Synthesis and characterization of BPQDs and R-BPQDs. BPQDs were prepared with a solvothermal method5,6, and then functionalized with Arg to obtain R-BPQDs (Fig. 1a). From the transmission electron microscopic (TEM) images, their average lateral sizes were measured to be about 4 nm (Figs. 1b,c). The lattice fringes of 0.23 nm could be ascribed to the (041) plane of the BP crystal (Inset in Fig. 1b)21. The measured heights of 1.1, 1.8 and 2.4 nm (Supplementary Fig. 1) corresponded to BPQDs with about 2-4 layers, respectively5,6. The Fourier transform infrared (FTIR) spectra showed the shifts of P−O and N−H bending peaks from 1070 cm-1 of BPQDs and 1721 cm-1 of Arg to 1089 and 1705 cm-1 of R-BPQDs, respectively (Fig. 1d)22,23, indicating the electrostatic and hydrogen bond interactions between the electron-withdrawing guanidine group in Arg and PxOy moiety24,25. The Arg modification also led to a slight red-shift of three prominent Raman peaks of BPQDs (Fig. 1e), identified as one out-of-plane phonon mode (A1g) at 360.1 cm-1 and two in-plane modes (B2g and A2g) at 435.5 and 462.7 cm-1 5,6, which manifested a decrease of the Raman scattering energy due to the interaction between guanidine group and PxOy moiety26. The R-BPQDs retained the characteristic P2p3/2 and P2p1/2 X-ray photoelectron spectroscopic (XPS) peaks of BP at 128.6 and 129.3 eV (Fig. 1f and Supplementary Figs. 2a,b), while BPQDs did not show these peaks, demonstrating that the introduction of Arg endowed BPQDs with better stability. The shifts of P−O and P=O XPS peaks from 132.6 and 133.6 eV of BP to 134.8 and 135.9 eV of R-BPQDs could be attributed to the decrease of the outer valence electron density of P atom in the presence of guanidine group (Fig. 1f and Supplementary Fig. 2b), which did not obviously change the N1s XPS peak due to the delocalized positive charge distribution over the guanidine group (Supplementary Figs. 2c,d), but increased the binding energy of O atom for P−O and P=O bonds due to the decreased outer valence electron density (Supplementary Figs. 2e-h). In a word, the presence of Arg on the PxOy moiety of BPQDs exerted a passivation effect against the oxidation defects of BPQDs (Fig. 1g).
Photophysical and electrochemical properties of BPQDs and R-BPQDs. The cathodic process of BPQDs/GCE showed an ambiguous reduction peak of BPQDs at around −1.22 V, which shifted to −1.15 V and became more distinct after introducing the electron-withdrawing guanidine group by Arg modification (Fig. 2a). In the presence of K2S2O8, R-BPQDs/GCE showed a cathodic ECL emission at −1.20 V, which was 25 folds stronger than that of BPQDs/GCE (Fig. 2b). The ECL emission followed a general K2S2O8 mediated co-reactant ECL mechanism containing the reduction of K2S2O8 and BPQDs or R-BPQDs to produce the excited state BPQDs* or R-BPQDs* (Inset in Fig. 2b). The reduction peak of K2S2O8 at bare GCE occurred at −0.95 V, and almost disappeared at BPQDs/GCE due to the greatly increased electron transfer impedance (Re) (Supplementary Fig. 3). The much lower Re than BPQDs/GCE led to obvious reduction peak of K2S2O8 at R-BPQDs/GCE, which overlapped with the reduction peak of R-BPQDs (Fig. 2a). The consistence of onset potentials for electrochemical reduction and cathodic ECL emission of R-BPQDs indicated that the electrogeneration of R-BPQDs•− was necessary for the formation of excited state R-BPQDs*27. Interestingly, the ECL depended on both Arg concentration and pH for preparation of R-BPQDs (Supplementary Fig. 4), implying that the amount of Arg assembled on BPQDs affected the ECL performance, though Arg did not participate in the ECL process (Supplementary Fig. 5).
The FL and PL emission of BPQDs centering at 505 and 580 nm (Fig. 2c) originated from the oxidation defects associated S1 and T1, respectively8,15, which was demonstrated by the photoluminescence decay spectrum 15 (PDS) (Supplementary Fig. 6). The PDS of both BPQDs and R-BPQDs showed biexponential functions with lifetime components of 3.08 and 12.27 μs and 1.82 and 13.14 μs, respectively(Supplementary Table 1), indicating the presence of two decay channels assigned to the two excited states for FL and PL emissions, respectively28. R-BPQDs displayed 45-nm and 20-nm blue shift and the increased FL and PL emission intensity, respectively. The hypochromic shift and intensity enhancement were related to the change of the two excited states due to the passivation of oxidation defects by Arg. Excitingly, the cathodic ECL spectra of both BPQDs and R-BPQDs displayed two emission peaks at 500 and 580 nm for BPQDs, and 460 and 570 nm for R-BPQDs (Fig. 2d). Compared with the FL and PL spectra of BPQDs and R-BPQDs (Fig. 2c), the ECL emission peaks at 500 and 460 nm could be attributed to the transition from S1, while the peaks at 580 and 570 nm originated from T1-to-S0 transition. Thus it could be concluded that ECL emissions of BPQDs and R-BPQDs contained two radiative transitions from both S1 and T1, as shown in Fig. 2e. After Arg modification, the hypochromic shifts of both S1-to-S0 and T1-to-S0 transitions were observed in the ECL spectra, which was consistent with larger experimental band gap of R-BPQDs than BPQDs (Supplementary Fig. 7). However, their lifetimes did not obviously change at temperatures from 170 to 310 K (Supplementary Fig. 8 and Table 2), so the possibility of thermally activated delayed fluorescence process could be excluded4,29, further indicating the existence of two radiative transitions from both S1 and T1 for ECL emission.
According to the electrostatic and hydrogen bond interactions between Arg and oxidation defects of BPQDs, TD-DFT computation was implemented to rationalize above conclusion. The existence of oxidation defects resulted in the localized HOMO of BPQDs at the defect sites (Fig. 2f), which hindered the charge transfer as a trap state, and thus weakened the FL, PL and ECL intensity. The Arg modification passivated the surface oxidation defects, accordingly leading to the delocalization of HOMO of R-BPQDs to the central zone (Fig. 2f), and thus the change of the electron transition channel, which significantly improved the emission oscillator strength and the charge transfer capability (Supplementary Fig. 3). Besides, R-BPQDs exhibited the strongest ECL emission and the most positive reduction potential among the BPQDs modified with 20 kinds of amino acids (Supplementary Table 3), demonstrating the significance of electron-withdrawing guanidine group, which stabilized the adjacent R-BPQDs•− anion radical after electrochemically injecting electron into LUMO of R-BPQDs30,31, and thus facilitated the cathodic ECL emission.
The BPQDs/GCE showed two anodic peaks at +0.85 and +1.45 V (Fig. 3a), which were attributed to the electrochemical oxidation of surface groups such as phosphite and hypophosphoric groups25,32. These peaks negatively shifted to +0.42 and +0.86 V after Arg modification (Fig. 3a) due to the much lower Re (Supplementary Fig. 3), which decreased the oxidation overpotentials. Although the oxidation of Arg could be observed at bare GCE at +1.62 V (Supplementary Fig. 9), it did not occur at R-BPQDs/GCE in the applied potential range due to the relative higher Re. Considering the low oxidation potential of R-BPQDs, N2H4∙H2O that can be electrochemically oxidized to produce N2H3• and N2H2 around +0.1 V was used as co-reactant to study the anodic ECL of BPQDs and R-BPQDs33,34. At bare GCE, the oxidation of N2H4∙H2O occurred near +0.10 V, which showed a peak at +0.55 V and a severe tailed anodic curve due to the further oxidation of N2H3• and N2H2 at higher potentials (Fig. 3b). Obviously, the anodic curve of N2H4∙H2O positively shifted due to the increased Re, and covered the oxidation peaks of BPQDs at both BPQDs/GCE and R-BPQDs/GCE. Furthermore, the hole-injected BPQDs (BPQDs•+) could oxidize the reducing N2H3•, N2H2 and N2H4 to form the excited state species BPQDs* or R-BPQDs* for ECL emission (Fig. 3c). The anodic ECL peak potential and intensity of R-BPQDs were 0.15 V lower and 2 times higher than those of BPQDs, which could be attributed to the better charge transfer capability, the greater spatial overlap between HOMO and LUMO, and the better stability under ambient conditions (Supplementary Fig. 10) after Arg modification. Similar to the cathodic ECL process, the anodic ECL spectra of BPQDs and R-BPQDs also displayed two emission peaks associating S1 and T1 transitions, along with the hypochromic shifts (Supplementary Fig. 11). Thus the co-reactant ECL mechanisms of R-BPQDs at the cathode and the anode could be illustrated in Fig. 3d.
ECL transient technology was further used to examine the stability of radical intermediates in two ECL processes of R-BPQDs. The ion annihilation ECL intensity at +1.60 V was stronger than that at −1.40 V in the absence of coreactant (Fig. 3e), indicating that the anion radical R-BPQDs•− was more stable than cation radical R-BPQDs•+35. Thus Arg stabilized the anion radical of BPQDs under ambient conditions, and thus led to the greater enhancement of cathodic ECL intensity than the anodic process.
Evaluation of integrin inhibitor with RRGDS-BPQDs. To implement the application of R-BPQDs/K2S2O8 ECL system, this work designed an ECL method to evaluate the inhibiting efficiency of integrin inhibitor, cyclo(RGDyK)36, by using Arg containing peptide RRGDS to modify BPQDs (RRGDS-BPQDs). The RRGDS-BPQDs functionalized carboxylated multi-wall carbon nanotubes (MWNTs) were coated on GCE to act as both the recognition unit and signal tag37. Compared to Arg-free peptide GGGDS, the presence of RRGDS could greatly enhance the ECL intensity (Fig. 4a), verifying the vital importance of Arg for improving the ECL emission of BPQDs. Upon the specific recognition of αV/β3 integrin on A549 cell membrane with RRGDS on GCE, the Re increased greatly (Supplementary Fig. 12), and thus the ECL intensity decreased obviously (Supplementary Fig. 13). In contrary, MCF-7 cells with low abundance of surface αV/β3 integrin showed little decrease. Under optimal conditions (Supplementary Figs. 14,15), the IC50 of cyclo(RGDyK) for 1 ´ 106 A549 cells mL-1 was obtained to be 12.0 nM from the ECL response plot of cyclo(RGDyK) treated A549 cells (Fig. 4b), which was comparable to 20 nM for immobilized αV/β3 integrin36 and 29.3 nM for B16-F10 cells38. Compared to general MTT method, this method was sensitive, simple and convenient. Moreover, this method could be extended for the evaluation of other inhibitors or the detection of cell surface groups by changing the Arg-containing peptide, showing the excellent practicability of the designed ECL modulating strategy.