Search

SignalP (https://services.healthtech.dtu.dk/services/SignalP-6.0/) is a very popular prediction method for signal peptides, the intrinsic signals that make proteins secretory. The SignalP web server has existed since 1995 and is now in its sixth major version. In this historical account, we (three authors who have taken part in the entire journey plus the first author of the latest version) describe the differences between the versions and discuss the various decisions taken along the way., SignalP ( https://services.healthtech.dtu.dk/services/SignalP-6.0/ ) is a very popular prediction method for signal peptides, the intrinsic signals that make proteins secretory. The SignalP web server has existed since 1995 and is now in its sixth major version. In this historical account, we (three authors who have taken part in the entire journey plus the first author of the latest version) describe the differences between the versions and discuss the various decisions taken along the way.

Secretory proteins are critically dependent on the correct processing of their signal sequence by the signal peptidase complex (SPC). This step, which is essential for the proper folding and localization of proteins in eukaryotic cells, is still not fully understood. In eukaryotes, the SPC comprises four evolutionarily conserved membrane subunits (Spc1-3 and Sec11). Here, we investigated the role of Spc2, examining SPC cleavage efficiency on various models and natural signal sequences in yeast cells depleted of or with mutations in Spc2. Our data show that discrimination between substrates and identification of the cleavage site by SPC is compromised when Spc2 is absent or mutated. Molecular dynamics simulation of the yeast SPC AlphaFold2-Multimer model indicates that membrane thinning at the center of SPC is reduced without Spc2, suggesting a molecular explanation for the altered substrate recognition properties of SPC lacking Spc2. These results provide new insights into the molecular mechanisms by which SPC governs protein biogenesis.

Human growth hormone (hGH) is a four-helix bundle protein of considerable pharmacological interest. Recombinant hGH is produced in bacteria, yet little is known about its folding during expression in Escherichia coli. We have studied the cotranslational folding of hGH using force profile analysis (FPA), both during in vitro translation in the absence and presence of the chaperone trigger factor (TF), and when expressed in E. coli. We find that the main folding transition starts before hGH is completely released from the ribosome, and that it can interact with TF and possibly other chaperones.

The central themes of Cell Boundaries concern the structural and organizational principles underlying cell membranes, and how these principles enable function. By building a biological and biophysical foundation for understanding the organization of lipids in bilayers and the folding, assembly, stability, and function of membrane proteins, the book aims to broaden the knowledge of bioscience students to include the basic physics and physical chemistry that inform us about membranes. In doing so, it is hoped that physics students will find familiar territory that will lead them to an interest in biology. Our progress toward understanding membranes and membrane proteins depends strongly upon the concerted use of both biology and physics. It is important for students to know not only what we know, but how we have come to know it, so Cell Boundaries endeavours to bring out the history behind the central discoveries, especially in the early chapters, where the foundation is laid for later chapters. Science is far more interesting if, as students, we can appreciate and share in the adventures-and misadventures-of discovering new scientific knowledge. Cell Boundaries was written with advanced undergraduates and beginning graduate students in the biological and physical sciences in mind, though this textbook will likely have appeal to researchers and other academics as well. Highlights the history of important central discoveries Early chapters lay the foundation for later chapters to build on, so knowledge is amassed High-quality line diagrams illustrate key concepts and illuminate molecular mechanisms Box features and spreads expand on topics in main text, including histories of discoveries, special techniques, and applications.

In recent years, it has become clear that many homo- and heterodimeric cytoplasmic proteins in both prokaryotic and eukaryotic cells start to dimerize cotranslationally (i.e., while at least one of the two chains is still attached to the ribosome). Whether this is also possible for integral membrane proteins is, however, unknown. Here, we apply force profile analysis (FPA)—a method where a translational arrest peptide (AP) engineered into the polypeptide chain is used to detect force generated on the nascent chain during membrane insertion—to demonstrate cotranslational interactions between a fully membrane-inserted monomer and a nascent, ribosome-tethered monomer of the Escherichia coli inner membrane protein EmrE. Similar cotranslational interactions are also seen when the two monomers are fused into a single polypeptide. Further, we uncover an apparent intrachain interaction between E14 in transmembrane helix 1 (TMH1) and S64 in TMH3 that forms at a precise nascent chain length during cotranslational membrane insertion of an EmrE monomer. Like soluble proteins, inner membrane proteins thus appear to be able to both start to fold and start to dimerize during the cotranslational membrane insertion process.

During SecYEG-mediated cotranslational insertion of membrane proteins, transmembrane helices (TMHs) first make contact with the membrane when their N-terminal end is ~ 45 residues away from the peptidyl transferase centre. However, we recently uncovered instances where the first contact is delayed by up to ~ 10 residues. Here, we recapitulate these effects using a model TMH fused to two short segments from the Escherichia coli inner membrane protein BtuC: a positively charged loop and a re-entrant loop. We show that the critical residues are two Arg residues in the positively charged loop and four hydrophobic residues in the re-entrant loop. Thus, both electrostatic and hydrophobic interactions involving sequence elements that are not part of a TMH can impact the way the latter behaves during membrane insertion.

The ribosome stalling mechanism is a crucial biological process, yet its atomistic underpinning is still elusive. In this framework, the human XBP1u translational arrest peptide (AP) plays a central role in regulating the unfolded protein response (UPR) in eukaryotic cells. Here, we report multimicrosecond all-atom molecular dynamics simulations designed to probe the interactions between the XBP1u AP and the mammalian ribosome exit tunnel, both for the wild type AP and for four mutant variants of different arrest potencies. Enhanced sampling simulations allow investigating the AP release process of the different variants, shedding light on this complex mechanism. The present outcomes are in qualitative/quantitative agreement with available experimental data. In conclusion, we provide an unprecedented atomistic picture of this biological process and clear-cut insights into the key AP-ribosome interactions.

Signal peptides (SPs) are short amino acid sequences that control protein secretion and translocation in all living organisms. SPs can be predicted from sequence data, but existing algorithms are unable to detect all known types of SPs. We introduce SignalP 6.0, a machine learning model that detects all five SP types and is applicable to metagenomic data. A new version of SignalP predicts all types of signal peptides.

Signal peptides (SPs) are short amino acid sequences that control protein secretion and translocation in all living organisms. SPs can be predicted from sequence data, but existing algorithms are unable to detect all known types of SPs. We introduce SignalP 6.0, a machine learning model that detects all five SP types and is applicable to metagenomic data.

Signal peptides (SPs) are short amino acid sequences that control protein secretion and translocation in all living organisms. SPs can be predicted from sequence data, but existing algorithms are unable to detect all known types of SPs. We introduce SignalP 6.0, a machine learning model that detects all five SP types and is applicable to metagenomic data.

Signal peptides (SPs) are short amino acid sequences that control protein secretion and translocation in all living organisms. SPs can be predicted from sequence data, but existing algorithms are unable to detect all known types of SPs. We introduce SignalP 6.0, a machine learning model that detects all five SP types and is applicable to metagenomic data.

Proteins commonly fold co-translationally at the ribosome, while the nascent chain emerges from the ribosomal exit tunnel. Protein domains that are sufficiently small can even fold while still located inside the tunnel. However, the effect of the tunnel on the folding dynamics of these domains is not well understood. Here, we combine optical tweezers with single-molecule FRET and molecular dynamics simulations to investigate folding of the small zinc-finger domain ADR1a inside and at the vestibule of the ribosomal tunnel. The tunnel is found to accelerate folding and stabilize the folded state, reminiscent of the effects of chaperonins. However, a simple mechanism involving stabilization by confinement does not explain the results. Instead, it appears that electrostatic interactions between the protein and ribosome contribute to the observed folding acceleration and stabilization of ADR1a., BN/Sander Tans Lab

In Gram-negative bacteria, periplasmic domains in inner membrane proteins are cotranslationally translocated across the inner membrane through the SecYEG translocon. To what degree such domains also start to fold cotranslationally is generally difficult to determine using currently available methods. Here, we apply Force Profile Analysis (FPA) - a method where a translational arrest peptide is used to detect folding-induced forces acting on the nascent polypeptide - to follow the cotranslational translocation and folding of the large periplasmic domain of the E. coli inner membrane protease LepB in vivo. Membrane insertion of LepB's two N-terminal transmembrane helices is initiated when their respective N-terminal ends reach 45-50 residues away from the peptidyl transferase center (PTC) in the ribosome. The main folding transition in the periplasmic domain involves all but the similar to 15 most C-terminal residues of the protein and happens when the C-terminal end of the folded part is similar to 70 residues away from the PTC; a smaller putative folding intermediate is also detected. This implies that wildtype LepB folds post-translationally in vivo, and shows that FPA can be used to study both co- and post-translational protein folding in the periplasm.

This volume of the Annual Review of Biochemistry contains three reviews on membrane channel proteins: the first by Szczot et al., titled The Form and Function of PIEZO2; the second by Ruprecht & Kunji, titled Structural Mechanism of Transport of Mitochondrial Carriers; and the third by McIlwain et al., titled Membrane Exporters of Fluoride Ion. These reviews provide nice illustrations of just how far evolution has been able to play with the basic helix-bundle architecture of integral membrane proteins to produce membrane channels and transporters of widely different functions.

Proteins commonly fold co-translationally at the ribosome, while the nascent chain emerges from the ribosomal exit tunnel. Protein domains that are sufficiently small can even fold while still located inside the tunnel. However, the effect of the tunnel on the folding dynamics of these domains is not well understood. Here, we combine optical tweezers with single-molecule FRET and molecular dynamics simulations to investigate folding of the small zinc-finger domain ADR1a inside and at the vestibule of the ribosomal tunnel. The tunnel is found to accelerate folding and stabilize the folded state, reminiscent of the effects of chaperonins. However, a simple mechanism involving stabilization by confinement does not explain the results. Instead, it appears that electrostatic interactions between the protein and ribosome contribute to the observed folding acceleration and stabilization of ADR1a. Wruck et al. investigate the folding of the small zinc-finger domain ADR1a inside and at the vestibule of the ribosomal tunnel, using optical tweezers, single-molecule FRET, and molecular dynamics simulations. They find that the ribosomal tunnel accelerates folding while stabilizing the folded state like chaperonins. This study provides insights into the role of the ribosomal tunnel in the folding dynamics of nascent polypeptides.

We follow the cotranslational biosynthesis of three multispanning Escherichia coli inner membrane proteins in vivo using high-resolution force profile analysis. The force profiles show that the nascent chain is subjected to rapidly varying pulling forces during translation and reveal unexpected complexities in the membrane integration process. We find that an N-terminal cytoplasmic domain can fold in the ribosome exit tunnel before membrane integration starts, that charged residues and membrane-interacting segments such as re-entrant loops and surface helices flanking a transmembrane helix (TMH) can advance or delay membrane integration, and that point mutations in an upstream TMH can affect the pulling forces generated by downstream TMHs in a highly position-dependent manner, suggestive of residue-specific interactions between TMHs during the integration process. Our results support the 'sliding' model of translocon-mediated membrane protein integration, in which hydrophobic segments are continually exposed to the lipid bilayer during their passage through the SecYEG translocon.

Proteins commonly fold co-translationally at the ribosome, while the nascent chain emerges from the ribosomal exit tunnel. Protein domains that are sufficiently small can even fold while still located inside the tunnel. However, the effect of the tunnel on the folding dynamics of these domains is not well understood. Here, we combine optical tweezers with single-molecule FRET and molecular dynamics simulations to investigate folding of the small zinc-finger domain ADR1a inside and at the vestibule of the ribosomal tunnel. The tunnel is found to accelerate folding and stabilize the folded state, reminiscent of the effects of chaperonins. However, a simple mechanism involving stabilization by confinement does not explain the results. Instead, it appears that electrostatic interactions between the protein and ribosome contribute to the observed folding acceleration and stabilization of ADR1a., BN/Sander Tans Lab

Cotranslational protein folding studies using Force Profile Analysis, a method where the SecM translational arrest peptide is used to detect folding-induced forces acting on the nascent polypeptide, have so far been limited mainly to small domains of cytosolic proteins that fold in close proximity to the translating ribosome. In this study, we investigate the cotranslational folding of the periplasmic, disulfide bond-containing Escherichia coli protein alkaline phosphatase (PhoA) in a wild-type strain background and a strain background devoid of the periplasmic thiol: disulfide interchange protein DsbA. We find that folding-induced forces can be transmitted via the nascent chain from the periplasm to the polypeptide transferase center in the ribosome, a distance of similar to 160 angstrom, and that PhoA appears to fold cotranslationally via at least two disulfide-stabilized folding intermediates. Thus, Force Profile Analysis can be used to study cotranslational folding of proteins in an extra-cytosolic compartment, like the periplasm.

The malarial parasite Plasmodium exports its own proteins to the cell surfaces of red blood cells (RBCs) during infection. Examples of exported proteins include members of the repetitive interspersed family (RIFIN) and subtelomeric variable open reading frame (STEVOR) family of proteins from Plasmodium falciparum. The presence of these parasite-derived proteins on surfaces of infected RBCs triggers the adhesion of infected cells to uninfected cells (rosetting) and to the vascular endothelium potentially obstructing blood flow. While there is a fair amount of information on the localization of these proteins on the cell surfaces of RBCs, less is known about how they can be exported to the membrane and the topologies they can adopt during the process. The first step of export is plausibly the cotranslational insertion of proteins into the endoplasmic reticulum (ER) of the parasite, and here, we investigate the insertion of three RIFIN and two STEVOR proteins into the ER membrane. We employ a well-established experimental system that uses N-linked glycosylation of sites within the protein as a measure to assess the extent of membrane insertion and the topology it assumes when inserted into the ER membrane. Our results indicate that for all the proteins tested, transmembranes (TMs) 1 and 3 integrate into the membrane, so that the protein assumes an overall topology of Ncyt-Ccyt. We also show that the segment predicted to be TM2 for each of the proteins likely does not reside in the membrane, but is translocated to the lumen.

The TIM23 complex mediates membrane insertion of presequence-containing mitochondrial proteins via a stop-transfer mechanism. Stop-transfer signals consist of hydrophobic transmembrane segments and flanking charges. Mgr2 functions as a lateral gatekeeper of the TIM23 complex. However, it remains elusive which features of stop-transfer signals are discriminated by Mgr2. To determine the effects of Mgr2 on the TIM23-mediated stop-transfer pathway, we measured membrane insertion of model transmembrane segments of varied hydrophobicity and flanking charges in Mgr2-deletion or -overexpression yeast strains. We found that upon deletion of Mgr2, the threshold hydrophobicity for membrane insertion, as well as the requirement for matrix-facing positive charges, is reduced. These results imply that the Mgr2-mediated gatekeeper function is important for controlling membrane sorting of marginal stop-transfer signals.

Signal peptides (SPs) are short amino acid sequences in the amino terminus of many newly synthesized proteins that target proteins into, or across, membranes. Bioinformatic tools can predict SPs from amino acid sequences, but most cannot distinguish between various types of signal peptides. We present a deep neural network-based approach that improves SP prediction across all domains of life and distinguishes between three types of prokaryotic SPs.

In bioinformatics, machine learning methods have been used to predict features embedded in the sequences. In contrast to what is generally assumed, machine learning approaches can also provide new insights into the underlying biology. Here, we demonstrate this by presenting TargetP 2.0, a novel state-of-the-art method to identify N-terminal sorting signals, which direct proteins to the secretory pathway, mitochondria, and chloroplasts or other plastids. By examining the strongest signals from the attention layer in the network, we find that the second residue in the protein, that is, the one following the initial methionine, has a strong influence on the classification. We observe that two-thirds of chloroplast and thylakoid transit peptides have an alanine in position 2, compared with 20% in other plant proteins. We also note that in fungi and single-celled eukaryotes, less than 30% of the targeting peptides have an amino acid that allows the removal of the N-terminal methionine compared with 60% for the proteins without targeting peptide. The importance of this feature for predictions has not been highlighted before., QC 20191126

In bioinformatics, machine learning methods have been used to predict features embedded in the sequences. In contrast to what is generally assumed, machine learning approaches can also provide new insights into the underlying biology. Here, we demonstrate this by presenting TargetP 2.0, a novel state-of-the-art method to identify N-terminal sorting signals, which direct proteins to the secretory pathway, mitochondria, and chloroplasts or other plastids. By examining the strongest signals from the attention layer in the network, we find that the second residue in the protein, that is, the one following the initial methionine, has a strong influence on the classification. We observe that two-thirds of chloroplast and thylakoid transit peptides have an alanine in position 2, compared with 20% in other plant proteins. We also note that in fungi and single-celled eukaryotes, less than 30% of the targeting peptides have an amino acid that allows the removal of the N-terminal methionine compared with 60% for the proteins without targeting peptide. The importance of this feature for predictions has not been highlighted before.

We have characterized the cotranslational folding of two small protein domains of different folds-the alpha-helical N-terminal domain of HemK and the beta-rich FLN5 filamin domain-by measuring the force that the folding protein exerts on the nascent chain when located in different parts of the ribosome exit tunnel (force-profile analysis, or FPA), allowing us to compare FPA to three other techniques currently used to study cotranslational folding: real-time FRET, photo induced electron transfer, and NMR. We find that FPA identifies the same cotranslational folding transitions as do the other methods, and that these techniques therefore reflect the same basic process of cotranslational folding in similar ways.

Helical membrane proteins are typically assumed to attain stable transmembrane topologies immediately upon co-translational membrane insertion. Here we show that unassembled monomers of the small multidrug resistance (SMR) family exist in a dynamic equilibrium where the N-terminal transmembrane helix flips in and out of the membrane, with rates that depend on dimerization and the polypeptide sequence. Thus, membrane topology can display rapid dynamics in vivo and can be regulated by post-translational assembly.

XBP1u, a central component of the unfolded protein response (UPR), is a mammalian protein containing a functionally critical translational arrest peptide (AP). Here, we present a 3 angstrom cryo-EM structure of the stalled human XBP1u AP. It forms a unique turn in the ribosomal exit tunnel proximal to the peptidyl transferase center where it causes a subtle distortion, thereby explaining the temporary translational arrest induced by XBP1u. During ribosomal pausing the hydrophobic region 2 (HR2) of XBP1u is recognized by SRP, but fails to efficiently gate the Sec61 translocon. An exhaustive mutagenesis scan of the XBP1u AP revealed that only 8 out of 20 mutagenized positions are optimal; in the remaining 12 positions, we identify 55 different mutations increase the level of translational arrest. Thus, the wildtype XBP1u AP induces only an intermediate level of translational arrest, allowing efficient targeting by SRP without activating the Sec61 channel.

Signal recognition particle (SRP) recognizes signal sequences of secretory proteins and targets them to the endoplasmic reticulum membrane for translocation. Many human diseases are connected with defects in signal sequences. The current dogma states that the molecular basis of the disease-associated mutations in the secretory proteins is connected with defects in their transport. Here, we demonstrate for several secretory proteins with disease-associated mutations that the molecular mechanism is different from the dogma. Positively charged or helix-breaking mutations in the signal sequence hydrophobic core prevent synthesis of the aberrant proteins and lead to degradation of their mRNAs. The degree of mRNA depletion depends on the location and severity of the mutation in the signal sequence and correlates with inhibition of SRP interaction. Thus, SRP protects secretory protein mRNAs from degradation. The data demonstrate that if disease-associated mutations obstruct SRP interaction, they lead to silencing of the mutated protein expression.

Ever since the signal hypothesis was proposed in 1971, the exact nature of signal peptides has been a focus point of research. The prediction of signal peptides and protein subcellular location from amino acid sequences has been an important problem in bioinformatics since the dawn of this research field, involving many statistical and machine learning technologies. In this review, we provide a historical account of how position-weight matrices, artificial neural networks, hidden Markov models, support vector machines and, lately, deep learning techniques have been used in the attempts to predict where proteins go. Because the secretory pathway was the first one to be studied both experimentally and through bioinformatics, our main focus is on the historical development of prediction methods for signal peptides that target proteins for secretion; prediction methods to identify targeting signals for other cellular compartments are treated in less detail.

Signal peptides (SPs) are short amino acid sequences in the amino terminus of many newly synthesized proteins that target proteins into, or across, membranes. Bioinformatic tools can predict SPs from amino acid sequences, but most cannot distinguish between various types of signal peptides. We present a deep neural network-based approach that improves SP prediction across all domains of life and distinguishes between three types of prokaryotic SPs.

Astrotactin 1 (Astn1) and Astn2 are membrane proteins that function in glial-guided migration, receptor trafficking, and synaptic plasticity in the brain as well as in planar polarity pathways in the skin. Here we used glycosylation mapping and protease protection approaches to map the topologies of mouse Astn1 and Astn2 in rough microsomal membranes and found that Astn2 has a cleaved N-terminal signal peptide, an N-terminal domain located in the lumen of the rough microsomal membranes (topologically equivalent to the extracellular surface in cells), two transmembrane helices, and a large C-terminal lumenal domain. We also found that Astn1 has the same topology as Astn2, but we did not observe any evidence of signal peptide cleavage in Astn1. Both Astn1 and Astn2 mature through endoproteolytic cleavage in the second transmembrane helix; importantly, we identified the endoprotease responsible for the maturation of Astn1 and Astn2 as the endoplasmic reticulum signal peptidase. Differences in the degree of Astn1 and Astn2 maturation possibly contribute to the higher levels of the C-terminal domain of Astn1 detected on neuronal membranes of the central nervous system. These differences may also explain the distinct cellular functions of Astn1 and Astn2, such as in membrane adhesion, receptor trafficking, and planar polarity signaling.

Signal peptides (SPs) are short amino acid sequences in the amino terminus of many newly synthesized proteins that target proteins into, or across, membranes. Bioinformatic tools can predict SPs from amino acid sequences, but most cannot distinguish between various types of signal peptides. We present a deep neural network-based approach that improves SP prediction across all domains of life and distinguishes between three types of prokaryotic SPs.

In bioinformatics, machine learning methods have been used to predict features embedded in the sequences. In contrast to what is generally assumed, machine learning approaches can also provide new insights into the underlying biology. Here, we demonstrate this by presenting TargetP 2.0, a novel state-of-the-art method to identify N-terminal sorting signals, which direct proteins to the secretory pathway, mitochondria, and chloroplasts or other plastids. By examining the strongest signals from the attention layer in the network, we find that the second residue in the protein, that is, the one following the initial methionine, has a strong influence on the classification. We observe that two-thirds of chloroplast and thylakoid transit peptides have an alanine in position 2, compared with 20% in other plant proteins. We also note that in fungi and single-celled eukaryotes, less than 30% of the targeting peptides have an amino acid that allows the removal of the N-terminal methionine compared with 60% for the proteins without targeting peptide. The importance of this feature for predictions has not been highlighted before.

Ever since the signal hypothesis was proposed in 1971, the exact nature of signal peptides has been a focus point of research. The prediction of signal peptides and protein subcellular location from amino acid sequences has been an important problem in bioinformatics since the dawn of this research field, involving many statistical and machine learning technologies. In this review, we provide a historical account of how position-weight matrices, artificial neural networks, hidden Markov models, support vector machines and, lately, deep learning techniques have been used in the attempts to predict where proteins go. Because the secretory pathway was the first one to be studied both experimentally and through bioinformatics, our main focus is on the historical development of prediction methods for signal peptides that target proteins for secretion; prediction methods to identify targeting signals for other cellular compartments are treated in less detail.

Signal peptides (SPs) are short amino acid sequences in the amino terminus of many newly synthesized proteins that target proteins into, or across, membranes. Bioinformatic tools can predict SPs from amino acid sequences, but most cannot distinguish between various types of signal peptides. We present a deep neural network-based approach that improves SP prediction across all domains of life and distinguishes between three types of prokaryotic SPs.

In bioinformatics, machine learning methods have been used to predict features embedded in the sequences. In contrast to what is generally assumed, machine learning approaches can also provide new insights into the underlying biology. Here, we demonstrate this by presenting TargetP 2.0, a novel state-of-the-art method to identify N-terminal sorting signals, which direct proteins to the secretory pathway, mitochondria, and chloroplasts or other plastids. By examining the strongest signals from the attention layer in the network, we find that the second residue in the protein, that is, the one following the initial methionine, has a strong influence on the classification. We observe that two-thirds of chloroplast and thylakoid transit peptides have an alanine in position 2, compared with 20% in other plant proteins. We also note that in fungi and single-celled eukaryotes, less than 30% of the targeting peptides have an amino acid that allows the removal of the N-terminal methionine compared with 60% for the proteins without targeting peptide. The importance of this feature for predictions has not been highlighted before.

In bioinformatics, machine learning methods have been used to predict features embedded in the sequences. In contrast to what is generally assumed, machine learning approaches can also provide new insights into the underlying biology. Here, we demonstrate this by presenting TargetP 2.0, a novel state-of-the-art method to identify N-terminal sorting signals, which direct proteins to the secretory pathway, mitochondria, and chloroplasts or other plastids. By examining the strongest signals from the attention layer in the network, we find that the second residue in the protein, that is, the one following the initial methionine, has a strong influence on the classification. We observe that two-thirds of chloroplast and thylakoid transit peptides have an alanine in position 2, compared with 20% in other plant proteins. We also note that in fungi and single-celled eukaryotes, less than 30% of the targeting peptides have an amino acid that allows the removal of the N-terminal methionine compared with 60% for the proteins without targeting peptide. The importance of this feature for predictions has not been highlighted before.

Ever since the signal hypothesis was proposed in 1971, the exact nature of signal peptides has been a focus point of research. The prediction of signal peptides and protein subcellular location from amino acid sequences has been an important problem in bioinformatics since the dawn of this research field, involving many statistical and machine learning technologies. In this review, we provide a historical account of how position-weight matrices, artificial neural networks, hidden Markov models, support vector machines and, lately, deep learning techniques have been used in the attempts to predict where proteins go. Because the secretory pathway was the first one to be studied both experimentally and through bioinformatics, our main focus is on the historical development of prediction methods for signal peptides that target proteins for secretion; prediction methods to identify targeting signals for other cellular compartments are treated in less detail.

Signal peptides (SPs) are short amino acid sequences in the amino terminus of many newly synthesized proteins that target proteins into, or across, membranes. Bioinformatic tools can predict SPs from amino acid sequences, but most cannot distinguish between various types of signal peptides. We present a deep neural network-based approach that improves SP prediction across all domains of life and distinguishes between three types of prokaryotic SPs.

Ever since the signal hypothesis was proposed in 1971, the exact nature of signal peptides has been a focus point of research. The prediction of signal peptides and protein subcellular location from amino acid sequences has been an important problem in bioinformatics since the dawn of this research field, involving many statistical and machine learning technologies. In this review, we provide a historical account of how position-weight matrices, artificial neural networks, hidden Markov models, support vector machines and, lately, deep learning techniques have been used in the attempts to predict where proteins go. Because the secretory pathway was the first one to be studied both experimentally and through bioinformatics, our main focus is on the historical development of prediction methods for signal peptides that target proteins for secretion; prediction methods to identify targeting signals for other cellular compartments are treated in less detail.

This article introduces the Protein Evolution and Design theme of the Annual Review of Biochemistry Volume 87.

My scientific career has taken me from chemistry, via theoretical physics and bioinformatics, to molecular biology and even structural biology. Along the way, serendipity led me to work on problems such as the identification of signal peptides that direct protein trafficking, membrane protein biogenesis, and cotranslational protein folding. I've had some great collaborations that came about because of a stray conversation or from following up on an interesting paper. And I've had the good fortune to be asked to sit on the Nobel Committee for Chemistry, where I am constantly reminded of the amazing pace and often intricate history of scientific discovery. Could I have planned this? No way! I just went with the flow …

The E. coli ribosome exit tunnel can accommodate small folded proteins, while larger ones fold outside. It remains unclear, however, to what extent the geometry of the tunnel influences protein folding. Here, using E. coli ribosomes with deletions in loops in proteins uL23 and uL24 that protrude into the tunnel, we investigate how tunnel geometry determines where proteins of different sizes fold. We find that a 29-residue zinc-finger domain normally folding close to the uL23 loop folds deeper in the tunnel in uL23 Delta loop ribosomes, while two similar to 100 residue proteins normally folding close to the uL24 loop near the tunnel exit port fold at deeper locations in uL24 Delta loop ribosomes, in good agreement with results obtained by coarse-grained molecular dynamics simulations. This supports the idea that cotranslational folding commences once a protein domain reaches a location in the exit tunnel where there is sufficient space to house the folded structure.

Integral membrane proteins are assembled into the ER membrane via a continuous ribosome-translocon channel. The hydrophobicity and thickness of the core of the membrane bilayer leads to the expectation that transmembrane (TM) segments minimize the cost of harbouring polar polypeptide backbones by adopting a regular pattern of hydrogen bonds to form a-helices before integration. Co-translational folding of nascent chains into an a-helical conformation in the ribosomal tunnel has been demonstrated previously, but the features governing this folding are not well understood. In particular, little is known about what features influence the propensity to acquire a-helical structure in the ribosome. Using in vitro translation of truncated nascent chains trapped within the ribosome tunnel and molecular dynamics simulations, we show that folding in the ribosome is attained for TM helices but not for soluble helices, presumably facilitating SRP (signal recognition particle) recognition and/or a favourable conformation for membrane integration upon translocon entry.

During ribosomal translation, nascent polypeptide chains (NCs) undergo a variety of physical processes that determine their fate in the cell. This study utilizes a combination of arrest peptide experiments and coarse-grained molecular dynamics to measure and elucidate the molecular origins of forces that are exerted on NCs during cotranslational membrane insertion and translocation via the Sec translocon. The approach enables deconvolution of force contributions from NC-translocon and NC-ribosome interactions, membrane partitioning, and electrostatic coupling to the membrane potential. In particular, we show that forces due to NC-lipid interactions provide a readout of conformational changes in the Sec translocon, demonstrating that lateral gate opening only occurs when a sufficiently hydrophobic segment of NC residues reaches the translocon. The combination of experiment and theory introduced here provides a detailed picture of the molecular interactions and conformational changes during ribosomal translation that govern protein biogenesis.

Proteins that fold cotranslationally may do so in a restricted configurational space, due to the volume occupied by the ribosome. How does this environment, coupled with the close proximity of the ribosome, affect the folding pathway of a protein? Previous studies have shown that the cotranslational folding process for many proteins, including small, single domains, is directly affected by the ribosome. Here, we investigate the cotranslational folding of an all-beta Ig domain, titin I27. Using an arrest peptide-based assay and structural studies by cryo-EM, we show that I27 folds in the mouth of the ribosome exit tunnel. Simulations that use a kinetic model for the force dependence of escape from arrest accurately predict the fraction of folded protein as a function of length. We used these simulations to probe the folding pathway on and off the ribosome. Our simulations-which also reproduce experiments on mutant forms of I27-show that I27 folds, while still sequestered in the mouth of the ribosome exit tunnel, by essentially the same pathway as free I27, with only subtle shifts of critical contacts from the C to the N terminus.

Large clostridial toxins (LCTs) are a family of homologous proteins toxins that are directly responsible for the symptoms associated with a number of clostridial infections that cause disease in humans and in other animals. LCTs damage tissues by delivering a glucosyltransferase domain, which inactivates small GTPases, across the endosomal membrane and into the cytosol of target cells. Elucidating the mechanism of translocation for LCTs has been hampered by difficulties associated with identifying marginally hydrophobic segments that insert into the bounding membrane to form the translocation pore. Here, we directly measured the membrane-insertion partitioning propensity for segments spanning the putative pore-forming region using a translocon-mediated insertion assay and synthetic peptides. We identified membrane-inserting segments, as well as a conserved and functionally important negatively charged residue that requires protonation for efficient membrane insertion. We provide a model of the LCT pore, which provides insights into translocation for this enigmatic family of a-helical translocases.

During the last five decades, studies of protein folding in dilute buffer solutions have produced a rich picture of this complex process. In the cell, however, proteins can start to fold while still attached to the ribosome (cotranslational folding) and it is not yet clear how the ribosome affects the folding of protein domains of different sizes, thermodynamic stabilities, and net charges. Here, by using arrest peptides as force sensors and on-ribosome pulse proteolysis, we provide a comprehensive picture of how the distance from the peptidyl transferase center in the ribosome at which proteins fold correlates with protein size. Moreover, an analysis of a large collection of mutants of the Escherichia coli ribosomal protein 56 shows that the force exerted on the nascent chain by protein folding varies linearly with the thermodynamic stability of the folded state, and that the ribosome environment disfavors folding of domains of high net-negative charge.

It is becoming increasingly clear that many proteins start to fold cotranslationally before the entire polypeptide chain has been synthesized on the ribosome. One class of proteins that a priori would seem particularly prone to cotranslational folding is repeat proteins, that is, proteins that are built from an array of nearly identical sequence repeats. However, while the folding of repeat proteins has been studied extensively in vitro with purified proteins, only a handful of studies have addressed the issue of cotranslational folding of repeat proteins. Here, we have determined the structure and studied the cotranslational folding of a beta-helix pentarepeat protein from the human pathogen Clostridium botulinum a homolog of the fluoroquinolone resistance protein MfpA-using an assay in which the SecM translational arrest peptide serves as a force sensor to detect folding events. We find that cotranslational folding of a segment corresponding to the first four of the eight beta-helix coils in the protein produces enough force to release ribosome stalling and that folding starts when this unit is similar to 35 residues away from the P-site, near the distal end of the ribosome exit tunnel. An additional folding transition is seen when the whole PENT moiety emerges from the exit tunnel. The early cotranslational formation of a folded unit may be important to avoid misfolding events in vivo and may reflect the minimal size of a stable beta-helix since it is structurally homologous to the smallest known beta-helix protein, a four-coil protein that is stable in solution.

Like all other secretory proteins, the HIV-1 envelope glycoprotein gp160, is targeted to the endoplasmic reticulum (ER) by its signal peptide during synthesis. Proper gp160 folding in the ER requires core glycosylation, disulfide-bond formation and proline isomerization. Signal-peptide cleavage occurs only late after gp160 chain termination and is dependent on folding of the soluble subunit gp120 to a near-native conformation. We here detail the mechanism by which co-translational signal-peptide cleavage is prevented. Conserved residues from the signal peptide and residues downstream of the canonical cleavage site form an extended alpha-helix in the ER membrane that covers the cleavage site, thus preventing cleavage. A point mutation in the signal peptide breaks the alpha helix allowing co-translational cleavage. We demonstrate that postponed cleavage of gp160 enhances functional folding of the molecule. The change to early cleavage results in decreased viral fitness compared to wild-type HIV.

The topologies of alpha-helical membrane proteins are generally thought to be determined during their cotranslational insertion into the membrane. It is typically assumed that membrane topologies remain static after this process has ended. Recent findings, however, question this static view by suggesting that some parts of, or even the whole protein, can reorient in the membrane on a biologically relevant time scale. Here, we focus on antiparallel homo- or heterodimeric small multidrug resistance proteins and examine whether the individual monomers can undergo reversible topological inversion (flip flop) in the membrane until they are trapped in a fixed orientation by dimerization. By perturbing dimerization using various means, we show that the membrane orientation of a monomer is unaffected by the presence or absence of its dimerization partner. Thus, membrane-inserted monomers attain their final orientations independently of dimerization, suggesting that wholesale topological inversion is an unlikely event in vivo.