Starch, L-cysteine (Cys), glycine (Gly), urea, glucose (Glu), calcium chloride (CaCl₂), sodium chloride (NaCl), copper chloride (CuCl₂), potassium chloride (KCl), iron (III) chloride (FeCl3), glutathione (GSH), sodium sulfate (Na₂SO₄), sodium nitrate (NaNO₃), sodium sulfite (Na₂SO₃), sodium phosphate (Na₃PO₄), sodium bisulfite (NaHSO₄) and sodium nitrite (NaNO₂) were purchased from Sinopharm Chemical Reagent Company (China). Dulbecco’s modified Eagle’s medium (DMEM) medium and fetal bovine serum (FBS) were purchased from HyClone (United States). WST assay kits were purchased from Energy Chemical (China). All regents were not processed or purified prior to use.

Fluorescence spectra were recorded on a fluorescent spectrophotometer (F97Pro, China). UV-vis absorption spectra were recorded on a U-3010 spectrophotometer (Hitachi, Japan). Fluorescence lifetime measurements were carried out on photoluminescence Spectrometer (FLS 1000, United Kingdom). An AXIS ULTRA DLD spectrometer was used to detect X-ray photoelectron spectroscopy (XPS). Freeze dryers (Scientz-10N, CHINA) was used to obtain CDs solid powders. Transmission electron microscopy (TEM, JEOL Ltd, Japan) was used to characterize the morphology of the CDs. Nano ZS/ZEN3690 (Malvern, United Kingdom) was used to investigate the particle size distribution and surface potential of the CDs. Fourier transform infrared (FT-IR) spectra was acquired using an FT-IR spectrometer (Agilent Cary 660, United States).

The synthesis of CDs was similar to our reported method (Wang et al., 2021a). Typically, 0.5 g starch was dissolved in 30 ml ultrapure water, stirred and ultrasonic vibrated for 10 min, and then heated at 200°C in a 100 ml stainless steel autoclave lined with polytetrafluoroethylene for 10 h. After the solution was cooled to room temperature, centrifuged at 10,000 rpm for 10 min to remove precipitate, filtered by a 0.22 μm filter membrane to further detach the aggregates and then dialyzed against pure water through a dialysis membrane (M_w = 1,000 Da) for 8 h. The product was lyophilized to obtain dark brown CDs and exhibited strong fluorescence under UV irradiation.

To detect of NO₂ ^-, different concentrations of NO₂ ^- solutions (10 mM, final concentration 0–700 μM) were added systematically into 3 ml aqueous solutions of CDs (20 μL; the final concentration is 20 μg ml^-1), then the sample was oscillated for 5 min at room temperature with a small oscillator at 1,000 rpm. Finally, the emission spectrum of the sample was measured by fluorescence spectrometer at the excitation wavelength of 360 nm. To verify detection selectivity of CDs toward NO₂ ^-, other ions solutions were examined in a similar way. For the assay of SO₃ ^2-, various concentrations of SO₃ ^2- (10 mM, final concentration 0–700 μM) were obtained by diluting the stock solution with ultrapure water. The subsequent experimental procedure is consistent with the NO₂ ^- detection process.

The fluorescence quantum yield is the efficiency of converting absorbed photons into emitted photons (Grabolle et al., 2009). For the unknown sample relative fluorescence quantum yield, we can according to the known absorption and emission of relatively perfect quantum yield standard such as rhodamine 101, quinine sulfate and rhodamine 6G to obtain (Wurth et al., 2013). The QY of CDs was measured using quinine sulfate (55%) as standard (Olmsted, 1979) and was calculated using following equation: Q Y s = Q Y s t ( A s t / A x ) ( I x / I s t ) ( η s / η s t ) 2

Where A_st refers to the absorbance of the standard, A_x is the absorbance of the sample to be tested, I represent the emission intensity integral, η represents the refractive index of the solution. The subscript st represents the standard (quinine sulfate), and s represents the sample to be tested (CDs). For more reliable results and to minimize errors, A_s and A_st were less than 0.05.

In briefly, HeLa cells were cultured in 0.4% penicillin streptomycin and 10% fetal bovine serum for 24 h in a 5% CO₂ incubator at 37°C, and then the cells were diffused into 96-well plates (100 μL per well, 5,000 cells) and treated with CDs at different concentration (0–500 ug mL^-1). After incubation for another 24 h, cytotoxicity of the CDs for HeLa cells was evaluated via a WST assay. The absorbance of each well was measured by a microplate reader at 450 nm after 4 h.

HeLa cells were seeded on the coverslips in 6-well plates and incubated at 37°C under 5% CO₂ in DMEM medium containing 10% FBS and 1% penicillin-streptomycin for 24 h. Subsequently, HeLa cells were treated CDs (200 μg ml^-1) for a period of 24 h and washed three times with PBS for imaging. For the detection of NO₂ ^- and SO₃ ^2-, these cells were incubated with 500 μM NO₂ ^- for 0.5 h. In order to restore the intracellular fluorescence, SO₃ ^2- (500 μM) was added and incubated for another 0.5 h. After washing the cells three times with PBS, the fluorescence images of the samples were observed using a confocal laser microscopy.

The morphology, surface functional groups, structure and composition of the CDs were investigated by transmission electron microscope (TEM), Fourier transform infrared spectroscopy (FT-IR), Particle size analyzer (ZS nano 90) and X-ray photo-electric spectrometry (XPS). As illustrated in Figure 2A, the TEM image and the corresponding histogram of size distribution (Supplementary Figure S1) illustrated that CDs with the average particle size of 5.6 nm were uniformly dispersed and spherical shape, indicating they were water-soluble. High resolution TEM image (insert in Figure 2A) showed that the particles have a lattice structure and lattice constant is 0.18 nm. In the FT-IR spectrum of the CDs (Figure 2B), the broad peak at 3,360 cm^-1is the telescopic vibration from O-H, the peak at 2,922 cm^-1 corresponds to the stretching vibration peak of C-H and the peak at 1710 cm^-1 comes from the stretching vibration of C=O (Ge et al., 2015; Li et al., 2015). The characteristic peak at 1,522 cm^-1, 1,202 cm^-1 and 1,022 cm^-1 corresponds to the stretching vibration peak of C=C bond, C-C bond and C-OH bond (Zhi et al., 2018), indicating the existence of hydroxyl and various other moieties (such as C-H, C=O and C=C) in the CDs.

Then, XPS was performed to further identify the structural information of CDs. The XPS full scan spectrum in Figure 3A contains two distinct peaks at 284.8 (C 1s) and 532.8 eV (O 1s). Further, the major peaks at 284.5, 285.7 and 287.2 eV in the high-resolution C 1s spectrum are respectively the signal peaks of C-C/C=C, C-O and C=O groups. The O 1s XPS spectrum of CDs can be decomposed into peaks at 531.8 and 532.9 eV corresponded to the C=O and C-OH groups. These results further confirm that the presence of carboxyl and hydroxyl functional groups on surface of CDs, which is consistent with the results of the FT-IR spectrum.

To further investigate the optical properties of CDs, UV-vis absorption and fluorescence spectroscopy were also performed. Figure 4 exhibits an intense absorption peak at 284 nm, which is mainly originated from the π-π transition of C-C bond. Moreover, the maximum fluorescence emission intensity of CDs is located at 435 nm and excitation wavelength at 360 nm. In addition, CDs preserved stable fluorescence in solution with a wide range of pH values from 1 to 10. As demonstrated in Figure 4, the fluorescence intensity of CDs changed slightly by 5% under extreme acidic conditions, which may be due to a large number of hydroxyl groups on the surface of CDs. However, strong alkaline conditions can seriously affect the intensity of CDs. The NO₂ ^- and SO₃ ^2- detections in this work were all performed under natural conditions. Besides, the stability of the fluorescence intensity of CDs solution after storage for different time periods was also evaluated. The fluorescence intensity of the CDs solution decreased by only 13% during 6-days storage period, which indicate the good fluorescence stability of the CDs. And the relative quantum yield of CDs is 12.2% by using quinine sulfate as a reference. The robust fluorescence stability makes the CDs suitable for further bioimaging applications.

The “on-off-on” fluorescent probes based on CDs were developed to detect NO₂ ^- and SO₃ ^2-. As shown in Figure 5A, CDs have a specific binding ability with NO₂ ^- and the emission fluorescence intensity of CDs at 435 nm was gradually quenched along with the increasing concentration of NO₂ ^-. Also, the fluorescence intensity of CDs was quenched over 60% after adding of NO₂ ^- at the concentration of 400 μМ, and then, the downward trend slows down with the NO₂ ^-. Furthermore, it is worth to point out that there was an excellent linear relationship (R ² = 0.999) between the fluorescence ratio (F₀-F)/F₀ and the concentration of NO₂ ^-, where F₀ and F are the fluorescence intensities of the CDs in the absence and presence of NO₂ ^-. In addition, the limit detection of CDs for NO₂ ^- was 0.425 μМ (LOD = 3σ/S, where σ is the standard deviation of the blank and s is the slope of the linear calibration plot). With the addition of SO₃ ^2-, the fluorescence intensity of CDs gradually recovered. As shown in Figure 5B, the emission intensities of this probe at 435 nm were recorded at 15 min after adding various concentration of SO₃ ^2-, which showed good linear relationship (R ² = 0.998) between the fluorescence ratio (F-F₀)/F₀ and the concentration of NO₂ ^- in the range of 200–600 μМ. The limit detection for SO₃ ^2- was determined to be 0.243 μМ. The detection performance of CDs based “on-off-on” fluorescent sensor was comparable to previous reports (Table 1), articulating the availability and simplicity of the proposed sensing probe. Therefore, the results show that CDs can be considered as a good fluorescent probe for monitoring the concentration of NO₂ ^- and SO₃ ^2- with excellent sensitivity.

In order to evaluate the selectivity in sensing response of CDs, various metal ions (Cu²⁺, Fe³⁺, Ca²⁺, K⁺, Na⁺), anions (NO₃ ^-, PO₄ ^3-, SO₄ ^2-, SO₃ ^2-) and organic molecules (Gly, GSH, Cys, Urea) were considered. Figure 6A displays the fluorescence intensities of CDs in the presence of NO₂ ^- as compared to multiple interfering ions. The fluorescence intensity was reduced by 90% by NO₂ ^- (500 μМ). Therefore, CDs show desirable selectivity for the detection of NO₂ ^-. Although, Fe³⁺ affect the fluorescence intensity of CDs. Fortunately, the concentration of Fe³⁺ in plasma is low (Supplementart Table 2) and the false signals can be effectively shield by triethanolamine. Thus, CDs have potential for directive and selective detection of NO₂ ^- ions. Moreover, to evaluate the selectivity in sensing response of CDs to SO₃ ^2-, various metal ions (Ca²⁺, K⁺, Na⁺), anions (SO₄ ^2-, HSO₃ ^2-) and organic molecules (Gly, GSH, Glu, Cys, Urea, Suc) were investigated for their impact on the fluorescence intensity of CDs/NO₂ ^-. As illustrated in Figure 6B, the fluorescent responses and the corresponding luminescence variations of organic molecules and metal ions was negligible compared to the presence of SO₃ ^2-. The degree of fluorescence intensity recovered by HSO₃ ^2- was equivalent to 57.8% fluorescence recovered by SO₃ ^2-. Therefore, the results confirmed that CDs had great potential for specifically detecting NO₂ ^- and SO₃ ^2-.

To understand the fluorescence quenching mechanism of CDs, fluorescence lifetime decay, zeta potential, UV-vis absorption (Supplementary Figure S2) and electrochemical behaviors (Supplementary Figure S3) were investigated. After adding NO₂ ^- and SO₃ ^2-, signals of UV-vis absorption were almost unchanged and these results were consistent with observation in the cyclic voltammogram (CV) after adding NO₂ ^- and SO₃ ^2- (Supplementary Figure S3). The CV of CDs just showed a reversible redox reaction. Moreover, the fluorescence lifetimes were respectively 1.8 ns (CDs only), 2.03 ns (in the presence of NO₂ ^-) and 2.55 ns (in the presence of NO₂ ^- and SO₃ ^2-) in the Supplementary Figure S4, which displayed no obvious change and it is different from a dynamic fluorescence quenching mechanism, suggesting a static fluorescence quenching effect occurred. In detail, the zeta potential of CDs solution was measured as -36.85 mV, which indicates that the nucleus of CDs is positively charged and the surface is rich in anions. After the introduction of NO₂ ^-, it replaces the anions on the surface of CDs. And the strong electron absorption of NO₂ ^- makes it difficult for the electrons in the CDs core to be excited, leading to fluorescence quenching. When SO₃ ^2- was introduced, SO₃ ^2- would reduce NO₂ ^- and destroy the charge transfer between NO₂ ^- and CDs, resulting in fluorescence recovery (Figure 7).

Before imaging, the cytotoxicity of CDs to HeLa cells was assessed using WST assay. Various concentrations of CDs (20, 50, 100, 200, 300, 400, 500 μg ml^-1) were added to Hela cells and cell viability was observed more than 90% after incubating HeLa cells with CDs for 24 h (Supplementary Figure S4). Owing to their low cytotoxicity and excellent biocompatibility, the fluorescent probe was used to image NO₂ ^- and SO₃ ^2- in live cells.

The possibility of CDs to be as a label agent for fluorescent bioimaging of NO₂ ^-and SO₃ ^2- was tested by a confocal laser microscopy. As shown in Figure 8, Hela cells treated with CDs solution (200 μg ml^-1) exhibit blue fluorescence and retained normal morphology. Upon incubation with NO₂ ^- for 20 min, the fluorescence in the living cells was significantly quenched. After adding SO₃ ^2-, the fluorescence recovered effectively. These results are consistent with the observed results in spectral experiments, which suggest that the fluorescent probe have the potential to detect NO₂ ^- and SO₃ ^2- in living cells.