Figure 1. Development of the Dual-NIR Switch. (a) Design of the odUCNPs. Schematic of the core–multishell architecture with spatially segregated lanthanide dopants (for example, Yb3+/Nd3+/Er3+ in the core and Yb3+/Tm3+ in an outer shell). (b) Orthogonal excitation pathways of odUCNPs. Under 980 nm NIR excitation, Yb3+ → Tm3+ upconversion in Shell 3 produces strong blue emission, while green emission from the core is largely suppressed. Under 808 nm NIR excitation, Nd3+ in Shell 1 harvests the photons and relays energy to the core, yielding intense green emission. (c) Energy-level diagram of the excitation and energy-transfer pathways. (d) TEM images acquired after successive shell-growth steps and corresponding normalized size distributions. Images share the same scale bar. (e) EDS elemental maps confirming the designed composition and dopant segregation. Images share the same scale bar. Comparison of mixed mUCNPs (f) and odUCNPs (g): photographs and emission spectra under 980 and 808 nm NIR excitation overlaid with SOUL activation (top) and deactivation (bottom) action spectra; gray shading highlights undesired crosstalk emission from mUCNPs under 980 nm NIR excitation, which is suppressed in odUCNPs. The two types of mUCNPs are NaYF4/Tm and NaYF4/Nd/Er. (h) In vivo workflow: delivery of SOUL-encoding AAV followed by local administration of odUCNPs in the targeted brain region. (i) Working principle of the Dual-NIR Switch: 980 nm-driven blue upconversion activates SOUL (ON state), producing sustained neuronal excitation (up to ∼30 min), whereas 808 nm-driven green upconversion deactivates SOUL (OFF state) and rapidly terminates excitation.#

研究团队首先在体外培养的表达SOUL的细胞中验证了光开关的功能。如图2所示，全细胞膜片钳记录显示，短时间的980纳米光照引发了明显的细胞膜去极化，而随后的808纳米光照迅速恢复了膜电位。实验进一步证实，纳米粒子表面经过谷胱甘肽修饰后能紧密粘附在细胞膜表面，相比未修饰的纳米粒子显著增强了局部光子通量，从而降低了所需的激发光功率密度。这一过程确保了光开关能够在较低的光剂量下实现高效的正交切换。

Fig.2 The Dual-NIR Switch modulates cell activities in vitro. (a) Patch-clamp configuration for SOUL-expressing HEK293T cells incubated with odUCNPs and stimulated upon sequential NIR illuminations. (b) Brightfield image of HEK293T cells during whole-cell patch-clamp recording. (c) Confocal image of SOUL–tdTomato expression in HEK293T cells. (d) SEM images of HEK293T cells after incubation, showing odUCNPs attached to the plasma membrane (arrows); right, magnified view. (e) Representative voltage traces from cells incubated with mixed mUCNPs or odUCNPs upon sequential NIR illumination at 980 nm (red) followed by 808 nm (orange). (f) Quantification of NIR-illumination-induced membrane potential changes under indicated conditions. Data are shown as mean ± s.e.m. (n = 16 cells for (+)/(+)/(+) and n = 8 cells for others, one-way ANOVA with Tukey’s multiple comparison test). (g) Quantification of NIR-illumination-induced membrane potential changes between mUCNPs and odUCNPs. Data are shown as mean ± s.e.m. (n = 9 cells for mixed mUCNPs and n = 16 cells for odUCNPs, t-test). (h) Representative voltage traces showing responses to varying durations of NIR illumination (1, 3, 5 s). (i) Onset/offset delay (left) and rate of NIR-illumination-induced membrane potential changes (right). Data are shown as mean ± s.e.m. (n = 16 cells). (j) Representative voltage traces at indicated NIR power densities of 5, 10, and 20 mW/mm2. (k) NIR-illumination-induced absolute membrane potential changes as a function of NIR power density (n = 15 cells, one-way ANOVA with Tukey’s multiple comparison test). (l) Representative current traces showing sustained activation after 980 nm NIR illumination (red arrow) and deactivation by 808 nm NIR illumination (orange arrows). (m) Normalized current decay after cessation of 980 nm NIR stimulation fitted with a single-exponential function (τ = 25.3 min). Unless otherwise stated, 980 and 808 nm NIR were applied at 20 mW/mm2, CW.#

随后，研究人员在自由活动小鼠的多个脑区验证了该系统在体无线调控行为的能力。如图3所示，在小鼠次级运动皮层注入光开关系统后，短暂的980纳米光照诱发了神经元高频放电，并显著增加了小鼠的运动速度、移动距离和单侧旋转行为；随后的808纳米光照则使运动指标迅速恢复至基线水平，展现出优异的秒级响应与重复调控能力。

Fig.3 Modulation of short-duration locomotion behavior in the M2 brain region by the Dual-NIR Switch. (a) Schematic of the injection timeline and in vivo electrophysiological recordings of M2 under modulation by the Dual-NIR Switch. (b) Confocal image showing SOUL expression in M2. (c) Bio-TEM images showing the distributions of odUCNPs in M2. Magnified views highlight clusters in the axon and soma. (d) Raster plots and corresponding firing rates in response to sequential NIR illumination at 980 nm followed by 808 nm under the indicated conditions. Each row of the raster plot represents a single trial (n = 6 mice, N = 18 trials). Both 980 and 808 nm NIR were applied at 24 mW/mm2, CW. (e) Quantification of the normalized firing rate within “During” period under the indicated conditions. Data are shown as mean ± s.e.m. (f) Quantification of the firing rate before 980 nm NIR illumination (Pre, 0–15 s), after 808 nm NIR illumination (Post, 65–80 s), and in between (During, 35–50 s). In (e) and (f), each point/point-set indicates a unit from one mouse averaged over three trials, n = 6 mice for each group, one-way ANOVA with Tukey’s multiple comparison test. Full set p-values are provided in Supporting Table S2. (g) Average firing rate dynamics under the indicated conditions. Data are shown as mean ± s.e.m. (h) Representative neuronal firing rate dynamics following cycles of neuromodulation by the Dual-NIR Switch. (i) and (j) Experimental timeline (i) and NIR stimulation pattern (j) for the locomotion enhancement test. 980 nm (at 15 s) and 808 nm (at 40 s) NIR lights were each applied for 5 s at 40 mW/mm2, CW. (k) Representative trajectory of a mouse before 980 nm NIR illumination (Pre, 0–15 s), in between (During, 15–45 s), and after 808 nm NIR illumination (Post, 45–60 s). (l) Kinetics of instant velocity (black) and cumulative distance (red) from a representative trial. (m)–(o) Changes in distance moved (m), mobility (n), and circling number (o) between “During” period and baseline (sum or average of “Pre” and “Post” periods), respectively. A circle was defined as a 180° counterclockwise rotation (around the body or arena center). Data are shown as mean ± s.e.m. (n = 12 mice for (+)/(+)/(+) and n = 6 mice for others, each mouse performed three trials, with data averaged across trials; one-way ANOVA with Tukey’s multiple comparison test).#

为了检验光开关在深部脑区进行分钟级调控的效果，研究团队将靶点设为外侧下丘脑区。如图4所示，激活外侧下丘脑的谷氨酸能神经元可抑制摄食行为。短暂的980纳米刺激使小鼠的进食量和在进食区停留的时间大幅减少，这种抑制作用在数分钟内保持稳定；在给予808纳米刺激后，小鼠的摄食行为迅速恢复正常。组织学检测表明，目标区域内表达SOUL的神经元中c-Fos阳性比例从5.3%显著提升至23.3%，证实了深部脑区回路的高效激活。

Fig.4 Modulation of medium-duration feeding behavior in a deep-brain region LHA by the Dual-NIR Switch. (a) Experimental scheme and circuit context for free-access feeding: SOUL and odUCNPs were targeted to LHA glutamatergic neurons, whose activation reduces food intake. (b) Arena schematic: the feeding zone (yellow dashed box) was defined as the region of interest (ROI). (c) Behavioral timeline: after habituation, mice were subjected to sequential NIR illumination at 980 nm followed by 808 nm (10 s each at 50 mW/mm2, CW). (d) Representative heatmaps of “During” period under the indicated conditions. (e) Quantification of relative food intake (food intake normalized to body weight; left) and time in ROI of “During” period (right). Data are shown as mean ± s.e.m. (f) Representative heatmaps of “Pre,” “During,” and “Post” periods of one mouse in the (+)/(+)/(+) group. (g) Quantification of the time spent in ROI of “Pre,” “During,” and “Post” periods. Each point set represents one mouse. In (e) and (g), n = 12 mice for (+)/(+)/(+) and n = 6 mice for others, one-way ANOVA with Tukey’s multiple comparison test. Full set p-values are provided in Supporting Table S2. (h) c-Fos immunostaining in the LHA to validate neuronal activation; SOUL (red), c-Fos (green), and merge. Images share one scale bar. (i) Quantification of the density of c-Fos+ cells under the indicated conditions (one-way ANOVA with Tukey’s multiple comparison test). (j) Specificity of activation: percentage of SOUL-expressing neurons that were also c-Fos+ (unpaired t-test). In (i) and (j), data are shown as mean ± s.e.m. Each point represents the average of fluorescence measurements from three brain slices, n = 5 mice in the (+)/(+)/(+) group and n = 3 for others.#

研究团队还评估了该系统在更长时间跨度下的行为调控潜力。如图5所示，在腹侧顶盖区条件位置偏好实验中，研究人员在训练阶段给予小鼠30秒的周期性光照刺激。结果显示，光开关激发的奖励效应使小鼠在测试阶段对训练隔间的停留时间从365.9秒显著增加至562.4秒，成功建立了长达亚小时尺度的上下文奖励记忆关联，且活性标记c-Fos阳性率提升至39.8%。

Fig.5 Modulation of long-duration reward conditioning behavior in a deep-brain region VTA by the Dual-NIR Switch. (a) Experimental scheme (top) and confocal validation of SOUL expression in the VTA (bottom). (b) Schematic of the#

CPP apparatus: compartment B is conditioned by sequential NIR illumination at 980 nm followed by 808 nm. The mid compartment separates compartments A and B and can be gated. (c) Timeline of the CPP test: in the Pre- and Post-tests, the mice freely moved in all compartments. During conditioning, mice were confined to compartment B and received sequential NIR illumination. Both 980 and 808 nm NIR illumination were applied for 30 s at a power density of 60 mW/mm2, with a pulse duration of 0.1 s and a repetition rate of 5 Hz. (d) Representative trajectories (left) and the corresponding heatmaps (right) for a mouse in the (+)/(+)/(+) group during the Pre-test and Post-test. (e) Quantification of time spent in compartment B during the Pre- and Post-tests under the indicated conditions (box plots: min–max with interquartile range; n = 6 mice per group; paired t-test; full set p-values are provided in Supporting Table S2). (f) CPP scores under the indicated conditions. Data are shown as mean ± s.e.m. (n = 6 mice per group, one-way ANOVA with Tukey’s multiple comparison test). (g) c-Fos immunostaining in the VTA to validate neuronal activation; SOUL (red), c-Fos (green), and merge. Images share one scale bar. (h) Quantification of the density of c-Fos+ cells under the indicated conditions (one-way ANOVA with Tukey’s multiple comparison test). (i) Specificity of activation: percentage of SOUL-expressing neurons that were also c-Fos+ (unpaired t-test). In (h) and (i), data are shown as mean ± s.e.m. Each point represents the average of fluorescence measurements from three brain slices, n = 5 mice in the (+)/(+)/(+) group and n = 3 for others.

最后，研究人员全面评估了双波长近红外光开关的生物安全性。如图6所示，电感耦合等离子体质谱与组织成像分析表明，注射入脑内的纳米粒子会在数周内逐步代谢清除，6周后残留量仅剩5.5%左右。由于调控过程仅需短暂光照，小鼠颅内温度保持在安全范围内，未引起明显的组织发热。行为学测试证实，该处理不会损害小鼠的运动协调性与认知记忆功能，免疫荧光染色也未观察到明显的神经元丢失或胶质细胞异常激活，展示出良好的生物相容性。

Fig.6 Biocompatibility assessments of the Dual-NIR Switch. (a) Treatment paradigms: injection of odUCNPs and NIR illumination patterns (delivered at 80 mW/mm2, CW). (b) Timeline for biocompatibility assessments: treatments followed by behavioral tests and IHC. (c) Representative open-field trajectories (central zone outlined). (d) Quantification of distance moved (left) and midzone entries (right). Data are shown as mean ± s.e.m. (n = 5 mice in each group, one-way ANOVA with Tukey’s multiple comparison test). (e) Rotarod test schematic and timeline. (f) Quantification of rotarod performance (falling speed and falling latency). Data are shown as mean ± s.e.m. (n = 5 mice per group, two-way ANOVA with Tukey’s multiple comparison test). (g) Contextual fear conditioning test schematic and timeline. (h) Quantification of freezing time during fear recall and in a new context. Data are shown as mean ± s.e.m. (n = 5 mice/group, one-way ANOVA with Tukey’s multiple comparison test). (i) Representative M2 immunofluorescence 1 week after treatment: NeuN (neurons), Iba1 (microglia), GFAP (astrocytes), and DAPI (nuclei). Images share one scale bar. (j****Fig.6 ) Quantification of marker-positive percentage (NeuN+/Iba1+/GFAP+ of DAPI+, left column) and cell densities. Data are shown as mean ± s.e.m. (n = 9 mice per group, one-way ANOVA with Tukey’s multiple comparison test).#