Mechanobiological Modulation of Blood-Brain Barrier Permeability by Laser Stimulation of Endothelial-Targeted Nanoparticles

Abstract

The blood–brain barrier (BBB) maintains an optimal environment for brain homeostasis but excludes most therapeutics from entering the brain. Strategies that reversibly increase BBB permeability are essential for treating brain diseases and are the focus of significant preclinical and translational interest. Picosecond laser excitation of tight junction-targeted gold nanoparticles (AuNPs) generates a nanoscale mechanical perturbation and induces a graded and reversible increase in BBB permeability (OptoBBB). Here we advanced this technique by showing that targeting endothelial glycoproteins leads to >10-fold higher targeting efficiency than targeting tight junctions both in vitro and in vivo. With both tight-junction and glycoprotein targeting, we demonstrate that OptoBBB is associated with a transient elevation and propagation of Ca²⁺, actin polymerization, and phosphorylation of ERK1/2 (extracellular signal-regulated protein kinase). These collectively activate the cytoskeleton resulting in increased paracellular permeability. The Ca²⁺ response involves internal Ca²⁺ depletion and Ca²⁺ influx with contributions from mechanosensitive ion channels (TRPV4, Piezo1). We provide insight into how the excitation of tight junction protein (JAM-A)-targeted and endothelial (glycocalyx)-targeted AuNPs leads to similar mechanobiological modulation of BBB permeability while targeting the glycocalyx significantly improves the nanoparticle accumulation in the brain. The results will be critical for guiding the future development of this technology for brain disease treatment.

Introduction

The vast brain neural network and supporting glial cells are protected from circulating neurotoxins by a tightly regulated blood–brain barrier (BBB).¹ The BBB is formed by tight junctions (TJ) and adherens junctions associated protein complexes, severely restricting paracellular diffusion.² That, coupled with low endothelial transcytosis,³ results in the highly restricted movement of molecules from the circulation to the brain interstitium. While the BBB plays a protective role for the brain, it also represents a significant barrier to drug delivery. Overcoming the BBB to facilitate brain drug delivery is critical for treating central nervous system (CNS) diseases and has drawn significant interest.^4-7 Our recent work has demonstrated that transcranial picosecond-laser stimulation of TJ-targeted gold nanoparticles (AuNPs) can temporarily increase BBB permeability by paracellular diffusion, referred to as OptoBBB.⁸ OptoBBB further allows the delivery of various therapeutics, including human IgG, viral vector, and liposomes, into the brain in vivo. Thus, OptoBBB provides a promising avenue for improving brain drug delivery.

There are several important questions that need to be addressed in order to move the OptoBBB method toward clinical translation. First, while TJ targeting allows anchoring the nanoparticles on the vessel wall, the targeting efficiency is low due to its limited distribution on cell boundaries. Glycoprotein is one of the critical components of endothelial glycocalyx, a dense and brush-like structure on the luminal surface of the BBB.^9,10 Therefore, glycoprotein may be a more effective target for OptoBBB due to the significantly higher coverage on the luminal surface compared with the TJ proteins. Importantly, the Lycopersicon esculentum lectin (LEL) can efficiently label the cerebral vessel wall by binding to glycoprotein.^8,11 Second, it remains unclear which cellular responses are involved in increasing the BBB permeability. Picosecond-laser irradiation of AuNPs causes pressure generation due to thermoelastic expansion of AuNPs,^12,13 known as the photoacoustic effect. The resulting mechanical wave can lead to several cellular responses, including G-actin polymerization,¹⁴ and activation of mechanosensitive channels that increase intracellular calcium (Ca²⁺).^15,16 Ca²⁺ is an important second messenger and plays a crucial role in the signaling pathways that regulate endothelial permeability.¹⁷ For example, activating Ca²⁺-sensitive signaling can increase BBB permeability by several pathways, including ERK1/2 phosphorylation and the resulting activation of actomyosin filament contraction.¹⁸ Therefore, elucidating the relevant mechanobiological signaling pathways would help clarify the mechanism for OptoBBB and guide future clinical translation.

In this study, we found that glycoprotein-targeting AuNPs have more than 10-fold higher accumulation on the BBB than TJ-targeting AuNPs (21-fold in vitro and 12-fold in vivo). With TJ and glycoprotein targeting, we show that optoBBB is associated with endothelial Ca²⁺ signaling, and pharmacological inhibition of Ca²⁺ signaling blocks the BBB permeability change, an accepted surrogate of the BBB regulation. Importantly, our results suggest that Ca²⁺ signals can propagate among endothelial cells and thus extend and coordinately regulate BBB opening. Further investigation of the Ca²⁺ signaling revealed the contribution of both internal Ca²⁺ release via inositol 1,4,5-trisphosphate (IP3) signaling and external Ca²⁺ influx with contributions from mechanosensitive ion channels (TRPV4, Piezo1). Ca²⁺ signaling further leads to the phosphorylation of ERK1/2 and actin polymerization, which leads to an activation of the cytoskeleton resulting in an increase in paracellular permeability. In summary, our study demonstrates that targeting glycocalyx can significantly increase the brain accumulation efficiency of the AuNPs compared to targeting the TJ protein and elucidates a similar mechanobiological mechanism of OptoBBB exploiting both targets. Our results establish a solid basis for the future development of this technology and CNS disease treatment.

Experimental details

Materials

Anti-JAM-A antibodies BV16 and BV11 were provided by Dr. Monica Giannotta at the FIRC Institute of Molecular Oncology Foundation. Lycopersicon esculentum (Tomato) Lectin (LEL, L-1170-2) was purchased from Vector Laboratories. Human cerebral microvessel endothelial cell/D3 cell line, EndoGRO^™-MV Complete Media Kit, FGF-2, trypsin-EDTA, collagen type I, and millicell ERS-2 System were purchased from Millipore. LF PVDF transfer kit stain-free protein gel and western ECL substrate were purchased from Bio-Rad. P-p44/42 ERK1/2 and p44/42 ERK1/2 were purchased from Cell Signaling Technology. Gold(iii) chloride, FITC-dextran, DMSO, hydroquinone, sodium citrate tribasic, BSA, Tween 20, Triton-X 100, sodium carbonate, sodium bicarbonate, phalloidin-RF, QuantiPro^™ BCA Assay kit (QPBCA), and sucrose were purchased from Sigma-Aldrich. OPSS-PEG-SVA and mPEG-SH were purchased from Laysan Bio, Inc. Biotin-PEG-SH was purchased from NANOCS. Penicillin–streptomycin and donkey anti-mouse IgG (H + L) 488, RIPA buffer, protease inhibitor cocktail, phosphatase inhibitor cocktail, BCA protein assay kit, fluo-4 AM, Cy3-labeled streptavidin, peroxidase-conjugated affiniPure goat anti-rabbit IgG (H + L), donkey serum, goat serum, Trypan blue, gold reference standard solution, WST-1 kit, U0126, GSK2193874, GsMTx4, BAPTA-AM, fluo-4, Hoechst dye 33342, Dulbecco's phosphate-buffered saline, 20 kDa dialysis membrane, LI-silver enhancement kit, 6-, 24-, 96-well plates, and transwell inserts were purchased from Thermo Fisher Scientific. All other chemicals were analytical grade. Adult mice were ordered from Charles River Laboratories. All animal procedures were performed in accordance with the Guidelines for Care and Use of Laboratory Animals of The University of Texas at Dallas and approved by the Institutional Animal Care and Use Committee (IACUC) committee.

Formation and characterization of cellular monolayers

Human cerebral microvessel endothelial cell/D3 (hCMEC/D3) cells were cultured in EndoGRO^™-MV Complete Media supplemented with FGF-2 on a collagen-coated porous membrane. The D3 cells were seeded in transwell inserts placed in a 24-well plate and cultivated for approximately 6 to 7 days to form cellular monolayers at 37 °C and 5% CO₂. The hCMEC/D3 monolayers were characterized by trans-endothelial electrical resistance (TEER), permeability, and immunocytochemistry (ICC) staining.

Transendothelial electrical resistance (TEER) measurement

The electrical resistance of hCMEC/D3 monolayers was measured in Ohms (Ω). A cellular monolayer was formed on transwell inserts (0.3 cm², pore size 8 μm), including upper and bottom compartments. For TEER measurement, cellular monolayers were detected by an epithelial voltmeter (Millicell ERS-2, Millipore, USA). The electrodes were rinsed with cell culture medium between the blank and sample well. The TEER values of cellular monolayers were measured before the laser (−0.5 hours) and at various time delays after laser irradiation (0, 0.5, 1, 1.5, 3, and 6 hours). The resistance value of blank culture inserts coated with collagen on the top side of the membrane was used as the baseline. To obtain the TEER, the baseline value was subtracted from the resistance measured from the cell monolayer samples. The resulting resistance value multiplied by the effective membrane area gives the TEER value in Ω cm².

Permeability measurement

All the medium of cellular monolayers was replaced by a cell culture medium without 5% FBS from the top and chambers, followed by 30 minutes of incubation at 37 °C and 5% CO₂. After that, a medium mixed with 0.3 mL FITC-dextran (40 kDa, 1 mg mL⁻¹) was added to the upper compartments. At particular time points, 100 μL of the medium was aspirated from the bottom well and added to 96 black well plates. 100 μL of culture medium was then replaced in the bottom chamber to keep the volume constant. We then measured the fluorescent intensity for the collected samples in the 96-well plate (excitation at 490 nm and emission at 540 nm) to obtain the FITC-dextran concentration. The quantity ($Q$) was obtained by multiplying the concentration and volume. The apparent permeability ($P_{\text{app}},\text{cm}{\mathrm{s}}^{-1}$) was calculated by the rate of FITC-dextran quantity change over time ($\mathrm{d}Q∕\mathrm{d}t$), divided by the initial concentration of dextran ($C$) and the membrane area ($A$).

$$P_{\text{app}}=\frac{\frac{\mathrm{d}Q}{\mathrm{d}t}\times 1}{A\times C}$$

Immunocytochemistry (ICC) staining

The JAM-A expression on hCMEC/D3 cells was characterized by ICC staining. Cellular monolayers were fixed for 5 minutes in pure methanol on ice and then washed in PBS 3 times on a shaker. A blocking buffer (5% donkey serum, 2% BSA in PBS, and 0.05% Tween) was applied at room temperature (RT) for 1 hour. Then the cells were incubated with primary antibodies, mouse anti-human JAM-A (BV16: 3–5 μg mL⁻¹), overnight at 4 °C or RT for 1 hour, followed by a second antibody incubation at RT for 1 hour. Finally, the samples were incubated with Hoechst dye for 10 minutes in the dark to stain the nuclei. The samples were washed in PBS every time before changing reagents.

For F-actin staining, all samples were fixed with 4% paraformaldehyde (PFA) for 10 minutes at 4 °C. After washing with PBS, treated monolayers were incubated with phalloidin–RFP (diluted ratio: 1 : 1000) at RT for 1 hour, followed by incubation of Hoechst dye. The monolayers were then mounted on glass slides after washing.

To compare the distribution of AuNP–BV16 and AuNP–LEL on hCMEC/D3 cells, AuNP–BV16 and AuNP–LEL were backfilled by biotin–PEG–SH (PG2-BNTH-1K, NANOCS) to stabilize the nanoparticles. AuNP–BV16–biotin and AuNP–LEL–biotin were incubated with monolayers for 0.5 hours at 37 °C. After washing with PBS 3 times, the treated monolayers were fixed with 4% PFA for 10 minutes at 4 °C. Cy3-labeled streptavidin (1 : 200) was used to detect biotin and thus, the distribution of targeting AuNPs. Confocal microscopy (FV3000RS or SD-OSR) was utilized to take fluorescent images after ICC staining.

Gold nanoparticle conjugation

AuNPs were synthesized following a previously reported method.¹¹ BV16 (anti-JAM-A antibody) was diluted to 0.5 mg mL⁻¹ in PBS, followed by dilution in aqueous 10 mM NaHCO₃ at pH 8.5. OPSS–PEG–NHS was dissolved in NaHCO₃ and quickly added to the diluted antibody at a 125 : 1 molar ratio. The mixture was vortexed briefly and kept shaking on ice for 3 hours, followed by dialysis to remove free OPSS–PEG–NHS through a 20 kDa MW membrane. The thiolated BV16 was reacted with concentrated AuNPs at a molar ratio of 270 : 1 for 1 hour on ice. Polyethylene glycol (PEG) was added at 6 PEG nm⁻³ to backfill¹² AuNPs and kept on ice for 1 hour to stabilize the particles. Finally, the modified AuNPs were washed 3 times and characterized by dynamic light scattering (DLS) and UV-vis spectroscopy. For AuNP conjugation with antibody BV11 and LEL, NaHCO₃ buffer was replaced by 2 mM borate buffer at pH 8.5. The number of proteins per AuNP was determined using a QuantiPro^™ BCA Assay kit. Briefly, the protein standard solution (0–30 μg mL⁻¹) was prepared in 2 mM borate buffer and the BCA working reagent. To measure the added protein amount, the OPSS–PEG–protein conjugates were diluted 10 times with borate buffer. To measure the remaining protein, the AuNP–protein conjugates were centrifuged at 5000g for 10 min, and the supernatant was collected for dialysis (6 kDa MW membrane, overnight) to remove the excess amount of reducing agents in the supernatant. The samples were mixed with BCA working reagent in 1 : 1 ratio, and then 300 μL of each standard and samples were placed into a 96-well plate, followed by incubating at 37 °C for 2 hours. Then the plate was cooled to room temperature and the absorption was measured at 562 nm. The total number of proteins on the AuNPs was calculated as the difference between the number of proteins added to the AuNP suspension and the number of proteins left in the supernatant. With a known AuNP number, the number of proteins per AuNP can be estimated. The distribution of AuNP modified by anti-JAM-A antibodies and LEL on monolayers was performed by ICC staining to confirm AuNP targeting.

OptoBBB in vitro

For the BBB transient opening in vitro, hCMEC/D3 cells were seeded on transwell inserts for about 1 week to form monolayers with a TEER value of around 60 Ω cm², P_app of 8.4 ± 1.2 × 10⁻⁷ cm s⁻¹ for 40 kDa FITC-dextran. The TEER value is highly dependent on the cell type composition and the properties of the transwell inserts (e.g., the surface area of the porous membrane and the pore size). Our TEER value is consistent with literature using the same cell line and transwell insert.¹⁹ The expression of JAM-A on D3 cells was confirmed by ICC staining. The culture was incubated with AuNP–BV16 or AuNP–LEL for 0.5 hours at 37 °C and 5% CO₂. The treated cellular monolayer was washed three times with PBS before a 28-picosecond laser (532 nm) was applied to the cellular monolayer. The BBB opening in vitro was characterized by TEER and permeability measurements.

WST-1 assay

The hCMEC/D3 cells were seeded in 96-well plates in 100 μL of culture medium for about 1 week to form monolayers. The electron mediator solution and developer reagent were mixed in equal volumes. 10 μL WST-1 mixture was added to monolayers, followed by 2 hours of incubation at 37 °C and 5% CO₂. After incubation, the 96-well plates were wrapped in foil and shaken for 1 minute at room temperature. Finally, the absorbance of samples was detected at a wavelength of 450 nm with a microplate reader (Synergy2, BioTek).

AuNP biodistribution in vivo

Briefly, the mice were intravenously administered AuNP–BV11 or AuNP–LEL, or the control group AuNP–PEG, with a dose of 18.5 μg g⁻¹. At 1 hour, we performed cardiac perfusion PBS and collected the brain tissue. The tissue was then digested in fresh-made aqua regia and centrifuged at 5000g for 5 minutes to collect the supernatant. Then the gold concentration was measured by Inductively Coupled Plasma Mass Spectrometry (ICP-MS). Moreover, the AuNP accumulation was visualized by silver enhancement staining. Briefly, 5 μm thickness brain slices were stained with LI silver enhancement kit for 20 min, followed by rinsing in water.

OptoBBB in vivo

OptoBBB in vivo was performed as described previously.⁸ The mice were briefly anesthetized and intravenously administered AuNP–BV11 (targeting JAM-A) or AuNP–LEL (targeting glycoprotein) with a dose of 18.5 μg g⁻¹. Then, the scalp was carefully removed to expose the skull, followed by transcranial picosecond laser stimulation (1 pulse, 28 picosecond pulse duration, 6 mm beam diameter). The molecular tracer Evans blue (66 kDa/albumin-bound, 2% in PBS, 100 μL) was injected intravenously 5 minutes before the laser. After 0.5 hours, transcardial perfusion was performed with 25 mL PBS and 4% PFA. The brains were extracted to visualize the extravasation of Evans blue and then snap-freeze on dry ice, followed by cryo-sectioned to 30 μm thickness slices. Fluorescent images of Evans blue dye extravasation were captured with Slide Scanner Microscopy (Olympus VS120). Evans blue leakage was quantified by total fluorescent volume (total fluorescent area×thickness) using Image-J.

Ca²⁺ signal detection

The fluorescent change of fluo-4 represents the cytosolic Ca²⁺ concentration change. The fluorescent intensity of fluo-4 was detected by an inverted microscope (IX 73, Olympus) combined with a fluorescent illumination system and HCImage Live software. The illumination is from X-Cite 110LED Illumination System (Lumen Dynamics Group Inc.). Fluo-4 was excited at 475 nm, and the emission was collected at 509 nm. All the videos were recorded with the same illumination conditions. The images were captured at 1 frame per second with a digital camera (Hamamatsu ORCA-Flash4.0 LT C11440). For Ca²⁺ signal detection, D3 cells were first seeded on transwell inserts to form monolayers, as described above. Monolayers were then incubated with AuNP–BV16 or AuNP–LEL at desired concentration for 0.5 hours and then with 3 μM fluo-4 indicators after washing away unbound AuNPs. The inserts were then placed in the 35 mm dishes. Next, the picosecond laser with a focused beam was aligned with the 10× objective. Lastly, the intensity of fluo-4 was read about 30–50 seconds before laser stimulation as a baseline and about 120–180 seconds after laser excitation, while the camera was blocked from the laser during laser excitation.

Ca²⁺ data analysis

The Ca²⁺ imaging data were analyzed using custom Python (https://www.python.org/) code implemented inside a Jupyter notebook.²⁰ Images were processed and analyzed using the scikit-image library.²¹ The analysis also made use of the NumPy and pandas libraries.^22,23

First, regions of interest corresponding to individual cells were identified by an iterative image segmentation process. Starting with the image three frames before the laser was applied, every other consecutive image in the data set was segmented to identify a set of cell regions of interest at that time point corresponding to increases in the fluorescence intensity over the background noise and the resting fluorescence levels before application of the laser. The segmentation procedure consisted of identifying local maxima in the fluorescence intensity after subtracting the resting fluorescence and computing an elevation map using a Sobel filter; included local maxima were required to be at least 20 pixels (12.8 μm) from other maxima. The local maxima were used as markers along with the edges detected by the Sobel filter to segment the image using a watershed algorithm. Then distinct regions were filled and labeled to identify the cell regions of interest. Small regions with an area of less than 50 pixels were then pruned from the set to avoid including regions significantly smaller than the expected cell size. At each time point, the identified regions were merged with the set of regions identified at the previous time point to iteratively update the set of regions for cells as they are activated over time. When merging the sets of regions, any two regions with a centroid less than 20 pixels (12.8 μm) from one another were combined into a single region, assuming they are highly overlapped and thus correspond to the same cell. The final set of regions was then further filtered based on several region properties to remove any spurious regions, including the removal of regions with area less than 100 pixels or greater than 750 pixels, major or minor axis width less than 8 pixels, eccentricity greater than 0.9, or extent less 0.5.

Next, the change of fluorescence ($\Delta F∕F$) of fluo-4 for each identified cell region was computed by the change in fluorescence, $\Delta F$ (the intensity at a time $t$, $F_t$, minus the baseline intensity, $F$), divided by the baseline intensity $F$. The image domain was then divided radially into bins of width 50 μm with origin at the center of the laser spot, and the $\Delta F∕F$ trace of cells within each radial bin was averaged to estimate the Ca²⁺ signal as a function of distance from the center of the laser spot.

$$\text{Fluorescence}\text{change}=\frac{F_t-F}{F}=\frac{\Delta F}{F}$$

Western blot

To study the ERK1/2 phosphorylation (p-ERK1/2), we incubated monolayers with targeting AuNPs for 0.5 hours, followed by laser stimulation. We then performed total protein extraction at various time points. For the blocker U0126 (inhibitor of p-ERK1/2) application, the monolayers were pretreated with 10 μM U0126 for 1 hour before AuNP incubation. Without AuNP incubation, proteins were collected at 0.5 hours after laser stimulation as only laser control. We followed the manufacturer's instructions to extract proteins. Briefly, total proteins from hCMEC/D3 cells were extracted using RIPA buffer (89900, Fisher Scientific) following the manufacturer's instructions. D3 cells were washed twice with cold PBS, followed by incubation of cold RIPA buffer mixed with a protease inhibitor cocktail (78410, Fisher Scientific) and a phosphatase inhibitor cocktail (78420, Fisher Scientific) for 5 minutes on ice. The lysate was centrifuged to collect the supernatant. Total protein concentration was determined by the BCA assay kit (23225, Fisher Scientific). The exact amounts of proteins (10 μg) were loaded to stain-free protein gel (5678093, Bio-Rad) and separated with electrophoresis under 110 V. The proteins were then transferred to the LF PVDF membrane (1704275, Bio-Rad) with the trans-Blot Turbo system (Bio-Rad), followed by blocking with 5% BSA in TBS buffer at room temperature. After 1 hour, the membranes were incubated with primary antibodies at 4 °C overnight. After washing the primary antibodies with TTBS buffer, corresponding secondary antibodies were applied at room temperature for 1 hour. The blots were enhanced with chemiluminescence (1705060, Bio-Rad), then visualized with ChemiDoc Touch Imaging System (Bio-Rad). The immunoblots were analyzed using ImageLab (Bio-Rad).

Statistical analysis

Statistical analyses were performed using OriginPro 2020. All experiments were repeated a minimum of three times. Unpaired t-test was performed for statistical analysis. Data are means ± SD, where applicable *p < 0.05, **p < 0.001, or ***p < 0.0001 was considered a statistically significant difference. The specific n value per group is provided in the figure captions.

Results

Glycoprotein targeting leads to significantly improved AuNP accumulation on BBB

First, we tested the hypothesis that glycoprotein is a more effective target for the vasculature due to its abundant expression on the luminal surface of the microvasculature relative to TJ proteins. We exploited a transwell model using human cerebral microvascular endothelial (hCMEC) D3 cells^24,25 for our study. We validated the model by characterizing the trans-endothelial electrical resistance (TEER), permeability, and TJ formation (Fig. 1a and Fig. S1).¹⁹ We synthesized 50 nm spherical AuNPs, and functionalized the TJ targeting AuNPs with anti-JAM-A antibody (AuNP–BV16, BV16: mouse anti-human JAM-A antibody) and glycoprotein-targeting AuNPs with lectin (AuNP–LEL) (Fig. S2), both were subsequently backfilled with polyethylene glycol (PEG).²⁶ When incubating AuNP–BV16 and AuNP–LEL with D3 monolayers using the same concentration (0.5 nM), a significantly higher amount of AuNP–LEL binds to the cell surface than AuNP–BV16, as indicated by the pink color differences after washing with PBS to remove free AuNPs (Fig. 1b). The immunocytochemistry (ICC) staining confirmed the result (Fig. 1c), and it was found that AuNP–LEL (0.5 nM) displays a 21-fold higher accumulation on the D3 cells than AuNP–BV16 at the same concentration (Fig. 1d). The protein number per AuNP was quantified using a QuantiPro^™ BCA Assay Kit. The results showed that the number of BV16 or LEL per AuNP was 66 ± 5 and 73 ± 8, respectively. Therefore, the higher accumulation efficiency of AuNP–LEL compared to AuNP–BV16 was primarily due to more available glycoprotein than JAM-A protein for nanoparticle targeting, rather than more protein on the nanoparticle surface. We then selected a much lower AuNP–LEL dose (0.02 nM) for the following experiments as it shows a similar gold accumulation to 0.5 nM AuNP–BV16.

Fig. 1. Laser stimulation of endothelial-targeted nanoparticles leads to reversible BBB permeability change (OptoBBB).

(a) Schematic of OptoBBB in vitro. Red dots indicate the molecules that are impermeable to the intact BBB and become permeable to the leaking BBB after laser stimulation. (b) Images of the well plate after 0.5 nM AuNP–LEL and AuNP–BV16 incubation with hCMEC/D3 monolayers. (c) Distribution of AuNP–LEL and AuNP–BV16 (red) on hCMEC/D3 monolayers using ICC staining. Blue: nuclei. (d) Quantitative gold accumulation of AuNP–LEL and AuNP–BV16 on D3 monolayers using inductively coupled plasma mass spectrometry (ICP-MS), n = 6. (e) Normalized TEER change over time by laser stimulation of AuNP–LEL (0.02 nM) and AuNP–BV16 (0.5 nM), n = 3. (f) Comparison of TEER change at 0.5 hours after laser irradiation of different targeting AuNPs, n = 3. (g) Comparison of permeability change after laser irradiation of different targeting AuNPs, n = 3. 35 mJ cm⁻², 5 pulses, 5 Hz. Data expressed as mean ± SD. Unpaired t-test was performed individually between Only laser and the other groups. *: P < 0.05, or **: P < 0.01, was considered a statistically significant difference. Scale bar: 20 μm.

We then compared the efficiency of OptoBBB with AuNP–BV16 and AuNP–LEL. We tested different conditions for BBB opening in vitro using AuNP–BV16 and selected an optimized combination (35 mJ cm⁻², 5 pulses, 5 Hz, and 50 nm AuNPs) for the following experiments to achieve a high BBB opening efficiency (indicated by larger drops in the TEER and increases in permeability, respectively) with minimal cell injury (Fig. S3-S7). We used a 532 nm picosecond laser since the surface plasmon resonance peak of 50 nm AuNPs matches well with the laser wavelength. Importantly, our results showed that only laser excitation of AuNP–BV16 could cause the TEER drop, compared to free antibody BV16 or AuNP–PEG as controls (Fig. S5a). Under the same laser irradiation, 20-fold different AuNP doses (0.02 nM AuNP–LEL and 0.5 nM AuNP–BV16) showed comparable TEER and permeability changes after OptoBBB (Fig. 1e-g) without causing significant cell injury (35 mJ cm⁻², 5 pulses, 5 Hz, Fig. S8). The TEER value returned to the baseline at 6 hours, indicating a reversible BBB opening. Furthermore, we tested the BBB modulation in vivo with these two targets. The gold element analysis shows that AuNP–LEL accumulation was 12-fold higher in the brain than TJ-targeting AuNPs (AuNP–BV11, BV11: rat anti-mouse JAM-A antibody) (Fig. S9a), while the total BV11 antibodies per AuNP was 68 ± 6, similar to the total LEL per AuNP (73 ± 8). The total gold content found in mice by the percentage of injection dose (%ID) was 5 ± 2% (AuNP–PEG), 53 ± 19% (AuNP–BV11), and 45 ± 9% (AuNP–LEL), respectively (Fig. S9b). In particular, the reduced accumulation of AuNP–LEL in the liver could be attributed to lower glycocalyx coverage in this organ.²⁷ We used silver enhancement staining to visualize the AuNP–LEL accumulation in the brain blood vessels, while no targeting of AuNP–PEG was observed (Fig. S9c). The in vivo optoBBB results showed that under the same AuNP dose and same laser pulse energy, AuNP–LEL leads to a more efficient BBB opening, indicated by a larger volume of Evans blue leakage (Fig. S9d-f). Thus, targeting glycoproteins increases AuNP accumulation on the BBB and reduces the AuNPs dosage to reach comparable efficacy with TJ-targeting AuNPs. Furthermore, we investigated the influence of AuNP–LEL dose on the in vitro and in vivo optoBBB efficiency. The results demonstrated that decreased nanoparticle doses showed a lower BBB opening extent in terms of in vitro TEER drop and permeability change (Fig. S10a and b) and in vivo Evans blue dye leakage (Fig. S10c and d).

Laser-induced elevation of Ca²⁺ propagates among endothelial cells and leads to BBB opening

Second, we examined the Ca²⁺ signaling and propagation during OptoBBB. We imaged intracellular Ca²⁺ signals using fluo-4 as an indicator before and after a focused laser stimulation (35 mJ cm⁻², 1 pulse. Fig. 2a). Our results show that the Ca²⁺ level started to increase at 2 seconds after laser excitation with AuNP targeting (Fig. 2b and c, S11, and video S1). In contrast, laser excitation without AuNPs shows no increase in fluo-4 intensity over the baseline (Fig. 2c). Interestingly, the image analysis shows Ca²⁺ propagation from laser-irradiated regions to adjacent and then distant regions (Fig. 2b and d, S11c and d, and video S1). We further observed an oscillatory Ca²⁺ signal in some cells (Fig. S11e and f). Previous studies reported that cell membrane deformation by mechanical stimulation triggered an increase in cytosolic inositol 1,4,5-trisphosphate (IP3). The Ca²⁺ wave is possibly stimulated by the IP3 diffusion and the receptor (IP3R) binding. As long as the IP3 concentration is sufficient to activate IP3R, Ca²⁺ release will continue to be initiated, and a Ca²⁺ wave will appear to propagate without degradation.²⁸ Importantly, incubation of the monolayers with BAPTA (Ca²⁺ chelator, 10 μM) blocks Ca²⁺ signaling and completely abolishes the TEER changes (35 mJ cm⁻², 10 pulses, 5 Hz. Fig. 2e). These findings confirm the involvement of transient Ca²⁺ elevation in the reversible BBB opening and that blocking Ca²⁺ is sufficient to block the BBB opening.

Fig. 2. Ca²⁺ elevation and propagation among endothelial cells.

(a) Schematic of Ca²⁺ imaging with a focused laser beam. (b) Representative images of the fluorescent intensity of fluo-4 (Ca²⁺ indicator) before and after laser stimulation of AuNP–LEL. Laser stimulation time is defined as 0 seconds. Green: fluo-4. The illustration graph of spatial binning shows the position of the center of the laser spot and the bins for Ca²⁺ signaling analysis. (c) Transient Ca²⁺ elevation in the laser spot after laser stimulation, 35 mJ cm⁻², 1 pulse. (d) Analysis of Ca²⁺ signaling shows the propagation after laser stimulation of AuNP–LEL and AuNP–BV16. 35 mJ cm⁻², 1 pulse. 30–60 cells were analyzed from 3 experiments. The distances refer to the radial distances for the edges of each bin relative to the center of the laser spot for the fluorescent analysis. (e) Ca²⁺ chelator (BAPTA) prevents TEER drop. 35 mJ cm⁻², 10 pulses. (f) The schematic shows only the right semi-well receives light. (g) A parallel circuit model for predicting the resistance of semi-well excitation is no propagation case. (h) Comparison of TEER with semi- and full-well laser stimulation and predicted resistance. 35 mJ cm⁻², 10 pulses. Ca²⁺ data corresponding to 30–60 cells was analyzed from 3 experiments. TEER data correspond to 3 replicates. Data expressed as mean ± SD. Unpaired t-test was performed individually between Only laser and the other groups. **: P < 0.01 was considered a statistically significant difference. Scale bar: 100 μm.

Ca²⁺ propagation has been reported in endothelial cells^29,30 and can amplify the BBB opening.³¹ To investigate the contribution of Ca²⁺ propagation on OptoBBB, we examined the TEER changes by blocking the laser irradiation on half of the transwell (Fig. 2f). A parallel circuit model (Fig. 2g) predicts the resistance change by calculating the parallel resistance across the two semi-wells. The experimental result shows that semi-well laser excitation leads to a similar TEER drop with the full-well excitation. However, the predicted resistance change of semi-well excitation is much less than the experimental measurement (35 mJ cm⁻², 10 pulses, 5 Hz. Fig. 2h). This finding indicates that Ca²⁺ propagation extends the area of BBB opening. Taken together, our results suggest that the elevation and propagation of cytosolic Ca²⁺ contribute to OptoBBB.

OptoBBB involves both internal Ca²⁺ depletion via IP3 signaling and extracellular Ca²⁺ influx

Next, we investigated the role of internal Ca²⁺ depletion and extracellular Ca²⁺ influx during OptoBBB, as both can lead to Ca²⁺ signaling for endothelial cells (Fig. 3a). To study the contribution of intracellular Ca²⁺ release, we removed the extracellular Ca²⁺ by incubating the cells with a Ca²⁺-free medium. Ca²⁺ elevation persists for both AuNP–LEL and AuNP–BV16 cases, although with a smaller amplitude (Fig. 3b). It has been reported that mechanical stimulation can trigger the cell membrane deformation to increase in cytosolic IP3. IP3 subsequently binds to its receptor IP3R on the endoplasmic reticulum (ER) to trigger Ca²⁺ release from the ER Pool.^32-35 To further confirm the role of IP3 signaling, we pretreated the monolayers using 2-aminoethoxydiphenyl borate (2-APB, 200 μM), an IP3R blocker, and observed no Ca²⁺ elevation after laser irradiation in Ca²⁺-free medium (Fig. 3b). As a control experiment and previously discussed, pretreatment of the cells with BAPTA, a high-affinity Ca²⁺ chelator, ablates the Ca²⁺ response (Fig. 3b).

Fig. 3. OptoBBB involves internal Ca²⁺ depletion and Ca²⁺ influx.

(a) The schematic of internal Ca²⁺ depletion via IP3 signaling and Ca²⁺ influx. (b) Internal Ca²⁺ depletion after laser stimulation of AuNP–LEL and AuNP–BV16. 35 mJ cm⁻², 1 pulse. Ca²⁺(+): FBS-free D3 culture medium that contains Ca²⁺ ions. Ca²⁺(−): HBSS solution without Ca²⁺ ions to remove the Ca²⁺ source from the extracellular environment. Ca²⁺(−) with 2-APB: HBSS solution without Ca²⁺ ions and pretreatment with IP3R blocker (2-APB) to suppress internal Ca²⁺ release. Ca²⁺(+) with BAPTA: FBS-free D3 culture medium and pretreatment with BAPTA (Ca²⁺ chelator). (c) Ca²⁺ influx after laser stimulation of AuNP–LEL and AuNP–BV16. 35 mJ cm⁻², 1 pulse. Ca²⁺(+) with 2-APB: FBS-free D3 culture medium and pretreatment with IP3R blocker. Ca²⁺(+) with 3 blockers: FBS-free D3 culture medium and pretreatment with 2-APB, GSK2193874, and GsMTx4 to block IP3 signaling, TRPV4, and Piezo 1, respectively. 30–60 cells were analyzed from 3 experiments. Data expressed as mean ± SD.

We further studied the contribution of extracellular Ca²⁺ influx on the cytosolic Ca²⁺ elevation. We blocked the internal Ca²⁺ source using 2-APB and observed a transient Ca²⁺ increase in the Ca²⁺-containing medium. We observed no Ca²⁺ elevation when switched to a Ca²⁺-free medium. This result suggests that Ca²⁺ influx from the extracellular solution (35 mJ cm⁻², 1 pulse. Fig. 3c) contributes to the observed cytosolic Ca²⁺ response. TRPV4 and Piezo1 are two mechanosensitive ion channels for Ca²⁺ influx, we pretreated monolayers with TRPV4 and Piezo 1 inhibitors (GSK2193874: 10 μM, GsMTx4: 10 μM) in the presence of the IP3R blocker, and the Ca²⁺ elevation persisted although with a smaller magnitude. This result indicates that other ion channels in addition to TRPV4 and Piezo 1 or changes in the plasma membrane itself may contribute to the observed Ca²⁺ influx. These studies suggest that OptoBBB involves internal Ca²⁺ depletion through IP3 signaling and extracellular Ca²⁺ influx through multiple mechanisms including the mechanosensitive ion channels (TRPV4 and Piezo1).

OptoBBB involves ERK1/2 phosphorylation and actin polymerization

Lastly, we explored the Ca²⁺-related downstream signaling that increases BBB permeability, including actin polymerization and ERK1/2 phosphorylation (p-ERK1/2)^36-38 (Fig. 4a). Western blotting results show increased p-ERK1/2 signal following laser stimulation with onset at 5 minutes, peak at 0.5 hours, and a return to baseline at 6 hours (35 mJ cm⁻², 1 pulse. Fig. 4b and c). Application of p-ERK1/2 blocker U0126 suppressed the p-ERK1/2 level (Fig. 4b and c), consistent with previous reports.³⁹ Rapid endothelial cytoskeletal rearrangement facilitates BBB disruption.³⁸ To assess the actin skeleton response, we stained the F-actin by fluorescent phalloidin. The results show that the F-actin fiber signal increases at 5 minutes, maximizes at 0.5 hours, and returns to baseline at 6 hours (35 mJ cm⁻², 1 pulse. Fig. 4d and S12). Quantitative analysis shows an increase in F-actin immunofluorescence and a decrease in the anisotropy score of F-actin at 5 minutes and 0.5 hours, both recovering at 6 hours (Fig. 4e and f). The increased immunofluorescence indicates more F-actin formation after laser stimulation. The decreased anisotropy score suggests F-actin is less aligned after laser stimulation. Such cytoskeleton reorganization could contribute to the increased BBB permeability.³⁸ However, a strongly correlated temporal profile of p-ERK1/2 signaling and F-actin does not suggest a direct causal link. These findings suggest OptoBBB involves phosphorylation of ERK1/2 and actin polymerization.

Fig. 4. OptoBBB involves ERK1/2 phosphorylation and actin polymerization.

(a) Proposed signaling pathway for OptoBBB. (b) Phosphorylation of ERK1/2 (pERK) after laser excitation detected by western blot. 35 mJ cm⁻², 1 pulse. (c) Quantification analysis of data in (b). Left panel: pERK normalized by control, no treatment. Right panel: ratio of pERK and total ERK1/2 (ERK) normalized by control, no treatment. (d) F-actin staining by phalloidin after laser stimulation of AuNP–LEL. 35 mJ cm⁻², 1 pulse. (e) Quantification of fluorescent intensity of F-actin normalized by nuclei. Red: F-actin. Blue: nuclei. (f) Quantification of anisotropy score of F-actin. The black dashed line indicates the baseline. Western blot data correspond to 3 replicates. F-actin data correspond to 21–30 field of views (FOVs) from 3 experiments. Data expressed as mean ± SD. Unpaired t-test was performed between No treatment and the other group, respectively. *: P < 0.05, **: P < 0.01, or ***: P < 0.001 was considered a statistically significant difference. Scale bar: 20 μm.

Discussion

Our results lead us to propose the following mechanobiological framework for OptoBBB (Fig. 5). Pulsed laser excitation of vascular-targeting AuNPs produces a nanoscale mechanical perturbation. This perturbation triggers (1) actin polymerization; (2) Ca²⁺-influx from mechanosensitive ion channels (TRPV4, Piezo1); (3) GPCR activation and the downstream production of cytosolic IP3. IP3 activates the receptors on the endoplasmic reticulum and leads to the Ca²⁺ release from the ER Pool. The elevation of Ca²⁺ from the extracellular influx and intracellular IP3 pathway activates ERK1/2 phosphorylation. The phosphorylation of ERK1/2, together with the actin network, leads to an activation of the cytoskeleton resulting in an increase in paracellular permeability. Moreover, our previous work demonstrated that under current experimental conditions, the plasmonic heating generated by the interactions between picosecond laser and AuNPs is not a key factor to be responsible for the BBB opening.^8,40 The above results are consistent with our in vivo investigation that showed increased BBB permeability via paracellular diffusion through the tight junctions.

Fig. 5. Summary of the proposed mechanism for OptoBBB.

The mechanical pressure generated by laser excitation of AuNPs produces 3 effects: (1) the mechanical pressure gently deforming the cell membrane triggering actin polymerization, (2) Ca²⁺-influx from mechanosensitive channels, and (3) mechanosensitive GPCR activation resulting in G protein activity, which in turn increases the production of cytosolic IP3. IP3 activates the receptors on the endoplasmic reticulum (ER), leading to the Ca²⁺ release from the ER. Then the elevation of Ca²⁺ activates the downstream ERK1/2 phosphorylation. Finally, the phosphorylation of ERK1/2 together with actin networks causes a cytoskeletal contraction to increase the paracellular space between adjacent cells, eventually increasing the BBB permeability. Laser/−: AuNPs targeting without laser stimulation. Laser/+: AuNPs targeting and picosecond laser stimulation.

Glycoprotein is one of the molecular components of the glycocalyx, a thick matrix highly expressed on the luminal surface of the BBB.^8,41,42 Targeting glycoproteins is expected to increase the vasculature targeting efficiency since glycoproteins display a much larger surface area on the BBB than the tight junction. Our results show significantly higher AuNPs accumulation for AuNP–LEL than AuNP–BV16 in vitro (21-fold) and in vivo (12-fold). We observed that targeting glycoprotein leads to more efficient BBB opening than targeting JAM-A with the same AuNP dose. The high targeting efficiency enables a much lower AuNP dose required for BBB opening (0.02 nM AuNP–LEL versus 0.5 nM AuNP–BV16) to achieve similar BBB opening efficiency. Future studies may investigate intracarotid infusion of AuNP–LEL to increase the targeting efficiency of AuNP–LEL in vivo and further reduce the systemic administration of AuNP, as well as validate the Ca²⁺-dependent mechanisms of optoBBB in vivo. Currently, we are exploiting near-infrared (NIR) laser and NIR light-absorbing nanoparticles to open the BBB, since NIR light has deeper penetration depth into the brain compared to visible light. We will further evaluate the BBB opening mechanisms for NIR laser and NIR light-absorbing nanoparticles. We will also investigate the mechanisms of BBB modulation in deep brain regions using an optical fiber.

Conclusion

In this study, we found that targeting the glycoprotein improves the vascular targeting efficiency compared with targeting tight junctions by more than 10-fold both in vitro and in vivo. With both tight junction and glycoprotein targeting, we demonstrate that laser excitation of vasculature-targeting AuNPs leads to elevation and propagation of Ca²⁺ among endothelial cells, actin polymerization, and Ca²⁺-dependent phosphorylation of ERK1/2. The Ca²⁺-sensitive signaling and actin polymerization lead to cytoskeletal activation and increase the paracellular space and permeability. Our study demonstrates a method to significantly increase BBB targeting efficiency and elucidates the Ca²⁺-dependent mechanobiological mechanism for OptoBBB, opening new avenues for future development of this technology and CNS disease treatment.