Ultrabright ratiometric Raman-guided epilepsy surgery by intraoperatively visualizing proinflammatory microglia

Summary

Resective surgery is an effective approach for long-term seizure control in drug-resistant focal epilepsy when the epileptic focus (EF) can be accurately delineated and removed. However, intraoperative mapping of EF with electrocorticography is laborious, time-consuming, and highly vulnerable to the effects of anesthesia. Here, we demonstrated that activated microglia can be reliable biomarkers for EF localization. Leveraging a newly developed ratiometric Raman nanosensor, ultraHOCls, we successfully visualize proinflammatory microglia in live epileptic mice, allowing for precise EF delineation without the interference of anesthesia. Compared to electrocorticography-guided surgery, ultraHOCl-guided surgery results in a substantial 61% reduction in total seizure burden in epileptic mouse models. Notably, ultraHOCls sprayed on freshly excised human brain tissues can effectively discriminate epileptic regions from non-epileptic tissues with high sensitivity (94.89%) and specificity (93.3%). This work provides an alternative strategy for delineating the EF intraoperatively, potentially revolutionizing surgery outcomes in epilepsy patients.

Graphical abstract

Highlights

• •The activated proinflammatory microglia provide robust biomarkers for EF localization
• •Developing an ultrabright ratiometric SERS nanosensor for imaging proinflammatory microglia
• •UltraHOCls effectively discriminate epileptic regions from non-epileptic tissues

Wang et al. develop a ratiometric SERS nanosensor, ultraHOCls, for detecting proinflammatory microglia in epileptic regions by responding to HOCl. Their ratiometric Raman signal variations enable precise delineation of the epileptic focus with high target-to-background ratios, making ultraHOCls a promising tool for intraoperative guidance in epilepsy surgery.

Introduction

Epilepsy affects more than 50 million people globally, and 30%–40% of patients fail to achieve seizure control with medication despite the fact that there are more than 20 antiseizure drugs available.^1,2 Drug-resistant epilepsy (DRE) is associated with numerous neuropsychological comorbidities, decreased quality of life, social discrimination, and an increased risk of injury and death.³ For patients with focal onset DRE, surgical resection of the epileptic focus (EF), the area of cerebral cortex indispensable for generating clinical seizures, offers the best chance at reducing seizure burden and achieving seizure freedom, when the EF can be accurately identified.^4,5 Even though the proportion of individuals with DRE who could potentially benefit from surgical intervention ranges from 10% to over 50%, less than 3% of patients with DRE are referred to surgical treatment annually due to the complexities of EF localization and high medical costs.^4,6 Furthermore, 30%–50% of DRE patients continue to experience recurrent seizures post-surgery largely due to incomplete resection of the EF.⁷

Various non-invasive preoperative techniques are employed to localize the EF, including semiology analysis, scalp telemetry electrophysiology, and magnetic resonance imaging (MRI), aided by additional technologies such as interictal positron emission tomography (PET) and ictal single-photon-emission-computed tomography (SPECT).⁵ If the EF cannot be accurately defined from non-invasive investigations, further investigation using stereoelectroencephalography (SEEG) may be required.⁸ However, the constrained spatial resolution of SEEG can still pose challenges, notably in accurately delineating the extent of the EF. Preoperative SEEG often identifies a region of interest rather than a precise boundary, thus necessitating additional intraoperative investigations.⁸ Intraoperatively, factors such as brain shift can lead to inconsistent alignment between preoperative and intraoperative localization information,⁹ further making complete EF resection difficult. Therefore, an intraoperative, real-time strategy that can accurately predict the EF and its boundary could revolutionize epilepsy surgery by greatly improving the efficacy of entire EF resection.

Currently, intraoperative electrocorticography (ECoG) is the sole technique used for intraoperative mapping of the EF by detecting epileptiform discharges.¹⁰ Although providing good temporal resolution, ECoG has the following limitations^11,12,13,14: (1) the epileptiform discharges recorded during surgery are highly vulnerable to anesthetic agents, which compromises the reliability of the collected EEG data; (2) prolonged sampling time is often required in order to obtain reliable result and may increase surgical risk, complications, and surgeon fatigue; (3) the interpretation of ECoG results can be subjective, leading to empirical surgical resection. In the past decades, optical intrinsic signal imaging,^15,16 voltage-sensitive dye imaging,¹⁷ and photoacoustic imaging have been developed for mapping the EF.¹⁸ However, these imaging strategies still rely on capturing the vulnerable seizure activity, making intraoperative application difficult. Recently, we reported a SPECT/surface-enhanced Raman scattering (SERS) dual-mode imaging probe for epilepsy imaging.¹⁹ However, the compromised epilepsy-to-normal tissue ratio caused by the heterogeneous probe distribution and the “always on” probe signal hamper its application in vivo.¹⁹ Therefore, there is an urgent need for alternative intraoperative approaches for locating the EF that are stable with reproducible result and has high specificity and sensitivity.

Microglia are resident immune cells in the brain and play physiological roles in maintaining brain homeostasis and mediating humoral and cellular immunity.^20,21 Prolonged or excessive microglial activation may decrease the seizure threshold and facilitate epileptogenesis by releasing proinflammatory molecules or disrupting neuronal circuits.^22,23 Chronic activation of microglia is a prominent feature of EF in both epilepsy patients and animal models.^22,24,25 Furthermore, myeloperoxidase (MPO), a key inflammatory enzyme actively involved in oxidative tissue damage and immune defense, is significantly upregulated in activated microglia in the brain.²⁶ MPO generates hypochlorous acid (HOCl)/hypochlorite ions (OCl⁻), which are particularly implicated in seizure induction and epileptogenesis.^27,28 Therefore, we hypothesize that imaging activated microglia by sensitizing the HOCl may provide a new strategy for locating the EF intraoperatively.

Here, we confirm that proinflammatory microglia accumulation and concomitant biosynthetic MPO can serve as robust biomarkers for EF localization in DRE patients. Next, we develop an ultrabright ratiometric SERS nanosensor, which shows picomole sensitivity and a nearly 100-fold increase in the Raman intensity ratio after HOCl treatment. With the assistance of this nanosensor, we can locate the EF intraoperatively and alleviate seizure activity effectively by visualizing activated proinflammatory microglia in rodent epilepsy models. Notably, this SERS nanosensor also shows effectiveness in locating the epileptic region in the freshly excised brain tissues of patients with epilepsy when applied topically. Compared to the ECoG-based strategy, the SERS-guided surgery strategy shows the following advantages: (1) defining the EF without the interference of anesthesia; (2) providing an extended and stable time window (>1 h) for EF visualization; and (3) locating the EF in a more objective and straightforward manner by transforming the spatial distribution of proinflammatory microglia into visible Raman signals; (4) can easily be incorporated with other adjuncts such as intraoperative neuro-monitoring, potentially allowing maximizing the EF resection while preserving functional area. Overall, this work proposes an alternative strategy for guiding epilepsy surgery and may potentially improve surgical outcomes in DRE patients.

Results

Proinflammatory microglia accumulate in the EF of DRE patients

To elucidate the existence of activated microglia in the seizure focus, noninvasive presurgical evaluations, including scalp EEG, noninvasive MRI, and PET imaging were collected, and a series of experiments were conducted on resected brain tissue samples (Figure 1A). As shown in Figure 1B, typical clinical manifestations and characteristic EEG recordings verified that DRE patients experienced uncontrolled seizures. PET/computed tomography (CT) images were collected from DRE patients after administration of ¹⁸F-DPA-714, a radiotracer targeting 18-kDa mitochondrial translocator protein (TSPO) that has been widely used as a biomarker of activated microglia in neuroinflammatory processes.²⁹ As expected, we observed that in DRE patients, the uptake of ¹⁸F-DPA-714 was increased in the region found to exhibit metabolic abnormalities by interictal ¹⁸F-FDG PET and structural abnormalities by T1-weighted (T1W) MRI (Figure 1C).³⁰ To confirm the co-occurrence of microglial activation and neuron hyperexcitability in the excised tissues from DRE patients, whole-cell current-clamp studies were performed. The electrophysiology results showed that compared with those in control brain tissues from patients without a history of seizures, neurons in DRE patient specimens were more hyperexcitable, demonstrating a higher number of action potentials (APs), a lower AP threshold, and an obvious decrease in rheobase current (Figures 1D and S1). Importantly, significant increases in the mRNA levels of the proinflammatory cytokines interleukin-6 (IL-6) and biomarker CD86 were observed in epileptic tissues compared to the control tissues (Figures S2A and S2B). Additionally, activated microglia in the specimens from DRE patients exhibited enlarged soma and process thickening (Figure S2C). In addition, the proportion of proinflammatory microglia (CD86⁺CD11b⁺CD45^mid+) was increased 4.4-fold in epileptic tissues compared with control tissues (Figures 1E and 1F). Next, we found that MPO immunofluorescence in epileptic tissues was significantly higher than that in control tissues, and MPO was highly localized in activated microglia, which suggests that activated microglia may be the main source of MPO in the EF (Figures 1G, S3A and S3B). Consistently, the average HOCl concentration in fresh epileptic tissue homogenates was determined to be 0.45 μmol g⁻¹, which was three times higher than that in control tissues (Figure S3C). Collectively, the aforementioned data showed that the high HOCl level generated by proinflammatory microglia may serve as a reliable biomarker for identifying the EF.

Figure 1.

Proinflammatory microglia accumulate in epileptic lesions in patients with refractory epilepsy

(A) Experimental workflow of the validation that activated microglia and its biosynthetic MPO are upregulated in EF patients with epilepsy.

(B) Schematic of scalp EEG recordings in patients with epilepsy (left top) and locations of the electrodes of the international 10−20 system used for EEG recording (left bottom). Representative ictal EEG recording of a 22-year-old DRE patient with complex partial seizure (right). The EEG onset (red dotted line) of frequent spikes was found in the right parieto-occipital region at electrodes O2 and P4.

(C) Preoperative T1W MR image, PET images of glucose metabolism (¹⁸F-FDG), and microglial activation (TSPO) in the DRE patient. Postoperative CT images verified that the epileptic regions were removed by surgery. The arrows indicate the right parieto-occipital region.

(D) Left, number of APs induced by current injections ranging from 0-pA to 140-pA (n = 8 patients per group); middle, a bright-field image of a patch pipette attached to the membrane of the neuron in neocortical slices; right, photographs of fresh specimens from patients with DRE.

(E and F) Representative flow cytometry analysis (E) and quantification (F) of proinflammatory microglia (CD86⁺CD11b⁺CD45^mid+) in excised tissues from epileptic and control patients.

(G) Colocalization of MPO and Iba1 (a marker of microglia) in epileptic and control brain tissues was assessed by immunofluorescence staining. Scale bars, 50 μm and 5 μm (inset). Con., control; Ep., epilepsy; Norm., normalized.

For (D), data were expressed as means ± SD (n = 8). For (F), data were expressed as mean ± SD (n = 3), statistical analysis was calculated via unpaired two-tailed Student’s t test.

Design and characterization of the HOCl-responsive SERS probe

Developing a ratiometric SERS probe with simultaneously high sensitivity and multiplexing capability is challenging. Here, with a computer-aided Raman reporter design (CARRD) strategy, we screened 36 compounds belonging to 17 structural skeletons and obtained two reporter molecules that have high Raman activity and non-overlapping Raman spectrum (Figures 2A and 2B, and S4). We developed an ultrabright HOCl ratiometric SERS nanosensor (hereafter referred to as ultraHOCls) by conjugating functional groups including HOCl-responsive Raman reporter molecule Lip-Cy7S,³¹ reference Raman molecule Lip-NB, and blood-brain barrier (BBB)-penetrating peptide angiopep2 on gold nanoparticle surface³² (Figures S5–S16). After modifying polyethylene glycol (PEG) and angiopep2 on the surface of Au surface, the surface charge of ultraHOCls transitioned from negative (−19.0 ± 3.4 mV) to nearly neutral ζ-potential (5.5 ± 1.6 mV) (Figures S17A and S17B). Both Fourier transform infrared spectroscopy (FTIR) and Raman spectrum showed the successful synthesis of ultraHOCls (Figures S17C and S17D). UltraHOCls exhibited well-defined star-shaped morphology, accompanied by a hydrodynamic diameter of 110.3 ± 5.9 nm in distilled water (Figures S17E–S17G). Moreover, ultraHOCls demonstrated high storage stability and photostability in vitro (Figures S17H–S17L). Under the excitation of a hand-held Raman scanner, ultraHOCls exhibited an extremely low detection limit of 2.0 pM in mouse plasma (Figures 2C and S18), which is orders of magnitude higher than current molecular probes for HOCl (usually ∼ nM).³³ UltraHOCls also showed a high tissue penetration depth of 3.0 mm, making it convenient for visualizing the microscopic EF without removing the cortex in model mice (Figure 2D). The response characteristics of ultraHOCls to HOCl were investigated, and we found HOCl caused an obvious reduction in SERS intensity at 518 cm⁻¹ (I₅₁₈), whereas the SERS intensity at 586 cm⁻¹ (I₅₈₆) was barely changed (Figure 2E). Thus the stable SERS band at 586 cm⁻¹ was employed as an internal standard, allowing built-in corrections for ratiometric quantitation of HOCl (Figure 2F). UltraHOCls showed a high response velocity to HOCl and more than 90% signal variation was completed within 1 min (Figure 2G). Importantly, ultraHOCls showed high specificity to HOCl and an over 6- to 20-fold selectivity for HOCl over endogenous metal ions, biomolecules, and other types of reactive oxygen species (ROS)/reactive nitrogen species (RNS) (Figure 2H). Importantly, the I₅₈₆/I₅₁₈ SERS intensity ratio of ultraHOCls remained constant when the pH value decreased from 10.5 to 1.5, indicating that the responsiveness of ultraHOCls to HOCl was physiological pH independent (Figure S19). Together, the aforementioned data demonstrated that ultraHOCls can measure HOCl concentrations with high accuracy and temporal resolution.

Figure 2.

Design and characterization of the HOCl ratiometric SERS nanosensor ultraHOCls

(A) A CARRD strategy for construction of ultrabright ratiometric SERS nanosensor by screening and optimizing Raman molecules.

(B) Schematic of the proposed mechanism by which ultraHOCls senses HOCl with ratiometric Raman signals. ultraHOCls was synthesized by incorporating an HOCl-responsive dye, Lip-Cy7, and a reference dye, Lip-NB, onto the gold nanoparticle surface. In the presence of HOCl, Lip-Cy7 underwent structural degradation while lip-NB remained unchanged, leading to a ratiometric signal variation between characteristic Raman peak 1 (green color) and peak 2 (blue color).

(C) Concentration-dependent Raman signal intensity of ultraHOCls in PBS.

(D) Raman signal intensities of ultraHOCls covered by porcine muscle with different thicknesses.

(E) Raman spectra of ultraHOCls as a function of HOCl concentration (0−24 μM).

(F) Plots of Raman intensity ratios (I₅₈₆/I₅₁₈) versus HOCl concentrations.

(G) The dynamic change of the I₅₈₆/I₅₁₈ ratio before and after the addition of HOCl (30 μM). HOCl was added at ∼40 s.

(H) I₅₈₆/I₅₁₈ ratios of ultraHOCls in the presence of various ions, biomolecules or ROS/RNS (30 μM). Vc, L-ascorbic acid; GSH, glutathione; Cys, cysteine; NADPH, nicotinamide adenine dinucleotide phosphate.

For (C), (D), and (H), data were expressed as means ± SD (n = 4). For (F) and (G), data were expressed as means ± SD (n = 3).

UltraHOCl image proinflammatory microglia in vitro

The intriguing Raman properties of ultraHOCls inspired us to explore its capability to identify microglia phenotypes in cell cultures. Primary microglia were prepared from neonatal mouse brains and then conditioned by cytokine interferon γ (IFNγ) and/or the endotoxin lipopolysaccharide (LPS) to polarize them toward the proinflammatory phenotype or stimulated with IL-4 to polarize them toward the anti-inflammatory phenotype in vitro³⁴ (Figure 3A). Strikingly, we found both the mRNA (3.4-fold) and protein (2.0- to 14.8-fold) expression levels of MPO increased robustly in LPS+IFNγ-treated microglia in comparison to PBS-treated microglia (Figures 3B, 3C, and S20). Then, we investigated the cellular uptake kinetics of ultraHOCls in microglia and found that obvious uptake occurred as early as 30 min post-incubation (Figure S21). Moreover, a higher I₅₈₆/I₅₁₈ ratio was observed in LPS+IFNγ-treated microglia than in PBS-treated controls, and the ratio reached a plateau at 2 h. To test the capability of ultraHOCls for visualizing microglia of different phenotypes, we subsequently performed ratiometric SERS mapping of microglia exposed to different stimulation conditions. The I₅₁₈ gradually decreased and the I₅₈₆ showed an almost constant in cells stimulated with gradient concentrations of LPS (with or without IFNγ) compared with PBS-treated microglia (Figures 3D–3F). The maximum increase in the I₅₈₆/I₅₁₈ ratio was observed upon incubation with 600 ng/mL LPS plus 100 ng/mL IFNγ, suggesting the HOCl level was increased in proinflammatory microglia. In contrast, only a small change in I₅₈₆/I₅₁₈ ratio was observed in cells stimulated with IL-4. Furthermore, the increase in I₅₈₆/I₅₁₈ ratio induced by LPS+IFNγ was significantly abated by salicylhydroxamic acid (SHA), an MPO inhibitor, suggesting that ultraHOCls was highly sensitive to the elevated MPO levels (Figures 3D–3F). The aforementioned findings were confirmed by flow cytometry analysis, and all the aforementioned results indicated that our probe could specifically recognize elevated HClO in proinflammatory microglia (Figures 3G and 3H). Importantly, the I₅₈₆/I₅₁₈ ratio derived from ultraHOCls enabled precise estimation of the proportion of pro-inflammatory microglial cells (Figure S22). Furthermore, we found the expression of MPO and the I₅₈₆/I₅₁₈ ratio of ultraHOCls were the highest in proinflammatory microglia among various types of brain cells (Figure S23). Together, all these in vitro cell-based studies showed the feasibility of using ultraHOCls to visualize proinflammatory microglia by ratiometric sensing intracellular MPO.

Figure 3.

Correlation between the HOCl concentration and proinflammatory microglial proportion

(A) Schematic of isolation and polarization of the primary mouse microglia.

(B) Immunofluorescence staining images of Iba1 and MPO in the activated microglia. Scale bars, 50 μm.

(C) Quantification of MPO immunostaining in microglia.

(D) Confocal Raman microscopic imaging of endogenous HOCl with ultraHOCls in primary mouse microglia after corresponding treatments. IFNγ-Low: 40 ng/mL IFNγ; IFNγ: 100 ng/mL IFNγ; LPS+IFNγ: 600 ng/mL LPS plus 100 ng/mL IFNγ; IL-4: 40 ng/mL IL-4; LPS+IFNγ+SHA: 600 ng/mL LPS plus 100 ng/mL IFNγ followed by SHA (500 μM). Images displayed in pseudocolour represent the Raman signal intensities collected at 518 cm⁻¹, 586 cm⁻¹ and the I₅₈₆/I₅₁₈ ratio. Scale bars, 30 μm.

(E) Typical Raman spectra collected at the region indicated in the bright field images in (D).

(F) Average I₅₈₆/I₅₁₈ ratios in (D).

(G) Representative flow cytometry analysis of phenotypic polarization of cultured primary microglia after various treatments.

(H) A plot of the I₅₈₆/I₅₁₈ ratio against the percentages of CD86⁺ microglia. Tryp., trypsinization; Pro, proinflammatory; Anti, anti-inflammatory; Int, intensity, Norm, normalized.

For (C), (F), and (H), data were expressed as means ± SD (n = 4). For (C), statistical analysis was calculated via one-way ANOVA with Tukey’s post-hoc test.

The mechanism by which ultraHOCls cross the BBB was further studied using an in vitro model. As shown in Figure S24, the BBB permeability of ultraHOCls decreased by 69.2% at 4°C and reduced by 50.3% following pre-treatment with chlorpromazine. In contrast, pre-treatment with methyl-β-cyclodextrin or cytochalasin D did not significantly affect the BBB permeability of ultraHOCls. These results strongly suggested that clathrin-dependent endocytosis is involved in the transcytosis of ultraHOCls for BBB crossing.³⁵

UltraHOCls cross the BBB in kainic acid model

Kainic acid (KA) models were selected based on their strong resemblance to individuals with focal epilepsy in terms of behavioral performance, neuropathological alterations, and electrophysiological aberrations.³⁶ Due to its high sensitivity, unlimited penetration depth, and superiority in quantifiability, we labeled ultraHOCls with ^99mTc to investigate its biodistribution in vivo.³⁷ The radiolabeled nanosensor ^99mTc-ultraHOCls showed high radiochemical purity, stability, and physicochemical properties similar to those of ultraHOCls (Figure S25). SPECT/CT showed the specific uptake of ^99mTc-ultraHOCls in the EF. In chronic KA models, the uptake of ^99mTc-ultraHOCls was 1.8–2.2-fold higher in the ipsilateral cortex and hippocampus than in the contralateral hemisphere, which was corroborated by ex vivo biodistribution studies (Figure S26). The higher uptake of ^99mTc-ultraHOCls during the acute phase relative to the chronic phase may be attributed to the temporary breakdown of the BBB, which is mostly restored in the chronic phase.³² These experimental results show that ultraHOCls is capable of crossing the BBB and accumulating in the EF.

The biocompatibility and long-term safety of ultraHOCls were systematically assessed. Using the CCK-8 assay, we found ultraHOCls exhibited minimal cytotoxicity in all tested brain cells, even at concentrations significantly exceeding those used for imaging (Figure S27A). Furthermore, ultraHOCls demonstrated biocompatibility without inducing cellular inflammatory responses (Figure S27B–S27D). Standard hematological markers were measured to assess potential infection or inflammation in vivo, and all parameters remained within normal ranges after ultraHOCls treatment (Figure S28A). Hemolysis assays confirmed that ultraHOCls caused negligible hemolytic activity (<10%) (Figure S28B), while blood biochemical analyses revealed no evidence of hepatic or renal dysfunction (Figure S28C). Most notably, H&E staining at both 60 and 180 days showed no inflammatory infiltration or cell death, even at dosages exceeding five times the amount used in our imaging protocol (Figure S28D).

UltraHOCls enable the localization of the EF in vivo

To assess the performance of ultraHOCls in locating the EF, ECoG was used as the “gold standard” in the study by Shi et al, .³⁸ As shown in Figure 4A, a 64 electrode (64E) microarray was fabricated in an area of ≈11 × 7 mm², and the EF in the chronic seizure mouse could be located with a high spatial resolution (Figure 4B). We collected the Raman spectra at the locations of the electrodes after intravenous administration of ultraHOCls. Significantly, brain areas with a high I₅₈₆/I₅₁₈ ratio colocalized well with the epileptic lesions identified by the 64E array, and a cut-off I₅₈₆/I₅₁₈ value of 0.6 was used to distinguish normal and epileptic mouse brain tissues (Figures 4C, 4D, and S29). By simply not adding the Lip-Cy7S dye and otherwise following the same preparation method as the ultraHOCls probe, we generated a control probe, Ang-NB, which exhibits an “always-on” Raman signal. In contrast, the control probe Ang-NB with “always-on” Raman signal showed a low E/N ratio and off-target chemical binding (Figure S30). To further substantiate the specificity of ultraHOCls, we treated KA-induced epileptic mouse models with PLX5622 to achieve microglial depletion.³⁹ The Raman imaging results demonstrated a marked E/N ratio reduction in the microglia-depleted group compared to the control group (Figure S31). The aforementioned studies indicated the effectiveness of ultraHOCls in locating the EF by targeting activated microglia.

Figure 4.

UltraHOCls shows better performance than ECoG in intraoperatively locating the EF

(A) Schematic of the customized 64E assay for in situ monitoring of ECoG signals in live KA-induced epilepsy model mice.

(B) The seizure focus was identified as the bright yellow region (top) by the 64E assay by collecting ECoG signals in KA-induced epilepsy model mice. The stars represent the EF.

(C) An identical seizure focus was located by the ultraHOCls, in which the Raman signals at the corresponding 64 lattices demarcated by the electrode arrays were collected by a handheld Raman scanner within approximately 3 min.

(D) Unsupervised k-means clustering of data from the EF location determined by ultraHOCls and ECoG was performed.

(E) Dynamic ECoG mappings showed transient hyperexcitability of neurons in the seizure focus, followed by a fast decay to the baseline within 10 s after seizure onset (top). In contrast, ultraHOCls sustainably located the epileptic lesion during a 1-h operation time window (bottom).

(F) Temporal evolution of E/N ratios and ECoG amplitudes.

(G) Raman/ECoG signal topographies of the brain during the “on-off” cycles of anesthesia in live model mice. +, anesthetized with 2% isoflurane; −, removal of anesthesia; Raman+/ECoG+: positive Raman/ECoG signals; Raman+/ECoG−: positive Raman signals but only background ECoG signals.

(H) The E/N ratios of ultraHOCls and ECoG in KA-induced epilepsy model mice treated with or without anesthesia. Norm, normalized; T_max, time of maximum ECoG signal from seizure onset.

For (F) and (H), data were expressed as means ± SD (n = 4).

One of the main limitations of ECoG is its long sampling time, which can be attributed to the difficulty in capturing occasional and transient neural activities.¹¹ As expected, although time-dependent ECoG maps revealed epileptic activity with high bioelectrical potentials at the electrodes covering the seizure focus, the instantaneous ECoG potentials decreased rapidly and returned to baseline levels within tens of seconds (Figures 4E and 4F). In comparison, a high I₅₈₆/I₅₁₈ ratio in the EF was detected and persisted for at least 1 h, which is convenient for surgeons by providing a stable imaging time window (Figures 4E and 4F).

Distortion of the ECoG signal by anesthesia is another challenge in epileptic surgery.^13,14 To investigate the influence of anesthesia on the performance of ultraHOCls in locating the EF, we tailored the presence and absence of isoflurane for anesthesia and sobriety testing. As shown in Figure 4G, when discontinued isoflurane administration, we recorded bioelectrical potential with high ratiometric values (average normalized ECoG amplitude in seizure focus points/average normalized ECoG amplitude in all acquired 64 points; >3.3) at the seizure focus. However, the ECoG potential decreased markedly, and only background signals with ratiometric values (<1.0) were recorded after the restoration of isoflurane anesthesia. In contrast, uncompromised Raman signal intensity and I₅₈₆/I₅₁₈ ratio (>2.5) of ultraHOCls in the EF were observed without suffering the interference of isoflurane anesthesia (Figures 4G and 4H). We further evaluated the effect of intravenous anesthetics and varying anesthesia depths on the Raman signal. As shown in Figure S32, varying doses of propofol administered intravenously all interfered with ECoG detection. In contrast, altering the dosage of propofol demonstrated that neither the presence nor depth of anesthesia affected the Raman signal. These results highlight the advantage of ultraHOCls-guided surgery in accurately identifying epileptic foci without anesthetic interference.

We further explored how ultraHOCls work at the molecular level. Brain areas with a high I₅₈₆/I₅₁₈ ratio (>0.6) were identified as epileptic tissues and surgically resected, while those with a low I₅₈₆/I₅₁₈ ratio (<0.6) were excised as normal controls (Figures 5A and 5B). Ex vivo electrophysiology studies confirmed the neurons in epileptic tissues were more hyperexcitable than those in the control tissues (Figure 5C). Furthermore, a 10.86-fold increase in the percentage of proinflammatory microglia was observed in epileptic tissues compared to control tissues (Figure 5D). Marked upregulation of MPO expression and HOCl concentration were also observed (Figures 5E, 5F, and S33). Transmission electron microscopic (TEM) images showed the average density of individual ultraHOCls in the EF was 4.77 ± 3.11 pieces μm², which was 2.5 times higher than that in normal brain tissues (Figure 5G). Specific delivery of ultraHOCls into epileptic tissues further benefitted proinflammatory microglia location with high sensitivity and selectivity in vivo. Taken together, all these results demonstrated the feasibility of using our nanosensors to intraoperatively locate the EF via imaging proinflammatory microglia.

Figure 5.

UltraHOCl-guided epilepsy surgery via visualization of proinflammatory microglia

(A) Schematic of the ultraHOCl-guided epilepsy surgery strategy. Resective surgery was performed 6–24 h after intravenous administration of ultraHOCls.

(B) Schematic of the in vivo experimental procedure of ultraHOCl-guided EF resection (left panel); representative Raman spectra collected during the surgery (right panel).

(C) Patch-clamp recording from neurons in resected slices from epilepsy model mice (n = 14 for the control, n = 19 for the epilepsy). Left, number of APs induced by current injections ranging from 0-pA to 140-pA. Right, representative AP trains induced by injecting 40-pA and 80-pA currents.

(D) Representative flow cytometry results and quantification of proinflammatory microglia (CD86⁺CD11b⁺CD45^mid+) in the excised epileptic and normal tissues from KA-induced epilepsy model mice.

(E) Western blot analysis of MPO expression in excised epileptic and normal tissues from epilepsy model mice.

(F) Determination of HOCl concentrations in epileptic and normal brain homogenates.

(G) Representative TEM images and quantification of ultraHOCls in the excised epileptic and normal tissues from epilepsy model mice. Scale bars, 200 nm. The red arrows indicate the gold nanoparticles retained in the brain.

For (D)–(F), data were expressed as means ± SD (n = 4). For (G), data were expressed as means ± SD (n = 5). For (D)–(G), statistical analysis was calculated via unpaired two-tailed Student’s t test.

UltraHOCl-guided surgery alleviates seizure activity in KA model

To assess the antiseizure effect of ultraHOCls in guiding epilepsy surgery, we randomly divided the chronic seizure model mice into five groups: (1) sham craniotomy without surgery (EP-C) group, (2) ECoG-guided surgery (simulating clinical scenarios) (ECoG-S) group, (3) the ultraHOCls-guided surgery (ultraHOCls-S) group, (4) the Ang-NB-guided surgery (Ang-NB-S) group, and (5) the ultraHOCls-C group that involved only intravenous ultraHOCls injection (Figure 6A). The EEG recording results showed both SERS and ECoG-guided surgery significantly decreased seizure severity (Figures 6B and 6C). Compared with the EP-C group, the ultraHOCls-S group exhibited an overall 83% decrease in seizure frequency and 83% reduction in the total seizure time per mouse over an 8-h period (Figures 6D, 6E, and S34). Compared with the Ang-NB-S group, the ultraHOCls-S group showed a 67% decrease in seizure frequency and a 66% reduction in total duration of seizures. Notably, compared with ECoG-guided surgery, ultraHOCls-guided surgery resulted in significant decreases in seizure frequency (63%) and total seizure time (61%), possibly due to its convenience in locating the EF intraoperatively, especially for deep and microscopic lesions. Notably, the ultraHOCls-C group showed no significant difference in seizure frequency and total seizure duration compared to the EP-C group, indicating ultraHOCls alone had no obvious impact on neural activity (Figures 6B–6E). Patch-clamp experiments conducted on post-surgical tissue margins revealed that ultraHOCls surpasses other techniques in guiding the thorough resection of epileptic foci (Figure S35). Taken together, the ultraHOCl-guided surgery can more efficiently alleviate seizure activity in chronic seizure model mice than the previous strategy.

Figure 6.

UltraHOCl-guided surgery alleviates seizure activity and preserves neurological functions

(A) Diagram of experimental timeline.

(B) Representative EEG traces, energy spectra and horizontal expansions of the EEG traces recorded from the experimental groups.

(C) Quantification of the EEG coastline index.

(D) Heatmaps of seizure activity per individual mouse in an 8-h time frame over 14 days.

(E) Quantification of seizure frequency of paroxysmal discharges in every 8-h time frame.

(F) Experimental scheme for behavioral tests and corresponding key dominant brain regions.

(G) Representative 10-min track plots of mice in the open field test. See also Video S1.

(H) Total distance traveled in the open field test.

(I) Screenshot of a representative CatWalk gait analysis test (top) and schematic illustration of the CatWalk gait parameters (bottom). See also Video S3.

(J) Quantification of print areas of the four paws in the CatWalk gait analysis.

(K) Escape latency in six sessions of the hidden-platform task in the Morris water maze test.

(L) Representative swimming paths of mice in the probe trial for spatial memory retention test.

(M and N) Quantification of the number of platform crossings (M) and time spent in the targeted quadrant (N) in the probe trial.

(O) Schematic illustration of the spontaneous alternation (correct alternation), incorrect alternation and perseveration in the Y-maze test.

(P) Quantification of the percentage of spontaneous alternation in the Y-maze test. RH, right hind; RF, right fore; LH, left hind; LF, left fore.

For (C), (E), (H), (J), (K), (M), (N), and (P), data were expressed as means ± SD (n = 9). For (C), (E), (H), (M), and (P), statistical analysis was calculated via one-way ANOVA with Tukey’s post-hoc test.

To assess the effect of surgery on behavioral and functional alterations, a battery of behavioral tests was conducted to investigate affective dysregulation, motor coordination, and memory (Figure 6F). A group of normal mice (Nor-C) was added for comparison. In the open field test, ultraHOCl-S-guided surgery showed effective therapy in alleviating hyperactivity with a 49% reduction in total distance and a 50% reduction in edge velocity compared with EP-C (Figures 6G, 6H, and S36; Video S1). Similarly, higher locomotor activity, such as total distance and total velocity, was observed in the EP-C and ECoG-S groups compared with the ultraHOCls-S groups in the elevated plus-maze test (Figure S37; Video S2). In tail suspension test, we found no significant change in either immobility time or active time in comparison to EP-C (Figure S38). These results signified that ultraHOCls was effective in alleviating epilepsy-related hyperactivity.

In CatWalk gait analysis (Figures 6I, 6J, and S39; Video S3), we found no statistical differences in the analyzed gait parameters, including cadence, speed, and swing among the tested groups except Ang-NB-S groups. In contrast, we found an obvious decrease in print area and max intensity, which indicated that Ang-NB-guided surgery may damage the motor function of mice. Similar trends of motor coordination were further corroborated in the accelerating rotarod test (Figure S40). The aforementioned results showed that Ang-NB-guided surgery may affect the motor function due to nonspecific imaging and removal of EF, which can be avoided by our radiometric surgery with ultraHOCls.

In the Morris water maze test, the mice in all the other groups exhibited markedly longer escape latencies, fewer platform crossings, and longer escape latencies than those in Nor-C group, indicating impaired hippocampus-dependent spatial learning and memory (Figures 6K–6N, and S41). In sharp contrast, there were no statistical differences in the alterations, the number of arm entries or the distance traveled among these groups in the Y-maze test, indicating that mice with chronic seizures and those that underwent surgical treatment showed no obvious impairment of working memory (Figures 6O, 6P, and S42). Not surprisingly, all behavioral test results revealed minimal differences between EP-C and ultraHOCls-C. These results indicated that the mice receiving only ultraHOCls injection did not affect working memory, spatial memory, motor coordination, and anxiety. Taken together, all these behavioral and functional data indicated that our ultraHOCl-guided surgery strategy can alleviate epilepsy-associated hyperactivity without obviously impairing motor function or aggravating memory deficits.

UltraHOCls define epileptic regions in excised tissues from patients

To test the feasibility of the ultraHOCls for intraoperatively locating epileptic regions in a more clinically relevant scenario, we first validated the existence of epileptic discharge with patch-clamp on fresh brain specimens resected from patients (Figure 7A). Then we sprayed a solution of ultraHOCls aqueous on the brain tissue slice and observed a more than 2-fold increase in I₅₈₆/I₅₁₈ ratios in epileptic tissues compared to control tissues (Figures 7B–7E). In addition, the pathological examination ascertained that the area of the higher I₅₈₆/I₅₁₈ ratios corresponded to the epileptic tissues (Figure 7D). Next, we performed receiver-operating characteristics (ROC) analysis to differentiate epileptic lesions from nonepileptic human brain tissues using I₅₈₆/I₅₁₈ ratios and identified a cut-off point of 0.71 with a sensitivity of 94.89% and specificity of 93.3% (Figures 7F and S43). The area under the curve value of 0.98 (95% confidence interval [CI]: 0.975–0.988) indicated a strong discriminative ability between the epileptic and control tissues with ultraHOCls. Of note, a strong positive correlation was observed between the I₅₈₆/I₅₁₈ ratios, the electrical activities, and the MPO expression level (Figure 7G). The aforementioned studies demonstrated the feasibility of ultraHOCls for locating the EF in clinically relevant scenarios.

Figure 7.

UltraHOCls imaging epileptic lesions in fresh specimens resected from patients with DRE

(A) Schematic illustrating the experimental workflow.

(B) Photographs and corresponding AP trains induced by injecting 0-pA current at selected ROIs in fresh specimens from patients with DRE and control patients.

(C) Raman signal topographies and typical Raman spectra at selected ROIs in the brain slices after spraying ultraHOCls.

(D) Immunohistochemistry of MPO in excised tissues from patients with DRE and control patients. Scale bars, 100 μm.

(E) Violin plots of I₅₈₆/I₅₁₈ ratios in the brain slices after spraying ultraHOCls.

(F) ROC curves showing the performance of ultraHOCls in discriminating epileptic tissues from non-epileptic tissues.

(G) The I₅₈₆/I₅₁₈ ratios increase proportionally with the expression of MPO and the average number of AP trains (induced by current injections ranging from 0-pA to 140-pA) in fresh specimens from patients with DRE and control patients. ROI, regions of interest; AOD, average optical density.

Discussion

Despite the massive theoretical population demand for epilepsy surgery, its widespread implementation remains limited even in developed nations, which can be attributed to the high economic and technical costs required. Currently, patients who are referred for epilepsy surgery typically undergo a comprehensive presurgical evaluation for the localization of the EF, which includes high-resolution MRI, video scalp EEG, and a detailed neuropsychological assessment.⁵ However, in cases where the results are inconclusive or ambiguous, additional imaging modalities such as PET or SPECT are employed to identify areas of focal hypometabolism during the interictal period or hypermetabolism following a recent seizure.⁵ In some cases, further evaluation with SEEG may be conducted to localize the suspected EF during a phase 2 evaluation⁵ (Figure S44).

Accurate localization of EF is crucial for improving surgical outcomes in DRE patients. In our previous study, we reported an electric field-responsive MR probe that enhances preoperative localization of EF by detecting changes in T1 relaxation rates in response to abnormal neuronal discharges.⁴⁰ However, the importance of intraoperative localization of the EF cannot be overstated. Even in well-selected patients with DRE, 30%–50% of them exhibit seizure recurrence after surgery, despite congruent presurgical evaluation.^7,41 One of the main reasons for surgical failure is the inability to completely resect the EF.⁴¹ Intraoperative EF localization is crucial for addressing challenges such as brain shift and boundary determination, which can significantly affect surgical outcomes.⁴² Although intraoperative MRI is used for localization during surgery, its high cost and prolonged imaging time limit widespread application. Moreover, similar to ECoG, the MR probe enhances signals by responding to neuronal discharges, which can be disrupted by intraoperative anesthesia. Additionally, various optical imaging techniques have been reported for intraoperative localization of EF, but these also require the occurrence of epileptic seizures.^15,16,18,43 To increase its effectiveness, reduce overall cost, and improve access to epilepsy surgery, alternative techniques need to be developed.

Our ultraHOCl technology represents one such advancement, offering superior intraoperative capability by recognizing the molecular biological signature of seizures, rather than relying solely on traditional electrical signatures. This work demonstrates its superiority for intraoperative use in the following ways: (1) providing a stable and reproducible signal, independent of concurrent seizure activity; (2) functioning as an independent technology with little interference with existing adjuncts; and (3) potentially offering greater accuracy in predicting the boundary of the EF. This study also has limitations. The tissue penetration depth of ultraHOCls is limited to 3 mm, requiring its use in combination with existing preoperative evaluations, such as SEEG, which remains the gold standard for seizure localization. However, if ultraHOCls proves clinically effective, it could become the gold standard intraoperative tool for seizure focus identification, much like 5-aminolevulinic acid for glioblastoma surgery.⁴⁴ Additionally, once its efficacy in localizing epileptic foci is validated, the strategy that visualizing proinflammatory microglia could be refined for preoperative imaging through non-invasive techniques. This advancement may allow certain patients to bypass SEEG and proceed directly to surgery, streamlining the treatment process. Notably, the technique can be adapted for other neurological diseases, expanding its potential use beyond epilepsy surgery.

Limitations of the study

While this study confirms the biocompatibility of ultraHOCls, the limited degradability of gold nanoparticles and their in vivo metabolism require further investigation. Localized spraying may serve as a potential alternative. Moreover, given the diverse etiologies of epilepsy, validating this strategy across multiple animal models will be essential for assessing its broader applicability. Another limitation is the slow acquisition speed of Raman imaging due to its point-scanning nature. Implementing a multi-point scanning system or developing faster intraoperative imaging techniques could help overcome this limitation.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to the lead contact, Cong Wang (cong_wang@fudan.edu.cn).

Materials availability

Materials generated in this study will be available upon completion of a material transfer agreement (MTA).

Data and code availability

The main data supporting the results of this study are available within the article and its supplemental information. The datasets generated and analyzed in the current study are available from the corresponding author upon reasonable request. The raw MATLAB codes are available within the supplemental information. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. All original NMR data have been deposited at website (https://bmrbig.org/released).

Acknowledgments

We thank Dr. Xunyi Wu and Lu Wang for their helpful discussions. We also thank Guorong Lin for his help with animal studies. This work was supported by the National Key Research and Development Program of China (2023YFA1801200, 2023YFF0714200, 2019YFA0709504, and 2021ZD0202805), the National Natural Science Foundation of China (92159304, 82227806, 82272116, 32471083, 82101526, 82202224, and 82472038), the Shanghai Explorer Program (23TS1401100, China), National Science Fund for Distinguished Young Scholars (82025019), Shanghai Municipal Health Commission (2022ZZ01006), National Postdoctoral Program for Innovative Talents (BX20200095), the Shanghai Rising-Star Program (23QA1407700), the Shanghai Sailing Program (23YF1448600), the construction project of Shanghai Key Laboratory of Molecular Imaging (18DZ2260400), Science and Technology Innovation Plan of Shanghai Science and Technology Commission (23Y31900300), Shanghai Municipal Health Commission Collaborative Innovation Group (2024CXJQ03), the Innovative Research Team of High-level Local Universities in Shanghai, 111 Project (B18015), Shanghai Municipal Science and Technology Major Project (2018SHZDZX01), Shanghai Center for Brain Science and Brain-Inspired Technology, and the Greater Bay Area Institute of Precision Medicine (Guangzhou). The funders had no role in study design, data collection and analysis, decision to publish, or writing of the manuscript.

Author contributions

Conceptualization, C.W., C.L., and Y.M.; methodology and investigation, C.W., Z. Li, W.D., Y.J., M.C., Y.C., Z.C., J. Zhao, F.Z., K.Y., and Q.W.; resources, L.C. and Y.M.; writing—original draft, C.W., D.W., H.Z., X.X., L.C. C.L., and Y.M.; supervision, Y.M.; writing—review & editing, C.W., D.W., H.Z., X.X., L.C. C.L., and Y.M.

Declaration of interests

L.C., C.W., Z. Li, W.D., and W.S. are inventors of a patent application (Chinese patent application no. 202210869400.0) that covers the ultraHOCls system.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Iba1 antibody (clone Polyclonal)	Novus Biologicals	Cat. #NB100-1028, RRID: AB_521594
Anti-myeloperoxidase antibody (clone EPR20257)	Abcam	Cat. #ab208670, RRID: AB_2864724
Anti-CD163 antibody (clone EPR19518)	Abcam	Cat. #ab182422, RRID: AB_2753196
Anti-iNOS antibody (clone EPR16635)	Abcam	Cat. #ab178945, RRID: AB_2861417
Anti-CD31 antibody (clone 390)	Arigo	Cat. #ARG20901, RRID: AB_2940966
Anti-ZO-1 antibody (clone Polyclonal)	Thermo Fisher	Cat. #61–7300, RRID: AB_2533938
PE anti-human CD45 antibody (clone 2D1)	Biolegend	Cat. #368510, RRID: AB_2566370
APC anti-human CD11b antibody (clone ICRF44)	Biolegend	Cat. #301309, RRID: AB_314161
FITC anti-human CD206 antibody (clone 15-Feb)	Biolegend	Cat. #321103, RRID: AB_571904
PerCP/Cyanine5.5 anti-human CD86 antibody (clone IT2.2)	Biolegend	Cat. #305419, RRID: AB_1575070
PE/Cyanine7 anti-mouse CD45 antibody (clone 30-F11)	Biolegend	Cat. #103114, RRID: AB_312979
PerCP/Cyanine5.5 anti-mouse/human CD11b antibody (clone M1/70)	Biolegend	Cat. #101227, RRID: AB_2565948
Alexa Fluor® 488 anti-mouse CD206 antibody antibody (clone C068C2)	Biolegend	Cat. #141710, RRID: AB_893232
PE anti-mouse CD86 (clone GL-1)	Biolegend	Cat. #105007, RRID: AB_313150
Recombinant anti-myeloperoxidase antibody (clone EPR20257)	Abcam	Cat. #ab208670, RRID: AB_2864724
Anti-GAPDH antibody-loading control (clone Polyclonal)	Abcam	Cat. #ab70699, RRID: AB_1209569
Biological samples
Human DRE specimen	Department of Neurosurgery, Huashan Hospital.	N/A
Nonepileptic brain specimen	Department of Neurosurgery, Huashan Hospital.	N/A
Chemicals, peptides, and recombinant proteins
HAuCl4	Sigma-Aldrich®	Cat. #16903-35-8
Citrate	Sigma-Aldrich®	Cat. #854
Ascorbic acid	Sigma-Aldrich®	Cat. #1043003
AgNO3	Sigma-Aldrich®	Cat. #248762
mPEG-SH	Sigma-Aldrich®	Cat. #QBD10792
IL-4	MedChemExpress	Cat. #HY-P70644
LPS	MedChemExpress	Cat. #HY-D1056
IFN-γ	MedChemExpress	Cat. #HY-P7071
Salicylhydroxamic acid	Sigma-Aldrich®	CAS: 89-73-6
PLX5622 in AIN-76A diet	SYSE BIO	Cat. #D20010801
Control diet in AIN-76A	SYSE BIO	Cat. #PD1001
Kainic acid monohydrate	Sigma-Aldrich®	CAS: 58002-62-3
Isoflurane	Sigma-Aldrich®	CAS: 26675-46-7
Propofol	Sigma-Aldrich®	CAS: 2078-54-8
Bovine serum albumin	Solarbio Life Sciences	CAS:9048-46-8
2-(4-amidinophenyl)-6-indolecarbamidine dihydrochloride (DAPI)	Beyotime	Cat. #C1002
Transwell filter insert	Corning	Cat. #3460
Fluoro-Jade B (Fjb)	Warbio	Cat. #TR-150-FJB
Intracellular staining perm wash buffer	Biolegend	Cat. #421002
Cell staining buffer	Biolegend	Cat. #420201
Fixation buffer	Biolegend	Cat. #420801
RIPA lysis buffer	Beyotime	Cat. #P0013B
QuickBlock™ blocking buffer	Beyotime	Cat. #P0220
18F-DPA714	Huashan Hospital PET Center	N/A
18F-FDG	Shanghai Cancer Center	N/A
Critical commercial assays
Cell Counting Kit-8	MeilunBio®	CAS: 65162-13-2
Super electrochemiluminescence (ECL) detection reagent	Yeasen Biotechnology	CAS: 36222ES60
SYBR Premix Ex Taq kit	TAKARA Bio	Cat. #639676
PrimeScript™ RT master mix	TAKARA Bio	Cat. #RR036B
Bicinchoninic acid (BCA) protein assay kit	Beyotime	Cat. #P0009
Deposited data
Raw NMR data of Cy7S	This study	BMRbig ID: bmrbig119
Raw NMR data of Lip-Cy7S	This study	BMRbig ID: bmrbig120
Raw NMR data of Lip-NB	This study	BMRbig ID: bmrbig121
Raw NMR data of Lip-PEG-Angiopep2	This study	BMRbig ID: bmrbig123
Raw NMR data of Lip-PEG-MAL	This study	BMRbig ID: bmrbig122
Experimental models: Cell lines
bEnd.3 cells	Aoruisai	Cat. #ORC002988
Primary microglia	This study	N/A
Primary astroglia	This study	N/A
Experimental models: Organisms/strains
Mouse: C57BL/6J	Slac Lab Animal Ltd.	http://www.slaccas.com/
Software and algorithms
OceanView software (Version 1.6.7)	Ocean Optics	https://www.oceanoptics.com
LabSpec5 software (Version 5.58.25)	HORIBA	https://www.horiba.com/
WITec Project software (Version 5.2)	WITec	https://raman.oxinst.com/
PowerLab system	AD Instruments	Cat. #PL3516
LabChart (Version 8.1.19)	AD Instruments	https://www.adinstruments.com/
MATLAB (Version R2013b)	MathWorks	https://ww2.mathworks.com
EthoVision XT (Versoin 15.0.1418)	Noldus	https://noldus.com/
Graphpad Prism (Version 7.00.)	GraphPad	https://www.graphpad.com/
OriginPro 2021b (Version 9.8.5.201)	OriginLab	https://www.originlab.com/
ImageJ (version 1.53k)	ImageJ	https://imagej.net/
FlowJo software (Version 10.4.2)	Becton, Dickinson & Company	https://www.flowjo.com/
DigitalMicrograph software (Version 1.71.38)	Gatan Company	https://www.gatan.com/cn
Zetasizer software (Version 7.11)	Malvern Panalytical	https://www.malvernpanalytical.com/
FieldTrip (Version 20200130)	Fieldtriptoolbox	https://www.fieldtriptoolbox.org/
3D Slicer software (Version 5.2.2)	3D Slicer image computing platform	https://www.slicer.org/
Inveon Research Workplace software (Version 4.2.0.15)	Siemens	http://www.siemensmedical.com/

Experimental model and study participant details

Patient selection and sample acquisition

The written informed consents of all participants were obtained and their clinical, imaging and neuropathological information are provided in Tables S1 and S2. The Ethics Committee of Huashan Hospital approved all patient-related studies (Shanghai, China) (Approval No. KY2020-065). For DRE patients who received resective surgery, a comprehensive pre-surgical work-up, including video-EEG monitoring, MRI and PET imaging, was performed. Human specimens were obtained directly from the surgery room at the Department of Neurosurgery, Huashan Hospital. Nonepileptic brain specimens obtained from the patients who did not display epileptic features, as assessed by scalp EEG and other related clinical examinations, were used as controls. The in vivo TSPO imaging in patients, collection and use of human resected tissues were approved by the human ethical committee in Huashan Hospital, Fudan University, China (Approval No. KY2022-030).

Mice and animal housing

The C57/BL6 mice were used for cell (1−2 days old) and animal (male; 4−8 weeks old) studies, respectively. Mice were housed in the SPF facility under a constant 12 h cycle of light-dark and allowed free access to regular chow and water. The temperature and humidity of the SPF facility were controlled (23°C and 50% humidity). For brain microglia pharmacological ablation, the mice were fed with a PLX5622 formulated diet ad libitum for 10 days. During the entire surgical procedure, the body temperature of anesthetized mice was kept at 37°C with a heating pad. After surgery, mice were housed individually for better recovery. At the end of the follow-up, mice were euthanized via inhalation overdose of isoflurane, followed by cervical dislocation to ensure death. Experimental procedures for animal experiments were performed in accordance with institutional animal welfare guidelines and approved by the ethics committee of Fudan University.

Seizure mouse model establishment

For the establishment of the intra-hippocampal KA model, the mice were anesthetized with 1–2% isoflurane and head-fixed in a stereotaxic apparatus (SJK-001, Jiekai Seiko, China). An incision was made to expose the skull surface, and burr holes were then stereotactically made on the skull after scraping the pericranium away with H₂O₂. KA (0.6 μg/μL, 0.5 μL) was microinjected into the dorsal hippocampal (anteroposterior = −2.0 mm, mediolateral = −1.8 mm, dorsoventral = −1.8 mm) using bregma as the reference. Seizure behaviors were rated by an experienced investigator according to our reported Racine scale.⁴⁰ Mice that reached stages 4–6 were selected for the subsequent studies. Seizure stages 1–3 indicate nonconvulsive focal seizures (FSs) and stages 4–6 are generalized seizures (GSs). The mice injected with isometric saline under the same conditions were used as controls. In all experiments, the mice with spontaneously recurrent seizures were confirmed as chronic epileptic mice.

Isolation of primary microglia and astroglia

Primary mixed glial cultures were prepared from neonatal mouse cerebral cortices with a mild trypsinization method.⁴⁵ Briefly, we removed the meninges of the cortical tissues, digested them with 0.25% trypsin-EDTA (37°C, 10 min), and then mechanically triturated in 10% FBS-DMEM/F12. Single-cell suspensions were obtained after passing the mixed cortical cells through a cell strainer with nylon mesh (70-μm pores). The obtained cells were seeded into a flask in 10% FBS-DMEM/F12 and cultured at 37°C in humidified 5% CO₂/95% air. We completely replaced the medium every 4–5 days and obtained the confluent mixed glial cells after 12–14 days of culture. Microglia were isolated from the mixed glial cultures with a diluted trypsin solution (0.05% trypsin-EDTA) for 20–40 min. The digested astroglia and the microglial cells that remained attached to the flask were harvested separately for further studies. Mycoplasma contamination testing was performed and all cultures tested negative.

Method details

Whole-Cell patch-clamp

The freshly resected brain tissues were immediately placed into ice-cold oxygen-saturated (95% O₂ mixed with 5% CO₂) cutting solution containing (in mM): 220 sucrose, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 0.5 CaCl₂, 7 MgCl₂, and 10 D-glucose, (pH 7.4, ∼350mOsm). Slices with a thickness of ∼300 μm were cut in this solution with a vibratome (VT1000S, Leica Instruments Ltd., German), and then immediately transferred to an incubation chamber filled with artificial cerebral spinal fluid (ACSF, in mM), 125 NaCl, 2.5 KCl,1.3 NaH₂PO₄, 25 NaHCO₃, 2 CaCl₂, 1.3 MgCl₂, and 15 D-glucose (pH 7.4, 300−310 mOsm) for 30–60 min at 35°C. During experiments, the brain slice was transferred to a submersion-recording chamber and continuously perfused with oxygenated ACSF at a rate of 2–2.5 mL min⁻¹ at r.t. Whole-cell patch clamping was performed using an EPC 10 USB amplifier (HEKA Instruments, Germany), with a low-pass filter (2.9 kHz) and a sample rate of 1 kHz. Patch pipettes (4–9 MΩ) were filled with the following internal solution (in mM): 145 K-gluconate, 5 NaCl, 1 MgCl₂, 10 HEPES, 2 Mg-ATP, 0.1 Na₃-GTP, 0.2 EGTA, 10 phosphocreatine disodium (pH 7.25, 280–290 mOsm). With the injection of hyperpolarizing current (20 pA increments from 0-pA to 140-pA, 1000 ms), the evoked action potential was recorded. We discriminated putative GABAergic neurons and pyramidal neurons in the hippocampus on the basis of their waveform and autocorrelogram of firing characteristics according to previously published criteria.^46,47

RT-qPCR

Total RNA was extracted from brain tissues or cultured microglia according to the manufacturer’s protocol. Then the quantity and quality of the RNAs were analyzed on Synergy HTX Multi-Mode Reader (HTX, BioTek Gen5, Vermont). The extracted RNA samples were then reverse-transcribed to complementary DNA (cDNA). PCR amplifications were carried out in duplicate on the QuantStudio 3D Digital PCR Systems (Applied Biosystems, CA) with the following thermal cycling (40 cycles): initial denaturation (95°C, 30 s) and then amplification, 95°C, 5 s and 60°C, 34 s). Relative mRNA expression levels of MPO, CD86, IL-6, iNOS, CD206, Arg-1 and TGF-β were determined by normalizing to internal GAPDH expression. The list of DNA primer sequences was provided in Table S3.

Immunostaining studies

For immunofluorescence studies, the frozen sections (10 μm) or cultured cells were permeabilized with 0.3% Triton X-100 for 15 min and blocked for 1 h at r.t. They were then incubated with primary antibodies overnight at 4°C and then with the corresponding secondary antibodies at r.t. for 1 h. After staining with DAPI (0.5 mg/mL, 5 min), the slices or cells were washed thoroughly with PBS, air-dried and then sealed with an antifade mounting medium. The fluorescence images were acquired on a Confocal Laser Scanning Microscope (LSM 710, Carl Zeiss, Germany). Fluorescence intensities were quantified using ImageJ (version 1.53k). For immunohistochemistry studies, the histology sections from paraffin-embedded brain tissues (5 μm) were deparaffinized, rehydrated and then underwent antigen retrieval. The slices were blocked and then incubated with primary antibodies overnight at 4°C. After incubated with corresponding horseradish peroxidase enzyme-labeled secondary antibodies at r.t. for 1 h, the above slices were set on an optical microscope (Olympus BX51, Olympus, Japan) with the assistance of a BIOPAD digital camera (DP72; Olympus, Japan) for immunoreactivity images collection.

Immunoblotting

After harvesting brain tissues or cultured cells, we lysed them with RIPA buffer with a 1% protease inhibitor cocktail. The tissues were homogenized using a tissue grinder (MX-2018, Jinxin, China) (90 Hz, 60 s). The lysed samples were centrifuged (14,000 g, 4°C, 20 min). After determining protein concentrations with a BCA assay, we mixed the supernatants with loading buffer (5✕), boiled (95°C, 10 min), followed by cooling (0°C, 3 min). The proteins were separated by 10% SDS PAGE and transformed to a nitrocellulose membrane in an ice bath. The membranes were blocked with 5% skimmed milk in triethanolamine buffered saline-tween (TBS-T) at r.t. for 1 h. Then we incubated the membranes with primary antibodies overnight at 4°C, followed with HRP-conjugated secondary antibodies (1:10000, YEASEN) at r.t. for 2 h. The immunoblot images were visualized on ChemiDocTM MP Imaging System (Bio-Rad, Hercules, CA), and cumulative densitometric analyses were performed with ImageJ (Version 1.53k).

Flow cytometric analysis

For cultured cellular studies, cellular suspensions were harvested and centrifuged (2,800 g, 3 min) at 4°C. For cell-surface marker staining, we blocked the cells with 5% BSA-PBS at r.t. for 10 min. For intracellular marker staining, the cells were permeated and then blocked as described above. After staining with fluorophore-conjugated primary antibodies at r.t. for 20 min, the cells were fixed with fixation buffer, washed and resuspended with PBS. For tissue-based analysis, the brain tissues were collected from patients or mice, minced with scissors, and mechanically homogenated in Cell Staining Buffer. Single-cell suspensions were obtained by filtering with a 70-μm membrane filter. The following procedure and antibodies used were identical to the in vitro flow cytometry for cells described above. Flow cytometric data were collected on a BD LSR flow cytometer (Becton, Dickinson and Company, Maryland) and analyzed using FlowJo software (Version 10.4.2).

Synthesis of Raman reporters

Cy7S. The Cy7Cl was synthesized according to our previous publication.⁴⁸ Then the reaction material Cy7Cl (109 mg, 0.2 mmol), 3-mercaptopropionic acid (25 mg, 0.24 mmol) and triethylamine (TEA, 30 mg, 0.3 mmol) were dissolved in anhydrous DMF and stirred at r.t. for 24 h. Then the product was recrystallized in ether and obtained as green precipitate after filtrating. The solid was purified via silica gel chromatography with the gradient of CH₂Cl₂/CH₃OH (from 50: 1 to 5: 1) to afford Cy7S as a green powder. Yield: 169.3 mg (75%). ¹H NMR (400 MHz, Methanol-d₄): δ 8.90 (d, J = 14.2 Hz, 2H), 7.49 (dd, J = 7.6, 3.9 Hz, 1H), 7.43–7.35 (m, 2H), 7.31 (dd, J = 8.3, 3.8 Hz, 2H), 7.27–7.21 (m, 2H), 6.39 (d, J = 14.1 Hz, 2H), 4.26 (t, J = 5.1 Hz, 4H), 3.95 (t, J = 5.1 Hz, 4H), 3.03 (t, J = 7.5 Hz, 2H), 2.66 (s, 4H), 2.46 (s, 2H),1.90 (d, J = 6.4 Hz, 2H), 1.76 (s, 12H). ¹³C NMR (101 MHz, Methanol-d₄): δ174.90, 174.90, 158.23, 147.33, 144.27, 142.43, 134.86, 129.66, 126.14, 123.37, 112.29, 102.79, 60.05, 50.57, 47.95, 35.33, 39.03, 28.50, 27.26, 22.30. ESI-MS m/z: 613.2 [M]⁺; HR-MS m/z: 613.3100 (Calc. 613.3095) [M]⁺, C₃₇H₄₅N₂O₄S⁺.

Lip-Cy7S. The intermediate Cy7S (61 mg, 0.1 mmol), N, N′-dicyclohexylcarbodiimide (DCC, 62 mg, 0.3 mmol), 4-dimethylaminopyridine (DMAP, 37 mg, 0.3 mmol) and lipoic acid (103 mg, 0.5 mmol) were dissolved in anhydrous DMF and stirred in dark at r.t. for 18 h. The solution was removed to obtain the crude product and then purified via silica gel chromatography with the gradient of CH₂Cl₂/CH₃OH (from 50: 1 to 5: 1) to give pure Lip-Cy7S as a dark green solid. ¹H NMR (400 MHz, DMSO-d₆) δ (ppm): 8.75 (d, J = 14.1 Hz, 2H), 7.59 (d, J = 7.4 Hz, 2H), 7.50–7.37 (m, 4H), 7.26 (t, J = 7.0 Hz, 2H), 6.43 (d, J = 14.2 Hz, 2H), 4.54 (t, J = 4.8 Hz, 4H), 4.46 (t, J = 4.8 Hz, 4H), 3.46 (dq, J = 8.9, 6.2 Hz, 2H), 3.16–3.00 (m, 4H), 2.97 (t, J = 7.1 Hz, 2H), 2.67 (s, 4H), 2.48 (s, 2H), 2.33 (dq, J = 12.5, 6.2 Hz, 2H), 2.10 (t, J = 7.5 Hz, 4H),1.84 (s, 2H), 1.77 (dt, J = 12.9, 6.7 Hz, 2H), 1.68 (s, 12H), 1.55–1.38 (m, 4H), 1.34–1.26 (m, 4H), 1.18 (q, J = 7.6 Hz, 4H). ¹³C NMR (101 MHz, DMSO-d₆) δ (ppm): 172.72, 172.45, 172.45, 155.71, 145.20, 142.16, 140.73, 133.15, 128.35, 124.84, 122.36, 111.26, 101.81, 60.53, 55.87, 50.57, 48.82, 43.02, 38.06, 35.22, 33.92, 33.22, 33.00, 27.96, 27.33, 25.76, 23.79, 20.53. ESI-MS m/z: 990.2 [M + H]⁺; HR-MS m/z: 989.3759 (Calc. 989.3754) [M]⁺, C₅₃H₆₉N₂O₆S₅⁺.

Synthesis of Lip-NB. Nile blue (83 mg, 0.2 mmol), O-(7-azabenzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HATU, 91 mg, 0.24 mmol), N,N-Diisopropylethylamine (DIPEA, 39 mg, 0.3 mmol) and lipoic acid (103 mg, 0.5 mmol) were dissolved in anhydrous CH₂Cl₂ and vigorously stirred at r.t. for 24 h. The reaction solution was filtered to remove the undissolved solid, and the filtrate was evaporated and chromatographed on silica gel with gradient of CH₂Cl₂/CH₃OH (changed from 100: 1 to 10: 1) to afford Lip-NB as a blue powder. ¹H NMR (400 MHz, DMSO-d₆) δ 10.70 (s, 1H), 8.94 (d, J = 8.8 Hz, 1H), 8.66 (s, 1H), 8.29–7.70 (m, 4H), 7.43 (s, 1H), 3.94 (s, 4H), 3.67 (dq, J = 8.6, 6.1 Hz, 1H), 3.26–3.08 (m, 2H), 2.74 (t, J = 7.5 Hz, 2H), 2.44 (dt, J = 12.4, 6.2 Hz, 1H), 1.90 (dq, J = 13.4, 6.8 Hz, 1H), 1.82–1.63 (m, 4H), 1.50 (q, J = 7.6 Hz, 2H), 1.33 (t, J = 7.0 Hz, 6H). ESI-MS m/z: 506.2 [M]⁺; HR-MS m/z: 506.1935 (Calc. 506.1930) [M]⁺, C₂₈H₃₂N₃O₂S₂⁺.

Synthesis of Lip-PEG-MAL

A mixture of NH₂-PEG-Mal (500 mg, 0.1 mmol), HATU (190 mg, 0.5 mmol), DIPEA (65 mg, 0.5 mmol) and lipoic acid (103 mg, 0.5 mmol) in 10.0 mL anhydrous DMF was stirred thoroughly at r.t. for 24 h. Lip-PEG-MAL was purified by dialyzing against ultrapure water followed by lyophilization.

Synthesis of Lip-PEG-Angiopep2

Coupling of Lip-PEG-MAL (52 mg, 10 μmol) and cysteine-modified angiopep2 peptide (24 mg, 10 μmol) was carried out in a mixture solution of acetonitrile/PBS (0.01 M, pH 7.4) (15:85 V/V) at r.t. Lip-PEG-Angiopep2 was purified by dialyzing against ultrapure water followed by lyophilization.

Synthesis of ultraHOCls

All the glassware and magnetic stir bars were soaked in aqua regia solution (HCl/HNO₃, 3:1) overnight, and then rinsed thoroughly with ultrapure water prior to use. Initially, we introduced 15 mL of 1% citrate solution into a boiling HAuCl₄ solution under vigorous stirring. The solution was kept boiling for 15 min. During this process, the solution gradually turned from brown to wine red. For the synthesis of gold nanostars, a solution of HAuCl₄ (200 mL, 0.25 mM) was prepared, and then 1 mL of HCl (pH = 1) and 4 mL of the citrate-stabilized seed solution were sequentially added with moderate stirring. Under vigorous stirring, we added ascorbic acid (1.0 mL, 100 mM) and AgNO₃ (2.0 mL, 3.0 mM) simultaneously to the above mixture solution. After obtaining the gold nanostar solution, methanol solution containing Lip-NB, Lip-Cy7S, mPEG-SH and Lip-PEG-Angiopep2 were successively added and stirred for 2–4 h (molar ratio of Lip-NB/lip-Cy7S/mPEG-SH/Lip-PEG-Angiopep2/AuS ≈100,000/20,000/1000,000/100,000/1). Subsequently, the SERS probe ultraHOCls was obtained by centrifugation (11,000 rpm) and washed three times with ultrapure water. The final SERS probe can be easily dispersed in water for further use. A control SERS nanosensor (Ang-NB) without lip-Cy7S and another control SERS nanosensor uraHOCls without Angiopep2, were also developed.

Raman imaging

A portable hand-held Raman spectrometer (QE65, Ocean Optics, Florida) with a spectral resolution of 4.5 cm⁻¹ was employed for both in vitro and in vivo Raman imaging studies, excluding Raman confocal studies. Raman spectra were recorded using a 785 nm NIR laser with OceanView software (Version 1.6.7). Raman images were obtained using a Raman-point mapping method and the acquired raw SERS spectra were disposed using LabSpec5 software (Version 5.58.25).

Raman activity calculation

The theoretical calculations were performed via the Gaussian 09 suite of programs. The structures were fully optimized at the B3LYP/6-311G(d) level of theory. The vibrational frequencies of the optimized structure were carried out at the same level. No constraints to bonds/angles/dihedral angles were applied in the calculations, and all atoms were free to optimize.

Determination of HOCl concentration

For ex vivo HOCl measurement, fresh brain tissues were harvested and homogenized in ice-cold PBS (pH 7.4) with a ratio of 1:7 volumes (W/V) by using a tissue grinder (60 Hz, 120 s). The above solutions were ultrasound for 10 min and then centrifuged (10,000 g, 4°C, 10 min). For cellular HOCl measurement, the harvested microglia with different phenotypes were lysed with RIPA buffer and centrifuged (10,000 g, 4°C, 10 min). We collected the homogenate supernatants for the following tests according to the supplier’s instructions. We calculated the HOCl concentration according to a calibration curve.

Radio-synthesis of ^99mTc-ultraHOCls

To synthesize thiol diethylenetriaminepentaacetic acid (DTPA) derivative, a mixture of mercaptoethylamine (3.8 mg, 50 μmol), and DTPA dianhydride (36 mg, 0.1 mmol) in 5.0 mL anhydrous DMF was stirred thoroughly with TEA (10 mg, 0.1 mmol) at r.t. for 12 h, and the reaction solution was used for subsequent reactions without any processing. DTPA-modified ultraHOCls was synthesized with a similar procedure as ultraHOCls but added an appropriate amount of reaction solution containing the above thiol-DTPA derivative. For the radiolabeling of ^99mTc, DTPA-modified ultraHOCls and stannous chloride (SnCl₂, 100 μg, pH 2.0) were added to 2 mL PBS (pH 6.5) and vortexed. Then we added freshly eluted ^99mTechnetium pertechnetate (0.2–0.5 mL, ∼200 MBq) to the above mixture solution and incubated at 50°C for at least 30 min. The mixed solution was purified by centrifugation (11000 rpm, 15 min) three times to remove unlabeled ^99mTc. We determined the radiochemical purity of ^99mTc-ultraHOCls by paper chromatography with Xinhua No.1 filter paper and utilized saline as the mobile phase. The sheets were analyzed with a gamma counter (SN-684, Shanghai Hesuo Rihuan Photoelectric Instrument, China) and high-performance storage phosphor screen (Cyclone; Canberra-Packard, Ontario, Canada).

Fourier transform infrared spectroscopy (FTIR)

FTIR spectra were carried out with Thermo Nicolet AVATAR 360FTIR using the KBr pellet method.⁴⁹ Briefly, each sample (8−14 mg) was gently mixed with KBr (180−200 mg) to make a pellet under pressure (70−80 kN cm⁻²) for 3−5 min. FTIR spectra were recorded over a wavelength region of 400–4000 cm⁻¹ with a resolution of 2.0 cm⁻¹.

Transmission electron microscopy (TEM)

TEM, high-resolution TEM (HRTEM) images and selected area electron diffraction (SAED) patterns were recorded on a JEOL 2100F (JEOL) field emission. We spotted sample aliquots on a copper grid coated with amorphous carbon and removed the excess liquid by wicking it with filter paper. The grids were then air-dried at r.t. Images were captured using TEM operating at 300 kV and processed with the DigitalMicrograph software (Version 1.71.38). Freshly resected brain tissues were immersed into ice-cold glutaraldehyde fixing solution (special for the electron microscope, 2.5%, pH 7.2–7.4) at 4°C. Samples were stored at 4°C till analysis. The tissues were dehydrated, paraffin-embedded, sectioned (60–80 nm) and placed on 150 meshes cuprum grids with formvar film. TEM images were taken on a Hitachi TEM (HT7800, Hitachi, Japan).

Dynamic light scattering (DLS)

The size distributions and Zeta potential (ζ-potential) of nanoparticles were measured in disposable Zeta potential cells using Malvern Zetasizer (ZS90, Malvern Instruments Ltd., UK) at 25°C ζ-potential was calculated by the Zetasizer software (Version 7.11) using the Smoluchowski model.

Tissue penetration depth study

Fixed porcine muscles of 0–3.5 mm thickness were placed on the top of a solution containing ultraHOCls for the Raman penetration depth measurements. The Raman intensities of ultraHOCls were obtained by quantifying the peak 518 cm⁻¹.

Preparation of various radical species

The hypochlorite (OCl⁻) solution was delivered from a commercial NaOCl solution (4–5%). The H₂O₂ solution was delivered from a commercial solution (30%). The superoxide radical anion (O_2-⋅) solution was prepared by dissolving commercially available KO₂ in DMSO. We prepared the ⋅OH solution by the Fenton reaction from iron (II) chloride (FeCl₂) and H₂O₂. We obtained the peroxyl radicals (ROO⋅) solution by diluting commercially available tert-Butyl hydroperoxide (t-BuOOH) in water. Equal amounts of NaMoO₄ and H₂O₂ were mixed to yield ¹O₂ of 10 mM. The NO solution was generated from sodium nitroferricyanide-(III) dehydrate in water. Angeli’s salt (Na₂N₂O₃) served as the HNO source. NO₂⁻ was produced using NaNO₂. The peroxynitrite generation reaction was conducted in a mixed solvent of isopropyl alcohol (10–70%, v/v) and aqueous sodium hydroxide, containing 0.02–0.5 M alkyl nitrite, and 0.20–0.55 M H₂O₂. The incubation temperature was maintained at 25°C, with continuous stirring throughout the reaction.

Raman signal stability

We investigated the influence of storage time (0, 1, 2, 4, 8, 15, and 30 days) and the irradiation time (0, 1, 2, 5, 10, 30, 60, 120, 180, and 240 min) on the Raman signal stability of prepared ultraHOCls. The intensities of Raman bands at both 518 cm⁻¹ and 586 cm⁻¹ were measured at 4 random locations.

Radio-stability

We added 20 μL of ^99mTc-ultraHOCls (∼0.74 MBq) to 0.2 mL of PBS or mouse serum and incubated the mixed solution at 37°C for 12 h. The radio-stability of ^99mTc-ultraHOCls was determined at selected time points (0, 1, 4, 8, and 12 h) by paper chromatography.

In vitro microglia activation

The obtained microglial cells were seeded and cultured at 37°C. The media was supplemented with IL-4 (40 ng mL⁻¹), IFNγ (40 ng mL⁻¹), IFNγ (100 ng mL⁻¹), or IFNγ (100 ng mL⁻¹) plus LPS (600 ng mL⁻¹), respectively. After incubating for another 24–48 h, the activated microglial cells were harvested and used for further studies. The resting microglia were obtained without any treatment. Notably, a group of microglial cells activated with IFNγ (100 ng mL⁻¹) plus LPS (600 ng mL⁻¹) was treated with SHA (500 μM, 1 h) as an MPO-negative control.

Raman confocal studies

For cell samples, we acquired Raman spectra with a confocal Raman microscope (WITec Raman alpha300 R, WITec, Germany) operating in standard mode at −60°C. The Raman laser excitation source was a diode laser with a wavelength of 785 nm and the spectral resolution of the spectrometer is 4–5 cm⁻¹. Raman spectra were recorded with the following parameters: 300 mm spectrograph with 300 g/mm grating, an excitation power of 5 mW, 50 ✕ (0.75 NA, Zeiss) or 63 ✕ objectives (0.9 NA, Zeiss), a range of 0–3300 cm⁻¹. For Raman mapping, an area of 150 ✕ 400 to 700 ✕ 500 μm/map was pre-selected over the visible light image under the 63✕ objective with a step size of 1 μm. The acquired data were analyzed with the WITec Project software (Version 5.2) and LabSpec5 software (Version 5.58.25).

Toxicity

For in vitro toxicity evaluation, cells were cultured in 96-well plates at 37°C for 12–24 h for adherence to the plates. The cells were refreshed with a culture medium containing gradient concentrations of ultraHOCls (from 0 to 4 nM) and incubated for 24 h. Subsequently, 10 μL CCK-8 reagent was added and co-cultured for another 4 h. We measured the optical density with a Synergy HTX Multi-Mode Reader (HTX, BioTek Gen5, Vermont) at 450 nm. For in vivo toxicity evaluation, we intravenously injected ultraHOCls into the healthy mice at a dosage of 800 nmol/kg, while another group of four mice injected with an equal volume of PBS was used as control. The blood was collected at 28 days post-injection, and centrifugated to obtain the supernatant for biochemical analysis. The mice were euthanized at 60 days and 180 days after different treatments, then the major viscera were harvested and stained with hematoxylin and eosin (H&E).

In vitro cellular uptake studies

Primary microglial cells were seeded into 6-well-plates and incubated for 12–24 h for adherence to the plates. Then the media were supplemented with IFNγ (100 ng/mL) plus LPS (600 ng/mL) or isometric PBS. The media were refreshed with a culture medium containing ultraHOCls (0.4 nM) and incubated for 0.5, 2, 4 and 8 h. After washing with PBS three times, we collected the cells for Raman spectra acquisition.

In vitro BBB penetration studies

Initially, 5×10⁴–10×10⁴ endothelial bEnd.3 cells were seeded on the luminal surface of the transwell filter insert (Corning, 3460, 0.4 μm pore size). After incubating for 14 consecutive days, the tight junction formation was established when the trans-endothelial cell electrical resistance (TEER) values reached 190–210 Ω cm². To investigate the transport function, the apical wells were incubated at either 37°C or 4°C with ultraHOCls or uraHOCls in the absence or presence of inhibitors. To assess the effect of endocytic inhibition, the cells were pre-incubated with chlorpromazine (CPZ), cytochalasin D (CD) for 30 min, or methyl-β-cyclodextrin (MβCD) for 60 min, followed by treatment with ultraHOCls or uraHOCls at 37°C. Following a 4-h incubation period, the media in the basolateral well was gently mixed, and a 200 μL sample of the mixture was collected for Raman detection studies.

Neuronal survival evaluation

Fluoro Jade B (FJB) staining studies were performed to identify degenerated neurons. Briefly, the dewaxed paraffin sections (5 μm thickness) were rehydrated with an ethanol solution and potassium permanganate, followed by rinsing with ultrapure water. The sections were then stained with a solution containing 0.001% FJB. Images were captured using a META confocal laser scanning microscope (LSM 710, Carl Zeiss, Germany). We calculated the number of FJB-positive neuronal cells in the hippocampus. Nissl staining studies were performed to evaluate survived neurons. The 5-μm paraffin-embedded slices were de-waxed, hydrated, and then stained with 1% cresyl violet solution. After dehydrating with ethanol, washing with ultrapure water, and clearing with xylene, slices were sealed with neutral gum. We acquired the images with an optical microscope (Olympus BX51, Olympus, Japan) and counted the number of surviving intact hippocampal pyramidal cells.

In vivo SPECT/CT studies

KA models and sham models were used for SPECT/CT imaging after injecting intravenously ^99mTc-ultraHOCls (5.6–7.4 MBq) at 0.5, 2, and 12 h, respectively. Mice were anesthetized with 1–2% isoflurane during the scanning period. Small animal SPECT/CT scans were performed on a nanoScan@SC SPECT/CT system (Mediso Medical Imaging System, Hungary) equipped with four high-resolution conical collimators with multi-pinhole plates. Acquisition parameters of CT image: X-ray tube, 0.98 mA 50 kVp; exposure time, 300 ms; 720 projections in all. SPECT was acquired under the standard mode using the following parameters: energy peak, 140 keV; full width, 20%; resolution, 1 mm/pixel; matrix, 256 × 256; scan time, 30 s/projection; 128 projections in all. All acquired data were reconstructed and analyzed using dedicated fusion software (Version 3.00.021.000).

Bio-distribution studies

KA models and sham models were intravenously injected with ^99mTc-ultraHOCls (1.85–5.55 MBq). The mice were sacrificed at 0.5, 2 and 12 h post-administration and tissues of interest were excised and weighed. We measured the radioactivity with a gamma counter (SN-684, Shanghai Hesuo Rihuan Photoelectric Instrument Co., Ltd., Shanghai, China). The decay-corrected radioactivity was calculated and radioactivity uptake of ^99mTc-ultraHOCls was expressed as the percentage of the injected dose per gram of tissue (% ID/g).

EEG studies

For EEG recording studies, we anesthetized the mice and then implanted handmade stainless steel screw electrodes in the following epidural space: the recording electrode (anteroposterior = −1.0 mm, mediolateral = −2.5 mm), the reference electrode (anteroposterior = −2.0 mm, mediolateral = 1.8 mm), and the ground electrode (over the cerebellum). We fixed the electrodes on the skull with cyanoacrylate and dental acrylic cement. For mice that had received epilepsy surgery, the resection bed was filled with Kwik-Sil (World Precision Instruments, Sarasota, FL). The electrodes were implanted with a similar procedure as described above but implanting the recording electrode in the resection bed. The EEG and the synchronous video were continuously recorded using a PowerLab system (PL3516, AD Instruments, Australia). The raw EEG signals were collected and analyzed offline with LabChart (Version 8.1.19) and MATLAB software (Supplementary Code 1) using the following parameters: sampling rate, 1000 Hz; bandpass filters, 0.3–500 Hz; fast Fourier transform or wavelet transform for total power analysis, 0−50 Hz. Frequently, mice exhibited paroxysmal discharges characterized by high-voltage sharp waves (or burst of spikes) and no obvious behavioral alterations in the synchronous video recordings can be observed in these non-convulsive FSs. Occasionally, we observed tonic-clonic GSs with typical paroxysmal EEG activity and obvious post-seizure depression. Spontaneous seizure events were defined as regular spike clusters with a duration of ≥15 s, spike frequency of ≥2 Hz, and amplitude at least two times of the baseline EEG.

Cortical ECoG studies

The ECoG recording was operated with the g.Recorder system (g.tec medical engineering GmbH, Austria). An electrode array with 64 channels evenly distributed in a 1.1 ✕ 0.7 cm² area was made for ECoG monitoring according to the previous publication.³⁸ For ECoG recording, the mouse was anesthetized with 1–2% isoflurane and then underwent a craniotomy with stereotaxic apparatus to expose the cerebral cortex. The homemade electrode array was attached directly to the exposed cortex surface for real-time ECoG monitoring, and a piece of saline-soaked gelatin sponge was placed conformally on the array for better recording signals. The sampling frequency of original date was 4.8 kHz and transcoded to MAT formats at 1 kHz. The preprocessing and data analysis were conducted using custom-written MATLAB (Supplementary Code 2) scripts based on Fieldtrip Toolbox.⁵⁰ Each channel of the raw ECoG data were re-referenced to its adjacent channel on the same electrode shaft, so as to compute the bipolar re-reference. The re-referenced data were notch filtered to remove 50 Hz and its second harmonics as power line noise and then band-filtered to 1–30 Hz. The dynamics of brain activities within the first 10s after the seizure onset were examined in a temporal resolution of every half a second. Specifically, the mean ECoG signals of all channels within a window were normalized to 0–1, where the channel with the largest amplitude at this moment was labeled as 1 and the smallest one was 0. The peak time of a certain channel was defined as the moment of the maximum normalized amplitude. To examine the influence of anesthetic depth on Raman signals, we induced short-term anesthesia using 1–2% isoflurane, followed by tail vein catheterization. Propofol at an initial dose of 27 mg/kg was injected over 60 s, after which sevoflurane administration was discontinued. Different infusion rates of propofol (10.0, 5.0, and 2.5 mg/kg/min) were applied to maintain varying depths (high, medium and low, respectively) of anesthesia, ECoG recording and Raman imaging studies were conducted to evaluate the effects on Raman signals systematically.

Raman-guided epilepsy resection

The chronic seizure mouse models were injected intravenously via tail vein with ultraHOCls (100–150 nM/kg). After 6 h of intravenous injection, the mice were anesthetized with 1–2% isoflurane and underwent a craniotomy to expose the cerebral cortex for Raman imaging. The average HOCl level in each acquired Raman spectra was determined by quantifying the intensity ratio of I₅₈₆/I₅₁₈ followed by substitution into the fitting formula. Guided by the generated HOCl map in real-time, the neurosurgeon carefully performed the resection until no tissues with high I₅₈₆/I₅₁₈ ratios could be detected by changing the detection angles of the handheld scanner. To simulate the clinical situation, a group of mice was surgically guided with ECoG and the experience of the surgeon. Additionally, a group that received no brain resective surgery after craniotomy was used as a sham control. In order to ascertain the precision of margin delineation during surgical procedures, we conducted in vivo surgeries on KA mice using a double-blind protocol.

In vivo MRI studies

The MRI studies were performed on a 9.4 T small animal MR scanner (PharmaScan; Bruker Biospin, Billerica, MA) with a 23-mm-internal-diameter transmit/receive quadrature volume radiofrequency coil. T2W images were acquired using a turbo-rapid acquisition with relaxation enhancement (RARE) sequences with the following parameters: time of repetition/time of echo = 2500/33 ms, acquisition matrix = 196 ✕ 196; number of excitation = 4, field of view = 20 ✕ 20 mm; number of slices = 20; slice thickness = 0.8 mm, rare factor = 8. For the post-surgery mice, we imported the MRI data with 3D Slicer software for 3D visual fusion reconstruction and calculated the volumes of total resected brain tissues and resected hippocampal brain tissues.

In vivo PET/CT studies

For the ¹⁸F-FDG experiment, mice were fasted for more than 6−12 h before image acquisition. PET/CT images were acquired with an Inveon micro-PET/CT (Preclinical Solutions; Siemens Healthcare Molecular Imaging, Knoxville, TN). Each mouse received 7.4 ± 0.3 MBq ¹⁸F-FDG via the tail vein injection. Isoflurane (1−2%) in oxygen was used to anesthetize the mouse for the duration of imaging. The CT images were firstly acquired under the following parameters: 120 projections per bed position, 360° of rotation, 500 μA anode source transmitting 80 keV X-ray with an exposure time of 300 ms, X-ray detector of 3072 transaxial pixels ✕ 2048 axial pixels. 15-min static PET scans were then acquired in list mode. The PET images were reconstructed using a Fourier rebinning and ordered subsets expectation maximization 3D algorithm. Reconstructed computed tomography and PET images were fused and analyzed using Inveon Research Workplace software.

Behavioral analysis

All behavioral tests were performed in specially-equipped rooms under dim light. Prior to the test, animals were allowed to habituate in testing rooms for 1 h to minimize stress. All behavior experiments were ordered from the least to the most stressful to avoid carryover effects and conducted between 9:00 a.m. and 7:00 p.m. Between each trial, all apparatus was cleaned with a 75% alcohol solution to prevent the odors or footprint deposited by the previous mouse from influencing the performance of the following tested mouse. All assessments were conducted by an investigator blinded to the groups. For open field test, elevated plus maze (EPM) test, tail suspension test and Y-Maze test, the behavior was recorded by a camera (640✕480 pixels) (Sony Corporation, Japan), and analyzed using the behavioral tracking software EthoVision XT (Versoin 15.0.1418).

Accelerating rotarod test

The accelerating rotarod behavioral test was used for testing motor function (Yiyan Technology, China). The parameters used in the tests include start speed (1 rpm), acceleration (accelerate 0.5 rpm/s), and highest speed (40 rpm). Each mouse underwent four consecutive trials with a rest period of 30 min between each trial. The mean latency time to fall off the rotating rod for the last two trials was used for analysis.

Open field test

Mice were initially placed in one corner of the open field apparatus (40 cm length ✕ 40 cm width ✕ 35 cm height) (Noldus Information Technology, Netherlands) and recorded for 10 min. Total distance traveled, total velocity, velocity in the edge were used as measures of locomotor behavior. The entrance frequency, duration, and distance traveled in the center area (20 ✕ 20 cm) were used as measures of anxiety-like behavior.

EPM test

Mice have a natural tendency to actively explore a new environment, but fear open areas. The apparatus used for the EPM test consisted of two closed arms and two open arms (35 cm length ✕ 5 cm width ✕ 30 cm height) (Noldus Information Technology, Netherlands) joined perpendicularly and raised 1 m from the ground. Mice were individually placed in the center of the apparatus and were allowed to freely explore the EPM for 5 min. Quantification of total distance, total velocity, and the number of total entries were used for locomotor evaluation. The percentage of open-arm entries, the number of entries in open arms, and the average duration in the open arms were analyzed for evaluating anxiety-related behavior.

Tail suspension test

Briefly, mice were suspended by fixing their tail with medical tape from a hook connected with the horizontal steel wire. The mice were hung upside down approximately 15 cm above the ground covered with a soft sponge panel. The activity of the tested mice was recorded for 6 min using a video camera and the immobility time was scored for the last 4 min of the session.

Y-maze test

The theoretical basis of test is the innate tendency of rodents to alternate between exploring different arms of a Y-maze. The apparatus used for the test consisted of three closed arms [(35 cm (length) x 5 cm (width) x 25 cm (wall height)]. Mice were initially placed into the center of Y-maze and were allowed to freely explore the three arms of the maze for 5 min. Prefrontal cortex-dependent cognitive impairment was evaluated by quantification of the percentage of spontaneous alternation, spontaneous alternation, the number of incorrect alternations, the number of perseverations, the total number of arm entries, and total distance.

Catwalk gait analysis

Gait parameters of mice were collected and analyzed with the CatWalk XT system (CatWalk XT, Noldus Information Technology, Netherlands). Briefly, the glass floor of enclosed glass walkway (130 ✕ 7 cm) has a red backlight. When mice come in contact with the glass floor, green light is illuminated internally. A high-speed video camera was used to detect the green light when the mice traversed the tunnel freely to obtain footprint images. Testing was considered to be qualified if the animal crossed the walkway without stopping and at least 3–4 placements of each paw were recorded. Each mouse needed to complete at least 3 runs across the walkway. The definitions of gait parameters of interest were illustrated in Table S4.

Morris water maze test

The Morris water maze task was conducted in a round pool. Four different geometric figures were placed on the wall of the pool which evenly divided the pool into four quadrants. A platform with a 10-cm-diameter was placed in the center of the targeted quadrant. Water (20°C–24°C) was filled in the pool until the surface was 1–2 cm higher than the platform, and white, nontoxic paint was added to make it opaque. The start location was the opposite quadrant. Four training trials per day were conducted for six consecutive days before the experimental test for each group. In each trial, we recorded the time to reach the platform. If mice failed to find the platform within a maximal time limit of 60 s, the latency time was recorded as 60 s and the mice were gently guided and remained on the platform for additional 15 s. In the final test, the time that the mouse spent in each quadrant and track plots were recorded and analyzed with ANY-maze video tracking system (60000, Stoelting Co., Illinois).

Quantification and statistical analysis

All data acquired in this study are presented as the mean ± SD. from the indicated sample size. For all animal experiments, model mice were applied randomly. We performed statistical analyses and graphs processing using Prism (Version 7.00.) or OriginPro 2021b (Version 9.8.5.201). We performed a two-tailed unpaired Student’s t test for comparison of two groups, and one-way ANOVA followed by Tukey’s post-hoc test for multiple groups comparison. Values of p < 0.05 were considered statistically significant.

Published: May 30, 2025