Ligand‐dependent protein interactions of the juvenile hormone receptor captured in real time

Juvenile hormone (JH) signalling provides vital regulatory functions during insect development via transcriptional regulation of genes critical for the progression of metamorphosis and oogenesis. Despite the importance of JH signalling, the underlying molecular mechanisms remain largely unknown. Our current understanding of the pathway depends on static end‐point information and suffers from the lack of time‐resolved data. Here, we have addressed the dynamic aspect of JH signalling by monitoring in real time the interactions of insect JH receptor proteins. Use of two tags that reconstitute a functional luciferase when in proximity enabled us to follow the rapid assembly of a JH receptor heterodimer from basic helix–loop–helix/Per‐Arnt‐SIM (bHLH‐PAS) proteins, methoprene‐tolerant (Met) and taiman (Tai), upon specific JH binding to Met. On a similar timescale (minutes), the dissociation of Met‐Met complexes occurred, again strictly dependent on Met interaction with specific agonist ligands. To resolve questions regarding the regulatory role of the chaperone Hsp90/83 in the JHR complex formation, we used the same technique to demonstrate that the Met‐Hsp83 complex persisted in the agonist absence but readily dissociated upon specific binding of JH to Met. Preincubation with the Hsp90 inhibitor geldanamycin showed that the chaperone interaction protected Met from degradation and was critical for Met to produce the active signalling dimer with Tai. Thus, the JH receptor functions appear to be governed by principles similar to those regulating the aryl hydrocarbon receptor, the closest vertebrate homologue of the arthropod JH receptor.

JHs comprise a group of insect-specific, epoxidated and methylated sesquiterpenoids produced by the corpora allata glands [9,10]. Intracellular receptors that bind JHs with high affinity have been identified as protein products of the Methoprene-tolerant (Met) gene [11] and its orthologs in the flour beetle Tribolium castaneum [12], the fruit fly Drosophila melanogaster [13,14] and other insects [15,16]. Unlike nuclear receptors for other lipophilic hormones, the JH receptor (JHR) Met belongs to the bHLH-PAS family of transcription factors [17]. Upon binding an agonist ligand, such as a native JH or its synthetic mimic, Met dissociates from a homophilic (Met-Met) complex of unknown stoichiometry [12,18] and associates with another bHLH-PAS protein Taiman (Tai) [12,15,[19][20][21]. Tai, also called SRC for steroid receptor coactivator, was discovered in D. melanogaster as a coactivator to the ecdysone (insect steroid hormone) receptor [22]. The Met-Tai complex is a 1 : 1 dimer [23] that drives the transcription of genes containing JH response elements (JHREs) by binding to the corresponding DNA sequence [14,15,19,21,24].
While the above critical components of JHR signalling have been established for more than a decade [17], the regulatory mechanisms remain poorly understood. The structure of the JHR protein Met has not yet been resolved, but its homology to vertebrate bHLH-PAS family members such as the hypoxia factor (HIF-2a) or aryl hydrocarbon receptor (AhR) presents an opportunity to draw analogies regarding their signalling mechanism. One of the shared features is that AhR [25,26] and Met [27] both require the highly conserved chaperone Hsp90 (homologous to insect Hsp83) for their activity. Hsp90 is implicated in many physiological processes and has a broad range of clients, i.e. proteins whose folding and/or stability rely on interactions with Hsp90 and an underlying cochaperoning machinery. Hsp90 typically protects and stabilises the client molecule, such as AhR, in a form competent for ligand binding; upon contact with the ligand, conformational changes lead to the complex dissociation, nuclear import and interaction with partners involved in active signalling [25,[28][29][30]. A similar model has been suggested by a proteomic analysis of proteins pulled down with T. castaneum Met, which detected dissociation of Hsp83 and enrichment of Tai in the presence of a JHR agonist [23].
Most of the information obtained so far on JHR signalling has relied on insect genetics, cell-based reporter assays reliant on transcriptional activation, protein and chromatin immunoprecipitation and immunostaining in fixed tissues and cells, and in vitro binding assays (see [17] for a review). In all cases, the data represented a snapshot of an end-point status. A method providing temporal resolution to JHR signalling has been missing. That prompted us to seek an approach that would permit monitoring of JHR activity in real time and independently of de novo transcription and translation. To this end, we exploited the NanoLuc Binary Technology (NanoBiT) where two interacting proteins, each tagged with a fragment of the NanoLuc luciferase, bring to proximity the NanoLuc fragments, which complete the functional enzyme [31]. The luminescence then emits from live cells, preloaded with the furimazine substrate. The NanoBiT system has recently been employed to monitor the entry of an insect steroid hormone, 20hydroxyecdysone, into mammalian cells [32].
In this study, we implemented the NanoBiT technology to follow the time course of agonist-dependent assembly of the Met and Tai proteins into the active JHR complex and the dissociation of Met-Met complexes in response to JH. We additionally applied the method to monitor the hitherto poorly understood effect of JH binding on the dissociation of Met from a complex with the insect ortholog of Hsp90. Use of specific JHR agonists, mutated receptor proteins and Hsp90 inhibition helped us to further the current understanding of the molecular events triggered by JH, suggesting NanoBiT as a suitable approach for studying the dynamics of hormonal signalling in general.

Dynamics of agonist-dependent assembly of the Met-Tai JHR complex
Mutual interaction between the JHR complex components Met and Tai to a large extent depends on agonist binding to Met [12,14,23,33]. We initially used the NanoBiT system to monitor the JH-induced interaction between the Met and Tai proteins from the beetle T. castaneum. Met and Tai were N-terminally tagged with the large (LgBiT) and small (SmBiT) fragments, respectively, of the NanoLuc luciferase ( Fig. 1A and Table 1). Met and Tai versions with the LgBiT and SmBiT tags swapped (Table 1) have also been tested (Fig. S1A). The pairs of fusion proteins were expressed in transfected CHO cells, which were then permeated with a furimazine substrate. Dimerisation between the tagged Met and Tai partners resulted in the structural complementation of NanoLuc and in the subsequent emission of luminescence. Raw data for the luminescence traces before and after normalisation to baseline levels are shown in Fig. S1A. The protein pair with Met fused to LgBiT and Tai fused to SmBiT, hereafter abbreviated LgMet and SmTai, was chosen for further experimentation.
Exposure of cells expressing LgMet and SmTai to the native T. castaneum hormone (JH III) or the synthetic JHR agonist fenoxycarb triggered rapid Met-Tai dimerisation that approached maximum within 10-15 min after agonist addition and plateaued thereafter (Fig. 1B,C). Both agonists acted in a dose-dependent manner within the range of concentrations consistent with their expected activity. The K d for JH III binding to Met has been reported to be~3 nM [12], and fenoxycarb is typically one or two orders of magnitude more potent than the native hormone in cell-based assays [33]. Accordingly, appreciable NanoLuc activation was achieved with 1 nM JH III and with 10 pM fenoxycarb (Fig. 1B,C). These data showed that in the heterologous intracellular environment, Met and Tai dimerised within minutes of agonist presence.
To confirm that direct agonist binding was required in the NanoBiT system for Met to interact with Tai, we mutated a critical threonine 254 within the ligandbinding pocket of Met. Prior work has established that the T254Y substitution effectively prevents JH III binding without compromising the stability of the Met protein [12]. As expected, unlike the functional 'wildtype' version (LgMet WT ), the LgMet T254Y mutant protein failed to dimerise with SmTai upon the addition of JH III, although the WT and T254Y Met variants were expressed to a similar extent (Fig. 1D).
We next tested the NanoBiT assay for the capacity to discriminate between active and inactive stereoisomers of JH. Our recent work [33] has demonstrated  strong stereoselectivity of a D. melanogaster JHR protein Gce (a paralog of Met) towards the JH homologue JH I. While JH I is native to lepidopteran insects such as the silkmoth, the PAS-B domain of T. castaneum Met has been shown to respond equally to JH III and JH I [34]. Therefore, we were able to utilise a set of JH I stereoisomers with different binding affinities, agonist potencies and in vivo biological activities we had determined previously [33]. The stereoisomers of JH I were tested using two independent methods: the NanoBiT and an earlier established two-hybrid assay [35]. In the two-hybrid assay, VP16-Met and Gal4-Tai fusion proteins were co-expressed in CHO cells carrying a NanoLuc reporter under the UAS elements recognised by Gal4. In agreement with data for the D. melanogaster JHR [33], the JH I stereoisomer with the natural configuration 10R,11S-(2E,6E) proved most potent in stimulating the Met-Tai assembly (EC 50~0 .5 nM; Fig. 2A). Almost thousandfold higher concentration (EC 50 of~440 and 430 nM) was required for activation by JH I stereoisomers 10R,11S-(2E,6Z) and 10R,11S-(2Z,6E), whereas 10R,11R-(2Z,6Z) was inactive ( Fig. 2A). The concentration of 10 nM was selected for the NanoBiT assay, which confirmed the native 10R,11S-(2E,6E)-JH I as the best inducer of Met-Tai dimerisation. The geometric isomers (2E,6Z) and (2Z,6E) both displayed a substantially lower activity relative to the native hormone, and the biologically inactive compound 10R,11R-(2Z,6Z)-JH I had no appreciable effect in the NanoBiT assay (Fig. 2B). Therefore, data obtained from the two-hybrid and NanoBiT systems were highly consistent. Importantly, the NanoBiT data corroborated the relationships between the agonist ligand activities of the JH I stereoisomers [33] and their effects on Met-Tai dimerisation.

JH-dependent dissociation of a Met-Met complex
An early immunoprecipitation study [18] has indicated that in the absence of JH, D. melanogaster Met occurs within a complex containing Met itself or its paralog Gce. A Met-Met complex from T. castaneum has later been shown to dissociate in a manner dependent on agonist binding to Met [12]. To address the dynamics of Met-Met interaction in the context of living cells, we co-expressed two fusion variants of T. castaneum Met, one with the N-terminal LgBiT (LgMet) and the other with C-terminally attached SmBiT (MetSm) of the NanoLuc luciferase ( Fig. 3A and Fig. S1B, Table 1).
The relatively high luminescence produced by LgMet and MetSm in the absence of agonist, which likely represented a Met-Met dimer, declined within minutes of exposure to JH III (Fig. 3B), i.e. on a timescale similar to that required for Met to associate with Tai (Fig. 1B). The presumed Met-Met dimer also dissociated upon addition of natural JH I, but not its inactive stereoisomer 10R,11R-(2Z,6Z) (Fig. 3B), indicating that Met-Met dissociation required specific agonist activity.
To further demonstrate the dependence of Met dissociation from the homophilic complex on agonist binding, we introduced the point mutations that disable ligand binding to Met. When one of the  Only when both subunits carried the T254Y mutation, no dissociation occurred upon exposure to JH III ( Fig. 3D and Fig. S1B). NanoBiT assays performed with another mutated Met variant V297F lacking the JH-binding capacity [12] yielded similar data, as dimers composed of LgMet T254Y and Met V297F Sm also remained stable in the presence of JH III ( Fig. 3E and Fig. S1C). These results indicated that engaging at least one of the two binding sites within the dimer with the agonist was necessary and sufficient for the Met-Met complex to fall apart.

Interaction of Met with the chaperone Hsp83/90
By analogy with the homologous ligand-activated receptor AhR [25,26], Met lacking an agonist is expected to reside within a complex with Hsp83, the insect ortholog of Hsp90. Agonist binding should liberate Met from Hsp83 and make it available for interaction with Tai, as has been suggested by our recent proteomic study [23]. On the contrary, an earlier model proposed by He et al. [27] has suggested that Hsp83 binds Met in a ligand-dependent manner and engages in a transcriptionally active Met protein complex. To resolve this controversy and gain insight into the steps leading from the chaperone complex to the transcriptionally active Met-Tai dimer, we employed the NanoBiT system. Initially, we used geldanamycin, an inhibitor of the N-terminal ATPase activity of Hsp90 that affects client maturation and stability [28,36], to block the activity of endogenous Hsp90 in the host CHO cells. Preincubation of the cells expressing LgMet and SmTai proteins with geldanamycin for 45 min before JH application markedly decreased the rate of Met-Tai dimerisation (Fig. 4A), which appeared unaffected when JH III and geldanamycin were added simultaneously (Fig. 4B). Immunoblot analysis of CHO cells exposed to geldanamycin for 1 h (i.e. roughly a period including the 45-min preincubation plus the time required for monitoring the complex dynamics) showed that the LgMet protein levels were significantly lowered relative to those in control cells (Fig. 4C,D). Depletion of LgMet likely reduced the amount of dimerisation. These data suggested that Hsp90 activity was required to stabilise the transfected LgMet protein prior to its binding with an agonist ligand.
We next examined the interaction between the T. castaneum Met and Hsp83 proteins directly. Tribolium castaneum Hsp83 was N-terminally tagged with the LgBiT or SmBiT fragment of the NanoLuc luciferase ( Table 1). The protein pairs comprising either LgHsp83 plus MetSm or SmHsp83 plus LgMet were expressed in CHO cells (Fig. 5A). Either combination resulted in a robust baseline luminescence signal that started declining immediately after JH III or JH I application ( Fig. 5B and Fig. S1D). By contrast, the baseline luminescence was unaffected by the inactive JH I stereoisomer 10R,11R-(2Z,6Z) (Fig. 5B). The Hsp83-Met complex also remained stable when functional (WT) Met was replaced with the Met T254Y or Met V297F mutant variants incapable of binding JH ( Fig. 5C and Fig. S1D), indicating that dissociation of Met from Hsp83 specifically required JH binding to Met. Interestingly, preincubation with the Hsp90 inhibitor geldanamycin also decreased the ability of Met to dissociate from the LgHsp83-MetSm complex after the addition of JH III (Fig. 6), suggesting a possible role of the ATPase activity of the chaperone in ensuring the ligand-binding competence of Met.
Collectively, the present data support a model in which agonist binding likely results in a conformational change of Met that in turn leads to its dissociation from the unliganded Met-Met and Met-Hsp83 complexes. This JH-dependent release apparently enables Met to form a heterodimer with Tai, which becomes engaged in transcriptional activation (Fig. 7).

Discussion
Identification of Met as a JH receptor [12,14] and Tai as its partner essential for the hormonally regulated transcriptional activation [15,[19][20][21]23] paved the way for studies of the underlying molecular mechanisms that govern insect development and reproduction. However, present knowledge on JHR signalling mostly derives from transcriptional assays that provide a static description of an end-point situation in cell lysates.
Aiming to monitor the real-time dynamics of agonist-dependent assembly and dissociation of JHR protein complexes taking place in live cells, we utilised the NanoBiT technology. A related approach based on bioluminescent resonance energy transfer (BRET) has previously proven successful to study interactions of a crustacean ortholog of Met in response to JHR agonists [37,38]. However, the method has only been applied to afford end-point status rather than timeresolved information. The NanoBiT technology has quite recently been adopted to study signalling by another important arthropod hormone, the steroid 20hydroxyecdysone (20E) [32]. Although the assay was not functional with a native insect 20E receptor, a heterodimer of the ecdysone receptor (EcR) and ultraspiracle (Usp) proteins from D. melanogaster [39], it performed well with a dimer of EcR and the human Usp ortholog RXR and thus has found use in monitoring the entry of 20E into the cell [32].
Here, we have established the NanoBiT technology as a method reliably reflecting specific effects of agonist binding on JHR activity. This is evidenced by the strict selectivity of Met towards the native JH I stereoisomer and by the inactivity of the mutated Met variants incapable of binding JH. The NanoBiT assay faithfully recapitulated a former model in which a homophilic Met-Met complex dissociated while the transcriptionally competent Met-Tai dimer assembled as a function of agonist binding to Met [5,12]. In addition, our study revealed new information that, within the cellular environment, both processes occur on a similar timescale of several minutes. This time is required for JH to penetrate cells, bind Met, and induce a hypothetical conformational change leading to the switch from Met-Met and/or Met-Hsp83 complexes to Met-Tai interaction (Fig. 7). The data indirectly suggest that the Met-Tai dimer forms independently of binding to a target DNA response element. Together, the results generated using the NanoBiT system further our insight into the dynamics of JHR complex formation and functioning.
The composition of receptor complexes, their localisation, stability and dynamics of their formation all contribute to the signalling outcome [40][41][42]. Signalling of the multifaceted bHLH-PAS family members presents an added layer of complexity with some proteins having dual roles as transcription factors and receptors and can also be affected by events beyond the  immediate receptor complex formation. Essential roles of the chaperone Hsp90 in AhR or HIF maturation and activity ranging from proper folding and stabilisation to assuming a conformation of the client competent for ligand binding have been extensively documented [36,[43][44][45][46][47]. Studying the dynamics of these interactions provides a framework necessary for understanding mechanisms underpinning the formation of actively signalling receptor complexes. The ability to associate with the Hsp90/83 chaperone has previously been reported for Met orthologs of D. melanogaster [27] and T. castaneum [23]. The former study proposed a model where a JHR agonist stimulated the interaction of Met with Hsp83, which in turn facilitated the nuclear import of the liganded Met complex through the action of b-importin [48]. Hsp83, rather than Tai, was detected among proteins associated with the target DNA elements bound by Met [27]. However, this model partly contradicts the accepted AhR mechanism, where agonist binding leads to the separation of AhR from Hsp90 and consequent formation of the transcriptionally active dimer of AhR with its partner Arnt from which Hsp90 is excluded [29,30]. Consistent with the AhR model, the latter study [23] detected agonist-dependent loss of Hsp83 and some associated cochaperones from Met interactome, raising a question of the precise role of the Hsp83 in JH signalling.
By using the NanoBiT technology, we have now confirmed in live cells and in real time that the specific agonist binding to Met occurring in a complex with Hsp83 causes Met to dissociate from Hsp83 while, on a similar timescale, the Met-Tai dimer forms. This complies with the model for AhR action, where the receptor remains associated with the chaperones prior to agonist binding but is distinct from the mode of interaction proposed for Drosophila Met by He et al. [27]. Further studies are warranted to fill in the missing details on the pathway mechanisms.
With the limited structural information available on Met, Tai and insect Hsp83, the analogy with mammalian counterparts may provide clues useful in the search of mechanistic information. While there clearly are client-specific chaperoning interactions [49][50][51][52], a general mechanism emerges for ordered client proteins such as bHLH-PAS transcription factors or steroid hormone receptors. A nascent client, stabilised by interaction with Hsp70, is presented to the Hsp90 dimer, then the ATPase activity of Hsp90 triggers Hsp70 release. In the newly formed complex, the client retains a partially unfolded segment threaded through the two Hsp90 molecules while its remaining parts are folded and protrude from the complex. When the ATPase is inhibited, clients display reduced stability and become degraded by the proteasome. Thus, it has been shown that the interaction with Hsp70 keeps the glucocorticoid receptor (GR) in an inactive state unable to bind ligands; the ligand-binding competence only ensues after the ATP hydrolysis and the transition to the Hsp90 co-chaperoning complex [53]. Based on cryo-EM structures of Hsp90-Cdk4 [54] and GR-Hsp90 [55], the client-bound Hsp90 dimer is in a closed conformation, with the C-terminal domains connected but the N-terminal domains apart. In this conformation, the client can interact with its ligand. A similar mechanism can be envisioned for AhR, which is known to interact with Hsp90 via residues within the PAS-B domain that overlap with the ligandbinding site [29,43,56,57]. The PAS-B domain of AhR may also be partially unfolded in the Hsp90 chaperone complex but rearranges upon ligand binding, leading to exposure of the nuclear localisation signal, displacement of AhR from Hsp90 and formation of the AhR-Arnt dimer through some proposed transitional states [30].
In the context of the AhR model, our study suggests that in the absence of JH, Met is bound to the Hsp83/ 90 chaperone complex, presumably also with a partially unfolded segment threaded through, from which Met dissociates rapidly upon hormone addition. This release requires specific ligand binding and is prevented by geldanamycin, an established inhibitor of the Hsp90 ATPase, possibly as ATP hydrolysis is required for Met maturation and ligand-binding competence. In the presence of geldanamycin, Met may be retained within an inactive complex with Hsp70 and Hsp83 and eventually targeted for degradation. Indeed, we have detected a reduction in Met protein levels in geldanamycin-treated cells.
Unlike AhR, unliganded Met is known to form homo-oligomers [12,18,37]. The origin, stoichiometry and function of such homophilic complexes remain unknown. As shown here, JH caused the Met-Met complex to fall apart rapidly, similar to the dissociation of Met from Hsp83. In both cases, the dissociation depended on specific ligand interaction. Using two distinct mutated variants (Met T254Y and Met V297F ) incapable of binding JH, we demonstrated that both Met molecules within the binary complex had to be mutated to prevent their ligand-induced disassembly. These data perfectly corroborated an earlier report based on co-immunoprecipitation of functional and mutated (Met V297F ) Met proteins in lysates of cells exposed to a JHR agonist [12]. However, the mechanistic relationship between the Met-Met and Met-Hsp83 complexes cannot be ascertained with our current knowledge. In most reported cases, clients bind asymmetrically as monomers to a single site within the Hsp90 dimer, so it is unclear whether the chaperone complex can accommodate dimeric clients. The parallel existence of homodimers and Hsp90 complexes has been documented for the oestrogen receptor b, albeit without structural details [58]. Therefore, while Hsp83 could theoretically bind a Met homodimer, it is also possible that separate pools of Met coexist as Met-Met and Met-Hsp83 complexes. By monitoring binary interactions, the NanoBiT assay cannot resolve these alternatives.
The Met-Tai dimer probably forms in the nucleus as that is where Tai resides [22]. Met and Gce paralogs from D. melanogaster contain nuclear export signals and signals for constitutive and JH-stimulated nuclear import [59,60]. Nuclear localisation of T. castaneum Met requires a conserved bipartite basic motif [23]. Earlier studies have found that D. melanogaster Met enters the nucleus in a manner reliant on association with Hsp83 [27,48]. The nuclear import of Met may thus resemble that of AhR, as geldanamycin inhibition of Hsp90 has been shown to cause cytoplasmic retention of AhR even in the presence of a strong AhR agonist dioxin (TCCD) [28]. As the NanoBiT signal develops independently of intracellular localisation, the method does not address the role of nucleocytoplasmic trafficking in JHR protein interactions.
Despite its constraints, the NanoBiT technique has demonstrated its great potential to improve our grasp of the dynamics governing JHR signalling. In summary, this study links the chaperoning activity of Hsp83/90 and hormone binding to the Met protein with Met-Tai dimerisation. The chaperone interaction protects Met, as inhibition of Hsp90 leads to Met degradation. We have shown that prior to agonist exposure, Met occurs in the Hsp83 chaperone and/or Met-Met complexes. This also applies to mutated Met variants incapable of binding JH, which, however, remain trapped in the unproductive complexes even in agonist presence. Our experiments could not determine whether the Met-Hsp83 and Met-Met complexes overlap or exist as separate pools. In either case, JH binding causes Met to dissociate from these assemblies and dimerise with Tai with similar kinetics. These steps help to shape the molecular landscape during the formation of the transcriptionally active JHR signalling complex.

Compounds
Racemic 10R,S-JH III was purchased from Sigma-Aldrich (St. Louis, MO, USA). JH I in the natural configuration 10R,11S-(2E,6E)-JH I and its three stereoisomers were provided by Dr K. Slama and were previously verified using NMR [33]. Working JH stocks were prepared in ethanol. The synthetic JHR agonist fenoxycarb (Sigma-Aldrich) and the Hsp90 inhibitor geldanamycin (MedChemExpress USA, Monmouth Junction, NJ, USA) were dissolved in DMSO.

Cell culture
The NanoBiT and two-hybrid experiments were performed in CHO cells propagated in Dulbecco's Modified Eagle Medium (DMEM; ThermoFisher Scientific, Waltham, MA, USA) with 10% heat-inactivated fetal bovine serum and 1% penicillin/streptomycin (both Sigma-Aldrich) and maintained at 37°C, 5% CO 2 and 85% humidity. All transfections of plasmid DNAs were done using the FuGENE-HD reagent (Promega, Madison, WI, USA) following the manufacturer's instructions.

Construction of NanoBiT vectors
All constructs expressing fusions of Met with the NanoLuc luciferase fragments encoded the entire T. castaneum Met protein (M1-V516; NCBI Accession NP_001092812.1) with added N-terminal Myc epitope, separated from the Nano-Luc fragment by a linker of 18 amino acids (Fig. 1A). The Met coding sequence was optimised for mammalian codon usage and custom synthesised (GenScript, Piscataway, NJ, USA). Constructs expressing NanoLuc fragments fused with Tai contained a synthetic DNA sequence (GenScript) encoding amino acids M1-V505 of T. castaneum Tai (NCBI Accession XP_008193629.1) and an N-terminal Flag epitope (Fig. 1A). The C-terminally truncated Tai protein included the bHLH and both PAS domains required for dimerisation with Met [12,23]. Constructs with N-terminal tags of large or small NanoLuc fragments were generated in the pBiT1.1-N [TK/LgBiT] or pBiT2.1-N [TK/SmBiT] (Promega) vectors, respectively. Met constructs tagged with the small NanoLuc fragment at the C-terminus were created in the pBiT2.1-C [TK/SmBiT] vector (Promega). Vectors for expressing mutated variants of T. castaneum Met with the single residue substitutions T254Y and V297F were generated using quick-change mutagenesis with complementary PCR primers, replacing the corresponding original nucleotides. The forward and reverse primers for Met T254Y were 5 0 -CCGCGAGGAGTACGTGTACCGCCA CCTGATCGATGGCCGC-3 0 and 5 0 -GCGGCCATCGAT CAGGTGGCGGTACACGTACTCCTCGCGG-3 0 , respectively. Primers 5 0 -GCGAGGATGTGCGCTGGTTCATGA TCGCCCTGCGCCAGATG-3 0 and 5 0 -CATCTGGCGCA GGGCGATCATGAACCAGCGCACATCCTCGC-3 0 were used to clone the Met V297F variant. The mutated nucleotide sequences are underlined. The T. castaneum hsp83 coding sequence was isolated from cDNA originating from the TcA cell line [61] using reverse-transcription PCR with the  Table 1 lists all protein constructs and their combinations used in the NanoBiT system. For Met-Tai interaction, both combinations with the tags swapped (i.e. LgMet plus SmTai and SmMet plus LgTai) were functional in the assay, but for simplicity, only the former was used beyond initial testing. In the Met-Tai and Met-Hsp83 interaction assays, combining N-terminally tagged constructs provided a robust signal, which did not critically depend on the identity of the tags attached to the individual proteins. For Met-Hsp83 interaction, using Met with either C-terminal SmBiT or N-terminal LgBiT tags worked equally well. However, the detection of the homophilic interaction of Met required one of the partners to have a tag on the Cterminus (Table 1).

NanoBiT luciferase assay
Assembly or dissociation of the Met-Tai, Met-Met and Met-Hsp83 complexes was monitored using the NanoBiT technology (Promega). Construct pairs individually subjected to the assay are listed in Table 1. CHO cells were plated semiconfluent in flat-bottom, solid white 96-well plates (#3917, Corning Life Sciences, Tewksbury, MA, USA). After 24 h, the cells were transfected with 50 ng DNA per well of each vector for the Met-Tai pairs and double the amount for all of the remaining combinations. Another 24 h later, the cells were equilibrated for 15 min at room temperature in serum-free DMEM medium buffered with 20 mM HEPES, pH 7.2 (HDMEM). The Nano-Glo Live Cell Substrate (Promega) was added and the baseline luminescence was recorded twice for 10 min or until stable using the Orion II microplate luminometer (Berthold Technologies, Bad Wildbad, Germany). Tested compounds or corresponding vehicle controls alone were added manually from 109 stocks in HDMEM, and luminescence was recorded for further 20 min. Baseline raw luminescence values slightly varied, probably depending on the transfection efficiency and, therefore, the level of dimerisation was determined as fold activation by normalising the measured signal to the baseline luminescence values for each individual well. This dramatically reduced the apparent variability and resulted in excellent reproducibility between experiments (Fig. S1). Respective solvents (DMSO or ethanol) at 0.1% or lower final concentrations were used as vehicle controls. For Hsp90 inhibition, geldanamycin or the vehicle alone were added with the HDMEM so that the total incubation time was 45 min.

Two-hybrid assay
The pACT vector (Promega) expressing the VP16 transcription activating domain fused to the N-terminus of the T. castaneum Met protein (M1-A452; NCBI Accession NP_ 001092812.1) was described in [35] and kindly provided by Dr H. Miyakawa. Synthetic DNA sequence encoding amino acids M1-V505 of T. castaneum Tai (same as in the Nano-BiT constructs above) was cloned to a pDBD vector based on the pcDNA3.1/Zeo(+) plasmid (ThermoFisher Scientific) with an insert of DNA encoding amino acids M1-S147 of the Gal4 DNA-binding domain (DBD). A reporter construct pNL-9xUAS [16] carried nine repeats of the upstream activation sequence (UAS) before a minimal promoter in the pNL[NLucP/minP/Hygro] vector (Promega) encoding NanoLuc luciferase fused to the PEST destabilisation sequence (NLucP) (Promega). The pACT[VP16-Met] and pDBD[Gal4-Tai] constructs (each 27 ng DNA per well) were co-transfected along with 110 ng per well of the pNL-9xUAS reporter plasmid to semiconfluent CHO cells seeded in 48-well plates. At 24 h post-transfection, vehicle (EtOH) or JH I stereoisomers dissolved in EtOH were added for an additional 20 h. The luciferase activity was measured with the NanoGlo Luciferase reagent (Promega). EC 50 values were calculated using PRISM 6.0 GraphPad Software (San Diego, CA, USA) by nonlinear regression (least squares ordinary fit) with the 'sigmoidal dose-response (variable slope)' function. The data were obtained from three independent experiments and are presented as mean AE SD.
Immunoblotting CHO cells 24 h post-transfection were lysed in the Passive Lysis Buffer (Promega), briefly sonicated and separated by electrophoresis on 8% polyacrylamide-SDS gel under reducing conditions using LDS with 1.25% b-mercaptoethanol in the sample buffer. Proteins were transferred to a nitrocellulose membrane and processed in 5% nonfat milk and 5% BSA in TBST as a blocking and antibody incubation buffer. Expression of the WT and mutated Met proteins was detected with the monoclonal 9E10 c-Myc antibody (Ther-moFisher Scientific) diluted 1 : 3000, followed by incubation with HRP-conjugated secondary antibodies and capturing the chemiluminescent signal (ThermoFisher Scientific) using the LAS-3000 luminescent image analyzer (FujiFilm, Tokyo, Japan). The stability of Met expression in cells incubated with geldanamycin was assessed from immunoblots with additional normalisation to the levels of nascent Hsp90 detected with anti-Hsp90 antibody (#4874; Cell Signalling Technology, Danvers, MA, USA) diluted 1 : 1000.
We appreciate the generous gifts of the pNL-9xUAS reporter plasmid from David Sedlak, the pACT-Met vector from Hitoshi Miyakawa and the pure JH I stereoisomers from Karel Slama. This work was supported by EXPRO 20-05151X from the Czech Science Foundation to MJ.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.