Small Proteins; Big Questions

ABSTRACT

In recent years, there has been increased appreciation that a whole category of proteins, small proteins of around 50 amino acids or fewer in length, has been missed by annotation as well as by genetic and biochemical assays. With the increased recognition that small proteins are stable within cells and have regulatory functions, there has been intensified study of these proteins. As a result, important questions about small proteins in bacteria and archaea are coming to the fore. Here, we give an overview of these questions, the initial answers, and the approaches needed to address these questions more fully. More detailed discussions of how small proteins can be identified by ribosome profiling and mass spectrometry approaches are provided by two accompanying reviews (N. Vazquez-Laslop, C. M. Sharma, A. S. Mankin, and A. R. Buskirk, J Bacteriol 204:e00294-21, 2022, https://doi.org/10.1128/JB.00294-21; C. H. Ahrens, J. T. Wade, M. M. Champion, and J. D. Langer, J Bacteriol 204:e00353-21, 2022, https://doi.org/10.1128/JB.00353-21). We are excited by the prospects of new insights and possible therapeutic approaches coming from this emerging field.

INTRODUCTION

For the purposes of this discussion, we consider “small proteins” to be proteins of 50 amino acids (aa) or fewer in length for which there is evidence that the small protein accumulates to sufficient levels for it to have a function. It should be noted that this definition is not strict, and we mention some proteins that are somewhat longer, as do other articles in this special issue. In general, the main defining and unifying characteristics of these proteins are their small size and the fact that they commonly have been missed. There are several interrelated reasons why small proteins have remained hidden. First, while automated genome annotation pipelines are adept at detecting long open reading frames (ORFs) in genome sequences, the false discovery rates balloon as the allowable length of an ORF drops (reviewed in reference 1). Consequently, many pipelines for protein annotation in bacteria ignore proteins smaller than 50 aa (reviewed in reference 2). Second, proteomic workflows similarly are optimized for large proteins such that the majority of small proteins are not retained, resolved, or detected unless specialized conditions are used. Similarly, most mass spectrometric methods rely on the ability to obtain multiple tryptic peptides from a single protein, which is not possible for small proteins (reviewed in reference 3). A compounding problem is that many of the search engines for peptide sequences are based on genome annotations, in which, as mentioned above, small proteins are underrepresented. The latter issue also impacts the identification of small protein genes by mutational screens.

Although their small sizes limited their identification and characterization for a long time, the increasing discovery of small proteins that are associated with mutant phenotypes or interact with larger proteins establishes a precedent for function. As such, overlooking an entire class of proteins means that investigators may be missing a whole level of regulation, critical structural components, as well as unique mechanisms of action. Until now, small proteins generally have been studied from perspectives and methods focused on large proteins. The recognition that there is a group of proteins missed by multiple approaches is an important step in discerning the diverse functions of these small proteins.

Peptide chains of 50 aa or fewer can be encoded by a defined short ORF (sORF) positioned at a variety of genetic locations (Fig. 1). They can be entirely separate sORFs, not overlapping other genes but encoded as part of operons or as independent transcripts (Fig. 1A). A subset of operonic sORFs are regulatory upstream ORFs (uORFs) (also denoted leader peptides) (Fig. 1B), whose translation impacts the expression of the downstream gene (reviewed in reference 4). sORFs also can be encoded antisense to a longer ORF or partially or fully overlapping a longer ORF on the sense strand (Fig. 1C). sORFs that are completely overlapped by a longer ORF, in the same or an alternate frame, are sometimes denoted embedded (or nested). Additionally, sORFs can be encoded on a small regulatory RNA, either overlapping or adjacent to base-pairing sequences (Fig. 1D). These protein-coding regulatory RNAs are often designated dual-function RNAs (reviewed in reference 5). Finally, small proteins can be cleaved from larger polypeptides (Fig. 1E). We refer to these as small peptides to distinguish this group from the independently translated small proteins. Here, we primarily focus on small proteins encoded as separate sORFs given that more of these small proteins have been characterized.

FIG 1.

Locations of small protein genes. (A to C) Small proteins can be encoded by independent sORFs (A), uORFs that act as regulators of the downstream gene (B), or sORFs encoded antisense to a larger ORF or embedded or nested within a larger ORF in the same or an alternative frame (C). (D) They also can be encoded on a regulatory small RNA that has a dual function as a base-pairing RNA and an mRNA encoding a small protein. (E) Functional small peptides can be cleaved from larger proteins.

While a 50-aa limit is useful as a working definition, slightly larger proteins that are unannotated thus far can be discovered and characterized with approaches developed for small proteins. It is not yet clear if there will be a lower size limit. Is there a size that should be considered too small? The surface area available for stable interaction with other molecules diminishes with size. Until now, the smallest characterized proteins, such as the antibacterial microcin C7 protein encoded as a separate sORF (6) and microbial signaling peptides processed from larger precursors (reviewed in reference 7), are 5 to 10 aa.

HOW CAN SMALL PROTEINS BE IDENTIFIED?

Investing in the development of approaches and standards for the identification of small proteins will facilitate small protein research and help to unify the field. Thus far, most identification has come from approaches used to identify larger proteins, although more and more improvements that allow better detection of small proteins are being developed.

While computational annotation for long ORFs can rely almost exclusively on the detection of long continuous coding potential, additional parameters are needed to mitigate the bioinformatic prediction problem of the inflated false discovery of sORFs. Thus, many computational searches for small proteins weight multiple parameters, which can include the presence of a ribosome binding site and amino acid conservation (8), although the latter parameter is progressively less effective the shorter the sORF. More and more computational searches also incorporate information from extensive RNA sequencing or mass spectrometric data sets currently available (9–11). One caution in using these data sets is that the transcription and even the translation of an sORF do not guarantee that a stable small protein with function is generated. Computational approaches are less effective at identifying candidate sORFs that are conserved in only a few bacteria or that overlap longer ORFs. Nevertheless, as the computational approaches continue to be refined, small protein genes can be predicted for any sequenced genome, with the caveat that the numbers of false-negative and false-positive results generally are unknown.

Ribosome binding to an RNA suggests that the transcript is being translated. Thus, ribosome profiling (also called Ribo-seq or ribosome footprinting), which allows ribosome binding to be interrogated on a genome-wide basis upon the isolation of translating ribosomes, is increasingly being used to identify small protein genes. As described in an accompanying review (12), modifications of the initial ribosome profiling protocols, such as treatment with antibiotics that enrich for initiation and termination complexes, have greatly improved this approach for identifying small proteins, although investigators who use the resulting data still need to be aware of limitations. An advantage of ribosome profiling is that the approach can identify actively translated ORFs in intergenic regions as well as provide evidence for the translation of antisense or embedded sORFs.

Improved methods for the direct detection of the translated polypeptide are also crucial, particularly because these may be the only approaches for identifying small peptides cleaved from larger translation products. Proteomic detection by mass spectrometry is the best method thus far for identifying both small proteins and small peptides, but again, the small sizes pose barriers to identification. The challenges related to small protein identification by mass spectrometry as well as recent developments in this field are discussed in another accompanying review (13). We anticipate that innovative developments in direct detection will have a significant impact on small protein research. While both ribosome profiling and proteomic detection can provide insights into differential expression for samples collected under various growth conditions, each approach has advantages and disadvantages. Ribosome profiling is considerably more sensitive than the proteomic approaches, while the proteomic approaches document that a small protein actually accumulates to a meaningful level.

It should be noted that the number of small proteins identified by ribosome profiling and direct detection will vary with the growth conditions, particular strain examined, or experimental setup, leading to potential differences among reports of the proteins. Related to this variation is the important question of how much validation is required to be confident that a small protein is synthesized and accumulates. Until now, the gold standard has been the independent detection of chromosomally encoded epitope-tagged small proteins. However, this gold standard may not be achievable for all legitimate bona fide small proteins predicted, as protein tagging is difficult in some organisms. Phenotypes associated with reduced or elevated levels of a small protein, if carried out with appropriate stop codon controls, can provide additional evidence for the existence of a small protein.

We advocate that in the absence of the detection of a chromosomally encoded tagged small protein, at least two lines of evidence be required for the annotation of a small protein gene. Thus, a small protein may be independently identified by conservation and ribosome profiling or ribosome profiling and proteomic approaches. We note that there is nuance in evaluating the quality of different types of evidence, however, and that appropriate quantitation, reproducibility, and control groups should be expected for all data sets used. Different discovery methods have the added value that they can provide distinct but often complementary information. For example, conservation is an indication of function but does not provide information about expression. On the other hand, mass spectrometric analysis shows that a small protein is stable enough to be detected but not whether it has function.

A corollary to the limitations in identifying translated sORFs and small proteins is ambiguity and variability with respect to information in databases. Currently, there are both under- and overannotation of small protein genes. Once new small proteins are stringently vetted for their synthesis as described above, the genes that encode them need to be added to genome annotations. There is value to also capturing information about small proteins for which corroborating evidence is not available as long as the level of confidence is clearly denoted. Websites collating evidence for small proteins are being developed (14), along with expanding databases of ribosome profiling and mass spectrometric data. We encourage the establishment of reliable, rigorous data sets that the small protein research community can continue to build on as this nascent field develops.

The approaches described for the discovery of small proteins and their genes will also lead to the identification of more uORFs. In fact, uORFs and sORFs cannot be distinguished in ribosome profiling data, other than through information about the location of the ORF (15). The identification of uORFs is an added benefit of ribosome profiling studies, and the encoded products of regulatory uORFs may have independent functions as small proteins (16).

The number of small proteins encoded by a genome is not known for any organism. At least 150 small proteins have been documented as chromosomally encoded tagged proteins for the well-studied organism Escherichia coli, but that is not the complete count, and only a fraction of these have been characterized sufficiently to know about function (reviewed in reference 2). Even less is known about what small proteins are synthesized or their functions in other organisms.

HOW CAN SMALL PROTEIN FUNCTIONS BE ELUCIDATED?

Along with the problems associated with the discovery of sORFs and small proteins, studies of the biological roles of small proteins come with several challenges. The first major issue is whether the product of a translated sORF has a function in the cell. It is possible that ribosomes are associated with an ORF as a result of noisy translation initiation or incidental translation. Alternatively, as mentioned above, the translation of the ORF itself may be regulatory in cis, while the product of translation has no function in trans.

The conservation of an sORF and also surrounding genes is one clue that the translated small protein might have a function, and information about conserved nearby genes can help predict that function. For example, the genes encoding toxic small proteins belonging to toxin-antitoxin (TA) systems generally are genetically linked to the genes encoding their antitoxin counterparts and thus can be identified through gene synteny (reviewed in reference 17). The value of searches for conserved sequences is limited, however, when the putative small protein is produced from an sORF embedded in a larger gene (18). Here, the analyses become muddled by the features of the larger gene product. Dual-function RNAs harboring a small protein-encoding sORF as well as a base-pairing sequence element, which can overlap the sORF, similarly are difficult to predict using bioinformatics (5). By definition, conservation and gene synteny analyses also cannot be used to detect functional strain-specific small proteins.

Other clues that translated small proteins have function come from in vivo and in vitro experiments. For instance, differential expression of the small protein and the mRNAs that encode them, such as induction in response to a specific stress, is a hint that the protein might have a function under that condition. This differential expression can be detected through the generation and examination of transcriptomic or proteomic data sets. Existing data sets can also be interrogated to detect these differentially expressed small proteins. However, as many small protein ORFs are currently lacking in standard genome annotations, these analyses necessitate remapping of previous data sets. Given that large transcriptome repositories such as the Gene Expression Omnibus (GEO) (19) currently hold thousands of published data sets, it is likely that mining existing data could provide new information on the expression of putative small protein genes. Such expression analyses can aid in the identification of environmental conditions under which the small protein of interest is produced and thus is more likely to carry out its biological function.

Phenotypes associated with either the overexpression of a small protein or, better yet, the deletion of the corresponding gene suggest biological functions and can provide important hints about these functions. Collections of overexpression plasmids or genome-wide mutational libraries such as transposon insertion sequencing (Tn-Seq) libraries (20) can be screened for a variety of growth phenotypes under various conditions. The effects of over- or underexpression can also be monitored by global transcriptome and proteome analyses. Again, it is worth noting some caveats. These approaches typically rely on existing annotation, and thus, unannotated small protein genes might evade analysis. Additionally, due to their size, mutation of small protein genes in Tn-Seq experiments requires a much higher transposon density than for conventionally sized genes in order to obtain the number of insertions necessary to achieve statistical confidence that the altered recovery of mutations in an sORF is meaningful. These approaches again are more challenging for genes where there is overlapping coding potential. Nonetheless, reevaluating existing data sets focusing on sORFs could promote a better understanding of small protein functions in various microbial species.

Once a candidate small protein has been identified, it is crucial to confirm that the protein accumulates in the cell before carrying out more detailed functional studies. To this end, various technologies are employed. Most frequently, the addition of an epitope tag to the N- or C-terminal end of the small protein at the chromosomal locus followed by Western blot analysis is used to validate small protein accumulation and further examine regulated expression (15). This approach comes with the advantage that the required genetic modifications are typically quick and straightforward. However, it is possible that the added tag interferes with or alters the expression and/or the function of the protein. While this is true for all proteins, it is particularly a concern for small proteins where even a small epitope tag is disproportionately large relative to the small protein and may obstruct interactions or significantly change the overall charge or hydrophobicity. As a consequence, it is important to ultimately test whether the expression of the tagged protein phenocopies the expression of the untagged small protein in complementation assays. These assays can be simplified by constructing reporters to genes whose expression is contingent on the function of the small protein. It may be necessary to test a variety of small protein derivatives with different types of tags and linker sequences. Alternatively, the small protein may need to be tagged through the introduction of a nonnatural amino acid. Antibodies specific to the small protein of interest would also be useful but are often difficult to obtain due to the limited number of epitopes present in a small protein (reviewed in reference 21).

The next steps in elucidating the biological role of a small protein usually involve determining its subcellular localization and identifying cellular interaction partners. Similar to the transcriptomic data sets mentioned above, small proteins typically are missing from large-scale protein localization and interaction studies, such as those carried out for E. coli (22, 23). Hence, independent analysis is usually required. Coimmunoprecipitation experiments using functional tagged small protein variants or specific antibodies have proven most successful for identifying interacting proteins that can lead to strong hypotheses regarding the cellular function of a small protein. Given the examples of small proteins that act as DNA or RNA binding factors (2, 24–26), copurification with nucleic acids should also be considered. In addition, it is conceivable that small proteins could act by binding to other cellular components such as lipids or small metabolites; however, no dedicated searches for such interactions have been reported. The unique advantage that small proteins can be chemically synthesized in their entirety can also be exploited in biochemical assays. Thus far, different combinations of experiments have led to the elucidation of individual small protein functions and are setting the stage for structural and further mutational studies for a mechanistic understanding of small protein mechanisms of action.

WHAT SMALL PROTEIN FUNCTIONS ARE KNOWN?

Several bacterial small proteins have now been functionally characterized. While we are not able to discuss all of these proteins in the context of this summary article, it is useful to mention examples that represent the general functional categories discovered thus far. A surprisingly large percentage of characterized small proteins act at the membrane. More detail about these small proteins is provided by another review in this issue (27).

One obvious category is the proteins that act as toxins, including some TA system toxins, microcins, phenol-soluble modulins like Staphylococcus alpha-hemolysin that act as a defense against host cells, and holin proteins produced by lytic bacteriophage (reviewed in references 28, to ,31). Many of these proteins are highly hydrophobic and contain an α-helical transmembrane domain. The overexpression of these transmembrane domain toxins typically results in depolarized membranes or membrane disruption, which has been attributed to the insertion of the small proteins into the cell membrane generating a pore or otherwise disrupting the membrane or membrane proteins.

Toxic proteins are not the only small proteins that consist of a transmembrane α-helix. A number of other small proteins found in the membrane are not toxic but rather associate with larger membrane proteins to regulate their levels or activities. It is not yet clear what features distinguish toxic α-helical transmembrane small proteins from those that are not toxic. The regulation of magnesium homeostasis in enteric bacteria illustrates how small nontoxic transmembrane proteins can modulate the levels of larger proteins. The 30-aa MgtR protein from Salmonella interacts with MgtA, MgtB, and MgtC (32, 33), which are inner membrane proteins required for adaptation to low magnesium concentrations, and induces their degradation by the FtsH protease (34). In contrast, the 28-aa MgtU protein stabilizes the same proteins and thus increases intercellular magnesium concentrations in Salmonella (32). MgtR and MgtU are not found in E. coli. However, in this organism, the 31-aa MgtS protein binds and stabilizes the MgtA magnesium transporter by blocking FtsH-mediated degradation (35). MgtS additionally binds and negatively regulates the PitA cation-phosphate symporter, resulting in increased intracellular magnesium (36).

Additional examples of transmembrane small proteins that modulate the activities of transporters are the 30-aa PmrR protein found in Salmonella and the more broadly distributed 49-aa AcrZ and 37-aa SgrT proteins. PmrR, whose expression is under the control of the iron (Fe³⁺)-responsive PmrAB two-component system (TCS), downregulates the activity of LpxT, an inner membrane protein that increases the negative charge of the lipopolysaccharide (LPS) (37). A less electronegative form of the LPS attracts fewer Fe³⁺ ions to the outer membrane, thereby dampening PmrAB activation as part of a negative-feedback loop. AcrZ associates with and modulates the activity of the AcrAB-TolC drug efflux pump, which confers resistance to many antibiotics (38). In conjunction with the surrounding membrane lipids, AcrZ has an allosteric effect on AcrB, leading to a conformational change that facilitates the transport of certain compounds (39, 40). E. coli SgrT specifically interacts with the unphosphorylated form of the PtsG glucose transporter inhibiting glucose uptake (41).

Other transmembrane small proteins appear to be integral parts of larger enzymatic protein complexes where they can have more of a structural role. Several different small proteins have been found to interact with cytochrome oxidase complexes. In E. coli, the 37-aa CydX and 29-aa CydH proteins are part of the cytochrome bd oxidase and likely assist in the folding and stability of the complex (42–44). Work on photosynthetic bacteria has also provided several examples of transmembrane small proteins affecting the activity of metabolic enzyme complexes (45), where they impact complex assembly (46, 47) as well as protein turnover (48, 49).

Another general category of small proteins is composed of those that interact with the membrane as amphipathic helices or through other membrane anchors and thus can serve as tethers to larger proteins. Here, the best-characterized example is the 26-aa SpoVM protein of Bacillus subtilis, which specifically binds curved membranes to recruit the SpoIVA protein during spore formation (50). Detailed structural and biophysical studies have given significant insights into how SpoVM recognizes membrane curvature (51–53). Another example in the human pathogen Listeria monocytogenes is the 31-aa Prli42 protein, which is predicted to be a tail-anchored membrane protein. Prli42 interacts with the stressosome complex, inducing the complex upon oxidative stress (54).

Small proteins do not need to act at the membrane to affect enzyme activity. In Synechocystis, the 44-aa AcnSP protein binds to, and stimulates the activity of, aconitase, thus impacting flux through the citric acid cycle (55), while the 51-aa PirA protein modulates global nitrogen metabolism by binding regulatory protein P_II (56). Examples can also be found in archaea where, in Methanosarcina mazei, binding of the 23-aa sP26 protein to glutamine synthetase (GlnA₁) is predicted to stabilize the dodecameric oligomer and might also promote the association of GlnK₁ (57). Archaeal small proteins are described more extensively in another review in this issue (58).

In addition to regulating the activities of transporters and enzymatic complexes, small proteins can control the activities of other regulators. Another E. coli small protein impacting magnesium homeostasis is the 47-aa MgrB transmembrane protein, which modulates signaling by inhibiting the PhoQ sensor kinase part of the PhoPQ TCS (59). The mgrB gene itself is activated by PhoPQ, and thus, the small protein provides autoregulation to the PhoPQ TCS (60). We expect that other small proteins that modulate kinases will be identified.

Small proteins can also act by affecting the activity of transcription regulators. For example, small, secreted peptides (5 to 10 aa) produced by various Gram-positive bacteria bind to and activate transcription factors belonging to the RRNPP (named for prototypical members) protein family (reviewed in reference 7). These peptides are typically part of bacterial communication systems and are frequently processed from larger proteins by specific proteases. Intriguingly, a similar mechanism is employed by temperate phages to control lysis versus lysogeny (61). We anticipate that more proteins of fewer than 50 aa will be found to bind and modulate transcription factors. Additionally, it is likely that there are other small proteins that bind specifically to DNA or RNA and have independent regulatory functions, particularly since some well-studied transcription factors (the Cro repressor of bacteriophage λ is 66 aa) and RNA binding proteins (the global posttranscriptional regulator CsrA in E. coli is 61 aa) are relatively small. Finally, it is noteworthy that many core ribosomal proteins (r-proteins) as well as several r-proteins that are assembled into the ribosome at times of specific stresses fit our definition of small proteins (2, 62).

WHAT ARE THE FUNDAMENTAL PROPERTIES OF SMALL PROTEINS?

In conjunction with the need to identify functions for small proteins, much remains to be learned about the fundamental properties of these proteins. Some of these properties, such as how the proteins are localized within the cells, whether the proteins are posttranslationally modified, and how the small proteins are degraded, are the same questions asked about larger proteins. Even the fundamental question of the relative concentration of small proteins has not been answered. Here again, the small size poses challenges given the caveats associated with detection by custom antibodies and epitope tags. The basic properties that have been investigated most extensively are the subcellular localization and membrane translocation of small membrane proteins. Initial tagging with larger tags such as green fluorescent protein suggested that small E. coli proteins can have various orientations in the membrane (with the N terminus on either the cytoplasmic or periplasmic side of the membrane) and be translocated by several different mechanisms (63). More recent studies of the E. coli 27-aa YohP and 33-aa YkgR modified with much smaller tags and assayed in vitro showed that the posttranslational insertion of these proteins into the membrane requires the signal recognition particle (64).

Given that many small proteins identified thus far are thought to function as regulators induced under specific conditions, there need to be mechanisms to downregulate the levels or activities of the small proteins when they are no longer needed. This downregulation could be accomplished by the binding of small molecules, amino acid modifications, and regulated proteolysis. Unfortunately, most global approaches to monitor protein modification or proteolysis are blind to the small proteins. Hence, these techniques will need to be adapted to specifically capture small proteins before more conclusions can be drawn about how proteins of this class themselves are regulated.

Some questions about fundamental properties relate to features of small proteins that may differ from those of larger proteins. These include questions about the translation and folding of the small proteins. For example, are there any barriers to the release of a small protein from the ribosome if the protein is shorter than the length of the ribosome exit channel? In the case of nested ORFs, does the translation of the longer ORF affect the translation of the embedded shorter ORF or vice versa? It is also not yet clear whether or how small proteins fold independent of the molecules that they bind. For some approaches such as nuclear magnetic resonance (NMR) spectroscopy, small size is an advantage, and several groups have initiated studies to systematically examine the structures of small proteins in isolation (65). However, these structures need to be viewed with some caution in the absence of information about whether the small proteins examined in vitro are functional. Mutational analyses that test specific predictions should also be carried out to provide support for the physiological relevance of these structures.

Related to questions about structure is the expectation that small proteins will have a more limited number of protein folds. However, the range of small protein folds and whether some structural features are more likely to be associated with small proteins are not yet known given the limited data available. Most of the small protein structures solved in complex with their interacting partners until now correspond to α-helices (40, 44, 66). However, these helices still can have unusual features. For example, it is noteworthy that a proline residue in the middle of the transmembrane helix of AcrZ introduces a kink that is critical for AcrZ binding to the AcrB transporter (40).

HOW DO SMALL PROTEINS EVOLVE?

Another intriguing question is how small protein genes evolve (Fig. 2). Generally, small proteins are not broadly conserved, so it appears that small protein genes can evolve rapidly. This assumption is supported by studies in Saccharomyces cerevisiae, where small protein genes that have arisen de novo increase fitness (67). Some small protein genes may be derived from the degeneration of a larger protein gene where only a small segment of the original ORF and translation signals remain intact (Fig. 2A). Alternatively, it is possible that a small region of a larger protein, such as a transmembrane domain, is duplicated and acquires a separate function as well as the ability to be expressed. Systematic assessments of whether small protein components of larger complexes are fused to a larger protein in some organisms and comparisons of synteny across closely related organisms could help to distinguish between these possibilities.

FIG 2.

Models for the evolutionary emergence and establishment of small protein genes in bacterial genomes. (A) A large protein-coding gene (gray) is expressed from the combined activities of a transcriptional start site (raised black arrow) and initiation codon (M). Small protein genes can evolve from this large ORF through the degeneration of the large ORF leaving only the signal for the expression and synthesis of part of the larger protein. This may occur after the duplication of the original large ORF. The small protein gene can also result from the duplication of just part of the large ORF. In this case, the synthesis of the small protein from this partial duplication may require the evolution of signals for transcription and translation. (B) Stochastic evolution of promoter and translation signals can lead to the expression of small proteins from random ORFs found in the genomic sequence. (C) Similarly, the stochastic evolution of a new proteolytic cleavage site can lead to the generation of a functional small peptide.

Small proteins may also be encoded by sORFs that arise by the de novo expression of a formerly noncoding nucleic acid sequence (Fig. 2B). Transcriptomic studies indicate that much of a genome can be transcribed. Additionally, there is increasing evidence for pervasive translation (reviewed in reference 12). Thus, there is not a high barrier to the chance occurrence of a combination of features resulting in the incidental expression of a functional small protein that confers a selective advantage. Sequence changes that result in the generation of a new protease cleavage site similarly can lead to the generation of a new functional small peptide (Fig. 2C).

It is interesting to ponder observations about small proteins in the context of evolutionary considerations. For example, although E. coli and Salmonella are quite closely related, the MgtB and MgtU small proteins are found only in Salmonella, and the Salmonella homolog of MgtS does not appear to act in the same way as the E. coli protein (32), even though all three of the small proteins bind to the MgtA magnesium importer. Strikingly, when an expression library allowing the expression of diverse synthetic proteins was used in selection for resistance to the antibiotic colistin, the six de novo colistin resistance-conferring (Dcr) peptides identified were hydrophobic and interacted with and activated the PmrB sensor kinase of the PmrAB TCS (68). Both of these findings suggest that it may be relatively facile for cells to evolve new small proteins, particularly transmembrane small proteins, with advantageous functions.

An even more challenging task is understanding the evolution of sORFs nested within larger ORFs, particularly since fewer examples have been characterized (18). The availability of more and more related genome sequences allowing expanded, fine-scale genome comparisons tuned to detecting additional conservation in intergenic regions as well as within ORFs could be a starting point for addressing some of these questions.

HOW CAN SMALL PROTEINS BE EXPLOITED?

Some of the characterized small proteins have already been shown to have medical consequences: the TA small proteins impact the occurrence of bacterial cells that persist under antibiotic exposure (reviewed in reference 69), and the modulins manipulate host cells. Small proteins are being detected in mixed microbial communities such as the human microbiome, which could well affect the composition of the microbiome and the interaction with host cells (9, 70). Additionally, small proteins that regulate transporters or regulators of key metabolic processes such as magnesium homeostasis could have important roles during bacterial infection (71). Thus, future experiments with an eye toward exploiting small proteins, which have a greater potential to cross membranes, should be considered in new antibacterial strategies as well as for optimizing bacteria for biosynthetic purposes.

OUTLOOK

As the discovery and functional assessment of small proteins are gaining momentum, a multitude of interesting and important questions await answers. Addressing these questions will require appropriate vigilance regarding the limitations of current approaches as well as the development of new technologies and more extensive studies of small proteins in organisms beyond E. coli and Salmonella enterica. However, even in the absence of global approaches, which generally have been more challenging for small proteins, the characterization of an expanding number of individual small proteins by many researchers studying various organisms undoubtedly will reveal more commonalities, general principles, as well as novel mechanisms. We expect that innovations in methods and ideas will drive an exponential increase in exciting fundamental discoveries about small proteins.