Synthetic cytokine receptors transmit biological signals using artificial ligands

Cytokine-induced signal transduction is executed by natural biological switches, which among many others control immune-related processes. Here, we show that synthetic cytokine receptors (SyCyRs) can induce cytokine signaling using non-physiological ligands. High-affinity GFP- and mCherry-nanobodies were fused to transmembrane and intracellular domains of the IL-6/IL-11 and IL-23 cytokine receptors gp130 and IL-12Rβ1/IL-23R, respectively. Homo- and heterodimeric GFP:mCherry fusion proteins as synthetic cytokine-like ligands were able to induce canonical signaling in vitro and in vivo. Using SyCyR ligands, we show that IL-23 receptor homodimerization results in its activation and IL-23-like signal transduction. Moreover, trimeric receptor assembly induces trans-phosphorylation among cytokine receptors with associated Janus kinases. The SyCyR technology allows biochemical analyses of transmembrane receptor signaling in vitro and in vivo, cell-specific activation through SyCyR ligands using transgenic animals and possible therapeutic regimes involving non-physiological targets during immunotherapy.

S ynthetic biology deconstructs and reassembles biological bits and pieces to construct biological devices for applications such as biological sensors, releasers, and switches 1 . Cytokine-induced signal transduction is executed by natural biological switches among many other functions control immune-related processes 2 . In principle, cytokine receptors are in an off-state in the absence of cytokines and in an on-state in the presence of cytokines. The on-state might be interrupted by negative feedback mechanisms or depletion of the cytokine and cytokine receptor. In the past, we reported ligand-independent synthetic receptors based on fusions of leucine zippers or IL-15/ sushi and the IL-6-signal transducer gp130, which are locked in the on-state, but were not switchable 3,4 . Interestingly, a marked activation of IL-6/IL-11 signaling in inflammatory hepatocellular adenomas was directly caused by gain-of-function mutations within the gp130-receptor chain, leading to ligandindependent constitutively active gp130 receptors 5 . Others described switchable synthetic cytokine receptors, resulting in gp130-induced signaling by stimulation with the cytokine erythropoietin (EPO) 6 . The major drawback of this system was that EPO has cross-reactivity with its natural EPO-receptors limiting its applications both in vitro and in vivo. Also, higher ordered multi-receptor complexes cannot be assembled using natural ligands such as EPO, which only induces receptorhomodimerization. Direct intracellular activation of signal transduction and induction of cell death was achieved using cell permeable, synthetic ligands (FK506), and binding proteins (FKBP12) resulting in homodimerization and homooligomerization 7,8 . The extent of oligomerization was, however, not controllable. Various formats of synthetic transmembrane receptors have been designed to optimize engineered chimeric antigen receptor (CAR) T-cell responses, including costimulatory receptors [9][10][11] , notch-based receptors 12 , and antigen-specific inhibitory receptors 13 . However, a switchable and background-free synthetic cytokine receptor system with full control over the assembly modus of the receptor complexes, e.g., hetero/homodimeric, -trimeric, or -multimeric is not available. Such a specific system would be a valuable tool to study receptor activation, their kinetics, stoichiometry, and biochemical properties. Moreover, background-free activation of cytokine receptors opens a great potential for novel therapeutic regimes involving non-physiological ligands during immunotherapy. Recently developed nanobodies specifically recognizing GFP and mCherry fail to bind endogenous ligands 14,15 and thus qualify as binding partners of synthetic cytokine receptors. The N-terminal region of Camelidae heavychain antibodies contains a dedicated variable domain, referred to as VHH or nanobodies, which binds to its cognate antigen. Nanobodies are single-domain antibodies of about 110 aminoacid residues generated from the variable regions of these heavy-chain antibodies 16 . Here, these nanobodies were used as extracellular sensors for homo-and heteromeric GFP:mCherry fusion proteins as part of Synthetic Cytokine Receptors (SyCyRs), ultimately leading to the formation and activation of homo-and heterodimeric and heterotrimeric receptor complexes. As biological read-out system, we use IL-23-and IL-6/ IL-11-signaling. Consequently, the extracellular sensors were fused to intracellular IL-23-and gp130-receptor chains. Using this set-up, we design a switchable synthetic cytokine receptor system, which resembles IL-23-and IL-6/IL-11-signaling and reveal that homodimeric IL-23R were biologically active. Moreover, we demonstrate that the Janus kinase activity and STAT3 phosphorylation-binding site in the intracellular domain (ICD) of the receptor can be separated on two different receptor chains, a phenomenon, which is referred to as transphosphorylation.
As depicted in Fig. 4e, 3xGFP-stimulation of Ba/F3-SyCyR(IL-6) cells resulted in time-dependent fast activation and slight inactivation of STAT3 phosphorylation after 120 min, which was accompanied by upregulation of SOCS3. Overall, 3xGFPinduced signal transduction was undistinguishable from HIL-6 induced STAT3 phosphorylation and SOCS3 expression. Next, we expressed SyCyR(IL-6) in liver tissue of C57BL/6 mice. Twenty-four hours after injection of cDNAs coding for G VHH -gp130 and 3xGFP alone or in combination, we observed STAT3 phosphorylation when G VHH -gp130 was coexpressed with 3xGFP ( Fig. 4f, g). Interestingly, G VHH -gp130 expression was found to be much higher in mice injected only with the cDNA coding for G VHH -gp130 as compared to mice expressing both, G VHH -gp130 and 3xGFP. The data suggest that co-expression of G VHH -gp130 and 3xGFP resulted in activation-dependent degradation of G VHH -gp130. Moreover, detection of 3xGFP in serum samples by western blotting showed strong accumulation of 3xGFP in mice injected only with the cDNA coding for 3xGFP, whereas 3xGFP was hardly detectable in mice injected with cDNAs coding for G VHH -gp130 and 3xGFP. These findings suggest that not only G VHH -gp130 but also 3xGFP protein was efficiently internalized and degraded in liver cells after binding and activation of G VHH -gp130 (Fig. 4f). Consistently, expression of the acute phase response gene Saa1 was increased following injection of cDNAs coding for G VHH -gp130 and 3xGFP (Fig. 4h), in sharp contrast to injection of cDNA coding for G VHH -gp130 or 3xGFP alone. Overall, our data showed that the tested SyCyRs phenocopied IL-6 and IL-23 signaling in vitro and in vivo.
During trans-phosphorylation a kinase-active receptor is able to trans-phosphorylate a kinase-negative mutant receptor. Since 2xGFP-mCherry was able to induce functional hetero-and homodimerization of SyCyRs, we wondered whether 2xGFP-mCherry can induce trans-phosphorylation of STAT3 via synthetic trimeric receptor complexes. Hence, a C-terminally truncated IL-23R (Δ503), lacking the canonical STAT-binding motifs but retained JAK, ERK, and AKT activity (IL-23R-ΔSTAT) was selected and fused with G VHH (Fig. 7a and Supplementary  Fig. 9a). Janus kinases interact with peptide motifs within the IL-23R localized between amino acid 403-479 and complete or partial deletion results in disabled JAK activity 18 . Accordingly, we created three deletion variants of the IL-23R intracellular domain with disabled JAK activity, ΔJAK-A (Δ403-417), ΔJAK-B (Δ455-479), and ΔJAK-C (Δ403-479) 20 fused to the mCherrynanobody ( Fig. 7a and Supplementary Fig. 9a). Cell surface expression of these SyCyRs in Ba/F3-gp130 cells was verified by flow cytometry (Supplementary Fig. 9b). As expected, stimulation of G VHH -IL-23R-ΔSTAT in Ba/F3-gp130 cells with 2xGFP-mCherry resulted in JAK and ERK phosphorylation but defective STAT3 activation (Fig. 7b, c). Consequently, 3xGFP-induced proliferation of Ba/F3-G VHH -IL-23R-ΔSTAT cells was drastically reduced as compared to Ba/F3-SyCyR(IL-23R) cells (Fig. 7d). Interestingly, only the assembly of a 2xGFP-mCherry-induced trimeric complex consisting of two G VHH -IL-23R-ΔSTAT receptors and one C VHH -IL-23R-ΔJAK receptor resulted in increased STAT3 trans-phosphorylation and cellular proliferation (Fig. 7c, e). Dimerization of G VHH -IL-23R-ΔSTAT with all C VHH -ΔJAK receptors by GFP-mCherry did not induce STAT activation, demonstrating that two biologically active JAKs in G VHH -IL-23R-ΔSTAT were needed for STAT3 trans-phosphorylation. Of note, also stimulation with a 2xmCherry fusion protein and formation of dimeric C VHH -IL-23R-ΔJAK did not result in STAT3 phosphorylation, whereas homodimers of C VHH -IL-23R were biologically active (Supplementary Fig. 10), demonstrating that the C VHH -IL-23R-ΔJAK variants were not biologically active as dimers. Cytokine receptors can be classified into high-affinity binders with longer ICDs and lower-affinity binders with shorter ICDs. Interestingly, high-affinity receptors with STAT-binding sites often heterodimerize with JAK1 or JAK2, whereas low-affinity receptors often pair with TYK2 or JAK3, and minimally contribute to STAT recruitment and activation 21,22 . Combining the short ICD-and TYK-binding-receptor G VHH -IL-12Rβ1 with C VHH -IL-23R-ΔJAK-B, in Ba/F3-gp130 cell lines did not result in trans-phosphorylation and proliferation (Fig. 8b, c and Supplementary Fig. 11), demonstrating that this combination fails to induce receptor activation.

Discussion
Here, we describe the development of a synthetic cytokine receptor system based on nanobodies directed against GFP and mCherry fused to truncated cytokine receptors. Nanobodies are versatile tools widely used in molecular biology, exhibiting highaffinity and antigen specificity. We chose nanobodies against GFP and mCherry because these fluorescent proteins are non-toxic to mammalian cells, will not cause unspecific binding to endogenous receptors and are, therefore, considered as side-effect/background-free 24, 25 . As receptor system, we used heterodimeric and homodimeric cytokine receptor compositions exemplified by IL-23 and IL-6 receptor signaling complexes. IL-23 signals via a heterodimeric receptor complex consisting of IL-12Rβ1 and IL-23R, whereas IL-6 signals via the non-signal-transducing IL-6R and the signal-transducing homodimer of gp130 19 . Both receptor complexes induce signals via receptor-associated Janus kinases that activate STAT, ERK, and AKT pathways 26,27 . JAKs are constitutive but non-covalently associated with class I and II cytokine receptors, which upon cytokine binding bring together two JAKs to create an active signaling complex. JAK interact with receptor peptide motifs, which are present in the intracellular domain of cytokine receptors. During receptor activation, JAKs switch into the "on''-status by reciprocal phosphorylation and subsequent phosphorylation of receptor-tyrosines and signaling molecules such as STAT3 28 .
The synthetic receptor complex mimicking IL-23-signaling was activated by a heterodimeric synthetic GFP-mCherry ligand but not by single GFP or mCherry or multimeric GFP fusion proteins, whereas the synthetic receptor complex simulating IL-6signaling was specifically activated by homodimeric synthetic GFP-ligands. Importantly, GFP-mCherry and 3xGFP fusion proteins did not activate cellular responses in cells lacking synthetic cytokine receptors.
A recent report showed that not only the intracellular domains determine the signaling strength but also the mode of extracellular receptor complex assembly 29 . Specifically, a point mutation in EPO was shown to change EPO receptor dimerization, which resulted in reduced STAT1 and STAT3 phosphorylation three times, starved, and stimulated with GFP-mCherry (100 ng/ml), HIL-23 (25 ng/ml), and HIL-6 (25 ng/ml) for 0-480 min. Stimulation with HIL-23 and HIL-6 were used as control. Cellular lysates were prepared, and equal amounts of total protein (50 μg/lane) were loaded on SDS gels, followed by immunoblotting using specific antibodies for phospho-STAT3 and STAT3, SOCS3, and β-actin. Western blot data show one representative experiment out of three. b Ba/F3-IL-12Rβ1-IL-23R cells were washed three times, starved, and stimulated with HIL-23 (25 ng/ml) for 0-480 min. Stimulation with GFP-mCherry (100 ng/ml) and HIL-6 (25 ng/ml) were used as control. Cellular lysates were prepared, and equal amounts of total protein (50 μg/lane) were loaded on SDS gels, followed by immunoblotting using specific antibodies for phospho-STAT3 and STAT3, SOCS3, and β-actin. Western blot data show one representative experiment out of three  but did not affect STAT5 activation 29 . This implies, that replacing the extracellular part of a cytokine receptor by another binding domain, such as nanobodies might influence signaling strength and kinetics, which ultimately lead to an altered intracellular response of the chimeric receptor. To exclude such effects for our synthetic cytokine receptors, apart from general analysis of typical signal transduction pathways (JAK/STAT, ERK, AKT), we verified that the time-dependent activation profile of IL-23 and IL-6 is identical with those of synthetic ligands. These findings were supported by transcriptome comparison of Ba/F3-IL-23R-IL-12Rβ1 and Ba/F3-SyCyR(IL-23/2A) cells, activated by HIL-23 and GFP-mCherry, respectively, in which almost all regulated genes for both cytokines were identical. Even though more regulated genes were detected after HIL-23 stimulation as  compared to GFP-mCherry stimulation, this difference was within expected fluctuations when using cell lines, which have been transduced with different receptor complexes and have been independently selected and cultivated for several weeks. Importantly, pathway analysis of the natural and synthetic IL-23 receptor complexes highlighted that naturally and synthetically induced signal transduction was basically identical. To the best of our knowledge, this is the first study using chimeric receptor complexes, which included a detailed analysis of signal transduction pathways and expression profiles. Importantly, the synthetic receptors appear to be active in vivo, since we demonstrated the activation of G VHH -gp130 by 3xGFP in the liver of mice after hydrodynamic injection. So far, no signaling role of IL-12Rβ1 in the receptor complex apart from activation of a Janus kinase has been assigned 20 . Consistently, using our synthetic receptors, we induced IL-23R homodimeric receptor complexes, as we have recently suggested 30,31 . Although the general activation of signaling pathways appeared to be similar to IL-23 signaling, gene-array analysis revealed a reduced number of regulated genes when compared to heterodimeric signaling. This phenomenon could be caused by the slightly reduced proliferation observed following SyCyR(IL-23R) signaling when compared to natural IL-23R/IL-12Rβ1 complex. Moreover, reduced affinity or biochemical features of the synthetic ligand may affect SyCyR signaling. This effect might also contribute to the observation, that the natural and synthetic IL-23 receptor activation was different in terms of the absolute number of regulated genes.
The modular nature of the synthetic ligands, with one receptor binding site per GFP or mCherry allows an exact composition of the receptor stoichiometry, which clearly will be interesting for many if not all other cytokine receptors. Moreover, this system will enable the combinatory assembly of uncommon receptor combinations with desirable signaling potentials and capacities. The number of recruited synthetic receptors is only limited by the maximal number of ligands connected in one GFP:mCherry fusion protein or by alternative GFP/mCherry multimerization strategies 32,33 .
Two recent reports describe surrogate dimeric ligands, which specifically bind and activate natural cytokine receptors supporting our hypothesis that receptor activation might be generally possible by dimeric ligands 34,35 . For IL-6 and IL-23 signaling, we used homo-and heterodimeric GFP:mCherry fusion proteins, but we were also able to generate homo-and heterotrimeric GFP: mCherry variants. Using these synthetic ligands, we analyzed, if biologically active trimeric receptor complexes could also be functionally assembled among the cytokine receptor family with associated kinases. Tyrosine-receptors and receptors with associated kinases are typically active as dimers 36 to juxtapose and subsequently activate at least two receptor kinases. This implies that these receptor systems naturally did not require a third receptor. To generate trimeric receptor complexes for IL-6 and IL-23-simulations, we deleted the STAT3-binding motifs in the synthetic G VHH -IL-23 and G VHH -gp130 receptors and combined these receptors with JAK-deficient receptors containing STATbinding motifs fused to C VHH . We showed that in these trimeric receptor combinations, STAT3 activation is mediated by transphosphorylation. The formation of a trimeric receptor complex with two JAK-proficient but STAT-deficient receptors and one STAT-binding motif receptor by GFP-GFP-mCherry (2xGFP-mCherry) resulted in STAT3 trans-phosphorylation. Assembly of one JAK-proficient receptor with one STAT-binding motif receptor by a GFP-mCherry fusion did, however, not lead to trans-phosphorylation, confirming that one Janus kinase is not sufficient for receptor activation. Of note, trans-phosphorylation was thus far only described for tyrosine-kinase-receptors of the PDGF and EGF family [37][38][39][40][41] , in which a kinase-active receptor was able to trans-phosphorylate a second kinase-inactive mutant receptor after receptor dimerization. In these cases, the kinasenegative mutant receptor was able to activate the functional kinase of the other receptor. Here, we describe for the first receptor-chain trans-phosphorylation for cytokine receptors with associated Janus kinases.
In summary, the synthetic cytokine receptor system allows tailor-made activation and analysis of cytokine signaling by recruitment of defined numbers and compositions of receptor chains. Receptor assembly is determined by the number and sequence of GFP-mCherry units in the ligand fusion proteins. This system simulates signal transduction without relevant background activation that has been described previously with chimeric receptor systems 6 . The lack of toxicity of fluorescent proteins in vitro and in vivo allows a widespread area of potential applications for studying cell-type specific receptor activation by synthetic ligand application in transgenic mice. Importantly, our system is easily on/off-switchable, because signal activation can be rapidly inhibited by application of soluble nanobody-fusion proteins directed against the synthetic GFP:mCherry ligands and will open up therapeutic regimes involving non-physiological targets during immunotherapy.
Expression and purification of G VHH -C VHH from E.coli. The cDNA coding for the bispecific antibody G VHH -C VHH was generated and subcloned in pet23a. The resulting bispecific antibody sequence was flanked by an N-terminal PelB leader sequence for periplasmic expression and a 3' hexahistidine sequence for purification. Proteins were expressed in the E.coli strain BL21-Rosetta. Bacteria were incubated in 2 l LB-media containing ampicillin 1:1000 (100 µg/ml) and chloramphenicol 1:1000 (34 µg/ml) at 37°C, until optical density reached 0.6-0.9. Then 1 mM IPTG was added. Bacteria were harvested by centrifugation (5000 × g, 30 min, 4°C) 4 h after IPTG induction. A cOmplete protease inhibitor tablet (Roche, Mannheim, Germany) was added and supernatant was filtered through a 0.45 µm bottle top filter. Proteins were purified via IMAC chromatography and eluted with 500 mM imidazole.
Cell viability assay. To remove the cytokines, Ba/F3-gp130 cell lines were washed three times with sterile PBS. In all, 5 × 10 3 cells were suspended in DMEM supplemented with 10% FCS, 60 mg/l penicillin and 100 mg/l streptomycin and cultured for 3 days in a final volume of 100 μl with or without cytokines/fluorescent proteins as indicated.  Imaging Instruments GmbH, Göttingen, Germany) were used for signal detection.
For re-probing with another primary antibody, the membranes were stripped in 62.5 mM Tris-HCl (Carl Roth, Karlsruhe, Germany) pH 6.8, 2% SDS (Carl Roth, Karlsruhe, Germany) and 0.1% β-mercaptoethanol (Sigma Aldrich, Munich, Germany) for 30 min at 60°C and blocked again. The band intensities of the western blots (Fig. 4f) were quantified using ImageJ software. Uncropped images of western blots are presented in Supplementary  Fig. 13.
Cell surface detection of cytokine receptors. To detect cell surface expression of the synthetic cytokine receptors, stably transduced Ba/F3-gp130 cells were washed with FACS buffer (PBS containing 1% BSA) and incubated at 5 × 10 5 cells/100 μl FACS buffer supplemented with a 1:100 dilution of anti-myc (71D10), (cat. #2278) or 1.2 µg anti-Flag (DYKDDDDK) (cat. F7425) mAbs for 1 h on ice. After a single wash with FACS buffer, cells were incubated in 100 μl FACS buffer containing a 1:100 dilution of Alexa Fluor 488 conjugated Fab goat anti-rabbit IgG (cat. A11070) for 1 h on ice. Finally, cells were washed once with FACS buffer, suspended in 500 μl FACS buffer and analyzed by flow cytometry (BD FACSCanto II flow cytometer, BD Biosciences, San Jose, CA, USA). Data was evaluated using the FCS Express 4 Flow software (De Novo Software, Los Angeles, CA, USA).