## Introduction

An ideal nanocarrier for drug delivery would be the one that can selectively collapse to release preloaded drugs in response to endogenous reporters overexpressed in disease tissues1,2,3,4,5,6,7,8,9. Since adenosine triphosphate (ATP) is known to be present at high levels in cancer tissues10, ATP-responsive nanocarriers might be a promising candidate3,4,5,6,7. In 2013, using partially modified biomolecular machine chaperonin GroEL as a monomer, we succeeded in developing a one-dimensional supramolecular polymer that can be depolymerized by the action of ATP to release its cargo3. However, ATP is also present in normal cells at rather high concentrations (>1 mM)11, and thus disease-selective drug delivery using ATP as the endogenous reporter cannot always be ensured. In the present work, we developed a nanocarrier (CLNCGTP/GTP*; Fig. 1e) that selectively responds to guanosine triphosphate (GTP). GTP is an intracellular molecule involved in many essential biological processes12,13,14,15,16,17,18,19,20,21,22,23,24,25, such as cell division12, nucleotide synthesis13, and cell signaling14. In the cell division process, the tubulin heterodimer (THD), which constitutes microtubules (MTs), uses GTP as an energy source to induce its polymerization and depolymerization15,16,17,18,19,20. GTP is also used as a component for the self-replication of RNA viruses26,27,28,29 such as coronaviruses. Notably, GTP is abundant in certain diseased cells (1.5–4.5 mM)30 such as rapidly proliferating cancer cells31 and RNA virus-infected cells32, whereas the concentration of GTP, unlike that of ATP, is negligibly low in normal cells (<0.3 mM)33. Therefore, GTP-responsive nanocarriers have the great potential to efficiently cure cancer and RNA virus-induced diseases including coronavirus disease 2019 (COVID-19)29. Although GTP-responsive carriers have already been reported, those carriers also respond to ATP5. So far, nanocarriers capable of responding solely to GTP have never been reported.

The nanocapsule (NC) that selectively responds to GTP to release a preloaded drug consists of THD. As shown in Fig. 1a, THD is composed of α-tubulin (green) and β-tubulin (cream), both of which bind to GTP. Notably, GTP attached to the α-tubulin unit is neither hydrolysable into GDP nor replaceable with other nucleoside phosphates. In contrast, GTP attached to the β-tubulin unit is known to be hydrolysable to GDP, which can be replaced with, e.g., GTP*, a nonhydrolysable GTP analogue (guanylyl 5’-α,β-methylenediphosphonate), affording THDGTP* (for convenience, only variable nucleoside phosphates attached to the β-tubulin unit are shown as a subscript). Both THDGTP and THDGTP*, when heated at 37 °C, have been reported to self-assemble into microtubules MTGTP and MTGTP*, respectively (Fig. 1b)34,35. Although MTGTP depolymerizes into THDGDP synchronously with the hydrolysis of hybridized GTP to GDP, MTGTP* does not depolymerize into THDGDP* because of the nonhydrolysable nature of GTP*. Therefore, our original motivation was to tackle a challenge of modulating the stability of MTs against depolymerization by changing the THDGTP/THDGTP* molar ratio. However, we unexpectedly found that the coassembly of THDGTP and THDGTP* at a certain mixing molar ratio resulted in the formation of a leaf-like 2D nanosheet (NS) rather than MT (Fig. 1e). Because of the increasing importance of 2D objects36,37, this finding prompted us to functionalize NS using the molecular glue technology38,39,40, which we developed for noncovalently functionalizing biomolecules such as proteins, nucleic acids, and phospholipid membranes, and also inorganic materials. Molecular glues are designed to carry multiple guanidinium ion (Gu+) groups and strongly adhere to such biomolecules under physiological conditions by taking advantage of a multivalent salt-bridge interaction with their oxyanionic functionalities (Fig. 1d). For this purpose, we chose $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ (Fig. 1c) and incubated it with NS. To our surprise, NS was transformed into a spherical nanocapsule NC (Fig. 1e). Using its photochemically modified version (CLNCGTP/GTP*; Fig. 1e), we successfully encapsulated and delivered doxorubicin (DOX)41, an anticancer drug, into GTP-overexpressing cancer cells to cause cell death.

## Results

### Reconstitution of MTGTP into NCGTP/GTP*

Figure 1e illustrates the overall procedure for the synthesis of NCGTP/GTP* from microtubule MTGTP. As a typical example of the procedure depicted in the flow chart in Fig. 2a, a 1,4-piperazinediethanesulfonic acid (PIPES) buffer (pH 6.8) solution of MTGTP (5.8 mg ml–1, Fig. 2c) was cooled at 4 °C, whereupon MTGTP underwent complete depolymerization within 3 h to yield THDGDP quantitatively (Fig. 2d)42. As observed by dynamic light scattering (DLS), the characteristic polydisperse feature of one-dimensional (1D) MTGTP (Fig. 2b, gray) changed to a monodisperse feature with a reduced hydrodynamic diameter of 8 nm (Fig. 2b, blue). Then, THDGDP (0.3 mg ml–1) was immersed in a PIPES buffer solution of a mixture of GTP and GTP* (300 µM in total) with a GTP* content of 83 mol% at 37 °C for 30 min. Under the present conditions, THDGDP was converted via the exchange events of GDP → GTP and GDP → GTP* into a mixture of THDGTP and THDGTP*, which then spontaneously coassembled into NSGTP/GTP* (Fig. 2e). The small-angle X-ray scattering (SAXS) profile of NSGTP/GTP* showed that its scattering intensity was proportional to q–2 in a small q region, which is characteristic of two-dimensional (2D) structures (Supplementary Fig. 9). As determined by atomic force microscopy (AFM), the average thickness of leaf-like NSGTP/GTP* was 5 nm (Fig. 2f). Here, the content of GTP* in the mixture of GTP and GTP* employed for the assembly of THDGDP was critical for its successful transformation into NSGTP/GTP*. When the content of GTP* ranged from 85–100 mol%, THD preferentially assembled into MT rather than NS (Supplementary Fig. 10), whereas THD barely assembled when its GTP* content was in the range of 0–70 mol% (Supplementary Fig. 11). Namely, the optimum THDGTP/THDGTP* molar ratio for the coassembly into NS is narrow, roughly with a GTP* content of 70–85 mol%. By means of nuclear magnetic resonance (NMR) spectroscopy in DMSO, NSGTP/GTP* prepared at a GTP* content of 83 mol% was found to contain 65 mol% of THDGTP* (Supplementary Fig. 12). It is known that MTGTP and MTGTP*, prepared from THDGTP and THDGTP*, respectively, are formed by edge-closing of NSGTP and NSGTP* as transient precursors43. Note that the longer axis of THDGTP is shorter than that of THDGTP*44. We suppose that this mismatch possibly affords unfolded NSGTP/GTP* rather than folded MTGTP/GTP*. Indeed, when THDGTP* (GTP* content of 83 mol%) was coassembled with THDGTPγS (THD hybridized with guanosine 5’-O-(3-thiotriphosphate), GTPγS), whose length is likewise shorter than THDGTP*44, NSGTPγS/GTP* was formed (Supplementary Fig. 13), whereas the coassembly of THDGTP and THDGTPγS (GTPγS content of 83 mol%), whose longer axes are close in length to each other44, resulted in MTGTP/GTPγS (Supplementary Fig. 14). NSGTP/GTP* was transformed into NCGTP/GTP* when it was incubated with GlueCO– (100 µM) in PIPES buffer at 37 °C for 30 min (Fig. 1e). This anomalous transformation was accompanied by a large change in the hydrodynamic diameter from 65 nm (Fig. 2b, green) to 660 nm (Fig. 2b, orange) with a slight increase in the zeta potential from –42.2 to –39.0 mV. Transmission electron microscopy (TEM) showed that the newly formed object NCGTP/GTP* was a hollow sphere (Fig. 2g). When MTGTP* and THDGDP instead of NSGTP/GTP* were likewise treated with $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$, ill-defined agglomerates resulted (Supplementary Figs. 15 and 16).

The physical stability of NCGTP/GTP* is important for its utilization as a carrier for drug delivery. Through several different experiments, we noticed that NCGTP/GTP* immediately collapsed upon incubation with albumin or serum in buffer, indicating its insufficient stability as a drug carrier. Here, we would like to point out a great advantage of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ and its homologues that their multiple benzophenone (BP) groups upon photoexcitation enable covalent crosslinking with adhering proteins (Fig. 1d). Successful examples so far reported include microtubule and kinesin39, whose dynamic behaviors could be attenuated by the reaction with photoexcited molecular glues. In the present work, by using fluorescent FITC-appended GlueFITC (Fig. 1c, FITC; fluorescein isothiocyanate) derived from $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$, we first confirmed that $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ has a sufficient photoreactivity with the constituent (THD) of NCGTP/GTP*. As shown in Supplementary Fig. 17, the reaction mixture, after being exposed to UV light (300 nm) in PIPES buffer, showed the presence of a fluorescence-emissive covalent adduct between THDGDP and GlueFITC in sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Then, we investigated whether this photochemical approach can provide NCGTP/GTP* with a sufficient physical stability by crosslinking the shell. Thus, a PIPES buffer solution of NCGTP/GTP* was exposed to UV light for 2 min, where TEM (Fig. 2h) and AFM imaging results (Supplementary Fig. 18) and DLS profiles (Fig. 2b, red) showed that crosslinked (CL) CLNCGTP/GTP* was spherical and remained intact even upon incubation with albumin (0.1 mg ml–1) or serum (0.01%) (Supplementary Figs. 19 and 20). CLNCGTP/GTP*, when prepared using GlueFITC instead of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$, was fluorescent (Supplementary Fig. 21), indicating the presence of the molecular glue in CLNCGTP/GTP*.

### Computational simulation of the assembly of NSGTP/GTP*

Considering that tubulin nanosheets NSGTP/GTP* are, on average, 0.04 µm2 wide and 4.2 nm thick, the formation of NCGTP/GTP* (surface area; ~6.2 µm2, membrane thickness; 50 nm) requires at least 1000 pieces of NSGTP/GTP* to assemble. Note that $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ carrying both Gu+ and $${{{{\mathrm{CO}}}}_2-}$$ groups in its structure can self-assemble via their salt-bridge interaction. In the initial stage of the transformation of NSGTP/GTP* into NCGTP/GTP*, we postulate that a certain number of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ molecules utilize their Gu+ groups to form a salt-bridged network with the surface $${{{{\mathrm{CO}}}}_2-}$$ groups on NSGTP/GTP* (Fig. 1d) as well as the focal-core CO2 group in $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$. This adhesion event can lower the surface charge density of NSGTP/GTP* and enhance its hydrophobic stacking, which is secured by possible reorganization of the salt-bridged polymeric networks on NSGTP/GTP* (Fig. 1e). We performed all atom molecular dynamics (MD) simulations45 to explore the adhesion of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ and the effect of this event on the tubulin assembly. From a full MT model (PDB code: 3J6E), we obtained its partial structure composed of three laterally assembled THDGTP* units ([THDGTP*]3) as a model of NS (Fig. 3a). The MD simulation suggested that $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ adopts a globular conformation in aqueous media with a hydrodynamic diameter of 1.5 nm (Fig. 3b, Supplementary Fig. 22). When exposed to 30 equivalents of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ (Fig. 3c, d), [THDGTP*]3 enhances its hydrophobic nature (Fig. 3e, f) as a result of the surface charge neutralization by adhering $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$. In the solvent-accessible surface area of [THDGTP*]3, the hydrophobic dominancy increases from 48% to 57% (Fig. 3g). Notably, when $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ was allowed to adhere onto [THDGTP*]3, the molecular simulations suggested that [THDGTP*]3 adopts a slightly more flattened conformation, characterized by a distribution angle with an average value of ~156° (Fig. 3h, i, blue), compared with that of native [THDGTP*]3 (red). The simulations also showed that, even after the $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ adhesion, [THDGTP*]3 preserved a certain level of flexibility (Fig. 3i). We also calculated radial distribution functions g(r) between the charged groups of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ and the amino acid residues of [THDGTP*]3. Supposedly, the CO2 groups in aspartic acid and glutamic acid are interactive with the Gu+ groups in $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$, while the cationic groups in lysine and arginine are interactive with the focal $${{{{\mathrm{CO}}}}_2-}$$ group in $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$. As expected, the g(r) data revealed that the Gu+ groups in $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ are largely populated near the $${{{{\mathrm{CO}}}}_2-}$$ groups on the [THDGTP*]3 surface (Fig. 3j, blue), whereas they are scarcely populated around the polar but nonionic hydroxyl groups in serine, threonine, and tyrosine (Fig. 3j, gray). Meanwhile, the focal CO2 group in $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ is not populated around the cationic groups on the [THDGTP*]3 surface (Supplementary Fig. 23). The computational calculation also showed that multiple adhering $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ molecules can interact and self-assemble via a salt-bridge interaction between their Gu+ and CO2 groups (Fig. 3j, red), which results in forming a dense Gu+/CO2 salt-bridged polymeric network on the [THDGTP*]3 surface (Fig. 3k). This may promote the self-assembly of flexible NSGTP/GTP and stabilize them in the gently curved multilayered configuration of NCGTP/GTP*, as observed experimentally46,47. As a control experiment, the use of $${{{\mathrm{Glue}}}}^{{{{{\mathrm{CO}}}}_{2}-}{\mathrm {Me}}}$$ (Fig. 1c) having a focal ester group instead of its ionized form for the transformation of NSGTP/GTP* into NCGTP/GTP* resulted in an ill-defined agglomerate (Supplementary Fig. 24).

### GTP-responsiveness of CLNCGTP/GTP*

We investigated whether photochemically stabilized CLNCGTP/GTP* is responsive to GTP or not. Notably, the concentrations of both extracellular and intracellular GTP are lower than 0.3 mM in normal cells33. However, as already described in the introductory part, rapidly proliferating cancer cells and RNA virus-infected cells contain GTP in a concentration range of 1.5–4.5 mM30. Therefore, drug-loaded CLNCGTP/GTP*, when taken up into such GTP-rich environments, might selectively collapse to release its preloaded guest. Upon incubation for 100 min at 37 °C in PIPES buffer with 0.2 mM GTP, CLNCGTP/GTP* still maintained its spherical shape, as observed by TEM (Fig. 4a). However, when the GTP concentration was increased to 0.5 mM, CLNCGTP/GTP* gradually collapsed (Fig. 4b), displaying a polydisperse DLS profile in 100 min (Fig. 4c, green). This minimum concentration threshold is important for achieving the error-free delivery to GTP-enriched sites. We added Biomol GreenTM as a phosphoric acid (PO4) detector to a mixture of CLNCGTP/GTP* and GTP (1 mM), and successfully detected PO4 by means of electronic absorption spectroscopy, indicating that CLNCGTP/GTP* has a GTPase activity (Fig. 4d). Although THDGTP*, the constituent of CLNCGTP/GTP*, has no GTPase activity, the product upon incubation of THDGTP* with GTP for 1 h in PIPES buffer at 37 °C showed a GTPase activity comparable to that of THDGTP (Supplementary Fig. 25), indicating the conversion of THDGTP* into THDGTP. Thus, under the GTP-rich conditions described above, GTP* in CLNCGTP/GTP* is likely replaced with GTP to afford CLNCGTP, which possibly collapses along with the hydrolysis of GTP in a manner analogous to the depolymerization of MTGTP. Of particular importance, CLNCGTP/GTP* remained intact to the treatment with other triphosphates (0.5 mM), such as ATP, cytosine triphosphate CTP, uracil triphosphate UTP (Fig. 4e, Supplementary Figs. 26 and 27).

### Guest encapsulation into CLNCGTP/GTP*

How to stably encapsulate guests inside nanocarriers is one of the important subjects for drug delivery. By using gold nanoparticles (NPAu; 14 pM, diameter 50 nm) as a guest, we succeeded in obtaining NPAu-encapsulated CLNCGTP/GTP* by adding $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ (100 µM) to a PIPES buffer solution of a mixture of NSGTP/GTP* (13 µg ml–1) and NPAu at 37 °C. After 30-min incubation, the resulting mixture was exposed for 2 min to UV light (300 nm) for crosslinking. Using TEM (Fig. 4f) and asymmetric field flow fractionation analysis (Supplementary Fig. 28), we confirmed that CLNCGTP/GTP* encapsulated NPAu (CLNCGTP/GTP*NPAu) in its hollow sphere. We also confirmed that CLNCGTP/GTP*, when treated with GTP, indeed released its preloaded guest. For this purpose, we first prepared FITC-labeled THDGDP with a mixture of GTP and GTP* (GTP* content: 83 mol%) in PIPES buffer, and further incubated the resulting fluorescent NSGTP/GTP* with $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ in the presence of doxorubicin (DOX) for 30 min. Then, the mixture was exposed for 2 min to UV light (300 nm) for transforming NCGTP/GTP*DOX into CLNCGTP/GTP*DOX, which was confirmed by confocal laser scanning microscopy (CLSM) to carry both FITC and DOX dyes (Supplementary Fig. 29, green and red, respectively). When CLNCGTP/GTP*DOX was incubated with 1 mM GTP in PIPES buffer for 100 min, DOX, as observed by CLSM, became much less fluorescent, indicating the disruption of CLNCGTP/GTP* to release DOX (Fig. 4g, (i)–(iii), lower panel). Upon incubation for 20 min, 50 min, and 100 min, the residues obtained by ultrafiltration (cut-off molecular weight = 5000) of the reaction mixtures contained 73%, 53%, and 21% of the total amount of preloaded DOX, respectively (Fig. 4h), while in the absence of GTP, DOX was not released (Fig. 4g, (i)–(iii), upper panel).

### Intracellular drug delivery with CLNCGTP/GTP*

As a proof-of-concept study, we investigated whether FITC-labeled CLNCGTP/GTP* can be taken up by human hepatocellular carcinoma Hep3B cells (Fig. 5a). The cells were incubated in Eagle’s minimum essential medium (EMEM) containing CLNCGTP/GTP* (0.5 µg ml–1) for 2.5 h, rinsed with Dulbecco’s phosphate-buffered saline (D-PBS), and further incubated in EMEM containing 10% fetal bovine serum (FBS) for 1.5 h. CLSM (Fig. 5b (i), left panel) together with flow cytometry analysis (Fig. 5c) revealed that most of the cells took up FITC-labeled CLNCGTP/GTP*. Upon subsequent incubation for 21.5 h in EMEM (10% FBS), the entire cytoplasm eventually became fluorescent (Fig. 5b (ii), right panel) as a possible consequence of the collapse of incorporated CLNCGTP/GTP*. In sharp contrast, FITC-labeled THDGDP and NSGTP/GTP*, the intermediates for constructing CLNCGTP/GTP*, were scarcely taken up into Hep3B cells (Supplementary Figs. 30 and 31). The high intracellular uptake of FITC-labeled CLNCGTP/GTP* is possibly due to a salt-bridge interaction between the Gu+ groups in adhering $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ and cell-surface oxyanionic groups48. We confirmed that the intracellular uptake was little affected by the presence of endocytosis inhibitor NaN3 (ref. 49), suggesting that the incorporation of CLNCGTP/GTP* into Hep3B cells was caused via an endocytosis-independent direct pathway (Supplementary Fig. 32). For the drug delivery application of CLNCGTP/GTP*, we conducted a cell viability assay with CLNCGTP/GTP*DOX. When treated with CLNCGTP/GTP*DOX ([CLNCGTP/GTP*] = 2.6 µg ml–1, [DOX] = 2 µM) in EMEM (Fig. 5d) for 2.5 h, Hep3B cells took up DOX as observed by CLSM after a subsequent incubation in EMEM (10% FBS) for 1.5 h (Fig. 5e (iii), left panel), and then died within next 21.5 h to form an ill-defined agglomerate (Fig. 5e (iv), right panel). We also confirmed that Hep3B cells took up a larger amount of DOX in CLNCGTP/GTP* (Fig. 5f, red) than DOX alone (Fig. 5f, orange). Accordingly, CLNCGTP/GTP*DOX successfully lowered the cell viability to 30 ± 6% (Fig. 5g, red), whereas that caused by DOX alone was only 48 ± 15% (Fig. 5g, orange). As expected, the cell viability decreased as the concentration of CLNCGTP/GTP*DOX was increased (Supplementary Fig. 33), while the viability upon incubation with CLNCGTP/GTP*DOX did not substantially increase when the incubation time was shortened from 2.5 h to 1.0 h (Supplementary Fig. 34). This is likely caused by the GTP-selective collapse of CLNCGTP/GTP*. The intracellular delivery of CLNCGTP/GTP*DOX was also successful with other cell lines such as A549 cell and HeLa cell (Supplementary Fig. 35). We also confirmed that neither the coexistence of THDGDP nor THDGDP/$${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ enhanced the efficacy of DOX (Supplementary Fig. 36). Together with the noncytotoxic nature of CLNCGTP/GTP* (Fig. 5g, green) and its stability in a range of pH at tumor tissue (Fig. 2h)50, these results allow us to expect that CLNCGTP/GTP* may have the potential to deliver preloaded drugs into cancer cells using GTP as an endogenous reporter.

## Discussion

Here, we have documented the successful reconstitution of MTGTP into a GTP-responsive nanocarrier (Fig. 1). MTGTP is depolymerized into THDGDP, which is incubated with a mixture of GTP* and GTP (content of GTP*: 70–85 mol%), thereby facilitating the in situ coassembly of the resulting THDGTP* and THDGTP monomers to form NSGTP/GTP*. Subsequently, NSGTP/GTP* is treated with molecular glue $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ to be transformed into spherical NCGTP/GTP*, followed by UV exposure to afford crosslinked CLNCGTP/GTP* capable of stably encapsulating guests (Fig. 2). In GTP-rich environments, CLNCGTP/GTP* collapses and releases preloaded guests through the transformation of CLNCGTP/GTP* into CLNCGTP followed by the hydrolysis of its bound GTP into GDP, analogous to the depolymerization of MTGTP (Fig. 4). Using CLNCGTP/GTP*, we successfully delivered DOX into cancer cells that overexpress GTP, and caused cell death more efficiently than DOX alone (Fig. 5). Most importantly, CLNCGTP/GTP* is a drug carrier that can selectively collapse in response to GTP rather than ATP that is abundant in normal cells. Since cells infected with RNA viruses such as coronavirus produce a large amount of GTP in their self-replication process, GTP is an endogenous reporter for RNA virus-infected cells. In vivo utilization of CLNCGTP/GTP* for curing RNA virus-induced diseases such as COVID-19 is one of the interesting subjects worthy of further investigation.

## Methods

MD simulation was performed using AmberTools 20, GROMACS 2020.5 package, and Visual Molecular Dynamics (VMD) package, and MD simulation methodologies are described in the Supplementary Information.

### Reconstitution of MTGTP into CLNCGTP/GTP*

THDGTP was obtained by purification from porcine brain51 by two cycles of polymerization and depolymerization in PIPES buffer (100 mM PIPES, 2 mM MgSO4, 0.5 mM GTP, 4 µg ml–1 leupeptin, and 0.4 mM PefaBlock, pH 6.8). A solution of THDGTP (5.8 mg ml–1) in PIPES buffer (100 mM PIPES, 5 mM MgCl2, 2 mM MgSO4, 1.5 mM GTP, and 10% DMSO, pH 6.8) was incubated at 37 °C for 30 min to afford MTGTP. The reaction mixture was centrifuged at 17,900 × g for 20 min at 24 °C. The resulting precipitate was dissolved in PIPES buffer (100 mM PIPES, 100 µM MgCl2, and 20 µM GDP, pH 6.8) and incubated at 4 °C for 3 h to afford THDGDP43. Subsequently, THDGDP (0.3 mg ml–1) thus obtained was incubated in PIPES buffer (100 mM PIPES, 1 mM MgCl2, 250 µM GTP*, and 50 µM GTP, pH 6.8) at 4 °C for 60 min and then at 37 °C for 30 min to afford NSGTP/GTP*. NSGTP/GTP* (13 µg ml–1) was incubated in a solution of $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ (100 µM) in PIPES buffer (14 mM PIPES, 1 mM MgCl2, and 200 µM GTP*, pH 6.8) at 37 °C for 30 min. The reaction mixture was exposed to UV light at 300 nm for 2 min, affording CLNCGTP/GTP*. FITC-labeled NSGTP/GTP* and CLNCGTP/GTP* were prepared using FITC-labeled THDGDP (14% labeling rate)51 under conditions that were otherwise identical to those listed above. Prior to the NMR measurement of the NSGTP/GTP* sample, unbound GTP and GTP* were removed by centrifugation (286,000 × g) of the reaction mixture at 37 °C for 60 min. Zeta potentials of NSGTP/GTP* (1.3 µg ml–1) and NCGTP/GTP* (1.3 µg ml–1) were measured at 37 °C in PIPES buffer.

### GTP-responsiveness of CLNCGTP/GTP*

A solution of CLNCGTP/GTP* (12 µg ml–1) in PIPES buffer (9 mM PIPES, 0.9 mM MgCl2, and 180 µM GTP*, pH 6.8) was incubated in the presence of GTP (0.1 mM, 0.2 mM, 0.5 mM, and 1 mM), ATP (0.5 mM), CTP (0.5 mM), and UTP (0.5 mM) at 37 °C for 100 min. For the evaluation of the GTP hydrolysis activities of CLNCGTP/GTP* (12 µg ml–1) and THDGDP (12 µg ml–1), Biomol GreenTM reagent (100 µl) was added to the reaction mixtures, incubated for 30 min at room temperature and subjected to electronic absorption spectroscopy at 620 nm.

### Preparation of CLNCGTP/GTP*⊃NPAu

CLNCGTP/GTP*NPAu was prepared after the incubation of a mixture of NSGTP/GTP* (13 µg ml–1), $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ (100 µM), and gold nanoparticles (NPAu; 14 pM) in PIPES buffer (14 mM PIPES, 1 mM MgCl2, and 200 µM GTP*, pH 6.8) at 37 °C for 30 min, followed by UV irradiation at 300 nm for 2 min. For the asymmetric field flow fractionation analysis, a sample solution of CLNCGTP/GTP*NPAu in PIPES buffer was subjected to ultrafiltration (1500 × g) for 5 min using a regenerated cellulose membrane (cut-off MW = 5000) prior to analysis. PEG-coated NPAu was used to avoid nonspecific adhesion of THD52.

### GTP-triggered release of DOX from CLNCGTP/GTP*

CLNCGTP/GTP*DOX was prepared after the incubation of a mixture of NSGTP/GTP* (13 µg ml–1), $${{{\mathrm{Glue}}}}^{{{{\mathrm{CO}}}}_2-}$$ (100 µM), and DOX (10 µM) in PIPES buffer (14 mM PIPES, 1 mM MgCl2, and 400 µM GTP*, pH 6.8) at 37 °C for 30 min, followed by UV irradiation at 300 nm for 2 min. The reaction mixture was incubated with GTP (1 mM) at 37 °C for 100 min and then subjected to ultrafiltration (2400 × g) using a regenerated cellulose membrane (cut-off MW = 5000) for 10 min. The resulting residue was subjected to fluorescence spectroscopy (λext = 470 nm). A reference sample without GTP was likewise prepared.

### Intracellular delivery

Hep3B cells (3.0 × 103 cells/well) plated onto an 8-well chambered cover glass were incubated in EMEM containing 10% FBS at 37 °C with 5% CO2 for 24 h. The cell samples were rinsed twice with D-PBS prior to use. Typically, the cells were treated with FITC-labeled CLNCGTP/GTP* (0.5 µg ml–1) and incubated at 37 °C with 5% CO2 for 2.5 h. Then, the cells were rinsed twice with D-PBS and further incubated at 37 °C for 1.5 h (4-h incubation in total) or 21.5 h (24-h incubation in total) with 5% CO2 in EMEM containing 10% FBS. Analogous cell samples treated with FITC-labeled THDGDP (0.5 µg ml–1), FITC-labeled NSGTP/GTP* (0.5 µg ml–1), CLNCGTP/GTP* (0.5 µg ml–1) with NaN3 (5 mM)49, CLNCGTP/GTP*DOX ([CLNCGTP/GTP*] = 2.6 µg ml–1, [DOX] = 2 µM), CLNCGTP/GTP* (2.6 µg ml–1), and DOX (2 µM) were likewise prepared. For a cell viability assay using CLNCGTP/GTP*DOX, CLNCGTP/GTP*, and DOX, the cell samples were incubated with Cell Counting Kit-8 reagents (10 µl) for 30 min, and subjected to electronic absorption spectroscopy at 450 nm. Hep3B cell samples treated with Tween 20 (0.2%) were used as a positive control.

### Statistics and reproducibility

All experiments including the preparation of CLNCGTP/GTP*, the investigation of its GTP-responsive collapse, and the intracellular delivery using CLNCGTP/GTP* were performed at least three times to check the reproducibility.

### Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.