Determining and interpreting protein lifetimes in mammalian tissues

Abstract

The orchestration of protein production, degradation, and the regulation of protein lifetimes play a central role in the majority of biological processes. Recent advances in proteomics have enabled the estimation of protein half-lives for thousands of proteins in vivo. What is the utility of these measures and how can they be leveraged to interpret the proteome changes occurring during development, aging, and disease? This opinion summarizes leading technical approaches and highlights their strengths and weaknesses. We also disambiguate frequently-used terminology, illustrate recent mechanistic insights, and provide guidance for interpreting and validating protein turnover measurements. Overall, protein lifetimes, coupled to estimates of protein levels, are essential for obtaining a deep understanding of mammalian biology and the basic processes defining life itself.

Relevance of studying protein turnover in vivo in whole mammals

The complex nature of mammalian tissues presents several analytical challenges for studying in vivo protein turnover (see Glossary). However, recent advances in liquid chromatography tandem mass spectrometry (LC-MS/MS) and proteomic data analysis have made high throughput studies of protein turnover in vivo a reality [1–8]. The results of these studies have begun to revolutionize our understanding of proteome fidelity and proteostasis. In this opinion, we highlight the importance of measuring protein turnover and half-lives in vivo in whole mammals and why correctly interpreting these results is critical for advancing the field. Specifically, we summarize leading analytical strategies, discuss recent discoveries, and disambiguate terms used to describe proteome-wide measures of protein turnover. The strengths and weaknesses associated with commonly used experimental designs are also presented while highlighting recent mechanistic insights gained from studying protein turnover in vivo.

Efficient protein degradation and robust protein turnover are critical for maintaining organ homeostasis. Accordingly, impaired protein turnover plays a key role in numerous human disorders, diseases, and during aging [9]. Historically, most early studies of protein half-lives have focused on tracking individual proteins, but currently a constellation of large-scale approaches can be used to monitor the turnover rate of several thousand proteins in a single experiment [10] (see Table 1). While these are exciting times for large-scale studies of protein turnover, an under appreciation of what is being measured in vivo and the impact of these findings has emerged. In our opinion, it is critical to emphasize the importance of careful experimental design, precise terminology, and accurate data interpretation.

Table 1.

Compendium of turnover studies with particular relevance for mammalian tissues and whole animals organized from oldest to most recent studies.

N	1st author / Year	Label	Animal	Labelling paradigm(s)	Data analysis strategy	Considerations	Biological relevance	Ref.
1	Wu 2004	15N	Rattus norvegicus	Short pulses	15N absolute quantification	Pioneering work, introducing 15N for metabolic labeling of mammals	Feasibility study	[19]
2	McClatchy 2007	15N	Rattus norvegicus	Short and generational pulses	15N enrichment	Pioneering work for slow turnover proteins. Analysis limited to unlabeled proteins	Tissue comparison	[62]
3	Bateman 2007	13C6-leucine 13C6-Phe	Homo sapiens	Short pulses	Stable isotope labeling tandem MS (SILT)	Pioneering work for the analysis of turnover in humans	Method for quantifying turnover of low abundance proteins	[63]
4	Price 2010	15N	Mus musculus	Several short pulses	Exponential fitting	Pioneering work, designed for defining the half-lives of rapidly turnover proteins	Tissue comparison, half-life determination	[1]
5	Kasumov 2011	2H2O	Rattus norvegicus	Several short pulses	Exponential fitting	Pioneering work, introducing the 2H2O labeling for protein turnover studies	Effects of feeding on albumin synthesis	[56]
6	Savas 2012	15N	Rattus norvegicus	Generational pulse and chase	15N/14N abundance	Pioneering work, identifying intracellular extremely long-lived proteins (ELLPs)	Identification of NUPs and histones as ELLPs	[27]
7	Price 2012	2H2O	Homo sapiens	Drinking 2H20 administration (single pulse)	Kinetic model to account for precursor enrichment	Pioneering work for the analysis of protein lifetimes in humans	Protein turnover values in human plasma	[51]
8	Guan 2012	15N	Modelling / data analysis	Several short pulses	Several computational approaches	Pioneering work dealing with compartment modeling for mammalian turnover studies	Simple exponential decays are not appropriate whole animals	[15]
9	Toyama 2013	15N	Rattus norvegicus	Generational pulse and chase	15N/14N abundance	Detailed work characterizing the long-lived proteome	Detailed analysis of nuclear pore complex	[26]
10	Lam 2014	2H2O	Mus musculus Homo sapiens	Several short pulses	Exponential fitting	Extensive work also providing turnover of the human plasma proteome	Study of isoproterenol effects on heart remodeling	[64]
11	McClatchy 2015	Azidoho moalanine (AHA)	Mus musculus	Short pulses	Enrichment of AHA (click and biotinylation)	Work aimed at identify the newly-synthetized proteome	Introduces a method for newly synthesized proteins in tissues	[22]
12	Karunadharma 2015	2H3-leucine	Mus musculus	Several short pulses	Exponential fitting	Several conditions and tissues analyzed in parallel (respiratory chain)	Tissue comparison for mitochondrial proteins	[32]
13	Hammond 2016	13C6-lysine	Myodes glareolus (bank vole)	Several short pulses	Exponential fitting	Turnover in a small rodent (bank vole) by cross-species matching to mouse	Analytical methodology may contribute to variance in turnover	[6]
14	Lau 2016	2H2O	Mus musculus	Several short pulses	Exponential fitting	Comprehensive dataset of half-lives in the heart	Heart hypertrophy studied across six mouse strains	[58]
15	Rahman 2016	15N and 2H2O	Modelling / data analysis	Several short pulses	Several approaches including a stochastic model	One- and two-compartment models are used to analyze data from other studies	Models can be independent of the labeling isotope	[16]
16	Naylor 2017	2H2O	Modelling / data analysis	Short pulses	Exponential fitting	The described software platform simplifies analysis	Turnover rates are consistent across studies	[60]
17	Fornasiero 2018	13C6-lys 13C6-15N4-arginine	Mus musculus	Several pulses including pulse and chase	Exponential fitting / global modelling	Comprehensive dataset of half-lives in the brain and in other tissues including cell sorting and fractionation	Proteins have a reduced turnover at synapses Environmental enrichment changes specific lifetimes	[4]
18	Basisty 2018	2H3-leucine	Mus musculus	Several short pulses	Exponential fitting	Analysis of the antibody-enriched ubiquitinome at different mouse ages and fed with different diets	Aging increases bulk protein ubiquitination. Aggregated proteins are older	[7]
19	Heo 2018	13C6-lysine	Mus musculus	Short pulse and chase	Turnover ratios	Analysis of brain synaptic proteins	Proteins have a reduced turnover at synapses	[31]
20	Lau 2018	2H2O	Mus musculus	Several short pulses	Exponential fitting	Integrated omics: transcript abundance, protein abundance and turnover	Integrated omics provides several gene candidates for heart hypertrophy	[35]
21	Sadygov 2018	2H2O	Mus musculus and modelling / data analysis	Short pulses	Nonlinear fitting with outlier detection and removal	The software platform simplifies the analysis of protein lifetimes	Analysis of fatty liver disease reveals changes in ribosomal proteins	[59]
22	Ko 2018	15N	Mus musculus	Short pulse	Difference in labelling ratios	Aimed at understanding the effects of peripheral nerve injury on protein turnover	Peripheral nerve injury induces faster turnover of defined synaptic proteins	[37 ]
23	Alevra 2019	13C6-lysine	Protocol with data analysis	Short pulses	Exponential fitting / global modelling	Detailed indications for determining protein lifetimes	Protocol covering aspects of lifetime measurements	[3]
24	McClatchy 2020	AHA	Mus musculus	Short pulse and chase	Exponential fitting	Use of AHA for determining degradation dynamics in different tissues	Subcellular localization and activity influence protein stability	[8]
25	Bomba - Warczak 2021	15N	Mus musculus	Several pulses (up to 4 months)	15N/14N abundance	Pulse labelling shows that some mitochondrial proteins are exceptionally long-lived	Oxidative phosphorylation complexes are preserved with low subunit exchange	[2]
26	Krishna 2021	15N	Mus musculus	Pulse and chase	15N/14N abundance	NanoSims confirmation that some mitochondrial proteins are exceptionally long-lived	COX7C contributes to oxidative phosphorylation complex assembly	[25]
27	Hark 2021	15N	Mus musculus	Generational pulse and chase	15N/14N abundance	Three genetic models of Alzheimer’s disease (AD) were analyzed	The turnover of synaptic vesicle associated protein is altered in AD	[34]
28	Chepyala 2021	13C6-lysine	Mus musculus	Several short pulses	Exponential fitting	Three settings can be used for calculating lifetimes	Direct measurements of lysine pools improve data	[17]
29	Rolfs 2021	13C6-lysine	Mus musculus	Several short pulses	Exponential fitting	Analysis of half-lives across 5 mouse tissues	Postnatal tissue development complicates the analysis of results	[65]
30	Kluever 2022	13C6-lysine	Mus musculus	Several short pulses	Exponential fitting	Analysis of mean protein lifetimes in aged mouse brain	Aged brain proteins last longer than young ones	[33]
31	Hammond 2022	2H2O and 13C6-lys	Mus musculus	Several short pulses	Exponential fitting	Comparison between different labelling strategies	The AA label tested are suited for turnover studies	[44]

A cohesive terminology to define protein renewal parameters

Proteostasis refers to the processes that ensure the delicate balance of protein production, maintenance, and degradation and is vital for cellular and tissue function. While the general concepts underlying protein synthesis and degradation are well understood, the terms describing these processes are sometimes used ambiguously. In this section, we would like to bring clarity by providing a common set of terms for the field.

Cells contain proteins with abundances that roughly vary from thousands to tens of millions copies [11]. For an individual protein, the protein lifetime encompasses the entire time from synthesis (i.e., birth) to degradation (i.e., death). Historically, protein renewal (i.e. the replacement of old proteins), has been quantified in terms of average half-life. However, measures of protein half-lives in mammals cannot be reliably calculated for long-lived proteins (LLPs). Since for some LLPs the mean protein lifetime is years or even decades, even small differences in labeling ratio will greatly influence these values.

Often the term meaning “protein lifetime” is used interchangeably with “protein half-life” and, as this is not strictly correct, we advise to use the latter whenever possible. In a steady-state situation, protein half-life is the point in time when the degradation of a population of old proteins is equal to the newly synthesized population of proteins. Thus, by definition at the steady state, the mean protein lifetime is the same as the protein half-life (Box 1). Caution needs to be taken when non-steady state conditions are the subject of investigation as these two measures diverge. It is also important to emphasize that this terminology, in the context of LC-MS/MS-measures, reflects an average process (e.g., the mean protein lifetime, half-life, or turnover) of a pool of proteins with the same amino acid sequence. This is because these technologies typically do not achieve measures at the level of single molecules and rather measure a population of peptides after protein extraction and trypsin digestion (Box 2).

Text box 1: Analyzing and interpreting in vivo measures of protein kinetics.

In a steady-state situation, protein kinetics can be approximated with simple exponential decays (Fig. I), where the degradation and synthesis rate constants are equal. In reality, in a mammal the availability of precursor molecule might depend on the modalities by which the labels are provided (see Box 2), further complicating half-lives calculations.

When interpreting the results of metabolic labeling, it is critical that the age of the animal and period of labeling are carefully considered. This is exemplified by the fact that some of the long-lived proteins (LLPs; such as extracellular matrix components or nuclear scaffolds in postmitotic cells) are synthesized at a specific age of the animal development [52,53]. If the goal is to study protein half-lives, using relatively short labeling periods with heavy amino acids is often appropriate, especially for measuring the components of the proteome that are replaced frequently. However, if the goal is to identify LLPs, then multi-generation labeling followed by a chase in unlabeled food is probably more appropriate.

Protein turnover measurements in isotopic labeling experiments are typically fractional, and the intact heavy and light labelled peptides are captured in the same MS1 scan. Thus, they are internally normalized within individual samples analyses and are not impacted by sample-to-sample technical variation. Attention must be exerted for the interpretation of these calculations since, depending on the workflow used, the fractional abundance of isotopically labeled proteins can reflect either an “older” or a “newer” population of proteins. However, bottom-up proteomic analysis depends on the ability to identify the peptide sequence in the MS2 scan. For some studies, both the heavy and light peptide pairs are identified, while in others only one isotopologue is selected for MS2 and the abundance is solely based on inferring its’s sequence based on the m/z an peak intensities. MS-based imaging of metabolically labeled tissue sections can be achieved with MALDI and NanoSIMS, providing spatial information.

Several labeling and analysis strategies have been deployed to investigate protein turnover and measure protein lifetimes in vivo. A problem that needs to be addressed when studying protein turnover in vivo is that amino acids (essential and non-essential) are recycled within animals to preserve energy and increase metabolic performance. All these approaches are based on theoretical predictions of kinetic influx and efflux of pools of amino acids and proteins and several computational approaches exist [1,4,15,17,54,55]. It is important to underline that these approaches are based on assumptions, which are required to allow mathematical modeling (see Table 1). The most common assumption is that the protein of interest does not change their level during the analyzed period, which may confound interpretation of the results.

Box 1, Figure I.

Schematic representation of exponential decay and the basic equations at the steady state.

Text box 2: Labeling rodents with stable isotopes for studying protein turnover.

Stable isotopes can be incorporated into rodent proteins by metabolic labeling with custom chow enriched with “heavy” nitrogen (i.e., ¹⁵N) or carbon (i.e., ¹³C). The rodent chow is formulated without “light” atoms for one or more defined molecular species, thus allowing to obtain specific labeling.

There are currently two practical strategies used for protein metabolic labeling: 1) Providing “heavy essential amino acids”, such as lysine (e.g., ¹³C₆-Lys), that cannot be synthesized and are thus solely provided from the diet and incorporated during protein biosynthesis. 2) Providing “labeled amino acid precursors”, that are incorporated into biomolecules and proteins through enzymatic reactions occurring within cells. As an example, this is what is achieved when employing ¹⁵N diets. In this case ¹⁵N atoms are slowly incorporated into all the nitrogen-containing molecules, such as amino acid sidechains and backbones. Since mice cannot efficiently incorporate ¹⁵N as a derivatized salt, historically the ¹⁵N diet is based on blue-green algae (i.e., Spirulina platensis) which can use ¹⁵N as the sole nitrogen source.

An alternative strategy is to deliver heavy atoms by subcutaneous injection or in the drinking water in the form of deuterium oxide (i.e., ²H₂O) [56]. Due to the large difference in relative mass with respect to ¹H (protium), deuterium (²H) is the only stable isotope which exerts a sizeable “kinetic isotope effect” that slows enzymatic reactions and results in toxicity for concentration higher than 30% in animals and eukaryotic cells [57]. Nevertheless, due to its relatively low price, heavy water is an attractive solution for protein turnover studies not only in rodents but also in humans [51]. In practice, the toxicity issues are mitigated by using low concentrations of heavy water and relying on robust bioinformatic approaches for data interpretation [58–60].

There are advantages and disadvantages to using global isotopic labeling (e.g.¹⁵N) labeled essential amino acids (Fig. I). Briefly, while atom-based tracers can provide reliable measures of relative labeling, analysis of the MS spectra is challenging due to the presence of heterogeneous populations of peptides with the same chemical composition differing only by their isotopic composition (i.e., isotopologues). It is important to mention that recently, this aspect has been addressed to simplify the analysis of the isotopolog distribution by either forcing a light isotope labelling shift or by using water labelling and considering only the enrichment from two mass isotopomers [18,61]. The use of heavy essential amino acids simplifies the analysis but allows protein turnover measures only for peptides containing the heavy amino acids. At the same time, as these have predictable mass shifts, which can be precisely separated and accurately measured by MS, their modelling is easier to handle.

Box 2, Figure I. Theoretical peptide mass spectra from a metabolic labeling experiment (left), commonly used experimental labeling schemes, and list of applications (right).

What is the main difference between “continuous labelling” vs. “pulse-chase” labelling? In the “continuous labelling” strategy, since there is only one type of measurement, more complex trajectories in protein degradation dynamics are not captured. At the same time, it is simpler to analyze and model the data for quantitative purposes if they are approximated to first-order degradation kinetics. For proteins that are long-lived and possibly stabilized in aggregates, a pulse-chase approach might become useful in order to decrease the background noise arising from the short-lived (non-aggregated) counterparts. In this case a chase allows to washout the short-lived proteins and reveal more reliably the longer-lived species. Both strategies can be combined to obtain different set of data and for checking possible labeling inconsistencies.

Notably, the previously mentioned measures are currently limited because proteins often exist in multiple independent pools. Unfortunately, we currently lack a sufficient repertoire of probes or analytical tools needed for distinguishing the turnover of different protein sub-populations in mammals in vivo. The same protein is frequently present in multiple protein complexes localizing to various organelles with dissimilar turnover rates. To be more accurate, we thus propose to refer to proteins with multiple binding partners and functions (i.e., present in distinct pools) that show different turnover rates in a single cell type as multi-lifetime proteins (Fig. 1A). Currently, in whole mammals and in order to properly analyze multi-lifetime proteins, we need to combine bulk measurements with biochemical purification of intact protein complexes and organelles in order to accurately determine their half-lives and distinguish protein subpopulations [12]. This is mainly due to technical limitations as for whole animals we lack simple and reliable experimental workflows for independently determining protein production and degradation constants as those available in cell culture [13].

Figure 1: Assessing and understanding in vivo protein turnover on multiple scales.

(A) Theoretical cartoon depiction of single and multi-lifetime proteins and how these factors impact global measures of protein turnover. (B) Cartoon depiction of possible changes in protein abundance and turnover upon a manipulation in vivo (C) Bulk measures of protein turnover are influenced at multiple scales during in vivo experiments. This biological complexity can occur at the molecular, organellular, cellular, organ, or organismal (i.e., individual) level. While many of these validations discussed are straightforward, they are rarely performed and we encourage their inclusion whenever possible.

Metabolic labeling with stable isotopes for studying protein lifetimes

The standard experimental strategy to study protein lifetimes on a proteome-wide scale uses metabolic labelling with heavy stable isotopes (i.e., ²H, ¹³C, or ¹⁵N) coupled to LC-MS/MS-based proteomic analysis. Variations of this experimental paradigm have been reported, including the use of light isotope labelling (¹²C) [14], but the general concept is simple and can be summarized as follows: small mammals, typically rodents, are metabolically labeled through chow or water enriched with select stable heavy isotopes that have extremely low levels in nature. As the animals consume the provided food, the supplemented isotopes are gradually incorporated into newly synthesized proteins during labeling periods that can span from days to months at a rate reflecting protein turnover (Fig. 1 and Box 2). In practice, a short period of metabolic labeling for a few weeks is sufficient to measure levels of incorporation for several organs with LC-MS/MS-based proteomics. However, to achieve near complete labeling (99%) that is required for some experimental workflows, labeling of mice for two generations is required. Following the measure of the metabolic labelling in vivo, several strategies can be used for the analysis of data to extract protein turnover measures either considering or not the reusage of the isotopic labels. Since detailed data analysis of metabolic labelling largely exceeds the purpose of this Opinion, readers might refer to several excellent works covering this aspect in detail [4,6,15–18].

Amino acids or amino acid precursor molecules enriched with heavy atoms are now the most commonly used chemical tracers used to study protein turnover [1,4,5,19–21]. Alternative chemical strategies for measuring protein turnover leverage biorthogonal labeling and can be performed with amino acid analogs. These can be either directly incorporated instead of methionine such in the case of L-azidohomoalanine (AHA) [22], or through the expression of a modified methionyl-tRNA synthetase such as in the case of azidonorleucine (ANL) [23]. While click-chemistry based strategies provide an opportunity to enrich the labeled proteome and measure the newly synthesized proteins, they can be toxic to animals at high concentrations, limiting their applicability for studies of protein turnover. In contrast, stable isotopes are particularly powerful for metabolic labeling since they are almost indistinguishable from naturally occurring atoms and provide the rare opportunity to confidently measure protein half-lives under nearly endogenous conditions [24]. Furthermore, protein labeling with these minimal tracers surmounts the limitations associated with the introduction of exogenous over-expressed fluorescent proteins or epitopes that may alter natural protein production, folding, complex formation, and degradation.

Studying protein lifetimes in mammals accelerates biological discoveries

Proteomics provides a rare opportunity to probe relationships between groups of proteins with similar turnover rates and thus extract biological information about complex biogenesis and degradation directly from mammalian tissues. For example, it has been established which subunits of several large protein complexes, such as the core of the nuclear pore and parts of the mitochondrial oxidative phosphorylation chain, show similarly exceptionally long lifetimes [2,4,25]. It has also been discovered that these LLPs are maintained together as single units for months and even years while several other components of the same complexes are turned over on much shorter time frames [2,4,25–27]. These discoveries represent the basis of new research avenues addressing differences in macromolecular complexes occurring during aging and pathologies that might affect their stability. Moreover, selectively impaired protein turnover can be observed after stress or by the expression of mutant proteins that become misfolded and accumulate [28][29,30]. Changes in protein half-lives can also reflect physiological responses, which modulate protein turnover [4,31]. For example, changes in subcellular localization and posttranslational modifications can influence protein stability. For this reason measures of protein half-live can be used for discovery purposes to reveal previously unknown interactors and molecular mechanisms (Fig. 1C).

In the context of aging, the study of protein lifetimes has shown that different tissues might affect protein turnover in slightly dissimilar manners, although the relative half-lives of mitochondrial components are very precisely coordinated across tissues [32]. The brain of aged animals specifically shows reduced protein turnover, increased half-lives, and specific alterations that might point to metabolic differences that affect proteome composition. Interestingly, these alterations are also linked to biological processes that are observed in neurodegenerative diseases [33]. These studies only start addressing the implication of different pathways that regulate protein stability and more systematic approaches will be necessary to understand the precise molecular causes of these observations and the implication of proteasomal and lysosomal mechanisms.

To demonstrate the usefulness of protein turnover measurements, we will summarize some exemplary case studies. To better understand the mechanisms underlying neurodegeneration, pulse-chase labelling of amyloid precursor protein (App) knock-in mice was used and showed that amyloid accumulates over a time period of months [34]. The amyloidogenic processing of App also caused a selective turnover impairment of synaptic vesicle associated proteins. It is important to underline that there is not a standardized “protein turnover analysis workflow”, and different strategies need to be tailored for each biological question. As an example, to study amyloid deposition, an initial long-term pulse was necessary to label the majority of the slowly accumulating protein pool and a successive chase was required for tracking the degradation of the aggregated-long lived pool and the non-aggregated protein that is shorter lived. This strategy increased signal and minimized noise, which would not be possible with a continuous labelling paradigm, which might be more informative for the determination of protein half-lives.

In the study of heart remodeling, protein turnover has been used in an elegant multiomic study integrating other omics measures (such as transcript and protein abundance), to reveal new disease gene candidates linked to heart hypertrophy [35]. In the same study, the integrated analysis of half-lives has confirmed that protein–protein interacting partners are coordinated in turnover and thus changes in these measures can be used to infer changes in protein complex composition.

Protein turnover has also been used for addressing changes in synaptic physiology, both in the central and in the peripheral nervous system [4,31,36,37]. One intriguing aspect reveled by these studies is that the localization to a specific cellular subdomain (the synapse) extends the half-lives of proteins [4,31], suggesting that local degradation mechanisms at the cell body or at the synapse are not equally active. One other intriguing aspect is that peripheral nerve injury decreases the turnover of synaptic proteins and suggesting that cellular damage could be one of the causes leading to inefficient neurotransmission [36].

Beyond the several examples from specific fields of biology, there are some overarching questions about biological systems that protein turnover measurements can help to solve. As an example, we still know very little about how proteome composition and homeostasis are maintained, and can only address these questions when several thousands of protein lifetimes and abundances are measured in parallel. These large-scale studies facilitate the identification of determinants of protein stability, such as biochemical parameters influencing protein turnover [38]. However addressing the role of post-translational modifications in protein stability remains challenging [39] (see outstanding questions).

Outstanding Questions.

Does physicochemical protein damage broadly interfere with protein turnover and how does this mechanistically contribute to aging and pathology?

Which are the molecular processes that regulate protein turnover, and can they be pharmacologically manipulated?

What is the impact of post-translational modifications (PTMs) such as phosphorylation on protein turnover? Will it be possible to differentiate at the molecular level the modifications that have causative effects in protein stabilization or degradation?

Specific PTMs (such as ubiquitination) are thought to regulate protein turnover, at the same time their role and regulation of protein stability in vivo remains largely unexplored. Will it be possible to obtain a comprehensive and quantitative atlas of these protein modifications combining information about their subtypes, abundance, and influence on protein stability?

Will the rapidly-evolving field of protein turnover be able to measure turnover at the “intact protein” level, transitioning from a peptide-centric to a more precise “top-down” approach?

Can in vivo protein turnover measurements become more common in the clinical context and be used to guide early diagnosis and personalized medicine?

Protein half-life and protein abundance: two orthogonal measures

What is the difference between measuring protein levels (i.e. protein abundances) and protein half-lives? In simple terms, classical protein abundance measures provide information about the relative quantity of protein, which in principle could be the consequence of changes in gene expression, in protein degradation, or in both processes. At the steady-state, protein production and degradation are equal by definition, so protein levels do not change (Fig. 1B). Protein half-lives at the steady state provide a substantially different set of data when compared to protein abundance. For example, if both production and degradation are doubled (and thus turnover is faster and the half-life becomes shorter), protein abundance would not change and protein abundance measurements would not be informative. One could still foresee that in such situation a protein could be used much faster and thus more rapidly degraded due to the increased molecular damage occurring. If the levels of that protein need to be kept constant for the optimal function of the organism, homeostatic mechanisms will continuously counterbalance the increase in degradation with higher protein production, very quickly reaching a new steady state situation where the overall protein levels are not changed.

In non-steady state situations, following metabolic labelling for a given period, half-lives measures are more complex to investigate, especially in whole mammals, where formal analysis of protein turnover trajectories is not as simple as in cell culture [13]. In any case, associated to protein abundance measurements, protein turnover measures provide additional information about the dynamic changes in protein expression. In this context, we would like to stress that from a practical point of view, protein abundance measures when changes are small (< 5-10% difference) are often too noisy to provide reliable information. In these cases, measures of metabolic incorporation of stable isotopes can be extremely quantitative as they often include an “internal control” which is the unlabeled version of the same protein, across multiple biological replicates. In other words, the fractional abundance (i.e. “old” / (“new” + “old”)) is a direct way to capture the protein properties [26]. In practice, this allows one to distinguish changes of protein renewal rates with high precision that cannot be currently obtained from conventional protein abundance measures. Ideally for discovery purposes, metabolic labelling dynamics and protein abundance should be monitored in parallel, allowing investigators to obtain a more complete picture of changes following perturbations.

Are protein lifetimes similar in in vitro and in vivo?

While simplified models are instrumental for addressing the regulation of protein lifetimes [40], recent data suggest that protein lifetimes measured in cultured cells underestimate those obtained from in vivo studies [4,41,42]. Immortalized cells have abbreviated cell cycles and may have lost key regulatory mechanisms necessary to ensure proper coordination of protein complex assembly and proteome stoichiometry. Ultimately, this situation may accelerate cellular processes and protein turnover. Additionally, proteins, which have a lifetime exceeding the duration of the cell cycle will complicate the interpretation.

In some cases, by studying primary cells we can overcome this limitation, with the caveat they frequently represent model systems that reflect developmental processes in vivo that require robust protein synthesis. This is problematic, since it can result in misestimating the protein turnover rates observed in vivo. On the other hand, there are cases where cells in culture do show very long lifetimes, for example, senescent cells undergoing contact inhibition [43]. These cells have minimal protein synthesis and show extended protein lifetimes that only capture minimal aspects of the dynamic situations observed in vivo.

While all these cellular models have their strengths and weaknesses, we suggest using them only when an in vivo alternative does not exist. In general, we wish to emphasize that researchers need to understand that to obtain the most biologically meaningful results; the biological question should match the protein labeling paradigm (Fig. 1C). If the goal is to determine absolute protein lifetimes, one needs to be very careful in considering the experimental limitations. However, determining relative changes in protein lifetimes between two conditions (i.e. genetic, pharmacological) is generally straightforward as long as the protein measures are precise.

Possible strategies to troubleshoot and validate protein turnover measurements

Validation of protein turnover changes pose a challenge even if several possible strategies exist. At first, there is a need to discern technical validations from biological validations, which ultimately serve as a confirmation that the observed alterations of the biological process under investigation are reliable and ultimately correspond to a meaningful biological change.

In terms of technical validations, we would like to delineate several categories. One category addresses the rigor of the fractional abundance measurements. In this case, a spike-in experiment can be run in parallel with the biological samples, to provide an idea of the sensitivity of the actual measures. This is obtained by mixing tissue extracts from unlabeled and fully-labeled animals and measuring different ratios of incorporation in the mixed samples [4]. While the mixing is performed in vitro, the tissues that will be measured are the same used for the actual turnover measurement, providing useful information about the reliability of the measures. One other category addresses the reproducibility across cohorts of animals and should be addressed either by performing the same measurements with different animals and potentially in different labs, to ensure that the measurements are reproducible (although this is often practically unfeasible, due to the relevant costs of these experiments). One additional technical validation addresses the influence of a defined isotope or of a precise diet formulation on the measurements. To solve this issue, one option is to perform a second experiment and track protein degradation with a different isotopologue. Finally, to avoid measurement differences due to variability of the analysis workflow (typically the type of LC-MS/MS), several cross-validation methods based on classical biochemical analysis or imaging approaches can be used, although these usually can recapitulate changes for only a defined number of proteins. Cross validating results using orthogonal MS analysis and labelling strategies is an opportunity for integration of complementary data and reach more reliable conclusions. A recent work directly compared different strategies to analyze datasets that are obtained using different labeling techniques but on the same mouse strains [44]. Such integrated approaches will be key for establishing shared ground truth frameworks for analyzing and interpreting proteome wide turnover measurements.

To provide spatial information that is lost in other MS approaches, NanoSIMS or MS imaging (i.e., MALDI) can be used to study protein replacement in the endogenous environment using tissue sections [45–49]. Biological validations are more complicated and, although virtually infinite scenarios exist, we showcase a few examples in Fig. 1C and discuss additional interpretative aspects in Box 2 that might be useful for planning the correct validations.

Concluding remarks

It is our opinion that the proliferation of new technologies and the development of robust analysis workflows will culminate in a wealth of in vivo protein turnover measures that will allow researchers to address several open questions (see Outstanding Questions Box).

The main challenge that this field will face is how to obtain a “meaningful biological understanding” of the complex underpinnings of protein lifetimes. To overcome this, we will need to consider contemporarily technological aspects that can influence the biological interpretation of the results while also combining more advanced metabolic labeling-based measurements in a multi-level perspective. To provide a simple example, the influence of different cell types on these measurements is difficult to address in vivo and might require cell-sorting or genetic labeling strategies. Tissues that harbor only a few cell types will be less challenging than more complex tissues with highly heterogeneous cellular compositions. Although even a single cell type is likely to have differences that are only addressable at single-cell resolution. Furthermore, within cells, protein pools might have dissimilar lifetimes depending on their interactors, post-translational modifications or subcellular localization reflecting an opportunity for yet unimaginable biological discoveries.

Overall, we believe that, in combination with numerous experimental manipulations, all of these approaches will allow to systematically define the dynamic properties of the proteomes which will provide new avenues for human pathophysiology and biomedical research [50,51].

Highlights.

Robust proteome homeostasis (i.e., proteostasis) is critical for organismal health since proteome imbalance and accumulation of damaged molecules have negative effects on nearly all biological processes.

It has become clear that determining protein half-life measures in mammals provides vital information at the whole proteome level for understanding dynamic phenotypic changes across scales.

While methods and analysis frameworks for determining protein half-lives in vivo at the whole proteome level are becoming more popular, they require careful customization depending on the biological question.

Samples obtained from metabolic labeling schemes can be used to provide spatial turnover information through mass spectrometry imaging technologies such as matrix assisted laser desorption ionization (MALDI) or nanoscale secondary ion mass spectrometry (NanoSIMS).

Protein abundance and turnover measures can be obtained with similar mass spectrometry-based approaches but are fundamentally different and provide valuable and complementary insights.

Glossary

Degradation and synthesis rate constants

values that quantify the relative change in concentration of a certain molecule over time. Following metabolic labelling in animals, in a steady state situation these constants can be obtained from relative labelling at the level of single proteins.

Liquid chromatography tandem mass spectrometry (LC-MS/MS)

the most commonly used analytical strategy for proteomic studies. LC-MS/MS ultimately provides information about peptide sequence, abundance, and isotopic distribution that are all key aspects for quantifying protein turnover.

Matrix assisted laser desorption/ionization (MALDI)

an ionization technique allowing for in situ analysis of proteins and peptides within biological samples and providing valuable spatial information that is usually lost in conventional LC-MS/MS.

Metabolic labeling

a type of labeling cells or tissues based on introducing one or more tracer isotopes into the cell culture media or the animal diet. This approach allows a stable isotope that is present in protein precursor molecules to be processed into labeled amino acid(s) and to become incorporated into the proteome during mRNA translation.

Nanoscale secondary ion mass spectrometry (NanoSIMS), also known as multi-isotope imaging mass spectrometry (MIMS)

an analytical strategy that facilitates nanoscale resolution measure of isotopic composition of biological molecules in situ.

Mean protein lifetime

represents the average length of time a protein species persists, often formally defined as the time required for a protein to be reduced to 1/e of its initial quantity.

Protein half-life

average time required for a protein species to be reduced to half of its initial number of molecules.

Protein turnover

a dynamic process by which old proteins are degraded into amino acids and are subsequently replaced by new versions through protein synthesis within cells.

Proteostasis

a term comprising the integrated cellular activities determining/regulating protein homeostasis including synthesis, folding, trafficking, and degradation aimed at ensuring proteome fidelity and function.

Pulse-chase analysis

a technique used for examining changes in abundance and labelling over a defined time period. Organisms or cells are exposed to a labeled compound (i.e., pulse), then the labeled compound is removed, and replaced with an unlabeled version (i.e., chase). By monitoring the level of the labeled compound over time one can determine its degradation dynamics and half-life.

Stable isotopes

atoms generally containing extra neutrons that are stable and non-radioactive and hence often present in nature and not dangerous for cells and animals. Examples of practical usage of stable isotopes for protein metabolic labeling include essential amino acids containing one or more stable isotopes (e.g., ¹³C₆-lysine, containing 6 atoms of ¹³C).

Stable isotope labeling with amino acids in culture (SILAC)

a LC-MS/MS based proteomic technique using stable isotopes supplied in cell culture media to measure relative differences in protein abundance.

Stable isotope labeling in mammals (SILAM)

a method to metabolically label rodents with stable isotopes.