FeSe quantum dots for in vivo multiphoton biomedical imaging

See allHide authors and affiliations

Science Advances  06 Dec 2019:
Vol. 5, no. 12, eaay0044
DOI: 10.1126/sciadv.aay0044


An immense demand in biomedical imaging is to develop efficient photoluminescent probes with high biocompatibility and quantum yield, as well as multiphoton absorption performance to improve penetration depth and spatial resolution. Here, iron selenide (FeSe) quantum dots (QDs) are reported to meet these criteria. The synthesized QDs exhibit two- and three-photon excitation property at 800- and 1080-nm wavelengths and high quantum yield (ca. 40%), which are suitable for second-window imaging. To verify their biosuitability, poly(ethylene glycol)-conjugated QDs were linked with human epidermal growth factor receptor 2 (HER2) antibodies for in vitro/in vivo two-photon imaging in HER2-overexpressed MCF7 cells and a xenograft breast tumor model in mice. Imaging was successfully carried out at a depth of up to 500 μm from the skin using a nonlinear femtosecond laser at an excitation wavelength of 800 nm. These findings may open up a way to apply biocompatible FeSe QDs to multiphoton cancer imaging.


Transition metal chalcogenides are attractive in many areas of research in nanoscience because they are useful as magnetic semiconductors (1), superconductors (24), photovoltaics (5, 6), electrocatalysts (7), sensors (8), and quantum dots (QDs) (9, 10). Layered iron-based materials are, in particular, regarded as promising superconductor candidates owing to their low toxicity and cost, as well as their unexpectedly high superconducting transition temperature in spite of the strong magnetism of iron (11, 12). In addition, iron chalcogenide materials become fluorescent nanosemiconductors when their dimension reduces to zero, and the unique optical and electronic properties of the materials have broadened their application range into biological imaging and solar power conversion (13, 14).

Fluorescent biomedical imaging by semiconducting nanocrystals is among the most promising detection techniques because of the high photostability and tunability of the nanocrystals in both absorption and emission spectra compared to those of conventional organic dyes (15). In addition, QDs exhibit the property of multiphoton-excited photoluminescence (PL), whereby one luminophore absorbs more than two photons at the same time through a virtual state (16) and emits visible light. This phenomenon has attracted considerable attention as it gives a way to improve spatial resolution as well as penetration depth because the longer wavelengths of exciting photons would reduce tissue autofluorescence and scattering (17); therefore, this is beneficial for in situ biomedical fluorescence imaging in cancer surgery (18, 19). Thus, multiphoton microscopy (MPM) has been successfully used as a noninvasive in vivo deep-tissue imaging tool (2024).

Here, motivated by previous works (14, 19, 24, 25), biocompatible QDs with two- and three-photon luminescent properties were new synthesized using iron and selenium components. Iron and selenium precursors naturally exist in the human body and exhibit low toxicity in a nanoparticle form (26, 27). In vitro tumor cell targeting specificity was evaluated using humanized monoclonal human epidermal growth factor receptor 2 (HER2) antibody–conjugated iron selenide (FeSe) QDs (anti-HER2–QDs) in HER2-overexpressed MCF7 cells, and in vivo MPM imaging was conducted in a live xenograft mouse model of human breast tumor.


Water-soluble FeSe QDs were synthesized via a one-pot strategy (Fig. 1A). The average size of the QDs was 3.4 ± 0.3 nm based on 200 counts in bright-field transmission electron microscopic (TEM) images (Fig. 1, B to D), and their hydrodynamic diameter was ca. 10 nm (fig. S1A). The high-resolution TEM image and electron diffraction patterns of the QDs indicated that the interplanar distance of 0.229 nm matches the (111) plane of tetragonal FeSe [Powder Diffraction File (PDF) number 16-2901]. The morphology of FeSe QDs was assumed to be pucklike polygons, analogous to that of FeSe nanoparticles in a previous report (14). Structural analysis using a grazing incidence x-ray diffraction (GIXRD) spectrometer and x-ray photoelectron spectroscopy further proved that the QDs were made up of tetragonal FeSe (Fig. 1E). The binding energy of the QDs agrees well with the PbO-type FeSe (28, 29), in which the Fe 2p 3/2 and Se 3d peaks were shifted to ca. 711.0 and 55.7 eV, respectively (fig. S1, B and C). The conjugation between the QDs and glutathione (GSH) was confirmed by comparing the Fourier transform infrared (FTIR) spectra of pure GSH and GSH-capped QDs (Fig. 1F). The S─H stretching vibration band at 2524 cm−1 disappeared in the FTIR spectrum of the GSH-capped FeSe QDs by forming an Fe─S coordinate bond, which means that GSH molecules were conjugated to the FeSe QDs via thiol bonding. Besides, the absorption band at 1713 cm−1 corresponds to the >C═O stretching mode of the carboxyl group of GSH (30). Zeta potential assessment indicated that the FeSe QDs dissolved in deionized water, 0.01 M phosphate-buffered saline (PBS), and 0.1 M PBS had zeta potentials of −54, −18, and − 14 mV, respectively (Fig. 1G), suggesting that the QDs had enough charge to be dissolved in highly concentrated PBS for bioexperiments. Furthermore, neither substantial aggregation nor notable difference in the fluorescence of the QDs was detected after they were monitored by a digital camera and fluorescence spectroscopy, respectively, for 5 days (fig. S2). The bandgap of FeSe QDs was approximated to be 2.44 eV by calculating the Kubelka-Munk transformation from the ultraviolet-visible (UV-Vis) spectrum (fig. S1D) (31).

Fig. 1 Physicochemical characterization of one-pot–synthesized water-soluble FeSe QDs.

(A) Schematic illustration of the one-pot synthesis procedure for GSH-capped FeSe QDs, (B) bright-field TEM image (inset: histogram of size distribution), (C) high-resolution TEM image, (D) fast Fourier transform of high-resolution TEM image, (E) GIXRD patterns, (F) FTIR spectra, and (G) zeta potential of FeSe QDs. a.u., arbitrary units.

The PL properties of FeSe QDs were explored at 25°C (Fig. 2). The inset in Fig. 2A shows digital images of FeSe QDs dispersed in deionized water, and the relative quantum yield of the FeSe QDs was around 40%. The PL lifetime (τ) of the QDs was 3.23 ± 0.04 ns, and the emission peak (λem) appeared at approximately 440 nm under an excitation (λex) of 365 nm (Fig. 2B). Two-photon (2PL) and three-photon (3PL) excited PL were clearly observed, whereby light was emitted at around 440 nm (cyan blue) at an excitation of 800 and 1080 nm, respectively (Fig. 2, C and D). The insets of Fig. 2 (C and D) show representative fluorescence microscopic images of MCF7 cells stained with FeSe QDs from 2PL and 3PL. Figure 2E shows the polynomial fitting (plotted on a logarithmic scale) of the emission intensity depending on the input excitation laser power. The slopes at excitation wavelengths of 800 and 1080 nm were approximately 1.98 and 3.08, corresponding to 2PL and 3PL, respectively (Fig. 2F) (32). This multiphoton excitation property is considered a notable technology, especially for bioimaging because excitation light with a longer wavelength can penetrate a maximum tissue depth that is sevenfold deeper, with reduced phototoxicity. For example, the imaging depth limit of traditional confocal 1PL imaging is ca. 200 μm, but the maximum penetration depth of multiphoton imaging is much greater. Shi et al. reported 2PL imaging under 800-nm excitation, which enables deep-tissue imaging up to 450 μm in the rat brain tissue (33). Horton et al. and Nwaneshiudu et al. performed mouse neocortex imaging by 2PL (λex = 775 nm) and 3PL (λex = 1675 nm) at a tissue depth of up to 700 and 1400 μm, respectively (34, 35). Brain tissues have minimal absorbance in window III from 1600 to 1870 nm, which is the so-called “golden window” (36, 37).

Fig. 2 Optical characterization of FeSe QDs.

(A) PL lifetime (τ) of FeSe QDs at an excitation wavelength of 380 nm; inset: digital images of FeSe dispersion under white light and UV lamp (λex = 365 nm). Normalized PL excitation (PLE) spectrum (black line) and PL spectrum (red line) at λem of 440 nm and λex of (B) 365 nm for 1PL, (C) 800 nm for 2PL, and (D) 1080 nm for 3PL. (E) Power dependence of PL intensity for 2PL (black square) and 3PL (red square). The slope of the power-dependence function is 1.98 and 3.08 for 2PL and 3PL, respectively. (F) Jablonski diagram of single-, two-, and three-photon luminescence.

Before applying FeSe QDs to bioimaging experiments, the QD influence on cell viability was assessed by integrating the growth of different cell lines (AGS, MG63, and NCI-H460) with FeSe QDs at various culture durations and QD concentrations (Fig. 3, A to C). The various cell lines exhibited excellent viability in the presence of QDs during 7 days of culture, with >75% cell viability at a QD concentration of up to 70 μg ml−1. Fluorescence microscopic images of AGS, MG63, and NCI-H460 cells revealed that the FeSe QDs had a superior biocompatible nature regardless of their concentration. The QDs did not interfere with cell proliferation, as the cells treated with QDs showed similar proliferation as that of nontreated control cells for up to 7 days (fig. S3).

Fig. 3 In vitro and in vivo two-photon microscopic imaging of FeSe QDs targeted to breast tumor.

Flow cytometry evaluation of the viability of (A) AGS, (B) MG-63, and (C) NCI-H460 cells exposed to QDs at various concentrations (0, 25, 50, and 70 μg ml−1) for 3, 5, and 7 days. (D) Conjugation procedure to prepare anti-HER2–PEG-QDs. (E) In vitro two-photon microscopic imaging of MCF7 and HER2-overexpressed MCF7 cells (MCF7/HER2) stained with PEG-coated FeSe QDs or anti-HER2–conjugated PEG-QDs (anti-HER2–PEG-QDs, 2 μg ml−1), where the nuclei were dyed with propidium iodide, and the cell membrane and nuclei were imaged at λex of 800 and 500 nm. Laser power = 40 mW at the focal plane. (F) Comparison of the photostability of QDs and rhodamine 6G (Rh6G) in deionized water under two-photon excitation (λex = 800 nm, laser power = 50 mW), where relative PL intensity was monitored for 30 min. (G) Digital photograph of tumor xenograft for in vivo imaging. (H) MPM system. CH PMT, channel photomultiplier tube; OPO, optical parametric oscillator. (I) In vivo MPM images before and after anti-HER2–QD injection and (J) in vivo MPM images at different focal depths (450 to 500 μm). Scale bars, 20 μm.

To minimize nonspecific binding during the assembly (38, 39), FeSe QDs were encapsulated with poly(ethylene glycol) (PEG) before conjugation with HER2 antibodies to form anti-HER2–PEG-QDs (Fig. 3D). To evaluate the nonspecific uptake and selectivity of anti-HER2–PEG-QDs for human breast cancer cell targeting, both HER2-overexpressed MCF7 and HER2-nonexpressed cells were stained with anti-HER2–PEG-QDs and PEG-QDs that were not conjugated with anti-HER2. The nuclei of the cells were stained with propidium iodide. All samples exhibited red fluorescence signals from propidium iodide in the nuclei and only HER2-overexpressed MCF7 cells stained with anti-HER2–PEG-QDs displayed cyan fluorescence (Fig. 3E). Neither cellular uptake nor cell surface binding was observed in the other control samples, indicating that the anti-HER2–PEG-QDs specifically targeted the HER2 receptors (fig. S5 and movies S1 and S2). Furthermore, QD fluorescence was mainly observed in the cell membrane and cytoplasmic region, whereas it was negligible in the nucleus. Without PEGylation, both the physically adsorbed and covalently bound anti-HER2–QDs did not show distinct selectivity to the HER2-overexpressed MCF7 cells against HER2-nonexpressed cells, suggesting the nonspecific binding behavior of anti-HER2–QDs at concentrations of 20 and 100 μg ml−1 and the internalization of QDs in the cell cytoplasm. In other words, anti-HER2–QDs were internalized in normal and HER2-overexpressed cells depending on the QD concentration without selectivity (fig. S4). These in vitro imaging results implied the potential of the PEGylated QDs as candidate in vivo imaging agents. The anti-HER2–PEG-QDs were physiologically stable and maintained their optical properties for 7 days in serum and in various buffer solutions (fig. S6), with <1% hemolysis rate at a concentration of 100 μg ml−1 (fig. S7). Inorganic QDs usually show superior photostability to any organic dye, but FeSe QDs were highly photostable when subjected to two-photon excitation. Their change in emission was negligible with laser excitation (λex = 800 nm) at 50 mW over a period of 30 min, which is important in biological imaging for long-term tracking of targeted cells. Note that rhodamine 6G showed almost 100% loss in intensity after laser exposure for 30 min under the same conditions (Fig. 3F).

Breast cancer is the second highest cause of cancer death for women with significant recurrence rates (40). Minimally invasive surgery aided by sensing and imaging techniques is a crucial tool for identifying the disease. In this study, imaging using FeSe QDs and nonlinear femtosecond laser was demonstrated. In vivo intravital MPM imaging was performed via intravenous injection of anti-HER2–PEG-QDs in an MCF7 xenograft animal model. The subcutaneous xenograft mouse model of breast cancer was established by injecting MCF7 and MCF7/HER2 cells into the flank of mice. After 4 weeks, the tumor volume reached 200 mm3, and 100 μl of anti-HER2–PEG-QDs (0.5 mg ml−1) was intravenously injected into the mice (Fig. 3, G and H). The data acquired at a depth of 480 μm from the surface showed two-photon signals in the vicinity of the MCF7 cells, which were recognized by green fluorescence protein (GFP) and indicated successful QD targeting in the tumor (fig. S8). The blue signals indicated the second harmonic generation (SHG) induced by fibrillar collagens in the dermis. The FeSe QDs were observed as a magenta signal, which was converted from cyan by a color threshold method for better distinction (Fig. 3I and fig. S9). The 2PL signal at different depths (450 to 500 μm from the surface) in the tumor area was acquired in the internal dermal layer at regular 10-μm intervals (Fig. 3J and movie S3). The SHG signal from collagen was dominant in the superficial area, and the PL signal from the QDs was clearly distinguishable near the breast cancer cells, which have green fluorescence from the GFP, at a depth of 480 to 500 μm.


In conclusion, biocompatible FeSe QDs were successfully synthesized, and high cell viability was observed at high QD concentrations (up to 70 μg ml−1). The QDs were used in two- and three-photon fluorescence imaging, whereby in vivo live animal multiphoton imaging at a depth of up to 500 μm from the skin surface was clearly visualized to monitor tumor cells using a nonlinear femtosecond laser. The biocompatible FeSe QDs and multiphoton imaging possibly open a way to realize noninvasive in situ bioimaging of live subjects.



High-resolution TEM images and energy-dispersive spectra were obtained using a JEOL microscopy (JEM-2100F; ZEOL Ltd., Tokyo, Japan) operating at an accelerating voltage of 200 kV at the Korea Basic Science Institute. The 1PL and 1PLE spectra were acquired by using fluorescence spectroscopy (Hitachi F-7000, Tokyo, Japan). The extinction spectra of the QDs were measured by a UV-Vis spectrometer (S310; SCINCO Inc., Daejeon, Korea). The PL lifetime (τ) of the QDs was measured using a lifetime spectrophotometer 3113 (PTI Inc., Princeton, NJ).

One-pot synthesis of water-soluble fluorescent FeSe QDs

All reagents used in this work including Fe(ClO4)2⋅xH2O (98%), GSH (≥98%), and Na2SeO3 (≥98%) were of analytical grade (Sigma-Aldrich, St. Louis, MO). In a typical synthesis of FeSe QDs, Fe(ClO4)2⋅XH2O (0.066 g, 0.26 mmol) and GSH (0.32 g, 1.04 mmol) were added to dimethyl sulfoxide (40 ml) in a three-neck flask (100 ml). Then, the system was heated to 140°C with vigorous stirring (700 rpm) under N2 atmosphere. After 10 min of magnetic stirring, Na2SeO3 solution (5 ml, 52 mM) was added and maintained at the same temperature with continuous stirring for 20 hours. After cooling the reaction mixture to 25°C, dark brown precipitates were obtained. The precipitates were then purified with isopropanol by centrifugation (15,000 rpm, 30 min). The resulting samples were dried into powder form in a vacuum oven at 25°C for further usage.

PEGylation of FeSe QDs

PEG was covalently bound to GSH-capped FeSe QDs through typical N′-ethylcarbodiimide hydrochloride (EDC)–N-hydroxysuccinimide (NHS) chemistry. PEG diacid (Mn: 6000 g/mol) was used to produce GSH-capped FeSe QDs. EDC solution (2 ml, 50 mM) was added to PEG diacid (2 ml, 50 mM), and after 2 min of sonication, NHS solution (2 ml, 125 mM) was added to the reaction mixture and vortexed for 5 min. This reaction solution was added to the FeSe QD dispersion (10 ml, 20 μM), and the mixture was stirred for 1 hour. Then, the dispersion was dialyzed for 48 hours using a dialysis bag [molecular weight cutoff (MWCO): 1 kDa].

Conjugation of FeSe QDs with HER2 antibodies

A humanized monoclonal HER2 antibody (trastuzumab, Herceptin; Roche, Mannheim, Germany) was purchased. PEG-QDs were covalently bound with the amine terminal of the HER2 antibody. EDC and NHS (5 mg) were dissolved in MES buffer (100 μl, 0.1 M), and the solutions were added to the FeSe dispersion (10 ml, 15 μM). The mixture was vigorously stirred for 30 min, and anti-HER2 (1 mg) dissolved in PBS (300 μl, 0.1 M) was mixed and stirred for 1 hour. On the basis of the calculation from the geometrical fully packaging model, a maximum loading ratio of anti-HER2 to the single PEG-FeSe QDs is 18:1 (41, 42). Last, the anti-HER2–PEG-QD dispersion was dialyzed for 48 hours (MWCO: 1 kDa).

In vitro MPM imaging of breast cancer cells using FeSe QDs

MCF7 is a human breast cancer cell line isolated from malignant breast adenocarcinoma tissue. The HER2 gene was introduced into MCF7 cells via lentiviral transfection so that they stably expressed HER2, and the transfected cells were denoted as MCF7/HER2. To identify the cell surface–expressed HER2, fluorescent images of cells were obtained using a confocal microscope (LSM780, Zeiss, Germany). The cells (3 × 104) were seeded on an eight-well–chambered slide and incubated at 37°C for 48 hours. After several washes with PBS, the cells were fixed with paraformaldehyde (4%) and stained (anti-HER2–Alexa647; BioLegend, San Diego, CA). The mounted cells were examined by confocal microscopy and analyzed using Zen software. To determine whether anti-HER2–PEG-QDs targeted HER2 expressed on the cell membrane, MPM images were obtained using a custom-built MPM microscope. MCF7 and MCF7/HER2 cells (3 × 104 cells per well) were seeded on an eight-well–chambered slide and incubated at 37°C for 48 hours. After washing three times with PBS, the cells were fixed with paraformaldehyde (4%) and stained with anti-HER2–PEG-QDs and PEG-QDs (as a negative control). Then, the mounted cells were examined using a custom-built MPM system.

Two- or three-photon excitation fluorescence images were acquired using a custom-built MPM system, which contained a Ti-sapphire laser-tunable femtosecond pulse laser (680 to 1080 nm, 140 fs, 80 MHz) and an optical parametric oscillation-based tunable femtosecond pulse laser (1050 to 1600 nm, 200 fs, 80 MHz). Two-dimensional (2D) imaging was achieved by raster scanning the laser beam with X-Y galvanometers. At different depths, 2D images were obtained by translating the location of the focal plane along the z axis using a motorized stage. Consequently, an entire 3D image was created in this manner by integrating a series of 2D images. The laser beam was reflected toward the objective lens and focused onto the sample using a 20× water-immersion objective with a high numerical aperture (XLUMPLFLN 20XW, NA 1; Olympus, Tokyo, Japan). The 2PL and 3PL signals were generated by the focused beam, and the PL signal from the sample was recollected in the opposite direction. The recollected signals were divided by a long-pass dichroic mirror set and band-pass filters for three channels: red (560 to 630 nm), green (490 to 550 nm), and blue (420 to 480 nm). Signals separated by color were lastly detected by photomultiplier tubes (H10682-210; Hamamatsu Photonics, Shizuoka, Japan), which are photosensitive detectors. The photomultiplier tubes were used for photon counting of fluorescence signals. Fluorescence signal acquisition was carried out using a personal computer–based data acquisition system.

In vivo MPM imaging of breast tumor xenograft using FeSe QDs

Female Hsd:Athymic nude–Foxn1 nude mice (from 6 weeks old) were purchased from Envigo (Gannat, France). All animal studies were approved by the Institutional Animal Care and Use Committee of the National Cancer Center Research Institute (NCCRI) (NCC-16-163C and NCC-16-163D). The NCCRI is a facility approved by the Association for Assessment and Accreditation of Laboratory Animal Care International and abides by the guidelines of the Institute of Laboratory Animal Resources (accredited unit: NCCRI unit number 1392). The subcutaneous xenograft mouse model of breast cancer was established via injection of MCF7 and MCF7/HER2 cells (5 × 106) with Matrigel (Corning Inc., NY) into the flank of the experimental mice. Four weeks after inoculation, the tumor volume reached 200 mm3 and anti-HER2–PEG-QDs (100 μl, 0.5 mg ml−1) were intravenously injected into the mice.

In the custom-built MPM system, imaging data were recorded separately for the blue, green, and red channels. The data obtained by each color were grayscale data that carried only information on intensity. Therefore, the red, green, and blue (RGB) multicolor imaging was performed using imaging software. The FeSe QDs were identified in cyan around HER2-overexpressed MCF7 cells. The color replacement method was applied, whereby cyan was extracted using the color threshold method and separated from the original data. The intensity of the extracted cyan FeSe QDs was replaced by magenta and combined with the original data. This color replacement method was applied in images of both HER2-overexpressed and normal MCF7 cells, and the final images showed better contrast and distinction (Fig. 3, I and J). Moreover, the color threshold method was used to remove the SHG signal of collagen caused by a laser with a wavelength of 800 nm (figs. S8 and S9).


Supplementary material for this article is available at

Supplementary Texts

Fig. S1. Physicochemical characterization of FeSe QDs.

Fig. S2. Stability assessment of FeSe QDs dissolved in deionized water and PBS at 0.01 and 0.1 M.

Fig. S3. Fluorescein diacetate assay.

Fig. S4. Tumor cell targeting specificity of anti-HER2–QDs.

Fig. S5. Tumor cell targeting specificity of anti-HER2–PEG-QDs.

Fig. S6. Evaluation of physiological stability of anti-HER2–PEG-QDs in various conditions.

Fig. S7. Hemocompatibility assay of FeSe QDs.

Fig. S8. In vivo two-photon microscopic images before and after SHG signal removal.

Fig. S9. Schematic diagrams presenting color replacement and removal process by using color threshold method.

Movie S1. In vitro two-photon microscopic images of HER2-overexpressed MCF7 cells (positive control) after treatment with anti-HER2–PEG-QDs (20 μg ml−1).

Movie S2. In vitro two-photon microscopic images of MCF7 cells (negative control) after treatment with anti-HER2–PEG-QDs (20 μg ml−1).

Movie S3. In vivo two-photon microscopic images (Z-scan) of cancer part were obtained 30 min after intravenous tail vein injection of anti-HER2–PEG-QDs by moving the excitation focal plane from 450 to 525 μm from the skin in 5-μm steps (λex = 800 nm, power = 100 mW).

Reference (43)

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank the Korea Basic Science Institute for helping with TEM and GIXRD analysis. Funding: This work was financially supported by the National Research Foundation of Korea (NRF) under the Ministry of Science, ICT & Future Planning (grant nos. NRF-2017R1A4A1015627 and NRF-2019R1A2C2007825); the Korea Healthcare Technology R&D Project (grant no. HI17C1260) of the Ministry for Health, Welfare and Family Affairs; and Samsung Electronics. Author contributions: J. Kwon, S.W.J., and S.I.C. designed the study. J. Kwon synthesized and characterized the FeSe QDs. S.W.J. performed femtosecond laser–based multiphoton bioimaging. S.I.C. prepared a tumor xenograft mouse model. X.M., J. Kim, and E.K.K. performed experiments of synthesis, antibody conjugation, and cytotoxicity test, respectively. Y.-H.K., S.-K.K., D.Y.H., C.-S.K., and J.L. supervised the research. All authors contributed to the manuscript writing and editing. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the corresponding author according to the material transfer agreement.

Stay Connected to Science Advances

Navigate This Article