Four-dimensional reconfigurable vascular tunneling machine of spatially programmed liquid crystal elastomers Authors: Yi Li, Yuan Gao, Yifei Li, Naiyu Guan, Zizheng Wang, Yongyu Lu, Weiqiang Hong, Chuchao Wang, Zeshi Yang, Ran Yang, Jingran Wang, Tao Li, Shize Sun, Lei Zhang, Jining Sun, Xu Guo, Zhaoqian Xie, Yuchen Cai, Dan Ge, Mengxi Wu, Junshan Liu Published: 2026 ABSTRACT AbstractThrombosis in tortuous vessels causes blockages and pathological conditions. Current thrombolytics are inefficient, and mechanical thrombectomies lack adaptability, hindering thrombus clearance in complex geometries. Here, we propose a 4D reconfigurable vascular tunneling machine (VTM) capable of synchronous rotation and elongation, mimicking teredo locomotion. The VTM operates via the reversible nematic-to-isotropic phase transition of liquid crystal elastomers, triggered by the reorientation of liquid crystal mesogens during a coupled contraction-torsion motion. Functionalized with magnetic liquid metals, the VTM achieves remote reconfiguration within 20 s via induction heating, with a temperature rise of only 1.1 °C and uniform heat dissipation. This rapid and safe thermal response facilitates the adaptive locomotion of the VTM within a cerebral artery model. Comprehensive biocompatibility evaluation confirms that the VTM exhibits no cytotoxicity and good hemocompatibility, and the VTM achieves the disruption of thrombus simulants across a stiffness range of 1 kPa to 1 MPa and sample retrieval up to 7.13 mm³. These demonstrations establish the VTM as a promising proof-of-concept platform, highlighting the potential for future development toward thrombus interventions and for operation as an adaptive soft robot in constrained environments.Subject terms: Actuators, Polymers, Liquid crystals Introduction Thrombus is highly prone to accumulate in arterial and venous tortuosity1–3, causing blockages and tissue lesions that pose severe threats to human health4–6. The commonly used methods for thrombus intervention are thrombolytics and mechanical thrombectomies. Thrombolytics are often constrained by narrow therapeutic windows, safety concerns, and limited efficacy against organized clots7,8. Mechanical thrombectomies employ rigid components that lack the adaptability to navigate complex vascular geometries, increasing the risk of vascular injury during procedures9,10. These inherent limitations hinder effective thrombus debulking in tortuous vasculatures, underscoring the need for devices that combine high adaptability with sufficient efficacy for thrombus therapeutic intervention. Recently, soft robots constructed from stimulus-responsive materials like shape memory polymers11,12, magnetic composites2,13, hydrogels14,15, and liquid crystal elastomers16–18 have demonstrated exceptional adaptivity and shape morphing capabilities, offer promising avenues for addressing the need of next-generation thrombus therapeutic devices. For example, Zhao et al.2,19 proposed a magnetic continuum robot capable of navigation and disrupting thrombi in arterial models. However, it is susceptible to instability and buckling when encountering obstacles, resulting in inefficient force transmission. Therefore, soft robots in constrained geometries face an inherent challenge: they must remain sufficiently compliant for safe, adaptive locomotion in confined spaces, yet simultaneously generate adequate mechanical force for targets such as effective thrombus interventions within tortuous vasculatures. Nature provides an elegant solution to this dilemma. The teredo is a mollusk that tunnels within submerged wooden structures20,21. It drills into wood by rotating a hardened anterior shell while synchronously elongating body muscles to tunnel in multiple directions. This tunneling process essentially constitutes a precise coupling of spatial rotation and axial elongation, enabling the teredo to adaptively navigate and generate mechanical output regardless of curved or highly confined spaces. Inspired by the tunneling mechanism of the teredo, we propose a four-dimensional (4D) reconfigurable vascular tunneling machine (VTM) capable of synchronous rotation and elongation. The VTM is constructed from a representative stimulus-responsive material: liquid crystal elastomers (LCEs). LCEs exhibit large, reversible shape morphing, governed by the programmed alignment of liquid crystal (LC) mesogens. Therefore, we developed a facile and versatile 4D mechanical alignment technique for LCEs, enabling synchronous elongation and rotation through spatially programmed coupled contraction-torsion orientation of LC mesogens. Enabled by this alignment technique, the VTM operates via the reversible nematic-to-isotropic phase transition of LCEs, which is triggered by the reorientation of liquid crystal mesogens during coupled contraction-torsion motion. Heating and cooling across the phase transition temperature (Tni) enable the VTM to reversibly undergo synchronous elongation and rotation. To demonstrate the versatility of this approach, we created a family of 4D reconfigurable VTMs and investigated structural design through both experiments and finite element analysis (FEA) simulations. Furthermore, we developed a composite consisting of oxidized liquid metals and magnetic particles. It remains melting‑free under induction heating, showing no delamination in the bloodstream at cerebral arterial flow rate22. Coated with this composite, the VTMs can be remotely actuated via induction heating. Comprehensive biocompatibility evaluation confirms that the VTM exhibits no cytotoxicity, hemolysis rates below 1%, no interference with coagulation pathways, and minimal platelet adhesion (<1 × 105 cells/cm2) with inactivated platelet morphology. Thermal safety validation in a physiologically relevant flow model demonstrates that blood temperature increase during actuation is only 1.1 °C, with gradual and spatially uniform heat dissipation—well within tolerable limits23. Under actuation, the VTMs exhibit robust adaptive locomotion within a cerebral artery model and generate highly tunable mechanical outputs, exhibiting effective capabilities for disrupting and sampling thrombi. Furthermore, quantitative assessment of contact forces under constrained conditions reveals maximum contact stress of 19.4 kPa, within reported ranges for cerebral artery tissue tolerance24,25. These demonstrations establish the VTM as a proof-of-concept platform that integrates programmable 4D reconfiguration, remote actuation, and validated biosafety. It provides a foundational framework for LCE-based soft actuators intended for operation in constrained environments, and highlights the potential for future development toward endovascular interventions. Results Principle and mechanism Figure 1a schematically illustrates the atraumatic disruption and sampling of thrombi within tortuous cerebral arteries using a soft, highly adaptive, and mechanically robust reconfigurable vascular tunneling machine (VTM). The design principle of the VTM is derived from the structural composition, locomotion mechanism, and mechanical properties of the teredo. Fabricated from liquid crystal elastomers (LCEs), the VTM achieves expected mechanical outputs through synchronous rotation and elongation while ensuring no additional damage to arteries. This core feature is enabled by the programmable alignment of liquid crystal (LC) mesogens, which dictates the intended shape morphing of LCEs. Figure 1b depicts the orientation of LC mesogens after undergoing axial contraction and torsion from the polydomain, respectively. When these two types of mechanical stresses are applied to LCEs, the LC mesogens adopt an alignment that couples axial contraction with spatial torsion. Following the initial crosslinking step of LCEs through two-stage thiol-acrylate Michael addition-photopolymerization (TAMAP) reaction26,27, Fig. 1c shows axial contraction and torsional force are applied to LCEs to form the nematic domain, resulting in a coupled contraction-torsion alignment of LC mesogens. Ultraviolet (UV) light (wavelength: 365 nm) exposure triggers the second-stage polymerization in pre-programmed LCEs, permanently stabilizing the central-helical shape. Heating above Tni (45.68 °C, see Fig. S1) disrupts the mesogenic order, generating internal stresses that drive a programmed shape morphing: synchronous elongation (25.7%) and rotation (360°), see Movie S1. This nematic-to-isotropic transition is reversible upon cycling through Tni, enabling fully reversible 4D reconfiguration—defined in this work as the pre-programmed shape change between distinct 3D states over time (Movie S2). To enable remote actuation, a composite consisting of liquid metals (oxidized Ga-In) and ferromagnetic particles (NdFeB) was synthesized. This flexible composite exhibits high induction heating efficiency. Coating it onto the surface of the VTM achieves remote thermal actuation, as schematically shown in Fig. 1d. In Fig. 1e, we also designed two drill bits inspired by the hardened anterior shell of the teredo: i) a solid conical and ii) a hollow conical structure with anchoring spines. These features can empower the VTM to generate sufficient mechanical output during reconfiguration, enabling effective thrombus disruption and pathological sample retrieval. Design principle and reconfiguration mechanism of the vascular tunneling machine (VTM). a Schematic of the teredo-inspired VTM (light flesh color) performing atraumatic disruption and sampling of thrombi (dark red color) within tortuous cerebral arteries (red color); black arrows present rotation and elongation; the right blue device is the VTM. b Schematic of spatial orientation of liquid crystal (LC) mesogens after contraction and torsion alignments. c Programming and stabilization of the central-helical shape of the VTM, enabling reversible reconfiguration upon cycling nematic-to-isotropic transition through phase transition temperature (Tni); light flesh ribbons present liquid crystal elastomers; red arrows present elongation and rotation; blue arrows present contraction and counter rotation; max means VTM achieving maximum reconfiguration. d Remote actuation of the VTM via the induction heating of the magnetic liquid metals (LMs). e Two drill bit designs to empower the VTM (silver color) for disruption and sampling in tortuous vasculature (red color). Fabrication and evaluation Guided by this bio-inspired design principle and 4D reconfiguration mechanism, we fabricated a scalable family of VTMs with single helix and double helices. The schematic in Fig. 2a and the experimental setup in Fig. S2 illustrate the complete fabrication process of the VTM. First, an LCE film is uniaxially stretched along the x-axis. Due to the Poisson effect, the film contracts along the y-axis, and the contraction ratio (0% to 36%) of LCE film along the y-axis is controlled by adjusting the stretching magnitude. Second, after stretching, the LCE film is removed from the stretcher and cut into ribbons of various dimensions along the y-axis using a CO₂ laser, yielding LCE ribbon precursors. Finally, the two ends of the LCE ribbon precursor are fixed along its central axis. The fixed ends are then rotated around the central axis by a stepper motor. The motor provides a torque of 400 mN·m, which is sufficient to induce rotation for VTMs of all tested dimensions. This rotation is coupled with the contraction from the previous step, and the central helix is formed and locked under UV exposure. FEA simulations in Fig. 2b, c reveal distinct differences in stress distributions within LCEs, subjected to pure torsion versus coupled contraction-torsion. Under pure torsion, stress initially concentrates at the center and spreads outward as torsional strain increases. However, the stress concentration region progressively narrows and eventually localizes along the axis under coupled contraction-torsion, which is highly consistent with the experiment (Fig. S3). This localized stress concentration along the axis is also reflected in the microscopic orientation of the LC mesogens, and the polarized optical microscopy images in Fig. 2d indicate that LC mesogens spatially aligned in a coupled contraction-torsion pattern along the axis. Therefore, the VTM is critically dependent on the spatial alignment of LC mesogens and the resulting localized stress concentration along the axis, which collectively determines the helix and reconfiguration of the VTM. Previous studies28–31 have reported that the width, thickness, and length influence helical structures such as twisted crystals, amyloid fibrils, and nematic ribbons, leading to various shapes. Consequently, we systematically recorded how thickness (t), width (w), and length (L) govern the helical formation of the VTM. As shown in the color-coded design table in Fig. 2e, we obtained 56 data points corresponding to different thickness-to-width (t/w) ratios under the LCE ribbon length fixed at 20 mm. Each color-coded point represents the helical state of the LCE ribbon under a given t/w condition. For example, when the thickness is 1 mm, and the width is 4 mm (t/w = 0.25), the corresponding point is colored light blue, indicating that the LCE ribbon can form both a single helix and double helices along its central axis. However, when the width is reduced to 3 mm with the same thickness (1 mm), the LCE ribbon fails to form either a full single helix or a full double helices, see Fig. 2f. Furthermore, by adjusting either the thickness or width, the LCE ribbon can be constrained to form either a single helix or a double helices along its central axis. When the width is fixed at 5 mm, and the thickness increases from 1.25 mm to 1.5 mm, the LCE ribbon can no longer form a full single helix (Fig. 2g) and is constrained to form double helices. Conversely, when the thickness is fixed at 0.5 mm, and the width increases from 6 mm to 7 mm, the LCE ribbon can no longer form full double helices (Fig. 2h) and is constrained to form a single helix. To investigate the influence of length on the helix formation of the VTM, we take LCE ribbons with a width of 4 mm and a thickness of 1 mm as an example. When we increase the LCE ribbon length from 20 mm to 40 mm, both FEA simulations (Fig. S4a) and experimental results (Fig. S4b) reveal that the helix distance is increased. This leads to slight buckling in the unconstrained segment between the two helices during rotation in the 40-mm-long ribbon, as shown in Fig. S4c. Additionally, due to the anisotropic deformation feature of LCEs, longer ribbons are more prone to non-uniform deformation during rotation, as shown in Fig. S4c. Furthermore, when the length of the LCE ribbon is reduced to 10 mm, the LCE ribbon forms a coiling-like shape, failing to achieve a full helix (Fig. S4d). Therefore, we selected 20 mm as the standard length for the VTM. After the fabrication of the VTM, we subjected it to 30 thermal actuation cycles. Throughout these cycles, both the elongation and rotation angle exhibit long-term cyclic stability with negligible hysteresis, as shown in Fig. S5. In Fig. S6, we also plotted the dynamic curves of both the rotation angle and elongation over a single actuation cycle. During the heating process, the rotational motion slightly leads the elongation from the onset. In the initial stage of the cooling process, the contraction begins marginally earlier; thereafter, the counter-rotation consistently leads the contraction until both motions conclude simultaneously, completing the actuation cycle. Overall, these results reveal that the rotation and elongation remain synchronized throughout the entire cycle. Notably, the coupled deformation is not the simple linear combination of individual elongation and rotation. Experimental results in Fig. 2i indicate that the rotation period of the coupled-programmed LCEs under stimulus shows no significant difference from that of the pure-torsion LCEs. However, the axial elongation of the contraction-torsion programmed LCEs is slightly reduced compared to that of the pure-contraction LCEs (Fig. 2i). Figure 2j quantitatively shows the elongation decreased by approximately 10% for single-helix and 20% for double-helix VTMs relative to their pure contraction counterparts. The observed trade-off suggests a complex interplay within the nematic domain network, where the spatially programmed torsion component partially constrains the full realization of axial strain, and thereby VTM with three or more helices are not considered in this work. Vascular tunneling machine (VTM) fabrication and structural evaluation. Scale bars, 5 mm. a A schematic of the VTM fabrication process shows the central-helix forming during the alignment of LC mesogens from a polydomain to a nematic state, and experimental images display the programmed VTMs; light flesh films and ribbons present liquid crystal elastomers. b FEA simulations compare the stress concentrations in the VTM under pure rotation versus coupled contraction-torsion, and c schematically shows the stress concentration region (purple color) changes. d The polarized optical microscopy images show the spatial orientation of the LC mesogens before and after adopting coupled contraction-torsion (C-T) alignment. e The color-coded table guides the VTM design and fabrication; t is thickness, w is width; black arrows present the helix formation transitions. f–h The representative cases show helical formation of the VTM under different width and thickness combinations; black arrows present the helix formation transitions. i The rotation of coupled-programmed LCEs under actuation is equivalent to that of pure-torsion LCEs, but the axial elongation of coupled-programmed LCEs is compromised, j decreasing by 10% (single-helix) and 20% (double-helix) compared to pure-contraction LCEs; Ec is abbreviation of elongation of LCEs after pure-contraction alignment, Ec-t is abbreviation of elongation of LCEs after coupled alignment of contraction and torsion; max means VTM achieving maximum reconfiguration; red arrows present elongation and rotation; blue arrows present contraction and counter rotation; black arrows present the combination of contraction alignment and torsion alignment; black gray arrow presents failed helix formation transition. The similar results are repeated in three independent experiments (d). Data were presented as mean ± SD, with points representing the mean and error bars denoting the SD of three independent samples (j). Remote actuation and performance characterizations To enable the remote and contactless actuation of the VTM, we adopted a wireless induction heating strategy. A magnetic LM composite, formed by mixing oxidized Ga-In with NdFeB particles, exhibits high induction heating capability. Figure 3a, b compare the heating performance of composites containing 0 wt% to 40 wt% NdFeB particles. The composite with 20 wt% NdFeB exhibits the best induction heating effect, reaching 90.7 °C within 10 s under an induction power of 1100 W, and was therefore selected as the actuation coating for the VTM. To avoid overheating of the magnetic LMs during induction heating, we reduced the induction power so that its temperature can reach or slightly exceed the Tni of the LCE in about 15 s, and is maintained around Tni thereafter, thereby ensuring complete actuation of the VTM with an additional 5 s (see Movie S3). Figure 3c shows that the magnetic LMs exhibit stable induction heating performance with only a variation within 3.3% (1.6 °C) after 20 actuation cycles, and the magnetic LMs remained melting-free before, during, and after 1 min actuation, as shown in Fig. 3d. In addition, even after 3 h of perfusion with sterile anticoagulated porcine blood at cerebral arterial flow rate22, the magnetic LM coating shows no visible delamination at flow-junction regions subjected to both prominent parallel and perpendicular shear stresses (Fig. 3e). Furthermore, although the VTM operates near the Tni (45.68 °C) for around 5 s, we evaluated its thermal safety in vascular environment. As shown in Fig. 3f, the induction heated VTM (initially at 25 °C) reaches blood temperature (36.5 °C) within 10 s and attains Tni of LCE after 15 s. Following the VTM maintaining at Tni for an additional 5 s, the maximum temperature rises in the blood and vascular tissue were only 0.35565 °C and 1.3506 °C, respectively, which is within the tolerable temperature range of the human body23. To experimentally verify these simulation results, we constructed a flow model (Fig. S7) and monitored blood temperature during thermal actuation. As shown in Fig. 3g, the measured temperature increase after 20 s is 1.1 °C, closely matching the simulation predictions. Infrared imaging in Fig. S8 further confirms that the temperature elevation is spatially uniform throughout the blood without localized heating. These experimental results verify the simulations and assess that the VTM poses no thermal risk. Remote actuation and performance characterization of the vascular tunneling machine (VTM). a, b Induction heating of magnetic liquid metals (LMs) with different NdFeB particle concentrations, under induction heating of 1100 W. Scale bars, 2 mm. c Temperature variation of magnetic LMs with 20 wt% NdFeB particles for 20 heating-cooling cycles. d Melting-free demonstration of magnetic LMs (gray silver color) under 1 min induction heating; black arrows present actuation process. Scale bars, 3 mm. e Bloodstream (red color) durability test for magnetic LM coating; Vf presents bloodstream at cerebral arterial flow rate. Scale bars, 150 µm. f Finite element analysis (FEA) simulations of the heat dissipation of VTM in a vascular tube. g The thermocouple recorded the temperature change of blood flow over a 20 s actuation period. The similar results are repeated in three independent experiments (e). Data were presented as mean ± SD, with the plot representing the mean and shaded area denoting the SD of three independent samples (g). Biocompatibility evaluation To evaluate the biocompatibility of the VTM, we systematically conducted cytotoxicity tests, hemolytic activity tests, anticoagulant tests, and platelet adhesion tests. As shown in Fig. 4a, compared to the negative control group, the cell viability of all groups treated by extracts of the LCE ribbons with magnetic LMs presents no statistically significant difference. This demonstrates that the VTM is non-toxic. In Fig. 4b, the hemolysis rates of all LCE ribbons coated with magnetic LMs exhibit below 1%, which are significantly lower than the ASTM F756-08 standard requirement (<5%) for non-hemolytic biomaterials32,33, indicating good erythrocyte compatibility. In Fig. 4c, coagulation function tests reveal no obvious differences in activated partial thromboplastin time (APTT), prothrombin time (PT), or thrombin time (TT) between all sample groups and the control group, demonstrating that the VTM does not interfere with coagulation pathways. Furthermore, platelet adhesion comparison in Figs. 4d, S9 reveals that the LM coating containing 20 wt% NdFeB particles, selected for VTM actuation, shows minimal platelet adhesion (<1 × 105 cells/cm2) with platelets retaining a rounded, inactivated spherical state32,34–36. This value is lower than the platelet adhesion levels reported for materials with good biocompatibility32,37–39, indicating no thrombogenic risk. Collectively, these results demonstrate that the VTM meets essential biosafety requirements for potential intravascular applications. Biocompatibility evaluation of vascular tunneling machine (VTM). a Cytotoxicity evaluation. NC refers to negative control (normal medium), and PC refers to positive control (0.2% Sodium Dodecyl Sulfate). b The hemolysis rate. NC refers to negative control (normal saline), and PC refers to positive control (DI water). c Anticoagulant performance, including prothrombin time (PT), activated partial thromboplastin time (APTT), and thromboplastin time (TT), NC refers to negative control of normal platelet-poor plasma (PPP). d Platelet adhesion numbers on the LCE ribbons coated with magnetic LMs. Data were presented as mean ± SD, with bars representing the mean and error bars denoting the SD of three independent samples (a, b, d). Statistical analysis by one-way ANOVA with Dunnett multiple comparisons test (**** P < 0.0001 indicates statistical significance, ns indicates no statistical significance) (a, b). Disrupting and sampling To empower the VTM, a conical drill bit was designed as shown in Fig. 5a. The length of this drill bit is 5 mm, with a base diameter ranging from 3 to 10 mm to accommodate different VTM widths, thus providing the pressure necessary to disrupt thrombi across 1 kPa to 1 MPa. The fabrication process is illustrated in Fig. 5b: (i) applying coupled contraction-torsion programming to LCEs forms a 4D central-helical structure; (ii) a magnetic LM layer (20 wt% NdFeB) is uniformly coated onto the LCE surface to form an efficient electromagnetic induction heating circuit; (iii) a 3D-printed drill bit is firmly bonded to the helical structure to construct the impowered VTM. Figure 5c–i exhibits that the VTM maintains stable mechanical disruption upon engagement with a high-stiffness thrombus simulator (>1000 kPa), see Movie S4. In a separate demonstration (Fig. 5c–ii), the VTM also successfully traverses a softer thrombus simulator (∼50 kPa) while retaining the actuation capability. As shown in Fig. 5d40–47 and Table S1, the modulus of the object disrupted by VTM exceeds those of various thrombus types, demonstrating that VTM can disrupt different types of thrombi. Additionally, Fig. 5e displays the axial elongation and rotation of the VTM (20 mm × 4 mm × 1 mm) inside the tube under remote actuation, effectively disrupting blockages. This reveals that external constraints have negligible influence on the 4D reconfiguration of our VTMs across varying dimensions, see Fig. S10. To further assess the feasibility of 4D reconfiguration of our VTM in geometrically constrained spaces, tests were conducted inside complex tubing with multiple bending angles (see multi-curvature tubes in Fig. S11a). Figure 5f schematically illustrates the expected deformation of the VTM within tubing bent at 30°, 120°, and U-shaped turns. Experiments in Fig. 5g, Movie S5 confirm that the VTM can smoothly navigate through curved paths while maintaining continuous and stable elongation and rotation. Figure S11b also documents the successful performance of the VTM in other tubes across 4 additional bending angles. Notably, during reconfiguration within such constrained geometries, contact with the tube wall occurs and generates an interaction force, which requires careful evaluation to ensure it remains within a safe range and does not cause damage to vascular tissue. To quantitatively assess the contact force and its potential impact, we designed a simplified benchtop setup (Fig. S12a). It consists of two holders with central apertures slightly wider than the VTM to allow rotation. Holder #1 is fixed, while holder #2 translates at controlled speeds via a stepper motor, tuning the constraint on the elongation of VTM during actuation and thus inducing varying degrees of bending (Fig. S12b). A force recorder, positioned within 500 µm of the VTM, serves as a surrogate arterial wall, enabling measurement of the contact force under different bending conditions (Fig. S12c). The measured contact force varies with the degree of constraint imposed on the elongation of VTM. Under full constraint (both ends fixed), the peak force reaches 210 mN (Fig. 5h). When a partial constraint is applied by translating one holder, the force decreases significantly. For example, at a translation speed of 0.0625 mm/s (1/8 of the VTM’s elongation speed), the peak force drops to 58.2 mN with a contact duration under 10 s (Fig. 5i). At 0.125 mm/s (approximately 1/4 of the VTM’s elongation speed), the contact duration shortens to about 3 s. At higher translation speeds (≥0.1875 mm/s, approximately 3/8 of the VTM’s elongation speed), the VTM elongates freely without contact, without measurable force (Fig. 5j). In realistic vascular navigation, the reconfiguration of VTM is dynamic, making any wall contact transient and comparable to the partial constraint cases with low-force, short-duration. Under these conditions, the contact between the LCE ribbon and the force recorder primarily relies on the surface contact, resulting in a maximum contact stress of 19.4 kPa, which falls within the stress tolerance range of cerebral arteries24,25, indicating a low risk of mechanical injury to the arterial wall during actuation. These results affirm the exceptional spatial adaptability of the VTM and provide potential for efficient operation within tortuous vasculature. Disruption performance of the vascular tunneling machine (VTM). a A 3D printed conical drill bit and two types of thrombus-mimicking materials with elastic moduli of ~ 1 kPa and ~1 MPa, respectively. Scale bars, 5 mm. b Fabrication process of the empowered VTM, including i) LCEs programming, ii) applying magnetic LM coating, and iii) integration of the drill bit. Scale bars, 5 mm. c Stable disruption of (i) a hard thrombus simulator (>1000 kPa) and (ii) a soft thrombus simulator (50 kPa). Scale bars, 2 mm. d The scatter graph shows a comparison of the modulus of the object disrupted by VTM with that of various reported thrombi. e The VTM maintains synchronous rotation and elongation, and completes the disruption inside the constrained tube. Scale bars, 5 mm. f Schematic of the predicted configurations of the VTM when navigating tubing bent at 30°, 120°, and in a U-turn. g Experimental demonstration of the VTM successfully navigating a tube with multiple curvatures while maintaining continuous elongation and rotation. Scale bars, 5 mm. h The curve shows the contact force of the VTM under a fully constrained case in Fig. S12c. i The curves show the contact force of the VTM under partially constrained cases in Fig. S12c. j The curves show the contact force of the VTM under minimally constrained cases in Fig. S12c. Figure 6a, b schematically illustrate the disruption and retrieval of pathological samples by the VTM via reversible elongation and rotation, facilitated by a barbed threaded drill bit 5 mm in length with a variable width of 3–10 mm. We initially deployed the VTM for disruption and sampling without external constraints. The VTM successfully achieved rapid disruption and sampling of the thrombus simulator within a reversible reconfiguration cycle, as demonstrated in Fig. 6b, Movie S6. After three actuation cycles, the VTM (the base diameter of the drill bit = 5 mm) achieved a sampling volume range of 1.93–7.13 mm3 for artificial thrombi of varying moduli (1 kPa–1 MPa). In addition, we observed that the highest sample yield occurred during the first retrieval attempt, while the second and third attempts showed a significant reduction in retrieval volume, as comparatively illustrated in Figs. 6c, S13 and Movie S7. In Fig. 6d, the VTM is designed to adaptively elongate and rotate when encountering highly tortuous vasculature, and the VTM achieves precise and effective disruption and sampling within two representative arteries with tortuosity (curvature > 0.3 mm−1, see location 1 and location 2). To prevent injury from drill bit contact with the arterial wall, its diameter is designed to be slightly smaller than the width of the VTM. Upon completion of thrombus disruption and sampling in the arterial model (Fig. 6e, f, Movie S8), the VTM retracts and counter-rotates to the original configuration without visible LM residue on the arterial tube, which enables in-situ intervention and sample retrieval while precluding any ancillary tissue damage. The above capabilities present a promising approach to mitigate the risk of life-threatening intravascular blockages and prevent irreversible vascular lesions. Sampling performance of the vascular tunneling machine (VTM). a Schematic illustration of the sample retrieval of the VTM via reversible elongation and rotation; red color clot presents thrombus; gray color device presents VTM. b The VTM successfully completes rapid disruption and sampling of a thrombus simulator within a single reversible reconfiguration cycle. Scale bars, 5 mm. c Sampling volume of 3 different artificial thrombi across 3 actuation cycles. d The VTM adaptively elongates and rotates within highly tortuous arteries; the red region presents the cerebral artery; the gray device presents VTM. e, f The precise disruption and sampling at 2 representative positions in (d). Scale bars, 3 mm. Data were presented as mean ± SD, with points representing the mean and error bars denoting the SD of three independent samples (c). Discussion In summary, this study presents a 4D reconfigurable vascular tunneling machine (VTM) based on liquid crystal elastomers (LCEs), with a design inspired by the biomechanics of the teredo. The mechanism of spatially coupled contraction-torsion orientation in LC mesogens has been systematically investigated, thereby enabling the VTM to mimic teredo locomotion through synchronous elongation and rotation. A systematic structural design guide table was first established to identify viable dimensional parameters, ensuring high‑fidelity central‑helix formation and achieving a maximum elongation of 25.7% and a full 360° rotation. Remote actuation of the VTM is enabled by a magnetic LM coating containing 20 wt% NdFeB particles, which remains intact without visible delamination in the bloodstream. Thermal safety of the VTM was also evaluated, confirming that tissue temperature rise remains within physiologically tolerable limits, and biocompatibility tests confirm that the VTM exhibits no cytotoxicity and good hemocompatibility. During reconfiguration, the VTM successfully performed synchronous rotation and elongation in tubular geometries featuring bending angles from 0° to 180°. Functionally, the VTM can navigate tortuous paths with curvatures larger than 0.3 mm−1, disrupt thrombi across the entire spectrum of hardnesses (1 kPa to 1 MPa), and perform sampling with volumes up to 7.13 mm3. Collectively, these demonstrations establish the VTM as a proof-of-concept platform that integrates programmable 4D reconfiguration, remote actuation, and validated biosafety within a single soft robotic system. The work provides a foundational framework for LCE-based actuators intended for operation in constrained environments, from fundamental mechanisms to functional validation. Prospectively, this platform could be further developed through integration with sensing modalities, advancing toward potential endovascular applications. Methods Liquid crystal elastomers (LCEs) synthesis The synthesis of LCEs in this work was completed through a two-stage thiol-acrylate Michael addition-photopolymerization (TAMAP)26,27. The LC mesogens RM257 (1,4-Bis-[4-(3-acryloyloxypropyloxy) benzoyloxy]−2-methylbenzene) was obtained from Henan Daken Chemical Co., Ltd. and utilized as received without further purification. 2,2’-(ethylenedioxy) diethanethiol (EDDET), pentaerythritol tetrakis (3-mercaptopropionate) (PETMP), 2-hydroxy-2-methylpropiophenone (HHMP), dipropylamine (DPA), and toluene were sourced from Shanghai Macklin Biochemical Technology Co., Ltd. and used as received. The synthesis began with dissolving powdered RM257 in toluene (40 wt%) at 80 °C. After RM257 is fully mixed with toluene and becomes a transparent solution, EDDET and PETMP were introduced at a molar ratio of 15:85 to initiate the first-stage reaction. Then, HHMP (0.38 wt%), a photoinitiator, was incorporated to enable the subsequent photopolymerization step. A diluted solution of DPA in toluene (molar ratio 1:50) was prepared separately to serve as a catalyst and was then introduced into the mixture. The resulting solution was mixed thoroughly and promptly transferred into a square polytetrafluoroethylene (PTFE) mold to produce LCE films with controlled thicknesses. Following 12 h of curing at room temperature, the films were subjected to an oven at 80 °C for another 12 h to remove residual toluene completely. With this step, the first-stage thiol-acrylate Michael addition-photopolymerization was completed. Magnetic liquid metals synthesis Liquid metals (eutectic gallium-indium alloy, Ga 75.5%, In 24.5%) were purchased from Dongguan Qiaotai Metal Technology Co., Ltd. and used as received. Ferromagnetic particles (NdFeB, particle size: D50 = 5 μm) were obtained from Shanghai Mingjue Trading Co., Ltd. and used without additional treatment. The LMs were first mixed in air for 30 s to promote the formation of surface oxides, after which NdFeB particles were added and mixed for an additional 10 min. The LM oxides adhered to the surface of the NdFeB particles, facilitating their incorporation into the LMs and finally obtained the magnetic LMs. The final magnetic LMs consisted of EGaIn alloy, LM oxides, and dispersed NdFeB particles. Artificial thrombi preparation Two types of thrombus-mimicking materials with elastic moduli of approximately 1 kPa and 1 MPa were initially prepared. The softer 1 kPa thrombus simulator was fabricated using Sylgard 184 silicone elastomer mixed at an 80:1 mass ratio of base to curing agent. The harder thrombus simulator consisted of dried loess soil blocks. By blending these two materials at different mass ratios, thrombus simulators with desired elastic moduli were produced. Induction heating A handheld AC (50 kHz) induction coil with the max power of 1100 W was utilized. The coil was held approximately 2 cm away from the VTM coated with magnetic liquid metals (LMs) to enable remote actuation. Cerebral artery model preparation The cerebral artery model was fabricated using photopolymerization 3D printing. The material is CRYSTA (crystal stereolithography resin), which exhibits good optical transparency. The printed model replicates human cerebral arteries at a 1:1 scale. Cytotoxicity test To assess the cytotoxicity of the VTM (magnetic LM coating applied onto the LCEs), we uniformly coated 5 LCE ribbons (20 mm × 4 mm × 1 mm) with the magnetic LM composite containing different mass fractions of NdFeB particles. The cytotoxicity of their extracts was evaluated using mouse microvascular endothelial cells (bEnd.3), which were purchased from HySigen Biosciences Technology Co., Ltd. All cells were revived, passaged and cultured according to standard cell culture protocols for subsequent material toxicity assays. First, the 5 sample types were sterilized for 30 min and then incubated with culture medium at 37 °C for 72 h to obtain the corresponding extracts. Subsequently, bEnd.3 cells were co-cultured with either the negative control (normal medium), the positive control (0.2% Sodium Dodecyl Sulfate), and the 5 sample extracts for 24 h. Cell viability was assessed using the MTS assay. Blood samples All blood samples in this work were obtained from ICR female mice (6–8 weeks old) purchased from Changsheng Biotechnology Co., Ltd. Animals were housed at ambient temperature on a 12 h light/dark cycle with access to food and water ad libitum. Animals were grouped using a completely randomized design. Hemolytic activity test Two methods were employed. In the first method, fresh mouse whole blood was diluted (Vblood: Vsaline = 1: 1.25), and 0.3 mL of the diluted whole blood was added to 2.7 mL of saline solution (0.9% NaCl) in which LCE ribbons with magnetic LM coating (five different mass fractions of NdFeB) were soaked. DI water and saline solution served as positive and negative controls, respectively. After incubation in a water bath at 37 °C for 1 h, each group was centrifuged at 102.78 × g for 5 min, and the optical density of the supernatant was measured at 541 nm using a SpectraMax®i3x multi-mode microplate reader (Molecular Devices, USA). The second method differed from the first method only in that 0.3 mL of diluted whole blood and 2.7 mL of saline solution were directly used to soak the LCE ribbons with magnetic LM coating (five different mass fractions of NdFeB). The control groups and subsequent steps were identical to those in the first method. Finally, the hemolysis rate was calculated using the formula: where ODS, ODN, and ODP refer to the mean absorbance values of the LCE ribbons with magnetic LM coating, the negative control, and the positive control, respectively. In the labels of Fig. 4b, the suffix “E” in the sample group names denotes the sample measured by the first method, while the suffix “S” denotes the sample measured by the second method. Anticoagulant test Fresh mouse blood was collected with an anticoagulant vacuum tube (Vsodium citrate: Vblood = 1:9), and the collected fresh blood was centrifuged at 925.02 g·min−1 for 15 min to obtain platelet-poor plasma (PPP). The activated partial thromboplastin time (APTT), prothrombin time (PT), and thromboplastin time (TT) of LCE ribbons with magnetic LM coating were evaluated using an automatic coagulation analyzer (CP3000, Japan), respectively. Briefly, 2 mm × 2 mm of LCE ribbons were placed in 1 mL of PBS solution (pH = 7.4) and incubated in a 37 °C water bath for 24 h, respectively. Then the PBS solution was removed, 1 mL PPP was added, and incubation continued at 37 °C for 30 min before transferring the PPP to the sample tube for use. Normal platelet-poor plasma (PPP) is the control group. Platelet adhesion test Fresh mouse blood was collected and centrifuged at 411.12 g·min−1 for 5 min to obtain platelet-rich plasma (PRP). LCE ribbons with magnetic LM coating were soaked in PBS solution at 37 °C for 24 h, then LCE ribbons were removed and placed in 500 μL of PRP and incubated at 37 °C for 2 h. The PRP was removed, and LCE ribbons were rinsed three times with PBS solution before being fixed with 4 wt% glutaraldehyde in PBS for 24 h at 4 °C. After being washed with PBS solution, the LCE ribbons were dried and dehydrated through a series of gradient physiological alcohol/PBS solutions (25, 50, 75, 85, 95, and 100 V/V%), the adhesion of platelets on the surface was observed by scanning electron microscopy (SU3900, Japan)48. Bloodstream durability test First, the LCE ribbon (20 mm × 4 mm × 1 mm) uniformly coated with the magnetic LMs (20 wt% NdFeB particles) was placed inside a cerebral artery model tube. Tertile anticoagulated porcine blood (purchased from Zhengzhou Biotechnology Co., Ltd.) was continuously perfused through the tube at a flow velocity of 0.4 m/s to simulate physiological flow in cerebral arteries22. After 3 h, the flow circuit was flushed with deionized water at the same flow rate to remove residual blood. Heat dissipation First, the LCE ribbon (20 mm × 4 mm × 1 mm) uniformly coated with the magnetic LMs (20 wt% NdFeB particles) was placed inside a cerebral artery mode tube. A thermocouple was inserted into a cerebral artery model tube and sealed it. The tube was continuously perfused with thermally anticoagulated porcine blood maintained at 37 ± 0.1 °C by a constant-temperature heater, at a flow velocity of 0.4 m/s to simulate physiological flow conditions in cerebral arteries. Recorded the captured values of the thermocouple under 20 s induction heating. FEA simulation of contraction and rotation of LCE ribbon ABAQUS 2023 (Dassault Systèmes, France) software was employed to perform a simplified simulation of the macroscopic deformation behavior of the LCE ribbon. The simulation compares the stress concentration in the LCE ribbon under pure torsion versus coupled contraction-torsion conditions. In the analysis step, the geometric nonlinearity option is enabled to account for the potentially large deformations of the LCE ribbon. A small damping factor (10−6) was also adopted to facilitate convergence. The simulation for pure torsion was conducted as follows: a fixed constraint was applied to one end of the LCE ribbon, while a torsional displacement around its central axis was applied to the opposite end. The stress distribution of the LCE ribbon was then obtained by solving the model. For the coupled contraction-torsion simulation, the setup was as follows: a fixed constraint was applied to one end of the LCE ribbon. The opposite end was subjected to a corresponding displacement to simulate a 25% contraction strain, while simultaneously undergoing the same torsional displacement around its central axis. The resulting stress distribution of the LCE ribbon was subsequently solved. The deformation and stress distribution of the LCE ribbon corresponding to changes in its length (20 mm, 30 mm, and 40 mm) were then obtained by solving the model. Simulation of heat dissipation In COMSOL Multiphysics, this model simulates the non‑isothermal flow of blood in a cerebral artery passing through a central-helical heat-source structure. By coupling the Laminar Flow and Heat Transfer in Fluids interfaces, the spatiotemporal evolution of the temperature field and flow characteristics under transient heating are investigated. The geometry consists of a cylindrical vessel (diameter: 5 mm) containing a helical internal structure (20 mm × 4 mm × 1 mm) that represents the complex flow path for heat exchange. Blood is modeled as a non‑Newtonian fluid using the Carreau model, with a density of 1050 kg/m3, a specific heat capacity of 3617 J/(kg·K), a thermal conductivity of 0.50 W/(m·K), and Carreau parameters: zero‑shear viscosity of 0.056 Pa·s, infinite‑shear viscosity of 0.00345 Pa·s, relaxation time of 3.313 s, and power‑law index of 0.3568. The Non‑Isothermal Flow multiphysics coupling is applied to account for bidirectional interactions between the flow and temperature fields. Boundary conditions include an inlet flow velocity of 0.4 m/s, an outlet pressure condition, and a temperature profile defined by a piecewise function: linear heating from room temperature (293.15 K) to a final temperature Tf during 0–15 s, followed by a constant Tf from 15–20 s. The vessel outer wall is set as thermally insulated, and the blood inlet temperature is fixed at 36.5 °C. A transient study is performed from 0 to 20 s with a 1 s time step, and a parameter sweep over Tf is conducted, with a key value of 45.68 °C (Tni of LCE). In this simulation, we did not account for the process of the structure reconfiguring from a helical to a flat shape during the final 5 s constant Tf period. If this process had been considered, heat would be transferred to the blood more efficiently, thereby reducing heat accumulation and the local temperature rise. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Supplementary Movie 4 Supplementary Movie 5 Supplementary Movie 6 Supplementary Movie 7 Supplementary Movie 8 Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgments The authors thank Bingbing Sun and Xin Li from the School of Chemical Engineering, Dalian University of Technology, for their firm support and kind help in the biocompatibility evaluation of this work. The authors acknowledge Li Chen, Penghe Yin, Pan Sun, Xi Zhang, Qing Shan and Guihong Bian from Key Laboratory for Micro/Nano Technology and System of Liaoning Province for their firm support and kind help. Author contributions The research was conceptualized and designed by Y.L., J.S.L., and M.X.W., who also acquired funding, administered the project, and provided supervision. Y.L., Y.Y.L., Z.Z.W., and Y.G. developed the methodology. Investigation and data collection were performed by Y.L., Y.G., Y.F.L., N.Y.G., Z.S.Y., X.G., L.Z., J.N.S., J.R.W., T. L., Y.C.C., and D.G. Visualization was carried out by Y.L., Y.F.L., W.Q.H., C.C.W., R.Y., and S.Z.S. The original draft was written by Y.L., Y.G., and N.Y.G., with review and editing contributions from Y.L., J.S.L., M.X.W., Z.Q.X., and L.Z. Peer review Peer review information Nature Communications thanks Guangchao Wan and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This work is supported by the National Key Research and Development Program of China (grant no. 2022YFB3204600), the Rejuvenating Liaoning Talents Plan of Liaoning Province (grant no. XLYC2403098), and the Dalian Science & Technology Innovation Fund (grant no. 2024JJ12GX023). Data availability All data supporting the findings of this study are available within the article and its supplementary files. Any additional requests for information can be directed to and will be fulfilled by the corresponding authors. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Yi Li, Email: yili@dlut.edu.cn. Mengxi Wu, Email: mengxiwu@dlut.edu.cn. Junshan Liu, Email: liujs@dlut.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-73495-1. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Supplementary Movie 4 Supplementary Movie 5 Supplementary Movie 6 Supplementary Movie 7 Supplementary Movie 8 Reporting Summary Transparent Peer Review file Source Data Data Availability Statement All data supporting the findings of this study are available within the article and its supplementary files. Any additional requests for information can be directed to and will be fulfilled by the corresponding authors. Source data are provided with this paper.