Flexible laser-induced graphene biosensor enables real-time, in vivo profiling of wound healing cytokine dynamics

Abstract

Laser-induced graphene (LIG) provides a scalable route to high-performance carbon nanomaterials, but its biomedical translation has been hindered by limited robustness, biocompatibility, and in vivo validation. Here, we report a flexible, elastomer-integrated LIG biosensor that enables continuous monitoring, label-free monitoring of inflammatory cytokines within living tissue. By systematically optimizing the laser processing parameters, we generated stable, low-resistance graphene networks that maintained electrical fidelity under repeated bending and adhesion stress. Antibody functionalization conferred molecular specificity, allowing picogram-per-milliliter detection of interleukin-6 (IL-6), CXCL12, and TGF-β1 with performance comparable to ELISA. In a chronic wound model, the biosensor resolved cytokine-specific temporal dynamics across inflammatory, proliferative, and remodeling phases, validated by conventional assays. This platform provides minimally invasive and longitudinal profiling of inflammatory signaling, establishing LIG biosensors as clinically translatable tools for precision monitoring in wound healing, neuroinflammation, and other inflammation-driven disorders. Such direct and mechanically stable interfacing with the wound microenvironment enables continuous in vivo cytokine profiling beyond conventional wearable or adhesive biosensing approaches.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04676-9.

Introduction

Inflammatory cytokines play pivotal roles in orchestrating immune activation [1–4], tissue repair [5–7], and pathological progression, including chronic wounds [8–10], autoimmune disorders [11–13], and neurodegenerative diseases [14–19]. Continuous and real-time profiling of key cytokines, such as IL-6 [20, 21], TGF-β1 [22, 23], and CXCL12 [24, 25], is essential to resolve dynamic immune processes and inform targeted therapeutic strategies. However, current approaches including ELISA, immunohistochemistry, and flow cytometry require invasive sampling, involve laborious processing, and provide only static snapshots. These constraints preclude continuous monitoring and limit the ability to capture dynamic cytokine fluctuations within living tissues [26–29].

Efforts to overcome these challenges have increasingly focused on bioelectronic sensors, but existing platforms face persistent trade-offs among electrical performance, mechanical compliance, and molecular specificity [26–34]. Rigid electrodes impede seamless integration with soft tissues, while solution-processed nanomaterials often suffer from poor reproducibility and instability under physiological strain [27, 28, 30–34]. Graphene has attracted attention as a promising alternative, offering exceptional conductivity, a large surface-to-volume ratio, and inherent biocompatibility [35–46]. However, the conventional fabrication processes remain restrictive. Specifically, chemical vapor deposition requires high temperatures and complex transfer procedures [47, 48], whereas solution-processed graphene frequently exhibits structural discontinuity and variability [49, 50]. Thus, the development of scalable, biocompatible graphene-based biosensing strategies capable of longitudinal cytokine monitoring in vivo remains a major challenge.

Notably, laser-induced graphene (LIG) is generated by directly converting polymer substrates into porous, conductive graphene networks through a rapid, maskless, and solvent-free process [51–56]. This technique combines scalability with tunable electrical properties and structural porosity, making it inherently well-suited for biointerfacing [57–65]. Although conventional LIG systems often exhibit sheet resistance in the kΩ/sq range depending on precursor type and laser processing conditions, our optimized LIG achieved a sheet resistance below 300 Ω/sq, representing a substantial improvement while preserving the simplicity and scalability of direct laser carbonization. Although this value remains higher than that of CVD graphene, which can reach the Ω/sq regime, the present approach avoids the high-temperature growth and transfer steps required for conventional CVD-based fabrication [66, 67]. Despite extensive demonstrations in environmental and electrochemical sensing, however, the application of LIG to biologically integrated [56–58, 61, 62, 68, 69], cytokine-specific biosensors have not yet been realized. A key barrier lies in simultaneously achieving stable conductivity on flexible substrates, mechanical robustness under repeated deformation, and molecular specificity through reliable biofunctionalization.

Herein, we present a flexible LIG biosensing platform that addresses these challenges by uniting optimized laser processing, elastomeric integration, and antibody functionalization. The resulting devices maintain stable electrical performance under strain, achieve picogram-per-milliliter sensitivity for IL-6, CXCL12, and TGF-β1, and operate reliably in vivo. In a chronic wound mouse model, the biosensor resolved stage-specific cytokine dynamics across the inflammatory, proliferative, and remodeling phases, validated against ELISA and flow cytometry benchmarks. By integrating scalability, flexibility, and molecular specificity, this work establishes LIG biosensors as a clinically translatable platform for time-resolved in vivo sensing, cytokine monitoring and minimally invasive diagnostics.

Results

Flexible LIG biosensor enables minimally invasive cytokine monitoring in vivo

The flexible LIG biosensor enables minimally invasive and longitudinal monitoring of inflammatory cytokines during wound healing. Figure 1 schematically illustrates the complete workflow from device fabrication to in vivo application and cytokine analysis. As shown in the upper right panel, porous graphene electrodes were fabricated using the LIG technique and subsequently transferred onto a polydimethylsiloxane (PDMS) substrate, yielding a sensing interface that combined mechanical flexibility with biocompatibility. The modular design also allowed facile assembly, detachment, and reuse, while ensuring stable tissue contact and minimizing external interference during signal acquisition.

Fig. 1.

Schematic illustration of the flexible LIG-based biosensor for monitoring wound healing

When applied as a wound-healing monitoring module, the biosensor was tuned to capture cytokine activity corresponding to the wound size and healing stage. As illustrated in the lower right panel of Fig. 1, IL-6, CXCL12, and TGF-β1 emerged as dominant cytokines during the inflammatory, proliferative, and remodeling phases, respectively, and were selectively detected by functionalized LIG electrodes. Furthermore, biosensor measurements were cross-validated with ELISA and flow cytometry results, confirming that the platform reliably resolves cytokine-specific dynamics in continuous monitoring within living tissue.

LIG electrodes were fabricated by direct laser carbonization and transferred onto a PDMS substrate, forming a soft and biocompatible sensing interface. The modular device enabled stable tissue contact and multiplexed antibody functionalization for simultaneous detection of IL-6, CXCL12, and TGF-β1 during wound healing. Biosensor readouts were cross-validated with ELISA and flow cytometry results, confirming reliable in vivo cytokine profiling. Fig. created with BioRender. Chou, N. (2026) https://BioRender.com/l8yyb4u.

Optimization of the CO₂ laser processing enables high-conductivity and structurally ordered LIG networks

Optimized CO₂ laser processing yielded continuous and conductive graphene networks on flexible substrates while maintaining the sheet resistance after transfer. Additional optical image analysis under the optimized condition showed reproducible pattern formation, with 20 independently patterned LIG lines exhibiting a mean line width of 123.75 ± 2.73 μm and clear edge definition (Fig. S1). Figure 2a illustrates the laser-induced carbonization process that converts SU-8 polymer films into porous graphene structures, forming the conceptual basis for parameter optimization and subsequent analysis. Optical inspection of LIG transferred onto PDMS revealed that both the laser power and scan speed critically influenced the density and continuity of the graphene networks (Fig. 2b, S2). Correspondingly, sheet resistance mapping highlighted processing windows where conductive networks were achieved (Fig. 2c). Regions in gray corresponded to cases where no graphene was formed, whereas white regions indicated graphene visible by optics but electrically insulating. Conductive domains were confined to the colored areas, defining the optimal parameter ranges. Although the sheet resistance modestly increased after transfer compared with SU-8-based LIG (Fig. S3), values consistently remained < 300 Ω/sq, confirming stable conductivity on flexible substrates.

Fig. 2.

Morphological and chemical characterization of transferred LIG on PDMS. a Schematic of the laser-induced carbonization process generating porous graphene. b Optical images of LIG patterns transferred onto PDMS, fabricated under varying laser powers and scan speeds. c Sheet resistance heatmap of transferred LIG as a function of processing parameters. d, e XPS survey spectra and elemental compositions, showing increased carbon and reduced oxygen contents under optimized conditions. f–h High-resolution XPS spectra of the C 1 s, N 1 s, and O 1 s core levels, confirming graphitization and bonding configurations. i, j Raman spectra and quantitative analysis (I_D/I_G ratio, L_a, and 2D FWHM), demonstrating improved crystallinity under optimized processing. k SEM images of LIG on PDMS, showing the porous morphology and strong interfacial integration in top view

The chemical composition and structural order of the optimally processed LIG were further characterized. Processing windows highlighted in Fig. 2c were selected for detailed X-ray photoelectron spectroscopy (XPS) analysis, including the lowest scan speed (40%), which yielded the most continuous networks. Survey spectra confirmed that an increased carbon content and reduced oxygen functionalities were obtained under optimized conditions (Fig. 2d), and elemental ratios supported effective graphitization (Fig. 2e). High-resolution C 1 s spectra further confirmed the progressive graphitization of the laser-induced network, featuring a prominent asymmetric peak at ~284.4 eV assigned to sp²-hybridized graphitic carbon (Fig. 2f). Concurrently, the higher-binding-energy C–O component (~286.5 eV) was significantly attenuated, signifying successful deoxygenation. The N1s spectra displayed a distinctive graphitic N feature at ~401.1 eV, indicating substitutional nitrogen atoms covalently incorporated into the graphitic lattice (Fig. 2g), which is instrumental in enhancing the carrier density of the framework. Furthermore, the O1s spectra showed a concomitant reduction in oxygen-containing groups, particularly the carbonyl-related (C = O) peak at ~531.1 eV (Fig. 2h). Collectively, these spectroscopic signatures demonstrate the restoration of the conjugated π-network, facilitating the formation of a highly reduced and electrically conductive LIG platform. In addition, Raman spectroscopy corroborated these findings, showing distinct D, G, and 2D bands characteristic of optimized graphitic domains (Fig. 2i). Quantitative analysis demonstrated an I_D/I_G decrease from ~0.59 under suboptimal conditions to ~0.08 using optimized parameters, corresponding to an increase in in-plane crystallite size (L_a) from ~18 to ~230 nm (Fig. 2j). The relatively low I_D/I_G ratio observed under the optimized condition indicates reduced structural disorder and fewer defect sites, reflecting improved graphitic crystallinity and the formation of higher-quality graphene-like domains within the porous LIG network. Additionally, the I_2D/I_G ratio provides further insight into the graphitic stacking structure and effective layer number. The marked decrease in I_2D/I_G from the lower-power condition indicates an increase in multilayer graphitic character (Table. S1). However, under the optimized condition, the 2D band retains a relatively sharp and symmetric profile resembling a single-Lorentzian feature, rather than the broadened and split profile typically associated with highly ordered AB-stacked graphite. This interpretation is further supported by the extracted FWHM (full width at half maximum) values of the 2D peak (Fig. 2j), which remained broader than those of monolayer graphene while retaining a relatively symmetric profile. Together, these features indicate that the material is not true monolayer graphene, but rather turbostratic few-/multilayer graphene, where physically stacked graphene layers remain rotationally misaligned. This structural feature is commonly observed in LIG generated by rapid localized photothermal carbonization, where the carbon precursor is heated almost instantaneously and does not have sufficient time to form highly ordered AB-stacked graphite [70, 71]. Taken together, these Raman trends, together with the corresponding morphological observations, define the practical processing boundaries for LIG formation. Based on this analysis, the laser processing condition of 90% power and 40% speed was selected as the optimized parameter for subsequent device fabrication. In particular, this condition yielded a balanced combination of low defect density, developed graphitic domains, and a continuous porous network, which are essential for stable electrical conductivity and sensing performance.

To further validate the graphitic domain formation and successful transfer, Raman spectra were acquired from bare SU-8 and PDMS substrates, as well as SU-8/LIG and PDMS/LIG hybrids (Fig. S4). The appearance of the D, G, and 2D bands in SU-8/LIG and PDMS/LIG confirmed effective laser-induced carbonization and the retention of graphene features following transfer onto PDMS. Scanning electron microscopy (SEM) revealed porous, interconnected graphene morphologies tightly integrated with PDMS (Fig. 2k). Cross-sectional SEM and EDS mapping analysis revealed a distinct bilayer structure with an approximately 50 μm porous LIG layer beneath the PDMS overlayer and a total hybrid thickness of approximately 95–110 μm (Fig. S5, S6). While SEM clearly distinguished the PDMS and LIG layers structurally, the presence of Si signals within the porous LIG region in the EDS maps suggests partial PDMS infiltration into the LIG network. These findings indicate that laser-induced carbonization was primarily confined to the upper portion of the SU-8 precursor, while also promoting interfacial integration with PDMS.

Quantitative image analysis of binarized SEM images yielded an areal porosity of 4.45 ± 0.34%, further supporting the formation of a reproducible and spatially homogeneous porous morphology under the optimized condition (Fig. S7). The compositional differences were further confirmed by energy-dispersive X-ray spectroscopy (EDS), as shown in (Fig. S8, S9). LIG fabricated under optimal conditions (90% power, 40% speed) exhibited a markedly higher carbon content (~76 at%), whereas non-graphitized controls (30% power, 40% speed) showed substantially lower carbon contents with elevated oxygen and silicon signals.

Mechanical durability and interfacial stability of the LIG–PDMS hybrids

Mechanical durability and interfacial stability of the LIG–PDMS hybrids were separately assessed, as shown in (Fig. 3a–d), with schematic diagrams of each test configuration provided in (Fig. S10). The enhanced mechanical stability of the LIG–PDMS system can be attributed to the partial infiltration of uncured PDMS into the porous graphene network, forming an interpenetrating and mechanically interlocked interface that suppresses crack propagation and delamination under repeated deformation. Unlike transferred or solution-processed graphene films, which often suffer from weak substrate adhesion and structural discontinuity, the porous LIG–PDMS hybrid in this work forms a mechanically interlocked interface through partial PDMS infiltration, thereby improving interfacial adhesion and suppressing delamination during repeated deformation. Cyclic bending tests at angles ranging from 0° to 180° (Fig. 3a and b) showed that devices fabricated under optimized conditions and aided by surface lubrication maintained stable performance, with the normalized resistance changes (ΔR/R₀) remaining below ~20% after 1,000 inward/outward bending cycles, with SEM analysis confirming no obvious structural degradation or delamination at the LIG/PDMS interface (Fig. S11). In contrast, non-lubricated devices exhibited resistance increases exceeding 100% under the same conditions. Angle-dependent bending (Fig. 3c) further revealed that resistance changes remained < 10% for lubricated devices at 180°, whereas non-lubricated controls showed progressive increases surpassing ~30%. Although the wound site itself undergoes relatively limited mechanical deformation during monitoring, the interconnection region can still be vulnerable to localized strain and friction-induced stress. To mitigate this, a lubricant was applied at the device–skin interface, which reduced interfacial shear stress and helped preserve the mechanical and electrical stability of the sensor during use. The relatively large error bars observed under some bending conditions are likely attributable to the intrinsic structural heterogeneity of the porous LIG network and minor variations in interfacial contact during deformation. However, because the intended application of the sensor is on relatively flat wound surfaces rather than highly curved or mechanically dynamic regions, such resistance fluctuations are not expected to be substantial under practical use conditions. In addition, lubricant was applied to the interconnection region to minimize friction-induced stress and reduce unintended resistance variation. Adhesion was validated using tape-peeling tests (Fig. 3d); lubricated devices showed only ~12%–15% resistance increases after 100 cycles, whereas non-lubricated controls exhibited significantly greater losses. Collectively, these results confirm that optimized laser parameters yield conductive, crystalline, and mechanically resilient LIG networks with robust adhesion to elastomeric substrates, providing a strong foundation for flexible bioelectronic interfaces. This resistance increase is attributed to the formation of an insulating biomolecular layer upon antibody–antigen binding, which alters the local interfacial charge environment and partially perturbs conductive percolation pathways within the porous LIG network. Such chemiresistive behavior is consistent with previously reported graphene-based biosensors, where biomolecular adsorption modulates carrier transport and interfacial conductivity.

Fig. 3.

Mechanical durability and interfacial stability of LIG on PDMS. a ΔR/R₀ during repeated inward and outward bending with or without lubricant coating. b Enlarged view of the boxed region in (a), comparing inward versus outward bending with and without lubrication. c Resistance change as a function of bending angle with and without lubrication. Bending performance of the circular LIG sensor was evaluated under progressive bending conditions, reaching a minimum estimated 5 mm radius of curvature, demonstrating its mechanical adaptability to highly curved biological surfaces. d Resistance change over repeated tape-peeling cycles, confirming interfacial stability

Biofunctionalized LIG electrodes achieve molecularly specific cytokine detection

Building on the optimized LIG networks described above, antibody functionalization enabled label-free and highly sensitive cytokine detection with molecular specificity, validated by electrochemical, spectroscopic, and imaging analyses. Figure 4a illustrates the 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC)/N-hydroxy succinimide (NHS) coupling chemistry used to immobilize the capture antibodies onto the porous graphene surface. Electrochemical characterization confirmed stepwise I–V shifts during the sequential immobilization of bovine serum albumin (BSA), antibodies, and antigens (Fig. 4b), consistent with well-established electrochemical biosensing principles in which the formation of insulating biomolecular layers increases interfacial resistance and modulates charge transfer characteristics. Such progressive modulation of current response has been widely reported in electrochemical immunosensor literature as an indicator of successful surface functionalization and target binding [72–74]. Calibration curves further demonstrated IL-6 detection down to 1 pg/mL, showing that ΔR/R₀ linearly increased with concentration (Fig. 4c). The sensing sensitivity is inherently influenced by the effective thickness of the LIG layer, as the electrical response arises from modulation of conductive pathways within the porous network. Thicker LIG layers provide multiple parallel conduction channels, which can attenuate the relative resistance change upon surface binding events. Conversely, excessively thin layers may increase baseline resistance and noise. The optimized laser parameters used in this study yielded a balanced LIG thickness that preserves conductivity while maintaining high surface sensitivity for biomolecular interactions. Cross-sectional analysis revealed that the fabricated structures had a total thickness of approximately 95–110 μm, with an LIG layer thickness of ~50 μm (Fig. S5). This configuration enabled stable electrical conduction while preserving a porous surface morphology suitable for biomolecular interactions.

Fig. 4.

Molecular-level detection of biofunctionalized LIG under physiological conditions. a Schematic of EDC/NHS coupling for antibody immobilization on porous LIG. b I–V characteristics of BSA-coated, antibody-functionalized, and antibody–antigen (IL-6) bound surfaces. c ΔR/R₀ as a function of IL-6 concentration, demonstrating label-free sensitivity down to the pg/mL range. d Schematic of multiplex cytokine monitoring using antibody-functionalized LIG electrodes. e–g)FT-IR analysis comparing pristine and IL-6–bound LIG (e), with differential spectra f–g highlighting antigen–antibody vibrational signatures (amide I, amide II, and C–N/C–O bands). h Fluorescence microscopy confirming IL-6 binding on the LIG surface after antigen binding. i SEM images of in vivo IL-6 binding on LIG, revealing aggregated immunocomplexes within the porous graphene network. Scale bars: 300 μm (i1), 30 μm (i2), and 2 μm (i3).

Multiplexed cytokine monitoring was achieved by electrode-specific antibody functionalization. Figure 4 d schematically illustrates the simultaneous detection of IL-6, CXCL12, and TGF-β1 using spatially separated electrodes. Fourier-transform infrared (FT-IR) spectroscopy provided supporting evidence of cytokine binding. As shown in (Fig. 4e), pristine LIG (black solid line) and IL-6 (red solid line) exhibited distinct spectral differences across the 500–4000 cm⁻¹ range. In particular, modest changes in (Fig. 4f) were detected near amide I (~1650 cm⁻¹) and amide II (~1550 cm⁻¹) (Fig. 4 g), in addition to absorbance in the 1000–1200 cm⁻¹ range corresponding to C–N and C–O stretching (Fig. 4 h). These spectral signatures, although subtle, are consistent with established protein vibrational modes for antigen–antibody interactions [75, 76] and align with prior biosensor studies employing attenuated total reflection FT-IR to confirm immunorecognition [77–79].

Direct imaging further verified immunocomplex formation on porous LIG. As a control, fluorescence imaging performed prior to antigen exposure showed only uniform low-level background signals (Fig. S12). In contrast, post-incubation with IL-6 yielded distinct fluorescence signals colocalized with the porous LIG surface (Fig. 4 h), confirming that the observed fluorescence originates from specific antigen–antibody interactions rather than nonspecific background. SEM further revealed micro-to-nanoscale immunocomplexes embedded within the graphene network (Fig. 4i1–i3).

To further disentangle in vivo effects from in vitro effects, we characterized IL-6 binding under controlled conditions. EDS of pristine, unexposed LIG showed only carbon-dominated elemental compositions with minimal proteinaceous signatures (Fig. S13). Conversely, in vitro exposure of antibody-functionalized LIG to purified IL-6 produced distinct protein aggregates and additional biological elements, including Na, P, Cl, and K (Fig. S14), providing molecularly specific evidence of cytokine–antibody interactions. These findings, together with in vivo results, demonstrate that the observed immunocomplexes arise from targeted IL-6 binding rather than nonspecific adsorption. Ultimately, the biofunctionalized LIG electrodes achieve both sensitivity and specificity, enabling multiplex cytokine monitoring under physiologically relevant conditions.

Device packaging integrates LIG biosensors into implantable formats for in vivo applications

Using optimized and biofunctionalized LIG electrodes, we developed a modular packaging system suitable for in vivo operation. Figure 5a shows the fabrication and transfer of LIG onto elastomeric substrates, recalling the optimized carbonization process that allowed tunable electrode geometries and stable conductivity. The device consists of three circular components: a sensor, a body, and a detachable cover (Fig. S5). The sensor had an overall thickness of 95–110 μm, including a ~ 50 μm-thick LIG layer. The sensing region comprised three electrodes (IL-6 and CXCL-12: 400 × 350 μm; TGF-β: 200 μm radius) with an inter-electrode spacing of 350–500 μm. The sensor integrates the transferred LIG array onto a flexible substrate and provides electrical routing through patterned polyimide traces coated with conductive epoxy, ensuring stable electrical interfacing with the external connector. The body was designed with a diameter of approximately 1.5 cm and a thickness of 1.3 cm, serving as the main structural support for the device. The detachable cover, with a diameter of approximately 1.3 cm and a thickness of 0.3 cm, functions as a protective enclosure that mechanically secures the sensor while leaving the central sensing area exposed for direct tissue contact. A spring structure was incorporated into the center of the body to improve stable attachment and conformal contact of the sensor with the wound surface. The circular geometry was selected to conform to the wound site and to minimize mechanical stress concentration at the edges during movement. This layered configuration electrically isolates the contact pads from the biological interface while preserving multiplexed sensing capability. The detachable architecture further enables replacement of the LIG sensing layer without reconstructing the entire device, supporting modular reuse. The pre-optimized electrodes were assembled into circular modules using patterned polyimide traces coated with conductive epoxy to ensure reliable electrical interfacing (Fig. 5b and c). Figure 5c illustrates the design of the modular device body and detachable cover, which enclose the LIG array while maintaining exposed sensing sites for direct tissue contact. As shown in (Fig. 5d), multiplex cytokine monitoring was enabled through electrode-specific functionalization, permitting simultaneous detection of IL-6, CXCL12, and TGF-β1. The modular architecture further allowed adjustment of the electrode form and size to align with cytokine release dynamics. Practicality and robustness were validated through sequential demonstrations of fabrication, assembly, and cover replacement (Fig. 5e), underscoring the reusability of the system. Finally, Fig. 5f shows the in vivo application, where the packaged device was affixed to the dorsal skin of a mouse. The compact and flexible design minimized discomfort during freely moving behavior, enabling longitudinal cytokine monitoring under physiologically relevant conditions. Furthermore, the transferred LIG, patterned according to size or shape, can be installed on the bottom side of the module and easily replaced (Fig. 5e), allowing the module to be reused (Fig. 5g).

Fig. 5.

Modular device packaging enables integration of LIG biosensors for in vivo applications. a Fabrication of LIG patterns on SU-8/PDMS substrates by laser processing and transfer. b Assembly of patterned LIG electrodes into modular housings with optional conductive coating. c Design of the modular device body and detachable cover, with photographs of top and bottom views. d Schematic of multiplex cytokine monitoring enabled by electrode-specific functionalization, allowing simultaneous detection of IL-6, CXCL12, and TGF-β1 with tunable electrode geometries. e Photographs and schematics of the detachable connector design, showing the top view, detachment process, and bottom view with contact pads. f In vivo demonstration of the packaged system attached to a mouse for recording. g Schematic of the recycle–reuse workflow, highlighting device recycling for repeated experiments and replacement of single-use parts

In vivo profiling reveals cytokine-specific dynamics during wound healing

In a chronic wound model, the LIG biosensor enabled time-resolved in vivo sensing, longitudinal profiling of cytokine-specific dynamics across distinct healing stages. Figure 6a outlines the experimental design, showing that sequential wounds were created and then monitored over 8 days using a single multiplexed sensor containing three independently biofunctionalized sensing regions for IL-6, CXCL12, and TGF-β1. The sensing layout was designed by considering the gradual reduction in wound size during healing, allowing the active sensing regions to remain positioned over the wound area throughout the monitoring period. Representative images at days 1, 3, 5, and 8 (Fig. 6b) documented progressive wound closure. Throughout the 8-day monitoring period, the same implanted biosensor remained in place and enabled continuous signal acquisition without observable device detachment, abrupt baseline failure, or loss of channel-specific temporal resolution, supporting its operational stability under chronic in vivo attachment conditions. Consistent with this, short-term exposure to serum and whole blood resulted in a rapid resistance increase within the first 20 min, reaching approximately 7.0% and 12.0% dR/R, respectively, followed by a more gradual increase of ~2.2% and ~3.2% from 20 to 60 min, ultimately reaching ~9.2% and ~15.2% at 60 min, whereas PBS showed only a minimal overall change (~1.5%) (Fig. S15). These results suggest that the LIG interface can retain electrical functionality under biofouling-relevant conditions. Histological analysis of wound tissues using H&E staining showed no significant differences between sensor-treated and control groups. Both conditions exhibited comparable tissue architecture, without noticeable increase in inflammatory cell infiltration, tissue damage, or abnormal structural disruption. These results indicate that the presence of the sensor does not adversely affect wound healing or tissue integrity (Fig. S16).

Fig. 6.

In vivo profiling of cytokine dynamics during wound healing using a flexible LIG biosensor. a Experimental timeline in a chronic wound mouse model with sequential dorsal injuries and cytokine measurements through day 8. b Representative wound images at days 1, 3, 5, and 8. Magnified images are shown below each panel. Scale bar: 1.5 cm. c ELISA quantification of IL-6, CXCL12, and TGF-β1, showing stage-specific expression. d, e LIG module wound healing monitoring (LMWHM) (ΔR/R₀), consistent with ELISA while revealing subtle temporal differences. Data represent mean ± SEM (n = 5 independent animals).

Cytokine dynamics quantified by ELISA, as indicated in Fig. 6c and d, revealed stage-specific profiles consistent with canonical wound healing phases. IL-6 peaked early during the inflammatory phase, reaching ~327 pg/mL on day 1 before decreasing to ~179 pg/mL by day 3. CXCL12 rose during the proliferative stage, peaking at ~106 pg/mL on day 5, whereas TGF-β1 progressively increased during the remodeling stage, reaching ~537 pg/mL by day 8.

Figure 6e shows that biosensor readouts exhibited highly similar temporal patterns. The in vivo measurements were performed across n = 5 independent animals, and consistent temporal cytokine response patterns were observed, demonstrating good signal reproducibility of the LIG biosensor under physiological conditions. The normalized resistance changes ((R − R₀)/R₀, %) showed distinct temporal patterns across cytokines, with IL-6 peaking at Day 1 (111.66 ± 12.84%), CXCL12 peaking at Day 5 (203.65 ± 14.72%), and TGF-β1 reaching its highest level at Day 8 (188.98 ± 14.93%). ΔR/R₀ results mirrored ELISA trends, with IL-6 showing an early peak (~110% on day 1), CXCL12 increasing during the mid-phase (~200% on day 5), and TGF-β1 rising later (~180% by day 8). The distinct temporal responses observed across the three independently functionalized sensing regions also provide indirect evidence of minimal cross-channel interference, as substantial cross-reactivity would be expected to produce more uniform signal changes across all channels regardless of healing stage. While minor discrepancies were observed, likely reflecting the difference between continuous electrical monitoring and discrete biochemical sampling, the overall concordance between the biosensor and ELISA validates the ability of the LIG platform to resolve cytokine-specific dynamics in vivo with both high sensitivity and temporal fidelity.

Notably, these results were further supported by independent flow cytometry analysis (Fig. 7), which confirmed the stage-specific cytokine expression patterns in immune cell populations. The agreement across the ELISA, biosensor, and flow cytometry results establishes a robust.multimodal validation of cytokine-specific dynamics during wound healing.

Fig. 7.

Cytokines from immune cells present in wounded skin quantified by flow cytometry. a Gating strategy for intracellular cytokine quantification. Skin wounds collected 8 days post-injury were digested to a single-cell suspension, and the cells were labeled with antibodies prior to analysis. Initial gating was performed on single, viable cells, followed by the exclusion of debris and doublets. CD45 + immune cells were selected, and intracellular cytokine expression was quantified within this population. b Quantitative comparison of cytokine levels across time points, illustrating the relative contribution of each cytokine during the wound healing process c–e Representative flow cytometry plots showing cytokine expression in distinct wound phases (IL-6: inflammatory phase (c), CXCL12: proliferative phase (d), TGF-β1: remodeling phase (e)

Flow cytometry independently validates cytokine patterns detected by the LIG biosensor

Flow cytometry confirmed that the temporal cytokine dynamics captured by the LIG biosensor accurately reflected in vivo immune responses. For independent validation, wound tissues were harvested at sequential time points, dissociated, and analyzed for intracellular cytokine expression (Fig. 7a). Representative dot plots revealed dynamic changes in cytokine-positive immune cell populations (Fig. 7b). Quantitative analysis demonstrated that IL-6 peaked during the early inflammatory phase (days 1–5; maximum 9.64%), CXCL12 rose during the proliferative phase (days 3–5; median 4.70%), and TGF-β1 increased during the remodeling phase (days 5–8; maximum 18%), each exhibiting distinct temporal expression patterns (Fig. 7c–e). Specifically, IL-6 levels rose on day 1, reached a maximum at day 5, and then gradually declined. This trajectory differed from that of the biosensor and ELISA measurements, likely because flow cytometry detects intracellularly retained cytokines within local immune cells, rather than secreted cytokines in the extracellular milieu, thereby resulting in delayed stabilization of the measured levels (Fig. 7c). Similar trends were observed for CXCL12 and TGF-β1 (Fig. 7d and f) [80, 81]. Given that the trajectories measured by the LIG sensor closely aligned with those obtained by ELISA, the biosensor readouts can be regarded as accurate representations of target cytokine concentrations, further highlighting the suitability of the LIG platform for time-resolved in vivo monitoring of immune responses.

Discussion

This study establishes flexible LIG biosensors as a robust platform for minimally invasive, label-free cytokine monitoring in vivo. Despite recent progress in flexible graphene biosensors, simultaneously achieving mechanical durability, stable electrical performance under physiological deformation, and molecular specificity in complex tissue environments remains challenging. By optimizing the laser processing parameters, we engineered porous graphene networks that combine electrical conductivity with mechanical resilience, maintaining stable performance under repeated strain and adhesion stress. Antibody functionalization conferred molecular specificity, enabling multiplex detection of IL-6, CXCL12, and TGF-β1 at picogram-per-milliliter levels. Control experiments with single-cytokine exposure confirmed that non-corresponding antibody-functionalized electrodes exhibited negligible resistance changes within baseline variation levels, indicating minimal cross-reactivity among channels. A modular packaging strategy further ensured stable implantation and direct tissue interfacing, marking a practical step toward translational bioelectronic applications Although a dedicated long-term baseline drift or anti-biofouling assay was not separately performed in this study, the biosensor maintained stable and interpretable cytokine-specific temporal response patterns throughout the in vivo monitoring period without abrupt signal collapse or apparent loss of sensing functionality. Further systematic evaluation of chronic biofouling effects and long-term interfacial stability will be important in future studies. This porous graphitic structure arises from CO₂ laser (10.6 μm)–induced photothermal pyrolysis, where efficient absorption in SU-8 enables controlled carbonization. Unlike shorter-wavelength systems that rely on photochemical ablation, this process produces a uniform and interconnected porous graphene network, which contributes to both electrical conductivity and mechanical robustness of the LIG biosensor.

Although the optimized laser parameters identified here are specific to the present SU-8/CO₂ laser system, the underlying optimization principle is broadly applicable across LIG platforms. For other precursor materials or laser sources, the optimal process window is expected to shift depending on material properties and laser–matter interaction characteristics [82, 83].

The SU-8 and PDMS thicknesses were selected by considering the balance between laser-induced carbonization depth, transfer reliability, and mechanical handling. A ~ 100 μm SU-8 precursor provided sufficient thickness to support stable LIG formation without excessive unconverted bulk, whereas a ~ 80 μm PDMS layer offered an effective balance between conformability and structural robustness during transfer and use.

Unlike ELISA or flow cytometry, which require invasive sampling and provide only static snapshots, the LIG biosensor directly captures cytokine-specific temporal dynamics within intact tissue. Unlike most wearable or adhesive biosensors designed for epidermal or surface-level measurements, the present platform directly interfaces with the wound microenvironment and combines tissue-level sensing capability with mechanically stable device integration under physiological conditions. In a chronic wound model, IL-6, CXCL12, and TGF-β1 exhibited stage-specific trajectories consistent with canonical immune processes, providing biosensor readouts that closely aligned with ELISA benchmarks. These findings highlight the translational potential of LIG biosensors for dynamic inflammation monitoring.

Beyond wound healing, time-resolved cytokine profiling may transform disease management in neuroinflammation, autoimmune disorders, and cancer immunotherapy, in which immune signaling critically dictates outcomes. The modular and scalable design supports adaptation across diverse anatomical sites, from subcutaneous tissue to neural interfaces. Future directions include integration with wireless telemetry and closed-loop therapeutic systems, as well as expansion to broader biomarker panels. Collectively, these advances position LIG biosensors as a versatile foundation for minimally invasive, precision diagnostics, with clear potential for clinical translation in monitoring inflammatory diseases and guiding personalized therapeutic interventions.

Experimental section

Fabrication of LIG

LIG was fabricated on oxidized silicon wafers coated with SU-8 films. To enhance SU-8 adhesion, the wafers were pretreated with hexamethyldisilazane by spin coating at 500 rpm for 30 s. SU-8 2075 (MicroChem) was then spin-coated to a thickness of ~100 μm, followed by soft baking at 65 °C for 10 min and 95 °C for 20 min. Next, the films were exposed to UV light at 240 mJ/cm² and post-exposure baked at 65 °C for 5 min and 95 °C for 15 min to complete the crosslinking reaction.

LIG was patterned on the SU-8 surface using a CO₂ laser system (Beamo, FLUX Inc.; 10.6 μm wavelength, 30 W max power). A CO₂ laser (10.6 μm wavelength) was selected due to its strong absorption in SU-8, enabling efficient photothermal carbonization through vibrational excitation of molecular bonds without requiring additional absorbers. The laser power and scanning speed were both varied from 10% to 100% to identify optimal carbonization conditions without damaging the underlying substrate. The resulting graphene exhibited a porous, interconnected morphology suitable for transfer and sensing applications.

For transfer, PDMS (Sylgard 184, Dow Corning; 10:1 base-to-curing-agent ratio) was spin-coated over the LIG-patterned surface to a thickness of ~60 μm. The PDMS was cured at 90 °C for 90 min and carefully peeled off, lifting the LIG pattern with 90% yield. (Fig. S17). The transferred LIG retained its structural integrity and conductivity, and was used for subsequent characterization and device integration.

The cytokine sensing region was designed as three concentric and spatially separated zones to reflect the dynamic spatial progression of wound healing. The outermost, intermediate, and central circular regions were assigned for IL-6, CXCL12, and TGF-β sensing, respectively. The IL-6 and CXCL12 sensing regions had identical lateral dimensions of 400 × 350 μm, while the TGF-β sensing region was patterned as a circular electrode with a radius of 200 μm. The CXCL12 and IL-6 sensing regions were positioned radially at approximately 350 μm and 500 μm from the central sensing region, respectively, with 300 μm spacing between adjacent sensing units and 120 μm-wide interconnection lines.

The effective sensing areas of the individual LIG electrodes were quantified from optical micrographs using ImageJ. The three sensing electrodes exhibited active areas of approximately 140,000 μm², 140,000 μm², and 125,664 μm², respectively, corresponding to an average sensing area of approximately 1.35 × 10⁵ µm². These dimensions were used to define the active electrochemical sensing regions in the multiplexed device layout (Fig. S5). These dimensions were used to define the active electrochemical sensing regions in the multiplexed device layout.

Evaluation of the electrical, chemical, and mechanical properties

The multifunctional properties of the LIG were systematically evaluated after transfer onto PDMS substrates. Sheet resistance was measured using a four-point probe configuration and a 2450 SourceMeter^® (Keysight Technologies). Three measurements were taken at different locations on each sample, and the average was reported as the representative sheet resistance.

Chemical composition and structural characteristics were analyzed by XPS and Raman spectroscopy. XPS was used to determine the elemental composition and bonding states associated with laser-induced graphitization. Raman spectra were acquired to evaluate the D, G, and 2D bands, confirming the presence of sp²-hybridized graphitic domains and assessing the overall material quality.

Mechanical robustness was assessed by performing cyclic bending, angle-dependent bending, and tape-peeling tests on the PDMS-supported LIG. Prior to testing, lubricant oil was applied to the LIG surface to form a thin coating, minimizing friction and preventing surface damage during deformation. In all tests, changes in the electrical resistance (ΔR) were recorded to evaluate the effect of mechanical stress. Bending cycle tests were used to assess resistance stability under repeated flexing and bending angle tests quantified resistance variation at different curvatures. Peel tests were conducted to evaluate interfacial adhesion and durability under shear force. Across all conditions, the LIG maintained structural integrity and stable electrical performance, demonstrating its suitability for flexible and wearable sensing applications.

Analytical characterization techniques

The LIG samples were characterized using a combination of spectroscopic, microscopic, and cytometric techniques to evaluate the chemical composition, structural features, molecular interactions, and biological specificity of the system.

XPS was performed using an ESCALAB 250Xi system (Thermo Fisher Scientific) equipped with a monochromatic Al Kα source (hν = 1486.6 eV). Survey and high-resolution spectra were collected to determine the surface elemental composition and bonding states. Survey spectra were acquired at an energy step of 1.0 eV to identify the elemental composition, and high-resolution scans were performed at 0.1 eV for detailed bonding analysis.

Raman spectroscopy was conducted using a LabRAM HR800 system (Horiba) with a 532 nm excitation laser. Spectra were acquired over the range of 500–4000 cm⁻¹ to analyze the D, G, and 2D bands. The I_D/I_G ratio was used to assess the degree of graphitization and defect density in the carbon lattice.

SEM was performed using a Hitachi FB-2100 system to observe the porosity of the LIG and the morphology of surface-bound IL-6. Prior to imaging, the samples were sputter-coated with a thin layer of platinum to minimize charging effects. EDS was simultaneously conducted to analyze the elemental composition of the LIG surface and confirm the presence of cytokine-associated elements after biomolecule binding.

FT-IR spectroscopy was performed over the range of 500–4000 cm⁻¹ using a Perkin Elmer instrument to detect surface functional groups and chemical bonding changes before and after biomolecule immobilization.

Total internal reflection fluorescence (TIRF) microscopy was employed to visualize IL-6 binding to the LIG surface in real time. A Nikon Ti-E microscope equipped with a TIRF module and a 100× oil-immersion objective was used to capture high-resolution fluorescence images of labeled cytokine interactions at the sensor interface.

Flow cytometry analysis was carried out using a FACSymphony S6 (BD Science). Single-cell suspensions were prepared from tissue samples and stained with fluorophore-conjugated antibodies against IL-6 and CXCL12. Fluorescence signals were analyzed to confirm cytokine identity and quantify expression levels during wound healing.

Implementation of a modular packaging system

A basal module was fabricated to enable connection with the modified LIG patterns. A polyimide film, patterned at the contact regions, was placed onto the module and uniformly coated with conductive material. Prior to coating, electrical connections were established inside the module by interfacing the conductive layer with an Omnetics PCB connector. To couple the coated surface with the transferred LIG, the module was gently pressed against the LIG-patterned substrate and then detached, effectively transferring the LIG to the module. The resulting module was subsequently applied to the wound site for cytokine measurement.

Biofunctionalization of the LIG surface

The surface of the LIG transferred onto PDMS was biofunctionalized to enable specific detection of target cytokines. Prior to functionalization, the LIG surface was rinsed with deionized water and dried under air. Carboxyl groups on the surface were activated by treating the samples with an aqueous solution of 0.2 M EDC and 0.05 M NHS for 30 min at room temperature. Monoclonal antibodies against IL-6, TGF-β1, and CXCL12 were then introduced at concentrations ranging from 1 pg/mL to 100 ng/mL and incubated for 1 h. The immobilization of antibodies was achieved through carbodiimide coupling, where EDC/NHS activation of carboxyl groups on the LIG surface forms reactive NHS esters that subsequently react with primary amine groups of the antibodies, resulting in stable covalent amide bond formation. All three antibodies (IL-6, CXCL12, and TGF-β1) were immobilized using the same coupling protocol to ensure consistent surface chemistry and comparable sensing behavior. Following immobilization, the sensor surface was blocked with 1% BSA for 1 h to minimize nonspecific adsorption during subsequent measurements.

Biomolecule detection

The biofunctionalized LIG sensors were used to detect inflammatory cytokines, namely IL-6, TGF-β1, and CXCL12, in vitro. Cytokine standards were prepared in phosphate-buffered saline (PBS; pH 7.4) at concentrations ranging from 1 pg/mL to 100 ng/mL. Each solution was drop-cast onto the sensor surface and incubated for 2 h and 30 min at room temperature to allow for specific antibody–antigen interactions. After incubation, the surface was gently rinsed with PBS to remove unbound molecules and dried under ambient conditions.

Electrical measurements were performed using a 2450 SourceMeter by sweeping the voltage from −3 to +3 V in 0.01 V increments while recording the corresponding current. The sensor resistance was calculated from the slope of the linear region of the resulting I–V curve. The relative change in resistance (ΔR/R₀) was used as the sensing signal, where R₀ represents the baseline resistance obtained from blank LIG sensors (measured from three independent sensors), and R is the average resistance measured at each cytokine concentration. This ratio quantifies the relative increase in resistance upon analyte binding; for example, a ΔR/R₀ value of 100% indicates that the sensor resistance has doubled compared with the baseline.

Excisional skin wound healing mouse model

Mouse experiments were approved and performed according to the Institutional Animal Care and Use Committee at the Korea Brain Research Institute (KBRI, IACUC-24-00018-M4). Herein, 8- to 10-week-old wild-type male mice (C57BL/6J) were purchased from Orient Bio (Korea) and housed in individually ventilated cages with a 12-h light/dark cycle and ad libitum access to food and water. Skin wound excision procedures were conducted following previously reported protocols [84]. Briefly, mice were anesthetized in a chamber with 2% (vol/vol) isoflurane, and the dorsal skin hair was trimmed using an electric shaver and further removed with depilatory cream. The intended wound shape was lightly marked using a 5 mm biopsy punch and subsequently incised with scissors to generate the precise wound. At 1, 3, 5, and 8 days post-injury, each mouse was used for in vivo validation of the LIG-based sensing platform. Wound tissues and blood samples were collected and subjected to ELISA and flow cytometry analysis.

In vivo validation

To evaluate the sensing performance of the LIG platform under physiological conditions, in vivo experiments were performed using 8-week-old C57BL/6 mice. A chronic wound model was established by creating three circular skin wounds (approximately 2 mm in diameter) on the dorsal surface of each mouse. Biofunctionalized LIG sensors, each functionalized with monoclonal antibodies against IL-6, TGF-β1, or CXCL12, were gently affixed to the wound sites and maintained in contact for 2 h to allow cytokine binding. Temporal cytokine expression profiles were monitored by conducting measurements on sequential days after injury. IL-6 levels gradually increased and peaked on day 2 post-injury, reflecting sustained inflammatory activity in the early wound healing phase. In contrast, TGF-β1 and CXCL12 signals began to emerge from day 3 onward, consistent with their roles in the proliferative and remodeling stages of tissue repair.

Electrical measurements were conducted ex vivo immediately after sensor removal. A voltage sweep from −3 to +3 V in 0.01 V increments was applied using a 2450 SourceMeter, and the corresponding current was recorded to derive the I–V characteristics. Sensor resistance was calculated from the linear region of the I–V curve, and ΔR/R₀ was used to quantify cytokine presence.

ELISA

Wound tissues dissected from mice were homogenized in 300 µL of RIPA buffer using a tissue homogenizer, followed by sonication (30 s, 10 cycles). For serum preparation, 1 mL of blood was obtained via cardiac puncture and supplemented with 10 µL of 0.5 M ethylenediaminetetraacetic acid (EDTA). After centrifugation, supernatants from both tissue homogenates and blood samples were collected for subsequent analysis. Mouse IL-6 (DY406, R&D Systems), mouse TGF-β1 (DY1679, R&D Systems), and mouse CXCL12 (DY460, R&D Systems) DuoSet ELISA kits were used according to the manufacturer’s instructions. Briefly, 96-well plates were coated overnight at room temperature with the capture antibody diluted in PBS. After washing and blocking, standards and diluted samples were added and incubated for 2 h at room temperature. Plates were then incubated with the biotinylated detection antibody, followed by streptavidin-HRP. The reaction was developed using TMB substrate and stopped with 2 N H₂SO₄, and absorbance was measured at 450 nm with wavelength correction at 540–570 nm.

Intracellular cytokines analysis

Wound tissues (approximately 1.5 cm ×1.5 cm) were excised from mice and immediately placed in serum-free RPMI1640 (LM011-01, Welgene) on ice. Subcutaneous tissue was carefully removed, and each sample was minced using sterile scissors. The minced tissue was digested in whole skin digestion solution containing 1 mg/mL collagenase D (11088858001, Roche) and 0.4 U/mL DNase I (M0303L, New England Biolabs) at 37 °C for 30 min with gentle agitation. Enzymatic activity was neutralized by the addition of ice-cold complete RPMI1640 supplemented with 10% fetal bovine serum (FBS; 16000044, Gibco), 1% penicillin/streptomycin (15140122, Gibco), and 55 µM 2-mercaptoethanol (21985023, Gibco), followed by mechanical dissociation using a 12 mL syringe. The suspension was filtered through 100 and 40 μm strainers and centrifuged. Cell pellets were resuspended in complete RPMI1640.

For stimulation, a cell stimulation cocktail (00–4975, Invitrogen) and protein transport inhibitor (00–4980-93, Invitrogen) were added to each well, and cells were incubated for 4 h at 37 °C in 5% CO₂.

For flow cytometry, cells were stained using the LIVE/DEAD Fixable Violet Dead Cell Stain Kit (L34955, Invitrogen) for 20 min on ice, washed with 2% FACS buffer (2% FBS, 1mM EDTA in PBS), and blocked with anti-mouse CD16/32 antibody (101320, Biolegend) for 5 min on ice. Cells were then fixed with IC fixation buffer (00–8222, Invitrogen) and permeabilized with 1× permeabilization buffer (00–8333, Invitrogen). Intracellular cytokines were stained for 30 min at room temperature with the following antibodies: PE anti-mouse IL-6 (12–7061-81, Invitrogen), PerCP/Cy5.5 anti-mouse LAP (TGF-β1) (141410, Biolegend), and APC anti-mouse CXCL12 (IC350A, R&D Systems). In addition, FITC anti-mouse CD48 (Biolegend, 103108) was used as a leukocyte surface marker. Finally, cells were washed and resuspended in 2% FACS buffer prior to acquisition.

Author contributions

S. L., J. E., AH. L, N. C. conceived of the main idea. S. L., J. E., AH. L designed experiments. S. L., J. E., AH. L, J. L., H. S. conducted experiments. S. L., J. E., AH. L, J. W. K, N. C. analyzed the data. All the authors discussed the results and commented on the manuscript. S. L., J. E., AH. L, and N. C. wrote the manuscript. J. W. K. and N. C. supervised the project.

Funding

This work has been supported by grants of the National Research Foundation (NRF-2021-NR062074, RS-2024-02243041, RS-2025-02243041) and Korea Brain Research Institute research program (grant no.: 26-BR-02-02, 26-BR-04-01, 26-BR-07-01), funded by Ministry of Science and ICT, Republic of Korea. In addition, EVG620NT, H-TIRF unit w/Ti-E, and FACsymphony S6 were supported by Brain Research Core Facilities in KBRI.

Data availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Information.

Declarations

Ethics approval and consent to participate

All animal-related procedures were conducted at the same facility and adhered strictly to institutional guidelines for the ethical treatment of laboratory animals, as approved by the Institutional Animal Care and Use Committee (IACUC) of Korea Brain Research Institute (Approval No. IACUC-24-00018-M4).

Competing interests

The authors declare no competing interests.