Use of Proximity Labeling in Cardiovascular Research.

Protein-protein interactions are of paramount importance in regulating normal cardiac physiology. Methodologies to elucidate these interactions in vivo have been limited. Recently, proximity-dependent biotinylation, with the use of BioID, TurboID, and ascorbate peroxidase, has been developed to uncover cellular neighborhoods and novel protein-protein interactions. These cutting-edge techniques have enabled the identification of subcellular localizations of specific proteins and the neighbors or interacting proteins within these subcellular regions. In contrast to classic methods such as affinity purification and subcellular fractionation, these techniques add covalently bound tags in living cells, such that spatial relationships and interaction networks are not disrupted. Recently, these methodologies have been used to identify novel protein-protein interactions relevant to the cardiovascular system. In this review, we discuss the development and current use of proximity biotin-labeling for cardiovascular research.

Proximity proteomic methodologies have been developed to enzymatically label nearby proteins while still maintaining information about the proteins’ location and interactions with other proteins within a cell (1, 2, 3, 4, 5, 6). With the rapid advancement of proximity labeling technologies comes the ability to collect substantial data and answer longstanding scientific questions that would otherwise have been impossible. In this review, we discuss using these techniques for cardiovascular research, highlighting the use of proximity labeling and quantitative mass spectrometry to identify the mechanisms underlying adrenergic regulation of Ca2+ current in the heart (7).

Development of Proximity-Dependent Biotinylation Methodology

With the advent of transcriptomics and the development of extensive lists of candidate genes, identifying critical protein interactors has remained a serious bottleneck in advancing our understanding of physiology and pathophysiology. To elucidate the components of macromolecular complexes, biochemical fractionation followed by mass spectrometry can be performed (8, 9, 10). Another approach uses affinity purification of a protein, typically by antibody, followed by mass spectrometry (11). These methodologies, however, have important limitations (12,13). Some proteins, especially membrane-bound proteins, are relatively inaccessible and difficult to solubilize. Coimmunoprecipitations require high-quality antibodies that are often not available. Important interactions may have fast-off kinetics and therefore will not persist after solubilization and during affinity purification (14). Also true of yeast-hybrid approaches, the biological process of interest may only operate in the proper milieu, as part of a larger more extensive protein interaction network that only exists in terminally differentiated mammalian cells. In addition, tissue-specific post-translational modifications can reduce the effectiveness of these methods.

Cognizant of these limitations, Roux et al. (15) adapted an Escherichia coli biotin ligase, BirA, to perform in situ labeling in eukaryotic cells. This system has several advantages. Posttranslational modification with biotin, that is, vitamin B7, was chosen in part for its low toxicity and frequent use, alongside its ultra-high-affinity binding partner streptavidin, in molecular biology and biochemistry. There is also a relative paucity of biotin in eukaryotic cells: generally, there is only a single biotin ligase per species and these are known for their exquisite specificity, targeting only a single lysine residue in up to 5 different protein targets depending on the species (16). Roux et al. (15) fused different variants of BirA to lamin A, a nuclear envelope intermediate filament protein, mutation of which can cause a number of diseases, including dilated cardiomyopathy (17). In HeLa and human embryonic kidney (HEK) 293 cells, BirA R118G, also known as BirA∗ and named BioID for “biotin identification,” was especially promiscuous. This variant was capable of a marked increase in nonspecific but proximity-based biotin ligation with the addition of excess biotin to the culture media (Figure 1A) (15). Combined with streptavidin-based affinity purification of biotinylated proteins and liquid chromatography/mass spectrometry, numerous known and novel interactors of lamin A were identified (Central Illustration). Depending on the experimental conditions and labeling time, the proximity-labeling radius may be as low as a few nanometers (4). For these reasons, BioID has been used in diverse model systems, and within 4 years of its introduction it was coupled with injection of adeno-associated virus to characterize a tissue-specific subcellular proteome in mice in vivo: postsynapse cortical inhibitory neurons in newborn mice (18).

Figure 1

Proximity Proteomics Labeling Methods Used in Cardiovascular Studies

(A) Biotin labeling of sterically available basic residues by the Escherichia coli biotin ligase BirA, or engineered derivatives (BirA∗ [BioID], BioID2, TurboID, and miniTurbo). BirA∗ is conjugated to the protein of interest. With biotin supplementation, biotin is covalently linked to both interacting and neighboring proteins. (B) Soybean ascorbate peroxide (APEX) or its derivative (APEX2) is conjugated to the protein of interest, which catalyzes the generation of biotin phenoxyl radicals that form covalent linkages with sterically available electron-rich residues. Proteins within the “cloud” of biotin phenoxyl radicals are labeled with biotin, whereas those outside of the radius are not labeled. (C) Complementation of N- and C-terminal APEX2 fragments, “AP” and “EX,” respectively, induced by rapamycin-mediated interaction between FK506 binding protein (FKBP) and its binding partner, the FKBP-rapamycin binding (FRB) domain of the kinase mammalian target of rapamycin. Biotin labeling of neighboring proteins is enabled in cellular locations where both protein A and protein B interact. Figure created with BioRender.com.

Central Illustration

Proximity Labeling Approaches Used in Cardiovascular Research

Proximity labeling and quantitative mass spectrometry have been used in cardiomyocytes to elucidate the interactome and neighborhood of proteins such as voltage-gated Ca2+ and Na+ channels (CaV1.2 and NaV1.5, respectively), junctophilin (JPH2), N-Cadherin (CDH2), and titin. They have also been used to elucidate the interactome of junctional adhesion molecule C (JAM-C) and K+ channels (Kir2.1) in cultured cells (not pictured). A workflow schematic for proximity labeling experiments, with α1C-APEX as an example, follows.

The next generation of biotin ligases overcame at least some of the limitations of BioID, especially its slower kinetics, which requires longer labeling times. Branon et al. (4) used error-prone polymerase chain reaction and 5 generations of yeast display-based directed evolution to produce a more promiscuous and faster BirA, TurboID, capable of markedly greater biotin labeling in heterologous expression and in live Caenorhabditis elegans and Drosophila melanogaster. An alternate fifth-generation BirA contained an N-terminal truncation with slightly less activity than the full-length form but also decreased labeling in the absence of biotin supplementation (referred to as miniTurbo). This has important implications in the development of transgenic animals. For example, the fraction of fruit flies expressing miniTurbo surviving to adulthood did not differ from control; however, fewer than one-half of TurboID-expressing flies survived to adulthood, an effect that was completely eliminated with dietary supplementation of biotin.

One important limitation of BioID proximity labeling is the time required to detectably biotinylate neighboring proteins—hours to days of biotin supplementation. This limitation has been somewhat reduced with the advent of TurboID, although this engineered biotin ligase still requires a minimum of several minutes to biotinylate protein neighbors (4).

The introduction of soybean ascorbate peroxidase (APEX) and its engineered derivatives represented a major advance in the field with the generation of significant enzyme-catalyzed proximity labeling within seconds, allowing the capture of dynamic signaling events at high resolution with subcellular localization (1,2). APEX2 was engineered with yeast display-based evolution to improve its catalytic efficiency, increasing sensitivity in cells with lower transgene expression. In the presence of biotin-tyramide (biotin-phenol) and hydrogen peroxide, these peroxidases form biotin-phenoxyl radicals that form covalent linkages to electron-rich residues (primarily tyrosines; Figure 1B) (1). The short half-life of these radicals (<1 ms) and their rate of diffusion ensures that labeling occurs only within ∼20 nm of APEX (19). Addition of an antioxidant cocktail quenches the reaction. The membrane impermeability of these radicals ensures that proximity labeling is limited to membrane-contained or membrane-excluded spaces.

There are several limitations, namely that the labeling by APEX2 requires hydrogen peroxide, which can induce brief oxidative stress, and that the high labeling activity of APEX2 may preclude the precise spatial resolution required to identify the local interaction network of a target protein (20). For example, clouds of APEX2-generated biotin-phenoxyl radicals may surround APEX2 fusion proteins in various subcellular compartments, not limited to the ultimate trafficking destination of the protein of interest, tagging bystander proteins along the way (e.g., ribosome and rough endoplasmic reticulum [ER], smooth ER, Golgi network, and so on). Proteins in these compartments may be important for folding, trafficking, and/or degradation, but deconvolving these neighborhoods is challenging. To address these limitations, quantitative proteomics were combined with either spatial references (20) or isobaric tandem mass tags (21) to enable tracking the activation, internalization, and subcellular protein complexes of G-protein–coupled receptors when expressed in HEK cells.

Use of Apex-Based Proximity Proteomics in the Heart

Ca2+ channels, which are the major pathway for entry of Ca2+ ions into excitable cells, transduce membrane depolarization into multiple cellular functions (22). In the heart, the influx of Ca2+ is essential for activation of excitation-contraction coupling and modulating pacemaker activity in sinoatrial nodal cells. In adult ventricular cardiomyocytes, the predominant type of Ca2+ channel, CaV1.2, localizes to transverse tubules in close proximity to ryanodine receptors at dyadic junctions (23). Despite considerable research, the molecular mechanisms responsible for adrenergic stimulation of Ca2+ influx via voltage-gated Ca2+ channels have remained stubbornly obscure for decades (24, 25, 26, 27, 28, 29, 30). For many years, the mechanism was thought to involve protein kinase A (PKA) phosphorylation of residues on CaV1.2 α1C- and/or β2B-subunits (31, 32, 33, 34, 35, 36, 37). More recent studies in mice demonstrated, however, that adrenergic stimulation of CaV1.2 did not involve direct phosphorylation of these residues on α1C or β2 subunits (Figure 2A) (7,38, 39, 40, 41, 42, 43, 44).

Figure 2

Proximity Labeling Methodology to Define Mechanisms of Adrenergic Modulation of Ca2+ Channels

(A) Schematic depicting adrenergic regulation of CaV1.2 in heart. A β-agonist binds to the β-adrenergic receptor, thereby activating protein kinase A (PKA). Before discovering the elusive protein via proximity proteomics, the target of PKA was unknown. ATP, adenosine triphosphate; cAMP, cyclic adenosine monophosphate. (B) Generation of transgenic mice with ascorbate peroxide (APEX2) fused to the N-terminus of either the α1C or the β2B subunits. (C) Proximity labeling can be completed in isolated myocytes or in intact whole heart ex vivo through retrograde perfusion of biotin-phenol and H2O2. (D) Schematic shows the localization of APEX2-conjugated CaV1.2 channels in the dyadic space of cardiomyocytes. After β-adrenergic receptor activation, PKA phosphorylation of Rad depletes Rad from the CaV1.2 complex, making it less accessible to APEX2-mediated biotinylation. Figure created with BioRender.com.

There are many hundreds of proteins localized to the dyadic space, many of which could transiently interact with the Ca2+ channel and serve as the functional PKA target (Figure 2A). Based on the observations that β-adrenergic regulation of CaV1.2 cannot be reconstituted when the core subunits are heterologously coexpressed (45), that PKA phosphorylation of the α1C- and/or β2B subunit is not sufficient (43), and that CaV1.2 is the epicenter of a large macromolecular complex, we aimed to identify proteomic neighbors of CaV1.2 in the heart that may be the true PKA target. The specialized ultrastructure of the cardiac dyad is spanned and maintained by junctophilin-2 (Jph2), which brings the sarcoplasmic reticulum and its Ca2+ release channels within 12–18 nm of the transverse tubules and its membrane-bound CaV1.2, permitting excitation-contraction coupling with every single heart beat (46, 47, 48, 49). It is therefore a membrane-enclosed space well suited to proximity-based proteomic characterization with the use of proximity labeling.

Transgenic mice with doxycycline-inducible cardiomyocyte-specific expression of dihydropyridine-resistant α1C or β2B proteins with APEX2 and a V5 epitope conjugated to the N-terminals were generated (Figure 2B). This inducible transgenic methodology (50,51) was perfectly suited for these experiments as we could both temporally and spatially limit expression of modified channels in cardiomyocytes (38,52,53). The subcellular localization and function of CaV1.2 in cardiomyocytes was not affected by fusing APEX2 to α1C and β2B (7). Importantly, the Ca2+ currents imparted by these APEX-conjugated Ca2+ channels were stimulated by β-adrenergic agonists, implying that conjugating APEX2 to the channels did not perturb PKA-mediated regulation of this process. Incubating isolated ventricular cardiomyocytes with biotin-phenol for 30 minutes followed by exposure to hydrogen peroxide for 1 minute induced biotinylation of proteins in a striated Z-disk pattern, consistent with proper localization of Ca2+ channels within the T-tubules of cardiomyocytes (Figures 2C and 2D). Affinity purification of the biotinylated proteins was then performed with the use of streptavidin-sepharose (Figure 3). The high affinity of biotin and avidin enables washing of the avidin-sepharose under denaturing conditions (3). Via mass spectrometry, we identified nearly all previously reported members of the cardiac CaV1.2 complex, as well as many proteins known to be in close proximity to CaV1.2 (7). A detailed description of sample preparation has been thoroughly reviewed elsewhere and is beyond the scope of this review (2,3,54).

Figure 3

Post-Biotinylation Processing and Analysis

After ascorbate peroxide–mediated biotinylation, the biotinylation is quenched and cells are lysed. Biotinylated proteins are affinity purified with the use of streptavidin magnetic beads and washed with denaturing conditions. Peptides are identified by means of mass spectrometry. Figure created with Bio.Render.com.

We found that proximity labeling of the Ca2+ channel neighborhood can also be performed in beating hearts, capturing the full effect of extracellular matrix, myocyte-myocyte, and myocyte-nonmyocyte interactions, the importance of which has become increasingly recognized over time (55,56). For these experiments, Langendorff-perfused hearts were retrograde perfused with Krebs solution for 10 min before the addition of biotin-phenol for 15 min (Figure 2C). Electrocardiograms were monitored throughout the infusion protocol to ensure viability of the preparation. Streptavidin–fluorescein isothiocyanate immunofluorescence of tissue sections of Langendorff-perfused hearts also showed biotinylation of proteins in a striated Z-disk pattern (7).

In our APEX2-CaV1.2 experiment, we also labeled nuclear, mitochondrial, ribosomal, ER, and Golgi proteins (7). This is because α1C and β2B APEX fusion proteins not only exist in the transverse tubule, but rather are constantly being created at the ribosomes and trafficked there from the rough ER. Wherever there are transgenic channels in the cell, nearby proteins are biotinylated. Moreover, it is challenging to differentiate direct physical interactors from bystanders even within the subcellular area of interest. Suspecting that the transducer of adrenergic regulation of Ca2+ channels would alter its interaction with CaV1.2 after exposure to β-adrenergic agonist, whereas most bystander proteins would not, we compared the CaV1.2 neighborhood in the absence and presence of isoproterenol in isolated cardiomyocytes and Langendorff-perfused beating hearts, similarly to studies using heterologously expressed G-protein–coupled receptors (20,21).

We preincubated isolated cardiomyocytes from α1C-APEX and β2B-APEX mice with biotin-phenol for 30 min, and during the final 10 min the cells were also exposed to either isoproterenol or vehicle (Figure 2D). The biotinylated proteins were purified and quantified with the use of tandem mass tag triple-stage mass spectrometry (TMT SPS MS3) (Figure 3). We also probed the effect of isoproterenol using Langendorff-perfused whole hearts. We were careful only to include in our analysis hearts with demonstrable adrenergic activation, assessed with the use of electrocardiography and phosphorylation of phospholamban. The protein quantification changed for several proteins: We detected an isoproterenol-induced recruitment of the PKA catalytic subunit to CaV1.2 and a 30% to 50% decrease in the amount of the small Ras-like G protein Rad in the neighborhood of CaV1.2. Ten-minute exposure of isolated nontransgenic control cardiomyocytes to isoproterenol had a minimal effect on proteins quantified with TMT SPS MS3 and specifically no effect on Rad compared with untreated cardiomyocytes. These experiments indicate that β-adrenergic stimulation depletes Rad from the neighborhood of Ca2+ channels, but not from the cell as a whole.

Rad is a member of the Rad, Gem, Kir family of GTP-binding proteins that are well known for their capacity to potently inhibit all high-voltage-activated Ca2+ channels (57, 58, 59, 60, 61, 62, 63). Applying the adenylate cyclase activator forskolin to HEK293T cells expressing Rad and wild-type α1C and β2B increased the maximal conductance and shifted the V50 for activation of barium currents, consistent with PKA activation of CaV1.2. We found that β-adrenergic regulation of CaV1.2 requires Rad binding to the CaVβ subunit under basal conditions and release of Rad binding to β after adrenergic stimulation. Taken together, Rad potently dampens CaV1.2 current while phosphorylation of Rad allows channels to operate as though they were devoid of Rad, similar to the way phospholamban regulates sarco/endoplasmic reticulum Ca2+-ATPase (64). This study demonstrated the utility of proximity labeling in vivo, especial for defining physiological mechanisms that could not be reconstituted heterologously.

Recently, we generated transgenic mice with doxycycline-inducible cardiomyocyte-specific expression of tetrodotoxin-sensitive NaV1.5 proteins with APEX2 and a V5 epitope conjugated to the N-terminus (Figure 4) (65). The NaV1.5 subcellular localization and function, as assessed by cellular electrophysiology and anti-V5 antibody immunofluorescence were not affected by fusing APEX2 to the Na+ channel. We incubated isolated ventricular cardiomyocytes with biotin-phenol and then hydrogen peroxide, which induced robust biotinylation of proteins at the sarcolemma, intercalated disk, and in a striated Z-disk pattern, consistent with the pattern of localization of the NaV1.5 channels. Immunoblots of streptavidin-purified biotinylated proteins showed both calmodulin and fibroblast growth factor (FGF) 13, which was consistent with prior studies showing that calmodulin and FGF13 interact with NaV1.5 (66, 67, 68, 69).

Figure 4

Proximity Proteomics Performed in Cardiomyocytes

Schematic of dyadic space of cardiomyocyte, showing position of T-tubules, sarcoplasmic reticulum, and actin-myosin filaments and titin. Proximity-labeled studies have been performed with the use of ascorbate peroxide derivative (APEX2)–conjugated CaV1.2 channels, BioID-conjugated junctophilin (Jph2), N-cadherin (CDH2), and titin. RyR = ryanodine receptor. Figure created with Bio.Render.com.

Proximity-Dependent Biotinylation Using BioID in the Heart

The dyadic junction proteome was further characterized with in vivo proximity labeling of the Jph2 subdomain proteome (Figure 4) (70). Targeted knock-in of engineered BioID (BioID2) fused to Jph2 had no adverse effects on Jph2 expression and function. Intraperitoneal biotin injection once daily for a week at a dose of 24 mg/kg body weight enabled proximity labeling within the dyad. Using label-free mass spectrometry, 550 proteins were identified in the avidin-purified samples from the homozygous knock-in mice. The top cellular components, based on gene ontology enrichment analysis, were sarcolemma, cation channel complexes, and transverse tubules. The authors also identified potentially novel cardiac dyad proteins that may play roles in cardiac dyad formation, maintenance, and function (70).

In another study, BioID was coupled to the structural protein titin (Figure 4) (71). Rudolph et al. (71) generated knock-in mice with BioID spliced into titin exon 28, corresponding to the location of the Z disk–I band transition. Mice were continuously supplied with supplemental biotin dissolved in their water. The authors compared this subdomain of the titin proteome, described as the sarcomeric proteome, in quadriceps and heart, finding 100 proteins in common (81% of labeled proteins in the heart and 41% in skeletal muscle). They reported that 5% of isolated protein was biotin labeled and, using titin’s immunoglobulin-like domains as reference points, calculated that the BioID labeling radius was 7–15 nm. As in the Jph2-BioID knock-in, no untoward phenotypic or functional effects of the knock-in were noted.

In another study, BioID was fused to N-cadherin and was expressed in neonatal murine cardiomyocytes using an adenovirus (Figure 4) (72). The N-cadherin–BioID fusion localized to the cell-cell contacts, as would be expected, and after biotin was added to the cell culture for 24 hours, proteins along the cell-cell contacts and Z disks were labeled. Mass spectrometry identified 917 proteins, which after setting a threshold fold change of ≥10 and p < 0.001 was reduced to a final list of 365 proteins, comprising primarily adaptor and adhesion proteins that promote junction specialization (72).

The BioID methodology has also been applied to elucidating the interactome of heterologously expressed wild-type Kir2.1 and a trafficking-deficient type 1 Andersen-Tawil syndrome mutant Kir2.1δ314–315 channel in HEK cells (73). Kir2.1 is an inward rectifying K+ current, playing important roles in setting the membrane potential and controlling excitation and action potential duration of ventricular cardiomyocytes (74). Mass spectrometry identification of avidin-purified biotinylated proteins revealed a total of 218 high-confidence Kir2.1 neighbors, including 75 proteins that were preferentially proximate with wild-type channels, 66 proteins that were preferentially proximate with the mutant Kir2.1 channels, and 77 proteins that were proximate with both wild-type and mutant Kir2.1 channels (73). Based on these BioID-identified interaction studies (73), the authors concluded that the pathogenesis associated with the Kir2.1Δ314–315 mutation involves malfunctions in the mechanisms governing Kir2.1 trafficking from the Golgi to the cytoplasmic membrane, in agreement with prior studies (75, 76, 77). Several desmosome-associated proteins, including plakophilin-4 (PKP4) were found to interact with both wild-type and Kir2.1Δ314–315 channels in HEK cells. Kir2.1 and PKP4 colocalize at intercalated disks and Z disks in cardiomyocytes, and PKP4 is a positive regulator of Kir2.1 current density when coexpressed in HEK cells (73). Taking these findings together, this study suggested that proximity labeling may reveal how and why specific mutations affect the trafficking and function of ion channels.

Comparison of BioID and Apex Labeling Methods

Whereas we identified more than 3,800 proteins and multiple peptides for over 3,000 proteins using APEX2-conjugated Ca2+ channels (7), Feng et al. (70) identified 550 proteins with at least 2 unique peptides. These differences speak to the promiscuity of APEX-catalyzed biotin-phenoxyl radicals diffusing out to sterically available or “exposed” tyrosines, relative to the BioID approach where the proximity-labeling enzyme itself must directly tag nearby basic residues, such as lysine. The labeling substrate in APEX experiments is biotin-tyramide, which can diffuse easily across membranes, although the radicals are membrane impermeable. The available substrate for BioID labeling depends on the vehicle, route of administration, and pharmacokinetics in the model system. Biotin is a carboxylic acid and is less membrane permeable than biotin-tyramide. Our experimental approach made use of mice with doxycycline-inducible cardiac-specific overexpression of the labeling enzyme (7). Feng et al. (70) instead used a knock-in approach with the native gene dosage of their gene of interest. As noted before, APEX experiments permit the characterization of rapid subcellular changes during signaling events and it is feasible to label in a Langendorff preparation ex vivo, whereas BioID can take several days to fully label the proteome subdomain. On the other hand, the BioID approach should not alter reactive oxygen species signaling that may occur in APEX-labeled cells, although any changes related to exposure to hydrogen peroxide in APEX labeling would likely be minimal because they would have to occur during the 1-minute labeling interval before quenching of the labeling reaction.

Use of Horseradish Peroxidase–Based Proximity Labeling

Horseradish peroxidase (HRP) is inactive in a reducing environment such as the cytosol of a cell (78). HRP is active in the oxidizing secretory pathway (Golgi and ER lumen) and extracellular surface, and catalyzes the same labeling chemistry as APEX2 with biotin-phenol (19,79). HRP was fused to the extracellular region of the junctional adhesion molecule C (JAM-C), designed to define the receptors localized with JAM-C at the cell junction and to determine which ones co-traffic with JAM-C in endothelial cells (80). Transfected human umbilical vein endothelial cells were exposed to biotin-phenol for 30 min, and proteins neighboring JAM-C were biotinylated during a 1-minute incubation with hydrogen peroxide. To specifically identify neighboring proteins within intracellular stores of JAM-C, ascorbate, a membrane-impermeant inhibitor of HRP, was used to block the proximity-labeling reaction at the extracellular side of the membrane. In addition, APEX2 was conjugated to the intracellular domain of wild-type JAM-C or a mutant JAM-C with 4 arginine substitutions for ubiquitinated lysine residues. Using these multiple proximity-labeling approaches, the authors concluded that JAM-C cotraffics with receptors associated with changes in permeability, such as vascular endothelial cadherin, neuropilin (NRP) 1, NRP-2, and some integrin subunits, but not with junctional proteins associated with transmigration of leukocytes, such as the lateral border recycling compartment components (80). This study demonstrated the full utilization of proximity labeling to explore both extracellular and intracellular neighborhoods of a protein of interest as well as the trafficking of the protein.

Next-Generation Proximity Labeling

Recent innovations in proximity labeling, primarily complementation, can help to address the issue of “false positives.” Han et al. identified 24 potential “cut sites” in the wild-type APEX (5). These 24 pairs were conjugated with FK506-binding protein (FKBP) and FKBP-rapamycin–binding domain to permit rapamycin-induced complementation. Ultimately, the 200 amino acid N-terminal fragment pairs and 50 amino acid C-terminal pairs were found to have the most favorable characteristics and were referred to as “AP” and “EX,” respectively (Figure 1C). These underwent multiple rounds of positive and negative selection with yeast display-based directed evolution to increase the strength of labeling and to ensure that labeling occurs only with complementation. Han et al. (5), with the use of rapamycin induction, were able to label the proteome of the mitochondrial-ER interface, and through fusing AP and EX to ribonucleoproteins specific to a particular noncoding RNA, label the proteome of this particular RNA. Although several split BioIDs have been developed, a faster split TurboID, developed and validated in the same manner as split APEX, is now available, allowing for rapid in vivo proximity labeling with high spatial specificity (81).

Conclusions

Proximity proteomics, an avenue of investigation developed only in the past 10 years, holds great promise for advancing our understanding of cardiovascular biology. As these methods continue to mature, several concepts should be kept in mind to ensure high-quality work and worthwhile investment in time and resources. A worthwhile question might be, “What will we do with these results?” Situations well suited to these methods may involve suspected though missing interacting elements that have avoided detection, and after labeling identification can be evaluated in some manner. We found that having a functional or biochemical readout to confirm the effect of our perturbation (β-adrenergic stimulation)—heart rate elevation and phospholamban phosphorylation in our case—allowed us to cull undertreated samples and increase the sensitivity of the method to detect changes in signaling. Examining the pathophysiological alterations of the proteome in vivo lends itself particularly well to biotin ligase labeling, whereas more acute changes, such as signal transduction, may favor APEX labeling.

Funding Support and Author Disclosures

This work was supported by National Institutes of Health grants R01 HL121253, R01 HL126735, and R01 HL146149. Dr. Kushner was supported by T32 HL007343 and the New York Academy of Medicine Glorney-Raisbeck Fellowship. Ms. Papa was supported by grants T32HL120826 and NSF 1644869. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.