Metazoan-like signaling in a unicellular receptor tyrosine kinase

Metazoan receptor tyrosine kinases (RTKs) respond to a variety of extracellular stimuli, initiating signals that regulate important cellular and developmental processes [1–4]. The phosphotyrosine-based signaling system (consisting of tyrosine kinases, tyrosine phosphatases, and pTyr-binding modules) evolved relatively recently [5–7]; pTyr signaling was originally thought to be unique to metazoans. Remarkably, recent genomic analyses have demonstrated that several unicellular organisms possess numbers of receptor and nonreceptor tyrosine kinases that are comparable to higher metazoans. Choanoflagellates are free-living aquatic protists that represent the closest unicellular relatives to metazoans [8, 9]. Tyrosine kinases are abundant in the choanoflagellates Monosiga brevicollis, Monosiga ovata, and Salpingoeca rosetta[10–13]. Tyrosine kinases are also found in even more ancient opisthokonts, including the filasterean Capsaspora owczarzaki[14]. Although the physiological functions of these unicellular tyrosine kinases are not yet known, they are presumably involved in the responses to extracellular signals such as the presence of nutrients, ions, or chemical messengers.

The genome of Monosiga brevicollis is estimated to encode as many as 128 tyrosine kinases and approximately 380 total protein kinases [10]. (For comparison, the human kinome contains 518 protein kinases, of which 90 are tyrosine kinases). Most of the Monosiga brevicollis tyrosine kinases have no identifiable metazoan homologs (the exceptions are nonreceptor tyrosine kinases of the Src, Csk, Abl, and Tec families). As in metazoans, Monosiga brevicollis tyrosine kinases never appear as isolated catalytic domains; instead, each tyrosine kinase possesses a repertoire of associated signaling domains. Many of the domain combinations are distinct from any observed in metazoans. There are predicted to be 88 RTKs in M. brevicollis; 73 of the RTKs cluster into 15 families (designated RTKA, RTKB, and so on) [10]. While these RTKs possess many of the same features as metazoan RTKs (putative extracellular ligand-binding modules, conserved catalytic residues, and potential sites of autophosphorylation), their overall sequences show limited homology to the families of metazoan RTKs. It is not yet known whether the M. brevicollis RTKs (or any unicellular RTKs) are enzymatically active as tyrosine kinases.

The RTKB family of tyrosine kinases from M. brevicollis consists of nine members. While the sequences of the RTKB kinase catalytic domains are related, the extracellular regions are quite diverse [10]. Three of the nine RTKB kinases appear to lack important catalytic residues and may function as pseudokinases or scaffolding proteins. Of the six remaining RTKB kinases, four possess a unique modular domain (designated RM2) in their cytoplasmic tails (Figure 1). The RM2 domains contain potential Src-family kinase phosphorylation sites and SH2-binding sites, suggesting that they may link RTK activation with downstream cytoplasmic signals [10]. In this paper, we characterized Monosiga brevicollis RTKB2, a kinase with six RM2 domains in its C-tail. We cloned, expressed, and purified the RTKB2 kinase, and conducted biochemical studies of its activity. These studies are the first of an RTK from a unicellular organism.

Monosiga brevicollis RTKB2 is predicted to be a type I transmembrane protein, with a single membrane-spanning region [10] (Figure 1A). The extracellular domain organization of RTKB2 has several features in common with families of metazoan RTKs, although the exact combination of domains has not been observed in any metazoan family. RTKB2 contains a Cys-rich domain similar to those seen in the TNF receptor and to the furin-like domains of EGFR and insulin receptor [4]. Two divergent repeats similar to hyalin are present; these modules are structurally related to the immunoglobulin-like fold seen in many metazoan RTKs [16]. RTKB2 contains a short consensus repeat (SCR)/complement control protein domain, as seen in a variety of complement and adhesion proteins. Towards the C-terminus of the extracellular domain, RTKB2 contains five EGF-like modules, as seen in (for example) the Tie and Eph families of metazoan RTKs. As is true for all Monosiga brevicollis RTKs, the extracellular ligand is unknown.

The predicted intracellular portion of RTKB2 contains a single tyrosine kinase catalytic domain. The kinase domain of RTKB2 has limited homology to known families of metazoan RTKs, but possesses the catalytically important residues found in all tyrosine kinases (Figure 1B). RTKB2 has a single tyrosine residue in the predicted activation loop, C-terminal to the kinase-conserved DFG motif (Figure 1B). The C-terminus of RTKB2 contains six copies of a unique domain designated RM2. RM2 domains are composed of approximately 80 amino acid residues. They are found in the cytoplasmic tails of four of the nine Monosiga brevicollis RTKB tyrosine kinases, but they are not present in any other sequence in the protein database, whether from Monosiga, metazoans, or any other organism [10]. RM2 domains contain tyrosine residues that are predicted to serve as phosphorylation sites as well as SH2-binding sites (described in more detail below).

To test whether the predicted RTKB2 protein is an active tyrosine kinase, we used PCR with kinase-specific primers to amplify the cDNA encoding the catalytic domain (residues 1450–1724) from a Monosiga brevicollis cDNA library. We cloned the RTKB2 kinase sequence into a baculovirus expression vector, infected Spodoptera frugiperda (Sf9) insect cells with the recombinant baculovirus, and purified the protein (Additional file 1: Figure S1). Using the phosphocellulose paper binding assay, we tested RTKB2 kinase activity towards several synthetic peptides (Figure 2). RTKB2 was highly active, with a specific activity similar to other purified RTK kinase domains (e.g., the kinase domain of insulin-like growth factor I receptor [17]). The highest activity was observed with a peptide (E4YM4) containing the EEEEYMMMM motif that was selected from a synthetic peptide library as a highly efficient insulin receptor substrate [18]. This activity (73 pmol/min/μg) is comparable to the activity of Monosiga brevicollis MbSrc1 toward its preferred substrates (145 pmol/min/μg; ref. 10). We also tested two peptides with sequences corresponding to predicted phosphorylation sites within RTKB2 RM2 domains. RTKB2-1 is derived from the fifth RM2 domain, while RTKB2-2 is derived from the sixth RM2. Both peptides were phosphorylated by RTKB2 kinase, consistent with the possibility that these sites on the RTKB2 C-tail serve as autophosphorylation sites. We previously showed that these peptides are also phosphorylated by the Monosiga Src-like kinase MbSrc1 [10].

Many RTKs are regulated by autophosphorylation at one or more tyrosines within the activation loop, a flexible segment between the two lobes of the kinase catalytic domain [2]. RTKB2 has a single tyrosine in the predicted activation loop (Figure 1B). To test for RTKB2 autophosphorylation, we first treated the purified enzyme with Yersinia tyrosine-specific YOP phosphatase. Next, we incubated the dephosphorylated RTKB2 with magnesium and [γ-³²P]-ATP and followed the reaction by SDS-PAGE and autoradiography (Figrue 3A, top panel). Labeled phosphate was incorporated into RTKB2 after 5–15 minutes under these conditions. The stoichiometry of phosphorylation after 15 minutes was 0.93 mol phosphate/mol protein. In a parallel experiment, the time-course of autophosphorylation was followed by Western blotting with anti-phosphotyrosine antibody (Figure 3A, bottom panel). In this experiment, the signal for pre-YOP treated RTKB2 was stronger than the signal for autophosphorylated RTKB2, raising the possibility that multiple sites are phosphorylated in Sf9 cells. To determine whether autophosphorylation influences RTKB2 kinase activity, we carried out a peptide phosphorylation assay with purified RTKB2, YOP-dephosphorylated RTKB2, and enzyme that had been allowed to re-phosphorylate as described above. Dephosphorylated RTKB2 lost significant kinase activity as compared to the starting sample, but the activity was regained upon re-phosphorylation (Figure 3B). These results are consistent with a role for RTKB2 autophosphorylation in the control of enzymatic function.

The six RM2 domains in the C-tail of RTKB2 contain tyrosines near their N-termini that are predicted to serve as phosphorylation and SH2-binding sites (Figure 4) [10]. As shown above in Figure 2, synthetic peptides derived from two of the domains are indeed substrates for RTKB2 kinase. To study an intact RM2 domain, we synthesized the gene corresponding to the sixth RM2 domain (the sequence of RM2-6 is presented in Figure 4). The sixth RM2 domain was chosen because it contains a predicted Src phosphorylation/SH2-binding site that conforms well to the consensus (EEVYEAI), and because the RTKB2-2 peptide (Figure 2) corresponds to this site. We cloned the cDNA into a bacterial expression vector, expressed the protein in E. coli cells, and purified it by metal affinity chromatography (Additional file 2: Figure S2A). Despite a predicted molecular weight of 10,089 daltons, purified RM2-6 migrated with an apparent molecular weight of approximately 20,000 daltons on SDS-PAGE (Additional file 2: Figure S2A). The aberrant mobility of RM2-6 may be due to the highly acidic amino acid composition (pI = 3.9). We confirmed the molecular weight of the bacterially-expressed RM2 domain by matrix-assisted laser desorption/ionization mass spectrometry (Additional file 2: Figure S2B).

To determine whether the isolated RM2 domain could serve as a substrate for RTKB2 kinase, we incubated the two purified proteins with Mg-ATP and followed the reaction by anti-pTyr Western blotting (Figure 5A, left panel). Tyrosine-phosphorylated RM2-6 was detected after 5–10 minutes in this reaction, consistent with the role of RM2 domains as autophosphorylation sites. The purified RM2 domain could also be phosphorylated by M. brevicollis MbSrc1 kinase (Figure 5A, right panel). We determined the stoichiometry of RM2-6 phosphorylation by MbSrc1 in a parallel reaction with [γ-³²P]-ATP. After 5 minutes, the stoichiometry was 2.02 ± 0.13 moles of phosphate per mole of RM2-6 domain. RM2-6 has three potential tyrosine phosphorylation sites; these results raise the possibility of multisite phosphorylation in the six RM2 domains in the C-terminus of RTKB2. The Tyr-Glu-Ala-Ile sequence within RM2-6 fits the consensus sequence for binding to the SH2 domains of Src-family kinases [19]. The MbSrc1 SH2 domain has a similar binding preference as its metazoan counterparts [20]. To test for MbSrc1 SH2 binding, we first phosphorylated RM2-6 using RTKB2 and [γ-³²P]-ATP. Next, we incubated the pY-RM2 with an immobilized protein consisting of glutathione S-transferase (GST) fused to the SH2 domain of MbSrc1 (Figure 5B). The phosphorylated RM2 domain bound to the MbSrc1 SH2 domain, as detected by autoradiography. When the binding reaction was carried out in the presence of a high-affinity SH2 binding peptide (pYEEI), binding between RM-6 and the MbSrc1 SH2 domains was drastically reduced (Figure 5B). These results are consistent with a model in which RTKB2 autophosphorylation initiates the formation of a signaling complex, potentially containing MbSrc1 (by analogy to metazoan RTK signaling) and additional SH2-containing Monosiga proteins.

To study RM2 domain phosphorylation in a cellular context, we amplified a cDNA encoding the C-terminal tail of the RTK (all six RM2 domains), and cloned it into a mammalian expression vector to produce a fusion with green fluorescent protein (GFP). We expressed the protein in Src/Yes/Fyn triple knockout fibroblast cells (SYF cells), which lack all Src family kinases [21]. Fluorescence microscopy showed that the GFP-RTKB2-CT has a diffuse cytoplasmic localization (Additional file 3: Figure S3). Next, we expressed GFP-RTKB2-CT in SYF cells alone or in the presence of Flag-tagged MbSrc1 kinase (Figure 6). We isolated RTKB2 by anti-GFP immunoprecipitation and analyzed tyrosine phosphorylation by Western blotting. RTKB2-CT was weakly tyrosine-phosphorylated in these experiments, and the addition of MbSrc1 gave only a modest increase in phosphorylation (Figure 6). We were unable to detect a complex between RTKB2-CT and MbSrc1 by co-immunoprecipitation (data not shown). While these experiments confirm the ability of the RTKB2 C-tail to be phosphorylated, it is likely that ligand-stimulated RTKB2 kinase would produce much higher levels of tyrosine phosphorylation. Furthermore, the lack of membrane-anchoring motifs in the RTKB2-CT and MbSrc1 constructs likely reduces the efficiency of their interaction.

The unique sequences of the RM2 domains suggest that they may serve as novel signaling modules. Many eukaryotic signaling domains adopt defined structures, even when removed from their parental proteins [22]. In order to evaluate the solution structure of ¹⁵N labeled RM2, a ¹H-¹⁵N HSQC spectrum was acquired (Additional file 4: Figure S4). The spectrum of RM2 illustrates that the chemical shift of the backbone amide resonances display narrow dispersion, cluster and overlap around the center of the spectrum, while the line shapes are sharp. These features are typical of a disordered protein. Furthermore, the number of backbone amide resonances is lower than the expected count of 103 (excluding the amino terminal residue and prolines) for His10 tagged RM2, presumably because the resonances are extensively overlapped due to the disordered state of the RM2 domain. We attempted to model the structure of the RM2-6 domain using the web-based Protein Homology/Analogy Recognition Engine (Phyre) (http://www.sbg.bio.ic.ac.uk/phyre2). Phyre predicted that 52% of the RM2-6 domain is disordered, and secondary structural elements were predicted with low confidence. Collectively, our data suggest that the RM2 domains of RTKB2 serve as kinase/SH2 binding sites, and that pTyr-SH2 binding is the main determinant for interaction, rather than a well-defined protein-protein interface.

The abundance and diversity of receptor and nonreceptor tyrosine kinases in the unicellular choanoflagellate Monosiga brevicollis rival that of any metazoan [10, 11, 23]. The 88 RTKs found in the genome of M. brevicollis possess a wide variety of domain organizations. The divergent architectures of the choanoflagellate RTKs were likely generated by gene duplication and domain shuffling [24, 25]. The M. brevicollis RTKs have no direct homologs in multicellular organisms. In contrast, sponges, which are regarded as the oldest surviving metazoan lineage, possess most of the RTK families found in higher metazoans [26–28]. The genome of the sponge Amphimedon queenslandica contains 150 RTK genes, including kinase domains from six known animal families: epidermal growth factor receptor (EGFR), Met, discoidin domain receptor (DDR), ROR, Eph, and Sevenless [26]. The sponge Oscarella carmela possesses a similar array of homologs, and RTKs with homology to the receptors for insulin-like growth factor I and fibroblast growth factor [28]. This is consistent with a model in which the common ancestor between choanoflagellates and metazoans had RTKs, but the animal cell-specific families of RTKs developed after the split between the two groups [10].

RTKB2, the tyrosine kinase studied here, is one of the nine RTKB-family kinases from M. brevicollis. As the most primitive RTK to be yet studied, these results shed light on the evolution of biochemical function in the receptor tyrosine kinase superfamily. RTKB2 is active as a tyrosine kinase, and the intrinsic enzymatic function of the RTKB2 catalytic domain is high towards synthetic peptides, particularly the IR/IGF1R family peptide substrate E4YM4 (Figure 2). RTKB2 also catalyzes autophosphorylation, an event that increases the activity of the enzyme (Figure 3). RTKB2 possesses a single tyrosine residue within the predicted activation loop (Figure 1B). Activation loops are one of the distinguishing features of eukaryotic protein kinases [29]. They are flexible, dynamic segments that are often stabilized in the active conformation by addition of one or more phosphates (either through autophosphorylation or through the action of another kinase). The control of RTKB2 activity by autophosphorylation indicates that this mode of regulation was present in primitive RTKs before the evolution of multicellular animals over 600 million years ago.

The C-terminal tail of RTKB2 contains 6 copies of the RM2 domain, a unique region of ≈ 80 residues that has not been found outside of the RTKB family in M. brevicollis (Figure 4). RTKB1, RTKB3, and RTKB4 each have one copy of the RM2 domain C-terminal to their tyrosine kinase domains [10]. Each of the RM2 domains has a tyrosine residue preceded by one or more negatively-charged amino acids. Scansite prediction indicated that these could serve as Src phosphorylation sites and/or SH2-binding sites [10]. Our data show that an isolated RM2 domain is phosphorylated by the RTKB2 kinase or by MbSrc1, a Src-family nonreceptor tyrosine kinase from M. brevicollis (Figure 5). Small synthetic peptides derived from the potential phosphorylation sites from two of the RTKB2 RM2 domains are also substrates (Figure 2). The tyrosine phosphorylated RM2 domain binds specifically to the SH2 domain of MbSrc1 (Figure 5B), suggesting that phosphorylation of one or more RM2 domains can recruit cytoplasmic tyrosine kinases and other cellular proteins to propagate the RTKB2 signal. In contrast to other modular signaling domains (e.g., SH2 or SH3), the isolated RM2 domain appears to lack an ordered structure when it is removed from the context of the surrounding protein (Additional file 4: Figure S4). These binding sites at the C-terminus of RTKB2 may serve a similar recruitment function as the tyrosine motifs found in the short, unstructured cytoplasmic tails of the thrombopoietin or erythropoietin receptors [30, 31] or the phosphorylation sites in the C-termini of RTKs such as the epidermal growth factor receptor.

Binding of the M. brevicollis MbSrc1 kinase to the phosphorylated RM2 domain raises the possibility that receptor and nonreceptor tyrosine kinase signaling are linked, as in metazoans. A classic example of the linkage in mammalian cells is observed in the role for Src in relaying the signal through the platelet-derived growth factor (PDGF) family of receptors. Src stably associates with the cytoplasmic portion of the PDGF receptor, leading to enhanced Src activity [32–34]. For PDGFRa, Src is required for the phosphorylation of the adaptor protein Shc [32]. After binding to PDGFRb, Src phosphorylates a tyrosine residue on the receptor; this inhibits a signaling pathway leading to motility, but increases mitogenic signaling [33]. MbSrc1 phosphorylates the C-tail of RTKB2 weakly in the absence of an activating signal (Figure 6). Our results are consistent with a model in which RTKB2 activation (by an unknown signal) stimulates receptor autophosphorylation within the RM2 domains. This leads to MbSrc1 recruitment and further phosphorylation, which may serve to increase or dampen specific downstream signals. Identifying the nature of these signals will require the development of methodology to manipulate gene function in choanoflagellates.

We conducted the first biochemical study of a unicellular receptor tyrosine kinase. We cloned, expressed, and purified the RTKB2 kinase, and showed that it is enzymatically active. The activity of RTKB2 is regulated by autophosphorylation. The receptor possesses six copies of a unique domain (designated RM2) in its C-terminal tail. An isolated RM2 domain was a substrate for RTKB2 kinase, and the phosphorylated RM2 domain bound to the SH2 domain of a Src family kinase. Thus, this unicellular signaling system contains many of the features found in metazoan RTK pathways.

Data supporting the results of this article are included within the article and in Additional file 1: Figure S1, Additional file 2: Figure S2, Additional file 3: Figure S3 and Additional file 4: Figure S4.