H2S-activatable near-infrared afterglow luminescent probes for sensitive molecular imaging in vivo

Afterglow luminescent probes with high signal-to-background ratio show promise for in vivo imaging; however, such probes that can be selectively delivered into target sites and switch on afterglow luminescence remain limited. We optimize an organic electrochromic material and integrate it into near-infrared (NIR) photosensitizer (silicon 2,3-naphthalocyanine bis(trihexylsilyloxide) and (poly[2-methoxy-5-(2-ethylhexyloxy)-1,4-phenylenevinylene]) containing nanoparticles, developing an H2S-activatable NIR afterglow probe (F12+-ANP). F12+-ANP displays a fast reaction rate (1563 ± 141 M−1 s−1) and large afterglow turn-on ratio (~122-fold) toward H2S, enabling high-sensitivity and -specificity measurement of H2S concentration in bloods from healthy persons, hepatic or colorectal cancer patients. We further construct a hepatic-tumor-targeting and H2S-activatable afterglow probe (F12+-ANP-Gal) for noninvasive, real-time imaging of tiny subcutaneous HepG2 tumors (<3 mm in diameter) and orthotopic liver tumors in mice. Strikingly, F12+-ANP-Gal accurately delineates tumor margins in excised hepatic cancer specimens, which may facilitate intraoperative guidance of hepatic cancer surgery.

F luorescence imaging probes are indispensable in biomedical research and clinical diagnosis owing to their low cost, simple operation, and noninvasiveness; these characteristics enable real-time detection of physiological and pathological processes at the molecular and cellular levels [1][2][3][4] . However, because fluorescence imaging requires real-time light excitation, the inevitable light absorption, scattering, and autofluorescence of biological tissues can limit penetration depth and elicit a low signal-to-background ratio (SBR), compromising the sensitivity for in vivo imaging [5][6][7] . Afterglow luminescent probes (also called persistent luminescent probes), which can trap excitation energy in defects and slowly release photons after cessation of light excitation, have recently emerged as promising tools for overcoming the limitations of fluorescence probes in biosensing and molecular imaging [8][9][10][11] . Afterglow imaging with these luminescent probes does not require real-time excitation with an external light, causing no autofluorescence interference from biological tissues and offering greatly improved SBR and sensitivity to visualize disease sites in vivo [12][13][14][15] .
During the past few years, several prominent afterglow probes have been developed using rare-earth heavy metal ion (e.g., europium or praseodymium)-doping inorganic nanoparticles 9,[16][17][18][19][20] or organic and polymeric afterglow materials 4,[21][22][23][24] . Richard and coworkers 25 optimized chromium-doped zinc gallate (ZnGa 1.995 -Cr 0.005 O 4 ) to develop inorganic nanoparticles that can emit intense long-lasting near-infrared (NIR) afterglow luminescence for noninvasive tracking of labeled cells and imaging of vascularization and grafted tumor cells. Pu and co-workers 26 proposed poly[2methoxy-5-(2-ethylhexyloxy)-1,4-phenylenevinylene] (MEH-PPV)based semiconducting polymer nanoparticles (SPNs) to develop biocompatible organic afterglow probes for sensitive detection of tumor metastasis and monitoring of hepatotoxicity in vivo 26 . Despite encouraging results, these afterglow probes encounter limitations when applied for in vivo diagnosis. Inorganic nanoparticlebased afterglow probes may be hampered by systemic toxicity concerns related to potential leakage of heavy metal ions 27,28 . Current MEH-PPV-based organic afterglow probes could be compromised by poor targeting ability to facilitate efficient delivery and accumulation in disease sites (e.g., tumors) after systemic injection [29][30][31] . In addition, most afterglow probes are designed as "always-on" luminescent probes; a continuous signal may produce a high background signal regardless of proximity or interaction with the molecular target of interest, thereby interfering with sensitivity and real-time imaging capacity 13,25,27,29 . By contrast, activatable afterglow luminescent probes with "off-on" signals in response to a biological target of interest can substantially reduce the background signal, improving SBR for real-time imaging. However, given a lack of effective activation approaches, activatable afterglow probes have been rarely developed for in vivo imaging 26,32 .
Hydrogen sulfide (H 2 S) is a highly reactive endogenous signaling molecule, playing a critical role in diverse physiological functions [33][34][35] . Aberrant concentration and tissue distribution of H 2 S are associated with many diseases (e.g., liver inflammation, hypertension, diabetes, and cancers) [36][37][38][39] . Therefore, precise spatiotemporal detection of H 2 S in living subjects is essential to the study of the biological functions and accurate diagnosis of H 2 Srelated diseases. Although numerous H 2 S-activatable fluorescent probes have been developed, the insufficient reaction kinetics, poor in vivo targeting ability, and limited tissue penetration of fluorescence have restricted the capacity to detect H 2 S' concentrations and locations in living animals [40][41][42][43][44][45][46] . Recently, we reported the engineering of an organic π-electron structure-based electrochromic material (EM 1 2+ ) as an H 2 S-responsive chromophore, which was amenable to build H 2 S-activatable NIR fluorescent probes for noninvasive imaging of H 2 S in living mice 47 .
In this paper, we optimize the structure of EM 1 2+ into a bis (pentafluorophenyl)-substituted EM F1 2+ that exhibits longer absorption wavelengths and significantly improved reaction kinetics toward H 2 S under physiological conditions. We showcase its feasibility to control afterglow luminescence from MEH-PPV (580 nm) and doping NIR photosensitizer (silicon 2,3-naphthalocyanine bis(trihexylsilyloxide), NIR775; 780 nm) when integrated into an MEH-PPV-and NIR775-containing afterglow nanoparticle (F1 2+ -ANP). Significant activation of afterglow luminescence by a factor of~122-fold can be achieved by incubating F1 2+ -ANP with H 2 S within 1 min, followed by irradiation with an 808-nm laser, allowing for rapid quantification of the H 2 S concentration in blood samples of healthy persons and patients diagnosed with hepatic cancers (HCCs) or colorectal cancers (CRCs). Moreover, through incorporation of β-galactose (β-Gal) ligands on the surface, we build a hepatic tumor-targeting and H 2 S-activatable afterglow probe (F1 2+ -ANP-Gal) that offers sensitive detection of subcutaneous and orthotopic liver tumors in living mice and delineation of liver tumor lesions in clinically excised HCC specimens. The demonstration of H 2 S-activatable afterglow luminescent probes feasible for sensitive and reliable detection of endogenous H 2 S underscores the potential of smart activatable afterglow luminescent probes for in vivo imaging.

Results
Design of H 2 S-activatable NIR afterglow probe. Figure 1 shows the design of an H 2 S-activatable NIR afterglow probe (F1 2+ -ANP). NIR775 was chosen as an NIR-excitable photosensitizer due to its stronger ability to generate singlet oxygen ( 1 O 2 ) compared to MEH-PPV, amplifying the afterglow of MEH-PPV upon pre-irradiation by 808-nm light 26,47 . Furthermore, efficient intraparticle energy transfer (ET) from MEH-PPV to NIR775 can promote a red shift of afterglow wavelength into the NIR window, permitting improved penetration depth in biological tissues. We employed EM F1 2+ as a H 2 S-responsive chromophore because it is designed to have longer ultraviolet-visible-NIR (UV-visible-NIR) absorptions (550 and 758 nm) than EM 1 2+ (546 and 685 nm), ensuring extensive spectral overlaps with MEH-PPV and NIR775 emissions (Fig. 1a). Additionally, EM F1 2+ possesses a more positive reduction potential (E red = + 0.40 V vs. SCE) than EM 1 2+ (+0.25 V vs. SCE), allowing for faster reaction to H 2 S via two-electron reduction. Within the compact nanoprobe F1 2+ -ANP, the fluorescence and 1 O 2 generation ability of MEH-PPV and NIR775 are each quenched by EM F1 2+ due to efficient fluorescence resonance ET (FRET) from MEH-PPV and NIR775 to EM F1 2+ ; the afterglow of F1 2+ -ANP is substantially quenched at the "off" state ( Fig. 1b). In the presence of H 2 S, the EM F1 2+ can be rapidly reduced into diene EM F2 with distinct blue-shift absorption (λ abs < 500 nm); the FRET process is hence eliminated, leading to recovery of 1 O 2 production in F2-ANP under 808-nm light irradiation. Therefore, NIR afterglow is activated at "on" state through 1 O 2 -dependent oxidation of MEH-PPV, followed by an efficient intraparticle ET process. Ultralow background noise and deep-tissue penetration are achieved given advantages of long-lasting afterglow luminescence and rapid reaction kinetics. The probe thus presents a high SBR for rapid and sensitive detection of H 2 S levels in human blood samples, clinical liver tumor resections, and living mice.
We next investigated the afterglow luminescence of F1 2+ -ANP in the absence or presence of NaHS in PBS buffer (pH 7.4) at 37°C. After optimizing irradiation conditions ( Supplementary Fig. 11), the solutions were pre-irradiated by the 808-nm laser at a power density of 1 W cm −2 for 1 min. F1 2+ -ANP exhibited little afterglow in the absence of NaHS, whereas NaHS-treated F1 2+ -ANP displayed obvious afterglow signals in the visible and NIR regions. Signal quantification from the images (Fig. 2e, inset) showed that afterglow intensities increased by~122-fold. Notably, the afterglow signals in the NIR region were much higher than those in the visible region, presumably attributable to the ET process from MEH-PPV to NIR775 within activated F1 2+ -ANP as demonstrated by the spectral overlaps between them ( Supplementary Fig. 12a, b). The ET process between MEH-PPV and NIR775 was further confirmed by the shorter fluorescence lifetime of MEH-PPV in activated F1 2+ -ANP (~49 ps) than that in the absence of NIR775 (~106 ps), with the ET efficiency of~53.8% (Supplementary Fig. 12c and Supplementary Note. 2). These results suggest that H 2 S can effectively turn on the NIR afterglow luminescence of F1 2+ -ANP. Through continuous acquisition of afterglow images, the afterglow could persist for over 40 min after laser cessation, with a half-life of~6.6 min ( Fig. 2f and Supplementary Fig. 13). Moreover, the afterglow signal could be recharged by multiple irradiation with the 808-nm laser; the afterglow signal of activated F1 2+ -ANP was still~50-fold higher than that of unactivated F1 2+ -ANP after irradiation for 15 cycles over 3 days, while their size and morphology were negligibly changed after irradiation, reflecting the potential for long-term imaging (Supplementary Figs. 14 and 15).
We subsequently examined the sensitivity and specificity of F1 2+ -ANP toward H 2 S (Fig. 2g, h). Afterglow images of F1 2+ -ANP became brighter with NaHS concentration, showing a linear correlation between afterglow intensities and NaHS concentrations from 0 to 50 μM (limit of detection:~0.1 μM (3σ/k). Only H 2 S could turn on the afterglow of F1 2+ -ANP ( Fig. 2i), consistent with those measured by fluorescence ( Supplementary Fig. 16), demonstrating that F1 2+ -ANP was a specific probe for H 2 S. In addition, F1 2+ -ANP was highly stable under different physiological conditions (Supplementary Figs. 17 and 18); the fluorescence and afterglow of F1 2+ -ANP were similarly activated by NaHS in different cell culture mediums ( Supplementary Fig. 19) or in PBS buffers with pH ranging from 5.0 to 8.0 ( Supplementary Fig. 20). Taking advantage of highsensitivity, specificity, and ultrafast reaction kinetics toward H 2 S, F1 2+ -ANP was applied to quantify H 2 S concentration in mouse blood ( Supplementary Fig. 21). The H 2 S concentration in mouse whole blood was measured to be 24.8 ± 2.2 μM (the concentration is expressed as mean ± standard deviation of three samples) using NaHS as an internal standard, which was within reported values 35,48,49 .
Detection of H 2 S in HCC cells. Since increased H 2 S level is found in malignant liver tumors 50,51 , we investigated the potential to detect H 2 S in human HCC HepG2 cells through afterglow imaging. The overexpression of cystathionine β-synthase (CBS) and cystathionine γ-lyase (CSE) for the biosynthesis of endogenous H 2 S in  Fig. 27). We then collected cell pellets (~1 × 10 5 cells) and irradiated with the 808-nm laser (1 W cm −2 , 1 min) to produce afterglow. As presented in Fig. 3c, HepG2 cells incubated with F1 2+ -ANP-Gal showed a much brighter afterglow image compared to that of F1 2+ -ANP, β-Gal-pre-treated HepG2 cells, or β-Gal receptor-deficient human lung cancer A549 cells incubated with F1 2+ -ANP-Gal. The afterglow intensity in HepG2 cells incubated with F1 2+ -ANP-Gal was more than 7-fold higher than that with F1 2+ -ANP (Fig. 3d). We subsequently applied F1 2+ -ANP-Gal to monitor H 2 S fluctuation in HepG2 cells (Fig. 3f, g). Afterglow signals in F1 2+ -ANP-Galtreated HepG2 cell pellets (~3 × 10 4 cells) increased noticeably when pre-treated with extraneous NaHS; pre-treatment with ZnCl 2 to scavenge H 2 S reduced afterglow substantially. HepG2 cells stimulated with L-Cys (a precursor for biosynthesis of H 2 S) to upregulate intracellular H 2 S levels could augment afterglow, which was greatly inhibited when adding DL-propargylglycine (PAG; 50 μg mL −1 ) and aminooxyacetic acid (AOAA; 20 μM) to inhibit endogenous CSE and CBS activities. These results were confirmed by fluorescence imaging of H 2 S in individual HepG2 cells (Fig. 3e). We also found that afterglow signals in F1 2+ -ANP-Gal-treated HepG2 cell pellets increased with the cell number, which could allow to detect 1000 and 5000 HepG2 cells in vitro and in vivo, respectively, much lower than fluorescence (Supplementary Figs. 28 and 29). Moreover, the afterglow luminescence of F1 2+ -ANP-Gal could still be clearly observed in HepG2 cells even after irradiation for 12 cycles over 24 h, suggesting the potential for longterm afterglow imaging in live HepG2 cells ( Supplementary  Fig. 30). Thus, F1 2+ -ANP-Gal was an efficient probe capable of sensitively detecting HepG2 cells through afterglow imaging of endogenous H 2 S.
Afterglow imaging of H 2 S in subcutaneous liver tumors. Prior to afterglow imaging of H 2 S in vivo, the penetration depth of afterglow luminescence emitted from the H 2 S-activated F1 2+ -ANP-Gal was examined. The afterglow and fluorescence signals each declined with the thickness of chicken tissues (Supplementary Fig. 31). However, because tissue autofluorescence was minimized without real-time excitation, the afterglow exerted significantly higher SBR than NIR fluorescence. At a thickness of 4 cm, the SBR of afterglow remained as high as 18.7 ± 4.6 (SBR is expressed as mean ± standard deviation, n = 3), whereas the SBR of NIR fluorescence was close to 1. These results indicate that the penetration of afterglow luminescence was much deeper than that of NIR fluorescence, which was also confirmed by the larger SBR of afterglow (~201) versus fluorescence (~2.5) when detecting signals through a mouse body (~1.9 cm). We next determined the blood half-life (t 1/2 ) of F1 2+ -ANP-Gal following intravenous (i.v.) injection into mice. The t 1/2 was measured to be~7 .5 h, permitting efficient circulation to facilitate delivery into tumor tissues. The biodistribution study showed that F1 2+ -ANP-Gal accumulated mainly in the liver of healthy mice at 12 h post injection (%ID/g ≈ 23.6%, Supplementary Fig. 32).
We next applied F1 2+ -ANP-Gal for afterglow imaging of endogenous H 2 S in subcutaneous (s.c.) HepG2 tumors (Fig. 4a). After i.v. injection into mice, afterglow and fluorescence signals both increased gradually and peaked after 12 h ( Fig. 4b and Supplementary Fig. 33). However, owing to the low background signal of afterglow, the SBR of afterglow was much higher than that of fluorescence (Fig. 4c). At 12 h, the average SBR of afterglow was 71.9 ± 8.8,~30-fold higher than that of fluorescence (2.4 ± 0.4). When L-Cys was injected into tumors to upregulate H 2 S levels, afterglow and fluorescence signals were both significantly enhanced, whereas intratumoral (i.t.) injection of ZnCl 2 to scavenge tumor H 2 S reduced tumor afterglow and fluorescence. The SBR of afterglow 12 h post injection of F1 2+ -ANP-Gal and L-Cys increased to 133.8 ± 24.9,~1.9-and 2.9-fold higher than that of F1 2+ -ANP-Gal alone and F1 2+ -ANP-Gal plus ZnCl 2 (46.7 ± 1.6), respectively. In addition, mice injected with untargeted F1 2+ -ANP showed much lower afterglow signals and SBR (44.2 ± 3.1) in HepG2 tumors compared to F1 2+ -ANP-Gal. These results aligned with ex vivo fluorescence imaging of resected tumor tissue slices ( Supplementary Fig. 34), affirming that F1 2+ -ANP-Gal could deliver into HepG2 tumors and efficiently detect tumor H 2 S levels.
We applied F1 2+ -ANP-Gal to noninvasively monitor HepG2 tumor growth in living mice. After s.c. injection of HepG2 cells (2 × 10 6 ) at~4,~8, and~15 days, average tumor sizes grew  (Fig. 4d, e). For a tumor measuring only~12 mm 3 , the SBR of afterglow was 47.8 ± 5.9; conversely, the SBR of fluorescence was only 1.1 ± 0.4 (Fig. 4f). These results indicate that the afterglow emitted from F1 2+ -ANP-Gal after activation by endogenous H 2 S was more appropriate than fluorescence when applied to detect tiny s.c. HepG2 tumors in vivo. Notably, we identified a strong correlation between the SBR of afterglow or fluorescence produced from F1 2+ -ANP-Gal and apparent tumor size growth (Pearson's r = 0.99). Therefore, F1 2+ -ANP-Gal can offer valuable information about tumor growth in living mice.
Imaging of orthotopic liver tumors in mice. By virtue of the deep-tissue penetration of afterglow imaging, F1 2+ -ANP-Gal was applied to detect orthotopic HepG2 liver tumors through afterglow imaging of H 2 S. Orthotopic liver tumors were established by inoculation of luciferase-transfected HepG2 cells (HepG2/Luc) into the lube of livers of mice (Fig. 5a). Strong bioluminescence (BL) in liver indicated the successful establishment of orthotopic liver tumors after 15 days (Fig. 5b). We then injected (i.v.) F1 2+ -ANP-Gal into mice followed by irradiation of each mice body (face up) with the 808-nm laser (1 W cm −2 , 1 min) after 12 h (Fig. 5a). NIR fluorescence and afterglow images both showed that mice with orthotopic liver tumors had brighter signals relative to control mice (Fig. 5b, c), with locations closely matching those of BL images. The SBR of afterglow in orthotopic liver tumors was 124.5 ± 12.5, 2.5-fold higher than that in livers of control mice (50.0 ± 10.9) (Fig. 5d). The SBR of afterglow was also much higher compared to that of fluorescence in liver tumors (4.3 ± 1.2). We subsequently conducted ex vivo imaging of main organs (e.g., liver, lung, heart, kidneys, intestines, stomach, and spleen) resected from control mice and orthotopic liver tumor mice. Figure 5e shows that tumor tissues in the resected liver exhibited the brightest afterglow image among all resected organs. The afterglow intensity in tumor tissues was~5.3-fold higher than that in surrounding liver tissues and~3.6-fold higher than normal liver tissues of control mice (Fig. 5f). NIR fluorescence imaging of the same tissues verified these findings (Supplementary Fig. 35), which were further confirmed via fluorescence imaging of tumor and normal tissue slices ( Supplementary  Fig. 36). Subsequent WB analysis revealed that enhanced afterglow and fluorescence in orthotopic HepG2/Luc tumors concurred with the higher expression of CBS and CSE enzymes compared to normal liver tissues ( Fig. 5g and Supplementary  Fig. 37). These findings demonstrate that F1 2+ -ANP-Gal could offer a large SBR to accurately and noninvasively position orthotopic liver tumor foci via afterglow imaging of H 2 S.  , respectively. f Comparison of the average afterglow intensities of tumors and main organs resected from control (blue) and orthotopic liver tumor (red) mice. g WB analysis shows the relative CBS and CSE protein levels in the liver tissues resected from control mice and HepG2 tumor tissues resected from orthotopic liver tumor mice. All afterglow luminescence images were acquired for 60 s with an open filter, after pre-irradiation of mouse body or organs with the 808-nm laser (1 W cm −2 , 1 min). All fluorescence images were acquired with λ ex/em = 740/790 nm. Data denote mean ± s.d. (*P < 0.05, **P < 0.01, n = 3). Statistical differences were analyzed by Student's t test. Source data are provided as a Source Data file.
Detection of H 2 S in clinical blood samples. We employed the activatable afterglow to detect H 2 S levels in human blood samples collected from 10 healthy persons, 10 HCC patients, and 10 CRC patients in clinics. Following incubation of 2-fold diluted blood with F1 2+ -ANP for 1 min, afterglow signals in HCC and CRC patients' bloods were much higher than that in healthy persons (Fig. 6a, b). We took a standard curve established by the addition of NaHS into healthy persons' blood as the internal standard, and the average H 2 S level in the whole blood of healthy persons was 27.6 ± 2.7 μM (the H 2 S level is expressed as mean ± standard deviation, n = 10); the level increased to 47.0 ± 5.0 and 75.9 ± 11.9 μM in the whole blood of HCC and CRC patients, respectively ( Fig. 6c and Supplementary Fig. 38). These results revealed that blood H 2 S levels in HCC and CRC patients were upregulated compared to healthy persons, suggesting that H 2 S could serve as a potential blood marker for certain malignant cancers (e.g., HCC and CRC). Thus, F1 2+ -ANP presents high potential to quantify blood H 2 S levels for cancer diagnosis.
Detection of liver tumor tissues in clinical specimens. Having demonstrated the higher blood H 2 S levels in HCC patients' bloods, we employed afterglow to delineate tumor margins in clinically excised HCC specimens. Excised liver tissues from four HCC patients were incubated with F1 2+ -ANP-Gal in PBS buffer (pH 7.4) for 3 h to allow efficient internalization and activation by H 2 S. After being washed with PBS buffer, whole specimens were irradiated with the 808-nm laser (1 W cm −2 , 1 min); the resulting afterglow and fluorescence images were acquired immediately (Fig. 7a). A strong afterglow image appeared in the left section of the specimen, corroborating the NIR fluorescence image of the tissue slice (Fig. 7b, c). Subsequent hematoxylin and eosin (H&E) staining revealed that the left section of the specimen was characterized by tumor tissue (Fig. 7d). Lesions with clear afterglow and fluorescence signals resulting from F1 2+ -ANP-Gal staining accorded with the tumor location was indicated by H&E staining, as was observed in the other three specimens from different HCC patients ( Supplementary Figs. 39-41). These results suggest that F1 2+ -ANP-Gal was effective in delineating liver tumor lesions in clinical specimens. Upon dividing the signal intensity in tumor lesions by that in adjacent normal liver tissue, the average SBR in afterglow imaging of the four clinical specimens was 58.6 ± 13.2, significantly higher than that in fluorescence (2.2 ± 0.7) (Fig. 7e). Such a high afterglow SBR could promote detection of tumor lesions in the specimen covered with 1.5cm-thick chicken tissue ( Supplementary Fig. 42). These findings imply that F1 2+ -ANP-Gal may have promise for detecting deeply located tumor cells, which could be applied to identify tumor margins and accurately guide liver tumor surgery in the future.
In vivo clearance and biosafety evaluations. To evaluate the in vivo clearance and biosafety, we injected (i.v.) F1 2+ -ANP-Gal at the same dose as the imaging dose (211/100/8 μg F1 2+ (BF 4 − ) 2 / MEH-PPV/NIR775, 200 μL). The NIR fluorescence images showed that the NIR fluorescence was gradually observed in the liver, with intensity maximized at 12 h (Supplementary Fig. 43). After that, the NIR fluorescence in mice decreased overtime, and nearly returned to the background after 28 days, indicating that F1 2+ -ANP-Gal could be cleared out after systemic administration. We then monitored the body weights, showing little difference between F1 2+ -ANP-Gal-treated and saline-treated mice for 28 days (Supplementary Fig. 44a). Moreover, the subsequent biochemistry and hematology test of mice 1, 7, 14, and 28 days post injection of F1 2+ -ANP-Gal showed that all the markers related to liver or kidney functions were within the reference  Fig. 44b, c); F1 2+ -ANP-Gal did not cause obvious toxic effect on mice, which was further confirmed by the histologic analysis of resected main organs (e.g., liver, lung, kidney, heart, and spleen) (Supplementary Fig. 45).

Discussion
Afterglow luminescence holds great promise for sensitive and noninvasive imaging of biomolecules in living subjects 26,31 ; however, real-time afterglow imaging remains rare due to lack of afterglow probes that can be rapidly activated by a molecular target to produce intense afterglow luminescence. Herein, we optimized the structure of an organic π-electron system-based EM 1 2+ into EM F1 2+ and engineered it as an effective H 2 Sresponsive chromophore capable of integrating with NIR775 and MEH-PPV to develop H 2 S-activatable NIR afterglow probe (F1 2+ -ANP) for in vivo imaging. F1 2+ with two electron-withdrawing pentafluorophenyl groups showed strong absorptions at 550 and 758 nm, overlapping more extensively with MEH-PPV (580 nm) and NIR775 emissions (780 nm) than 1 2+ . F1 2+ also possessed a more positive value of E red (+0.40 V vs. SCE) than 1 2+ (+0.25 V vs. SCE), permitting much faster reaction kinetics toward H 2 S (k 2 = 3600 ± 130 M −1 s −1 ). These features offer F1 2+ noteworthy advantages over 1 2+ in designing activatable afterglow probes. The longer absorption of F1 2+ could (a) allow it to prevent NIR775 from generating 1 O 2 , which is key to eliciting afterglow luminescence (Supplementary Fig. 9) and (b) directly quench visible and NIR afterglow luminescence of F1 2+ -ANP. Accordingly, afterglow signals could be substantially quenched in an "off" state. Upon reaction with H 2 S, rapid reaction kinetics between F1 2+ and H 2 S could allow H 2 S to efficiently reduce F1 2+ into F2 within 1 min, thereby recovering afterglow luminescence by a factor of~122-fold (Fig. 2e). Such activatable afterglow probes with large turn-on ratio and ultrafast reaction kinetics could facilitate sensitive and real-time detection of H 2 S in vitro and in vivo. H 2 S is recognized as the third signaling gasotransmitter, participating in many essential physiological and pathological processes [36][37][38][39] . Throughout the past few years, many colorimetric and fluorescent probes have been developed for H 2 S measurement [52][53][54][55][56][57] . However, due to the low concentration and transient nature of H 2 S in biology 58 , the detection of endogenous H 2 S with these probes has been compromised by insufficient sensitivity, selectivity, or reaction kinetics. F1 2+ -ANP possesses high sensitivity, selectivity, and an ultrafast response to H 2 S, rendering it suitable for quantification of H 2 S concentration in mouse blood samples ( Supplementary Fig. 21). Parallel measurements of H 2 S levels in human blood freshly collected from healthy persons, HCC patients, and CRC patients were taken following incubation with F1 2+ -ANP for 1 min. Findings showed that blood H 2 S levels in HCC and CRC patients were substantially higher than in healthy persons (Fig. 6), suggesting that H 2 S could be a potential blood marker for cancer diagnosis.
In addition to detecting blood H 2 S in vitro, F1 2+ -ANP was engineered into a hepatic tumor-targeting and H 2 S-activatable afterglow luminescent probe, F1 2+ -ANP-Gal, for afterglow imaging of hepatic tumor H 2 S in vivo. With a long blood circulation half-life and β-Gal receptor-mediated active delivery, F1 2+ -ANP-Gal could efficiently enter s.c. HepG2 tumors after i.v. injection into mice, turning on afterglow luminescence for real-time and noninvasive monitoring of H 2 S fluctuation in tumors. Benefiting from low tissue autofluorescence in the absence of real-time light excitation, F1 2+ -ANP-Gal with activated NIR afterglow presented a much higher SBR (47.8 ± 5.9) than NIR fluorescence (1.1 ± 0.4), permitting greater sensitivity to detect tiny s.c. HepG2 tumors (~12 mm 3 , Fig. 4d, f). The SBR resulting from afterglow images also increased linearly with tumor size, suggesting the potential of F1 2+ -ANP-Gal for (a) accurate detection of earlystage liver tumors and (b) noninvasive monitoring of HepG2 tumor growth in living mice through afterglow imaging of H 2 S. Moreover, the enhanced penetration depth of NIR afterglow luminescence could allow F1 2+ -ANP-Gal to sensitively detect orthotopic liver tumors, affording a large SBR (124.5 ± 12.5) to precisely position liver tumor foci in mice (Fig. 5). Of note, the H 2 S-induced rapid and substantial afterglow enhancement in F1 2+ -ANP-Gal could also offer a large SBR (58.6 ± 13.2) to identify liver tumor lesions in excised HCC specimens, with tumor margins aligning with H&E staining (Fig. 7). These results may offer valuable information to guide surgical resection of liver tumor tissues in intraoperative HCC patients.
In conclusion, we report the design and development of H 2 S-activatable NIR afterglow luminescent probes by doping EM F1 2+ and NIR775 into MEH-PPV-based organic afterglow nanoparticles. Our findings demonstrate the probe's utility in noninvasive and real-time imaging of H 2 S in living mice. F1 2+ -ANP can be selectively and rapidly activated by H 2 S, leading to a large afterglow luminescence turn-on ratio (~122fold) after pre-irradiation with an 808-nm laser, which is useful for quantifying the H 2 S concentration in the blood of healthy persons (27.6 ± 2.7 μM), HCC patients (47.0 ± 5.0 μM), and CRC patients (75.9 ± 11.9 μM). A hepatic tumor-targeting and H 2 Sactivatable afterglow probe (F1 2+ -ANP-Gal) is also built, offering a high SBR and deep-tissue penetration for highsensitivity imaging of s.c. and orthotopic liver tumors via afterglow imaging of H 2 S in mice. Moreover, F1 2+ -ANP-Gal can accurately delineate liver tumor margins in excised HCC specimens, which may be especially useful for image-guided HCC surgery. This study presents the potential of using H 2 S-activatable afterglow luminescent probes for in vivo imaging, which could detect H 2 S levels in other diseases (e.g., liver inflammation and hypertension) and promote early disease diagnosis in the future.

Methods
Chemicals and characterization. 1 H and 13 C nuclear magnetic resonance (NMR) spectra were recorded on a BRUKER Ascend TM 400 ( 1 H/400 MHz and 13 C/100 MHz) spectrometer. Mass spectra were recorded on a JEOL JMS-T100GCV spectrometer in FD mode (GC-MS&NMR Laboratory, Research Faculty of Agriculture, Hokkaido University). Redox potentials (E ox and E red ) were measured by CV in dry CH 2 Cl 2 containing 0.1 M Bu 4 NBF 4 as a supporting electrolyte at a scan rate of 100 mV s -1 . A Pt disk and wire electrodes were used as the working and counter electrodes, respectively. The working electrode was polished using a water suspension Al 2 O 3 (0.05 µmol) before use. Column chromatography was performed on a silica gel I-6-40 (YMC, 40-63 µm), a silica gel 60 N (Kanto Chemical Co., Inc., spherical neutral, 40-50 µm), or aluminum oxide 90 standardized (Merck, 63-200 µm). The DFT calculations were performed with the Gaussian 09 W program package. The geometries of the compounds were optimized by using the B3LYP method in combination with the 6-31G* basis set, unless otherwise indicated. Melting points were measured on a Yamato MP-21 melting point apparatus and are recorded uncorrected. Elemental analyses were performed on an Exeter Analytical CE440 (the Open Facility, Global Facility Center, Creative Research Institution, Hokkaido University). HPLC analysis was carried out on Thermo Fisher U3000 HPLC systems. Absorption spectra were performed an Ocean Optics UV-visible spectrometer. Fluorescence spectra were obtained on a HORIBA Jobin Yvon Fluoromax-4 fluorescence spectrometer. An Olympus IX73 fluorescent inverted microscope was used for fluorescence imaging for cells and tissue slices. DLS and zeta potential analysis were measured by a 90 Plus/BI-MAS equipment (Brookhaven, USA). Transmission electron microscopy images were conducted on a JEM-1011 transmission electron microscope (JEOL Ltd., Japan) with an accelerating voltage of 100 kV. MTT assay was performed on a microplate reader (Tcan). The afterglow luminescence spectra, afterglow, and fluorescence images in vitro or vivo were acquired with the IVIS Lumina XR III imaging system. An 808-nm high-power NIR laser (FC-808-10W-MM, Xilong Company, China) was used to irradiate the specified irradiation sites to generate the afterglow luminescence, unless otherwise noted. Measurement of the fluorescence and afterglow luminescence. To measure the fluorescence of F1 2+ -ANP or F1 2+ -ANP-Gal in response to H 2 S, F1 2+ -ANP, or F1 2+ -ANP-Gal (2.2 μg NIR775, 28 μg MEH-PPV, 58 μg F1 2+ (BF 4 − ) 2 ) in 1.0 mL PBS buffer (pH 7.4) were incubated with NaHS (200 µM) at 37°C for 1 min, and the fluorescence spectra were recorded on a HORIBA Jobin Yvon Fluoromax-4 fluorometer with fluorescence synchronous scanning mode (λ ex = 400-800 nm, offset = 100 nm).
To measure the afterglow luminescence of F1 2+ -ANP or F1 2+ -ANP-Gal in response to H 2 S, F1 2+ -ANP, or F1 2+ -ANP-Gal (2.2 μg NIR775, 28 μg MEH-PPV, 58 μg F1 2+ (BF 4 − ) 2 ) in 1.0 mL PBS buffer (pH 7.4) were incubated with 200 μM NaHS at 37°C for 1 min. Then, the incubation solutions were illuminated by an 808-nm laser (1 W cm −2 ) for 1 min. The laser was removed, and the afterglow luminescence images were acquired on an IVIS Spectrum imaging system equipped with an open filter or a 790 nm emission filter, with an acquisition time of 60 s. The afterglow luminescence intensity in each image was quantified by applying region of interest (ROI) over the image, using the Living Image Software (4.5.2, PerkinElmer, MA, USA).
Cell culture. Human HCC HepG2 cells and human lung cancer A549 cancer cells were purchased from Stem Cell Bank, Chinese Academy of Sciences (Shanghai, China). All cell lines were routinely tested to exclude infection with mycoplasma. They were authenticated by the supplier using short tandem repeat test. Human HCC HepG2 cells were cultured in Dulbecco's modified Eagle's medium (DMEM). Human lung cancer A549 cancer cells were grown in F-12 medium. All the mediums were supplemented with 10% fetal bovine serum (FBS), 1% penicillin, and streptomycin. All cells were maintained at 37°C in a humidified atmosphere with 5% CO 2 .
For afterglow imaging of cells, HepG2 cells were plated at a density of 3 × 10 4 (or 1 × 10 5 ) cells into culture dish. Cells were treated according to the abovementioned procedures. Cells were harvested from the culture dish by digestion with trypsin. The cell pellets were then collected after centrifugation at 161 × g for 4 min. The cell pellets were irradiated with the 808-nm laser (1 W cm −2 ) for 1 min. After removal of the laser, the afterglow luminescence images were immediately acquired on an IVIS Spectrum imaging system equipped with an open filter, with an acquisition time of 60 s. The afterglow luminescence intensities in the cell pellets were quantified by applying ROI over the images, using the Living Image Software To establish tumors, HepG2 cells (2.0 × 10 6 cells, 50 μL) suspended in 50 v/v% mixture of Matrigel in supplemented DMEM (10% FBS, 1% pen/strep (100 U mL −1 penicillin and 100 µg mL −1 streptomycin)) were injected subcutaneously into the selected positions of nude mice. The tumor volume (V) was determined assuming ellipsoid shape with the formula of V = (L × W 2 )/2, where the length (L) and width (W) of each tumor was measured using a caliper. When the tumor volume reached about 12-100 mm 3 , fluorescence and afterglow imaging were conducted. To establish an orthotopic HCC model, a midline incision of the anterior abdominal wall was made. HepG2 cells transfected with luciferase (HepG2/Luc) (2 × 10 6 , 100 μL) in serum-free culture medium were carefully injected into the lube of liver of mice under anesthesia by pentobarbital sodium. Tumor growths were monitored by BL imaging. After 2 weeks, mice with HCC model were successfully established for afterglow and fluorescence imaging. The fluorescence images were acquired on an IVIS Lumina XR III imaging system, with λ ex /λ em = 740/790 nm. To acquire afterglow images, the mouse body at 0, 4, 8, 12, and 24 h post i.v. injection of F1 2+ -ANP-Gal was irradiated under the 808-nm laser (1 W cm −2 ) for 1 min. After cessation of laser, the afterglow luminescence images were immediately acquired on an IVIS Spectrum imaging system equipped with an open filter, with an acquisition time of 60 s. Every experiment was conducted in three mice. ROIs were drawn over the tumor and the thigh of the mouse to quantify the fluorescence and afterglow intensities using the Living Image Software. The SBRs were calculated by dividing the fluorescence and afterglow intensities in tumor to that in thigh.

WB analysis of CSE and CBS expression.
To conduct the WB analysis of mouse liver and orthotopic HepG2 tumor tissue homogenates, three normal mice and mice with orthotopic liver tumors were euthanized, and the livers from normal mice and the HepG2 tumor tissues from the orthotopic liver tumor-bearing mice were resected. Liver and HepG2 tumor tissues were then homogenized with lysis buffer. The homogenates were incubated on ice for 30 min, and the lysates were obtained by centrifugation (13,000 × g, 15 min) at 4°C. Aliquots of proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (10%) and subsequently transferred to nitrocellulose membranes by electroblotting. The membrane was blocked in 5% skim milk powder in 0.1% Tris-buffered saline/ Tween-20 (TBST, 20 mM Tris-HCl, pH 7.4, 137 mM NaCl, and 0.1% Tween) at r.t. for 2 h, and then incubated with antibody raised against CBS (Abcam, ab135626, lot: GR3265732-10, dilution 1 : 1000) for 2 h at r.t. After three washes with TBST, the membrane was incubated with a secondary antibody horseradish peroxidase (HRP)-conjugated goat anti-rabbit immunoglobulin G (IgG) for 2 h at r.t. The films were developed with the ECL system. The CBS protein and markers were visualized using G:BOX chemiXR5. Afterwards, the membrane was rinsed in D.I. water at r.t. for 5 min, and antibody stripping buffer (weak alkaline) was added. The membrane was then rinsed with D.I. water at r.t. for 5 min, blocked in 5% skim milk powder in 0.1% Tris-buffered saline/Tween-20 at r.t. for 2 h, and incubated with antibody raised against CSE (Abcam, ab151769, lot: GR3257101-3, dilution 1 : 1000) for 2 h at r.t. After three washes with TBST, the membrane was incubated with a secondary antibody HRP-conjugated goat anti-rabbit IgG at r.t. for 2 h. The films were developed with the ECL system. The CSE protein and markers were visualized using G:BOX chemiXR5. The resulting band intensities were quantified using a Gel-Pro32. Each experiment was repeated for three times.
Measurement of the H2S concentration in human blood. Whole-blood samples were collected from healthy volunteers and the patients with HCC and CRC at Affiliated Drum Tower Hospital of Nanjing University (Nanjing, China). For the detection of H 2 S in blood samples of healthy persons and patients diagnosing with HCC or CRC, freshly collected bloods were 2-fold diluted with PBS buffer (1×, pH 7.4), and then incubated with F1 2+ -ANP (2.2/28/58 μg mL −1 NIR775/MEH-PPV/F1 2+ (BF 4 − ) 2 ) at 37°C for 1 min, followed by irradiation with the 808-nm laser (1 W cm −2 ) for 1 min. After removal of laser, the afterglow luminescence images were immediately acquired for 60 s with a 790 nm filter. The afterglow luminescence intensities at 790 nm for F1 2+ -ANP were quantified by applying ROI over the image using the Living Image Software (4.5.2, PerkinElmer, MA, USA). With constructing a standard curve in healthy blood by an internal standard method, the concentrations of H 2 S in whole bloods of healthy persons and HCC or CRC patients were obtained.
Imaging of human HCC specimens. HCC specimens were collected from the patients with HCC at Affiliated Drum Tower Hospital of Nanjing University (Nanjing, China). This study was approved by institutional review board of Affiliated Drum Tower Hospital of Nanjing University, and all subjects provided written informed consent under institutional review board prior to sample collection. HCC specimens were resected from four HCC patients. HCC specimens were infiltrated with F1 2+ -ANP-Gal (58/28/2.2 μg mL −1 F1 2+ (BF 4 − ) 2 /MEH-PPV/ NIR775) in PBS buffer for 3 h. The specimens were then rinsed with PBS buffer (1×, pH 7.4) three times to remove nonspecifically absorbed F1 2+ -ANP-Gal on the tissues. The specimens were irradiated with the 808-nm laser (1 W cm −2 , 1 min). After removal of laser, the afterglow images were acquired on an IVIS Spectrum imaging system equipped with an open filter, with an acquisition time of 60 s. The fluorescence images were acquired with λ ex/em = 740/790 nm. ROIs were drawn over the tumor and adjacent normal liver tissues to quantify the fluorescence and afterglow intensities using the Living Image Software. The SBRs were calculated by dividing the fluorescence and afterglow intensities in tumor to that in normal liver tissues.
Immunohistochemistry studies. HCC tissues were fixed in 4% formalin and then embedded in paraffin before 10-µm sectioning. Histology samples were stained by H&E using a standard protocol. White light images were acquired using an IX73 optical microscope equipped with a color camera.
Statistical analysis. Results are expressed as the mean ± standard deviation unless otherwise stated. Statistical differences among experimental groups were analyzed by Student's t test. P < 0.05 was considered statistically significant. All statistical calculations were performed using GraphPad Prism 6 including assumptions of tests used (GraphPad Software Inc., CA, USA).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.