Intracellular Delivery Using Nanosecond-Laser Excitation of Large

Mar 14, 2017 - Efficiently delivering functional cargo to millions of cells on the time scale of minutes will revolutionize gene therapy, drug discove...
0 downloads 10 Views 7MB Size
Intracellular Delivery Using Nanosecond-Laser Excitation of Large-Area Plasmonic Substrates Nabiha Saklayen,*,†,& Marinus Huber,‡,§,& Marinna Madrid,‡ Valeria Nuzzo,∥ Daryl I. Vulis,‡ Weilu Shen,‡ Jeffery Nelson,⊥ Arthur A. McClelland,# Alexander Heisterkamp,*,7 and Eric Mazur*,†,‡ †

Department of Physics, ‡John A. Paulson School of Engineering and Applied Sciences, ⊥Division of Science, and #Center for Nanoscale Systems, Harvard University, Cambridge, Massachusetts 02138, United States § Department of Physics, Ludwig Maximilian University of Munich, 80539 Munich, Germany ∥ ECE Paris Ecole d’Ingénieurs, 75015 Paris, France 7 Institute for Quantum Optics, Leibniz University Hannover, 30167 Hanover, Germany S Supporting Information *

ABSTRACT: Efficiently delivering functional cargo to millions of cells on the time scale of minutes will revolutionize gene therapy, drug discovery, and highthroughput screening. Recent studies of intracellular delivery with thermoplasmonic structured surfaces show promising results but in most cases require time- or cost-intensive fabrication or lead to unreproducible surfaces. We designed and fabricated large-area (14 × 14 mm), photolithographybased, template-stripped plasmonic substrates that are nanosecond laser-activated to form transient pores in cells for cargo entry. We optimized fabrication to produce plasmonic structures that are ultrasmooth and precisely patterned over large areas. We used flow cytometry to characterize the delivery efficiency of cargos ranging in size from 0.6 to 2000 kDa to cells (up to 95% for the smallest molecule) and viability of cells (up to 98%). This technique offers a throughput of 50000 cells/min, which can be scaled up as necessary. This technique is also cost-effective as each large-area photolithography substrate can be used to deliver cargo to millions of cells, and switching to a nanosecond laser makes the setup cheaper and easier to use. The approach we present offers additional desirable features: spatial selectivity, reproducibility, minimal residual fragments, and costeffective fabrication. This research supports the development of safer genetic and viral disease therapies as well as research tools for fundamental biological research that rely on effectively delivering molecules to millions of living cells. KEYWORDS: plasmonic intracellular delivery, template-stripping, flow cytometry, pulsed lasers, thermoplasmonic substrates

T

directly into cells will transform biomedical research. However, no current intracellular delivery method, either biological, chemical, or physical, can offer a collection of advantageous features at once: (1) high efficiency, viability, and throughput, (2) diverse cargo delivery capability, (3) spatial selectivity (delivering to specific cells on a surface), scalability and reproducibility, (4) minimal postdelivery cytotoxicity, and (5) cost-effectiveness. Viral transduction is the most popular biological method due to decades of extensive research but has limitations. For example, viral methods offer limited cargo-carrying capacity, only deliver genetic cargo, require customization of the virus for each cargo and cell type, and include immunotoxicity risks.12−14

he direct intracellular delivery of functional proteins, enzymes, or genetic material is a powerful way to manipulate cell behavior for biomedical research and disease intervention.1,2 For example, functional proteins are delivered to ablate genes in hematopoietic stem cells with high precision for research and therapeutic applications.3,4 Enzymes are delivered for their ability to bind targets with high affinity and specificity.5,6 siRNA delivery facilitates gene silencing for biomedical applications.7,8 In the case of blood disorders such as human immunodeficiency virus (HIV) or leukemia, delivering functional molecules into a patient’s stem cells for transplantation therapy could cure them of the disease, circumventing the side effects of chemotherapy, transplantation rejection by the immune system, and the search for a matching donor.9−11 The ability to effectively deliver large and diverse cargos such as amino acids, peptides, proteins, protein cages, antibodies, polysaccharides, nucleic acids, viruses, and DNAs/RNAs © 2017 American Chemical Society

Received: December 6, 2016 Accepted: March 14, 2017 Published: March 14, 2017 3671

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

www.acsnano.org

Article

ACS Nano

Figure 1. Nanosecond-laser excitation of large-area plasmonic substrates for intracellular delivery. A nanosecond laser is scanned across the sample for high-throughput delivery to 1 million cells cultured on the substrate. The laser spot illuminates thousands of cells simultaneously. Laser illumination causes thermoplasmonic nanoheating at the apex of each pyramid. The heating forms bubbles that create temporary pores in cells at close contact. The pores enable fluorescent cargo in the surrounding solution to diffuse directly into the cytosol.

to produce substrates to study millions of cells, and switching to a nanosecond laser makes the setup cheaper and easier to use. We show that template-stripped substrates, consisting of pyramidal nanoheaters, permit intracellular delivery with efficiencies of up to 95% for the smallest molecules, viability up to 98%, and scalable throughput of 50000 cells/min, as quantified by flow cytometry. Additional desirable features of this technique include cargo delivery capacity up to 2000 kDa, spatial selectivity, scalability, reproducibility, minimized cytotoxicity from gold fragment residue, and no requirement of biological or chemical customization. Intracellular delivery using large-area template-stripped thermoplasmonic substrates will support the ability to effectively study cellular function with improved speed and control, which in the long run could lead to the development of gene therapies to cure chronic diseases.

Among nonviral methods, electroporation techniques offer high throughput and efficiency but also lead to high cell death.15 Nucleofection offers improved viability but requires expensive customized reagents.16 Lipofection, a chemical method, offers high throughput but is cell-type specific and risks endosomal trapping of the cargo.17 Physical methods such as microinjection, nanowire-mediated delivery, and microfluidic squeezing are promising but offer limited throughput and/or reproducibility.13,15,18−23 Thermoplasmonic nanostructures have the potential to become excellent tools for biomedical applications.24−26 Laser-activated thermoplasmonic gold nanoparticles form heat-mediated bubbles to make transient pores in cells for intracellular delivery.27−35 Despite being effective intracellular delivery agents, the long-term cytotoxicity of gold nanoparticles still remains under debate.36,37 Recent studies of intracellular delivery with thermoplasmonic structured surfaces are encouraging as they do not utilize gold nanoparticles in solution but in some cases require time- or cost-intensive fabrication; in some instances, each substrate is fabricated using lithography methods, making it difficult to scale up to applications that require many large-area samples.20,38 Some substrate-based studies with gold nanoparticle layers are more easily fabricated but can be nonreproducible in nature and leave metal fragments behind after the experiment.39 Template stripping is a high-throughput and reproducible approach to producing plasmonic structures in the form of metallic films that are ultrasmooth and precisely patterned over large areas.40−43 This fabrication method is highly advantageous as lithography needs to be performed once to fabricate templates in silicon that can be repeatedly used to generate template-stripped metallic substrates in a few steps. In our previous work, tipless pyramid substrates exhibiting a high nearfield enhancement were irradiated with femtosecond laser pulses to show proof-of-principle intracellular delivery.44 However, only small areas could be fabricated due to template-stripping from an e-beam lithography template, limiting experiments to the study of tens of cells. The tipless pyramids required angular gold deposition using a thermal evaporator with rotating stage, which affected precision in reproducibility. Using a femtosecond laser made the optical setup complex. These challenges made it difficult to develop large-scale biomedical applications of the technique. We demonstrate that thermoplasmonic substrates made by template-stripping from large-area photolithography templates can be combined with nanosecond laser excitation to achieve intracellular delivery in large numbers of cells, which enables flow cytometry measurements. This technique is cost-effective as large-area photolithography templates can be used repeatedly

RESULTS AND DISCUSSION Nanosecond Laser Illumination of Large-Area Plasmonic Substrates. We use pulsed-nanosecond laser illumination of large-area template-stripped thermoplasmonic substrates to deliver cell-impermeable cargo directly into the cytosol of living cells (Figure 1). The thermoplasmonic substrate consists of microscale pyramids with a continuous gold thin film on top, fabricated using photolithography and template-stripping. We culture cells on the substrate, surround them with a solution containing the delivery cargo, and scan a nanosecond pulsed laser across the substrate to generate intense heating at the apex of each pyramid, forming hotspots. The hotspots lead to bubble formation, creating transient pores in cells adhering to the substrate. These pores enable cargo from the surrounding solution to enter the cytosol via diffusion. A large beam spot illuminates the substrate to generate pores in thousands of cells simultaneously (setup shown in Figure S1). Millions of cells are permeabilized when the laser is scanned across the substrate over the course of minutes. In our previous research, we used tipless pyramids that were template-stripped from e-beam lithography templates and femtosecond laser-excited to show proof-of-principle intracellular delivery.44 The tipless pyramids were designed to maximize the high nearfield enhancement with femtosecond pulses as the high nearfield enhancement is considered a possible driving mechanism for cell poration.45 We switched to a compact, turn-key nanosecond laser to bring down the cost of the laser setup by a factor of 10 and reduce the setup complexity. Since the nanosecond laser causes heating to be the dominant effect, we changed our design to regular pyramids (with tip in place) to ease the fabrication process. The nanosecond laser used has a wavelength of 1064 nm (which is advantageous for reduced scattering effects), repetition rate of 3672

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

ACS Nano

Figure 2. Large-area template-stripped thermoplasmonic substrates. (a) Repeated template-stripping on photolithography templates enables low-cost fabrication of large areas. (b) One substrate consisting of 10 million pyramids. (c) Top-down view of pyramids under scanning electron microscopy (SEM). (d) SEM-tilted view shows uniform base lengths (2.4 μm), heights (1.4 μm), and spacings (1.2 μm). (e) Finite element method simulations showing the thermoplasmonic pyramidal apexes reaching maximum temperature of 342 °C at a laser fluence of 45 mJ/cm2.

Figure 3. Cell morphology on substrate. (a) SEM of chemically fixed HeLa CCL-2 cells on a thermoplasmonic substrate. (b) Confocal laserscanning microscopy slice of cell with Calcein red-orange AM fluorescence at z = 8.85 μm, as measured from the bottom of the cell. The cell is 40 μm in length. (c) Grid pattern from pyramids. (d) 27 z-slices stacked (with spacings of 0.57 μm) to show the cross section of the same cell taken along the dotted line in (c).

50 Hz, beam spot diameter of 1.2 mm, and a flat top beam profile. We use a large beam spot to directly scan the surface of the substrate instead of focusing through an objective to enable rapid scanning of large areas without worrying about losing focus. Due to the low repetition rate, each pulse can be seen as having an individual effect. In order to obtain larger pyramid arrays, our challenge was to optimize fabrication for large areas while maintaining uniform precision for approximately 10 million pyramids on a single substrate, across dozens of substrates. We fabricated silicon master templates (14 × 14 mm) using photolithography and anisotropic etching. Each template consists of 10 million inverted pyramids of base length 2.4 μm and edge-to-edge separation 1.2 μm (details of the fabrication process are

discussed in the Supporting Information and Figure S2). We deposit a 50 nm thick gold film onto dozens of master templates and add UV glue and a coverslip on top. After the glue cures, we peel off (“template-strip”) the gold, glue, and coverslip composite using a razor blade (Figure 2A). This results in thermoplasmonic substrates comparable in size to a U.S. quarter dollar (Figure 2B). By repeating gold deposition and template-stripping, we fabricate dozens of thermoplasmonic substrates with high precision in a single batch. Scanning electron microscopy (SEM) confirms that the templatestripped substrates are highly uniform in base lengths, spacings, and heights across the length of an individual substrate and consistent from batch to batch (Figure 2C,D). The absorption, reflection, and transmission spectra measured on the substrates 3673

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

ACS Nano

Figure 4. Characterization of the efficiency and viability of delivery to cells and confirmation of no visible substrate damage. (a) Overlay of a bright-field and fluorescent image. Areas I and III have pyramids. Areas II and IV are flat gold. A laser is scanned below the dotted line. Area III is the only region with green molecules delivered to cells. (b) Delivery efficiency and viability as a function of laser fluence. Data represent mean ± standard error from n = 3 independent experiments. (c) Calcein green delivery. (d) Calcein red-orange AM viability. (e) Overlay of images (c) and (d). (f) SEM image after chemical fixing. (g) Overlay of images (e) and (f). (h) Zoom of (g). Inset of an individual pyramid (2.4 μm base length) confirms no visible damage on pyramid. Filopodia of a cell at the bottom right corner of the pyramid.

90% of this critical value, which corresponds to temperatures between 293−340 °C. Simulation results show that for fluences above 40 mJ/cm2, the aqueous environment within tens of nanometers surrounding the pyramidal apex reaches temperatures above 300 °C, which is within the temperature range for bubble formation. All laser experiments with cells are performed above the threshold laser fluence necessary for bubble formation (40 mJ/cm2) but below the point where we damage the gold film at the pyramidal apex (200 mJ/cm2). We imaged the morphology of HeLa cells on the thermoplasmonic substrate using SEM and confocal microscopy. Figure 3A is an SEM image of a chemically fixed cell on the pyramids. Because chemical fixation kills the cell and alters the cell’s lipid membranes, we additionally used confocal microscopy to image the morphology of a living cell that was fluorescently tagged with a cell-permeant dye for living cells (Figure 3B,C). A confocal slice at a height of 8.85 μm from the bottom of the cell reveals regular cell morphology (Figure 3B). A scan closer to the bottom of the cell (at 3.42 μm) makes pyramids appear in a dark gridlike pattern as the pyramids do not fluoresce in this channel (Figure 3C). A z-stack cross section of the cell along the dotted white line shows the membrane adhering to the pyramids (Figure 3D). Both imaging methods indicate that each cell adheres to approx-

are shown in Figure S3. The master templates can be reused hundreds of times for template stripping. This intracellular delivery technique is easily scalable as photolithography can be extended to make larger samples, and the scanning speed and/ or beam spot size can be increased with additional optical components. Plasmonic pyramids are effective nanoheaters due to the combined effect of a thin metal film and sharp apex, making them ideal for thermoplasmonic applications.22,41−43 Sharp metallic nanostructures are ideal for absorbing laser energy and concentrating the energy to hotspots that form bubbles in an aqueous environment.24,46−48 We performed numerical simulations to confirm that the incoming laser energy is concentrated at the pyramidal apex, in a region of hundreds of nanometers, leading to localized heating (Figure 2E). At fluences above 200 mJ/cm2, we observe damage at the apex of the pyramid, confirming experimentally that the heating is concentrated at the apex of the pyramids (Figure S4). Using numerical simulations, we determined the threshold laser fluence required for the temperature at the pyramidal apex to reach the temperature range for bubble formation. The critical temperature of water at 1 atm is 367−377 °C, and thermodynamical theory and experimental studies46,49 show that explosive boiling occurs where temperatures reach 80− 3674

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

ACS Nano

Figure 5. Quantifying delivery efficiency and viability of different-sized molecules using flow cytometry. (a, b) Representative data from the FITC−dextran 150 kDa experiment. The experimental sample (a) is laser-scanned while the control (b) is not. A density plot of side-scatter area vs Calcein red-orange AM fluorescence determines viable cells. The viable cells are plotted in a side-scatter vs FITC−dextran 150 kDa density plot to measure the delivery efficiency. The background signal is less than 1% in the control in (a). The delivery efficiency is 74% in the experiment (b). (c) Viability and efficiency for FITC-cargo ranging in size from 0.623 kDa to 2000 kDa. Data represent mean ± standard error from n = 3 independent experiments.

laser-scanned, while areas I and II were not. Only area III, the area containing gold-coated pyramids that is irradiated by the laser, demonstrates the successful delivery of cell-impermeable Calcein green molecules. This observation confirms that goldcoated pyramids in combination with laser irradiation leads to the delivery of molecules into cells. Figure 4A demonstrates that the technique offers spatially selective delivery and works only when cells are cultured on pyramids and laser-scanned. Note that the pyramidal surface on the left reflects less light than the flat gold on the right, making the surface appear darker.

imately 40−50 pyramids, which enables the excitation of specific regions of the cell by targeting local hotspots. Optimizing Laser Fluence for Intracellular Delivery without Damage. Experiments demonstrate that laserscanning a thermoplasmonic substrate covered in HeLa cells leads to intracellular delivery of cargo that is dissolved in the surrounding solution in a spatially selective manner. In Figure 4A, we overlay a bright-field image with a fluorescent image with excitation and emission wavelengths of 495/515 nm, respectively. Areas I and III contain gold-coated pyramids, whereas areas II and IV are flat gold. Areas III and IV were 3675

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

ACS Nano

after laser scanning (Table S6). Cells that undergo thermoplasmonic intracellular delivery have a signal that is representative of background noise, as do cells that were not in contact with gold. ICP-MS measurements on cells that are incubated with gold nanoparticles (and then washed several times to remove floating nanoparticles) exhibit greater gold content than background noise. ICP-MS verifies that no gold residue remains in the cells post-experiment as compared to controls, therefore minimizing risks of gold fragment-induced mutagenesis. Quantifying Delivery Efficiency and Viability with Flow Cytometry. We performed flow cytometry experiments to quantify the delivery efficiency and viability of cells for delivery of different cargo sizes with triplicate experiments performed on different substrates on the same day. We delivered a range of cargos (Calcein green, dextran 10 kDa, dextran 70 kDa, dextran 150 kDa, dextran 500 kDa, and dextran 2000 kDa) to 1 million HeLa cells by laser scanning the entire thermoplasmonic substrate in 3 min. We then used trypsin to detach the cells from the substrate to prepare them for flow cytometry measurements. Flow cytometry measured the forward and side-scattered light and the fluorescence of cells passing through a beam of light. The forward and side scattering gave information about cell size and internal granularity, respectively.50 Figure 5 compares viability and delivery efficiency of cells that were not laser-scanned (Figure 5A) and cells that were laser-scanned (Figure 5B). For the substrate that was not laserscanned, the viability is 98.3%, and the background signal for delivery is less than 1% (Figure 5A). For the laser-scanned substrate, the viability is 98.2% and the delivery efficiency is 74% (Figure 5B). The delivery scatter plots show a range of fluorescence signals, indicating that different amounts of FITCcargo entered the cell. We performed triplicate independent experiments with flow cytometry for cargo sizes ranging from 0.6 to 2000 kDa (Figure 5C and Figures S9 and S10). The largest cargo, FITC−dextran 2000 kDa, is delivered with an efficiency of 16% and a viability of 97% (Figure 5C). The largest increase in efficiency is between FITC−dextran 500 kDa (24%) and FITC−dextran 150 kDa (68%). Calcein green (0.623 kDa) is delivered at 95% efficiency and 98% viability. We attribute the increased efficiency with decreasing cargo size to faster diffusion of smaller molecules. The cargos we delivered match functional proteins in molecular weight (13−150 kDa).51 Flow cytometry results were highly reproducible, which is important for reliable biological studies that require robust statistical data, indicating a batch-to-batch consistency in substrate fabrication. The viability with and without cargo in solution combined with and without laser scanning was determined to be between 97.3 and 98.6% for all cases using flow cytometry (Table S5). Consistent fabrication enables different substrates to undertake intracellular delivery to cells at reproducible efficiencies. The viability of cells after different cargo sizes were delivered to them in Figure 5c is consistent, even though experiments were performed on different days, indicating that results are reproducible over extended periods of time.

We determined the efficiency of Calcein green delivery to HeLa cells (Figure 4C) and used a second dye, Calcein redorange AM, to check the postexperimental viability of the cells 4 h after the experiment (Figure 4D) in a small region of the substrate. The overlay of the efficiency and viability in Figure 4E demonstrates that cells with Calcein green delivered to the cytoplasm survive the experiment. To determine how many cells had Calcein green delivered to the cytoplasm (efficiency) and how many of those cells were alive in the Calcein redorange AM channel (viability) we used automated cell counting on fluorescent images. We determined the optimum laser fluence for maximum delivery efficiency and viability by repeating each experiment described above on three separate substrates in three separate dishes at different laser fluences. Figure 4B shows the fluence dependence of the efficiency and viability. At the lowest fluence (48 mJ/cm2) in Figure 4B, the viability is 98% and the efficiency is 70%. This means that most of the cells are surviving the laser treatment, but only 70% of them have cellimpermeable cargo (Calcein green) inside them. As the laser fluence is increased to 54 mJ/cm2, 95% of the cells have Calcein green in them, and 98% of them are still viable. We define this point as our optimum fluence as viability and efficiency are maximized. Simulation results (Figure S4) show that the local temperature reaches 370 °C at this fluence, which is in the temperature range for bubble formation. Increasing the laser fluence beyond this point leads to the onset of cell death, as can be seen starting at 56 mJ/cm2, where the viability drops to 80%. The efficiency is also around 80% and continues to follow the viability curve closely at higher fluences. This trend occurs because Calcein green can only remain within a cell if the cell is alive and has an intact cell membrane. Since a fraction of the cells begin to die starting at 56 mJ/cm2, the efficiency decreases accordingly with viability. Past the optimum fluence, cells either survived the laser irradiation, had cell-impermeable cargo within, or were not viable anymore. To determine if the pyramids experience any visible damage or melting during laser exposure at 54 mJ/cm2, we chemically fixed the cells after laser scanning and took an SEM image of the substrate. As Figure 4F illustrates, the pyramids in the lasertreated area experiences no visible damage despite undergoing intense thermoplasmonic heating at the apex (inset Figure 4H). We observe no damage to the pyramids in the fluence range 48−90 mJ/cm2 (Figure S6). All subsequent experiments are performed at the optimum fluence of 54 mJ/cm2 because it offers the highest efficiency and viability and no visible damage. In the Supporting Information, we show that this technique delivers cargo with uniform distribution within each cell, which is advantageous for applications that need to avoid endosomal trapping (Figure S7). We observe cells proliferating after thermoplasmonic treatment, indicating high cell viability at 48 h after laser treatment (Figure S8). Dextran 150 kDa is retained in the cytoplasm during proliferation over 48 h (Figure S8). The viability with and without cargo in solution combined with and without laser scanning was determined to be between 97.7 and 99.3% for all cases using cell counting (Table S1). Even though cells may appear viable after undergoing intracellular delivery with this technique, there is a potential risk of DNA mutations if gold nanofragments remain in the cell.36 We performed inductively coupled plasma mass spectroscopy (ICP-MS) on laser-scanned cells, along with unirradiated controls, to measure the gold content in the cells

CONCLUSIONS We demonstrate that nanosecond-laser excitation of templatestripped thermoplasmonic substrates offers an alternative intracellular delivery technique for cargos of different sizes. 3676

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

ACS Nano

700-1 photoresist was spin-coated onto the wafer at 3000 rpm for 45 s (ramp = 1000 rpm/s). The wafer was then softbaked at 115 °C for 60 s (Figure S2B). The entire area of the silicon wafer was exposed in the autostepper to form a grid-based pattern. A postexposure bake was performed at 115 °C for 60 s. The wafer was developed in CD-26 developer for 1 min and then rinsed with DI water for 20 s. Development was repeated until no residue was released into the developer. The plasma stripper was then used to descum the wafer for 15 s (Figure S2C). A chromium etch was performed for 12 s (1.5 nm/ s etched at room temperature) to remove the Cr in the exposed squares. The sample was washed with DI water and dried with a N2 gun. The photoresist was removed in acetone. An O2 plasma clean was performed (conditions: 100 W at 20 mTorr for 3 min) to completely remove residual photoresist (Figure S2D). KOH etch: A 4.9% HF etch of 15 s was used to remove oxide formed on the silicon. A KOH etch consisting of 2 parts water and 1 part 45% KOH was performed at 80 °C on a hot plate with a thermometer for 3 min to form the inverted pyramid substrate (Figure S2E). Chromium etching was performed at room temperature for 20 s to remove the hard mask (Figure S2F). Fabrication of Template-Stripped Substrates. Gold (50 nm) is deposited via electron-beam evaporator (Figure S2G). A no. 1.5 coverslip is glued to a master template with UV-curable glue (Norland Adhesive 61) and cured under the UV lamp overnight (Figure S2H). The thermoplasmonic substrate is peeled off from the template using a razor blade (Figure S2I) to produce the final template-stripped substrate (Figure S2J). Temperature Simulations. The light−substrate interaction was studied in an aqueous environment using a Finite Element Method (Comsol, Multiphyiscs 4.4). The three-dimensional electromagnetic interaction was calculated using the scattered field formulation under the assumption that the spot size is large compared to the periodic pyramid structure (plane wave approximation) and that the optical properties of the polymer, gold film, and adjacent water do not change during the interaction with the linearly polarized 11 ns (fwhm) Gaussian pulse. The geometrical parameters (pyramid base length, 2.4 μm; spacing, 1.2 μm; thickness of the gold layer, 50 nm) were chosen to resemble the actual fabricated sample. Periodic boundary conditions were applied in the x and y directions. The simulation domain was truncated in the z direction using perfectly matched layers (PML). The calculated laser energy absorbed was used as a transient heat source for a one-temperature model. The spatial and temporal evolution of the gold and water temperature was calculated for different laser fluences. More details about the setup of the simulation can be found in Demesy et al.53 The physical properties of gold and water were taken from Ekici et al.54 Properties of the UV glue were provided by the manufacturer. Seeding Cells on Substrates. HeLa CCL-2 cells are cultured in DMEM containing 10% (FBS) and 1% penicillin streptomycin and are incubated at 37 °C and 5% CO2. Cells are passaged every other day and used for experiments at 80% confluence between passage numbers 15 and 30. Cells are washed with 8 mL of PBS and incubated for 3 min with trypsin (5 mL) before being neutralized with cell media (13 mL). Pipetting is used to wash the bottom of flask five times with the cell mixture before the mixture is transferred to a 15 mL tube and centrifuged for 5 min at 125g. The supernatant is removed gently with vacuum pipet, and cells are resuspended in 8 mL of fresh media and pipetted 30 times (up and down counted as 1 time). We use a Countessa cell counter to measure cell density and viability. Healthy cells have 90−99% viability. Eight template-stripped substrates are taped lightly to the bottom of a 100 mm Petri dish with double-sided Kapton tape. Five million cells are suspended in 15 mL of fresh cell medium and added to the Petri dish for overnight incubation. Laser-Scanning Experiments. Substrates with cells on the surface are transferred to a 35 mm Petri dish, and 2 mL of prewarmed PBS (37 °C) solution containing the molecules to be introduced into the cells (Calcein green at 500 μM or FITC−dextran at 25 mg/mL) is added. The laser setup is shown in Figure S1. A Nd:YAG laser emits 11 ns pulses at a wavelength of 1064 nm with a repetition rate of 50 Hz. The laser beam passes through an optical isolator, a half-wave plate (HWP), and a polarizer (P) controlling the beam energy passing

We use laser irradiation to form hotspots at the apex of goldcovered pyramids on a substrate. These hotspots make the surroundings, within a nanoscale region (tens of nanometers), reach high temperatures required for bubble formation. These bubbles permeabilize the membranes of cells adhering to the substrates and the temporary pores allow for cell-impermeable cargo to diffuse into the cytoplasm. The ability to deliver cargos ranging in size from 0.623 to 2000 kDa to cells makes this technique applicable to different research and clinical applications. The cargos delivered (especially the 70 and 150 kDa) are the same size as biologically relevant cargos, including proteins and antibodies, making this technique valuable for studying cellular function, accelerating drug discovery, enhancing gene therapy, and improving high-throughput screening technologies. We characterized the following features of this intracellular delivery technique: (1) efficiency, viability, and throughput, (2) diverse cargo delivery, (3) spatial selectivity, scalability and reproducibility, (4) gold fragment residue, and (5) costeffective fabrication and laser setup. We quantify delivery efficiency (up to 95% for 0.623 kDa cargo) and viability (98%) using two methods, fluorescent microscopy and flow cytometry. We fabricated highly uniform template-stripped thermoplasmonic substrates over large areas using cost-effective templatestripping of templates made with photolithograpy. The large size of the substrates enabled us to perform flow cytometry measurements on a large number of cells, making our technique useful for further biological studies, which is not possible with many substrate-based intracellular delivery techniques due to complex and/or irreproducible fabrication.13 Even though we achieved promising delivery throughput rates of 50000 cells/ min, this rate can increased by several factors by building scanning mirror setups and/or increasing the beam spot size. Additional desirable features of the technique include reproducibility across batches, scalability, and spatial selectivity, which can be important for the direct intracellular delivery of functional proteins, enzymes, or genetic material and not always offered by gold nanoparticles and other substrate-based intracellular delivery methods.38,39 We examined the reproducibility across samples by performing flow cytometry in triplicate on the same day and on different days and delivered cargos of wide-ranging sizes (0.6−2000 kDa). ICP-MS measurements confirmed that no gold residue remains in the cells postexperiment, therefore minimizing risks of gold fragment-induced mutagenesis in therapeutic applications, which is advantageous as compared to nanoparticle-mediated intracellular delivery.36,52 In the future, we plan to transpose this technique to other cell types. Intracellular delivery using nanosecond-laser activation of template-stripped thermoplasmonic substrates holds potential to advance personalized medicine for chronic diseases in a costeffective manner; for example, genome-editing cargo could be delivered directly to a patient’s own cells for treatment, bypassing an extensive donor search or the possibility of lifethreatening toxicity.

METHODS Fabrication of Master Templates. Silicon wafers were sonicated in acetone for 5 min and then methanol for 5 min before being rinsed in IPA. O2 plasma cleaning was performed with the conditions 100 W at 20 mT for 1 min. Cr (15 nm), used as a hard lithographic mask, was deposited via thermal evaporation (Figure S2A). The wafer was baked at 200 °C for 3 min to evaporate all solvents before processing. SPR 3677

DOI: 10.1021/acsnano.6b08162 ACS Nano 2017, 11, 3671−3680

Article

ACS Nano through the setup. A pelican beam splitter sends 8% of the beam to an energy detector (ED) to monitor the energy during an experiment. A lens (L) loosely focuses the beam (1.2 mm in diameter) on the sample, which sits on an x−y translational stage. The Petri dish contains the dye to be delivered in PBS solution. For the fluence experiments (Figure 4), the laser was scanned in a line across the sample (scan speed = 10 mm/s; line length Δx = 100 mm), with each line at a different fluence. We used the maximum scanning speed on our stage (10 mm/s) to make the experimental time as short as possible. We then optimized the fluence of the laser for this scanning speed. We used a seeding density to ensure 80% cell confluency at the time of experiments. Distances longer than the width of the sample were scanned to avoid cells being affected by the acceleration and deceleration of the stage at the end of each run, which changes the number of pulses hitting the sample and affects the viability and efficiency of the technique. For flow cytometry, experiments were done using the exact same procedure as fluence scanning experiments and the entire sample was scanned (scan speed = 10 mm/s, Δx = 100 mm, Δy = 0.5 mm, pyramid region scanning time is 3 min, total time scanning time is 10 min). A negative control of areas with pyramids and no laser irradiation and areas with no pyramids and no irradiation were performed on the same sample for each experiment. Before laser scanning, the cells were incubated overnight (16−20 h). After laser scanning, the cells were incubated for 4 h before microscopy and 2 h before flow cytometry. Fluorescent Microscopy and Cell Counting. Fluorescence microscopy was done on an upright microscope to image Calcein green and Calcein red-orange AM. Automated cell counting of fluorescent images was done using Fiji Image processing software using fluorescent images of the samples. The final cell count was checked to make sure that none of the cells were missed. Additional cells were added in the cell counter window. The following definitions were used to calculated percentages, viability = NCAMRed /Naverage and efficiency = NGreen /Naverage, where NCAMRed is the number of live cells (in red) in the field of view, Naverage is the average number of live cells in a laser-irradiated area, and NGreen is the number cells that appear green due to delivery of green molecules. Triplicate experiments were performed for all data sets. Cell Fixing. Cells were fixed immediately after fluorescence microscopy, which was about 4.5 h after laser scanning. Samples were soaked in fixing solution (1 part 25% glutaraldehyde, 1 part 1 M Hepes, 8 parts Millipore water) for at least 10 min. Samples were soaked in buffer solution (2 parts 1 M Hepes, 8 parts Millipore water) and wrapped with parafilm before overnight storage in the refrigerator at 5 °C. The next day, the samples were rinsed by soaking in Millipore water for at least 5 min. Samples were soaked in 1% osmium for at least 20 min and then rinsed with water for 5 min. The samples were then dehydrated with ethanol by soaking in 50% ethanol for 5 min, 70% ethanol for 5 min, 90% ethanol for 5 min, 100% ethanol three times, 7 min each (total: 20−30 min), and 100% ethanol with molecular sieves (grade: 3 Å) for 7 min. Samples were finally soaked in HMDS three times, at least 7 min each (total 20−30 min), before air drying at room temperature. Flow Cytometry. Flow cytometry measurements were performed to measure viability and delivery efficiency of different cargos to thousands of cells. Each FITC-cargo to be studied (Table S2) had six samples associated with it, including six control samples and six experimental samples using the conditions shown in Table S3. For flow cytometry, we used the maximum FITC−dextran concentration recommended by the manufacturer of 25 mg/mL to ensure the detection of the smallest levels of fluorescence in a cell in the flow cytometry scatter plots. The samples were washed with 10 mL of PBS and then 1 mL of PBS. Two mL of trypsin was added in a 35 mm dish and incubated for 7 min before being neutralized with 5 mL of prewarmed cell medium. Cells were transferred to a 15 mL tube and centrifuged at 125g for 5 min. The supernatant was removed, and cells were resuspended in 1 mL of PBS and pipetted up and down 30 times before being transferred to a 5 mL round-bottom test tube with a cell strainer snap cap. Flow-cytometry measurements were done using BD LSRFortessaSORP cell analyzer running BD FACSDiva software

version 6.1.3, using the optical configuration in Table S4. Measurements were done following the daily instrument QC, which utilizes BD CS&T beads (BD Biosciences catalog no. 642412) in the following order: B0, G0R0, G1, E1-E3. Appropriate gates were set for each sample using flow cytometry analysis software version FCS Express 5 Research Edition (DeNovo Software, Glendale, CA). Ten thousand events were recorded for each sample. Figure S8 shows representative data from dextran 10 kDa delivery experiments. The ungated side scattering area (SSC-A) vs forward scattering area (FSC-A) density plot represents the lightscattering distribution of cells. SSC-A corresponds to intracellular composition and FSC-A to size. The gate was set broadly for a monoculture population and serveed to eliminate debris. A sidescattering height (SSC-H) vs side scattering width (SSC-W) density plot, gated on gate 1, was used to exclude possible doublets (cells stuck together). A SSC-A vs Calcein AM red-orange (561 nm laser with a 582 ± 15 nm band-pass filter) density plot, hierarchically gated on gate 2, was used to gate on only live cells (Calcein AM red-orange positive). The resulting gate, gate 3, provided the viability statistics. A SSC-A vs FITC−dextran 10 kDa (488 nm laser with a 530 ± 30 nm band-pass filter) density plot, hierarchically gated on gate 3, was used to quantify the number of cells in the live population that had FITCcargo inside them. The control sample was used to set the gate for the experimental samples, where generally