Trans-interacting Plasma Membrane Proteins and Binding Partner Identification

Abstract

Plasma membrane proteins (PMPs) play critical roles in a myriad of physiological and disease conditions. A unique subset of PMPs function through interacting with each other in trans at the interface between two contacting cells. These trans-interacting PMPs (tiPMPs) include adhesion molecules and ligands/receptors that facilitate cell-cell contact and direct communication between cells. Among the tiPMPs, a significant number have apparent extracellular binding domains but remain orphans with no known binding partners. Identification of their potential binding partners are therefore important for the understanding of processes such as organismal development and immune cell activation. While a number of methods have been developed for the identification of protein binding partners in general, very few are applicable to tiPMPs, which interact in a two-dimensional fashion with low intrinsic binding affinities. In this review, we present the significance of tiPMP interactions, the challenges of identifying binding partners for tiPMPs, and the landscape of method development. We describe current avidity-based screening approaches for identifying novel tiPMP binding partners and discuss their advantages and limitations. We conclude by highlighting the importance of developing novel methods of identifying new tiPMP interactions for deciphering the complex protein interactome and developing targeted therapeutics for diseases.

Graphical Abstract

I. Introduction

Multicellular organisms rely on communications between cells of the same and different types to coordinate diverse physiological processes such as organ development and immune responses. On the other hand, abnormal cell-cell contact and signaling are involved in tumor development and metastasis. Mapping inter-cellular communications is therefore critical for understanding normal physiology and disease conditions. Cells can signal to each other in four different fashions: 1) autocrine, where the same cell releases and receives its own signals, such as cytokines; 2) paracrine, where a cell releases signals that diffuse locally over short distances to surrounding cells, for example, growth factors and neurotransmitters; 3) endocrine, where a cell releases signals to the whole organism via the circulatory system over long distances, e.g., growth hormones; and 4) juxtacrine, where signals are transmitted between two cells in direct contact through the interaction between plasma membrane-anchored ligands and receptors. In autocrine, paracrine, and endocrine, signals are carried by soluble molecules. Except for neurotransmitters released into the neuronal synapses, signals can be received by all cells expressing the corresponding receptors within certain distances. In contrast, juxtacrine signaling relies on direct cell-cell contact, and is therefore highly directional and precise. This level of precision is necessary for many critical physiological processes such as immune cell activation.

This review focuses on plasma membrane-anchored proteins (PMPs) that interact in trans at the interface between two opposing cells. A trans-interaction therefore always involves two PMPs on two different cells. As illustrated in the examples shown in Fig. 1, these trans-interacting PMPs (tiPMPs) include ligands and receptors directly involved in juxtacrine signaling and adhesion molecules that enable the necessary cell-cell contacts (1, 2). It should be noted that cis-interactions between PMPs on the same cell are beyond the scope of this discussion. Cis-interactions can facilitate the formation of PMP complexes on cell membranes but do not directly participate in binding to PMPs on another cells. In contrast to trans-interactions that involve only PMP extracellular domains, cis-interactions may involve PMP extracellular domains, transmembrane domains and local membrane structures such as lipid rafts. The exact number of tiPMPs is unknown but is expected to be large. There are approximately 6,000 membrane proteins that make up about one third of all human proteins (~20,000) (3). Out of all the membrane proteins, 1,492 have been experimentally identified as PMPs according to the Cell Surface Protein Atlas (CSPA) (4). Many PMPs function as tiPMPs but their binding partners are unknown. For example, the B7 immunoglobulin superfamily (IgSF) PMPs have nine members that are expressed on antigen presenting cells (APCs) and cancer cells. Four of the members, B7.1, B7.2, PD-L1 and PD-L2, bind the receptors CD28, CTLA-1 or PD-1 on T lymphocytes (T cells) to regulate T cell activation. The other five members, B7-H3, B7-H4, B7-H5, B7-H7, and BTLN2, are also known to control T cell activation but their receptors are unknown (5). Another example is the adhesion G protein-coupled receptors (aGPCRs), which are a group of B2 family GPCRs consisting of 33 members that have large N-terminal extracellular domains (6). aGPCRs are found on cells of most organ systems and are involved in diverse physiological processes, such as regulating cell size and shape, planar cell polarity, as well as cell adhesion and migration. Several aGPCRs are known to bind to cell surface ligands in trans, including teneurin-2 for ADGRL2, teneurin-3 for ADGRL3, WNT-7 and integrin-αvβ3 for ADGRA2, Frizzled for ADGRC1 and ADGRC2, neuroligin-1 for ADGRB1, stabilin-2 for ADGRB3, CD55 and CD90 for ADGRE5, and CD81 for ADGRG1. However, there are still 17 aGPCRs that may function as tiPMPs yet their PMP ligands remain unknown.

Figure 1. tiPMP interactions at the interface between an APC and a T cell.

Initiation of T cell activation requires alignment of membrane by sequential interactions between adhesion molecules, starting with LFA-1-ICAM-1 and α4 integrin-VCAM interactions with large extracellular domains followed by CD2-CD58 with smaller extracellular domains. The close membrane alignment excludes transmembrane proteins with very large extracellular domains such as CD45 and enables interactions between tiPMPs with small extracellular domains including pMHC-TCR that is key to T cell antigen recognition. T cell activation is modulated by interactions between auxiliary tiPMPs including co-stimulatory CD28-B7-1 and co-inhibitory PD1-PDL1.

Identifying binding partners for tiPMPs, including those described above, is crucial for understanding their diverse and important functions. A thorough understanding of multicellular organism physiology requires the identification of all tiPMP interactions to unlock their interactome. The unique nature of tiPMP interactions, however, poses significant challenges for conventional approaches. Specifically, the low intrinsic affinities of tiPMP interactions hinder classical methods such as immunoprecipitation or “pull-down” and avidity-based screening. It is also technically challenging to replicate the unique physical environment at the interface between two cells that may impact tiPMP interactions. For instance, the distance between the two opposing plasma membranes is influenced by the abundance and size of adhesion molecules and may have profound impact on the kinetics of tiPMP interactions taking place in the same space. Here, we review the physiological and pathological significance of tiPMP interactions, outline the challenges of binding partner identifications, and describe current approaches that overcome the low affinity of tiPMP interactions through avidity enhancement.

II. Physiological and Pathological Significance of tiPMP Interactions

tiPMP interactions are involved in the development and function of virtually all organ systems of multicellular organisms. Here, we highlight three aspects to illustrate the nature of tiPMP interactions and their significance.

II.1. tiPMPs in Organismal Development

The contribution of tiPMP interaction to multicellular organism development starts at the very first step–the fertilization of an egg cell by a sperm. The sperm head structure is highly organized and compartmentalized into outer acrosome membrane, inner acrosome membrane, equatorial segment, postacrosomal region, and middle piece (7). All these regions express tiPMP which, through interactions with counterparts on the oocyte, facilitate and regulate the fertilization process, including the attachment of sperm to egg cell (Zona Pellucida-binding), penetration of sperm into egg membrane (acrosome reaction), and fusion with the oolemma (8). As the fertilized egg goes through the complex and delicate stages of gastrulation, neurulation, and organogenesis, highly coordinated tiPMP interactions play essential roles in controlling cellular proliferation, differentiation, and spatial organization. For example, in heart development, interaction between fibronectin and α5 integrin is crucial for the formation of the pharyngeal region in cardiovascular morphogenesis (9). Interaction between Notch2 and Jagged1 plays a critical role in intrahepatic bile duct development in the liver (10). In brain development, homophilic neural cell adhesion molecule (NCAM) trans-interaction controls neuronal cell growth and synaptic stability (11). Conversely, disruptions in tiPMP networks have been shown to result in various organ-related pathological conditions, such as cardiovascular disease and Alzheimer’s disease (12, 13).

Through spatial-temporal regulation of various cell signaling pathways, tiPMPs play important roles in cell fate determination that is critical for body axis formation and organogenesis (14). One of the most studied tiPMP interactions is between Notch receptors (Notch 1-4) on one cell and Delta ligands (Dll1, Dll3, Dll4) or Jagged ligands (Jag1, Jag2) on a juxtaposed cell. Notch signaling was first characterized in Drosophila melanogaster for its roles in dictating neighboring cell fate and regulating developmental phenotypic pattern formation through a mechanism called lateral inhibition (15). Canonically, upon engagement by Delta ligands on a neighboring cell, Notch receptors undergo two successive proteolytic cleavage events mediated by ADAM-family metalloproteases and γ-secretase, leading to the release of the Notch Intracellular Domain (NICD). NICD is then transported into the nucleus and acts as a transcriptional regulator to activate transcription of downstream target gene Hey/Hes1, which in turn can downregulate the cell’s own Delta ligand expression. In lateral inhibition, a high Delta-expressing cell can suppress its neighboring cells’ Delta expression through Hey/Hes1 activation, ultimately leading to low Delta expression on the neighbor cell. Consequently, a divergent phenotype is established among neighboring cells, hence leading to different cell fates. In angiogenesis, optimal blood vessel formation is dependent on a similar proportion of two distinct endothelial cells: tip cells and stalk cells. Lateral inhibition determines cell fate by preventing neighboring cells of a tip cell from attaining the same tip cell fate (16).

II.2. tiPMPs in Immune Function

The mammalian immune system comprises the innate and adaptive subsystems, each containing multiple different cell types. The innate immune system includes neutrophils, basophils, eosinophils, mast cells, macrophages, dendritic cells (DCs), and natural killing (NK) cells. Serving as the first line of defense, it mounts rapid actions against pathogens. The adaptive immune system consists of B and T lymphocytes that can develop immune memories for pathogen-derived antigens to thwart repeated attacks. To mount an effective immune response, cells from innate and adaptive systems coordinate actions through a complex communication network. Immune cells in lymphoid organs, for example, communicate through autocrine and paracrine pathways through a variety of secreted cytokines and chemokines. Notably, juxtacrine signaling through tiPMP interactions play an especially prominent role. Thanks to its ability to deliver highly direct and specific information from one cell to another, juxtacrine communication is critical for ensuring the specificity of an immune response and for controlling its magnitude, which are critical for achieving effective control of infections while avoiding unnecessary inflammation and normal tissue damage.

Many tiPMPs are involved in juxtacrine signaling between immune cells in immune responses. tiPMP interaction is the main mechanism used by NK cells, B cells, and T cells to recognize antigens – the critical first step of an immune response. NK cells of the innate immune system use the receptor natural-killer group 2, member D (NKG2D) to bind MICA/B or RAET1/ULBP ligands expressed on the surface of infected, transformed, senescent, and stressed cells (17). T cells express T cell receptors (TCRs) to recognize antigens presented on the surface of APCs such as DCs, macrophages, and B cells. The antigens on APCs exist as a complex between antigen-derived short peptide and the cell surface molecule major histocompatibility complex, which binds TCR with low affinity (K_D ≈ 3 μM to 90 μM) (18) (Figure. 1). Strikingly, through this rather weak tiPMP interaction, T cells can detect as few as 1-10 pMHC molecules on the entire surface of an APC, enabling remarkable sensitivity of T cell antigen recognition (19, 20, 21).

tiPMP interactions also play a central role in regulating immune responses. In the innate system, interaction between 2B4 expressed by NK cells and CD48 on macrophages is critical for NK cell activation and secretion of cytokines (22). During inflammation, neutrophils express L-selectin which can bind to its ligand PSGL-1 on other neutrophils to recruit other immune cells to the site of injury (23). In the adaptive system, members of the B7 family ligands and receptors can either enhance or inhibit T cell activation (Figure. 1). For example, B7.1 and B7.2 expressed on APCs promote T cell activation and survival through binding to the receptor CD28 on T cells but inhibit T cell activation through interacting with CTLA-1 (24). PD-L1 has also been identified as a major inhibitor of T cell activation through binding to the receptor PD-1, which acts as a checkpoint for uncontrolled and potentially harmful T cell responses. It should be noted that information transfer through these tiPMP interactions at the interface between T cells and APCs are enabled by adhesion molecules such as LFA-1 and ICAM-1, which align the two opposing membranes within a certain distance. Additionally, LFA-1 and VLA-4 expressed on T cells also facilitate their interaction with nonimmune cells such as endothelial cells through binding to ICAM-1 and VCAM-1, respectively. On the B cell side, sufficient B cell activation and antibody production require the engagement of CD40 by CD40L expressed on helper T cells (25).

3. tiPMPs in the Development and Metastasis of Cancer

Human homeostatic controls are maintained through inhibiting aberrant cell proliferation and preventing the emergence of cancer cells. Cell-cell communication contributes to homeostasis by sending or receiving reciprocal signals at the right time and location. Intercellular communication between cancerous and normal cells are dynamically regulated by either direct cell-cell contact or secretion of soluble molecules such as cytokines, growth factors, and chemokines. In direct cell-cell communication of cancer, cells interact via different types of gap junctions, tight junctions, adherens junctions, desmosomes, tunnel nanotubes, and adhesion molecules such as cadherins, selectins, and integrins (26). tiPMP interactions are involved in all these mechanisms. Among them, interactions between E-cadherins are critical for maintaining contact inhibition of proliferation and for preventing epithelial-mesenchymal transition that enables cellular transformation and metastasis (27, 28).

In the tumor microenvironment, tiPMPs are recruited by cancer cells to promote tumor growth and inhibit immune attack. IgSF cell adhesion proteins, such as vascular cell adhension molecule (VCAM) and intercellular adhesion molecule (ICAM), are known to interact with integrin to promote angiogenesis (29). Notch mutations and aberrant signaling have been found to promote oncogenesis in a variety of cancers (30). tiPMPs involved in immune checkpoint regulation significantly contribute to the formation of immunosuppressive tumor microenvironment. In addition to the B7 family members mentioned above, cancer cells express MHC-II, CD112 and ceacam-1 to inhibit T cell function through engaging receptors LAG-3, TIGIT and TIM-3, respectively (31).

III. Challenges of Identifying tiPMP Binding Partners

Despite the importance of tiPMPs, progress in identifying tiPMP binding partners has been hindered by factors associated with the unique nature of tiPMP interactions. tiPMPs interact in a 2-dimensional (2D) fashion as both binding parties are anchored on the cell surface and have only 2D freedom in diffusing on the plasma membrane (32, 33). This contrasts with 3-dimensional (3D) interactions between, for example, cell surface hormone receptors and their soluble hormone ligands. The affinity and kinetics of 3D interactions, which are usually measured using methods such as surface plasmon resonance (SPR), are largely determined by the intrinsic physical and chemical characteristics of the binding sites, including buried area when bound, contact topology, and the number and types of non-covalent bonds formed between the side chains of amino acid residues. 2D binding kinetics, on the other hand, is additionally influenced by the alignment of two opposing membranes. To associate, the two membranes must be brought together through cell locomotion and changes in cell shape. The distance between two membranes may need to be further aligned and adjusted by adhesion molecules to accommodate the size and reach of tiPMP extracellular domains. On the other hand, once the two membranes are aligned at the right distance, binding may occur very rapidly due to the limited rotational freedom of the binding sites and very high local tiPMP concentrations in the narrow gap between two opposing membranes. This fast association also enables rapid rebinding after dissociation. Similarly, the dissociation of bound tiPMPs is determined not only by the intrinsic binding strength but also the stability of membrane alignment. When the two membranes separate during, for example, cell migration or morphological changes, the binding is inevitably stressed by mechanical forces. In this case, the binding may fall apart at a much faster rate than its intrinsic off-rate since the dissociation rate increases exponentially with force if it is a “slip bond” (34). For catch bonds, however, tensile force increases the bond lifetime by promoting new and stable atomic contacts at the interface (35). In other words, while a slip bond lasts the longest under zero mechanical stress, a catch bond achieves its maximum lifetime when pulled with a certain amount of force. Many tiPMP interactions have been found to display the characteristics of catch bonds, including those between CD44 and CD62L (L-selectin), integrins and ICAM-1, Notch1 and Jag1, and agonist pMHC and TCR (35, 36, 37, 38, 39, 40, 41). An increasing amount of evidence suggests that the ability to form a catch bond may be the defining feature of a cognate pMHC-TCR interaction that can lead to T cell activation (38, 40, 41, 43, 44, 45, 46). Taking this into account, the Garcia group screened a TCR library for TCRs that bind pMHC target with low intrinsic 3D affinities but can mediate T cell activation. The TCRs identified were shown to form catch bonds under force. Importantly, they identified a catch bond-capable mutant of MAGE-A3-specific TCR that lost the ability to recognize the TITIN peptide, which is expressed in normal cardiovascular tissue and responsible for cardiotoxicity in therapies using T cells expressing the wild type MAGE-A3-specific TCR. To directly identify TCRs that function under force, the Fordyce group developed a screening method called BATTLES (biomechanically-assisted T cell triggering for large-scale exogenous pMHC screening), which uses pMHC anchored on swelling polymer beads to apply force to T cells expressing a library of TCR mutants (46).

Taken together, the unique characteristics of 2D interaction, especially the fast association rate and rapid unbinding and rebinding, enables tiPMP interactions to accomplish physiological functions such as adhesion and receptor signaling through low intrinsic affinities (47). It should be noted that the key feature of catch bond is that the maximum bond lifetime is achieved under ~10pN of force rather than under no force. The maximum bond lifetime is still very short (~ 1s for pMHC-TCR interactions). In fact, high intrinsic affinity may be detrimental in 2D settings. Adhesion between cells may become irreversible, thus affecting cell migration and organ development. Very high affinity interactions between TCR and pMHC ligands have been shown to inhibit serial engagement of multiple TCRs by a few ligands, negatively impacting receptor signaling (49). Indeed, the intrinsic binding affinities measured using SPR with the tiPMP extracellular domains are generally low. For example, the dissociation constant (K_D) of TCR binding to pMHC ligands are in the range of 1 – 10 μM. The K_D of binding between adhesion molecules CD2 and CD58 is 10 μM (50). The inhibitory receptor CTLA4 on T cells binds to CD80 on APCs with a K_D of 12 μM (51).

The low intrinsic affinities of tiPMP interactions and the membrane location of tiPMPs pose significant challenges for identifying binding partners using conventional biochemical methods such as affinity purification/mass spectrometry (AP/MS). In this approach, bait protein molecules conjugated to an immobile phase, typically polymer beads, are incubated with lysates of cells expressing the putative binding partner, or the prey. The beads are then removed and washed. Proteins bound are separated using electrophoresis or chromatography and their identities are then determined using mass spectrometry. When applied to tiPMP interaction, the bait can be a recombinant version of the extracellular domain of a tiPMP if it is a simple type I or type II single pass transmembrane protein. For tiPMPs with multipass transmembrane domains, however, generating the bait can be complicated by the dependence of extracellular domain conformation on membrane interaction (52). Since the prey proteins in the lysate are no longer anchored on intact plasma membrane, this approach relies on interactions between the bait and prey proteins in a 3D fashion. The low intrinsic affinity of tiPMP interaction poses significant challenges to obtaining enough bead-bound prey for MS analysis. In the binding step, the effect of low affinity is exacerbated by the relatively low concentration of the prey in the lysate since tiPMPs tend to be expressed at relatively low levels on a per cell basis. In the washing step, the fast dissociation rate of low affinity tiPMP binding events may lead to significant loss of binding to beads. For example, for TCR-pMHC interactions with K_D of ~1 μM, the binding half-life is only a few seconds at 25 °C (53). Although this may be prolonged at 4 °C, it may still not be enough to withstand the washing step, which is often intensive in order to remove nonspecific bindings by the large variety of proteins in the lysates. Finally, the presence of detergents in the incubation and washing steps may interfere with the binding and further reduce the amount of target tiPMP bound on the beads available for analysis. An alternative to AP/MS is coelution or cofractionation coupled to MS (co-Frac-MS) (54). In this method, cell lysate is fractionated using chromatography or electrophoresis and proteins in each fraction are identified using MS. Proteins in the same fraction may interact directly or indirectly. Certain variations of the method, such as protein correlation profiling-stable isotope labeling by amino acids in cell culture (PCP-SILAC), allows high-throughput identification of interacting proteins in cytosol, organelles, and cis-interacting transmembrane protein complexes (55, 56, 57, 58). Methods based on co-Frac-MS, however, may not be suitable for identifying weak or transient interactions likely due to relatively high affinities being required for proteins to stay in the same fraction during fractionation processes (54, 59).

IV. Current Approach for Identifying tiPMP Interaction Partners: Avidity-Enhanced Screening of Protein Libraries

Current effort to identify tiPMP interactions relies on screening protein libraries using enzyme-linked immunosorbent assay (ELISA)-like approaches. In these methods, the bait tiPMPs are anchored on the plastic surfaces of multiwell plates. Proteins can be adsorbed onto plastics such as polystyrene through electrostatic and hydrophobic interactions or through specific interactions with proteins such as streptavidin and antibodies already adsorbed on the plate (60). Direct absorption is straightforward but may lead to protein denaturation and buried binding sites due to the random orientations of bound protein. Specific interaction requires tagged proteins, but the bound proteins are likely to have native conformation and more accessible binding sites. The plates with bait proteins are then used to screen a library of tagged candidate prey proteins by adding each candidate to a well. Bound prey proteins are detected using tag-binders conjugated with enzymes such as horseradish peroxidase (HRP), which enable detection of colored, fluorimetric, or luminescent derivatives. The fluorimetric method offers slightly higher sensitivity and wider dynamic range than the 2 to 4 OD limit of the colorimetric detection. Luminescent detection is by far the most sensitive with a wide dynamic range. Protein libraries are usually generated by expressing tagged extracellular domains of tiPMPs in eukaryotic expression systems. The proteins in culture supernatant may be used directly or purified with affinity chromatography.

Conventional ELISA relies on high affinity antibody-antigen interactions (K_D = 10⁻⁸ to 10⁻¹⁰ M). To detect low affinity interactions between tiPMPs, the sensitivity of the approach needs to be enhanced by using multimeric bait and prey proteins to increase binding valency. While affinity refers to the inherent binding strength conferred by the noncovalent bonds (hydrogen bonds, salt bridges, van der Waals’ interactions, and hydrophobic interactions) at the interface of a pair of binding proteins, avidity refers to the combined strength of all binding pairs formed between multimeric binding partners. In the setting of detecting low affinity binding, the main advantage of multivalent interaction is the increased overall binding stability (61), which allows the binding of candidate proteins to survive the washing steps before detection. The unbinding of multivalent interactions requires all binding pairs to dissociate at the same time, the odds of which are much lower than monovalent interactions. The unbinding of a single interaction within a multivalent interaction is likely to be followed by fast rebinding due to the close proximity of the unbound binding partners. An example application of exploiting multivalent interactions to detect weak tiPMP binding is the use of pMHC tetramers to stain T cells bearing pMHC-specific TCRs for flow cytometry analysis. While monomeric pMHC-TCR interaction is too weak to generate sufficient signals, pMHCs tetramers formed using fluorescently labeled streptavidin and biotinylated pMHCs enabled the detection of rare virus-specific T cells. The technique has significantly contributed to the understanding of the size and dynamics of pathogen-specific T cell repertoire, which can help design effective vaccines. It should be noted that in this case, TCRs on T cells exist as dimers or aggregates in lipid rafts of the plasma membrane, therefore acting as multimers when binding to pMHC tetramers.

For tiPMP identification, the most commonly used approach to create multivalent interactions is to pentamerize bait and/or prey proteins by using rat cartilage oligomeric matrix protein (COMP). COMP is a cell surface glycoprotein expressed as pentamers, which are formed through hydrophobic interactions and disulfide bonds between uniquely structured α-helices in each monomer (62). Fusing tiPMP extracellular domain (ECD) to the COMP α-helix therefore enable the formation of stable pentamers. The power of COMP pentamers to increase the binding stability of weak tiPMP interactions was demonstrated by characterizing the binding between CD200 and CD200R (K_D of ~1 μM) using SPR (63). While the monomeric interaction has a very short half-life of ~0.8 s, pentamerization of CD200 increased the half-life to more than 3000 s (64). Taking advantage of COMP-mediated avidity enhancement, Kerr and Wright developed the avidity-based extracellular interaction screen (AVEXIS) (Figure 2A) (65, 66). In this approach, the bait protein was fused to rat CD4 domains 3 and 4 followed by an Avi-tag. Avi-tag is a 15-amino acid peptide sequence in which the lysine residue can be specifically biotinylated by the BirA enzyme. To produce biotinylated bait proteins, HEK-293 cells were cotransfected with DNA constructs encoding the bait protein and BirA enzyme and cultured in medium containing biotin. The culture media were harvested and dialyzed to remove free biotin and the concentrations of bait proteins were normalized by detecting the CD4 domains using ELISA. The media were then incubated in streptavidin-coated plates to capture the biotinylated bait proteins. The prey proteins were expressed in HEK-293 cells as fusion proteins with the COMP helix and β-lactamase at the C-terminus to allow pentamerization and detection using the β-lactamase substrate nitrocefin. Prey proteins in the culture media were used directly for screening after normalizing the concentrations by measuring β-lactamase activity. Using the low affinity binding between CD200 and CD200R (Kd = 1 μM, t_1/2 = 0.9 s), the authors determined that pentamerization of the prey increased the screening sensitivity by at least 250-fold compared with monomeric prey proteins (66). Using a panel of tiPMP interactions with known affinities, the authors set the detection threshold to 8 μM by controlling the amount of prey proteins used in the assay. Using AVEXIS, Bushell et al screened the interactions among IgSF proteins, which are known to form receptor-ligand pairs between themselves. In total, 6105 interactions among 110 zebrafish IgSF proteins were screened, and 17 interactions were identified between 19 proteins. Four interactions had very low half-lives at, or below, the limit of SPR sensitivity (t1/2 ≤ 0.1 s), demonstrating the extreme sensitivity of the AVEXIS technology.

Figure 2. Approaches based on avidity-enhanced screening of protein libraries for identifying low-affinity tiPMP interactions.

A. Avidity-based extracellular interaction screen (AVEXIS). Library of prey proteins anchored on streptavidin-coated plate are screened against pentameric β-lactamase-tagged bait proteins. B. Extracellular Interactome Assay (ECIA). Library of bait proteins anchored on protein A-coated plate are screened against pentameric AP-tagged prey proteins. C. Avidity-enhanced screening using IgG Fc-based multimeric preys. AP-fused bait proteins anchored on anti-AP-coated plates are used to screen Fc-tagged prey protein library. Signal outputs are detected using HPR-conjugated anti-Fc antibodies. D. Avidity-enhanced screening using bait proteins on microbeads. Bait ECD-Fc fusion proteins anchored onto protein A microbeads are used to screen a prey library in the form of a protein microarray. Fc bait proteins are labeled with Cy5 fluorescent dye.

AVEXIS was also employed to screen interactions between 150 zebra fish leucine-rich repeat (LLR) receptors that are involved in neurodevelopment and function (67). Experiments were set up to screen 2,548 interactions between 49 baits and 52 preys. Seventeen interactions between 12 proteins were identified. The authors then screened the 52 preys against 97 zebra fish IgSF members. Out of the 5044 total interactions screened, 17 binding pairs were identified. The study revealed extensive cross-interaction between members of LLR and IgSF families and a high percentage (48%) of receptors having heterophilic binding partners.

Özkan and co-workers developed the Extracellular Interactome Assay (ECIA) that also utilizes COMP-mediated pentameric prey proteins (Figure 2B) (68). The preys were fused with alkaline phosphatase (AP) instead of β-lactamase for enzymatic detection. The bait protein, however, is fused to Fc and anchored on protein A-coated plates. Compared with the monomeric bait protein used in AVEXIS, the bivalent Fc fusion baits may bind preys with higher avidity. In AVEXIS, however, the bait proteins are anchored on tetrameric streptavidin on plates. Therefore, the real valency of bait protein may depend on the density of bait protein on the plate surface. Unlike AVEXIS, the amounts of bait and prey proteins in ECIA were not normalized. The authors argued that protein normalization would exclude poorly expressed proteins in the screen and lead to false negatives. Indeed, the high sensitivity of ECIA enabled the identification of Vein–Ihog/Boi interactions, despite nearly undetectable Vein expression. The false positives caused by “sticky” proteins were accounted for during data analysis through normalization of the data matrix within every bait and prey protein. SPR analysis of a subset of identified interactions showed no false positives. Using ECIA, 83 previously unknown interactions were identified by screening 20503 interactions between drosophila IgSFs, fibronectin type-III and leucine-rich repeat proteins. Among them, DIPc interacts with 19 molecule and acts as a hub of the interaction network. Applying the ECIA approach to tiPMP interactions in plants, the Smakowska group screened interactions among 200 leucine-rich repeat receptor kinases (LRR-RK) in the model plant Arabidopsis thaliana (69). An LRR-based Cell Surface Interaction network (CSI^LRR) was constructed based on 567 interactions identified and used to predict the function of uncharacterized LRR-RKs in plant growth and immunity.

Another example of successful implementation of avidity-enhanced screening is the identification of the receptor for human cytomegalovirus (HCMV) (70). Martinez-Martin et al established an automated high throughput platform to screen a library consisting of 1297 human single transmembrane proteins for binding to HCMV envelop protein. A unique feature of the setup is that single transmembrane prey proteins were fused to IgG Fc and anchored on protein A-coated plates to interact with HCMV envelop proteins as natural trimers or as pentamers through COMP fusion. Seven proteins were found to bind to the envelop protein, and neuropilin-2 (Nrp2) was identified as a new receptor based on its ability to mediate viral entry, demonstrated in overexpression, knockout, and antibody blocking experiments. Building on this success, the same group developed a modified approach called RDIMIS (Receptor-Display In Membranes Interaction Screen). In RDIMIS, the same plate-bound Fc-fusion prey library was screened for interaction with extracellular vesicles (EVs) displaying bait membrane proteins (71). To prepare the EVs, Expi293F cells were transfected to express HIV Gag-luciferase fusion protein and the bait protein. Gag drives the HIV budding process and was found to increase EV production by 4-fold. The EVs were isolated using ultracentrifugation and quantified based on protein content. EV binding to prey proteins was determined using luminescence assays. The RDIMIS method was validated by its ability to identify known binding partners for poliovirus receptor (PVR), PD-L1, CD80, and CD276. Notably, RDIMIS identified CD248 as the putative binder for the hard-to-purify cancer-associated fibroblast protein LRRC15, demonstrating the unique advantage of this EV-based approach for studying challenging plasma membrane proteins. Potential CD248-LRRC15 interaction was supported by their coexpression in the tumor microenvironment at the RNA and protein levels.

Avidity-enhanced screening has also been used to investigate binding preferences among structurally similar protein isomers and the underlining mechanisms. In Drosophila, Down syndrome cell adhesion molecule 1 (Dscam1) is a type I transmembrane protein with as many as 19008 isomers that can be expressed by neurons to distinguish between self and nonself and to form mutually exclusive receptive fields (72). The binding interface is composed of three distinct IgSF domains, Ig2, Ig3, and Ig7. Each domain can have multiple variants generated through alternative splicing of unique exons. To determine the preference of binding between isomers, the authors measured the interactions between 92 isomers. To anchor the isomers on plates, isomer ECDs were fused with AP and were captured on wells precoated with anti-AP antibodies. Another set of isomer ECDs were generated as Fc fusion proteins and their interactions to isomers in the wells were detected using HRP-conjugated anti-Fc antibodies (Figure 2C). Although ECD-Fc fusion proteins should exist as dimers, it was shown that higher orders of clustering by the detection antibody is likely to be critical for generating sufficiently strong signals. The proteins were generated by transfecting Drosophila S2 cells. Their presence in culture media was determined using SDS-PAGE followed by immunoblotting and normalization of protein concentrations. The results showed striking isoform-specific homophilic interactions among the isomers and that homophilic interactions are achieved in a modular fashion with each variable domain exhibiting highly specific self-binding.

An alternative approach to increase the valency of bait-prey interaction, developed by Ramani and colleagues, is to anchor bait ECD-Fc proteins onto protein A-coated microbeads (Figure 2D) (73). In this study, 89 bait human IgSF receptors were used to screen their interactions with 1334 highly diverse extracellular proteins that include transmembrane and secreted proteins from human and other species. Specifically, Fc-tagged bait proteins were expressed in CHO cells, purified, and labeled with Cy5 fluorescent dye at a 2:4 (dye:protein) ratio as determined by absorbance at 280 nm and 650 nm. Then, the Fc-tagged bait proteins were bound to protein A microbeads at the saturation level as determined using biolayer interferometry. On the prey side, a total of 1334 tagged proteins representing 686 genes were used to generate protein arrays on epoxysilane-coated glass slides. The levels of protein immobilization were confirmed using fluorescently labeled antitag antibodies. Then, microbeads carrying bait IgSF proteins were added to interact with the prey microarray, and bound beads were detected using a microarray reader. To assist hit calling and identification of nonspecific interactions, a statistical methodology was developed that includes a 0.0001% probability cutoff based on normal distribution of signals, duplication, and the exclusion of highly promiscuous binders with > 10% hit rates. The screen identified 105 bait-prey interactions, among which 11 were novel interactions confirmed using SPR. Compared with soluble protein baits, baits on microbeads showed 10 to 150 folds higher signal when tested using CD200-CD200R1, and PD1-PDL1 interaction pairs. Using the same CD200-CD200R1 pair, AVEXIS increased the signal 250-fold. The difference may be explained by several factors in these two distinct methodologies, including different nature of signals, prey protein formats, and data analysis. Although in theory, bait proteins anchored on microbeads may interact with prey at higher valency than the pentameric baits in AVEXIS, the real degree of bait protein oligomerization on beads should depend on protein density and distribution on bead surface, which were not described. The number of bait proteins on beads surface that can interact with preys on the array also depend on the curvature of the surface, which relates to the size of the beads.

V. Conclusions

tiPMPs play indispensable roles in physiological and disease processes ranging from organismal development, immune regulation, to oncogenesis and cancer metastasis. Thanks to their accessibility as proteins on the cell surface, tiPMPs represent attractive targets for small molecule drugs as well as biologics such as monoclonal antibodies. Despite their importance, tiPMP interactions remain poorly characterized due to technical difficulties associated with their unique binding characteristics, mainly low binding affinities and fast dissociation rates. The challenges have been partially addressed by recent development of methodologies such as AVEXIS and ECIA that screen protein libraries using oligomerized bait tiPMP proteins to enhance binding avidity. These approaches have identified many novel tiPMP interactions that are involved in various biological processes including viral entry, immune regulation, and neuron development.

Notwithstanding the successes reviewed here, avidity-enhanced screening approaches are not without limitations. Artificially enhanced avidity may lead to identification of tiPMP interactions that are not physiologically relevant. The identified interacting tiPMPs, for example, may exist as monomers in their natural forms and their monovalent interaction may not lead to signaling, adhesion, or other meaningful functional consequences. Interactions may also be identified between tiPMPs located in anatomically separate organ or tissue systems that cannot possibly interact. Careful functional analysis is therefore critical to identify true tiPMP binding partners. On the other hand, avidity-enhanced screening may still be unable to identify certain tiPMP interactions. First, avidity may not be enhanced to the level that enables detection of very low affinity interactions that still have functional consequences. Second, the extracellular domains of certain tiPMP proteins may not be produced in sufficient amounts or with correct conformation for the preparation of multivalent bait or prey. In natural settings, expression of these tiPMPs on cell surfaces may depend on the local membrane environment. In addition, tiPMPs expressed by insect cells (e.g., sf9 cells) may have biased glycosylation patterns that impact their interactions with binding partners. Finally, current methods are limited to screening one-on-one interactions between two proteins, but some tiPMP interactions may involve more than two players. For example, to interact with a tiPMP on a cell, two transmembrane proteins on another cell may need to work together to maintain conformation or to form the binding interface.

tiPMP interactions are fundamental to multicellular life. Avidity-enhanced screening methods have shed light on a small fraction of the vast and complex tiPMP interactome. To decipher the complete tiPMP interactome, the development of additional and improved high throughput approaches for the identification and characterization of tiPMP interactions are urgently needed.

Availability of data and material:

Not applicable.