Near‐infrared fluorophores with absolute aggregation‐caused quenching and negligible fluorescence re‐illumination for in vivo bioimaging of nanocarriers

Environment‐responsive fluorophores with aggregation‐caused quenching (ACQ) properties have been applied to track nanocarriers with reduced artefacts caused by unbound or free fluorophores but suffer from incomplete fluorescence quenching and significant re‐illumination, which undermine bioimaging accuracy. Herein, through structural modifications to reinforce the hydrophobicity, planarity and rigidity of fluorophores with an aza‐BODIPY framework, probes featuring absolute ACQ (aACQ) and negligible re‐illumination are developed and evaluated in various nanocarriers. aACQ probes, FD‐B21 and FD‐C7, exhibit near‐infrared emission, high quantum yield, photostability, water sensitivity, and negligible re‐illumination in blood, plasma and 1% Tween‐80 in contrast to ACQ probe P2 and conventional probe DiR. All nanocarriers can be labeled efficiently by the tested fluorophores. Polymeric micelles (PMs) labeled by different aACQ probes manifest similar biodistribution patterns, which however differ from that of DiR‐labeled PMs and could be ascribed to the appreciable re‐illumination of DiR. Significantly lower re‐illumination is also found in aACQ probes (2%–3%) than DiR (20%–40%) in Caco‐2, Hela, and Raw264.7 cells. Molecular dynamics simulations unravel the molecular mechanisms behind aggregation and re‐illumination, supporting the hypothesis of planarity dependency. It is concluded that aACQ fluorophores demonstrate excellent water sensitivity and negligible fluorescence re‐illumination, making themselves useful tools for more accurate bioimaging of nanocarriers.


INTRODUCTION
levels. [5][6][7][8] Nevertheless, fluorescence "cheats" sometimes, raising concerns over its reliability. [9][10][11] One of the biggest challenges rests with the interfering signals caused by free or dissociated fluorophores in the surroundings, which may bias the readout of nanocarrier-associated fluorescence. [3,4] In fact, the imaging accuracy of some commercial fluorescent probes has been challenged. For example, the commonly utilized near-infrared (NIR) fluorophore DiR (1,1′dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide) gives confusing and misleading signals when employed to probe nanocarriers. [12][13][14][15][16][17][18][19] In brief, quantitative correlation cannot be established between apparent fluorescence intensity and nanocarrier amount owing to interference derived from free fluorophores, fluorescence quenching/saturation, and re-illumination, which altogether undermine the accuracy of bioimaging. [13,14,20,21] In order to improve the accuracy of bioimaging, it is essential to discriminate particle-bound signals from interfering signals derived from unbound fluorophores. In recent years, the application of environment-responsive fluorescent probes based on Förster resonance energy transfer (FRET), aggregation-induced emission (AIE), and aggregation-caused quenching (ACQ) have demonstrated tremendous superiority in exploring the elaborate dynamics of nanocarriers. [4,[22][23][24] In general, commercial fluorophores such as DiR are reported to be less responsive to environmental factors and retain a high fraction of its original fluorescence after dissociation from the particles ( Figure 1A). [4,20] FRET-based fluorophores function by switching signals, changing emission wavelengths and intensity, whereas AIE-based fluorophores work based on a rationale of "restriction of intramolecular movement". [22][23][24] However, re-pairing of donor and receptor molecules, in the case of FRET, or aggregation and repartition into hydrophobic constructs, in the case of AIE, and other complex variables, work together to undermine the accuracy of bioimaging. [22,25] It is ideal that free or unbound fluorophores give no signals at all; therefore the fluorescence observed could be employed to represent nanocarriers. This has been partly realized in our previous studies by exploiting the ACQ properties of NIR fluorophores. [20,[26][27][28][29][30][31][32] ACQ is generally referred to as a phenomenon that fluorophores remain emissive in a dilute solution but quench at elevated concentrations or when aggregates are formed. [4,20] Briefly, the underlying mechanism behind the ACQ phenomenon is the formation of excimers and exciplexes induced by aromatic molecule interactions (π-π stacking). The nonradiative decay of dye aggregates of excited states leads to fluorescence quenching immediately. Other dyes such as cyanines can self-aggregate driven by intermolecular attractive forces at the solid-liquid interface or in solution, which also results in fluorescence quenching. [4] For detailed information, readers can refer to previous reviews. [4,33,34] In bioimaging, ACQ is usually regarded as unfavorable, yet it can be rendered useful when applied to track nanocarriers. [4] ACQ probes, such as those employed in our previous studies with boron-dipyrromethene (BODIPY) and aza-BODIPY frameworks, illuminate nanocarriers when encapsulated in the nanocarrier matrices but form aggregates spontaneously in water with simultaneous fluorescence quenching. The interference caused by free or released fluorophores could be reduced by a certain degree, thus improving the accuracy of bioimaging. [20,27] Nevertheless, previously developed ACQ probes suffer from incomplete fluorescence quenching when applied to in vivo bioimaging. [4,20] The retention of fluorescence could be attributed to (1) transferring of fluorescence to newly formed constructs such as mixed micelles during degradation of nanocarriers and (2) fluorescence re-illumination of quenched aggregates due to repartition into hydrophobic constructs such as protein pockets, membranes, and endogenous micellar/vesicular structures ( Figure 1B). [4,20,35] To eliminate fluorescence retention and further improve bioimaging accuracy, it is hypothesized to strengthen the responsiveness of ACQ probes to the aqueous environment and suppress the re-activation of quenched aggregates by physiological constituents ( Figure 1C).
Previous experience suggests that strengthening the hydrophobicity and planarity of the fluorophore molecules facilitates π-π stacking, [4,36] which may in turn boost water sensitivity and reinforce the cohesion of quenched aggregates. Therefore, the current study aims to develop novel probes with absolute ACQ (aACQ), that is, minimized fluorescence retention and re-illumination, by rigidifying the aza-BODIPY parent structure and/or introducing methoxy or acetoxy groups ( Figure 1D). In vivo and subcellular experiments were conducted to validate their applicability in various nanocarriers including polymeric micelles (PMs), polymeric nanoparticles (PNs), and solid lipid nanoparticles (SLNs). To better illustrate the advantages of novel aACQ probes, comparisons with previously developed ACQ probes and DiR were conducted both in vitro and in vivo. More importantly, as the mechanisms underlying the ACQ phenomenon and fluorescence re-illumination have not been explored at molecular levels, molecular dynamics (MD) simulations are employed to investigate the aggregation processes of the fluorophores in the background of interaction with water and the dissociation of aggregates in response to extraction by hydrophobic solvent. [37][38][39][40][41] It is envisaged that the development of aACQ probes offers a useful tool to improve the bioimaging accuracy of nanocarriers, and the hypothesis of reinforcing the rigidity and hydrophobicity of the fluorophores provides a solution to development of fluorophores with enhanced water sensitivity and negligible free-state interference.

Water sensitivity and in vitro fluorescence re-illumination
To enhance the hydrophobicity, planarity, and rigidity of fluorophores, structural modifications were applied to the aza-BODIPY backbone, such as alternation of substituent numbers and positions and introduction of heterocycles and fused-ring structures, forming a series of novel aACQ probe candidates (FD-B3∼FD-C7) ( Figure 2 and Figure S1, Supporting Information). To highlight the superiority of novel aACQ probes, their performance was compared with P2((T-4)-[4,5-Dihydro-7-methoxy-N-[5-(4-methoxyphenyl)-3-phenyl-2H-pyrrol-2-yliden-Κn]-3phenyl-1H-benz[g]indol-2-aminato-Κn 1 ]difluoroboron), an ACQ probe previously investigated, and DiR, a less environment-responsive commercial fluorophore, whose structures were also provided in Figure 2. The structural and F I G U R E 1 (A-C) Schematic illustration of the rationale of aggregation-caused quenching (ACQ) (B) and aACQ (C) probes in comparison with less environment-responsive probes (A). (A) Conventional less environment-responsive fluorescent probes give off steadfast signals ("on→on") even if they are dissociated from the particles, which makes it difficult to discriminate signals of free dyes from those of integral particles. (B) In contrast, environmentresponsive ACQ probes can improve the accuracy of bioimaging based on "on→off" fluorescence switching, illuminating integral nanocarriers while generating substantially reduced interference signals due to the formation of quenched aggregates in the aqueous environment. Nevertheless, complicated bio-factors can lead to fluorescence re-illumination of ACQ probes due to repartition into the hydrophobic core of micellar structures. (C) aACQ probes are developed to eliminate fluorescence re-illumination through inhibiting re-activation of quenched aggregates by physiological constituents. (D) Chemical modifications are introduced to aACQ probes with an aza-BODIPY parent structure such as FD-B21 and FD-C7 by enhancing hydrophobicity, planarity, and rigidity, which favor the formation of π-π stacking and eventually boost water sensitivity and aggregation cohesion of fluorophores spectroscopic data of tested compounds were provided in Tables S1 and S2 (Supporting Information).
High brightness, NIR emission, and good photostability are fundamental requirements for bioimaging agents. [42] In the present study, seven novel aACQ probes were screened to possess relatively high molar extinction coefficients and quantum yields for bright and clear imaging, as well as NIR emission wavelengths ( Figure 2A and Table S2, Supporting Information), allowing for fluorescence bioimaging in the NIR window with decreased photon scattering, absorption, and autofluorescence. [43] All tested probes showed good photostability with over 80% of their initial fluorescence in acetonitrile maintained after exposure to 150 mW xenon lamp for 10 min ( Figure 2E and Figure S2F, Supporting Information). Noteworthily, aACQ probes and P2 possessed better photostability than DiR. Then the ACQ properties were evaluated in an acetonitrile/water binary system. Although without an acknowledged definition, the water sensitivity of ACQ probes could be simply understood as the ability of hydrophobic fluorescent probes of a certain concentration to aggregate and quench in the water/organic solvent binary system at room temperature. Generally, the lower the water content, the more complete the fluorescence quenching denotes higher water sensitivity of fluorescent probes. The fluorescence of tested probes (FD-B21 and FD-C7) decreased gradually along with the increase of water content. Abrupt and complete fluorescence quenching was observed when the water content reached about 60%-80% ( Figure 2B-D, Figure  S2A-E, Supporting Information), confirming high water sensitivity, which sets the premise for further comparisons regarding fluorescence re-illumination ( Figure 2F,G and Figure S2G-I). According to the fluorescent spectra, all of the NIR dyes selected displayed expected ACQ effect despite the slight difference in sensitivity to water content. We also tested the Ultraviolet-Visible (UV-Vis) spectra of probes (250 nmol L −1 ) in the binary acetonitrile/water mixtures ( Figure S3).  It was noteworthy that in all cases, the absorption peak initially diminished slowly when water content increased and then a new absorption peak generated bathochromic shift when a certain content of water was reached. Significant spectra broadening in absorption were observed after reaching a certain threshold, and this phenomenon could represent the formation of aggregates. [20] The hydrophobicity and large conjugated coplanar structures that favor π-π stacking facilitated the formation of quenched aggregates with enhanced cohesion. [4,[44][45][46][47] Belonging to the same family of aza-BODIPY, P2 also displayed high water sensitivity. It is interesting to find that DiR demonstrated water contentdependent fluorescence quenching as well, contradicting with previous perception that DiR was unable to quench substantially in water. [20] The sensitivity of DiR to pure water could be attributed to its high hydrophobicity, [47] but the flexible long sidechains might weaken the rigidity and cohesion of the aggregate formed and lead to fluorescence retention in complex systems.
Apart from favorable water sensitivity, the newly developed fluorophores also possessed superior anti-reillumination properties in various media, compared with DiR and P2. It was not surprising that all fluorophores retained a fraction of the original fluorescence after being dispersed into and incubated with plasma, blood, and 1% Tween-80 ( Figure 2F,G and Figure S2G-I, Supporting Information). If the fluorophores are dispersed directly into 100% water, there would be no fluorescence retention or re-illumination at all because all tested fluorophores quenched completely at a water content above 60%-80%, as observed in the water sensitivity study ( Figure 2B). The observed fluorescence retention could be ascribed to the partitioning of fluorophore molecules into the hydrophobic cavities of biomacromolecules, lipid constituents in blood or plasma and hydrophobic cores of surfactant-derived micelles and subsequent activation to emit fluorescence. [4] As observed in Figure 2F, both FD-B21 and FD-C7 recovered less fluorescence (<10%) within 48 h than P2 and DiR, which gradually regained up to 15% and 25% of fluorescence, respectively. FD-B21 performed the best in plasma with negligible (<1%) fluorescence retention. Nevertheless, the fluorescence of FD-B3, FD-B12, and FD-B16 underwent more drastic increase within 48 h ( Figure S2G,H, Supporting Information). In the case of FD-B3, when a single methoxy was present at the ortho-position of the phenyl ring, the ring was twisted, undermining the molecular planarity and the formation of π-π stacking. As for FD-B12 and FD-B16, the introduction of pyridine rendered them relatively polarized, which made easier the disassembly of dye aggregates in water. Compared to methoxy, larger groups such as esters were less conducive to π-π stacking, as indicated by the slightly higher re-illumination percentages of FD-C1 and FD-C2 compared to P2. In terms of the influence of surfactants, all aACQ probes tested except FD-B3 exhibited less prominent fluorescence re-illumination (<30%) in 1% Tween-80 for 48 h than P2 and DiR with the highest fluorescence retention of 35.7% and 85.7%, respectively ( Figure 2G, Figure S2I, Supporting Information). FD-B21 also performed the best in 1% Tween-80 with near zero fluorescence retention, whereas FD-C1 and FD-C2 performed just next to FD-B21 with less than 2% fluorescence retention. DiR retained as high as 25% fluorescence at 8 h in blood/plasma and more than 60% fluorescence at all time points in 1% Tween-80. Although the mechanisms not fully known, this helps explain why DiR is usually regarded as a less environment-responsive fluorophore. Notably, FD-B21 and FD-C7 demonstrated minimal re-illumination in all test media, which endowed them with potential superiority to P2 and DiR in terms of substantially reduced interference. Considering the combined results of in vitro studies, FD-B21 and FD-C7 were selected for further investigations.

Preparation and characterization of fluorescently labeled nanocarriers
To test the applicability of aACQ probes in common nanocarriers, probe-encapsulated nanocarriers including methoxy poly(ethylene glycol) 2k -poly(D,L-lactic acid) 2.5k (mPEG 2k -PDLLA 2.5k ) PMs, polycaprolactone (PCL) PNs and Precirol ATO 5 SLNs were prepared and characterized. The particle size, polydispersity index, zeta potential and entrapment efficiency (EE%) of different nanocarriers were displayed in Tables S3-S5 and Figure S4 (Supporting Information). The particle sizes of PMs, PNs, and SLNs were approximately 60, 130, and 110 nm, respectively ( Figure 2H,I). Zeta potentials were within the range of ±10 mV. Transmission electron microscopy (TEM) images indicated spherical morphology and uniform size distribution for all groups of nanocarriers. The high EE% suggested efficient encapsulation of all probes into nanocarriers. Both fluorescence intensity and particle size remained stable in water for at least 48 h ( Figure  S5-S6, Supporting Information). The stable fluorescence is indicative of lack of fluorophore leakage and premature quenching due to infiltration of water into the nanocarrier matrix, both of which are factors that may result in premature quenching of fluorophores. [4] After incubating with fresh blood and plasma, the fluorescence declined for all three nanocarrier groups ( Figure S5A-F, Supporting Information). DiR-labeled nanocarriers underwent more significant attenuation compared with those labeled by aACQ probes. This is not unusual because the interaction of nanocarriers with blood or plasma components such as blood cells, lipids, and proteins may deteriorate the stability of nanocarriers or alter the surroundings of the encapsulated fluorophores, all resulting in attenuated fluorescence. [4,27,30,[47][48][49][50][51][52][53] On the whole, nanocarriers labeled with different probes demonstrated accordant fluorescence quenching patterns in both blood and plasma, which corresponded well with their gradual degradation under in vivo conditions.

In vivo live imaging and evaluation of re-illumination
To assess the applicability of novel aACQ probes in tracking nanocarriers in vivo, biodegradable PEG-PDLLA PMs were selected as model nanocarriers. The biodistribution of fluorescently labeled PMs was evaluated via an in vivo imaging system (IVIS) after intravenous (i.v.) injection, whereas prequenched fluorophores were administered intravenously to access the degree of fluorescence re-illumination ( Figure 3A). As shown in Figure 3B and Figure S7E-L (Supporting Information), FD-B21-, FD-C7-, and P2-labeled PMs exhibited similar biodistribution patterns within 48 h. To offset different fluorescence intensity of the administered PMs labeled by different probes, normalized average radiant efficiency (ARE) was calculated by dividing ARE by the total fluorescence intensity of administered PMs. Predominant accumulation of PMs was found in the liver region as a result of uptake by the mononuclear phagocyte system (MPS). Meanwhile, an average of 3.6-5.8 times higher fluorescence intensity was found in peripheral regions than the trunk regions in an order of genital > snout > hindlegs > trunk, as estimated by normalized ARE ( Figure 3C,E). For FD-B21-labeled PMs, the fluorescent signals in the peripheral organs peaked at around 1 h and underwent gradual declination afterward, while fluorescence peaked in the liver at 4 h post i.v. administration, which could be attributed to the rapid perfusion and growing hepatic uptake of PMs. [27] In the FD-C7-labeled group, hepatic fluorescence reached a plateau after 4 h, while the strongest signals appeared in the snout and hindlegs at 5 min postadministration ( Figure S7A, Supporting Information). Overall, these results suggested substantial accumulation and biodegradation of PMs in the liver as well as rapid perfusion to the peripheral tissues. The DiR-labeled group however took on a different biodistribution pattern. The fluorescent signals in peripheral tissues maximized at 5 min post administration and declined continuously for up to 48 h. Hepatic fluorescence intensity fluctuated within 2 h and peaked at 8 h ( Figure 3B,E). Furthermore, it is noteworthy that i.v. administration of quenched fluorophore dispersions of either FD-B21 or FD-C7 induced minimal fluorescent re-illumination within 48 h ( Figure 3D and Figure S7B, Supporting Information), indicating that fluorescent labeling by aACQ probes would incur negligible pseudo-positive influence at in vivo level. In contrast, the apparent fluorescence re-illumination in P2 and DiR dispersion groups indicated that released dye aggregates could exert more profound influence on their outcomes ( Figure 3F and Figure S7D, Supporting Information). It should be noted that the normalization of in vivo fluorescence by initial fluorescence of the administered nanocarriers was meant to enable comparison between different probes. However, the normalized percentages should not be utilized as a quantitative measure to evaluate in vivo re-illumination, which is yet unmeasurable by available technologies because of the dynamic nature of in vivo processes and fluorescence attenuation by tissues.

Ex vivo imaging
To further profile the biodistribution patterns of PMs following i.v. administration, animals were sacrificed and various organs were dissected for IVIS imaging. After 1 h, PMs  Figure  S16, Supporting Information mainly accumulated in MPS organs including the liver, spleen and lungs, as well as fatty tissues, heart and kidneys. In contrast, fluorescence intensity in the muscle and skin was minimal, and no distribution to the brain was observed within 24 h ( Figure 4A1-A4, Figure S16A,B, Supporting Information). Fluorescence quantification as normalized ARE demonstrated similar biodistribution patterns of PMs when labeled by aACQ probes (Figure 4B1-B3). Accumulation in the tested organs/tissues complied with an order of liver > lung > spleen > heart ≈ fatty tissue > kidney. Fluorescence, expressed as normalized ARE, peaked at 4 h post i.v. injection and declined gradually afterward, which indicated the gradual accumulation of PMs in the first 4 h, followed by fluorescence quenching of released ACQ probes owing to degradation of the nanocarriers. Although FD-B21-and FD-C7-labeled PMs displayed similar biodistribution patterns with the P2-labeled counterparts, the re-illumination percentages of quenched dye dispersions varied among three probes. Of note, minimal fluorescence re-illumination was observed for the quenched dispersions of FD-B21 and FD-C7 within 24 h ( Figure 4C1-C3). The re-illumination percentages in the liver and spleen were lower than that for P2 ( Figure 4D1-D4). Surprisingly, in sharp contrast, PMs labeled with DiR took up a distinct biodistribution and accumulation pattern. As observed, fluorescence intensity in the liver and spleen increased continuously within 24 h ( Figure 4A4,B4), which however was contradictory to the biodegradable nature of PEG-PDLLA. Moreover, fluorescent signal in fatty tissue was stronger than that in the lungs, heart, kidneys, and peripheral tissues. The intense fluorescence re-illumination of DiR in hydrophobic biological components may offer an explanation for this unusual phenomenon. As observed in Figure 4C4 and Figure S16D,H (Supporting Information), quenched DiR dispersion rapidly regained significantly stronger fluorescence in the liver and spleen only 1 h after i.v. injection, which even exceeded that of fluorescently labeled PMs. Therefore, significant re-illumination signals of DiR aggregates could be misinterpreted as integral carriers, concealing the true accumulation state of PMs in these organs. Taken together, it is safe to suggest that along with the degradation of polymeric nanocarriers and probe leakage in vivo, novel probes would form robust, nonemissive aggregates less inclined to repartition and re-illumination, which would exert negligible probe-associated interference compared to conventional probes. Nevertheless, the estimation was based on the condition that carriers underwent complete destruction and abrupt dye leakage upon i.v. injection, taking into account the in vitro stability of PMs, the actual quantity and interference of quenched aggregates would be lower than the calculated results. [27]

Cellular uptake and re-illumination
Caco-2, Hela, and Raw264.7 cell lines were employed to investigate the cellular uptake of fluorescently labeled nanocarriers and re-illumination of prequenched fluorophores at cellular levels. The cellular uptake percentages of fluorescently labeled nanocarriers fell into the range of 1%-2% in three cell models ( Figure 5A-C). The relatively large size of nanocarriers might be one of the factors that contribute to the relatively low cellular uptake efficiency. In agreement with in vivo results, prequenched FD-B21 and FD-C7 dispersions demonstrated minimal fluorescence re-illumination (about 2%-3%) in all cell types, compared to that of P2 (less than 8% in Caco-2, and about 3% in Hela and Raw264.7). However, DiR exhibited much more significant re-illumination in all cell models with about 25.7%, 40.4%, and 21.7% of re-illumination in Caco-2, Hela, and Raw264.7 cell monolayers, respectively ( Figure 5D-G), further strengthening the deduction that re-illumination of quenched dyes might produce artefacts that conceal the actual cellular uptake states. Although common calculation methods were used here to estimate cellular uptake efficiency, which might incur interference considering fluorescence re-illumination, especially for DiR-labeled group, the results helped to shed light on the potential influence of significant re-illumination and the inaccuracy of common calculation methods.

System construction and MD simulation of the aggregation/dissociation process
In order to investigate the mechanisms underlying the ACQ and fluorescence re-illumination process, theoretical approaches of molecular and quantum mechanics were employed to unravel the inner mechanics. In addition to the probes tested in above experiments, another fluorophore FD-B3, utilized as counter-evidence, with high re-illumination was also included in this experimental section. To balance the computing consumption and precision, five different probe systems with different unit numbers were constructed. The gyration of radius (Rg) of each duodenary unit showed that all probes could aggregate swiftly in pure water during 100 ns simulation ( Figure 6A, Video S1-S3, Supporting Information). Upon incorporating into pure acetonitrile, all aggregation clusters were ready to dissociate. The Rg revealed no aggregation between each probe, indicating the dispersive states of probes in pure acetonitrile ( Figure 6A,B).

Varied aggregation stability of different probes
Previous experimental results informed us that although the tested probes were completely quenched in the aqueous environment, quenched aggregates could transform to emissive states at both in vitro and in vivo levels to different degrees. Herein, MD simulations were performed to provide more evidence about fluorescence re-illumination from a microscopic perspective. With the addition of 30% acetonitrile to simulate the hydrophobic influencing factors, the aggregation states varied drastically among tested probes (Video S4-7, Supporting Information). At the beginning of the simulation, each system fluctuated slightly to adapt to a new solvent environment and became stable swiftly ( Figure S17, Supporting Information). No obvious fluctuation of root mean square deviation (RMSD) and Rg was observed in system FD-B21, FD-C7, and P2 after ∼30 ns, indicating their relative stable aggregation conformation (Figures 6C and 6E1,E2). However, the distinct and continuous increasing trend of RMSD and Rg for probe FD-B3 and DiR were captured as an indicator of system instability and increased volume. The RMSD and Rg of FD-B3 increased from 0.25 and 1.15 Å to 0.45 and 1.25 Å (Figures 6C and 6D1,D2), respectively. Moreover, DiR demonstrated the most significant changes in both RMSD and Rg, which underlined the gradual disassembly of DiR cluster under the effect of acetonitrile ( Figures 6C  and 6D1,D2). Noteworthily, the acetonitrile molecules were prone to aggregate around DiR, which led to the decline of adjacent water concentration and consequently decreased hydrophobic interaction, thereby facilitating and accelerating the disassembly of DiR cluster.

Planarity-dependent aggregation stability of aACQ probes
In view of the poor aggregation stability of DiR clusters, further comparisons were conducted among the ACQ probes to provide more insights about the anti-rekindling mechanisms. The quaternary units revealed the basic aggregation state of each system. The FD-B3 system was apparently different from the other systems with an odd aggregation state and lower planarity, with almost twice as high molecular planarity parameter (MPP) and span of deviation from the plane (SDP) compared with those of FD-B21, FD-C7, and P2. The binding free energy (ΔG) given by the umbrella sampling (US) method from the binary systems showed that the binding affinity in water was −55.8, −52.3, −50.8, and −48.2 kJ mol −1 for FD-B21, FD-C7, P2, and FD-B3, respectively. Hence, the whole probe cluster of FD-B3, which represented the lowest affinity, might not be as stable as the others. Herein, the calculation about FD-B3 with low affinity led us to new clues. Figure 7B1-B4 showed the Independent gradient model based on Hirshfeld partition (IGMH) analysis of wavefunction from the representative conformation of quaternary units. The large interaction area and thickness (indicating more overlap of electron cloud) of sign(λ 2 )ρ were found between benzene rings (so-called π-π stacking) in FD-B21, FD-C7, and P2. However, only a tiny and thin area among the molecule 3 to 1 and 4 in Figure 7B4 might result in the odd aggregation state of FD-B3 because these molecules were hard to form π-π stacking in this system. In sharp contrast, the small and dense contact surfaces were observed in DiR clusters ( Figure 7C), which indicated a general hydrophobic interaction rather than π-π stacking. The interaction strength of π-π stacking relied on the number of rings. Each of our probes has at least four rings, therefore, according to the noncovalent binding strength research of π-π stacking by Grimme the interaction energy of π-π stacking with four rings was around −16.0 kcal/mol (≈−66.9 kJ/mol), which is almost equal to the strongly charged h-bonds. [54,55] And due to the structural property, there is no classical hydogen donor and acceptor exposing to water. Considering the results of F I G U R E 6 Aggregation and dispersive states of tested probes. (A) Radius of gyration of tested probes in acetonitrile (above) and water (below). (B) The graphical representation of 12 probes for each system in acetonitrile. (C) The re-illumination snapshot of five probes (color in yellow: FD-B21, cyan: FD-C7, pink: P2, purple: FD-B3 and ice blue: DiR) at 100 ns with a fixed solvent box (blue cube). The fluctuation of (D1) Rg and D2) root mean square deviation (RMSD) during 100 ns and zoom in (E1) and (E2), respectively. The aggregation processes of FD-B21, P2, and DiR were provided in Video S1-3 (Supporting Information), and the re-illumination processes of FD-B21, FD-B3, P2, and DiR are provided in Video S4-7 (Supporting Information), respectively potential of mean force (PMF), the molecular flexibility and van der Waals repulsion, it is reasonable to hypothesize the ππ interaction plays an essential role in the aggregation process of probes.
Consequently, subsequent research on molecular planarity was conducted with this query. Figure 7D,E suggested the MPP and SDP of FD-B3 were higher than the other probes, and we could easily capture the deflected dihedral of the probe structure from Figure 7F. And the degree of this dihedral (FD-B3: atom C54-C55-C58-N67) was detected as 50.7 • , which represented an approximately 20 • increase compared with the other probes ( Figure 7G). To extensively discuss causal relationship between the structural planarity and position of methoxy groups, we also found a noticeable deviation from the plane at methyl (the bluest) on FD-B3 in Figure 7F, which was the ortho-position of the benzene. However, the methoxy groups on other probes were all located in the para-position. Interestingly, we did not observe the obvious relationship between the number of methoxy groups and probe properties compared to positions of methoxy groups. Figure 7F showed our four probes (FD-B21, FD-C7, P2, and FD-B3) have four, one, two, and two methoxy groups, respectively. The FD-B21 (four methoxy groups on para-position) and FD-C7 (one methoxy group on para-position) exhibit excellent fluorescence and re-illumination properties. And the P2 (two methoxy groups on para-position) also showed the good property. However, FD-B3 (one methoxy group on para-position, another on ortho-position) showed bad aggregation ability and planarity. Therefore, it is rational to deduce that the steric hindrance of the methyl affected the planarity of FD-B3 and in turn hindered the overlap of electron cloud. Overall, the location of functional groups and the basic binding pose were the key factors influencing the re-illumination of aACQ probes. According to the previous work, [56] there is a nodal plane near the 2,6-positions in the lowest unoccupied molecular orbital (LUMO) but not in the highest occupied molecular orbital (HOMO), while there are nodal planes near the 1,7positions in the HOMO but not in the LUMO. Although there are no nodal planes, there are larger MO coefficients at the 3,5-positions in the HOMO than in the LUMO. Therefore, it can be reasoned that the influence of the substituent group at the 3,5-positions is greater than that of substituent group at the 1,7-positions. Thus, the spectroscopic properties of P2 F I G U R E 7 Planarity-dependent anti-re-illumination properties of absolute aggregation-caused quenching (aACQ) probes via molecular dynamics (MD) simulations. A detailed microscopic analysis of (A) probes aggregation from binary units to duodenary clusters in water using bias/unbiased MD simulation and (B) independent gradient model based on Hirshfeld partition (IGMH) analysis to represent the weak interaction between probes with Gaussian16 and Multiwfn. For A1-A4, selected FD-B21 as an example. The boxes in a1 in turn represented the duodenary cluster state in water, the quaternary units with molecular planarity parameter (MPP) and span of deviation from the plane (SDP) parameter (Å), and the binary binding pose with binding free energy (ΔG, kJ mol −1 ). For b1-b4, sign(λ 2 )ρ colored IGMH δ g inter = 0.004 a.u. isosurfaces of representatively quaternary units. The units were defined as the four fragments in the IGMH analysis. The coloring method of sign(λ 2 )ρ was given in Figure S18 (Supporting Information). (C) Aggregation conformation of duodenary and quaternary units for probe DiR in pure water and relevant IGMH analysis (below, right). Quantitative and graphical investigation of molecular planarity using Multiwfn and visual molecular dynamics (VMD). The total change of (D) MPP and (E) SDP for each probe. (F) Graphical representation of probes with different planarity (Å) (the color to blue or red suggested a lower planarity, and the color to white showed absolute plane). (G) Dihedral change during whole simulation process (the bubble in the structure in 2C was used to emphasize the key dihedral) are similar to FD-B21. When a single methoxy is present in the ortho-position of the phenyl ring (FD-B3), low quantum yield was observed. This feature indicated that the phenyl ring in FD-B3 is twisted, and electron transfer might take place.
ACQ probes offer great prospects in bioimaging due to their "on→off" signal transition mode, which significantly reduces the interference of released probes and enhances the veracity of signal judgment. Although previous studies have suggested less significant influence of quenched P2 dispersions, [26,27] it is important to further enhance the ability to resist fluorescence re-illumination since multiple factors in the biological environment may reactivate quenched aggregates. Herein, through chemical modification of aza-BODIPY frameworks, novel aACQ probes were successfully synthesized. Investigation of the optical properties of novel probes confirmed their eligibility for ACQ-based bioimaging, and the interference of novel aACQ probe dispersions was thoroughly investigated, compared with P2 and DiR. By comparing the fluorescence re-illumination profiles in different media, FD-B21 and FD-C7 was preliminarily screened out for further evaluations. Animal and cellular studies combined to confirm their applicability in multilevel bioimaging, as well as the ability to withstand fluorescence reactivation in different biological contexts. MD simulations indicated varied aggregation mechanisms for ACQ probes and DiR and revealed the planarity-dependent aggregation stability of aACQ probes via π-π stacking. Taking into account all these investigations, it is safe to propose that the distinct molecular structure featuring enhanced hydrophobicity and planarity could endow FD-B21 and FD-C7 with tremendous superiority in terms of anti-fluorescence re-illumination properties to facilitate more accurate bioimaging. Meanwhile, the study emphasized the potential misunderstanding derived from DiR-based bioimaging, which proves to be a crucial consideration when investigating the in vivo fate of nanocarriers. Moreover, this phenomenon raises deep concern for DiR-based bioimaging of biodegradable nanocarriers that along with the degradation of nanocarriers in vivo, released DiR might quench transiently but be re-illuminated rapidly with fluorescence even stronger than that in their encapsulated state, resulting in misreading in stability, biodistribution and pharmacokinetic profiles of nanocarriers. The significant differences between DiR and aACQ probes at in vivo and in vitro levels are attributed to their different structures and properties. The hydrophobicity and large conjugated coplanar structures of aACQ probes facilitate π-π stacking, which in turn promotes the formation of quenched aggregates with increased cohesion, incurring negligible re-illumination at both in vitro and in vivo levels. In contrast, DiR, featuring highly hydrophobic structure and flexible long sidechains with small and dense contact surfaces, forms aggregates via hydrophobic interaction rather than π-π stacking, undermining the rigidity and cohesion of aggregates. Following the gradual disintegration of nanocarriers as well as release and aggregation of free probes, the disassembling of unstable DiR aggregates easily leads to more significant re-illumination in vivo. Combining the chemical and MD considerations, the distinct molecular structure especially hydrophobicity and planarity decides the antifluorescence re-illumination properties and consequently bioimaging performances. Nevertheless, further enquiry is required to illustrate the underlying mechanisms of DiR-induced interference, possibly by refining examining time points, investigating the degree of DiR re-illumination following different administration routes or using different loading contents.

CONCLUSION
To further diminish the potential impact of fluorescent probe-associated artefacts, novel aACQ probes FD-B21 and FD-C7 with strong hydrophobicity and planarity to form π-π stacking were successfully developed through chemical modification of the aza-BODIPY backbone. In vitro studies demonstrated reduced re-illumination of aACQ probes compared to P2 and DiR in blood, plasma and 1% Tween 80. In vivo studies demonstrated about 3.6-5.8 times higher fluorescence intensity of ACQ probe-labeled PMs in peripheral areas than trunk. Ex vivo studies exhibited a "rise-fall" distribution to vital organs/tissues with an order of liver > lung > spleen > heart ≈ fatty tissue > kidney. Nevertheless, DiR-labeled PMs demonstrated distinct biodistribution patterns with continuously increased fluorescence intensity in the liver and spleen within 24 h, which was contradictory to the biodegradable nature. A glimpse at the re-illumination offered a reasonable explanation, as aACQ probes led to significantly reduced re-illumination percentages particularly in the liver and spleen for 24 h, compared to P2, while quenched DiR rapidly regained intense fluorescence in these organs, which even surpassed that of PM signals. Moreover, aACQ probes led to significantly lower re-illumination (about 2%-3%) than DiR (20%-40%) in three different cell types. Overall, mutilevel comparisons not only confirmed the minimized fluorescence re-illumination of aACQ probes and their enhanced accuracy for bioimaging but also indicated the potential interference of commercial fluorescent probe DiR. Finally, mechanism studies via MD simulations further suggested a planarity-dependent capacity to form stable quenched dye aggregates and to resist fluorescence re-illumination, which will hopefully facilitate the rational design and selection of effective bioimaging probes.

Synthesis and characterization of aACQ probes
As shown in Figure S1 (Supporting Information), the aACQ probes (P2, FD-B3, FD-B12, FD-B16, FD-C1, FD-C2 and FD-C7) were synthesized according to previous procedures with modifications. [36] The structures of the synthesized compounds were identified with Varian Model Mercury 400 MHz spectrometer (Varian, USA). 1 H NMR chemical shifts (d) were given in ppm (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet) downfield from Me4Si, determined by chloroform (d = 7.27 ppm) and methanol (d = 3.3 ppm, 4.8 ppm). 13 C NMR spectra were recorded on a Varian Model Mercury 600 MHz spectrometer (Bruker, Germany). 13 C NMR chemical shifts (d) were reported in ppm with the internal chloroform at d 77.0 ppm as standard, respectively. The spectroscopic data of the probes were determined. Spectrometer UV-vis spectra were acquired on a 759S UV-visible spectrophotometer (Lengguang Tech, Shanghai, China) and U-2900 spectrophotometer (HITACHI, Japan). Fluorescence spectra were measured by F98 fluorescence spectrophotometer (Lengguang Tech, Shanghai, China). Fluorescent dyes for spectroscopy were prepared as stock solutions in acetonitrile and diluted such that the acetonitrile concentration did not exceed 0.5% (v/v). All measurements were taken at ambient temperature (25 ± 2 • C) in deionized water, unless otherwise noted. Fluorescent quantum yields of products were determined via Equation (1) in acetonitrile using P2 as a reference with Φ of 0.35. The quantum yield was calculated using the following method recommended by Varian (www.jobinyvon.com/usadivisions/Fluorescence/ applications/quantumyieldstrad.pdf, accessed on May 9th, 2022) and was compared with the method reported by Fery-Forgues et al. [57] Φu = Φs ⋅ Fu Φ, F, A, and η referred to the fluorescent quantum yield, fluorescent area, absorption, and solvent refractive index for the unknown (U) or standard (S), respectively. With designed compounds in hand, their photophysical properties were subsequently studied. The photostability of dyes was first tested by exposing their acetonitrile solutions (250 nmol L −1 ) to 150 mW xenon lamp for up to 10 min. Next, the ACQ properties were determined by measuring their UV-Vis spectra and fluorescence spectra in acetonitrile/water binary systems with different water contents on U-2900 spectrophotometer and F98 fluorescence spectrophotometer. The acetonitrile stock solutions with probe concentrations of 50 μmol L −1 were prepared, and 20 μL of the stock solution was added into acetonitrile/water binary systems with different water contents (total 4 mL), so that the final probe concentrations were consistent (250 nmol L −1 ). Respective fluorescent images were taken via an IVIS live imaging system (124262, PerkinElmer Inc., USA) after adding 100 μL of probe samples with the same water contents into 96-well plates. To test fluorescence re-illumination, pre-quenched fluorophore dispersions were prepared by diluting the dimethyl sulfoxide (DMSO) stock solutions with ultrapure water by 200 folds. The fresh blood and plasma were obtained from Sprague-Dawley (SD) rats. Fresh blood was collected from the abdominal aorta after anesthesia of the rats and stored in heparin-pretreated anticoagulant tubes. After thorough mixing, plasma was collected from the supernatant after centrifugation at 5000 rpm for 5 min. All blood and plasma samples were used immediately. Quenched samples were diluted by 10 folds in 1% Tween-80 solution, blood, and plasma, respectively (the final probe concentration was 0.857 μmol L −1 ), and incubated at 37 ± 0.5 • C to monitor the changes in their fluorescence intensity over time via an IVIS live imaging system. Excitation/emission wavelengths were set at 710/760 nm for ACQ probes and 745/800 nm for DiR.

Preparation of dye-loaded nanocarriers
The fluorescently labeled nanocarriers of different probes were prepared following similar procedures. Take FD-B21 as an example, dye-loaded mPEG 2k −PDLLA 2.5k PMs were prepared by a thin-film dispersion method. [27] In brief, mPEG 2k -PDLLA 2.5k (100 mg) and FD-B21(110.3 μg) were dissolved in DCM (10 mL), and the resultant solution was evaporated under vacuum at 60 • C to form a homogeneous film. The film was hydrated in preheated deionized water (60 • C, 20 mL) under mild stirring (500 rpm, 1 h). The dispersion was filtered through a 0.22 μm filter to obtain the PM solution. PCL NPs were prepared by an O/W emulsion-solvent evaporation method. [58] Briefly, PCL (200 mg) and FD-B21(110.3 μg) were dissolved in DCM (5 mL), which was added dropwise into the PVA solution (1%, 20 mL). Then the mixture was subject to probe sonication (SCIENTZ-IID, Ningbo Scientz Biotechnology Co., Ltd., Ningbo, China) (350 W, 3 min) and high-pressure homogenization (AH100D, ATS Engineering Inc., Toronto, Canada) (800 bar, 3 min) to form homogeneous primary emulsion. The resultant emulsions were subject to continuous stirring to evaporate DCM (150 rpm, 4 h) under ambient temperature and was filtered through a 0.45 μm filter to obtain the PCL PNs. SLNs were prepared utilizing a hot homogenization method. [28] Briefly, ATO 5 (2.5 g) was melted at 75 • C to serve as the oil phase, and FD-B21 DCM solution containing FD-B21 (275.8 μg) was added into the melted lipids. The mixture was heated under 75 • C to remove DCM. The obtained mixture was dispersed into Tween-80 solution (2%, 50 mL) and homogenized first by high-shear homogenizer (XHF-D, Ningbo Scientz Biotechnology Co., Ltd., Ningbo, China) (10,000 rpm, 3 min), followed by a microjet homogenizer (Nano DeBEE, DeBEE, South Easton, USA) (20,000 psi, 5 cycles). The resulted dispersion was cooled down rapidly and filtered through a 0.45 μm filter to obtain FD-B21-encapsulated SLNs. Other nanocarriers encapsulating different fluorophores were prepared following the same procedures with the same molarities of fluorophores. The water-quenched dye solutions of different probes were prepared by diluting their DMSO stock solutions with ultrapure water to the same molarity. All formulations were stored at 4 • C for future use.

Characterization of fluorescently labeled nanocarriers
The particle size and zeta potential of all fluorescent formulations were measured at ambient temperature (Malvern Zetasizer Nano ZSP, Malvern Instruments Ltd., Malvern, UK). All measurements were performed after diluting the formulations to an appropriate concentration with deionized water. The morphology of the nanocarriers were observed via TEM (JEM-1230 Electron microscope, JEOL, Japan). Samples were prepared by dropping diluted suspensions on the copper grids and negatively staining with 1% (w/v) uranyl acetate, after which samples were dried under ambient atmosphere before observation. The EE% was measured by an ultrafiltration method. [28] In brief, the DMSO stock solutions of different fluorescent probes were diluted to various concentrations, ranging from 0.002442 to 0.625 μg mL −1 , and the fluorescence intensity was measured by a Cary Eclipse fluorospectrophotometer (Agilent Technologies, Inc., Santa Clara, USA) at respective excitation/emission wavelengths (Table S2) to establish standard curves. The diluted formulations were centrifuged (6000 rpm, 20 min) (TG16-WS, Cence, Changsha, China) in an Amicon ultra-0.5 (MWCO 100 KDa) tube to obtain the filtrate. To extract and measure unencapsulated probes in the filtrate, add diethyl ether (2 mL), thoroughly vortex mix for 20 min, and wait until separate layers were formed. Take the diethyl ether layer (1 mL), blow dry with nitrogen under 40 • C, redissolve with DMSO (0.5 mL), and measure the fluorescent intensity. The amounts of free probes were then determined by fluorescent intensity standard curves. The EE% of fluorescent formulations was calculated by Equation (2): where W total and W free represented the weight of total and free probes, respectively. The in vitro stability of probe-encapsulated PMs, PNs, and SLNs was investigated by monitoring their fluorescence intensity and particle size at different time points. In brief, fluorescently labeled formulations were diluted by 10 folds with deionized water, blood, and plasma, respectively (the final probe concentration was 0.857 μmol L −1 ). The obtained dispersions were incubated under 37 • C, 100 rpm for 24 h. At each predetermined time point, samples (0.5 mL) were withdrawn to measure the particle size (0.4 mL) and fluorescence intensity (0.1 mL) with DLS and IVIS live imaging system, respectively.

In vivo Imaging
Based on primary experimental results, two novel aACQ probes, namely FD-B21 and FD-C7, were selected for further studies. In vivo live imaging of fluorescently labeled PMs was conducted following previous procedures. [27] All animals were fasted overnight but allowed free access to water before the experiment. The abdominal hair of rats was removed beforehand to reduce hair-derived autofluorescence.
Fluorescently labeled PMs and their respective pre-quenched dispersions were injected intravenously into rats through the tail vein (0.04285 μmol dye equiv kg −1 ). Fluorescence signals were captured at predetermined time intervals via an IVIS live imaging system. To highlight the superiority of novel aACQ probes to P2 and DiR, PMs labeled with these two probes were also inspected under the same procedures. Excitation/emission wavelengths were set at 710/760 nm for ACQ probes and 745/800 nm for DiR, respectively.

Ex vivo Imaging
Following similar experimental procedures, FD-B21 and FD-C7-labeled PMs as well as their quenched dispersions were given to SD rats via i.v. injection (0.04285 μmol dye equiv kg −1 ). At predetermined time points, the rats were anesthetized with chloral hydrate solution (10%) and subject to cardiac perfusion with saline to reduce the influence of bloodborne fluorescence. Then the rats were sacrificed and various organs and tissues were collected, especially the MPS organs as well as fatty tissues, which were more prone to incur fluorescence re-illumination. The fluorescence of different organs and tissues were measured and displayed as ARE.

Cellular uptake
Cellular uptake of various nanocarriers and fluorescence reillumination of the fluorophores were investigated in three cell models, that is, Caco-2, Hela, and Raw264.7 cell lines. Cells were seeded into blank 96-well plates at a density of 3.2 × 10 4 cells cm −2 for Caco-2 and Hela cells and 6.4 × 10 4 cells cm −2 for Raw264.7 cells. After culturing for 2-4 d, homogeneous cell monolayers were established. The fluorescently labeled nanocarriers (PMs, PNs, and SLNs) and prequenched fluorophore dispersions were diluted with HBSS, and 100 μL of the diluted samples were added to cell wells (4.285 μmol L −1 dye). After incubating for 2 h, the dispersions were discarded and the cell lines were washed with PBS three times. The total fluorescence of regions of interest in each well was measured by an IVIS live imaging system. The cell uptake percentage was determined by calculating the fluorescence ratio of the fluorescently-labeled preparation incubation group to that of the administered preparation. The re-illumination efficiency was determined by calculating the fluorescence ratio of the quenched dye suspension incubation group to that of the fluorescently-labeled preparation incubation group.

System generation for MD simulation
All MD simulations were performed using GROMACS 2020 package, [59,60] under GAFF and UFF forcefield and accelerated in high performance computing center with GPU Nvidia Tesla v100 and CPU Intel(R) Xeon(R) Gold 6148 CPU @ 2.40 GHz. The Restrained ElectroStatic Potential of probes derived from Multiwfn, [61] and Gaussian16 with B3LYP(D3)/6-311G** level, [62] and relevant topology parameters were generated using Sobtop. [63] In order to mimic the experimental methods of the probe aggregation process and to demonstrate the conformation of the aggregation system in pure water, acetonitrile and acetonitrile/water binary systems, five different probe systems (FD-B21, FD-C7, P2, FD-B3, and DiR) were constructed. For each probe system in pure water, three times simulations with binary, quaternary, and duodenary states were investigated. And the representative conformations from pure water were extracted and added to pure acetonitrile box and water/acetonitrile mixture (30% acetonitrile fraction). The detailed system information was provided in Table S6 (Supporting Information).

US
In the binary binding system, the US algorithm was performed to capture the aggregation details and to calculate the probe dimerization free energy in water. [64] The reaction coordinate was distributed over 47 sampling windows with 0.05 nm interval for each window (Total 2.3 nm center of mass [COM] distance between two probes) to calculate the PMF during the whole pulling route. The COM of the probe was fixed into the position of reaction coordinate through a harmonic potential with a force of a force constant of 5000 kJ mol −1 nm 2 . And the weighted histogram analysis method was employed to describe the PMF and relevant estimated error with the gmx wham tool in gromacs.

Unbiased MD simulation
The unbiased MD method was employed to describe the motion process of quaternary and duodenary systems. Initially, probes were put in a box (4 × 4 × 4 nm 3 for the quaternary and 6 × 8 × 8 nm 3 for the duodenary) with water model of simple point charge and acetonitrile filled in. In the energy minimization step, conjugate gradient algorithm was used to eliminate the irrational contacts of intramolecules and converge the maximum force <100 kJ mol −1 nm −1 . [65] And each task was performed in the isothermal-isobaric (This method defines simulated system with settled particle number (N) pressure (P), and temperature (T), NPT) ensemble with pressure 1.0 bar and coupled isotropically with 4.5 × 10 −5 compressibility and coupling constant of 12.0 ps under Parrinello-Rahman algorithm, [66] and temperature 298.15 K and coupled with coupling constant of 1.0 ps under v-rescale algorithm. [67] The cutoff distance of short-range interaction was set as 1.0 nm, and the Particle Mesh Ewald algorithm was hired to compute the long-range interaction. [68] Note that 20 ns annealing simulation of four time-and temperature-points (0 ps, 300 ps, 800 ps, 2000 ps; 0 K, 500 K, 500 K, 0 K) to obtain rational probe conformations in water. And 100 ns NPT simulation for each system in water and acetonitrile was executed to generate the trajectory of probe motion. For the solvation of the water/acetonitrile mixture, the representative coordinates of probes in water were set as the initial conformation. And to supply the homogeneous solvent environment to mimic the re-illumination process of probes, we firstly added a fixed number of waters, acetonitrile, and probes in an empty box. Secondly, 10 ns preequilibrium with the position restraint of 50000 kJ mol −1 nm 2 force constant to vector x y z was put on probes to fix the motion of probes before complete solvation. Finally, 100 ns NPT ensemble was set to capture the probe re-illumination. Visual MDs (VMD) and Qtgrace were employed to visualize the MD results.

Independent gradient model based on Hirshfeld partition (IGMH) analysis and molecular planarity investigation
To better understand the physicochemical mechanism of our system, independent gradient model based on IGMH method to represent the weak interaction of inter/intra-molecules was hired to investigate the aggregation mechanism of tested probes. [69] And g inter in IGMH indicated the interaction among interfragment of our system. sign(λ 2 )ρ was used to distinguish the interaction type and strength of inter/intra molecules. In addition, a graphical representation of molecular planarity was also introduced to characterize the property of a single probe in an aggregation state. [70] The MPP and SDP parameter (Å) were both employed to measure the planarity of molecules. And the lower MPP and SDP indicated high planarity and deviation degree to plane from the overall perspective, respectively. Therefore, the 0 value showed the complete plane.

A C K N O W L E D G M E N T S
This work was financially supported by Shanghai Municipal Commission of Science and Technology (grant numbers: 21430760800 and 19XD1400300) and National Natural Science Foundation of China (grant numbers: 81872826, 81872815, 81973247, and 82030107). Molecular modeling was performed at the High-Performance Computing Cluster (HPCC), which is supported by the Information and Communication Technology Office (ICTO) of the University of Macau.

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.

D ATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.