A generic approach towards afterglow luminescent nanoparticles for ultrasensitive in vivo imaging

Afterglow imaging with long-lasting luminescence after cessation of light excitation provides opportunities for ultrasensitive molecular imaging; however, the lack of biologically compatible afterglow agents has impeded exploitation in clinical settings. This study presents a generic approach to transforming ordinary optical agents (including fluorescent polymers, dyes, and inorganic semiconductors) into afterglow luminescent nanoparticles (ALNPs). This approach integrates a cascade photoreaction into a single-particle entity, enabling ALNPs to chemically store photoenergy and spontaneously decay it in an energy-relay process. Not only can the afterglow profiles of ALNPs be finetuned to afford emission from visible to near-infrared (NIR) region, but also their intensities can be predicted by a mathematical model. The representative NIR ALNPs permit rapid detection of tumors in living mice with a signal-to-background ratio that is more than three orders of magnitude higher than that of NIR fluorescence. The biodegradability of the ALNPs further heightens their potential for ultrasensitive in vivo imaging.

O ptical imaging that utilizes photon-electron interactions to decipher biological processes has grown into an indispensable tool in biomedical research and clinical practice 1 . Complementary to tomographic modalities such as magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET), optical imaging has the unique advantages of high spatial-temporal resolution and low cost, permitting real-time investigation of pathological processes at molecular level and sensitive detection of diseases for intraoperative imaging-guided surgery [2][3][4][5] . However, most optical techniques detect fluorescent signals generated upon real-time light excitation, wherein the background noise from endogenous molecules in biological subjects are inevitable 6 . Such a flaw challenges reliable detection of signals, giving rise to minimized signal-to-background ratio (SBR), limited penetration depth and consequently compromised imaging sensitivity 7,8 .
Real-time light-excitation-free optical agents including chemiluminescent, bioluminescent, Cerenkov, and afterglow (or persistent luminescent) probes can circumvent the interference of tissue autofluorescence 9 . However, each has its own trade-offs. For instance, chemiluminescent and bioluminescent agents utilize chemical reactions that, respectively, require reactive oxygen species and enzyme to catalyze the decomposition of substrates to trigger luminescence 10,11 , and their imaging sensitivity is usually perturbed by cellular environment and substrate availability 12,13 . On the contrary, Cerenkov and afterglow agents do not require particular chemical mediator or exogenous enzyme, and thus have higher versatility for imaging applications. However, Cerenkov agents are basically radioisotopes and intrinsically limited to only emit visible light 14,15 , thus their biomedical applications are challenged by both the biosafety issue of radiotracers and shallow imaging depth due to short-wavelength emission. Differently, afterglow agents act as the optical battery to trap irradiated photoenergy in defects and then slowly release the stored energy by photonic emission upon physical (thermal, mechanical, etc.) activation, eliminating the need of invasive radiotracers or exogenous mediators 16 .
Herein, we report a generic approach to transform ordinary fluorescent agents into afterglow luminescent nanoparticles (ALNPs) for in vivo imaging. This approach relies on an intraparticle cascade photoreaction of three key components termed as afterglow initiator, afterglow substrate, and afterglow relay unit (Fig. 1a) to store the photoenergy as the chemical defects for delayed luminescence after cessation of light excitation. Within ALNPs, a photosensitizer serves as the afterglow initiator to absorb and convert photoenergy into signaling singlet oxygen ( 1 O 2 ); a 1 O 2 -reactive molecule serves as the afterglow substrate to absorb and react with 1 O 2 , forming the unstable chemiluminescent intermediate (1,2-dioxetane); and a fluorescent agent behaves as the afterglow relay unit to accept the energy from 1,2dioxetane via chemically initiated electron exchange luminescence (CIEEL), gradually releasing it in the form of photons. Depending on whether there is an efficient secondary energy transfer (SET) between the fluorescent agent and the photosensitizer, the ultimate afterglow emission spectrum could be close to that of the fluorescent agent or the photosensitizer.
The afterglow contrast agents were prepared with different combinations of afterglow initiator, substrate, and relay unit through co-nanoprecipitation with an amphiphilic copolymer PEG-b-PPG-b-PEG (Fig. 1f, Supplementary Fig. 8). The doping ratios for each component within the ALNPs were optimized (Supplementary Figs. 9-10). The solutions of resulted 50 kinds of ALNPs were translucent with no obvious precipitates after preparation (Supplementary Figs. 11&12). Dynamic light scattering (DLS) revealed the hydrodynamic diameters of the ALNPs ranged from 80 to 180 nm (Fig. 2a, Supplementary Fig. 11b, c), except for CQD-based ALNPs (12 nm). This should be attributed to the intrinsic hydrophilicity and surface charge of CQD. Furthermore, transmission electron microscope (TEM) revealed the spherical morphology of these ALNPs (Fig. 2a).
Afterglow luminescence of these ALNPs were recorded with optimized light irradiation time (5 s) ( Supplementary Fig. 13). Because of different absorption ( Supplementary Fig. 14), NCBSdoped ALNPs were irradiated with 808 nm laser (1 W cm −2 ) while RB-or TPP-doped ALNPs were irradiated with white light (0.1 W cm −2 ). As expected, luminescence was detected from ALNPs after cessation of laser irradiation (Fig. 2d), which was barely detectable for the nanoparticles consisting of only afterglow initiator and substrate or relay unit (Fig. 2e, f). This validated the proposed afterglow mechanism and the collaborative roles of three components. Dependent on the compositions of ALNPs, the afterglow luminescence spectra ranged from visible to NIR region (Fig. 2b, c). In general, the afterglow spectra of ALNPs were similar to the corresponding fluorescence spectra . If there was energy transfer from afterglow relay unit to initiator, the shape of afterglow spectrum could be more like that of the initiator. Otherwise, the afterglow emission was closer to that of the relay unit. For instance, ALNPs consisting of PFO, NCBS, and DO (termed as PFO-N-DO) had a strong NIR afterglow emission at 780 nm because of the secondary energy transfer from PFO to NCBS; whereas ALNPs consisting of PFO, RB, and DO (termed as PFO-R-DO) only had the afterglow emission from PFO due to the inefficient energy transfer from PFO to RB. Discrepancy in the fluorescence and afterglow spectral profiles was observed for several ALNPs (especially TPP-doped ones) such as GQD-N-DO, DiO-TPP-DO (termed as DiO-T-DO), CQD-R-DO, etc. This should be ascribed to the fact that the afterglow photophysical process was different from that of fluorescence ( Supplementary Fig. 18): in the fluorescence process, light excitation only led to the emission of the fluorescent agent, which was followed by the potential energy transfer to the photosensitizer; whereas, in the afterglow process, in addition to such a potential energy transfer, the photosensitizer could be directly excited through the energy released from the high-energy intermediate (1,2-dioxetane). Thus, the photophysical interplay between the afterglow initiator and the relay unit offered additional space to fine-tune the afterglow profiles of ALNPs, potentially enabling multiplexed imaging (Supplementary Fig. 19).
The afterglow intensities of ALNPs were different from each other ( Fig. 2d-f). Comparison of the nanoparticles with the same afterglow initiator (NCBS) and relay unit revealed that DO-doped nanoparticles had the brightest afterglow luminescence among three afterglow substrates (Fig. 2e), which was followed by SO and HBA doped ones. Moreover, among all the tested fluorescence agents, PFVA-based nanoparticles had the highest afterglow intensities provided that other two components were the same. For instance, PFVA-N-DO gave the brightest afterglow luminescence among all NCBS-doped ALNPs, which was 12-and 123-fold higher than that of PFVA-N-SO, and PFVA-N-HBA nanoparticles, respectively. Variation of photosensitizer also impacted the afterglow luminescence of ALNPs, because the ability to generate 1 O 2 was different. NCBS-doped ALNPs generally had brighter afterglow luminescence than TPP-and RB-doped ones when other components were the same (Fig. 2f). However, this was not the case when the fluorescent agent was PFVA. For instance, afterglow intensities of PFVA-R-DO and PFVA-T-DO were 4.6-and 4.1-fold higher than that of PFVA-N- DO, respectively. This was because of the additional amount of 1 O 2 generated by PFVA under white light irradiation but not 808 nm irradiation. Thus, the afterglow intensities of ALNPs were determined by all three components, but independent of particle size if the components were the same ( Supplementary Fig. 20). Moreover, the afterglow could be repeatedly induced (Supplementary Fig. 21), or restored after preservation of pre-irradiated ALNPs at −20°C ( Supplementary Fig. 22).
Quantitative analysis and prediction of afterglow intensity. To quantitatively analyze the factors governing the afterglow intensity of ALNPs, a mathematic model was proposed. Relative afterglow intensity (Φ Afterglow ) was defined as the ratio of the afterglow luminescence of individual ALNP to that of the control ALNP which consisted of NCBS and DO without afterglow relay units (termed as N-DO). Based on the detailed afterglow process ( Supplementary Fig. 23), four descriptors were retrieved and defined for simulation. First of all, because afterglow process started from photosensitization, the production of 1 O 2 by afterglow initiator (Φs 1 ) was defined as the first descriptor (Fig. 3a).
As previously reported 39,40 , chemiluminescent property of afterglow substrate and the oxidation potential of fluorescent agent (afterglow relay unit) are important to CIEEL. Therefore, the chemiluminescent quantum yield of afterglow substrate after reaction with 1 O 2 (Φ Cl ) was selected as the second descriptor (Fig. 3b); moreover, corresponding to oxidation potential, the energy level of highest occupied molecular orbital (HOMO) of afterglow relay unit (E HOMO ) with respect to frontier molecular orbital theory (Fig. 3c, Supplementary Fig. 24) was defined as the third descriptor. Because the afterglow emission in this generic approach involved the potential SET between the afterglow relay unit and the afterglow initiator, the relative fluorescence efficiency of ALNP (η Fl ) was defined as the last descriptor. This descriptor was determined by both initiator and relay unit and was quan- According to statistical results ( Supplementary Fig. 26), all descriptors were strongly correlated with Φ Afterglow (Supplementary Fig. 26c, P < 0.05). The calculated Eq. (1) showed an impressive coefficient of determination (R 2 ) of 0.944 (adjusted R 2 = 0.936), and the measured and simulated Φ Afterglow values involved in quantitative analysis showed close proximity to each other (Fig. 3d), both suggesting the excellent fitness of Eq. (1) to this afterglow model. Based on Eq. (1), it was apparent that Φ Afterglow demonstrated non-linear increment with four descriptors related to the cross-talk of three major afterglow components. Basically, increased production of 1 O 2 , chemiluminescence of afterglow substrate, HOMO energy level of afterglow relay unit, or fluorescent efficiency of fluorescent units in ALNP could contribute to brightening afterglow. Such a pattern corresponded well with the experimental data in Fig. 2. Thereby, these statistical data demonstrated the rationality of descriptor selection for quantitative analysis and implied the feasibility of Eq. (1) for afterglow prediction.
To test the predictive capability of the proposed equation, another fluorescent agent, CPV, was used as the afterglow relay unit. Note that CPV hardly emitted afterglow luminescence by itself 22 . However, after the nanoformulation through doping CPV with afterglow initiator (NCBS) and substrate (DO, SO, or HBA), intense afterglow luminescence was detected ( Supplementary  Fig. 27). The comparison of experimentally measured and Eq. (1) simulated Φ Afterglow of CPV-based ALNPs was shown in Fig. 3e. Impressively, no significant difference was observed between the measured and estimated Ln[Φ Afterglow ] (P > 0.05), validating the prediction reliability of Eq. (1) to estimate afterglow luminescence for this generic afterglow approach.
Tissue penetration of afterglow luminescence. To assess the imaging capability of ALNPs, we examined the imaging performance of ALNPs in comparison with NIR fluorescence at different tissue depths. Moreover, the afterglow performance was benchmarked against the reported afterglow agent SPN-NCBS5. SPN-NCBS5 was similarly prepared via nanoprecipitation, wherein MEHPPV was doped with 5 w/w% NCBS using PEG-b-PPG-b-PEG as the matrix 22 . Considering strong afterglow  [28][29][30]. With the increase of tissue depth, both NIR fluorescence and afterglow luminescence intensities from the buried ALNPs significantly decreased (Fig. 4a). Because of the minimized background noise of afterglow imaging, the SBRs of afterglow images were remarkably higher than those of NIR fluorescence images at all tissue depths (Fig. 4b). Notably, the NIR fluorescence was almost undetectable at 3 cm (SBR close to 1), whereas the afterglow luminescence could still be clearly visualized (SBR: 61 ± 8; SBR is expressed as mean ± standard deviation of three independent measurements). These data suggested the superior imaging performance of afterglow luminescence over NIR fluorescence. Moreover, depending on afterglow substrate, PFVA-N ALNPs showed similar (PFVA-N-HBA) or even higher (PFVA-N-DO/ SO) afterglow SBRs than SPN-NCBS5 at the same tissue depth.
These results not only indicated the advantage of PFVA-N ALNPs over SPN-NCBS5 for afterglow imaging, but also emphasized the design flexibility of ALNPs to further promote penetration depth and imaging sensitivity. The deep-tissue imaging capability of PFVA-N ALNPs was further validated in living animals. As illustrated in Fig. 4c, nanoparticle solutions were placed beneath a living mouse wherein the tissue depth was measured to be 1.5 cm. Because of the strongest afterglow intensity and low background noise (Fig. 4d), the afterglow of PFVA-N-DO had the highest SBR (1136 ± 19), followed by PFVA-N-SO (272 ± 3) and PFVA-N-HBA (138 ± 1) (Fig. 4e). Remarkably, the afterglow SBR of PFVA-N-DO exceeded that of SPN-NCBS5 (131 ± 3) by 8.7 times. On the other hand, the NIR fluorescence from PFVA-N-DO could hardly be differentiated from the background (tissue autofluorescence). These data thus corresponded well with in vitro tissue penetration study, validating the ability of ALNPs for ultrasensitive deep-issue afterglow imaging.  (Fig. 5a). In contrast, the NIR fluorescence SBR slightly increased (Fig. 5b), and thus the tumor was only detectable at 4 h post injection. Note that at 1 h post injection, the afterglow SBR in tumor region (2922 ± 121) was three orders of magnitude higher than that of NIR fluorescence (~1). Such a afterglow SBR was not only higher than fluorescence imaging in both first and second NIR window (SBRs up to~135) 43 , but also significantly exceeded the SBRs of other excitation-free imaging modalities including chemiluminescence (up to~20) 44 , bioluminescence (up to~1000) 10 , and Cerenkov luminescence imaging (up to~154) (Supplementary Table 1) 45 . This should be mainly attributed to the fact that chemiluminescence, bioluminescence, and Cerenkov luminescence imaging usually rely on visible emission. Ex vivo data revealed that the uptake of PFVA-N-DO in tumor was 0.58-fold of that in liver ( Supplementary Fig. 31), further confirming its ability to passively target tumor. These data highlighted that by PFVA-N-DO mediated afterglow imaging allowed for rapider detection of tumor with the superior contrast and sensitivity over other optical agents.
Biodegradation and in vivo clearance study were subsequently performed to examine the biosafety of PFVA-DO-N. To mimic in vivo environment, myeloperoxidase (MPO) abundantly expressed in phagocytes was used as the oxidative enzyme for in vitro biodegradation (Fig. 5c). In the presence of hydrogen peroxide (H 2 O 2 ), MPO catalyzes the production of hypochlorous acid (HClO) to digest foreign substances 46    at 450 nm assigned to PFVA significantly dropped (Fig. 5d), suggesting the fragmentation of PFVA by MPO. The biodegradation was further confirmed by gel permeation chromatography (GPC), as indicated by the evidently decreased molecular weight of PFVA after MPO treatment (Fig. 5e). Such an efficient degradation should be ascribed to the oxidation induced cleavage of double bonds in the conjugated backbones of PFVA (Fig. 5c), which was previously reported 22,30 .
To monitor the in vivo clearance of PFVA-N-DO, they were systemically administered into mice followed by long-term NIR fluorescence recording (Fig. 5f). After administration, NIR fluorescent signals from liver increased over time and reached the maximum at 3 days post injection. Later, the NIR fluorescence from liver continuously decreased to almost undetectable level at 33 days post injection (Supplementary Fig. 32). These results indicated the long-term clearance of PFVA-N-DO via hepatobiliary excretion in living animals. Furthermore, no noticeable histological damage was observed in the major organs of living mice after systemic administration of PFVA-N-DO for 33 days (Fig. 5g), suggesting the good biocompatibility of PFVA-N-DO.

Discussion
In summary, we reported a generic approach that transformed traditional fluorescent agents into a new library of afterglow agents (ALNPs). By virtue of an efficient intraparticle cascade photoreaction of three key components (afterglow initiator, substrate, and relay unit), ALNPs were able to chemically store the photoenergy and spontaneously emit long-lived luminescence after cessation of optical excitation. Such a facile approach was applicable to a wide range of compositions: RB, TPP, or NCBS for the initiator, DO, SO, or HBA for the substance, and inorganic or organic fluorophores for the relay units. The sophisticated but fairly controllable photochemical interactions within the nanoparticles allowed to fine-tune the afterglow emission from visible to NIR region by adjusting the ALNP compositions. To elucidate the factors involved in this generic afterglow process, a 4descriptor based mathematical model (Eq. (1)) was generated from supervised learning analysis, which accurately predicted the afterglow intensities of unknown ALNP composition. Using PFVA-N-DO ALNPs as an example, the afterglow achieved a maximal imaging depth at 5 cm in biological tissue, deeper than the reported afterglow agents (4 cm). As compared with NIR fluorescence, the afterglow of ALNPs exhibited three orders of magnitude higher SBR (2922 ± 121), allowing for rapider detection of tumor in living mice after systematic administration. To the best of our knowledge, this is the highest SBR achieved so far for in vivo optical imaging regardless of their optical modalities and detection wavelengths. In conjunction with the heavy-metalfree benign nature, the representative PFVA-N-DO ALNPs were enzymatically biodegradable and clearable with a good long-term biocompatibility, further ensuring their in vivo applications. Thus, our study showed a controllable nanoengineering approach nearly applicable to all kinds of fluorophores regardless of their composition for background-free molecular optical imaging.

Methods
Chemicals and characterization. All materials were purchased from Sigma-Aldrich Pte. Ltd. unless otherwise noted. PFVA, PFO, and PFBT were purchased from Luminescence Technology Corp. DO and SO were purchased from Aberjona Laboratories, Inc.
DLS profiles of nanoparticles were measured by Malvern Nano-ZS Particle Sizer. TEM images of nanoparticles were captured by JEOL JEM 1400 TEM with an acceleration rate of 100 kV. Proton nuclear magnetic resonance ( 1 H NMR) spectra were measured by Bruker Avance 300 MHz NMR. Electrospray ionization mass spectrometry (ESI-MS) spectrum of RB was measured by ThermoFinnigan LCQ Fleet MS equipped with Themo Accela LC and ESI source. Absorption spectra were measured on a Shimadzu UV-2450 spectrophotometer. Fluorescence spectra and fluorescence efficiency were acquired on a Fluorolog 3 spectrofluorometer (HORIBA, Ltd.). Chemiluminescence of afterglow substrates was recorded with a Luminometer (Promega, USA). Molecular weights of PFVA ALNPs in biodegradation studies were characterized by GPC using THF as the eluent and polystyrene as the standard. White light source for afterglow luminescence was supplied by an LED Fiber Optic Illuminator (L-150A) with an output power density of 0.1 W cm −2 (wavelength range: 400-800 nm). Fiber coupled 808 nm laser system was purchased from Changchun New Industries Optoelectronics Tech. Co., Ltd. NIR fluorescence and afterglow images were acquired by IVIS SpectrumCT In Vivo Imaging System (PerkinElmer, Inc.).
Synthesis of RB. Compound 1 (1.0 g; 1.0 mmol) and 2-ethylhexyl bromide (0.5 g; 2.6 mmol) were dissolved and magnetically stirred in N,N-dimethylformamide (DMF) at 80 ℃. After 6 h, excess 2-ethylhexyl bromide and DMF were removed by rotary evaporation. The residue was then dissolved in diethyl ether. To remove compound 6 and inorganic salts, the residue in diethyl ether was washed with water for three times followed by desiccation using anhydrous Na 2 SO 4 . The obtained residue was further purified by column chromatography using ethyl acetate as the eluent to afford the final product RB in deep purple color. 1  Synthesis of HBA. HBA was synthesized following the method in literature 37 . Briefly, synthesis of HBA was started from the commercially available 3hydroxybenzaldehyde by protecting with trimethyl orthoformate to afford 3-(dimethoxymethyl)phenol, which was followed by additional protection of the phenol group with tert-butyldimethylsilyl chloride (TBS-Cl) to give tert-butyl (3-(dimethoxymethyl)phenoxy)dimethylsilane. This compound was reacted with trimethylphosphite to produce a phosphonate derivative, and then condensed with 2adamantanone via the Wittig-Horner reaction to provide an enol ether precursor. At last, deprotection of the TBS group of the resulted precursor gave HBA. 1 Chemiluminescence quantum yields (Φ Cl ) of SO and HBA were measured and calculated relative to that of DO (Φ Cl = 0.021) referring to the reported method 38 .
Fluorescence efficiency of ALNPs (η Fl ) was calculated as the ratio of integrated  (4)). Afterglow substrates were excluded because of their negligible influence on fluorescent emission.
HOMO energy levels of afterglow relay units were collected from either references or computational calculation.
Calculation of energy levels. HOMO and LUMO energy levels of Reso and highenergy intermediates of DO, SO, and HBA were calculated by Gaussian 09 software based on density functional theory (DFT) with B3LYP/6-31 G(d) method.
Cell culture and cytotoxicity assay. 4T1 murine mammary carcinoma cells were purchased from ATCC (American Type Culture Collection). These cells were cultured in DMEM (Dulbecco's Modified Eagle Medium) supplemented with 10% FBS (fetal bovine serum) and 1% antibiotics (10 U/mL penicillin and 10 mg/mL streptomycin). Flasks seeded with 4T1 cells were placed in an incubator with 5% CO 2 and 95% O 2 humidified air atmosphere at 37 ℃.
To examine the cytotoxicity of ALNPs, 4T1 cells were seeded in 96-well plates (6000 cells in 200 µL supplemented DMEM per well). After culture for 24 h, PFVA-N, PFVA-N-DO, PFVA-N-SO, and PFVA-N-HBA (final concentration [PFVA] = 5, 10, 30, 50 µg mL −1 ) were added to cell culture medium, respectively. After incubation of nanoparticles with cells for 24 h, the culture medium was removed, and cells were gently washed with fresh sterile 1 × PBS buffer for three times. Fresh supplemented DMEM (100 µL per well) mixed with MTS (0.1 mg mL −1 , 20 µL per well) was then added to cells. After 3 h incubation, absorbance of culture medium at 490 nm was recorded by SpectraMax M5 microplate/cuvette reader. Because absorbance at 490 nm is proportional to the quantity of living cells, cell viability was calculated as the ratio of absorbance of sample treated cells to that of control cells. Imaging System (exposure time: 1 s). As for in vivo tissue penetration study, nanoparticle solutions were placed under a living mouse with a tissue depth of 1.5 cm, and other procedures was the same as in vitro tissue penetration study.
Signal-to-background ratio (SBR) is calculated as the ratio of luminescence (fluorescence or afterglow) in region of interest with ALNPs to that of tissue background without ALNPs.
Tumor mouse model. Animal experiments were carried out under the guidelines of Institutional Animal Care and Use Committee (IACUC), Sing Health. To establish tumor model, 2 million 4T1 cells suspended in supplemented DMEM were subcutaneously injected to the left shoulder of female NCr nude mouse (~6weeks old). Tumors were allowed to grow until 7~10 mm 3 before in vivo NIR fluorescence and afterglow luminescence imaging.
In vivo tumor imaging. NIR fluorescence and afterglow luminescence imaging of 4T1-tumor bearing mice were carried out using IVIS SpectrumCT In Vivo Imaging System. NIR fluorescence and afterglow images were at first captured before injection of ALNPs. NIR fluorescence was acquired with excitation at 710 nm and emission at 780 nm (exposure time: 0.1 s). Afterglow images were acquired Signal-to-background ratio (SBR) is calculated as the ratio of luminescence (fluorescence or afterglow) in tumor region after i.v. injection of ALNPs to that of tissue background before injection. , female NCr nude mice were euthanized by CO 2 asphyxiation and major organs (hearts, livers, spleens, lungs, and kidneys) were collected. These organs were then fixed in 4% paraformaldehyde followed by embedment in paraffin and 10-µm sectioning. H & E staining was then performed to tissue sections referring to standard procedure. Optical images were captured by Nikon ECLIPSE 80i microscope (Nikon Instruments Inc., NY, USA).