Microproteins: Overlooked regulators of physiology and disease

Ribosome profiling

Ribosome profiling, or Ribo-seq (ribosome sequencing), is a powerful method that generates an unbiased genome-wide snapshot of active translation which can be utilized to identify sORFs with protein-coding potential.³⁰ Ribosome “footprints”, typically ∼28–30 nucleotides in length, are generated from ribosome-bound mRNA that is protected from nuclease digestion (Figure 1A). These footprints are sequenced and mapped to the genome to identify regions undergoing active translation including putative sORF-encoded microproteins. Ribo-seq can capture instantaneous translation events in response to signals or stimuli, helping to identify sORFs that may only be translated under specific conditions.³¹ While Ribo-seq provides high sensitivity to map ORFs in the genome, these results must be experimentally validated as it has been demonstrated that ribosomes can associate with mRNAs without undergoing active translation.³² Despite this limitation, Ribo-seq is a robust tool that can be applied to identify sORFs, and there are ongoing efforts to increase the sensitivity of Ribo-seq to the single cell level to capture cell-specific transcripts,³³ which may aid in the identification of cell-type specific microproteins in the future.

Microprotein enrichment for MS

Mass spectrometry (MS) is the gold standard for protein identification and quantification. For MS analysis, proteins are first isolated from cells or tissue extracts and then processed by a series of steps (reduction, alkylation, trypsin digestion, de-salting, and peptide enrichment) that provide tryptic peptides for liquid chromatography with tandem MS (LC-MS/MS).³⁴ However, while MS is a powerful method used for protein identification, there are several factors that limit the ability of standard MS approaches to detect microproteins. First, many small proteins are often lost in the sample preparation steps leading up to MS.²⁷ Second, MS is inherently biased for large proteins that can produce hundreds of unique tryptic peptides, while microproteins often yield only a single peptide after protease digestion making their detection more difficult.^34,35 Third, many microprotein sequences are not contained in reference protein databases since they are not yet annotated, therefore unique peptide fragments often go unmapped.²⁹

To overcome some of these challenges, new microprotein enrichment techniques are being used to exclude larger proteins and enrich for microproteins prior to protease digestion to facilitate their identification (Figure 1B). Some of the most commonly used methods for microprotein enrichment include solid-phase extraction, urea-tricine polyacrylamide gel electrophoresis (PAGE), acid precipitation, and molecular weight cut-off (MWCO) filtration.^34,36 These methods have resulted in the identification of several microproteins,^37,38,39,40 and when combined with sequencing-based or bioinformatic approaches, they may be the most robust approach for high confidence microprotein discovery and identification.³⁴ This was recently highlighted with the implementation of comprehensive proteogenomic approaches combining peptidomics and RNA sequencing to filter out known peptides and exclusively identify new microproteins.^41,42 Lastly, it may be useful to implement data independent acquisition (DIA) during MS analysis. DIA analyzes the full spectrum detection range and can identify lower abundant peptides, such as microproteins, that are typically lost in traditional data dependent acquisition (DDA).^29,43

Proximity ligation-based assays

BioID proximity ligation methods have recently been used to identify microproteins and elucidate their interactomes.^44,45,46 These techniques make use of promiscuous biotin ligase enzymes, such as MicroID (TurboID)⁴⁴ or APEX,⁴⁵ which are fused to microproteins of interest or bait proteins and covalently biotinylate proteins in close proximity (1-10 nm) upon treatment with biotin (MicroID) or biotin phenol with hydrogen peroxide (APEX). Biotinylated proteins are subsequently purified using biotin-streptavidin-based systems and identified by MS analysis. For example, the APEX system was recently used to identify interacting partners of the uncharacterized microprotein C11orf98.⁴⁵ Additionally, a recent study elegantly applied the MicroID system to identify numerous microproteins and alternative proteins (alt-proteins) in subnuclear compartments both in vitro and in vivo⁴⁴ highlighting the ability of this method to be implemented in the discovery of microproteins when coupled with MS. While these proximity ligation assays have enabled the discovery of microproteins, it is important to note that these systems rely on treatment with biotin or hydrogen peroxide which are found endogenously and may yield false positives, thus necessitating experimental validation.⁴⁴

Publicly available bioinformatic tools and databases

Numerous bioinformatic tools and databases have been developed for the identification of microproteins and a subset of these tools is featured in Table 1. Additionally, a recent review highlighted many of the tools and databases that are publicly available²⁹ including SmProt⁴⁷ and sORFs.org,⁴⁸ which contain thousands of predicted sORFs with protein coding potential. However, concerns have been raised regarding the stark difference between both databases, though both used many of the same datasets for construction.⁵ This is presumably because each applied different criteria for classifying putative coding ORFs, further highlighting the need for more cohesive approaches for classifying sORFs. Collaborative efforts to standardize the annotation of translated sORFs are currently underway and will facilitate high confidence microprotein identification in the future.⁴⁹ Still, while there are a growing number of bioinformatic tools available to identify sORF-encoded microproteins, there is still a need to continue optimizing these tools and databases to specifically identify microprotein-coding potential in the genome.

Here, we will highlight one such tool, PhyloCSF, which is a conservation-based algorithm that has been successfully used to identify numerous microproteins.^{11,12,13,14,50,56,57,58,59} A hallmark feature of protein coding ORFs is their high degree of sequence conservation at the AA (codon) level, which differs from ncRNAs that are typically not highly conserved. These conservation signatures can be leveraged to identify sORF-encoded microproteins using PhyloCSF, which applies a phylogenetic sequence conservation algorithm to analyze and score codon substitution frequencies across >50 mammalian species.⁵⁰ PhyloCSF positively scores synonymous codon substitutions (those that encode the same AA) or AAs with similar properties (polarity, charge, hydrophilicity, etc.), while nonsynonymous or missense substitutions are scored negatively. Importantly, PhyloCSF can classify very short portions of coding sequences in isolation from the full sequence, which is necessary when considering individual exons and this inherent property of the program can be leveraged to help identify sORFs with microprotein-coding potential.^49,50 PhyloCSF is user-friendly and can be easily accessed and used via the UCSC Genome Browser.⁶⁰

Microprotein characterization and validation

It is important to note that while bioinformatic and computational approaches can be readily applied to assess microprotein-coding potential for genomic regions of interest, such predictions do not guarantee that identified coding ORFs will be translated into functional protein products. Therefore, identified microproteins must be validated by experimental approaches such as raising a sequence-specific antibody to detect the microprotein of interest. A caveat to antibody-based validation is that due to the inherent small size of microproteins, there are limited options for target antigen design for antibody production, and available sequences may not be suitable to generate strong antigenicity. Additionally, since many microproteins contain transmembrane domains or are associated with larger proteins, the antibody-binding site may not be readily available under native conditions for recognition using methods such as immunoprecipitation and immunocytochemistry. Alternatively, the CRISPR/Cas9 genome editing system can be used to insert a coding sequence for an epitope tag (e.g., HA, FLAG, myc) in-frame with the microprotein of interest within its endogenous locus via homology-directed repair (HDR).⁶¹ When using CRISPR/Cas9 HDR knock-in methods to generate epitope-tagged fusion proteins, it is ideal to screen knock-in clones with the tag inserted into the N- and C-terminus of the sORF of interest because the properties of the tags themselves may influence microprotein localization and function.⁶¹ Many of these methods have been successfully implemented to elucidate the microproteins described in detail below.

Plasma membrane

In addition to partitioning individual cells, the plasma membrane is responsible for mediating functions such as ion exchange, cell fusion, and signal transduction. Many of these processes are controlled by large protein complexes that are subjected to regulation by microproteins. For example, the Phe-Xaa-Tyr-Asp (FXYD) family (FXYD1-7) are type I single-pass transmembrane microproteins that regulate ion transport at the plasma membrane. FXYD proteins contain an invariant PFXYD motif in their extracellular N-terminal domain that enables them to serve as auxiliary subunits of the sodium (Na⁺)/potassium (K⁺) ATPase (NKA).^18,62 NKA is a plasma membrane enzyme that generates an electrochemical gradient for Na⁺ and K⁺ across the membrane by exporting 3 Na⁺ ions and importing 2 K⁺ ions for every ATP molecule hydrolyzed. This ubiquitous membrane transporter is essential for maintaining cell volume, pH, nutrients, and ion gradients that supply energy for secondary membrane transport. FXYD proteins were recognized for several decades as small proteins (61-95-AA, except FXYD5 at 178-AA) that could affect ion transport when overexpressed, though their function and interactions remained unknown until more recently.¹⁸ FXYD2 (γ-subunit of NKA) was the first member of the family defined to be a regulator of NKA. Expressed primarily in the kidney,⁶³ FXYD2 regulates NKA ion affinity depending on membrane potential. In addition, human mutations in FXYD2 have been linked to primary hypomagnesemia,⁶⁴ representing one of the few known instances of a pathology-causing microprotein. The other members of the FXYD family are expressed in a tissue-specific manner, allowing for precise regulation of NKA throughout the body. The cardiac-enriched FXYD1, also called phospholemman (PLM) is depicted in Figure 2A.

PLM is one of the best examples of the dynamic actions of microproteins and their ability to fine-tune cellular physiology. Originally described to interact with and regulate NKA, PLM was subsequently found to also be an endogenous regulator of the Na⁺/calcium (Ca²⁺) exchanger (NCX),¹⁹ which led to extensive investigation into its role in regulating skeletal and cardiac muscle contractility.⁶⁵ PLM activity, and therefore NKA and NCX function, is regulated by the phosphorylation state of PLM on cytoplasmic serine residues 63 and 68.^66,67 PLM is phosphorylated by protein kinase A (PKA) and protein kinase C (PKC) during stress and catecholamine release, which induces its dissociation from NKA and relief of inhibition. In contrast, phosphorylated PLM promotes inhibition of NCX via interaction with the intracellular loop.^66,68 The elegant nature of PLM activity is evident when viewed at the level of the cardiomyocyte: during stress responses, PLM is phosphorylated which relieves inhibition of NKA and decreases intracellular Na⁺.¹⁹ The relative loss of intracellular Na⁺ activates NCX, which in-turn reduces cytoplasmic Ca²⁺.¹⁹ However, because phosphorylated PLM inhibits NCX activity, intracellular Ca²⁺ loss is limited. These coordinated efforts simultaneously lower the risk for arrythmia while preserving contractility¹⁹ (Figure 2A).

Plasma membrane microproteins function beyond ion exchange to affect other critical processes such as fusion and cell signaling during development. For example, the 63-AA microprotein NEMEP (Nodal Enhanced MEsendoderm Peptide) is a transmembrane microprotein that helps regulate early mesendoderm development.⁶⁹ Discovered as a putative ncRNA target of Nodal, NEMEP localizes to the plasma membrane and interacts with the glucose transporters GLUT1 and GLUT3 to directly enhance glucose transport into cells. NEMEP loss-of-function studies demonstrate impaired mesodermal differentiation and glucose uptake in vitro and mesodermal developmental defects in vivo.⁶⁹

Similarly, the 84-AA microprotein myomerger,⁷⁰ also called myomixer⁷¹ and minion,⁷² is important for myoblast fusion during muscle formation. It is now known that two myoblast proteins, myomerger and the 221-AA protein myomaker, regulate fusion in a stepwise manner, requiring first myomaker on both cells and then myomerger on only one cell to complete the process.⁷³ Myomerger-null mice do not undergo myoblast fusion which leads to neonatal lethality,^70,71,72 marking myomerger as a microprotein that is essential for muscle formation and viability.

Endo/lysosomes

Endocytosis is a critical fundamental cellular process that utilizes membrane-bound vesicles to partition functions such as signal transduction, energy sensing, autophagy, and exogenous protein digestion within the endo/lysosome.⁷⁴ Microproteins contribute to several distinct aspects of these processes, such as the 83-AA cancer associated small integral membrane open reading frame 1 (CASIMO1). Using a microarray-based approach, CASIMO1 was discovered as a sORF-encoded microprotein upregulated in breast cancer samples.⁷⁵ Functional characterization revealed co-localization of CASIMO1 with the late endosomal marker LAMP1, leading to the discovery of CASIMO1’s role in actin cytoskeletal organization, cell migration, and proliferation, which are key features of cancer progression and prognosis.

SPAR is another endo/lysosomal-associated microprotein that was discovered through MS-based proteomic screening of ncRNAs.^76,77 The ncRNA LINC00961 was revealed as a candidate microprotein and validated to code for a 90-AA protein that localizes to the late endo/lysosome and interacts with the membrane-associated v-ATPase hydrogen (H⁺) pump. Results further showed SPAR as an inhibitor of mTORC1, a key player in cell growth and proliferation in response to skeletal muscle injury⁷⁸ (Figure 2B). SPAR is now recognized to have important implications for skeletal muscle regeneration and continues to be investigated as a regulator of tissue injury response.

Several human microproteins have been identified to have Drosophila melanogaster homologs, including the 88-AA hemotin⁷⁹ which functions at the late endosome. Hemotin is a necessary component of the Drosophila immune system via regulation of phagocytosis in macrophage-like cells, a function which is shared with its human homolog, stannin. Stannin promotes endosomal maturation and phagocytic processing in macrophages,⁷⁹ confirming the evolutionarily conserved function of hemotin/stannin in organismal immunity.

Sarco/endoplasmic reticulum

The S/ER regulates cellular homeostasis on multiple fronts including protein synthesis, protein folding, and storage of intracellular Ca²⁺. Ca²⁺ is a ubiquitous second messenger that participates in numerous cell-signaling events upon its release from the S/ER into the cytosol.⁸⁰ Ca²⁺ is actively transported back into the S/ER via the S/ER Ca²⁺ ATPase (SERCA) to restore basal cytoplasmic Ca²⁺ levels. SERCA enzymes are large (994–1042 AA) P-type ATPase pumps that are embedded in the S/ER membrane and their activity is potently regulated by a family of microproteins termed the “SERCA-regulins” including PLN (52-AA), ALN (65-AA), ELN (56-AA), MLN (46-AA), and SLN (31-AA)^{12,20,21,22,23} (Figure 2C). The SERCA-regulins are co-expressed with different SERCA isoforms (SERCA1a, SERCA2a, SERCA2b, and SERCA3) in a tissue-specific manner.^12,81 These small proteins share sequence homology within their TM domain, which enables direct interaction with SERCA.¹²

The potent activity of these SERCA-regulins can be appreciated in the context of cardiomyocyte contractility. During cardiac contraction, Ca²⁺ is rapidly released from the S/ER, which induces sarcomere shortening and activates SERCA2a (the primary SERCA gene product in the heart), which uses the energy generated from ATP hydrolysis to pump Ca²⁺ back into the S/ER to induce cardiac relaxation.⁸² This process is highly regulated by the cardiomyocyte-expressed SERCA-regulins (PLN, SLN, ALN). Under normal physiological conditions, a population of SERCA2a is inhibited by PLN to maintain a “cardiac reserve” for situations where heart contractility must be rapidly increased. In response to catecholamine induced β-adrenergic signaling, PLN is phosphorylated on its cytoplasmic serine 16 residue by PKA, causing dissociation from SERCA2a. This alleviates its inhibitory effects and potently increases SERCA2a Ca²⁺ transport activity and cardiomyocyte contractility (i.e., the “fight-or-flight” response).²² Notably, PLN-deficient mice exhibit maximal myocardial contractile performance at baseline and completely lose the ability to respond to β-adrenergic signaling.⁸³

In addition to the SERCA-regulins, dwarf open reading frame (DWORF, 34-AA) has been identified as a positive regulator of SERCA2a in cardiac and slow-skeletal muscle.¹⁴ The exact mechanisms that regulate DWORF’s stimulatory effects on SERCA2a are not yet fully understood, but DWORF has been shown to act in an antagonist manner to PLN by binding to the same residues on SERCA2a and actively displacing PLN from the pump to relieve its inhibitory effects, resulting in an increase in SERCA2a Ca²⁺ affinity.^14,84 Additionally, new evidence from heterologous cell culture systems suggests that DWORF may also directly stimulate the maximal Ca²⁺ transport activity (V_max) of SERCA2a.^85,86 Given its potent stimulatory effects on SERCA2a and cardiomyocyte contractility, DWORF has recently emerged as a potential heart failure therapeutic.^84,87,88 Additionally, there has been a recent report of another SERCA-interacting protein, 48-AA pTUNAR, that enhances SERCA activity during neural differentiation.⁵⁸ Both DWORF and pTUNAR directly interact with SERCA, but their transmembrane domains lack the SERCA-inhibitory motif that is present in the SERCA-regulins,^12,58 thus indicating that their unique transmembrane domains may mediate their distinct ability to activate SERCA.

Mitochondria

A large number of microproteins that have been identified thus far localize to the mitochondria.^89,90 Mitochondria are the primary organelles responsible for integrating cell signaling and survival cues to match changing metabolic and energetic demands. Several studies have identified microproteins that affect various aspects of tissue and systemic metabolism, implicating these microproteins in health and disease. One such protein is mitoregulin (Mtln),^91,92 also known as MOXI (micropeptide regulator of β-oxidation).¹³ Previously annotated as a non-coding RNA (LINC00116), three independent groups discovered the 56-AA mitoregulin and defined its role in fatty acid oxidation, lipid metabolism, and respiratory chain activity in heart and skeletal muscle.^13,91,92 Additional studies have described a role for mitoregulin in regulating lipolysis and mitochondrial β-oxidation in adipocytes⁹³ and contributing to systemic lipid metabolism.⁹⁴ The exact molecular mechanisms driving mitoregulin function are still under active investigation; however its involvement in lipid and fatty acid metabolism is well supported, suggesting it could be developed as a target for modifying these processes in disease.

Numerous studies have defined microproteins as essential core components, accessory subunits, and assembly factors for the electron transport chain (ETC).^{56,89,91,95,96} This is not surprising given that the ETC complexes are known to contain a disproportionately high number of small proteins. In fact, compared to the total cellular proteome where proteins under 100 AAs make up <2% of the known proteome, they constitute ∼28% of ETC proteins.⁸⁹ Microproteins that have been recently identified that are directly involved in ETC function and supercomplex assembly (i.e., respirasomes) include BRAWNIN (71-AA)⁸⁹ UQCC3 (93-AA),⁹⁶ UQCC4 (132-AA),⁹⁵ SMIM4 (70-AA),^95,97 mitolamban,⁵⁶ and mitoregulin⁹¹ (Figure 2D). Collectively, these studies highlight the important roles that microproteins play in the formation of the functional ETC and respiratory supercomplexes, processes that are currently not fully understood.

In addition to contributing to metabolism and respiration, microproteins have been implicated in organelle-organelle communication at the mitochondria. Such is the case for the 54-AA microprotein PIGBOS, which facilitates interactions between the outer mitochondrial membrane (OMM) and mitochondrial-associated ER membrane (MAM) during the stress-induced ER unfolded protein response (UPR).⁵⁷ The UPR has been implicated in numerous neurodegenerative diseases such as Alzheimer’s, Parkinson’s, amyotrophic lateral sclerosis, and others^98,99 and is thus an area of active investigation for targeting disease. PIGBOS increases resistance to ER-stress and apoptosis by interacting with the ER membrane protein chloride-channel CLIC-like 1 (CLCC1), suggesting a unique role for PIGBOS in inter-organelle communication and the cellular stress response.

Another mitochondrial microprotein is the anti-tumorigenic 133-AA miPEP133.¹⁰⁰ miPEP133 is translated from its microRNA (miRNA) precursor, miR-34a, which is known to suppress transcription of oncogenes in multiple cancers.^101,102,103 miPEP133 itself functions independently in an anti-tumorigenic positive-feedback loop to increase transcription of p53 and miR-34a. miPEP133 promotes anti-cancer effects by interacting with HSPA9 in the mitochondria where it inhibits interactions between HSPA9 and its binding partners. This causes changes in membrane potential, ATP production, and the cell cycle that induces apoptosis and limits tumorigenesis. miPEP133 levels correlate with favorable cancer prognosis, emphasizing the relevance of this microprotein as a prognosis factor.

Cytoplasm

One of the earliest discovered microproteins was humanin, a 24-AA secreted protein that has been shown to be beneficial in Alzheimer’s disease (AD).¹⁰⁴ Humanin is encoded by the mitochondrial genome, with its ORF found within the mitochondrial 16S rRNA gene. The beneficial effects of humanin in AD have been linked to its anti-apoptotic activity and prevention of neuronal cell death both in the cytoplasm and via cell surface receptors. Within the cytoplasm, humanin binds and stabilizes Bax, preventing downstream apoptotic signaling of the Bax/Bcl family of apoptosis-inducing proteins.¹⁰⁵ Humanin also exerts cytoprotection through binding to cell surface receptors including the cytokine receptor complex CNTFR-a/WSX-1/gp130¹⁰⁶ and the G-coupled protein receptor FRPL1.¹⁰⁷ The interest in humanin as a promising therapeutic expands beyond AD to other neurodegenerative diseases, diabetes, and cardiovascular disease among others.^108,109,110

The 68-AA microprotein NoBody functions within the cytoplasm to regulate mRNA stability.¹¹¹ Previously annotated as the ncRNA LINC01420, NoBody is expressed in multiple cell lines and shows high levels of AA conservation in mammals. NoBody localizes to processing-bodies (P-bodies) in the cytosol, which regulate mRNA degradation and translation repression.¹¹² P-bodies achieve partitioning though liquid-liquid phase separation, raising interesting questions about NoBody as a microprotein that functions within a cellular compartment that lacks a membrane. NoBody interacts with the mRNA decapping complex protein EDC4, leading to enhanced mRNA decapping, decreased mRNA stability, and increased mRNA degradation. Therefore, there is an inverse correlation between NoBody levels and the number of P-bodies, leading to effects on mRNA stability and overall gene expression.

Nucleus

The nucleus is the home of many highly regulated processes, and therefore, it is not surprising that numerous microproteins localize to this site of complex cellular function. Two such examples include the splice isoforms MRI-1 (modulators of retrovirus infection homolog 1) and MRI-2 which modulate distinct pathways of DNA repair. MRI-1 was first described as a 157-AA modulator of retroviral infection,¹¹³ though later studies discovered its role in regulating DNA repair, leading to its renaming to CYREN (Cell cYcle REgulator of NHEJ).¹¹⁴ CYREN suppresses the error-prone and possibly genotoxic non-homologous end-joining (NHEJ) DNA repair pathway in favor of the more accurate homologous recombination (HR). CYREN achieves this by inhibiting the NHEJ proteins Ku70 and Ku80, allowing HR to outcompete NHEJ in the S/G2 cell cycle phases. In contrast, its splice isoform MRI-2 codes for a 69-AA microprotein that localizes to the nucleus and associates with the Ku-DNA NHEJ complex.¹¹⁵ In instances of DNA double-strand breaks, MRI-2 increases the rate of NHEJ to promote DNA repair and prevent apoptosis. CYREN and MRI-2 represent microprotein splice isoforms that independently regulate DNA repair and overall cell survival.

Another microprotein that localizes to the nucleus is the 87-AA pTINCR, so named due to its associated lncRNA, TINCR, which plays a role in epithelial differentiation.^59,116 In mechanisms distinct from TINCR, pTINCR was found to promote epithelial differentiation by contributing to post-translational modification of CDC42. CDC42 is an epithelial pro-differentiation factor that has been associated with oncogenic phenotypes. However, the interaction of pTINCR with CDC42 was shown to be anti-tumorigenic and pro-differentiating, thus marking it as an anti-oncogenic factor for many epithelial cancers.

Targeting microproteins in heart failure, obesity, and diabetes

There are several microproteins that are abundantly expressed in the heart and regulate important processes such as Ca²⁺ homeostasis (the SERCA-regulins and DWORF), energy metabolism (MOXI, BRAWNIN, UQCC3, UQCC4, SMIM4, Mitolamban, MOTS-c), and ER-stress (PIGBOS). Therefore, there is great potential for these microproteins to serve as therapeutic targets to treat heart failure and obesity.

Targeting microproteins in cancer

Numerous microproteins have been implicated in cancer as key players in cell proliferation, tumor suppression, invasion, and metastasis.^75,137,138 Here we describe several candidate microproteins that represent potential therapeutic targets using overexpression (CIP2A-BP, MIAC, MP31) or inhibition (ASAP, APPLE) strategies (Figure 3).