges in top-down proteomics in a multi-pronged approach including new cleavable surfactants for protein solubilization, new strategies for multi-dimensional chromatography separation of proteins, novel nanomaterials for enrichment of low-abundance proteins. Additionally, we have been developing a new comprehensive user-friendly software package for native top-down proteomics. Moreover, we have made major advances in top-down proteomics for analysis of intact proteins directly purified from heart tissue, blood, and human pluripotent stem cell-derived cardiomyocytes (hPSC). Importantly, we have linked altered cardiac proteoforms to contractile dysfunction in heart diseases using animal models and human clinical samples. Furthermore, we are harnessing the power of innovative top-down proteomics with patient specific hPSC-derived cardiomyocytes (CMs) in engineered cardiac tissue to understand proteoform biology in cardiac diseases.

Lysosomes and mitochondria are membrane-bound organelles in the cell, responsible for trash-disposal and ATP production, respectively. Perturbations in lysosomes and mitochondria have been linked to numerous human diseases, such as cancer, neurodegeneration, and cardiovascular diseases. These organelles are highly dynamic and frequently interact with other cellular components in a transient fashion, which are difficult to capture and visualize in a high-throughput fashion. Here, I will describe our recent efforts in developing and improving proximity labeling, organelle isolation, and dynamic SILAC proteomic methods to characterize sub-organelle microenvironment and protein dynamics in live cells. Patient skin fibroblasts can be reprogrammed into induced pluripotent stem cells (iPSCs) and further differentiated into neurons, enabling the study of live human neurons that were previously inaccessible. Using multimodal proteomic strategies in human neurons, we aim to decipher the molecular mechanisms underlying lysosomal and mitochondrial dysfunctions in brain diseases. I will also introduce our newly established universal and sample type-specific contaminant libraries for both DDA and DIA proteomics that can benefit the broad proteomics community.

Since data-independent acquisition (DIA) was introduced in 2004, DIA acquisition and data analysis tools have been continuously improved, making DIA a crucial technology to identify and quantify thousands of proteins with high reproducibility and deep proteomics coverage. DIA data analysis, in general, relies on spectral libraries constructed from data-dependent acquisition (DDA). Alternatively, the library-free method searches DIA data directly against a protein database. We have developed CCS-enabled TIMS DIA-NN software to support both spectral library and library-free strategies. The tool was integrated into PaSER (Parallel Search Engine in Real-time) to analyze DIA data in real time. Moreover, we implemented and incorporated multiple deep-learning models in the software to improve identification, quantification, and performance.

While mass spectrometry (MS) imaging allows unparalleled insight into the spatial distribution of molecules in a sample, its data pose challenges for traditional machine learning methods. Spatial statistics and computer vision methods are not suited for such high-dimensional imaging datasets, and machine learning methods must be adapted to deal with MS imaging’s imperfect class labels, noisy data, and relatively small sample sizes. Furthermore, instrumentation improvements continue to produce larger datasets and greater file sizes, compounding the computational challenges.

It is more important than ever that MS imaging data is appropriately processed for downstream analysis, but the impact of preprocessing decisions on machine learning results is often unclear. Evaluating this impact is challenging due to imperfect pixel-level annotations for MS imaging. For classification, the goal is to correctly label subregions of a sample (for example, into healthy and disease classes). However, ground truth labels are rarely available on a pixel-by-pixel level, so machine learning methods that require complete labels can produce invalid results. We present our recent work using multiple instance learning with a convolutional neural network (miCNN) in the presence of uncertain class labels. We further examine the impact that different preprocessing workflows can have on the accuracy of the results.

Many of the methods we present are implemented in the open-source R package Cardinal which provides a full software workflow for MS imaging, including data import, pre-processing, visualization, and statistical and machine learning. We will describe how recent updates to Cardinal help provide accessible and scalable computing infrastructure for processing and analyzing MS imaging experiments.

Three-dimensional cell cultures are attractive models for biological research. They combine the flexibility of cell culture with some of the spatial and molecular complexity of tissue. For example, colon cancer cell lines form spheroids, in vitro mimics of poorly vascularized tumors. The spheroids are composed of a central necrotic core, a middle quiescent layer and an outer proliferative layer of cells, similar to a rapidly growing colon tumor. Our laboratory has characterized the distribution of endogenous proteins via MALDI imaging mass spectrometry in colon spheroids and determined that the molecular gradients correlate with the pathophysiological changes in the structure. We have developed Spatial SILAC, a method to selectively label the cells in the distinct chemical microenvironments in the spheroids and assess the proteomic changes in response to drug treatment. In this presentation, the spatially-localized proteomic changes in response to the kinase inhibitor Regorafenib will be described.

Protein post-translational modifications allow cells to perform physiological functions greater than the coding capacity of human genome. Dysregulation of PTMs is shown to relate to pathological dysfunctions. Studying multiple PTMs is undoubtedly fundamental in understanding protein functions and multi-PTM crosstalk, which is gaining traction in functional proteomic studies. Despite the importance of multi-PTM proteomics, limited studies investigate more than one PTM at a time. In this study, co-enrichments, multiplexing, fractionation, hybrid data acquisition (DDA + DIA), as well as combined data analysis were investigated for the application on tissue samples, focusing on phosphorylation, glycosylation, acetylation, and ubiquitination.

Looking for disease biomarkers in omics datasets is often compared to looking for a needle in a haystack; the number of potential markers that can now be quantified is immense.  The challenge for data scientists is to find the needles (true biomarkers), when they are present and to not be misled by potentially promising candidates that turn out to be no more useful than shiny pieces of hay.  The presentation will address this challenge, known as feature selection, from a variety of perspectives.  As a cautionary tale, we will show that an improperly executed feature selection strategy leads to identifying disease biomarkers that appear to offer “>90% accuracy”, but are, in fact, no better than flipping a coin for predicting disease.  Unfortunately, the improperly executed strategy is a pervasive mistake that currently occurs when researchers combine feature selection methods with proteomics datasets; we will demonstrate how the problem can be avoided.  A second vignette will illustrate the unequivocal importance of using racially diverse sample sets in biomarker studies.  We will show that biomarker candidates for Alzheimer’s Disease, which may be optimally beneficial for a broad, racially diverse population, can be overlooked if the underlying datasets used to identify them are not racially diverse.  Identifying useful disease biomarkers is a formidable challenge, but real successes are possible. Both the samples and the features are important, and carefully attending to the nuances gives the magic to the method.

The last years have shown how precise measurements of protein turnover – the interplay between protein production and loss - yield important insight about gene expression regulation. Moreover, several different approaches have been introduced to provide precise turnover measurements for thousands of proteins. These include dynamic SILAC measurements which were either followed by deep fractionation and data dependent acquisition (DDA) proteomics measurements, also sometimes combined with TMT labeling, or by data independent acquisition (DIA) measurements of unfractionated samples. We are systematically testing these different approaches in two mammalian in vivo model systems. First, we are determining protein turnover changes in mouse organoids derived from pancreatic tumor and metastases. Second, we are measuring protein turnover in brain tissue from wildtype mice and mice where an important regulator of protein homeostasis is knocked out. In my talk I will share insight about our ongoing work to optimize the yield of reliable protein turnover measurements in both these systems and the advantages and challenges associated with all the tested approaches to measure protein turnover.

By largely unknown mechanisms, COVID-19 patients with pre-existing chronic inflammatory diseases are vulnerable to severe symptoms. Excessive serum cytokines and abnormally activated macrophages are found in the bronchoalveolar lavage fluid of these severe COVID-19 patients. Thus, effective therapies to minimize mortality depends on a mechanistic understanding of SARS-CoV-2-induced pathogenesis. We report a novel gene-specific translation mechanism that suppresses synthesis of proteins associated with SARS-CoV-2 transmission, impaired T cell function, blood clotting, and the host hyperinflammatory response. First, we employed our chromatin-activity based chemoproteomics (ChaC) approach to dissect the interactome/pathways associated with G9a, a histone methyltransferase whose mRNA was upregulated with virus load in COVID-19 patient peripheral mononuclear cells. This ChaC analysis revealed that constitutively (enzymatically) active G9a interacted with multiple regulators of translation and ribosome biogenesis. This finding implicated a noncanonical function of G9a in the translational regulation of COVID-19 immunopathogenesis. Accordingly, using our translatome proteomics approach, we identified and profiled proteins whose translation depended on G9a activity, that is, G9a upregulated the widespread translation of a battery of COVID-19 pathogenesis-related genes. Mechanistically, based on ChaC identification of G9a interaction with METTL3, an N6-methyladenosine (m6A) RNA methyltransferase, we found that G9a methylates the nonhistone protein METTL3 to co-upregulate m6A-mediated translation (synthesis) of specific proteins functionally associated with immune checkpoint and hyperinflammation. Therefore, we conducted a correlated multiomics study on the m6A/METTL3 transcriptome, proteome, and phosphoproteome of SARS-CoV-2 infected human alveolar epithelial cells that overexpressed ACE2 (A549-hACE2) with or without treatment with inhibitors of G9a. Results indicated that G9a inhibition reversed the multiomic landscapes established by SARS-CoV-2 infection and from which proteins that showed G9a-dependent translation unite the networks associated with viral replication, virus-host interactions, T cell activation/proliferation, and systemic cytokine response. Inhibition of G9a suppressed SARS-CoV-2 replication, which validated our discovery of G9a-regulated translational mechanism of COVID-19 pathogenesis. Further, our correlated multiomics analysis revealed a mechanism of G9a inhibition action that showed multifaceted virus- and host-directed therapeutic effects. Because translational regulation is a precise and energy-efficient step of controlling the expression of proteins, inhibiting the G9a-mediated translation regulatory mechanism can be highly specific and effective by immediately suppressing the aberrant synthesis of COVID 19-related proteins without the need for altering transcriptional activation and mRNA processing steps.

Many biological processes are regulated through dynamic protein phosphorylation. Monitoring disease-relevant phosphorylation events in circulating biofluids is highly appealing but also technically challenging. We introduce here a functionally tunable material and a strategy, extracellular vesicles to phosphoproteins (EVTOP), which achieves one-pot EV isolation, extraction and digestion of EV proteins, and enrichment of phosphopeptides starting with only trace amount of biofluids. The streamlined, ultra-sensitive platform enables us to quantify 500 unique EV phosphopeptides with only a few μL of plasma and over 1,200 phosphopeptides with 100 μL of cerebrospinal fluid (CSF). We demonstrated its clinical application through evaluating the outcome of chemotherapy of primary central nervous system lymphoma (PCNSL) patients with small volume of CSF, presenting a powerful tool for broad clinical applications.

An underlying premise of precision medicine is that increasing the number of clinical and molecular features will improve accuracy diagnosing disease and predicting clinical outcomes (e.g., a biomarker predictive value, PPV). In other words, more is better. We propose that developing a parsimonious multi-omic model, composed of the minimal number of features able to provide similar or equal predictive performance to models produced with larger and more complex analyte compositions is required for traction in the clinical situation, especially in health care systems with economic burden. Furthermore, we propose that the features underlying a parsimonious model will represent the minimal mechanisms driving the disease outcome even in diseases with high heterogeneity in clinical manifestation. As a first test of our proposal, we analyzed 74 patients with pancreatic ductal adenocarcinoma (PDAC) obtaining >6500 features from clinical, computational pathology, and molecular (DNA, tissue RNA, tissue and plasma protein, and plasma lipid). Multiple independent machine learning models were developed and tested on curated single- and multi-omic analyte panels to determine their ability to predict clinical outcomes in patients. Interesting, the best performing multi-Omic models were comprised of different feature types even if they had equivalent PPV and accuracy  for survival (0.85, 0.87, respectively), suggestive of diverse disease mechanisms achieving the same clinical outcome. The parsimonious model with 589 multi-Omic features had the same PPV while <50 features comprised of only plasma lipids and plasma protein had only slightly lower PPV. Thus, the parsimonious model is both cost effective and easily deployable in clinical practice, regardless of health care system, especially if two tier analysis is deployed with an initial plasma sample followed, if positive, by a more invasive biopsy. The adoption of remote sampling devices, where an individual can take their own blood sample and submit by mail to practitioners, could reduce barrier in medical deserts leading to broader adoption of clinical care. 

Brain extracellular space contains a wide range of molecules including neurotransmitters, neuromodulators, and metabolites. The chemical milieu in this space is indicative of cellular activity and is involved in regulation of that activity. Samples from this space can be accessed by microdialsyis probes. Most work to date using microdialysis has focussed on measuring a small number of neurotransmitters or metabolites at a time. In this work we apply LC-MS based metabolomics approaches to monitoring the brain metabolome. Our goal is to both identify the chemicals present and begin to understand their relationship to phenotype, behaviour, drug effects, or disease state. In undirected metabolomics compounds can be identified by matching the mass spectra to a database. Often only a small fraction of the signals detected can be attributed to specific compound. For example, it is not uncommon to detect 104 “features” (signal at a given retention time and mass) but only identify a few hundred compounds. We show how deeper analysis can be performed by using advanced separations and concentrated samples. Once identified, compounds can be tracked using faster seprations for good throughput. We have identified over 300 compounds present in the brain extracellular space so far with potential for many more. We also show how directed metabolomics methods can be used to uncover chemical differences of phenotypes, in this case the HR/LR behavioural model, and relate these to differences in behavior. Finally, we describe a LC-MS based method to using stable-isotope tracing to track specific pools of glutamate as a neurotransmitter.

Intracellular long-lived proteins (LLPs) provide structural support for several highly stable protein complexes and assemblies that play essential roles in ensuring cellular homeostasis and function. Recently, using using  in vivo dynamic metabolic stable isotope (15N) labeling of rodents, we showed that a subset of the mitochondrial proteome can persist for months in mammalian brains. Interestingly, our analysis revealed that mitochondrial LLPs (mt-LLPs) are concentrated at the inner mitochondrial membrane, specifically the sub-compartment of mitochondria called cristae. The exceptional longevity of the mt-LLPs over the course of months prompted a question of how the old and newly synthesized proteins are integrated within mitochondria and whether same peptide chains are recycled and intermixed, or spatially restricted within the organelles. To address this question, we performed a pioneering crosslinking experiments on intact mitochondria isolated from cortex and heart extracts of dynamically 15N-labeled mice. Through a combination of mitochondria immunocapture, advanced cross-linking and hybrid MS2–MS3 fragmentation approach we were able to directly probe for protein-protein interactions within old, new, and mixed mitochondrial complexes and showed that mt-LLPs are not randomly dispersed throughout individual mitochondria, but are rather spatially restricted and co-preserved for months is brains of mice. 

In this presentation I will describe modifications to an Orbitrap hybrid mass spectrometer that allow for deposition of protein-protein complex cations onto transmission electron microscopy grids. These samples are then directly imaged using TEM and the three-dimensional structures of the particles solved. Here we describe the roles of the chemical matrix, the use of the mass filter for sample purification, and outline how the technique has potential to advance the exciting field of cryo-EM.

The glycocalyx is a highly interactive environment composed of glycans on lipids and proteins. Cell-cell interactions are mediated through the glycocalyx, however the characterization of the glycocalyx remains a considerable challenge. In this presentation, new methods for the characterization of the cell membrane will be described. We have developed enrichment methods and liquid chromatography – mass spectrometry techniques that yield the glycolipid and glycoprotein components of the  cell membrane. We have employed glycosyltransferase inhibitors that allow us to modify the glycocalyx and create specific glycoforms and glycotypes. We then developed new methods for characterizing glycan-mediated protein-protein interactions. Through these tools, we explore the most comprehensive characterization of interactions in the glycocalyx.

The Kidney Precision Medicine Project (KPMP) consortium, in part, aims to create a human kidney tissue atlas from evaluating healthy and diseased biopsies. We have developed and optimized matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI)-based spatial metabolomics, lipidomics, and N-glycomics assays for analysis of human kidney biopsy tissues obtained from tissue recruitment sites (TRS) across the USA. The KPMP TRS acquire tissue from patients with varying disease states (e.g., Acute kidney injury, Chronic kidney disease, and Diabetic kidney disease). We have also linked our data to other omics-based analyses being performed within the consortium. By combining data obtained within our tissue integration site (TIS) with data from other TISs, we have begun to identify key metabolite, lipid, and N-glycan markers of cell types and disease states.

Knowledge of protein–metabolite interactions enhances understanding of biochemical processes and facilitates drug discovery, but the generation of unbiased data is challenging. We developed a sensitive, high-throughput ligand-purification/mass spectrometry approach to map endogenous small-molecule metabolites associated with essential enzymes, putative transcription factors, and functionally-unannotated proteins in the Gram-negative bacterium, Escherichia coli.  We applied structure-based computational modeling to define high-confidence ligand binding pockets, and assessed metabolic pathway relationships and evolutionary conservation to determine functional significance. The resulting interaction network includes hundreds of chemically diverse compounds including reaction substrates, products, analogs and cofactors. This ligand-interactome landscape illuminates gene function, reveals unexpected pathway crosstalk and feedback mechanisms, and suggests scaffolds for antimicrobial leads. Our approach is scalable and may be equally informative for mapping similar interactions, functional dependencies and drug design in other organisms, including human.

Post-translational modifications control the structure, activity, localization, and lifetime of nearly all proteins and are often dysregulated in human disease. However, identification of post-translational modifications has far outpaced assignment of their biological functions, an endeavor that requires detailed information about where and when these modifications occur within the cell. We are integrating principles from organic chemistry, protein engineering, and mass spectrometry-based proteomics to develop innovative tools for spatially and temporally resolved mapping of protein modifications in living cells. These technologies will advance our understanding of how post-translational modifications program biological function, leading to the development of new therapeutic hypotheses for the treatment of human disease.

The field of proteomics research has been rapidly evolving during the last two decades. During the last few years, the field of single-cell proteomics has progressed substantially. However, deep proteomic profiling of limited samples (e.g., small populations of rare cells, individual cells, microneedle biopsies, subpopulations of extracellular vesicles (EVs) isolated from minute volumes of biofluids) and especially, characterization of post-translational modifications (PTMs) and non-covalent protein interactions at such sample amount levels have been a major challenge because of very low abundance and high heterogeneity of complex biological matrices. In this presentation, we will overview a combination of advanced sample preparation, pressure- and electric field-driven ultra-low flow (ULF) high-efficiency nanoscale liquid phase separations coupled to mass spectrometry (MS) via alternative interfacing techniques to evaluate the potential applicability for high sensitivity, robust and reproducible proteomic profiling of individual cells and low ng-/sub-ng-level complex biological samples, using bottom-up, top-down, and native proteomic and PTM profiling approaches.

PARP1 is a nuclear enzyme that is critically involved in mediating DNA damage response (DDR). Its enzymatic function is to catalyze a protein posttranslational modification (PTM) known as Poly-ADP-ribosylation (PARylation). During genotoxic stress, PARP1 is recruited to nicked DNA and is rapidly activated, resulting in the synthesis of many PARylated proteins and initiation of the DNA repair mechanisms. It has been proposed that PARP1 inhibitors (PARPi) kill tumors via two distinct but interconnected mechanisms (i.e., PARP1 inhibition and PARP1 trapping). However, all FDA-approved PARPi possess both of these two activities, and their relative contribution to the cytotoxicity of PARPi is poorly understood.

We recently developed a large-scale mass spectrometric approach for the proteome-wide and site-specific characterization of Asp/Glu-PARylation (Zhang et al., Nature Methods, 2013; Gibson et al., Science, 2016; and Zhen et al., Cell Reports, 2017). Using this technology, we comprehensively characterized the downstream signaling network of PARP1 and found that the PARylated proteins are involved in not only DDR, but a wide array of other nuclear functions.

To understand the role of PARP1 trapping, we recently developed a series of small molecule degraders of PARP1 (Wang et al., Nature Chemical Biology, 2019). Treatment with one such compound, iRucaparib-AP6, results in highly efficient and specific PARP1 degradation in primary neonatal rat cardiomyocytes. iRucaparib-AP6 blocks the enzymatic activity of PARP1 in vitro, and PARP1-mediated PARylation signaling in intact cells. These “non-trapping” PARP1 degraders mimic PARP1 genetic depletion, which enables the pharmacological decoupling of PARP1 inhibition from PARP1 trapping. Using these unique compounds, we showed that PARP1 trapping is a key determinant of the DNA damage, cytotoxic and immunomodulatory effects of PARP1 inhibitors (Kim et al., eLife, 2020 and Kim et al., Cell Chemical Biology, 2021). In summary, these compounds represent ‘non-trapping’ PARP1 degraders that block both the catalytic activity and scaffolding effects of PARP1, providing an ideal approach for the amelioration of the various pathological conditions caused by PARP1 hyperactivation.

Protein turnover is a critical regulatory mechanism for proteostasis. However, proteome-wide turnover quantification is technically challenging, and, even in the well-studied E. coli model, reliable measurements remain scarce. Here, we quantify the degradation of ~3.2k E. coli proteins under 12 conditions by combining heavy isotope labeling with complement reporter ion quantification and find that cytoplasmic proteins are recycled when nitrogen is limited. We used knockout experiments to assign substrates to the known ATP-dependent proteases. Surprisingly, we find that none are responsible for the observed cytoplasmic protein degradation in nitrogen limitation, suggesting that a major proteolysis pathway in E. coli remains to be discovered. Lastly, we show that protein degradation rates are generally independent of cell division rates. Thus, we introduce broadly applicable technology for protein turnover measurements. We provide a rich resource for protein half-lives and protease substrates in E. coli, complementary to genomics data, that will allow researchers to decipher the control of proteostasis.