Functional diversity and molecular interactions of small membrane proteins in bacteria

Abstract

Bacteria constantly adapt to changing environmental conditions through diverse processes that involve numerous regulator and effector proteins. In this regard, small proteins play a significant role in promoting stress adaptation in bacteria. Although they were largely overlooked in early genome annotations, recent technological advances and a growing recognition of their significance have paved the way for the increasing identification and characterization of this intriguing class of proteins. Many small proteins contain a transmembrane domain and are integral to the cytoplasmic membrane. Others interact with and modulate membrane protein complexes. In this review, we focus on the current knowledge of these small membrane proteins, with an emphasis on their interactions, membrane insertion pathways, and toxicity.

Membrane insertion and functional diversity of bacterial small membrane proteins.

Introduction

Small proteins in bacteria are typically defined as proteins of less than 100 amino acids, which are synthesized from small mRNAs and do not derive from larger proteins via proteolysis (Gray et al. 2022). A growing number of small bacterial proteins have been identified and validated through ribosome profiling and specialized mass spectrometry approaches (Cassidy et al. 2021, Ahrens et al. 2022, Meier-Credo et al. 2022, Vazquez-Laslop et al. 2022) in conjunction with customized bioinformatics tools (Gelhausen et al. 2021, Fesenko et al. 2025). In recent years, our understanding of prokaryotic small proteomes has expanded significantly, encompassing organisms that extend far beyond the traditional laboratory strain of Escherichia coli (Hemm et al. 2020). Small proteins have been systematically identified in diverse prokaryotes, including the plant symbiont Sinorhizobium meliloti (Hadjeras et al. 2023b) human pathogens such as Salmonella enterica Typhimurium (Venturini et al. 2020), Campylobacter jejuni (Froschauer et al. 2025), Streptococcus pneumoniae (Laczkovich et al. 2022), and Mycobacterium tuberculosis (Smith et al. 2022), as well as key archaeal species such as Methanosarcina mazei (Tufail et al. 2024) and Haloferax volcanii (Hadjeras et al. 2023a). However, archaeal small proteins are beyond the scope of this review and are not further covered here. Although several small proteins have been functionally characterized in bacteria, most newly discovered small proteins have unknown functions and their characterization remains a rate-limiting step in defining their importance for the bacterial physiology. This is particularly evident for small membrane proteins (SMPs), which constitute a considerable proportion of the small proteome. SMPs typically comprise a single α-helical domain characterized by a high proportion of hydrophobic regions. The overall hydrophobicity of SMPs appears to be more important for functionality than the individual amino acid residues, resulting in functional homologues that do not share exactly the same sequence (Steinberg and Koch 2021, Yadavalli and Yuan 2022). The hydrophobic nature of SMPs makes them challenging to detect by mass spectrometry, complicates their chemical synthesis, and renders them mainly unsuitable for the generation of specific antibodies. In particular, small membrane toxins cannot be overexpressed for downstream analyses due to their inherent toxicity. The small size and high hydrophobicity of small membrane proteins also poses a particular challenge for the cellular machineries required for their membrane targeting and insertion (Steinberg and Koch 2021). Nevertheless, significant progress has been made in understanding this class of proteins, leading to the identification of two strategies that support their proper insertion into the lipid bilayer and in defining their role in controlling membrane integrity, and regulating their membrane protein targets. Additionally, a detection method with higher sensitivity has been established to identify endogenous small proteins (Simoens et al. 2025). Furthermore, a novel cell-free approach has been developed to synthesize SMPs in vitro. This approach is compatible with high-throughput techniques for identifying SMP targets, potentially accelerating future functional discovery (Bhattacharya et al. 2020, Jiang et al. 2024).

The role of bacterial small proteins and their interactions with larger proteins have been recently covered in a comprehensive review by the Storz lab (Burton et al. 2024). In this review, we focus primarily on bacterial SMPs and recent advances in understanding the function of this subclass of small proteins. First, we will summarize how SMPs regulate their membrane targets, particularly kinases, respiratory complexes, and photosystems. Next, we will review new insights into the membrane insertion pathways of SMPs. Finally, we will discuss how small membrane toxins from type I toxin-antitoxin systems affect bacterial physiology.

Strategies for SMPs to regulate membrane protein targets

Membrane proteins constitute nearly one-third of the cellular proteome and perform a wide range of essential functions, such as nutrient transport, signaling, energy generation, cell division, extracellular matrix synthesis, and virulence. SMPs have emerged as modulators of these processes by directly targeting larger membrane proteins. For example, the SMP MgtS (31 aa) in E. coli and MgtR (30 aa) in S. enterica interact with magnesium transporters. MgtS inhibits, whereas MgtR promotes, FtsH-dependent degradation of their targeted transporters, thereby regulating magnesium homeostasis (Wang et al. 2017a, Choi et al. 2012, Jean-Francois et al. 2014, Yin et al. 2019). SMPs can also function as elements within large membrane complexes. For instance, in E. coli KdpF (29 aa) associates with the potassium transporter KdpABC, stabilizing the complex (Gassel et al. 1999, Huang et al. 2017, Stock et al. 2018). AcrZ (49 aa) binds the AcrAB-TolC multidrug efflux pump and modulates its substrate specificity (Wang et al. 2017, Hobbs et al. 2012, Du et al. 2020). Recent investigations of SMPs acting on membrane-associated kinases, respiratory complexes, and photosystems highlight the diverse cellular processes subject to small protein regulation, as summarized below.

Small proteins targeting kinases

MgrB (47 aa) is one of the most-studied SMPs that modulates the activity of its target sensor kinase. Recent investigations have provided deeper insights into its mechanism of action and its role in signaling. MgrB is widespread and highly conserved in E. coli, Salmonella, Klebsiella pneumoniae, and related species (Lippa and Goulian 2009). Like many SMPs, MgrB has a single transmembrane helix. It directly binds to the membrane-localized sensor kinase PhoQ, a master regulator of the bacterial virulence program (reviewed in (Groisman et al. 2021)). Kinase-active PhoQ promotes the expression of the phoPQ operon, mgrB, and genes involved in the stress response and virulence program. MgrB inhibits the kinase and stimulates the phosphatase activity of PhoQ, thus giving negative feedback and stabilizing the PhoQ/PhoP two-component signaling system (Lippa and Goulian 2009, Salazar et al. 2016, Yadavalli et al. 2020, Jiang et al. 2024). Deletion or mutation of the mgrB gene leads to increased PhoQ kinase activity and enhanced drug resistance via lipid A modifications (Poirel et al. 2015). This is particularly problematic for K. pneumoniae infections, because in the absence of a functional MgrB, the pathogen is resistant to colistin, the last-resort antibiotic used worldwide in clinical settings (Cannatelli et al. 2014, Khoshbayan et al. 2024). Recent mechanistic studies have revealed that MgrB interacts extensively with the periplasmic and the TM domains of PhoQ, triggering conformational changes (Fig. 1A). These changes include the movement of the cytosolic ATP-binding domain away from the autophosphorylation site, thus reducing PhoQ kinase activity (Jiang et al. 2023, Lazaridi et al. 2024). In the presence of cationic antimicrobial peptides, MgrB was found to dissociate from PhoQ and to de-repress PhoQ activity, mediating PhoQ to detect antimicrobial peptides, a crucial component of human innate immunity (Jiang et al. 2023). MgrB also acts as an entry point of redox signals via its two highly conserved periplasmic cysteine residues (Lippa and Goulian 2012). When the periplasmic oxidizing environment is disturbed, the disruption of the functionally important disulfide bond between the two cysteines leads to reduced MgrB function and the derepression of PhoQ activity (Lippa and Goulian 2009, Lippa and Goulian 2012). As a negative regulator, MgrB broadens the range of external stimuli sensed by the PhoQ sensor kinase, facilitating the integration of diverse signals and enabling appropriate cellular responses that enhance bacterial fitness and survival under adverse conditions (Fig. 1B).

Figure 1.

Bacterial small membrane proteins that regulate PhoQ kinase activity. (A) The small protein MgrB interacts with the periplasmic and the transmembrane domains of the PhoQ sensor kinase. The predicted complex structure has been verified by chemical crosslinking (Jiang et al. 2023). MgrB inhibits PhoQ kinase activity, while SafA activates PhoQ in E. coli. In S. enterica, UgtL has been shown to activate PhoQ, while UgtS antagonizes this activation. The structural information regarding PhoQ in complex with SafA, UgtS, and UgtL remain elusive. (B) MgrB serves as an entry point for additional inputs into the PhoQ/PhoP signaling pathway.

Multiple SMPs have been found to bind directly to the PhoQ sensor kinase and to regulate its activity. In E. coli, SafA (65 aa) activates the PhoQ kinase activity primarily through interactions with an internal cavity located in PhoQ’s periplasmic domain (Eguchi et al. 2007, Eguchi et al. 2012, Ishii et al. 2013, Yoshitani et al. 2019). In Salmonella enterica, UgtL (132 aa) activates PhoQ, while UgtS (34 aa), expressed from the same bicistronic operon, antagonizes the activation by binding to both UgtL and PhoQ (Choi and Groisman 2017, Salvail et al. 2022). In contrast to the wide distribution of MgrB, SafA is found primarily in E. coli (Eguchi et al. 2007) and UgtL and UgtS are only found in S. enterica and a subset of serovars in S. enterica, respectively (Choi and Groisman 2017, Salvail et al. 2022). Structural predictions with AlphaFold 3 (Abramson et al. 2024) support that all four PhoQ-regulating SMPs have alpha-helical elements in their structures, specifically in the transmembrane region. However, the binding sites of UgtS and UgtL on PhoQ and their mechanisms of action remain unknown. Comprehensive quantitative characterization of all four PhoQ-regulating small proteins—including their in vivo abundances and binding affinities to PhoQ—are needed to fully understand how the small protein/PhoQ interactions are integrated into the regulation of PhoQ activity.

PhoR is another sensor kinase recently found to be the target of small proteins. It is regulated by the small DUF1127 proteins in the plant pathogen Agrobacterium tumefaciens and E. coli (Kraus et al. 2020) (Fig. 2). The PhoR/PhoB two-component system regulates the expression of the high-affinity phosphate transporter PstSCAB and is essential for phosphate homeostasis (reviewed in (Gardner and McCleary 2019)). DUF1127 proteins (47–54 aa) are arginine-rich small proteins that bind to the membrane-localized PhoR (Remme et al. 2025). Deletion of all three copies of the corresponding genes in A. tumefaciens leads to increased expression of the PhoB regulon, upregulation of phosphate import, polyphosphate accumulation and impaired growth, suggesting that these DUF1127 proteins inhibit PhoR activity in A. tumefaciens. Homologs of the DUF1127 proteins from Sinorhizobium meliloti, Rhodobacter sphaeroides and E. coli (YjiS), but not that from S. enterica, have been found to complement the deletion in A. tumefaciens, indicating some degrees of functional conservation (Remme et al. 2025).

Figure 2.

Bacterial small proteins target membrane-localized kinases. The small DUF1127 proteins (SDP) interact with the PhoR sensor kinase and are involved in phosphate homeostasis regulation in A. tumefaciens (Kraus et al. 2020). The small protein YjiS was co-purified with the membrane fraction and might target the sensor kinase SsrA in S. enterica (Venturini et al. 2025). In E. coli, the SMP YoaI was induced under low magnesium stress. It interacts with the sensor kinase EnvZ and connects the two stress signaling systems, EnvZ/OmpR and PhoR/PhoB (Vellappan et al. 2025). In K. pneumoniae, the small protein RmpD preferably interacts with the octameric tyrosine kinase Wzc (with one monomer highlighted in blue), regulating the oligosaccharide chain length in the capsule (Walker et al. 2020). All small proteins are shown schematically with “+” indicating positive charges. AlphaFold3 predictions are used to show the dimeric PhoQ, SsrA, and EnvZ, which have not been verified experimentally. The cryoEM structure (PDB 7NHR) is used for the Wzc octameric complex. The Walker A and Walker B motifs in one Wzc protomer are in red. The C-terminal tyrosine tail from the neighboring protomer is in magenta.

The DUF1127 domain-containing protein, YjiS (54 aa), in S. enterica Typhimurium was first found to be strongly induced during infection (Venturini et al. 2020). Deletion of yjiS results in a delayed escape of Salmonella from macrophages, as evidenced by an increased fraction of vital bacteria inside mammalian cells (Venturini et al. 2025). Mutagenesis of YjiS reveals that individual arginine residues are not functionally essential and do not affect infection (Venturini et al. 2025). Other experimental results support that YjiS is associated with the bacterial inner membrane, where it is proposed to interact with the membrane-localized sensor kinase, SsrA (Fig. 2). Given the role of the SsrA/SsrB two-component system in regulating virulence in Salmonella, YjiS may contribute to infection regulation through its association with SsrA (Venturini et al. 2025). It is intriguing that the arginine-rich, positively charged YjiS proteins from E. coli appears to interact with PhoR like in A. tumefaciens, despite its sequence similarity to the YjiS in S. enterica. Given the functionally dispensable nature of single arginine residues and the absence of SsrA kinase in E. coli, these findings suggest that the distribution of the charged arginine residues is likely the key determinant of the function of the small DUF-1127-containing proteins. How these small proteins regulate their kinase targets remains an open question.

A recent report revealed the functional role of the small inner membrane protein YoaI, which is transcriptionally regulated by the PhoR/PhoB two-component system in E. coli (Yoshida et al. 2012). Its expression is strongly induced under magnesium limitation, yet this response occurs independently of PhoQ, the canonical sensor kinase for low magnesium (Vellappan et al. 2025). Bacterial two-hybrid and reporter assays support that YoaI interacts with EnvZ (Fig. 2), the sensor kinase for osmotic changes, and stimulates its activity (Vellappan et al. 2025). These findings suggest that YoaI functions as a molecular connector between the PhoR/PhoB and EnvZ/OmpR signaling pathways. The precise mechanism of this regulatory interplay remains to be elucidated.

The small protein RmpD (58 aa) was recently found to confer hypermucoviscosity, a phenotype strongly associated with K. pneumoniae virulence, by modulating polysaccharide chain length (Walker et al. 2020). The rmpD gene locates in the intergenic region between rmpA and rmpC. RmpA is a LuxR-like transcription factor, regulating the expression of rmpADC (Nassif et al. 1989, Cheng et al. 2010, Hsu et al. 2011, Walker et al. 2019). RmpC regulates capsule gene expression (Walker et al. 2019). Overexpression of rmpD and rmpC from a plasmid complements rmpA deletion and restores hypermucoviscosity and capsule synthesis, respectively (Walker et al. 2020). Deletion of rmpD leads to the loss of hypermucoviscosity but has nearly no change in capsule gene expression compared to the wild type, indicating RmpD is mainly involved in the hypermucoviscosity phenotype (Walker et al. 2020). RmpD is conserved in hypervirulent K. pneumoniae strains. It has an N-terminal TM helix and a C-terminal positively charged region with the C-terminus in the cytosol (Fig. 2). Co-purification and two-hybrid assay results support that RmpD interacts with Wzc, an inner membrane tyrosine autokinase. Wzc is proposed to interact with Wzy and Wza—the capsular polysaccharide polymerase and the outer membrane translocon, rescpectively—and is therefore involved in the polymerization and transport of polysaccharide chains from the periplasm to the outer membrane (Whitfield et al. 2020, Yang et al. 2021, Ovchinnikova et al. 2023, Weckener et al. 2023). Wzc has a five-tyrosine cluster in its cytosolic C-terminal end and one additional tyrosine residue (Tyr569) located upstream. Tyr569 is autophosphorylated intra-molecularly, while the five-tyrosine cluster is phosphorylated inter-molecularly (Grangeasse et al. 2002). Wzb, the cognate phosphatase, brings Wzc to an unphosphorylated state, and both of them are indispensable for capsule expression (Wugeditsch et al. 2001). Mutational and structural studies reveal that phosphorylation and dephosphorylation allow Wzc alternate between monomeric and octameric states. The octameric Wzc complex creates a central cavity in the membrane and connects to Wza for polysaccharide transport (Yang et al. 2021, Yang et al. 2025). Pertubation of Wzc phosphorylation status changes capsule polysaccharide chain length by changing the coordination between polysaccharide polymerization and transport (Doublet et al. 2002, Yang et al. 2021). RmpD is found to preferentially interact with the unphosphorylated octameric Wzc, resulting in the production of capsular polysaccharides with higher uniformity and longer average chain length (Doublet et al. 2002, Ovchinnikova et al. 2023). The precise mechanism underlying this effect remains unknown.

Studies on the small DUF1127 proteins, YjiS, YoaI, and RmpD have expanded the repertoire of kinase-regulating small proteins. Similar to MgrB, some function as negative regulators. While positive regulators accelerate signal transduction, negative regulators play equally critical roles by shaping spatial and temporal dynamics via suppressing system fluctuations, limiting the maximum output, and stabilizing the system to an equilibrium state (Becskei and Serrano 2000, Brandman and Meyer 2008). Moreover, negative regulators can provide entry points for additional inputs, as exemplified by MgrB (Fig. 1B) broadening the range of stimuli sensed by the sensor kinase PhoQ (Lippa and Goulian 2012, Jiang et al. 2023). Therefore, it is worthwhile to analyze whether the short DUF1127 proteins enable PhoR to sense intracellular phosphate-related signals, compensating the lack of a conventional periplasmic sensor domain in PhoR. Similarly, a recent report showed that human urine down-regulates rmpD and reduces Klebsiella mucoidy without changing capsule synthesis (Khadka et al. 2023), supporting the notion that negative regulators act as an entry point for additional signals.

Small membrane proteins in complex with respiratory oxidase complexes and photosystems

The physiological importance of SMPs was initially recognized because many protein complexes of the respiratory chain or the bacterial photosystems contained SMPs as accessory subunits (Zickermann et al. 2010, Kraus and Hess 2025). Examples are PetN, which is a 29 amino acid long subunit of the cyanobacterial cytochrome b6f complex (Schneider et al. 2007), or AtpE, which consist of 79 amino acids and forms the C-ring within the E. coli ATP synthase (Frasch et al. 2022). Another example is the cyanobacterial photosystem II, which contains several SMPs as subunits, which are required for electron transfer and stabilization (Zabret et al. 2021).

CioY (34 aa) is a newly validated SMP found to be in complex with the cyanide-insensitive quinol oxidase CioAB in the foodborne pathogen Campylobacter jejuni (Froschauer et al. 2025). CioY was first identified as a potential CydX homolog in a bioinformatics study (Allen et al. 2014). It is distributed in different Campylobacter species and has an α-helical structure predicted to integrate into the membrane. Reciprocal co-immunoprecipitation results support the interaction between CioY and CioA. Structural prediction and comparison suggest that CioY interacts with CioA in a highly comparable manner to the CydX/CydA complex in E. coli (Fig. 3).

Figure 3.

Bacterial small proteins forming membrane protein complexes with respiratory oxidases. The small membrane protein CioY in C. jejuni was proposed to form a complex with the CioAB terminal oxidase based on AlphaFold prediction (not yet verified experimentally) (Froschauer et al. 2025). The predicted complex is highly comparable to the E. coli complex (PDB 6RKO), where structural studies reveal that CydX and CydH form a complex with CydAB (Safarian et al. 2019, Thesseling et al. 2019). Structural comparison indicates that the position of CioY is similar to that of CydX in the complex. In R. capsulatus, the small membrane protein CcoQ is a subunit of the cytochrome cbb₃-type oxidase complex and stabilizes the complex by interacting primarily with the CcoP subunit. The formation of the Cu- and heme-containing catalytic center of CcoN is facilitated by another SMP, CcoS, which also interacts with the Cu chaperone SenC. (Peters et al. 2008, Rauch et al. 2025). An AlphaFold prediction is used to show the CcoNOQP complex, where only the core complex CcoNOP has been experimentally verified (Buschmann et al. 2010, Steimle et al. 2021).

E. coli CydAB represents a cytochrome bd quinol oxidase with high oxygen affinity (Unden et al. 2014). The two subunits form apseudo-symmetric dimer, which binds with two SMPs, CydX and CydH (also called CydY). CydX (37 aa) is encoded by a small open reading frame downstream of the cydAB operon, the same synteny as cioY (VanOrsdel et al. 2013). CydX is widespread in eubacteria based on homology analyses. CydX was co-purified with the CydAB complex and was shown to be essential for oxidase activity in vivo and in vitro by stabilizing the complex or contributing to the stability of the active site di-heme center (VanOrsdel et al. 2013, Hoeser et al. 2014). Consequently, a cydX deletion mutant was found to be more sensitive to β-mercaptoethanol and nitric oxide in S. enterica Typhimurium and showed reduced growth inside macrophages (Duc et al. 2020). β-mercaptoethanol removes copper ions indispensable for the cytochrome bo-type quinol oxidase, which exacerbate the effect of cydX deletion. The heme in cytochrome bd reacts with nitric oxide, directly mediating the resistance (52). In Brucella abortus, the ΔcydX mutant has a phenotype similar to that of cydB mutants including increased sensitivity to hydrogen peroxide and to the respiratory chain inhibitor sodium azide (Sun et al. 2012). Cryo-electron microscopy revealed the second small protein, CydH, in complex with the CydAB core dimer in E. coli (Safarian et al. 2019, Thesseling et al. 2019). CydH interacts with CydA at a site opposite from where CydX binds in the complex, blocking a spacious oxygen entry site previously identified in the Geobacillus thermodenitrificans enzyme (Safarian et al. 2016). The E. coli enzyme has instead a second narrow oxygen-conducting channel, which may confer higher substrate specificity (Safarian et al. 2019, Thesseling et al. 2019). No sequence homolog of CydH has been identified in C. jejuni. This is not unexpected as CioY also shows little sequence similarity to CydX despite sharing structural homology. A similar case is seen in G. thermodenitrificans, where CydS is structurally homologous to CydX but lacks detectable sequence similarity (Safarian et al. 2016). Since protein structure is generally more conserved than sequence, it is likely that additional small protein paralogs with strong structural but weak sequence similarity remain to be discovered.

Two SMPs, CcoQ (58 aa) and CcoS (52 aa) are found to be essential for the active cytochrome cbb₃-type cytochrome oxidase in Rhodobacter capsulatus (Fig. 3) (Peters et al. 2008, Rauch et al. 2025). Both SMPs have a single TM helix with an N-out/C-in topology, which was determined by in vitro protease protection assays (Kulajta et al. 2006, Peters et al. 2008, Rauch et al. 2025). CcoQ serves as subunit of the cbb₃-type cytochrome oxidase and its interaction with the CcoP subunit is essential for the stability of the cbb₃-type cytochrome oxidase (Peters et al. 2008). CcoS acts as an assembly factor for cbb₃-type cytochrome oxidase (Kulajta et al. 2006, Rauch et al. 2025). Mutagenesis studies revealed that the cytosolic C-terminal part of CcoS is functionally indispensable with D33 being the key residue. Cells expressing CcoS D33C variant showed oxidase activity almost as low as the ΔccoS strain. The available data also indicate that CcoS interacts with the oxidase subunits CcoN and CcoP, and facilitates, together with the Cu chaperone SenC, the formation of the heme b-Cu_B binuclear center in the catalytic subunit CcoN (Rauch et al. 2025).

SMPs are not only involved in the assembly of respiratory complexes, but are also important for their regulation in response to different environmental conditions. This is in particular evident in cyanobacteria, which have to adapt to high- and low-light conditions. One example is the widely conserved small amphipathic protein AtpΘ in Synechocystis, which acts as an inhibitor of ATP synthase in the dark and thus prevents it from pumping protons at the expense of ATP hydrolysis (Song et al. 2022). Another example is Slip4 (small light-induced protein of 4 kDa), which is suggested to favor cyclic electron flow at high-light conditions by stabilizing the contact between photosystem I and NDH1 in Synechocystis (Alvarenga-Lucius et al. 2023).

Strategies for membrane insertion of small bacterial membrane proteins

The canonical SRP-pathway for membrane protein insertion in bacteria

The transport of proteins generally requires signal sequences that serve as identification tags and define the final localization of proteins within the cell (von Heijne 1994). For bacterial proteins, these signal sequences are usually located at the N-terminus and have a tripartite structure. The positively charged N-region is followed by a hydrophobic H-region and a polar C-terminal region that contains the signal peptidase cleavage site (Hegde and Bernstein 2006). The hydrophobic H-region is particularly important for the recognition of signal sequences by protein-targeting factors, such as the signal recognition particle (SRP) or SecA (von Heijne 1994, Hegde and Bernstein 2006). These signal sequences are usually 25–30 amino acids long and do not contain particular sequence motifs, although exceptions exist, e.g. for the RR-motif within the signal sequence of substrates of the Twin-Arginine translocation pathway (Alami et al. 2003, Kudva et al. 2013) or the extended signal sequence of the autotransporter family (Type V secretion pathway) (Chevalier et al. 2004). The Twin-Arginine pathway is dedicated to specific bacterial cargos, such as cofactor-containing periplasmic proteins and a few C-tail anchored membrane proteins (Alami et al. 2003, Kudva et al. 2013, Gallego-Parrilla et al. 2024), while the type V secretion is used for outer membrane proteins and extracellular proteins (Clarke et al. 2022). Since there are so far no examples of SMPs using one of these two pathways, they are not covered here. The signal sequence of bacterial membrane proteins usually lacks the signal peptidase cleavage site and instead anchors the protein in the membrane (Boyd et al. 1987, Kuroiwa et al. 1996), therefore they are also referred to as signal-anchor sequences. Most of these signal-anchor sequences in bacteria are co-translationally recognized by the universally conserved signal recognition particle (SRP) (Bibi 2011, Akopian et al. 2013, Steinberg et al. 2018). The E. coli SRP consists of the protein Ffh and the 4.5S RNA (Koch et al. 2003) and represents a simplified version of the more complex eukaryotic SRP, which directs proteins co-translationally to the endoplasmic reticulum (Zimmermann et al. 2011). SRP binds to the ribosome close to the ribosomal peptide tunnel (Fig. 4A) (Gu et al. 2003, Halic et al. 2004, Schaffitzel et al. 2006, Bornemann et al. 2008) and traps emerging signal anchor sequences via the C-terminal methionine-rich M-domain of Ffh (Halic et al. 2006a, Bernstein et al. 1989). SRP then targets the ribosome-associated nascent peptide chain (RNC) to the membrane-bound SRP receptor FtsY. Although FtsY lacks a TM domain, it is stably associated with the membrane via two positively charged membrane binding motifs (Parlitz et al. 2007, Weiche et al. 2008, Braig et al. 2009, Mircheva et al. 2009, Erez et al. 2010). In addition, FtsY interacts with the SecYEG translocon and the YidC insertase (Angelini et al. 2005, Angelini et al. 2006, Kuhn et al. 2011, Draycheva et al. 2016, Jomaa et al. 2016, Jomaa et al. 2017, Petriman et al. 2018, Oswald et al. 2021), which facilitate the insertion of membrane proteins either separately or in cooperation (Kumazaki et al. 2014a, Van den Berg et al. 2004, Sachelaru et al. 2015, Kuhn et al. 2017, Sachelaru et al. 2017). Membrane proteins lacking large periplasmic domains can be inserted by either the SecYEG translocon or via the YidC-only pathway (Samuelson et al. 2000, Welte et al. 2012, Dalbey and Kuhn 2014), while more complex membrane proteins engage the SecYEG translocon or a holo-translocon consisting of SecYEG, YidC and the SecDF complex (Schulze et al. 2014, Komar et al. 2016). Additional proteins can associate with the holo-translocon to support the transport of particular proteins (Jauss et al. 2019).

Figure 4.

Targeting and insertion of bacterial SMPs. (A) The canonical insertion pathway for membrane proteins of > 45–50 amino acids is initiated when the ribosome-bound SRP binds co-translationally to the emerging signal-anchor sequence (red) of the synthesized membrane protein. SRP then targets the ribosome-associated nascent protein (ribosome-associated nascent chain, RNC) to the SRP receptor FtsY, which can bind to both the SecYEG translocon and to the YidC insertase. The SRP-FtsY contact reciprocally stimulates the GTPase activities of FtsY and SRP and induces their dissociation from the ribosome and SecYEG/YidC, respectively. This in turn allows docking of RNCs onto the SecYEG translocon or the YidC insertase, which facilitate the co-translational lipid insertion of the nascent membrane protein. The translational activity of the ribosome and lipid partitioning provide the driving force for membrane protein insertion. Insertion via YidC is limited to small membrane proteins and membrane proteins lacking extended periplasmic loops, while the SecYEG translocon can insert both small and more complex membrane proteins. However, the translocation of large periplasmic domains in membrane proteins requires the ATPase SecA and the proton-motif-force in addition to SRP/FtsY and the SecYEG translocon (Bornemann et al. 2008, Denks et al. 2017, Steinberg et al. 2018). (B) Most SMPs are too short for a co-translational recognition by SRP and are recognized only post-translationally, i.e. after their release from the ribosome (Noriega et al. 2014a, Denks et al. 2017). After their SRP-dependent targeting to FtsY and the GTP-dependent dissociation of the SRP-FtsY complex, SMPs are post-translationally inserted by either the SecYEG-translocon or the YidC insertase (Steinberg et al. 2020, Steinberg and Koch 2021). Lipid partitioning is apparently sufficient as driving force for membrane insertion of SMPs. (C) Many SMPs are produced when cells are exposed to stress conditions or when they enter stationary phase. However, this is also linked to the formation of the hyperposphorylated guanine nucleotides pppGpp and ppGpp by a mechanism called the stringent response (Steinchen et al. 2020). (p)ppGpp act as competitive inhibitors of GTPases, including SRP and FtsY (Czech et al. 2022). Thus, under those conditions the SRP pathway is impaired and SMPs engage an SRP-independent insertion pathway via mRNA targeting. Due to the intrinsic RNA binding activity of the SecYEG-translocon and the YidC insertase, they can tether the mRNAs of SMPs directly. Ribosomes then translate these SecYEG- or YidC-bound mRNAs and the SMPs are directly inserted via SecYEG or YidC into the membrane (Sarmah et al. 2023, Shang et al. 2024).

Binding of the SRP-RNC complex to the SecYEG-bound FtsY initiates the next step in membrane protein insertion. Associated conformational changes align the ribosomal peptide tunnel with the SecY channel (Halic et al. 2006b, Kuhn et al. 2015, Jomaa et al. 2017) and the reciprocal stimulation of the GTPase domains in SRP and FtsY induces substrate release and the dissociation of the SRP-FtsY targeting complex (Peluso et al. 2001, Egea et al. 2004). The nascent membrane protein can then co-translationally enter the lipid phase of the membrane via a lateral opening of the SecY channel, the so-called lateral gate (Collinson et al. 2015, Corey et al. 2016, Rapoport et al. 2017, Hegde and Keenan 2024). Correct folding of the nascent protein (Dalbey and Kuhn 2014) and its subsequent assembly into larger protein complexes is then also supported by YidC, which is located at the lateral gate and even enters the SecY channel (Sachelaru et al. 2015, Sachelaru et al. 2017). Targeting of SRP-RNCs to the YidC-bound FtsY likely follows the same strategy. However, whether YidC can form a protein conducting channel is under debate. The crystal structure of monomeric YidC does not show any channel, but rather indicates that membrane insertion occurs via a hydrophobic groove that is formed by two of YidC’s TM domains and a hydrophilic cavity (Kumazaki et al. 2014a,b, Chen et al. 2015, Chen et al. 2017). On the other hand, in the presence of ribosomes, an ion-conducting channel activity of a YidC dimer was observed (Knyazev et al. 2023). Thus, it is possible that YidC acts in a substrate-dependent manner as a monomeric insertase or a dimeric protein channel.

Membrane insertion of small membrane proteins

SRP-dependent co-translational targeting to either SecYEG or YidC is limited to membrane protein substrates with a minimal length of 45–50 amino acids, because the ribosomal peptide tunnel shields approx. 35 amino acids and SRP requires approximately 10–15 hydrophobic amino acids for stable binding to its substrates (Noriega et al. 2014a,b, Schibich et al. 2016, Denks et al. 2017). Thus, SMPs were initially not considered to be substrates for SRP. This view was changed when direct contacts between SRP and the SMPs YohP and YkgR were observed (Steinberg et al. 2020). These contacts occurred post-translationally and demonstrated that the bacterial SRP can recognize its substrates even after they were released from the ribosome (Fig. 4B). Like in the co-translational mode, SRP targets SMPs to FtsY and by using a reconstituted system, the complete insertion pathway from the initial SRP binding to the membrane insertion by either SecYEG or YidC was revealed (Steinberg et al. 2020). The ability to engage either the SecYEG translocon or the YidC insertase has also been observed for larger membrane proteins (Welte et al. 2012), and demonstrates that SRP can deliver substrates to both insertion sites in the membrane. This is in line with the observation that FtsY was found in contact with both the SecYEG translocon and the YidC insertase (Petriman et al. 2018) and also explains why the depletion of either SecYE or YidC does not necessarily impair membrane insertion of SMPs (Fontaine et al. 2011). Although it is likely that most SMPs are post-translationally targeted to the membrane by the SRP system, for an artificially truncated membrane protein of 56 amino acids, a SecA-dependent membrane targeting was observed (Deitermann et al. 2005). The ATPase SecA serves as motor protein for the translocation of secretory proteins across the SecYEG translocon (Gelis et al. 2007, Allen et al. 2016) and the contribution of SecA to SMP targeting needs to be further analyzed.

An alternative mRNA-based targeting system for small membrane proteins

The production of many SMPs is controlled by environmental conditions, such as membrane stress, nutrient starvation, or low oxygen concentrations (Hemm et al. 2010, Burton et al. 2024, Fesenko et al. 2025). This is explained by their ability to modulate the activity of sensor kinases and thus the perception of extracellular signals (Fig. 1). Additionally, some SMPs are involved in co-factor delivery and stabilization of quinol and cytochrome oxidases with high-oxygen affinity and therefore important for the adaptation to low-oxygen environments (Peters et al. 2008, Pawlik et al. 2010, Hoeser et al. 2014, Kohlstaedt et al. 2017, Rauch et al. 2025). However, nutrient starvation and stress conditions pose a particular challenge for insertion of membrane proteins because they induce a bacterial response program that is called the stringent response (Steinchen et al. 2020, Leiva et al. 2023). Characteristic for this response program is the production of the hyperphosphorylated guanine nucleotides ppGpp and pppGpp by the pyrophosphokinase RelA and the bifunctional (p)ppGpp synthase/hydrolase SpoT (Hauryliuk et al. 2015, Loveland et al. 2016, Haas et al. 2020, Pausch et al. 2020). These signaling molecules, collectively called alarmones (Potrykus and Cashel 2008, Anderson et al. 2021), regulate bacterial metabolism in a concentration dependent manner by controlling the promoter affinity of RNA polymerase and by acting as competitive inhibitors of GTPases (Lemke et al. 2011, Diez et al. 2020, Steinchen et al. 2020). Importantly, this also affects the bacterial SRP system and it was shown that (p)ppGpp occupies the GTP binding sites in both Ffh and FtsY (Czech et al. 2022). Binding of (p)ppGpp to Ffh and FtsY prevents the formation of a stable SRP-FtsY targeting complex and inhibits both co- and posttranslational membrane protein targeting (Czech et al. 2022). This then raises the question on how stress-induced membrane proteins are inserted into the membrane, when the SRP system is inactivated by accumulating (p)ppGpp, which can reach a cellular concentration of up to 1 mM (Potrykus and Cashel 2008, Varik et al. 2017). The K_D values for binding of ppGpp or pppGpp to Ffh or FtsY match the K_D values for GDP and GTP binding (Czech et al. 2022) and it is therefore unlikely that Ffh/FtsY can escape inactivation by binding preferentially to GTP. On the other hand, the number of (p)ppGpp targets in a bacterial cell is quite high (Steinchen et al. 2020, Haas et al. 2022) and it is therefore unlikely that all SRP/FtsY molecules present in a bacterial cell are inactivated by (p)ppGpp. Still, considering that the SRP concentration in E. coli is already low (0.3–0.5 µM) (Kudva et al. 2013) and easily saturated, alternative strategies for the insertion of membrane protein are likely to exist. The insertion of small phage proteins was initially considered to occur spontaneously, i.e. without the help of any targeting factors or membrane transport systems (Kiefer and Kuhn 1999, Kuhn et al. 2017). Although such a low-rate spontaneous insertion might be possible for some SMPs, the available data demonstrate that in vivo most SMPs require either SecYEG or YidC for insertion (Samuelson et al. 2000, Samuelson et al. 2001). In agreement with this, liposome studies did not find any indication for a spontaneous membrane insertion of the SMPs YohP and YkgR (Steinberg et al. 2020). Instead, recent data demonstrate that stress-induced SMPs can be inserted SRP-independently via an mRNA-targeting step (Sarmah et al. 2023, Shang et al. 2024). The existence of such an alternative insertion pathway for SMPs under stress conditions underscores their critical role in stress adaptation.

The intracellular localization of mRNA and local translation generally play an important role in maintaining the organellar sub-proteomes in eukaryotic cells (Kraut-Cohen et al. 2013, Weis et al. 2013) and also contribute to protein localization in bacteria (Amster-Choder 2011, Nevo-Dinur et al. 2011, Benhalevy et al. 2015, Benhalevy et al. 2017, Kannaiah et al. 2019, Irastortza-Olaziregi and Amster-Choder 2020, Mahbub et al. 2020). In a pioneering study, the translation-independent membrane binding of several E. coli mRNAs encoding membrane proteins was shown, while mRNAs encoding cytosolic proteins showed a random distribution within the cytosol (Nevo-Dinur et al. 2011). Although the underlying mechanisms are not completely resolved, the uracil content of the mRNAs appears to contribute to membrane targeting (Kannaiah et al. 2019), which is also reflected by the uracil bias in the coding sequence of membrane proteins (Prilusky and Bibi 2009). The mRNA encoding the SMP YohP was also found to be exclusively membrane localized and this localization was not influenced in the presence of antibiotics, such as kasugamycin or puromycin, which inhibit translation initiation or elongation, respectively (Steinberg et al. 2020, Sarmah et al. 2023). Although this is generally taken as evidence for a translation-independent mRNA localization, it does not necessarily proof that the mRNA reaches the membrane without being in contact with the ribosome or its subunits (Shang et al. 2024). However, in in vitro assays, the ribosome-independent membrane binding of the yohP mRNA was demonstrated and it was also shown that the yohP mRNA binds to the cytosolically exposed loops of SecY and YidC (Sarmah et al. 2023). Thus, SecY and YidC not only serve as insertion site for membrane proteins, but also as mRNA receptor for their respective mRNAs (Fig. 4C). During co-translational targeting, both SecY and YidC make extensive contacts to the ribosomal RNA (Frauenfeld et al. 2011, Kedrov et al. 2016) and RNA binding is likely a primordial and conserved ability of both proteins (Jagannathan et al. 2014, Bhadra et al. 2021, Lewis and Hegde 2021). Ribosomes then bind and translate these SecY- or YidC-bound mRNAs and the translation product is inserted into the membrane by either SecYEG or YidC. Importantly, when YohP was translated from already SecY-bound mRNAs, its insertion was independent of the SRP pathway (Sarmah et al. 2023). This suggests that mRNA targeting can bypass the need for SRP-dependent membrane protein insertion. In vivo, the SRP-dependent protein targeting and SRP-independent mRNA targeting probably act in parallel (Shang et al. 2024). However, when the SRP-dependent insertion is blocked by inducing the stringent response and (p)ppGpp synthesis, the insertion via mRNA targeting is essential for the insertion of SMPs (Sarmah et al. 2023), which ensures the efficient insertion of stress-responsive SMPs. However, it is important to emphasize that experimental evidence for an SRP-independent insertion of SMPs via mRNA targeting is so far only available for a few examples and that it is currently unknown whether mRNA targeting also eliminates the need for SRP during the insertion of larger membrane proteins. In addition, the contribution of other mRNA binding proteins, besides SecY and YidC, to localized translation and membrane insertion needs further analyses. One additional protein that appears to be involved in mRNA membrane binding is RNase E, which was shown to form a complex with the SecYEG translocon (Kuhn et al. 2011) and to influence the polar localization of mRNAs in concert with MinD (Kannaiah et al. 2024).

The diverse effects of small membrane toxins

Type I toxin-antitoxin systems in bacteria

Bacterial toxin-antitoxin (TA) systems were originally discovered on plasmids, where they contribute to plasmid maintenance by selectively inhibiting or killing plasmid-free offspring. Examples include ccdB/ccdA on plasmid F and hok/sok on plasmid R1 (Ogura and Hiraga 1983, Gerdes et al. 1986). However, TA systems have also been found in bacterial genomes, sometimes in astonishingly high numbers (Anantharaman and Aravind 2003, Pandey and Gerdes 2005, Fozo et al. 2010), suggesting that they have functions beyond plasmid stabilization. Bipartite TA modules consist of a toxin that impairs cellular functions and an antitoxin that prevents toxin activity either directly by interacting with the toxin, indirectly by having an antagonistic effect, or by repressing toxin translation. TA systems are classified into at least eight different types depending on the nature of the antitoxin (RNA or protein) and the particular mode of toxin inhibition (reviewed in (Jurenas et al. 2022, Shore et al. 2024)). Types I and II contain the most well-studied examples; the TADB 3.0 database (Guan et al. 2024) currently lists 505 experimentally validated type I (102 entries) and type II systems (403 entries) as compared to only 31 validated systems of the remaining types (accessed on 11 September 2025). Furthermore, there is an overrepresentation of TA systems from Enterobacteriaceae, Mycobacteriaceae, Staphylococcaceae, Pseudomonadaceae, Streptococcaceae, Lactobacillaceae, Vibrionaceae, Bacillaceae, and a few other bacterial families, indicating a knowledge gap regarding the abundance and phylogenetic distribution of TA systems.

Type I TA systems are of particular interest with regard to small proteins, as the corresponding toxins are usually SMPs. In these systems, the antitoxin is a small RNA that binds to the toxin mRNA, thereby preventing translation initiation by occupying either the ribosome binding site or by inducing structural rearrangement within the toxin mRNA (reviewed in (Brantl and Jahn 2015, Shore et al. 2024)). Additionally, primary transcripts of toxin genes typically require a processing step to become translationally active (Thisted et al. 1994, Darfeuille et al. 2007, Kristiansen et al. 2016, Arnion et al. 2017), resulting in uncoupling of transcription and translation (reviewed in (Masachis and Darfeuille 2018)). This intricate post-transcriptional regulation provides important checkpoints and restricts toxin production to specific situations. Indeed, it was observed for the tisB/istR-1 system in E. coli that removal of the post-transcriptional regulation mechanisms caused inadvertent toxin production, which was a selective disadvantage under certain conditions (Berghoff et al. 2017a, Edelmann et al. 2021a, Edelmann et al. 2021b).

Although plasmid-encoded type I TA systems are known to contribute to plasmid stability, the biological functions of chromosomal type I TA systems are not fully understood. Interestingly, some chromosomal type I TA systems are associated with other mobile genetic elements, such as prophages. In Clostridioides difficile, for example, the CD/RCd type I TA systems are located within the phiCD630 prophage regions and stabilize the prophages by inhibiting the growth of prophage-free cells through the action of the CD membrane toxins (Maikova et al. 2018, Peltier et al. 2020). Likewise, in Bacillus subtilis, it is assumed that the type I TA systems bsrE/SR5, bsrG/SR4, txpA/ratA, and yonT/SR6 contribute to prophage stabilization, probably by killing of prophage-free cells via membrane toxins BsrE, BsrG, TxpA, and YonT, respectively (Silvaggi et al. 2005, Durand et al. 2012, Jahn et al. 2015, Müller et al. 2016, Reif et al. 2018). Thus, type I TA systems promote the selfish propagation of mobile genetic elements, but also stabilize important functions encoded by these elements. The question about the function of type I TA systems that are not associated with mobile genetic elements still remains and is difficult to answer for many systems. This is primarily because the relevant physiological conditions have yet to be identified. However, some type I TA systems are integrated into certain stress responses and likely contribute to stress tolerance and survival strategies. For example, the TisB toxin is involved in the SOS response to DNA damage, while the HokB toxin responds to elevated (p)ppGpp levels mediated by the Obg GTPase in E. coli (Dörr et al. 2010, Verstraeten et al. 2015).

Small membrane toxins cause primary and secondary effects

Most toxins from type I TA systems are small hydrophobic or amphipathic proteins with a length of < 60 amino acids. They usually have one transmembrane helix and short N- and C-terminal extensions of only a few amino acids (Nonin-Lecomte et al. 2021). There are however exceptions. One of the smallest type I toxins, IbsC, has only 19 amino acids, which are almost entirely involved in transmembrane helix formation (Mok et al. 2010). By contrast, type I toxins from the Hok/Gef family have sizes of ∼50 amino acids and have rather long C-terminal extensions. These extensions may contain cysteine residues that are important for dimerization and functionality, as seen with HokB (49 aa) in E. coli (Wilmaerts et al. 2019, Wilmaerts et al. 2021). The transmembrane helices of type I toxins mediate localization to the cytoplasmic membrane, where they disturb membrane functions. This ultimately results in growth arrest, which is sometimes accompanied by phenotypic alterations. For instance, in Helicobacter pylori, expression of the AapA1 toxin is responsible for the morphological transformation from a spiral shape to coccoids (El Mortaji et al. 2020). Even though cell lysis is occasionally observed upon expression of small membrane toxins (Jahn et al. 2015), it is generally assumed that the main function is inhibition of cellular activity rather than cell death. It was observed that many of these toxins cause depolarization of the inner membrane, for example DinQ (Weel-Sneve et al. 2013), HokB (Verstraeten et al. 2015), IbsC (Fozo et al. 2008), ShoB (Fozo et al. 2008), TisB (Lobie et al. 2025), and ZorO (Bogati et al. 2021) in E. coli, TimP (Andresen et al. 2020) in Salmonella, or PepG1 (Fermon et al. 2023) in Staphylococcus aureus. It is assumed that depolarization results from oligomeric assembly of type I toxins within the inner membrane and formation of water-filled pores that are able to dissipate ion gradients across the membrane (Gurnev et al. 2012, Wilmaerts et al. 2018, Schneider et al. 2019). In particular, depolarization indicates proton gradient dissipation and disturbance of pH homeostasis, which was experimentally addressed for TisB in E. coli. By using pH-sensitive fluorescent probes it was shown that TisB was the major cause of proton gradient dissipation under DNA damage conditions (Cayron et al. 2024, Lobie et al. 2025). Dissipation of the proton motive force interferes with ATP production by ATPase, which explains why ATP depletion is commonly observed as a consequence of membrane toxin expression. Alternatively, membrane toxins might even cause direct ATP leakage as observed for HokB in E. coli (Wilmaerts et al. 2018). However, whether this is a common mechanism remains to be investigated. Finally, ATP depletion and the concomitant reduction of cellular activities are associated with the establishment of a stress-tolerant persister state (Dörr et al. 2010, Conlon et al. 2016, Shan et al. 2017, Edelmann and Berghoff 2022).

The consequences of membrane toxins are often studied using plasmid-based overexpression constructs. While this strategy is important to make initial observations, phenotypes need to be substantiated by experiments with wild-type strains in comparison to toxin deletion mutants under relevant stress conditions. Using TisB as a model for toxin biology in E. coli, several secondary effects were revealed over the past years. The tisB gene is part of the bacterial SOS response (Vogel et al. 2004) and is one of the genes with the strongest induction following DNA damage (Berghoff et al. 2017b). A common strategy to study TisB-dependent cellular effects is the induction of the SOS response using fluoroquinolone antibiotics, such as ciprofloxacin or ofloxacin, which inhibit topoisomerases and cause double-strand breaks. It was observed that tisB-expressing cells produce reactive oxygen species (ROS), in particular superoxide, leading to upregulation of the oxidative stress response. Upon ciprofloxacin treatment, ROS formation was significantly reduced in a tisB deletion strain, suggesting that membrane toxins disturb metabolic functions and elicit secondary stress responses (Fozo et al. 2008, Edelmann and Berghoff 2019). Another observation relates to protein aggregation and cytoplasmic condensation. Strong ATP depletion and cytoplasmic protein aggregates were observed upon moderate tisB expression. Notably, aggregates were visible in wild-type cells following ciprofloxacin treatment, whereas a tisB deletion strain showed no evidence of aggregate formation (Leinberger et al. 2024). Similarly, TisB-dependent cytoplasmic condensation occurred after treatment with ofloxacin, as evidenced by leakage of the cytosol and the appearance of clear regions within the cell. However, as with protein aggregation, there was no evidence that cytoplasmic condensation contributed to antibiotic lethality (Cayron et al. 2024). On the contrary, protein aggregation actually increased the duration of cellular dormancy, thus supporting long-term tolerance to antibiotic stress (Pu et al. 2019, Dewachter et al. 2021, Leinberger et al. 2024). It is therefore tempting to speculate that cytoplasmic condensation also contributes to dormancy and tolerance. An early study showed that tisB overexpression impairs major cellular processes, such as replication, transcription, and translation (Unoson and Wagner 2008). Whether this is an overexpression artifact remains unknown, but diminished protein production was also observed in wild-type cells in comparison to a tisB deletion strain following prolonged ciprofloxacin exposure (Edelmann et al. 2021a). Fig. 5 shows a model of primary and secondary effects that are commonly observed upon type I toxin expression. Collectively, these effects promote a state of cellular inactivity and increase stress tolerance (Edelmann and Berghoff 2022). The model is primarily based on observation made with TisB, HokB, DinQ, and some other well-characterized examples, but we suggest that most of the observed effects also apply to other membrane toxins.

Figure 5.

Primary and secondary effects of small membrane toxins. Primary transcripts of toxin genes must undergo processing before being translationally competent. Toxin monomers attach to the membrane (probably via positively charged amino acids) and eventually form water-filled pores (Gurnev et al. 2012, Wilmaerts et al. 2018, Schneider et al. 2019). The primary consequence of pore formation is the dissipation of the proton gradient, which in turn disables ATP production by ATP synthase. Further secondary effects include ROS production, protein aggregation, and cytoplasmic as well as nucleoid condensation. Together, these effects decrease cellular activity and support a stress-tolerant persister state. The model is mainly based on observations made with the well-characterized type I toxins TisB, HokB, and DinQ (Weel-Sneve et al. 2013, Edelmann and Berghoff 2019, Cayron et al. 2024, Leinberger et al. 2024, Lobie et al. 2025).

Another effect that is occasionally observed is the compaction of chromosomal DNA, i.e. nucleoid condensation, as seen with membrane toxins Fst, LdrD, and DinQ from the type I TA systems fst/RNAII, ldrD/rdlD, and dinQ/agrB, respectively (Kawano et al. 2002, Patel and Weaver 2006, Weel-Sneve et al. 2013). It is known that nucleoid-associated proteins (NAPs) such as Dps, HU, and Fis cause nucleoid condensation during the transition into stationary phase. This provides physical protection for the DNA via a coating mechanism (Holowka and Zakrzewska-Czerwinska 2020). In E. coli, the SMP YohP is induced during stationary phase and contributes to nucleoid condensation (Steinberg et al. 2020). Deletion of hupAB, which encodes the histone-like heterodimeric protein HU, reduces YohP-induced nucleoid condensation, suggesting that YohP acts indirectly and is at least partially mediated by other proteins, such as HU (Natriashvili et al., 2025). It remains to be investigated whether small membrane toxins from type I TA systems affect DNA compaction in a similar manner. Interestingly, increased production of the DinQ toxin in E. coli was found to be associated with higher sensitivity to ultraviolet light-induced DNA damage. It has been hypothesized that prolonged nucleoid condensation in the DinQ-producing strain impairs DNA repair processes (Weel-Sneve et al. 2013), suggesting that an elevated toxin production disturbs the balance between physical nucleoid protection and accessibility to DNA repair machineries.

Open questions about membrane toxins

The process by which membrane toxins insert into the cytoplasmic membrane is not yet fully understood. One possibility is that amphipathic toxins initially attach to the negatively charged membrane using positively charged amino acid residues, followed by hydrophobic interactions and the formation of water-filled pores due to oligomerization (Scheglmann et al. 2002, Cao et al. 2023). Indeed, the positively charged amino acid residues are often essential for toxin functionality (Korkut et al. 2020, Bogati et al. 2022, Fermon et al. 2023, Leinberger and Berghoff 2024). A self-contained insertion of membrane toxins is supported by in vitro experiments with model lipid bilayers and by molecular dynamics simulations (Gurnev et al. 2012, Wilmaerts et al. 2018, Schneider et al. 2019, Korkut et al. 2020). Alternatively, membrane insertion might be promoted by SRP and SecY/YidC as described above for YohP and other SMPs, but this has not yet been experimentally addressed. Another central question concerns potential interaction partners of membrane toxins. The idea of toxin oligomerization and the formation of water-filled pores that dissipate the proton gradient does not consider the role of interaction partners, and supports the view of toxins acting independently. Identifying interaction partners that modulate membrane toxins, or even enable their functionality, would change our understanding of how these small membrane toxins operate.

Outlook

SMPs are promising candidates for a variety of applications. In basic research, they can be used to modulate the activity of membrane complexes. For example, the activity of membrane machineries—such as components of electron transport chains or transport systems—can be turned on and off using an SMP-based regulatory strategy. This could be beneficial when deletion strains exhibit severely impaired fitness or cannot be constructed due to the essential nature of the genes. These considerations are based on the assumption that the relevant SMP is already known. Alternatively, synthetic SMPs can be designed from scratch using artificial intelligence to predict structures and SMP-protein interactions. Similarly, small membrane toxins can serve as a blueprint for designing synthetic toxic peptides that can be used as new antibiotics to treat multidrug-resistant pathogens.

Funding

J.Y., H.-G. K. and B.A.B. were supported by the German Research Council (DFG) in the framework of the priority programme SPP 2002 (YU 247/3-1 to J.Y.; KO 2184/9-1 and KO 2184/9-2 to H.-G. K.; BE 5210/3-1 and BE 5210/3-2 to B.A.B.). H.-G. K. also acknowledges funding via the SFB1381 (Project-ID 403222702).

Conflict of interest

The authors declare no conflict of interest.