Synthetic Biology Reveals the Uniqueness of the RIP Kinase Domain

Abstract

The RIP Kinases (RIPKs) play an essential role in inflammatory signaling and inflammatory cell death. However, the function of their kinase activity has been enigmatic, and only recently has kinase domain activity been shown to be crucial for their signal transduction capacity. Despite this uncertainty, the RIPKs have been the subject of intense pharmaceutical development with a number of compounds currently in preclinical testing. In this work, we seek to determine the functional redundancy between the kinase domains of the four major RIPK family members. We find that while RIPK1, RIPK2 and RIPK4 are similar in that they can all activate NF-kB and induce NEMO ubiquitination, only RIPK2 is a dual specificity kinase. Domain swapping experiments showed that the RIPK4 kinase domain could be converted to a dual specificity kinase and is essentially indistinct from RIPK2 in biochemical and molecular activity. Surprisingly, however, replacement of RIPK2’s kinase domain with RIPK4’s did not complement a NOD2 signaling or gene expression induction defect in RIPK2^−/− macrophages. These findings suggest that RIPK2’s kinase domain is functionally unique compared to other RIPK family members, and that pharmacologic targeting of RIPK2 can be separated from the other RIPKs.

Introduction

The RIP Kinases (RIPKs¹) play an essential role in inflammatory signaling and cell death(1, 2). RIPK1 is required for TNF-induced NF-κB activation and helps regulate the switch between TNF-induced apoptosis and necroptosis(1–3), partnering with RIPK3 to induce necroptosis(1, 2, 4). RIPK2 is an essential kinase regulating signaling downstream of the Crohn’s disease susceptibility protein, NOD2(5, 6). In this role, RIPK2 is part of the protein complex that recognizes intracellular bacterial infection and helps tailor the cytokine response to eradicate an offending pathogen(7, 8). While less well studied, RIPK4 is the causative gene in Popliteal Pterygium Syndrome, a disease characterized by early lethality with multiple developmental abnormalities(9). Given the collective influence of the RIPKs on innate immune and inflammatory signaling, there has been intense interest in manipulating these kinases pharmacologically for clinical gain. Pharmacologic RIPK1, RIPK2 and RIPK3 inhibitors have all been described and are in various states of clinical development for disorders as diverse as sepsis, inflammatory bowel disease and multiple sclerosis(10–19).

Despite this pharmaceutical interest, the function of the RIPKs’ kinase domains has been enigmatic with few bone fide substrates identified(1, 2, 20). In no case is this more true than in the case of RIPK2. Initial BLAST searches suggested that RIPK2 was a serine-threonine kinase, and indeed, RIPK2 was shown to autophosphorylate(6, 21, 22). In these initial descriptions, which were based largely on overexpression studies, RIPK2’s kinase activity was shown to be dispensable for signaling such that while the RIPK2 protein was essential for NOD1/2 signaling, its kinase activity was unnecessary(6, 21, 22). Hints to RIPK2’s kinase function began to emerge when it was shown that the joint p38 and RIPK2 inhibitor, SB203580, could cause decreased expression of RIPK2, presumably through a loss of protein stability(23). While this work was also supported by the fact that a genetic knockin of kinase-dead RIPK2 showed decreased expression, this feature is shared by many kinases in which a kinase-dead variant shows decreased expression(24). In fact, additional pharmacologic studies using a more diverse and specific panel of RIPK2 inhibitors has shown that inhibition of RIPK2 kinase activity does not have a universal role in RIPK2 protein stability(11, 19, 25, 26), thus the role of the kinase activity in RIPK2 protein stability still remains unanswered. A last mystery surrounding the RIPK family of kinases centers on which phosphoacceptor they prefer to phosphorylate. RIPK2 was initially misclassified as a serine-threonine kinase when in fact it is a dual-specificity kinase, capable of phosphorylating serines, threonines and tyrosines(11). Despite this advance in the NOD:RIPK2 field, the preferred phosphoacceptors of the other RIP Kinases remains unstudied.

Structural studies have also recently highlighted the differences between, and the importance of, the kinase domains of this family of proteins. RIPK2 contains an extended, deep ATP binding pocket, which allows a pharmacologic manipulation likely not afforded by the other RIPKs(11, 16, 18). While molecular modeling and crystal structures have shown largely superimposable kinase domains between RIPK1, 2 and 3, there are subtle structural differences between these three kinases, which can help explain pharmacologic specificity(18). Lastly, pharmacologic inhibitors for RIPK1, RIPK2 and RIPK3 have been developed that independently target these three kinases(10–19). While structural studies have elucidated subtle differences among the kinase domains in this family of proteins, they provide only a snapshot of the protein in the lowest energy state at a single point in time. In contrast, little functional work has been done to determine potential in vivo cellular redundancy between the RIPKs. How specific are the RIPK kinase domains for their cellular function? Can one RIP kinase domain substitute for another and does the signal transduction specificity of the RIPKs rely on the kinase domain or their C-terminal effector domains? In this work, we study these central questions in the field and show that RIPK2’s kinase domain is uniquely required for innate immune signaling and NOD2-driven gene expression.

Materials and Methods

Cell lines, Plasmids, Transfection, and Western Blotting

Transient transfection assays were performed using calcium phosphate transfection of HEK293 cells (ATCC, CRL-1573) which were grown in 10% FBS, 1% Pen/Strep. Myc-K399R NEMO and HA-ubiquitin were generated as previously described(7, 27). cDNA expression constructs for RIPK1 and RIPK3 were obtained from Vishva Dixit (Genentech), and a cDNA expression construct for RIPK4 was obtained from Shiv Pillai (Massachusetts General Hospital). The template for RIPK2 was used as described(7). Gibson subcloning technology was used to insert each of the RIPKs into the NTAP expression construct (Stratagene) (28). The NTAP expression construct contains an N-terminal calmodulin binding domain and a streptavidin-binding domain. For immunoprecipitation and pulldown assays, cell lysates were prepared with a buffer containing 50 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1 mM EGTA, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 5 mM iodoacetimide, 5mM N-ethylmaleimide, 1 mM PMSF, 1 µM sodium orthovanadate, and protease inhibitor cocktail. Streptavidin beads (Sigma) were blocked with 1% BSA and added to the lysate overnight when a RIPK was precipitated. Immunoprecipitates were washed 5× in lysis buffer before boiling in an equal volume of 2× Laemmli sample buffer. Western Blotting was performed as described previously(7). For NEMO precipitation assays assessing ubiquitination, lysates were boiled before immunoprecipitation to denature the lysate and allow direct assessment of NEMO ubiquitination. NEMO was precipitated via its N-terminal 3Xmyc tag (antibody 9E10 clone, Santa Cruz). The K399R NEMO construct was utilized as this limits background ubiquitination. For signaling experiments, 10 µg/mL L-18 MDP (Invivogen) was added to the media for the given amount of time before lysates were generated using the above buffer. Protein concentrations were standardized by the Bio-Rad protein assay, and western blots were performed as described. The HA antibody (16B12) was obtained from Covance. The phosphotyrosine antibody (P-Tyr-100) was obtained from Cell Signaling, as were the pIKK, total IKK, IkBa and phospho IkBa. The GADPH antibody was obtained from Genscript. The RIPK2 antibody (H-300) was obtained from Santa Cruz Biotechnologies and recognizes the C-terminus of RIPK2 and is thus able to blot the chimeric constructs.

Viral production and stable cell line generation

Immortalized RIPK2^−/− macrophages were obtained from Michelle Kelliher (UMASS Medical School) and were grown in 10% FBS, 1% Pen/Strep. Lentiviral Crispr V2 (Addgene) was used as a Gibson subcloning template to generate the empty lentiviral construct outlined in Figure 3A. Gibson subcloning was then used to generate the retroviral constructs containing full length NTAP-tagged RIPK2 or the NTAP-tagged RIPK3/2 and RIPK4/2 chimeric constructs. HEK293 cells were transfected via calcium phosphate with pMD.2 (Addgene), psPAX (Addgene) and the RIPK lentivirus in a 1:3:4 molar ratio. Two days later, supernatant was harvested, centrifuged at 1200 rpm for 5 minutes and filtered through a 0.45µm filter. Polybrene (8µg/mL) was added to the viral supernatant and this mixture was added to the RIPK2^−/− macrophages. Two days later, cells were selected in 500 µg/mL of Hygromycin-Gold (Invivogen). Selection continued for >2 weeks. Greater than 10,000 individual colonies were pooled, and western blotting showed roughly equal expression levels of the transduced construct.

Figure 3. The kinase domain of RIPK2 is uniquely required for NOD2 signaling.

A. Schematic showing novel lentiviral construct designed to express the RIPK chimeras. The hygromycin resistance gene is cloned in frame with the self-cleaving peptide, P2A, and the NTAP-tagged RIPK chimera. A single mRNA is generated under the EF-1 promoter and upon translation; the P2A sequence allows a translational skip such that during translation, 2 proteins (hygromycin^R and the NTAP-tagged RIPK) are generated from a single mRNA. B. Immortalized RIPK2−/− macrophages were transduced with lentivirus containing no RIPK (empty), RIPK2, RIPK3/2 and RIPK4/2. Two days after transduction, cells were selected with hygromycin. After 2 weeks of selection, over 10,000 individual cell colonies were pooled. Streptavidin bead isolation and western blotting showed that the stable cell lines expressed the gene of interest. C, D. The indicated RIPK cell line was treated with 10 µg/mL of the NOD2 agonist L-18 MDP for the indicated time period. Lysates were generated, and western blotting was performed. While the empty vector line showed no signaling (consistent with RIPK2 being genetically absent), cells reconstituted with RIPK2 show a strong signaling response. Neither RIPK3/2 nor RIPK4/2 reconstituted cells showed a NOD2-dependent signaling response. In D, the final two lanes are RIPK2 reconstituted such that a positive control is present on those blots. Each given experiment was performed in at least three biologic replicates with similar results in each.

RNA isolation and qRT-PCR

The stably transduced RIPK macrophages were treated with 10 µg/mL MDP for the indicated time. Cells were then harvested and RNA extracted using a Qiagen RNeasy kit using the manufacturer’s instructions. RNA was reverse transcribed using a Quantitect reverse transcription kit (Qiagen). The following primer pairs were used for amplification: mCXCL10-F 5’ TCCTTGTCCTCCCTAGCTCA 3’, mCXCL10-R 5’ ATAACCCCTTGGGAAGATGG 3’, mGPR84-F 5’ GGGAACCTCAGTCTCCAT 3’, mGPR84-R 5’ TGCCACGCCCCAGATAATG 3’, mIRG1-F 5’ GTTTGGGGTCGACCAGACTT 3’, mIRG1-R 5’ CAGGTCGAGGCCAGAAAACT 3’, mIL-6-F 5’ GCCTTCTTGGGACTGATGCT, mIL-6-R 5’ TGCCATTGCACAACTCTTTTCT, mGAPDH-F 5’ AGGCCGGTGCTGAGTATGTC 3’, mGAPDH-R 5’ TGCCTGCTTCACCACCTTCT 3’. Sybr Green was obtained from BioRad and the real-time PCR reactions were carried out using a CFX96 C1000 Real-Time Thermal Cycler from BioRad. RT-PCR data is presented as the mean ± standard error of the mean (SEM). RT-PCR experiments were performed in duplicate and repeated three times. Significance of comparisons shown was assessed by Student’s two-tailed t-test. Significance levels are shown in each graph.

Results

Despite the homology within the kinase domains, the RIP Kinases show differential molecular abilities

The RIP kinases have been classified into a family of kinases based on homology within the kinase domains. All of the kinase domains lie in the N-terminus of the protein. C-terminal to the kinase domain, however, their domain architecture differs significantly. While both RIPK1 and RIPK3 contain RHIM domains to allow for homotypic protein:protein interactions(29), only RIPK1 also contains a Death Domain(30). RIPK4 contains Ankyrin repeats(31), and RIPK2 contains a CARD domain(21, 22), which allows it to interact with NOD2 and serve as a sensor of intracellular bacterial exposure (Figure 1A)(6). Given that there is widespread interest in targeting this family of kinases pharmacologically for diseases as diverse as autoinflammation, sepsis and autoimmunity(10–19), we sought to formally compare the molecular and biochemical activities of the RIPKs to determine unique features and functional redundancy of this kinase family. NF-κB luciferase studies showed that RIPK1, RIPK2 and RIPK4 could all activate NF-κB while RIPK3 could not (Figure 1B). Surprisingly, only RIPK2 was confirmed as a dual specificity kinase as only RIPK2 could autophosphorylate on tyrosine (Figure 1C). Lastly, every RIPK except RIPK3 could induce the ubiquitination of NEMO, a key feature of NF-κB activation(32) (Figure 1D). These findings suggest that RIPK1, 2 and 4 share similar molecular abilities to activate the NF-κB signaling pathway while RIPK3 diverges. RIPK2 uniquely autophosphorylates on tyrosine, and under these biochemical conditions, is the only dual specificity kinase among this family.

Figure 1. Comparison of the molecular activities of the RIP kinases.

A. Schematic showing the RIP kinases’ domain structure. Homology lies within the kinase domain in the N-terminus while the C-termini have differing domain architecture. B. HEK293 cells were transfected with CMV-renilla, NF-κB-driven luciferase and with 1.5 µg of the indicated RIPK construct. Transfection efficiency was standardized to renilla expression and luciferase activities were measured. RIPK1, RIPK2 and RIPK4 could activate NF-κB, but RIPK3 could not. C. HEK293 cells were transfected as indicated and streptavidin bead association isolated the individual RIPK. In vitro kinase assays were performed in the presence or absence of ATP. Only RIPK2 was able to autophosphorylate on tyrosine. D. HEK293 cells were transfected with HA-tagged ubiquitin, myc-tagged NEMO and the indicated RIPK construct. NEMO was isolated by immunoprecipitation under stringent conditions, and western blotting was performed. RIPK1, RIPK2 and RIPK4 were all able to cause NEMO ubiquitination, while RIPK3 was not. Each given experiment was performed in at least three biologic replicates with similar results in each.

Domain switching reveals that RIPK2 and RIPK4’s kinase domains are functionally similar

Given that RIPK2’s tyrosine autophosphorylation is required for downstream NOD2 signaling(11), we were surprised that the other RIPKs didn’t show tyrosine autophosphorylation activity. To determine if this activity was unique to RIPK2’s kinase domain or if it required the specific spacial proximity to the substrate present in RIPK2’s C-terminus (in which Y474 is phosphorylated(11)), synthetic biology techniques were used to generate chimeric RIPK constructs. In each of these constructs, the C-terminus of RIPK2 (immediately downstream of the kinase domain) was held constant while the kinase domains were swapped. For example, the RIPK1/2 chimera contained the N-terminal RIPK1 kinase domain with RIPK2’s C-terminus while the RIPK3/2 chimera contained RIPK3’s kinase domain with RIPK2’s C-terminus (Figure 2A). We first determined if these chimeric molecules could maintain the interaction with NOD2. NOD2 is known to interact with RIPK2 through RIPK2’s C-terminal CARD domain(6), and thus, should interact with the chimeric kinases. Western blotting following co-immunoprecipitation from transfected cells showed that all three chimeric constructs as well as WT RIPK2 could interact with NOD2, suggesting that the chimeric proteins were folding correctly and could still interact with RIPK2’s key signaling partner (Figure 2B). To then test if kinase domain swapping could biochemically function, in vitro kinase assays were performed. Of the RIPKs, RIPK4 could autophosphorylate on tyrosine only when RIPK2’s C-terminal domain was present (Figure 2C). Neither the RIPK1/2 or RIPK3/2 chimeric kinases could autophosphorylate on tyrosine (Figure 2C). Domain swapping further revealed that the RIPK3/2 and RIPK4/2 chimeras could induce NEMO ubiquitination, albeit at lower levels relative to WT RIPK2 (Figure 2D). Lastly, ubiquitination of NEMO was not sufficient for NF-κB activation as only the RIPK4/2 chimeric protein, and not the RIPK3/2 chimeric protein, could activate NF-κB (Figure 2E). These findings suggest that, like RIPK2, RIPK4 also possesses tyrosine kinase activity, as well as the ability to induce NEMO ubiquitination and cause subsequent NF-κB activation. The chimeric RIPK4/2 protein is therefore functionally similar to WT RIP2 and gives us an important tool to now dissect the uniqueness of RIPK2’s kinase domain in signaling and gene expression systems. This line of research is especially important as numerous pharmaceutical companies have RIPK inhibitors in clinical development(10–19).

Figure 2. Domain swapping reveals similar molecular activities between RIPK2 and RIPK4.

A. Schematic showing the chimeric constructs generated and utilized. The C-terminus of the constructs is identical to the C-terminus of RIPK2 while the kinase domains have been swapped as indicated. B. Co-transfection into HEK293s with the indicated constructs followed by immunoprecipitation and western blotting shows that all chimeric RIPK proteins can bind to NOD2. C. HEK293 cells were transfected as indicated and streptavidin bead association isolated the individual RIPK. In vitro kinase assays were performed in the presence or absence of ATP. While RIPK2 could autophosphorylate on tyrosines, only the RIPK4/2 chimera retained this ability. D. HEK293 cells were transfected with HA-tagged ubiquitin, myc-tagged NEMO and the indicated RIPK chimera. NEMO was isolated by immunoprecipitation under stringent conditions, and western blotting was performed. RIPK2, RIPK3/2 and RIPK4/2 were all able to cause NEMO ubiquitination to a certain degree, while RIPK1/2 was not. E. HEK293 cells were transfected with CMV-renilla, NF-κB-driven luciferase and with the indicated RIPK construct. Transfection efficiency was standardized to renilla expression and luciferase activities were measured. Of the chimeric constructs, only the RIPK4/2 chimera could activate NF-κB. Each given experiment was performed in at least three biologic replicates with similar results in each.

RIPK2’s kinase domain is uniquely required for NOD2 signaling

To then answer if RIPK2’s kinase domain is uniquely required for NOD2 signaling, we utilized synthetic biology techniques to develop a novel lentiviral expression construct (generated from the lentiCRISPR V2 construct(33)) and then made use of immortalized RIPK2^−/− macrophages. This lentiviral expression construct contains standard lentiviral LTRs, however, the EF-1 promoter drives exogenous mRNA transcription. A hygromycin resistance gene (Hyg^R) was Gibson cloned in frame to a C-terminal P2A self-cleaving peptide cassette. Finally, NTAP-tagged RIPK2, RIPK3/2 and RIPK4/2 were Gibson cloned in to the vector in frame with the P2A cassette. The end result is an expression vector which is driven by EF-1 (a promoter insensitive to NF-κB activity) and which generates a single mRNA containing the resistance gene and our gene of interest. Upon translation, the single mRNA product is generated as two individual proteins (schematic shown in Figure 3A). While the RIPK4/2 chimeric protein can both tyrosine autophosphorylate and activate NF-κB, the RIPK3/2 protein can perform neither of these functions and was therefore used as an additional negative control. Lentivirus was produced and used to infect RIPK2^−/− macrophages. After infection, cells were selected in hygromycin for 2 weeks before >10,000 colonies were pooled. Western blotting showed that while RIPK4/2 was initially expressed at a slightly lower level (Figure 3B), upon stronger hygromycin selection, levels of the exogenous proteins normalized (Figure 3C and 3D). Signaling experiments were performed. While RIPK2 expression could rescue NOD2-dependent signaling in the RIPK2^−/− macrophages, expression of empty vector (Figure 3C), RIPK3/2 or RIPK4/2 could not (Figure 3D), suggesting that despite the biochemical similarities between RIPK2 and RIPK4/2, RIPK4’s kinase domain could not replace RIPK2 in NOD2 signaling. To then further determine the extent of the signaling defect in a manner more quantifiable, NOD2-driven gene expression was studied. Our lab has previously utilized NextGen sequencing technologies to identify the NOD2-driven genes most sensitive to RIPK2’s kinase activity(25, 26). We used these genes as read-outs for gene expression. In all cases, only RIPK2 expression could rescue NOD2-driven gene expression. This was true for IL-6 (Figure 4A), CXCL10 (Figure 4B), IRG-1 (Figure 4C) and Gpr84 (Figure 4D). Together, these data suggest that despite molecular and biochemical similarities between RIPK2 and RIPK4’s kinase domains, RIPK2’s kinase domain functions uniquely – a key feature if one hopes to pharmaceutically target RIPK2 for clinical gain.

Figure 4. The kinase domain of RIPK2 is required for NOD2-driven gene expression.

A-D. The RIPK-reconstituted cells were treated with 10 µg/mL MDP for 2.5 or 5 hours. qRT-PCR was performed using expression of GADPH as an RNA quantification control. Only cells reconstituted with full length RIPK2 allowed NOD2-driven gene expression (A) IL-6, (B) CXCL10, (C) Gpr84 and (D) IRG-1.

Discussion

Despite their homology and familial grouping, the RIP Kinases participate in varied biologic functions. RIPK1 and RIPK3 are important in dictating the mechanism of cell death in response to a variety of innate immune and inflammatory signaling (1, 2, 34). RIPK2 is critically required for NOD1/2 signaling in response to intracellular bacterial exposure (35), and RIPK4 is required for proper development (9). Despite this, recent structural work has shown that a number of broad-spectrum kinase inhibitors target RIPK1, RIPK2 and RIPK3 with similar potency (16–18). This same structural work has shown that while there are subtle structural differences that may help direct medicinal chemistry toward specific inhibitors; these three kinases overlap significantly in a 3D structural context, potentially make such efforts futile (18). Additionally, work presented in this manuscript shows that they have overlapping molecular and biochemical activities. RIPK1, RIPK2 and RIPK4 all induce NEMO ubiquitination and subsequent NF-κB activation. Despite this, only RIPK2 autophosphorylates on tyrosine and is the only RIPK proven to be a dual specificity kinase. Given the interest in pharmacologically targeting a family of kinases with both similar and divergent molecular activities (10–19), it was important to determine the functional redundancy of the kinase domain between the RIPK family members. To this end, domain-swapping synthetic biology approaches were used. In this context, the only kinase domain that could replicate RIPK2 kinase domain function in an in vitro system was the RIPK4 kinase domain. Substituting RIPK4’s kinase domain for RIPK2’s allowed NOD2 binding, tyrosine autophosphorylation, NEMO ubiquitination and NF-κB activation, all key molecular events in which RIPK2 is required downstream of NOD2 activation. Surprisingly, despite the molecular similarities between the two kinase domains, the RIPK4 kinase domain could not substitute for RIPK2 in an endogenous setting. It could not support NOD2-induced signaling in RIPK2^−/− macrophages and could not replace RIPK2’s role in driving NOD2-induced gene expression. These findings suggest that RIPK2’s kinase domain is uniquely required for NOD2 signaling and cannot be replaced by even its closest homologues, implying that unique pharmacologic targeting of the RIPK family members is readily achievable.

RIPK2’s role in innate immune signaling has largely centered on its scaffolding function. RIPK2 clearly helps nucleate signaling complexes to transduce signals from NOD1 and NOD2, and genetic loss of RIPK2 does not allow signaling through the NOD1 and NOD2 receptors(5, 34, 35). The fact that overexpression of kinase-dead RIPK2 could activate NF-kB suggested that the kinase domain might be dispensable for RIPK2’s major known function(6, 21, 22). Despite this, recent work uncovering specific inhibitors of RIPK2 suggested that while initial and acute NF-kB signaling didn’t require kinase activity, optimal NOD-stimulated cytokine and gene expression absolutely require it(11, 19). This finding is supported by our prior study utilizing NextGen RNAseq methods showing that a significant subset of NOD2-induced genes require RIPK2’s kinase activity for optimal expression(25, 26). A key question that remains centers on the scaffolding function of RIPK2’s kinase domain versus its actual kinase activity. To answer this question, we utilized domain-swapping experiments in which we replaced RIPK2’s kinase domain with its closest structural homologues (RIPK4 and RIPK3). Surprisingly, we found that despite the fact that a RIPK4/2 chimera could largely replace RIPK2’s function in overexpression systems, it could not replace RIPK2’s function in more endogenous, acute signaling experiments. This finding is surprising because pharmacologic experiments have shown that while RIPK2’s kinase activity is required for optimal gene expression, it is largely dispensable for acute NF-kB signaling(19). This experiment shows that there must be structural elements of the RIPK2 kinase domain independent of its kinase activity such that RIPK4 could not replace RIPK2’s role in acute signaling. The scaffolding function and acute NF-kB signaling of RIPK2 is dependent on its kinase domain but not its kinase activity, and this scaffolding activity cannot be replaced even by RIPK2’s closest homologue.

Another interesting finding in this study centers on tyrosine phosphorylation. RIPK2 is known to be a dual specificity kinase(11), but work in this manuscript shows that this feature is not shared by the other RIP Kinase family members. Given this, it is surprising that RIPK4 is able to autophosphorylate on tyrosine when its C-terminal Ankyrin repeats are replaced by RIPK2’s intermediate and CARD domains. While native RIPK4 cannot autophosphorylate on tyrosine residues, the RIPK4/2 chimera can autophosphorylate on tyrosines, and this activity matches RIPK2’s tyrosine kinase activity. This surprising result suggests that RIPK4’s kinase domain has the intrinsic ability to be a dual specificity kinase, however, its ability to phosphorylate on tyrosine is substrate-restricted rather than kinase-activity restricted. To our knowledge, this substrate driven dual-specificity kinase activity is unique and has broader implications for the kinase field as a whole, suggesting that phosphoacceptor preferences can be altered by substrate selection rather than by intrinsic kinase structure.

Thus, in addition to categorizing and comparing the RIP Kinases to one another in terms of their ability to activate NF-kB and perform NEMO ubiquitination, this study illustrates two key features of the RIP Kinase family. First, RIPK2’s kinase domain is uniquely structured in such a way as to nucleate signaling complexes independent of its kinase activity. For this reason, its closest homologous kinase domain, RIPK4, cannot replace it structurally despite having similar kinase activity. Secondly, RIPK4 has substrate-restricted dual-specificity kinase activity that can be induced by physically fusing the substrate to its kinase domain. In the context of subsequent pharmacologic targeting of this family, the work suggests that not only might a small molecule exclusively target RIPK2, but also that by exclusively targeting RIPK2, the function of the other RIPKs might not be affected. It also suggests that by developing Type III kinase inhibitors for RIPK2 and RIPK4, one might be able to identify substrate-specific inhibitors and limit substrate phosphorylation rather than eliminate all RIPK2 or RIPK4 phosphorylation.