A stable and biocompatible shortwave infrared nanoribbon for dual-channel in vivo imaging

Abstract

The shortwave infrared (SWIR) region is an ideal spectral window for next-generation bioimaging to harness improved penetration and reduced phototoxicity. SWIR spectral activity may also be accessed via supramolecular dye aggregation. Unfortunately, development of dye aggregation remains challenging. We propose a crystal-aided aggregate synthesis (CAASH) approach to introduce a layer of rationality for the development of J-aggregate and the successful development of a water-soluble SWIR JV-aggregate with a bisbenzannulated silicon rhodamine scaffold (ESi5). The resulting SWIR-aggregates exhibit excellent stabilities toward organic solvents, pH, sonication, photobleaching, thiols, and endogenous oxidative species. Notably, the aggregates have a high structure-dependent melting temperature of ca. 330-335 K. In fact, the heating/annealing process can be exploited to reduce aggregation disorder. The aggregates are biocompatible and have broad potential in in vivo fluorescence and photoacoustic imaging and more.

While the shortwave infrared (SWIR) region has potential for bioimaging applications, stable and biocompatible SWIR absorbers/emitters are challenging to access. Here, the authors report the approach of crystal-aided aggregate synthesis for the development of J-aggregates for in vivo bioimaging.

Introduction

Shortwave infrared (SWIR), i.e., 1000 nm and beyond, is an ideal spectral window for deep-tissue and high-contrast in vivo imaging, disease diagnosis, phototherapies, and surgery navigation^1–4. SWIR-active materials of diverse nature have been reported and found applications, e.g., inorganic low-bandgap materials in infrared optoelectronics^5–7 and electrochromic smart windows⁸, hybrid materials in optical filtering and laser shielding⁹, and organic dyes in infrared photography¹⁰, lasing¹¹, and OLED¹². Yet, all these existing SWIR materials have limitations to overcome to improve their potential in the aforementioned cutting-edge biomedical applications. Inorganic or hybrid SWIR materials contain heavy transition metals, the long-term toxicity of many of which are still to be systematically evaluated¹³. Though organic dyes typically exhibit good biocompatibility, SWIR-absorbing organic dyes have their own challenges^13,14. Imparted by their high-lying HOMO and low-lying LUMO, SWIR dyes inevitably suffer from poor stabilities, which inhibit their practical and clinical applications. First, their conjugative backbones are readily destructed, either irreversibly by endogenous/photo-induced oxidative species^15,16, or reversibly by ubiquitous nucleophiles¹⁷. Second, the electron-delocalization along the long conjugative backbone accounting for their SWIR spectral activity is readily disrupted by water via dipole-induced Peierls transition¹⁸. Concomitantly, their long-wavelength absorption band becomes weak, broad, and often significantly blue-shifted¹⁹. Rendering thermodynamically unstable SWIR dyes kinetically persistent in the biological milieu is the key to circumventing their instabilities. An obvious and convenient approach is supramolecular encapsulation in polymer²⁰, serum albumin²¹, or porous materials²². Molecular steric shielding approaches have also been developed for squaraines^23,24, cyanines^25,26, and polybenzannulated xanthenoids^27–29. Generally speaking, this approach becomes exponentially difficult when the absorption maximum marches deeper into the SWIR and their conjugative scaffold concomitantly becomes larger. Alternative robust strategies to develop stable and biocompatible SWIR-absorbing/emitting materials are a timely addition to the field.

The rationalization of aggregation-induced bathochromism led to the formulation of the exciton theory³⁰, the mechanistic elucidation of the natural wonder of photosynthesis³¹, and the later emergence of the nowadays prosperous field of dye aggregation^32–34. The highlight of this strategy for SWIR-active materials is that the resulting longer-wavelength J-aggregate can enjoy the higher stability of the shorter-wavelength monomer. Harnessing the power of J-aggregation for SWIR spectral activity is eminently promising, and yet remains an endeavor necessitating both rationality and serendipity³³. An elusive but useful guideline has been empirically established in designing suitable monomeric dyes for J-aggregation. The monomeric dye ideally exhibits a large transition dipole moment to enhance dipole coupling³⁵. Its electron push–pull headgroups should be flat to allow stacking. A molecular twist or a bulky group should exist at the center of its conjugative backbone to inhibit the undesired co-facial H-type aggregation^32,36. Beyond these electronic and structural nuances, dyes should self-assemble or be organized in a topologically and orientationally precise fashion. Mother Nature has mastered both approaches and crafted astoundingly elaborate functional complexes such as photosynthetic machineries³¹. The nature’s prowess in protein engineering is beyond our scope of synthetic access. Matrices like DNA³⁷, polymer³⁸, mesoporous silica³⁹, hydrogel⁴⁰, and liposome⁴¹, have also been explored to template the aggregate synthesis. Besides, a number of environmental parameters, e.g., counterions, solvent, pH, temperature, and ionic strengths, can also be tinkered with to influence the aggregation outcome^42–46. The implementation of this guideline has led to a number of elegant J-type aggregates based on, e.g., porphyrin(oid)s^47,48, PDI^49,50, BODIPY/aza-BODIPY^51,52, squaraine^53,54, and Cyanine^55,56. Unfortunately, the vast majority of these J-type aggregates typically exhibit an absorption maximum well within 1000 nm^55,57. Aggregates with a SWIR-absorption maximum are few and exhibit a rather broad absorption and a Stokes-shifted emission, suggesting disorder in aggregation due to the high structural flexibility of the corresponding NIR monomeric dyes^43,58–60. Additionally, most SWIR aggregates are synthesized in organic solvents and lack biocompatibility. Also, some J-aggregates, e.g., double-walled nano-tube, are not thermodynamically stable and may exhibit polymorphism^61–63. We recently reported a bisbenzannulated silicon-rhodamine (ESi5) dye²⁸. The ESi5 scaffold has flat electron-donating headgroups, bulky groups at the center, a rigid conjugative scaffold, and a deep-NIR absorption maximum at 867 nm, the combination of which structural and electronic properties promote us to develop high-performance J-aggregate for in vivo bioimaging applications starting from this ESi5 scaffold.

Herein, we report a stable and biocompatible SWIR aggregate via crystal-aided aggregate synthesis (CAASH). It is a two-layer nanoribbon exhibiting a V-type intra-chain packing and a J-type inter-chain packing. The spectral signature of this unusual packing is two red-shifted peaks along with one blue-shifted sharp peak. Notably, the aggregate is stable toward H₂O, pH, sonication, photobleaching, oxidative species, and sulfides. Its aggregation is temperature-dependent and fully reversible. Aggregation disorder could be reduced by repeated heating-annealing. The design and synthesis of robust molecular SWIR aggregate offer a fascinating photophysical playground to formulate/test photophysical principles and high-performance materials for numerous cutting-edge biotechnologies.

Results

Design guideline

Our design of SWIR-absorbing aggregate started with the crystallization of ESi5, a bright cationic fluorophore absorbing at 867 nm. Interestingly, two distinct crystal structures of ESi5 were obtained when different counterions were used. With Cl⁻, ESi5 stacks into a conventional brickwork J-aggregated wall, which further aligns with other walls in an H-fashion to form the 3D crystal packing (Fig. 1a). Following the literature convention, such a packing was referred to as an HJ-aggregate⁶⁴. With PF₆⁻, ESi5 adopts a different zig-zag intra-chain packing mode with a 121.1° angle between two proximal dyes (Fig. 1b). Such an oblique-dimer was pictorially referred to as a V-dimer^65,66. When two such walls are brought together in a co-facial fashion, their inter-chain packing mode is no longer H-type, but J-type with a slipping angle of 51.7°. For this reason, this aggregate is referred to as JV-aggregate. Suppose we can trim off the crystal layer by layer until the last two layers, which stack to each other via π–π interaction. To further render such two-layer stack stability and biocompatibility in aqueous medium, two flexible hydrophilic chains can be installed at the electron push–pull headgroups of the dye. Following this idea, we displaced the two chlorine atoms of the ESi5 with azido groups, and further clicked it to a terminal alkyne linked to various water-soluble groups (Fig. 1c). Such judiciously functionalized amphiphilic ESi5 is expected to spontaneously aggregate in an aqueous medium. However, the specific aggregation outcome, e.g., HJ-type, JV-type, or any other unexpected fashion, is serendipitous, but can be deduced by correlating their absorption spectral features with their packing style in the crystal structure. As for ESi5, a single red (or blue)-shifted peak would suggest a J- (or H-) aggregate, while a V-aggregate would result in a red-shifted peak along with a blue-shifted peak via energy-splitting. The relative intensities of the two peaks depend on the angle in between. Suppose the angle (α) remains at ca. 120°, the oscillation strengths of the J-band (f_J) are expected to be roughly three times that of the H-band (f_H), according to f_J/f_H = tg²(α/2)⁶⁷.

Fig. 1. The molecular packing of ESi5 in single crystals and the design of its JV-aggregates.

The top view and side view of the packing mode of ESi5•Cl⁻ (a) and ESi5•PF₆⁻ (b). c The crystal-aided aggregate synthesis (CAASH) with the ESi5 dye.

Photophysical properties

The fluorochromic core of ESi5 is cationic, and installation of anionic terminal results in amphiphilicity, imparting a driving force to aggregate in an aqueous medium. We prepared eight ESi5 derivatives with different water-soluble groups, i.e., ESi5-NS with a zwitterion, ESi5-N with a quaternary ammonium, ESi5-S with a sulfonate, ESi5-C1(-C5) with a carboxylate-caped linear alkyl chain of 1–5 carbon atoms (Fig. 2a). The surface electrostatic potential of each was mapped to qualitatively assess their potential to aggregate (Fig. 2b). An aliquot of each compound (5 mM, 2 μL) in MeOH was added to H₂O (2 mL) with stirring, and their UV-Vis absorption spectra were then acquired without further sitting. No spectral feature of aggregate formation was observed for ESi5-NS and ESi5-N. The gradual addition of NaCl (200 mM) to the solution containing ESi5-NS resulted in a spectral blue shift from 867 nm of ESi5 dye to 744 nm, suggesting the formation of an undesirable H-type aggregate (Table 1). This H-aggregate of ESi5-NS was non-fluorescent. ESi5-N did not aggregate even in the presence of NaCl (200 mM), likely due to its tri-cationic feature (Fig. 2c). In sharp contrast, the ESi5-S and ESi5-C1(-C5) aggregated instantaneously upon the addition its MeOH stock into H₂O (Supplementary Fig. 15). Their resulting absorption spectra exhibited similar spectral features, i.e., a lower-intensity red-shifted peak at 1098 nm, another red-shifted peak at 1038 nm of much higher intensity, a blue-shifted peak at 696 nm of comparable intensity to the peak at 1098 nm, and non-resolved vibronic bands between 696 nm and 1038 nm (Table 1).

Fig. 2. The spectral monitoring of the aggregation of ESi5 dyes with different water-soluble groups.

a The chemical structures of ESi5-NS, ESi5-N, ESi5-S, and ESi5-C1(-C5). b The electrostatic potential surfaces calculated for ESi5-NS, ESi5-N, ESi5-S and ESi5-C1 at B3LYP/6-31G level. c The UV-Vis absorption (Blue, solid line) and the fluorescence emission (Red, dash line) spectra of the resulting solution upon dilution of the MeOH stock of ESi5-NS, ESi5-N, ESi5-S, and ESi5-C1 in PBS (pH = 7.4, 10 mM). The UV-Vis absorption (Blue, space-filled) and emission spectra (Red, space-filled) of ESi5 in MeOH were plotted for comparison purposes.

Table 1.

The photophysical properties of ESi5 in MeOH and ESi5-NS, ESi5-N, ESi5-S, and ESi5-C1(-C5) in PBS (pH = 7.4, 10 mM)

The photophysical properties
	Monomer	ESi5-NS	ESi5-N	ESi5-S	ESi5-C1	ESi5-C2	ESi5-C3	ESi5-C4	ESi5-C5
λabs (nm)	867	744	866	1038	1037	1024	1022	1033	1019
λabs (nm)	–	–	–	1098	1095	1082	1080	1080	1075
λabs (nm)	–	–	–	696	699	700	708	701	709
λem (nm)	909	–	911	1106	1108	1090	1100	1097	1101
Stokes Shift (nm)	43	–	45	8	13	8	20	17	26
ε (×105 cm−1 M−1)	1.95	–	1.74	0.88	0.79	0.81	0.82	0.86	0.78
ϕ (%)	16	–	1.6	0.06	0.04	0.02	0.02	0.02	0.02
(ε × ϕ) (cm−1 M−1)	15,054	N/A	2784	52.8	31.6	16.2	16.4	17.2	15.6

The ESi5-NS solution contains NaCl (2 M) in PBS.

The unambiguous resolution of the packing mode of an aggregate remains a grand challenge for the field. An advantage of the CAASH approach is that the crystal structure of the monomeric dye provides a reliable source of reference. Such a three-peak pattern cannot possibly be the result of an HJ-type packing in Fig. 1a, but is in agreement with the JV-type packing in Fig. 1b. The two bands at 1038 and 696 nm are presumably due to the intra-chain V-aggregation of ESi5, based on the fact that their oscillation strengths are in a ca. 3:1 ratio (vide ante) (Table 1). The remaining reddest peak at 1098 nm has to be the result of an inter-chain two-dimensional J-aggregation. The peak width is an indicator of the dipolar coupling strength and calculated via spectral deconvolution. The FWHM of all three peaks are small at 574.0 cm⁻¹ for 696 nm, 452.6 cm⁻¹ for 1038 nm, and 233.3 cm⁻¹ for 1098 nm, respectively, suggesting a strong dipolar coupling between proximal ESi5 dyes. Upon photoexcitation at 696 nm, 1038 nm, or 1098 nm of a solution of ESi5-S agg or ESi5-C1(-C5) agg, the same resonant fluorescence emission band at ca. 1106 nm was observed with a very sharp FWHM of 481.9 cm⁻¹ (Table 1). The fluorescence quantum yield was 0.06% for ESi5-S, significantly reduced compared to 1.7% of the monomer. Through a transient absorption study, a fluorescence lifetime of 58 ps was calculated (Supplementary Fig. 31). The change of the terminal anion from sulfonate to carboxylate slightly led to a reduced degree of spectral resolution between the two red-shifted peaks, and enhanced minor vibronic peaks between 1038 and 696 nm. Increasing the length of the alkyl chain from the -CH₂- of ESi5-C1 to the -(CH₂)₅- of ESi5-C5 resulted in a slight broadening of the absorption bands (Supplementary Figs. 16–21).

Morphology of ESi5 aggregates

The morphology of the ESi5-S agg was different with respect to the sitting time of the solution. Without sitting time, the aggregate adopted the form of small particulate with a diameter of 7.7 ± 0.6 nm (Supplementary Fig. 29A). In 6 h, the particulate could still be seen while the aggregate predominantly existed in form of short nanoribbon with an average width of 39.6 ± 1.6 nm, and an average length of 0.44 μm (Supplementary Fig. 29B). In 24 h, the short nanoribbon further grew into a network. Some ribbons exhibited an increased width of up to 116.2 ± 2.2 nm (Fig. 3a, Supplementary Fig. 29C). Some ribbons folded back and forth suggesting a soft and flexible nature. The nanoribbon morphology of ESi5-S agg in solution was clearly captured by the cryo-electron microscopy (cryo-EM) image (Fig. 3b). The thickness of the ribbon was determined to be ca. 3.4 nm by atomic force microscopy (AFM), in agreement with the hypothesized double-layer aggregate model (Fig. 3c, d). The transmission electron microscopy (TEM) images of ESi5-C1/C2/C3 agg revealed that their morphologies were similar to that of ESi5-S agg (Supplementary Figs. 24–28). As for ESi5-C4/C5 agg, the ribbon was significantly shorter and wider. The existence of particulate in the TEM images of ESi5-C4/C5 agg suggested that their aggregation displayed inhomogeneity, likely due to the enhanced structural flexibility of the monomer dye (Supplementary Figs. 27 and 28). The thickness of aggregates increased with respect to the chain length from 3.4 nm of ESi5-C1 agg to 5.8 nm of ESi5-C5 agg (Supplementary Figs. 24–28). Despite the morphological change, it was interesting to note that the absorption spectra of the bulk aggregate solution remained unchanged. This indicated that the molecular packing of the monomer in the particulate at 0 h, short nanoribbon at 6 h, and long nanoribbon network at 24 h remained unchanged.

Fig. 3. Morphology of ESi5-S agg.

The TEM image (a) and the Cryo-EM image (b) of ESi5-S agg (200 μM) in H₂O. c The AFM image of ESi5-S agg with cross-section analyses (d) along red dashed lines. The dark spots are crystalline ice. Each experiment was repeated independently with similar results (n = 3).

Governing parameters on aggregation

With ESi5-S agg as an example, the stabilities of the JV-aggregates of ESi5 dyes toward organic solvent, dye concentration, sonication, pH, temperature, light irradiation, and oxidants were thoroughly evaluated. The UV-Vis absorption spectra of ESi5-S agg (5 μM) in a mixture of neutral PBS buffer and varying percentages of MeOH (0%–32%, v/v) were acquired (Fig. 4a, b). The absorption spectrum remained essentially unchanged when the MeOH percentage was below 6%. When the MeOH percentage went above 6%, the peaks of the JV-aggregate started to decrease, and the peak of the monomer at 867 nm started to increase accordingly. When the MeOH percentage was 22%, a 50% aggregation was observed. Raising the MeOH beyond 32% eventually totally inhibited the formation of JV-aggregate. ESi5-S exhibited a high tendency to aggregate. In the literature, a high dye concentration of 10⁻⁴–10⁻³ M⁻¹ was routinely used to promote aggregation. Yet, ESi5-S at even 1 μM is completely aggregated. Further increasing the dye concentration gave a linear increase in the absorbance at 696, 1038, and 1098 nm with respect to the dye concentration, tested up to 30 μM (Fig. 4c, d). This experiment could help to determine the ideal concentration for bioimaging.

Fig. 4. The stabilities of JV-aggregate toward different stimuli.

a Solvent-dependent changes of the absorption spectra of ESi5-S (5 μM) in PBS (pH = 7.4, 10 mM) with different MeOH fractions (f_MeOH) from 0% to 32% at room temperature. b The plot of molar absorptivities of ESi5-S monomer (867 nm) and JV-aggregate (1038 nm) with respect to the f_MeOH. c The concentration-dependent absorption spectra of ESi5-S in PBS (pH = 7.4, 10 mM, containing <2% MeOH) at room temperature from 1 to 30 μM. d The plot of absorbance at 696, 1038, and 1098 nm, respectively, as a function of the concentration of ESi5-S. e The temperature-dependent absorption spectra of ESi5-S agg (5 μM) in PBS (PH = 7.4, 10 mM) with 5% MeOH from 20 to 90 °C. f The plot of aggregation percentage (α_agg) with respect to the temperature. g The absorption spectra of ESi5-S (10 μM) in PBS (pH = 7.4, 10 mM) with 10% MeOH upon repeated heating/annealing between 20 and 75 °C. h The reversible absorbance changes of the monomer at 867 nm and the ESi5-S agg at 1038 nm with respect to heating and annealing cycles.

The temperature dependence of the aggregation was studied with a solution of ESi5-S (5 μM) in PBS with 5% MeOH. When the temperature gradually raised from room temperature to 90 °C, the aggregation degree (α_agg) decreased from near unity at 25 °C, to ca. 2.8% at 90 °C (Fig. 4e). Its melting temperature (Tm, when α_agg = 50%), was 74 °C (Fig. 4f, Table 2). When the solution temperature was further lowered back to 25 °C from 90 °C, the absorption of the aggregate was essentially restored. Interestingly, the absorption band at 1038 nm and the 1098 nm became narrower, and their separation became slightly wider. Likely, the heating/annealing process eliminated some of the aggregation defects and further enhanced dipolar coupling. We then repeated the heating/annealing processes over 20 times, and the system showed excellent reversibility (Fig. 4g, h). This experiment not only highlighted the stability of the aggregate, but also was a manifestation of the high stability of the monomeric ESi5 dyes at demanding conditions, i.e., 75 °C in PBS for hours. The aggregate equilibrium constant (K_e) and Gibbs free energy (ΔG) of ESi5-S aggregate were calculated to be 8.2$\times$ 10⁵ M⁻¹, and −36.6 kJ•M⁻¹, respectively. Changing the sulfonate of ESi5-S into the carboxylate of ESi5-C1 led to a slight decrease of T_m to 69 °C, K_e to 7.1$\times$ 10⁵ M⁻¹, and ΔG to −36.2 kJ•M⁻¹ (Table 2). Further elongation of the alkyl chain led to a notable further decrease of the aggregate stability (Table 2, Supplementary Fig. 36). The heating-annealing results with ESi5-S agg promoted us to wonder if the same method could be applied to the ESi5-C1(-C5) since their aggregation disorder seemed to be more pronounced than ESi5-S. The solutions of ESi5-C1(-C5) aggregates in PBS with 5% MeOH were heated to 90 °C and then cooled to room temperature (Supplementary Fig. 15). The effect was very obvious. Three major bands at ca. 1098 nm, ca. 1038 nm, and ca. 696 nm became sharper in shape. The two spectrally unresolved red bands were now totally resolved. The two red bands slightly red-shifted by ca. 8 nm, and the blue band slightly blue-shifted by ca. 12 nm. These were indications of the reduction of aggregate disorder. ESi5-S agg was also stable toward sonication and pH change (Supplementary Figs. 34 and 35).

Table 2.

The thermodynamic parameters of ESi5-S, and ESi5-C1(-C5) aggregation

The thermodynamic parameters
	Tm (K)	Ke (105 M−1)	DPN	ΔH (kJ mol−1)	ΔS (kJ mol−1)	ΔG (kJ mol−1)
ESi5-S	349	8.2	2.59	−74.8	118.3	−36.6
ESi5-C1	345	7.1	2.45	−79.9	135.3	−36.2
ESi5-C2	341	6.3	2.31	−86.7	157.6	−35.8
ESi5-C3	335	4.1	2.02	−96.3	190.6	−34.7
ESi5-C4	331	2.9	1.80	−100.3	206.0	−33.8
ESi5-C5	328	2.2	1.66	−111.9	243.9	−33.1

SWIR-absorbing dyes are usually very susceptible to nucleophilic attack of ONOO⁻. Yet, the stability of the JV-aggregate of ESi5-S and ESi5-C1(-C5) was all very high. The addition of up to ONOO⁻ (50 μM, ca. 100-fold of the maximum physiological concentration) into its solution resulted in only 10.5% decrease of the aggregate absorbance. Addition of H₂O₂ (5 mM), GSH (60 mM), cysteine (5 mM), and HS⁻ (tested to 1 mM) did not induce absorbance decrease (Supplementary Fig. 33). Also, the aggregate and monomer exhibited high photostability and irradiation of the solution at 808 nm or 1064 nm for 30 min resulted in an absorbance decrease of only ca. 4.8% and 11.8%, respectively, while the absorbances of ICG at 808 nm irradiation and IR1064 at 1064 nm irradiation decreased by 98.7% and 85.9%, respectively (Supplementary Fig. 32).

In vivo fluorescence imaging

ESi5-S agg exhibited good biocompatibility. ESi5-S agg was essentially non-cytotoxic when its concentration was smaller than 30 μM, since the cells viability was higher than 85% after a 24 h incubation with HeLa cells (Supplementary Fig. 38). Incubating ESi5-S agg (100 μM) with red blood cells for 24 h did not induce noticeable hemolysis (<5%) (Supplementary Fig. 39). An acute-toxicity study was carried out to verify that up to 2.5 mg/kg did not exhibit toxicity based on monitoring of the weight and post-mortal H&E histological analysis (Supplementary Figs. 45 and 46). We further tested the feasibility of ESi5-S agg for in vivo fluorescence imaging harnessing its intense and narrowband SWIR absorption and high stability in an aqueous medium. Upon i.v. injection, an aggregate is diluted and re-establishes the aggregation-disaggregation equilibrium in blood vessels. Therefore, a dual-channel imaging experiment may be carried out to monitor the fluorescence of the aggregate (λ_ex = 1064 nm) and that of the dissociated monomer (λ_ex = 808 nm). Since the aggregate and the monomer are drastically different in terms of their size, charge, and morphology, they are expected to exhibit distinct metabolic profiles and in vivo distribution. We further verified that the aggregate can be selectively excited by a 1064 nm laser and the monomer by an 808 nm laser, respectively, with minimal cross-talk (Fig. 5a, b). An aliquot of ESi5-S agg solution (500 μM in PBS, 100 μL) was injected into a BALB/c mouse through the tail vein to render a final vasculature concentration of 2.5 mg/kg, equivalent to a biocompatible blood concentration of ca. 30 μM. The first set of images was acquired 10 min post injection (Fig. 5e). The fluorescence signal of the aggregate was found from such organs as the liver, spleen, and bones, whereas the signal of the monomer localized in the liver (Fig. 5e–g). The blood vessels were not observed in either channel. This suggests that the aggregate exhibited a notable stability against dissociation even in blood and was distributed through circulation. Its affinity toward bone was likely due to its polyanionic nature. This bone affinity was also in agreement with many other polyanionic materials, due to its affinity to hydroxyapatite⁶⁸ (Supplementary Fig. 47). The ratio of the fluorescence intensities between the vertebrae and soft tissue reached a peak of 3.9 in 180 min (Supplementary Fig. 41). Besides, the analysis crossing the vertebrae and shin also showed good differentiation between bone and others, with FWHM of 0.89 mm and 1.80 mm (Fig. 5k, l).

Fig. 5. The in vivo dual-channel fluorescence (Fl.) imaging of BALB/c mouse using ESi5-S agg.

a The normalized absorption spectra of ESi5-S monomer and ESi5-S agg and the compatible common laser lines. b The epi-fluorescence images of the ESi5-S Monomer (Left, 500 μM in water with 1% tween 80) and ESi5-S agg (500 μM in PBS) solutions in EP tubes excited by 808 (60 mW•cm⁻²) and 1064 nm (80 mW•cm⁻²), respectively. c The timeline of the in vivo fluorescence mouse imaging experiment. 808 nm channel: λ_ex = 808 nm (60 mW•cm⁻²), long pass filter: 1200 nm, exposure time: 100 ms. 1064 nm channel: λ_ex = 1064 nm (80 mW•cm⁻²), long pass filter: 1200 + 1350 nm, exposure time: 500 ms. d A schematic diagram of major anatomical structures of mouse related to this experiment. Created in BioRender. Yang, Y. (2024) https://BioRender.com/q64i385. The two-channel fluorescence images of BALB/c mice in the supine position (e) 10 min and (f) 180 min post i.v. injection, and in the prone position (g) at 180 min post i.v. injection. The changes of the liver region fluorescence intensity (h) in the 808 nm channel and (i) in the 1064 nm channel. The line intensities of (j) the ROI 1 for the intestine in the 808 nm channel, (k) the ROI 2 for the sternum in the 1064 nm channel, (l) the ROI 3 for the tibia in the 1064 nm channel, and (m) the ROI 4 for the lumbar vertebra in the 1064 nm channel. Scale bar: 10 mm.

During a course of 180 min, the fluorescence images of these two channels were acquired intermittently. The distribution pattern was essentially unchanged except for a gradual decrease in signal intensity. The fluorescence images of the liver region of the two channels provided a good estimate of the metabolic kinetics of the monomer and the aggregate to be on the order of a few hours (Fig. 5h, i). This suggested a gradual dissociation and metabolism of the aggregate. In 180 min, the monomer was metabolized from the liver to the intestine (Fig. 5f). Line intensity profile could be used to calculate the size of various anatomical structures, e.g., sternum to be ca. 0.89 mm (Fig. 5k) and tibia to be ca. 1.80 mm (Fig. 5l). Likewise, imaging of the prone state (Fig. 5g) also displayed a high sensitivity and SBR in the normal bones, including the lumbar vertebra, ilium, caudal vertebra, and femur. Specifically, line-intensity analysis (red line) along the lumbar vertebra was extracted and afforded the dimensions of all the vertebrae in the range of 1.12–1.63 mm (Fig. 5m). The lack of the monomer signal in other organs except the liver suggested that the dissociation of aggregate occurred predominantly in the liver. Overall, ESi5-S agg is a specifically bone-binding nano-assemblies for visualizing the whole skeletal system in vivo, and its enterohepatic metabolism processes are revealed by dual-channel imaging.

In vivo photoacoustic imaging

An additional proof-of-concept photoacoustic imaging application was carried out to further showcase the potential of this series of aggregates with ESi5-S agg as an example for different modalities of bioimaging. PA signal is mainly related to the photothermal conversion and subsequent thermal-to-acoustic transformation of a light-absorbing material. The temperature increase profiles of the ESi5-S agg solution in PBS correlated with both the aggregate concentrations and laser power density (Fig. 6a, b). Fitting of the photothermal conversion profile yielded a photothermal conversion efficiency of 58.1% (Fig. 6c). Further, the thermal stability of the ESi5-S agg was evaluated by repeated heating/cooling cycles in the range of 25–46.6 °C (Fig. 6d). After five cycles, no sign of fatigue was observed, while IR-1061 did not survive two around of heating/cooling cycles under the same conditions. The photoacoustic effect of ESi5-S agg was further tested at various concentrations (0–50 μM) (Fig. 6e). A dose-dependence was observed between the photoacoustic intensity and the ESi5-S agg concentration. Notably, the photoacoustic signal intensity was 3.5 times higher than that of IR-1061 (50 μM in PBS).

Fig. 6. The in vivo photoacoustic (PA.) imaging of BALB/c mouse using ESi5-S agg.

The concentration (a) and laser intensity (b) dependent photothermal property of ESi5-S agg in PBS. c The calculation of the photothermal conversion efficiency of ESi5-S agg (30 μM) under continuous 1064 nm laser (1000 mW•cm⁻²) irradiation. d Photothermal stability of ESi5-S agg (30 μM) in PBS and IR-1061 (30 μM) in PBS under 1064 nm laser (1000 mW•cm⁻²) irradiation. e The photoacoustic intensity of the ESi5-S agg and IR-1061 in PBS excited by 1064 nm laser (n = 3, Data are presented as mean values ± SD). Photoacoustic imaging of BALB/c mice acquired (f) without ESi5-S agg and (g) 180 min post injection of ESi5-S agg (2.5 mg/kg in PBS). h, i Line intensities of the highlighted ROIs of the images of vertebra. Scale bar: 1 mm.

The feasibility of ESi5-S agg for in vivo photoacoustic imaging was carried out with BALB/c mouse. First, we acquired the background photoacoustic imaging with mouse without injecting ESi5-S agg. The presence of weak signals from the vasculature-rich area/organs was due to a weak absorption of hemoglobin at this SWIR region (Fig. 6f). Then, a solution of ESi5-S agg (2.5 mg/kg) in PBS was injected into shaved BALB/c mouse through tail vein. Due to its intense absorption at 1064 nm, a good contrast could be obtained over the signal from the hemoglobin. In 180 min, the photoacoustic experiment was carried out with a 1064 nm excitation and a three-dimensional photoacoustic tomography was reconstructed (Fig. 6g). The photoacoustic signal was concentrated in the liver and the bone (including sternum, costae, and lumbar), in agreement with the fluorescence imaging results. For lumbar, specifically, vertical and horizontal analysis (ROI 1 and ROI 2) along the lumbar vertebra showed eight and three peaks of photoacoustic intensity respectively, and the FWHMs were high enough to discriminate different vertebrae including posterior articular protrusion (Fig. 6h, i). From the above experimental results, it can be concluded that the ESi5-S exhibits an excellent PA imaging performance.

Discussion

In conclusion, we proposed an approach, i.e., CAASH, to make J-aggregate development more of a rational and less of a serendipitous endeavor. We chose ESi5 as the monomeric dye to develop SWIR aggregate because it fulfills the basic set of criteria: (1) deep-NIR absorption maximum at 867 nm due to a large transition dipole moment; (2) flat electron push–pull headgroups; (3) bulky groups at the center of the conjugative backbone. Implementation of the CAASH approach to the scaffold of ESi5 was a two-step process. First, the tendency of the ESi5 to form aggregate was evaluated by growing the crystals of the bare bone ESi5 fluorophore. Interestingly, we successfully obtained two distinct crystals of ESi5 when different counteranions were used. With the success of the first step, we then installed two hydrophilic chains to the electron push–pull headgroups of ESi5 scaffold to impart additional driving force to aggregate in an aqueous medium. Upon addition of the amphiphilic ESi5 dyes, e.g., ESi5-S, or ESi5-C1(-C5), into H₂O, spectral features suggesting the formation of an aggregate were immediately observed. The intense absorption at 867 nm of ESi5 monomer disappeared and three new peaks at 696, 1038, and 1098 nm appeared. Taking into account both the basic exciton theory and the crystal packing styles of ESi5 crystal, it was deduced that the 696 and 1038 nm bands were presumably from intra-chain V-aggregation, and the 1098 nm band from inter-chain J-aggregation. Upon photoexcitation, ESi5 aggregates exhibited by a sharp resonant fluorescence emission band and a greatly reduced fluorescence lifetime (τ = 58 ps). The JV-aggregate from ESi5 dyes exhibited unexpectedly high stability toward organic solvent, temperature, photobleaching, and reactive chemical species. Notably, we found that heating/annealing not only did not lead to destruction of the JV-aggregate, but also helped to reduce aggregation disorder and promote aggregation coupling strength. ESi5-S agg exhibited high biocompatibility, e.g., high aqueous stability, aqueous solubility, low toxicity, and low hemolysis capability. Together, these properties rendered the ESi5 aggregates broad potential for cutting-edge biotechnologies exploiting the SWIR region. In this work, we intravenously injected an aliquot of ESi5-S agg to BALB/c mouse. The ESi5-S agg exhibited good stability in the bloodstream and can be used for dual-channel fluorescence bioimaging. The 1064 nm laser line can be used to excite ESi5-S agg, which exhibits high affinities toward bones, along with some uptake in the liver and spleen. The 808 nm laser line can be used to excite the monomeric ESi5-S dye, which was found only to exist in the liver and metabolized to the intestine in the course of a few hours. ESi5-S agg also enabled photoacoustic tomography of mouse bones, including sternum, costae, and vertebra. Multiplexed imaging with minimal cross-talks between different channels is a challenge for the near-infrared region due to the broad absorption bands of the NIR dyes. Sletten et al. noted that the aggregates with sharp absorption bands are viable candidates for multiplexing in this spectral region³⁹. Therefore, developing aggregates that can pair with ESi5-S agg for multiplexed imaging is a viable direction for future research.

Methods

Animals

The animal experiment protocol was approved by the Laboratory Animal Welfare and Ethics Committee of Shanghai University of Traditional Chinese Medicine (PZSHUTCM220613018, June 13, 2022). The mice which purchased from Shanghai JieSiJie Laboratory. Six to eight-week-old female BALB/c mice with a body weight of ~20 g were chosen for research. The mice were housed in an environment with a temperature of ~25 °C and a 12 h light/dark cycle, suitable humidity (typically 40%–60%). Sex was not noted in the mouse imaging studies because the results are applicable to both male and female mice.

General materials

Chemicals were purchased from major venders based in China and used without further purification. Analytical grade solvents, including petroleum ether, ethyl acetate, dichloromethane, methanol, dimethyl sulfoxide (DMSO), and N,N-dimethylformamide (DMF), were obtained from Titan Scientific. IR-1061 (CAS: 155614-01-0) and ICG (CAS: 3599-32-4) was purchased from Sigma Aldrich. Anhydrous solvents for reactions (e.g. tetrahydrofuran and DMF) were procured from Adamas with activated 4A molecular sieves. Silica gel (200–300 or 300–400 mesh) for flash chromatography was acquired from Haiyang Chem, China. The HeLa cells were purchased from the Cell Bank of Type Culture Collection of the Chinese Academy of Sciences. Sheep red blood cells were purchased from SenBeiJia Biological Technology Co., Ltd.

Instruments

The ¹H-NMR and ¹³C-NMR spectra were acquired over Bruker Avance 400 or Avance 600 spectrometers. The ESI-HRMS spectra were obtained using Micromass GCT spectrometer at the Analysis and Testing Center of East China University of Science and Technology.

The UV-Vis absorption spectra were acquired using a SHIMADZU UV-2600 UV-VIS spectrophotometer and the fluorescence spectra were acquired with a PTI-QM4 steady-state fluorimeter with 1 cm cuvette and an InGaAs photodetector. The femtosecond transient absorption spectra were acquired with Helios fire, ultrafast systems. The single-crystal X-ray diffraction data were acquired with a Rigaku Synergy-R X-ray single-crystal diffractometer. The in vivo fluorescence imaging was carried out with MARS (Artemis Intelligent Imaging) equipped with an InGaAs camera (NIRvana, Teledyne Princeton Instruments). The photoacoustic imaging of solutions and in vivo photoacoustic imaging of mice were performed with the LOIS 3D imaging system (TomoWave Laboratories, USA).

The ¹H-NMR and ¹³C-NMR spectra were processed using MestReNova. The spectral data were processed using Origin 2024b. The microscopy images were analyzed with Image J 1.48. The X-ray crystal diffraction data (cif file) were analyzed using Mercury 3.8, CCDC.

Electrostatic potential (ESP) calculations

All theoretical calculations were accomplished by Gaussian 16. The structural optimization and ESP distribution were derived from density-functional theory calculations at the B3LYP/6-31G (d) level. And the surface density minima or maxima analysis was performed using Multiwfn (Comput. Chem. 2012, 33, 580–592).

Preparation of ESi5 aggregates

Take ESi5-S for example, a stock solution of ESi5-S (5.0 mM) was prepared in organic solvents, e.g., MeOH, and an aliquot was subsequently added to a PBS (pH = 7.4, 10 mM) solution with stirring. The aggregation was completed essentially instantaneously.

Fluorescence quantum yield (ϕ) determination

The fluorescence quantum yield of ESi5 dyes in organic solvent (i.e. DMSO, MeOH, MeCN, DMF) and ESi5-N in PBS were referenced to ECXb (ϕ = 13.3% in CH₂Cl₂), with excitation at 880 nm. The fluorescence quantum yield of ESi5 JV-aggregates in PBS (λ_em > 1000 nm) were referenced to IR26 (ϕ = 0.5% in 1,2-dichloroethane), with excitation at 980 nm. Specifically, the fluorescence quantum yield of the sample was calculated according to literature methods: Due to the fact that there still lacks a stable and bright dye absorbing in SWIR region with robustly calibrated fluorescence quantum yield, this quantum yield should be considered as an estimate.

Spectral deconvolution

Spectra deconvolution was conducted following literature methods (J. Am. Chem. Soc. 2022, 144, 14351–14362).

Electron microscopy characterization

The stock solution containing ESi5-S or ESi5-C1(-C5) (5.0 mM, 40 μL) was prepared in methanol (MeOH) and then added to double-distilled water (960 μL) to induce aggregation. The resulting aggregate solution (0.2 mM) was incubated in the dark for 24 h, and the percentage of MeOH in the solution was diluted to 4%.

TEM samples were prepared by depositing 5 μL of aggregate solution on a 400-mesh formvar-supported copper grids coated with ultra-thin carbon film. After a 10-min deposition period, excess solution was removed using filter paper. The aggregate samples were imaged using JEM-1400 and JEM-2100 at an acceleration voltage of 100 kV.

Three microliters of solution containing aggregate was incubated on a mica sheet for 5 min and then was blotted with the filter paper until only a thin solution film remained. The AFM images were captured using a Bruker Dimension Fast Scan, using ScanAsyst mode under the ambient conditions.

For cryo-EM, the solution of aggregate (5 μL) was applied to glow-discharged holey carbon copper grids before plunge-frozen in liquid ethane after blotting with filter paper using Vitrobot Mark IV(FEI). Cryo-EM micrographs were acquired on a Tecnai T12 TEM (FEI), operated at an acceleration voltage of 120 kV.

Chemo-stability toward thiols and reactive oxygen species

An aliquot of GSH (0–10 mM) in H₂O, Cys (0–100 mM) in H₂O, NaSH (0–1 mM) in H₂O, H₂O₂ (0–10 mM) in H₂O, NaClO (0–50 μM) in H₂O (pH = 10), ONOO⁻ (0–50 μM) in H₂O (pH = 10) were added into a solution of ESi5-S agg. The solution was stirred for 5 min before the absorption spectra were acquired.

Photostability tests

The ESi5-S monomer and ICG in DMSO with same absorbance (A = 0.2) in 808 nm were irradiated by an 808 nm laser (500 mW•cm⁻²) with vigorous stirring for 30 min. The ESi5-S agg in PBS and IR-1061 in DMSO with same absorbance (A = 0.2) in 1064 nm were irradiated by a 1064 nm laser (520 mW•cm⁻²) with vigorous stirring for 30 min. The absorption of solutions was measured every 1 or 2 min.

The thermodynamic parameters for aggregation

The degree of aggregation and thermodynamic parameters for aggregation were calculated following literature methods (Angew. Chem. Int. Ed. 2012, 51, 5615–5619).

In vitro hemolysis assay

Sheep red blood cells (2%, 0.5 mL) were incubated with Triton X-100 (2%, +), PBS buffer (pH = 7.4, 10 mM, −), and incremental concentration gradient of ESi5-S agg (20, 40, 60, 80, 100 μM) for 4 h on an oscillator at 37 °C, respectively. The suspension supernatant was concentrated by centrifugation (10,000 rpm, 5 min) and placed in a 96-well plate. The absorbance at 540 nm of each well was measured by enzyme marker, and the hemolytic percentage value was calculated according to Equation:

$$HP\left(\%\right)=\frac{{OD}_{Sample}-{OD}_{Saline}}{{OD}_{Triton}-{OD}_{Saline}}\times 100\%$$

Where OD_sample, OD_saline, and OD_Triton stand for the absorbance of sheep red blood cells incubated with different ESi5-S agg, PBS, and triton X-100, respectively.

In vivo dual-color fluorescence imaging by MARS system

Six to eight-week-old female BALB/c mice with a body weight of ~20 g were chosen for in vivo dual-color color fluorescence imaging. An aliquot of ESi5-S agg solution (500 μM in PBS, containing 4% DMSO, 100 μL) was injected into a BALB/c mouse through the tail vein to render a final vasculature concentration of 2.5 mg/kg, equivalent to a biocompatible blood concentration of ca. 30 μM. The whole-body images were acquired 10, 30, 60, 120, 180 min post injection. The images with excitation at 808 and 1064 nm were acquired sequentially. When the excitation of 808 nm (exposure time: 100 ms, 60 mW•cm⁻²) was used, a long pass filter with a cutoff of 1200 nm was used for emission collection. When the excitation of 1064 nm (exposure time: 500 ms, 80 mW•cm⁻²) was used, two long pass filters at 1350 and 1200 nm were used for emission collection. All mice were shaved before fluorescence imaging.

In vivo toxicity study

Six-week BALB/c female mouse (n = 6) was used as an animal model and randomly divided into three groups. An aliquot of PBS (10 mM, pH = 7.4, 0.1 mL) and ESi5-S agg (2.5 mg/kg, 5.0 mg/kg) solution was injected through the tail vein, respectively. The mice were raised for a period of 14 days, during which period the body weight and vitality of the mouse were recorded.

For H&E histology, the mice were sacrificed 24 h post i.v. injection of PBS (10 mM, pH = 7.4, 0.1 mL) and ESi5-S agg (2.47 mg/kg in PBS) and the major organs are harvested for H&E histology analysis for acute-toxicity features.

Photothermal conversion performance

The increase in temperature of ESi5-S agg solution (2 mL, in PBS) with different concentrations (0, 5, 10, 20, 30, 40, and 50 μM) under 1064 nm laser (1000 mW•cm⁻²) and ESi5-S agg solution (30 μM in 2 mL PBS) under different power densities (200, 400, 600, 800, and 1000 mW•cm⁻²) were measured using a temperature detector. The photothermal conversion efficiency (η) was calculated according to Equation:

$$\mathrm{\eta }=\frac{hA\left(T_{\max }-T_{surr}\right)-Q_{dis}}{I\left(1-{10}^{-A1064}\right)}$$

Where $h$ is the heat transfer coefficient and $A$ stands for the surface area of the quartz sample cell. $T_{\max }$ and $T_{surr}$ are the maximum steady-state temperature under laser and the surrounding environment temperature, respectively. $Q_{dis}$ is the heat dissipation from the light absorbed by the solvent and the quartz sample cell. $I$ was the incident laser power, and A1064 represent the absorbance intensity of the solution at 1064 nm.

In vivo photoacoustic imaging

The photoacoustic signals of ESi5-S agg and IR-1061 solutions with different concentrations (0, 5, 10, 20, 30, 40, and 50 μM) were carried out with LOIS 3D (TomoWave Laboratories, USA) under 1064 nm laser.

For in vivo photoacoustic imaging, 6–8-week-old female BALB/c mice with a body weight of ~20 g (n = 3) were injected with ESi5-S agg (2.5 mg/kg, in PBS containing 4% DMSO). The PA imaging was performed by LOSI 3D imaging system before injected and post-injected of 120 min under 1064 nm laser. All mice were shaved before photoacoustic imaging.

Calcium-binding experiments

Hydroxyapatite (HA), calcium phosphate (CP), calcium pyrophosphate salts (CPP), calcium carbonate (CC), calcium oxalate (CO), and manganese dioxide (MnO₂) (10 mg/mL) were incubated separately with ESi5-S agg (10 μM) in PBS (pH = 7.4, 10 mM). The mixtures were vortexed for 60 min and then washed three times with PBS before centrifugated to remove unbound JV-aggregate. The precipitate was diluted in 200 μL PBS and the fluorescence emission intensity was tested.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.Y. was responsible for synthesis, spectral studies, draft preparation, and editing. R.W., X.Z., and R.C. contributed to the synthesis. J.Z. and J.C. carried out transient absorption studies. Y.S. and J.R. contributed to TEM. W.X. and C.L. contributed to cryo-EM studies. D.Q. contributed to the vary-temperature experiment. Y.W. contributed to AFM studies. X.L. Y.D., and G.G. contributed to in vivo imaging. X.L. and X.Q. were responsible for advising and editing. Y.Y was responsible for advising, editing, and conceptualization.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Supplementary information is available. Requests for materials should be addressed to the corresponding author. The source data generated in this study have been deposited in the FigShare database under the accession code 10.6084/m9.figshare.27959742.v1. Source data are provided alongside the manuscript. Crystallographic data for the structures reported in this article have been deposited at the Cambridge Crystallographic Data Centre, under deposition numbers CCDC 2120399 (ESi5•Cl⁻), and 2364282 (ESi5•PF₆⁻). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures/. Data are available from the authors on request. Source data are provided with this paper.

Competing interests

A Chinese patent application (2024115961715) with Y.Y., X.Q., X.L., and C.Y. listed as inventors is filed for the structures, preparation, and potential applications of the aggregates described in this manuscript. Y.Y., X.Q., X.L., and C.Y. declare no other competing interests. Other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-55445-x.