Targeted Near‐Infrared Photoacoustic Probes for Dual‐Channel Cartilage and Bone Imaging

ABSTRACT

Simultaneous imaging of cartilage and bone with molecular specificity remains a significant unmet need in orthopedic research and intraoperative guidance. Here, we report the development of two tissue‐targeted near‐infrared photoacoustic probes (Cart‐670 for cartilage and Osteo‐750 for bone) that enable dual‐channel visualization of these adjacent tissues within a single imaging session. The probes were designed through a modular strategy: A systematic screen of NIR‐absorbing non‐fluorescent dyes identified QSY21 as the optimal photoacoustic scaffold (limit of detection: 0.5 µM in PBS; signal‐to‐noise ratio ≥ 3), which was then functionalized with a cationic cartilage‐targeting motif to yield Cart‐670. In parallel, Osteo‐750 was synthesized by conjugating a bisphosphonate moiety to Alexa Fluor 750 for bone targeting. Both probes exhibit strong NIR absorption, low detection limits (Cart‐670: 1 µM; Osteo‐750: 2–3 µM in PBS), and selective ex vivo tissue accumulation: Cart‐670 shows preferential retention in cartilage over bone, while Osteo‐750 produces ∼200× higher photoacoustic signal on bone relative to cartilage. Spectral unmixing at 680 and 750 nm enables artifact‐free two‐color imaging without cross talk, demonstrating the feasibility of multiplexed photoacoustic visualization of the cartilage‐bone interface.

Two near‐infrared photoacoustic probes enable simultaneous visualization of cartilage and bone within a single imaging session. Cart‐670 exploits electrostatic affinity for glycosaminoglycans to label cartilage, while Osteo‐750 uses bisphosphonate chelation to target bone mineral. Because the two probes operate at distinct wavelengths, dual‐channel photoacoustic imaging delineates cartilage and bone simultaneously without spectral interference.

1. Introduction

Imaging cartilage and bone in real time and with molecular or tissue specificity is critically important in orthopedic research and surgery [1, 2]. Bone and cartilage are anatomically adjacent yet chemically distinct tissues, commonly affected together in conditions like osteoarthritis or bone metastasis [3]. These tissues play distinct but tightly coupled roles in joint function and disease progression, and their interfaces are often the sites of early pathological change [4]. However, cartilage is thin, avascular, and compositionally heterogeneous, while bone is dense and highly mineralized, making simultaneous visualization of both tissues particularly difficult [5]. As a result, there is a persistent need for imaging approaches that can resolve cartilage and bone distinctly and sensitively within the same anatomical context.

Current clinical approaches for musculoskeletal imaging primarily rely on magnetic resonance imaging (MRI) and computed tomography (CT). MRI offers strong soft tissue contrast and is well suited for assessing cartilage morphology [6], while CT excels at imaging mineralized bones. Despite these strengths, both techniques are expensive, infrastructure‐intensive, and poorly suited for intraoperative deployment. Long acquisition times, limited portability, and the need for specialized facilities restrict their use to preoperative or postoperative settings, limiting real‐time assessment during orthopedic procedures [1, 7]. Ultrasound provides a more accessible alternative, offering real‐time imaging, portability, and low cost. It is widely used intraoperatively to visualize structural boundaries and tissue interfaces [8]. However, ultrasound contrast is governed largely by acoustic impedance mismatches, which can make it difficult to reliably distinguish cartilage from adjacent calcified or osseous tissue. Subtle compositional differences and early‐stage changes in cartilage are therefore still challenging to resolve using ultrasound alone. Fluorescence image‐guided surgery (FIGS) has also emerged as an intraoperative tool, with commercially available near‐infrared (NIR) fluorescence platforms now in widespread use for real‐time visualization of tissue perfusion, tumor margins, and anatomical landmarks (e.g., indocyanine green; ICG) [9, 10]. In orthopedic surgery, FIGS has shown promise for assessing bone perfusion to predict fracture healing capacity [11], guiding debridement in osteomyelitis [12], and delineating tumor boundaries during sarcoma resection, where ICG‐guided procedures have significantly reduced positive margin rates and local recurrence [13, 14]. However, FIGS remains fundamentally limited by the shallow penetration depth of NIR‐I fluorescence (typically 1–2 cm in soft tissue and substantially less in mineralized bone), the need to dim operating room lights to achieve adequate signal‐to‐background ratio, and the reliance on non‐targeted dyes such as ICG that accumulate via the enhanced permeability and retention effect rather than through specific tissue binding [10, 11].

Photoacoustic imaging (PAI) addresses several of these limitations by combining optical absorption contrast with ultrasonic spatial resolution, enabling deeper tissue penetration while retaining molecular sensitivity [15, 16]. In PAI pulsed laser excitation generates thermoelastic expansion in optically absorbing materials, producing ultrasonic waves that can be detected with conventional transducers—effectively overcoming the optical scattering limit that constrains fluorescence imaging [17, 18, 19]. While endogenous absorbers (e.g., hemoglobin) can provide intrinsic contrast [20], exogenous dyes expand PAI's capabilities to visualize specific molecular targets or tissues. Among exogenous agents, plasmonic nanoparticles (e.g., gold nanostructures) exhibit exceptionally strong optical absorption due to localized surface plasmon resonance [17, 21], often exceeding that of small‐molecule organic dyes by orders of magnitude. This high absorption cross‐section makes them highly efficient photoacoustic transducers. However, inorganic nanoparticles are generally not biodegradable and can bioaccumulate in organs such as the liver and spleen, raising long‐term safety and translational concerns. In contrast, small‐molecule organic dyes, while less optically efficient, are more readily cleared and metabolized, making them attractive for clinical translation.

However, currently available organic dyes present their own limitations (Figure S1). ICG, for example, is rapidly cleared (plasma half‐life ∼2‐4 min) [22], which limits effective accumulation. Alternatively, non‐fluorescent dyes (“dark” quenchers) are particularly well suited for PAI because they convert absorbed photon energy almost entirely into heat through nonradiative decay, maximizing thermoelastic signal generation [23, 24]. This principle has been validated using Black Hole Quencher (BHQ) dyes—Haedicke et al. demonstrated BHQ‐1 conjugated to cyclic RGD as a targeted optoacoustic contrast agent for integrin imaging in glioblastoma [25], while Maeda et al. used BHQ‐3‐labeled trastuzumab for dual photoacoustic and fluorescence imaging of HER2‐expressing breast tumors [26]. QSY21 dye has similarly been employed in activatable photoacoustic probes due to its strong absorption (λ _max ≈ 660 nm) and efficient nonradiative decay [27, 28]. However, these prior efforts have largely focused on oncologic targets using highly specific molecular recognition events—antibody‐antigen binding or integrin‐targeting peptides—and no non‐fluorescent dye‐based photoacoustic probes have been developed for musculoskeletal tissue discrimination.

To address this, we designed modular photoacoustic dyes that combine strong NIR absorbance with built‐in targeting moieties for bone and cartilage. We first conducted an evaluation of NIR dyes for photoacoustic performance, which identified QSY21 as exhibiting the lowest limit of detection and high photostability. This scaffold was therefore selected as the basis for further modification, enabling the introduction of tissue‐associated functionality while preserving favorable photoacoustic performance. We then investigated targeting capabilities. Recent advances in fluorescent imaging have identified scaffold‐based dyes that inherently target cartilage after systemic administration, presumably via favorable charge and polarity characteristics that promote binding to cartilage's glycosaminoglycan‐rich matrix [29, 30, 31]. For bone, bisphosphonate‐dye conjugates have shown strong affinity for hydroxyapatite [32], thus enabling long‐term bone labeling and enhanced imaging contrast. Inspired by these developments, we synthesized two new photoacoustic dyes: Cart‐670, a cartilage‐targeting NIR dye, and Osteo‐750, a bone‐targeting NIR dye (Scheme 1). Each contains a high‐absorbance chromophore core functionalized with a targeting moiety. The synthetic strategy follows a modular approach wherein the dye's NIR‐active core (absorption in the 650–750 nm range) is preserved while its periphery is tailored for tissue binding. We note that neither targeting mechanism relies on the high‐affinity molecular recognition characteristic of antibody‐antigen or enzyme‐substrate interactions; rather, both exploit physicochemical affinity—electrostatic and hydrogen‐bonding interactions with glycosaminoglycans for Cart‐670, and chelation of hydroxyapatite's calcium surface for Osteo‐750.

SCHEME 1.

Illustration of the targeting mechanisms: Cart‐670 preferentially accumulates in cartilage, consistent with electrostatic interactions with negatively charged proteoglycan‐rich matrices, while Osteo‐750 shows preferential retention in bone, consistent with interactions with calcium‐rich hydroxyapatite. These interactions are proposed to contribute to the observed tissue‐selective accumulation. The chromophore cores in both probes are derived from commercially available NHS esters whose complete chemical structures are proprietary and not disclosed by the vendor; accordingly, Scheme 1 depicts all non‐proprietary structural elements (targeting moieties, linkers, and the point of conjugation).

In this study, the optical and photoacoustic properties of Cart‐670 and Osteo‐750 were compared against non‐targeted benchmarks, including ICG and QSY21. We report in vitro characterization of absorption, photostability, and photoacoustic sensitivity, followed by ex vivo tissue imaging to assess relative accumulation in cartilage and bone. Together, these results demonstrate that high‐performance photoacoustic non‐fluorescent dyes can be adapted into tissue‐associated probes, supporting a modular design strategy for molecular photoacoustic imaging of musculoskeletal tissues.

2. Results and Discussion

2.1. Chromophore Design and Photophysical Properties

We began by evaluating a library of NIR dyes to identify an optimal PA contrast scaffold (see Experimental Section and Supporting Information Figures S2–S9 for synthetic information). ICG, a heptamethine cyanine dye bearing two sulfonate groups, was used as a “gold‐standard” control due to its extensive use as a clinically approved NIR dye [33, 34, 35]. Each candidate dye was tested in aqueous phantom conditions, defined here as dye‐loaded polyethylene tubes (0.86 mm i.d.) mounted in a 3D‐printed holder and immersed in a tissue‐mimicking bath containing titanium dioxide (28 mg in 40 mL, for optical scattering) and India ink (0.004%–0.01% v/v, for optical absorption), for its photoacoustic signal intensity, spectral stability, and detection sensitivity. The screen included NIR‐absorbing non‐fluorescent dyes with absorption maxima spanning 650–900 nm that are commercially available. Figure S10 summarizes the key findings. BHQ‐series quenchers were excluded from the screen because BHQ‐3 has already been characterized as a PA contrast agent in prior studies [36]. Importantly, it is also insoluble in aqueous media, has a molar absorptivity of only 42,700 M⁻¹ cm⁻¹ at 672 nm—less than half that of QSY21 (89,000 M⁻¹ cm⁻¹ at 660 nm)—and has documented in vivo instability with nonspecific dequenching occurring within minutes of injection [37, 38]. In short, completely non‐fluorescent dyes such as QSY21 and PL169‐74 (quantum yield 0.92%) reached peak PA intensities of ∼200 a.u. at saturating concentrations, approximately 4‐fold higher than fluorescent controls ICG (quantum yield 14%) and AF676 (quantum yield 20%), which plateaued near ∼50 a.u. under identical conditions (Figure S11). This disparity is consistent with the near‐complete non‐radiative relaxation of non‐fluorescent dyes, where absorbed photon energy is converted to heat rather than lost to fluorescence emission. ICG also exhibited poor photostability, retaining less than 15% of its initial signal within 24 h of sunlight exposure. In contrast, several candidate dyes, including PL‐160‐99 and PL‐160‐85, retained more than 50% of their signal even after 96 h (Figures S12 and S13).

Importantly, QSY21 (Figure 1A) emerged as the top‐performing dye, with a limit of detection (LoD) of 0.5 µM in PBS, defined as the lowest concentration at which the signal‐to‐noise ratio (SNR) exceeded 3, where noise is the standard deviation of the PA signal measured in a dye‐free PBS control region adjacent to the sample inclusion. ICG returned an LoD of 2.5 µM (Figure 1B) under identical excitation and detection conditions (680 nm excitation, 29 MHz transducer, 0 dB gain). This superior performance can be rationalized by considering the fundamental determinants of photoacoustic signal generation. The amplitude of the generated acoustic wave scales with the optical absorption coefficient and the fraction of absorbed energy that undergoes nonradiative decay within the thermal confinement time of the laser pulse [39]. QSY21 has a reported fluorescence quantum yield of ∼0%, ensuring nearly all absorbed energy converts to heat and thus photoacoustic signal, rather than being lost to radiative emission (Figure S15). The observed sub‐micromolar detection limit of QSY21 therefore reflects both its strong extinction coefficient and its optimized energy dissipation pathway (Table 1). This contrasts with ICG, which has a quantum yield of ∼14% and consequently partitions absorbed energy between fluorescence and internal conversion. The absorption spectrum of QSY21 was broad (FWHM ≈ 100 nm) with a maximum at 660 nm, in the near infrared window (650–900 nm) where tissue absorption is low (Figure 1C). Photoacoustic excitation was performed at 680 nm, which falls within the broad absorption envelope of QSY21 and corresponds to approximately 85% of its peak absorbance. After 10 min of continuous pulsed laser exposure, we identified no significant changes compared to the initial, indicating minimal photobleaching (Figures S16 and S17). In contrast, under the same conditions (680 nm, 10 Hz, 5 ns pulse width, 20 mJ cm⁻² fluence for 10 min), ICG exhibited a notable decrease (50%) in absorbance, consistent with photobleaching and degradation (Figure 1D). The absence of measurable photobleaching under repeated pulsed irradiation suggests that QSY21 maintains structural integrity during cyclic excitation, a property that is particularly important for longitudinal imaging or intraoperative applications requiring repeated frame acquisition [40, 41]. Importantly, both QSY21 and ICG maintained their spectral profile in biologically relevant media (50% plasma), with only minor broadening (ΔFWHM < 15 nm) and negligible redshift (<20 nm), suggesting they are not prone to immediate aggregation or quenching in blood.

FIGURE 1.

Absorbance spectra and photoacoustic behavior of the NIR dyes. (A) Absorption spectra of 1 µM QSY21 and (B) ICG in: PBS (pH 7.4) (red/ green, respectively); 50% plasma (maroon/ teal, respectively); after photoacoustic laser irradiation for 10 min (black). QSY21 absorbs broadly with λ _max at 660 nm and ICG has a peak at ∼780 nm but still exhibits ∼50% of max absorption at 680 nm. Both dyes show high molar extinction (> 10⁴ M⁻¹cm⁻¹). The spectra remain well‐defined (with minor red‐shift and broadening) in the protein‐rich environment of 50% plasma. Black curves are photostability test results: These data are absorption after 10 min of pulsed laser irradiation (680 nm, 10 Hz; pulse width: 5 ns; fluence: 20 mJ cm⁻²) showing no significant change, indicative of high photostability. (C) Photoacoustic signal intensity for QSY21 and (D) ICG. Each curve represents a concentration‐response profile (0–10 µM) acquired in PBS, PBS within a skin‐mimicking (SM) phantom, and 50% plasma. As expected, overall photoacoustic signal intensity decreases in more optically and acoustically attenuating environments due to increased scattering, absorption, and acoustic impedance; however, the signal remains concentration‐dependent and readily detectable across conditions. Note: concentrations above ∼10 µM approach detector saturation and were excluded from quantitative comparisons (Figure S14). These properties illustrate that QSY21 is suitable for consistent photoacoustic signal generation under biological conditions.

TABLE 1.

Photophysical properties of NIR dyes.

Dye	Role	LoD (µM) PBS, Plasma	Absorbance max (nm)	Molar absorptivity (M−1 cm−1)	Quantum Yield (%)	Brightness (M−1 cm−1)	Photoacoustic Loudness (M−1 cm−1)
ICG	Control	2.5, 5	787	223,000	14	31,220	191,780
QSY21	Cart‐670 Chromophore	0.5, 5	660	89,000	0	0	89,000
Alexa Fluor750	Osteo‐750 Chromophore	—	749	290,000	12	34,800	255,200

* PA Loudness is defined as ε ‐ Brightness, representing the molar absorptivity‐weighted fraction of absorbed energy available for nonradiative (thermal) conversion. This metric provides a first‐order estimate of relative photoacoustic efficiency under matched conditions.

2.2. Design of Cartilage‐ and Bone‐Targeted PAI Probes

Using QSY21 as the core chromophore, we next developed a tissue‐specific PAI contrast agent by attaching a cartilage‐targeting motif. Cartilage extracellular matrix is rich in negatively charged sulfated glycosaminoglycans (sGAGs) and proteoglycans, which create a high fixed negative charge density. Many cartilage‐targeted fluorophores and drug carriers explicitly exploit this feature: they bear cationic or zwitterionic groups that drive electrostatic partitioning into GAG‐rich cartilage [6, 42]. However, these existing probes are designed for fluorescence detection, and their high quantum yields inherently limit photoacoustic signal generation. We therefore adapted these established targeting strategies onto PA‐optimized chromophores. The Cart‐670 probe was synthesized by conjugating a cartilage‐affine ligand to the QSY21 NHS ester. Successful conjugation was confirmed via high performance liquid chromatography (HPLC) and mass spectrometry (MS) (see Experimental Section for synthetic details). We chose a small cationic moiety as the targeting ligand, taking inspiration from known cartilage‐targeting fluorophores that use permanent positive charges to bind the negatively charged cartilage matrix [43]. In particular, quaternary ammonium groups and cationic peptides have been shown to preferentially accumulate in cartilage by electrostatic association with GAGs [44]. In Cart‐670, the introduced cationic functionality confers a strong cartilage affinity without significantly altering the dye's optical properties. The absorption maximum of Cart‐670 remains around 660 nm (similar to unmodified QSY21), ensuring it operates in the first near‐IR window. We note that this absorption maximum is slightly blueshifted relative to the minimum excitation wavelength of our photoacoustic laser (680 nm); Cart‐670 is therefore excited on the red edge of its absorption band, where it retains approximately 70%–80% of its peak absorbance. Figure 2A shows the absorption spectrum of Cart‐670, which overlaps with the parent QSY21 spectrum, indicating that the conjugation did not diminish or shift the chromophore's absorbance. The probe retains a high extinction coefficient and produces robust PA signals comparable to the unconjugated dye.

FIGURE 2.

In vitro characterization of targeted NIR dyes. (A) Absorption spectra of Cart‐670 and (B) Osteo‐750 in: PBS (pH 7.4) (blue/ brown, respectively); 50% plasma (purple/orange, respectively); after photoacoustic laser irradiation for 10 min (black). (C) Schematic of the photoacoustic imaging setup and multi‐well phantom used for concentration‐dependent (0–100 µM) measurements and representative photoacoustic images acquired in PBS, illustrating signal intensity as a function of decreasing dye concentration. (D) Photoacoustic signal amplitude as a function of concentration for Cart‐670 and (E) Osteo‐750 in PBS, 50% plasma, and skin‐mimicking (SM) phantom conditions. Insets show the linear region with corresponding R² values. Error bars represent mean ± s.d. of two replicates.

In parallel, we designed a bone‐targeted PA probe termed Osteo‐750 to serve as a second imaging channel. Our goal was to create an analog of Alexa Fluor 750 that absorbs at a longer wavelength (∼750 nm) and specifically localizes to bone. Alexa Fluor 750 was selected because it is ∼100 nm redshifted relative to QSY21, enabling dual‐channel imaging. Notably, the pairing of QSY21 with cartilage and Alexa Fluor 750 with bone is not interchangeable. QSY21 succinimidyl ester is insoluble in aqueous media, whereas alendronate sodium (the bone‐targeting moiety) requires aqueous conditions for conjugation. This chemical incompatibility precludes the synthesis of a QSY21‐alendronate conjugate and thus dictates the assignment of each dye to its respective target. Although Alexa Fluor 750 is a fluorescent dye (quantum yield ∼12%) its high molar absorptivity (290,000 M⁻¹ cm⁻¹) yields a photoacoustic loudness of 255,200 M⁻¹ cm⁻¹ (Table 1), making it a strong PA emitter despite the energy lost to fluorescence. The term “photoacoustic loudness” has been defined as ε × (1 − Φf), where ε is the molar absorptivity and Φf is the fluorescence quantum yield [45]. This metric provides a first‐order estimate of the molar absorptivity‐weighted fraction of absorbed energy available for thermal conversion and thus thermoelastic signal generation. We note that this is an upper‐bound estimate: Not all nonradiative relaxation pathways (e.g., intersystem crossing to triplet states, photochemical side reactions) necessarily contribute to the rapid heating required for efficient photoacoustic transduction within the thermal confinement time of the laser pulse. For bone targeting, we selected a bisphosphonate (BP) moiety, as bisphosphonates are well known for their strong affinity to hydroxyapatite in bone mineral. Bisphosphonates have been widely used as bone‐seeking agents in both therapy and imaging, including clinically in osteoporosis treatments and preclinically as bone‐targeted fluorescent probes [32]. Bisphosphonate‐mediated targeting in Osteo‐750 is driven by high‐affinity chelation to calcium ions in hydroxyapatite crystals. The phosphonate groups coordinate Ca^{2 +} on the mineral surface, producing strong adsorption at bone interfaces [45, 46]. This interaction effectively concentrates the chromophore at the mineralized surface and limits diffusion‐mediated washout, explaining the persistent photoacoustic signal observed after multiple washing steps. Unlike electrostatic cartilage association, bone targeting through mineral chelation is spatially confined to regions of exposed hydroxyapatite, resulting in a surface‐localized signal that sharply delineates bone structures. Importantly, the absorption spectrum of Osteo‐750 is centered around 750 nm, providing a clear spectral separation from Cart‐670. Osteo‐750 maintained high absorption in biologically relevant media (Figure 2B), fulfilling the criteria for an efficient PA agent.

2.3. In Vitro Photoacoustic Performance

The photoacoustic signal generation of the modified dyes was evaluated in vitro and detection sensitivity (limit of detection, LoD) was assessed under various conditions (Figure 2C–E). All dyes had a photoacoustic signal intensity that was linear with concentration in PBS. These calibration curves showed that the dyes could be detected down to low micromolar concentrations (<2 µM) with 680 nm excitation and 29 MHz detection. QSY21 and ICG, serving as reference standards, had LoDs on the order of 1 µM in PBS (signal‐to‐noise ratio ≥ 3). The targeted dyes showed comparable sensitivity: Cart‐670 was detectable down to ∼1 µM, while Osteo‐750 had an LoD around 2–3 µM. These thresholds reflect the dyes’ high optical absorbance and efficient energy conversion. We note that at the lowest gain settings (to simulate high background scenarios), the LoDs increased (e.g. ∼25 µM for Osteo‐750 at gain ‐20), but using nominal or positive receiver gain allowed detection in the single micromolar range for all dyes (see Figure S18–S20).

The presence of the biological matrix can attenuate sensitivity, so the dyes were examined in 50% blood plasma, and in tissue‐mimicking agar phantoms. In plasma alone, the max absorptivity for QSY21, Cart‐670, and Osteo‐750 remain stagnant (statistically negligible red‐shift of less than 20 nm) in the protein‐rich environment while ICG shows signs of aggregation (> 100 nm blueshift). However, background absorption and protein binding caused an increase in LoD for all dyes: for instance, Cart‐670's LoD rose to ∼20 µM at default gain (vs 1 µM in PBS), though stronger amplification of the gain settings could partially offset this. Osteo‐750 showed a similar trend. ICG and QSY21 likewise showed reduced PA signals in plasma. For ICG, this likely reflects competing effects: while protein binding (primarily to albumin and lipoproteins) can enhance fluorescence by disaggregating dye stacks and increasing quantum yield, this same increase in radiative decay diverts energy away from the nonradiative thermal conversion that drives photoacoustic signal generation. The net result is lower photoacoustic output in protein‐rich media despite potentially higher fluorescence—a tradeoff that is inherently absent for non‐fluorescent quenchers like QSY21, whose reduced PA signal in plasma is instead attributable to background absorption and scattering. Still, all dyes remained detectable at tens‐of‐micromolar concentrations in plasma, indicating feasibility for in vivo use given typical dosing (Figure S21). Quantitatively, the signal‐to‐background ratio for all four dyes at 50 µM in PBS exceeded 10:1 at the inclusion locations compared to dye‐free control inclusions, confirming that even at moderate micromolar levels the contrast is readily appreciable. Bright PA signals are observed at the inclusion locations for all four dyes compared to a dye‐free control inclusion, confirming that even at low micromolar levels the contrast is readily appreciable. These in vitro results demonstrate that the new targeted dyes perform on par with—and in some respects beyond—standard dyes like ICG in generating strong photoacoustic signals at biologically relevant concentrations. The targeted dyes matched ICG's detection sensitivity in PBS (Cart‐670 LoD: 1 µM; Osteo‐750 LoD: 2–3 µM; ICG LoD: 2.5 µM) while exhibiting markedly superior photostability, retaining >95% of initial PA signal after 10 min of pulsed irradiation at 20 mJ cm^{− 2} compared to a ∼50% signal loss for ICG under identical conditions. Additionally, the signal‐to‐background ratio at 50 µM exceeded 10:1 for all targeted probes in PBS, compared to ∼8:1 for ICG.

Finally, before evaluating the imaging performance of Cart‐670 and Osteo‐750 in tissue, we confirmed their cytocompatibility to ensure suitability for biological applications. Cytocompatibility was assessed using an MTT assay: HEK‐293T cells incubated with Cart‐670 or Osteo‐750 (0‐1000 µM, 24 h) maintained >90% viability across all concentrations tested (Figure S22). Encouraged by these results, we next investigated the targeting specificity and imaging utility of Cart‐670 and Osteo‐750 using ex vivo cartilage and bone samples.

2.4. Ex Vivo Photoacoustic Performance

We used fresh chicken cartilage and bone as a model system and stained the samples with Cart‐670 (10 µM, 50 µL applied topically; 15 min incubation at room temperature). After washing (∼5 mL PBS rinse to remove unbound dye; see Figure S23 for representative images of incomplete vs. complete washing), the samples were imaged by a high‐frequency photoacoustic scanner. Functionally, Cart‐670 exhibits a preferential uptake in cartilage‐rich tissues. We attribute this selective enrichment to electrostatic attraction between the cationic Cart‐670 and the anionic cartilage matrix, consistent with previously reported GAG‐binding cationic dyes [47, 48]. Importantly, we avoid describing this interaction as irreversible binding; rather, Cart‐670 shows preferential association with cartilage, leading to higher local concentrations in cartilage than in non‐cartilaginous tissue (Figure 3A). When free dye was washed out of tissue sections, Cart‐670 remained associated with cartilage zones, whereas little residual signal was observed in other regions (Figure 3B).

FIGURE 3.

Ex vivo specificity. (A) Imaging plane and anatomical context. (B) Photoacoustic imaging demonstrating tissue‐selective signal localization for cartilage‐ and (C) bone‐associated probes. (Left) Co‐registered ultrasound images of the joint. (Right) Photoacoustic images acquired independently for each probe, showing preferential signal localization of Cart‐670 to cartilage (B, red) and Osteo‐750 to bone (C, blue). Pseudo‐color assignments (red: Cart‐670 channel; blue: Osteo‐750 channel) are arbitrary and do not correspond to the gross visual appearance of the tissue.

We repeated this work with Osteo‐750 using matched staining conditions (10 µM, 50 µL, 15 min incubation) (Figures 3C and S24). Here, a photoacoustic signal approximately 200x higher than that on cartilage was observed from the bone surface. We attribute this to the attraction of the bisphosphonate group in Osteo‐750 for calcium‐rich bone mineral, leading to surface deposition of dye molecules. The bone‐binding strength is such that even after multiple washes the dye is not removed. Notably, Osteo‐750 did not appreciably bind to cartilage (its signal on cartilage dropped to baseline after washing). The specificity is further supported by spectral analysis of the bone signal—the photoacoustic spectrum of the Osteo‐750‐treated bone matched the dye's known absorption profile, confirming the signal is indeed from bound dye and not, for example, blood residue (Figure S25).

2.5. Dual‐Channel Cartilage and Bone Imaging

Having developed Cart‐670 and Osteo‐750 as separate cartilage‐ and bone‐targeted PA probes, we then demonstrated their combined use for dual‐channel imaging. We first confirmed minimal cross talk between the two channels (Figure S26). Figure 4A illustrates the two‐channel photoacoustic imaging of a biological sample containing both cartilage and bone. The sample was treated with both Cart‐670 and Osteo‐750 by co‐incubation at varying ratios (0%–75%) of Cart‐670 (total dye concentration maintained at 10 µM; 50 µL applied topically; 15 min incubation), and PA images were acquired at two wavelengths corresponding to each probe's absorption peak (680 and 750 nm). Spectral unmixing was applied to differentiate the signals, yielding a composite image where cartilage‐associated and bone‐associated signals are pseudo‐colored separately at increasing ratios of Cart‐670 (Figure 4B–E). Cart‐670 produced a strong signal localized to the articular cartilage lining the joint. In the same field, Osteo‐750 highlighted the underlying bone structures, such as the epiphyseal bone or areas of calcified matrix. Notably, no appreciable Cart‐670 signal was detected in the bone regions and vice versa, confirming minimal cross talk between the channels. The use of well‐separated absorption peaks (∼100 nm separation) ensured that each probe could be excited and detected independently.

FIGURE 4.

Dual‐channel photoacoustic imaging. (A) Schematic of co‐administration of Cart‐670 and Osteo‐750 at varying molar ratios, followed by photoacoustic imaging. (B‐E) Representative co‐registered ultrasound (grayscale) and spectrally unmixed photoacoustic images acquired with varying ratios (Cart‐670:Osteo‐750 = 0:100 (B), 25:75 (C), 50:50 (D), 75:25 (E)) with total probe concentration held constant. Bone‐associated signal (blue) and cartilage‐associated signal (red) are shown overlaid after Ex.: 680 nm and 750 nm. Increasing Cart‐670 fraction results in progressive enhancement of cartilage‐localized signal while maintaining detectable bone signal, demonstrating tunable contrast between cartilage and bone within the same imaging field.

Finally, to simulate intraoperative scenarios in which bone and cartilage are partially covered by overlying soft tissue, we placed slabs of 2 and 3 mm thickness over stained bone/cartilage samples. PA signals from both Cart‐670 and Osteo‐750 remained detectable through 3 mm of simulated tissue coverage, with signal attenuation of ∼40% at 2 mm and ∼70% at 3 mm (Figure S27). These findings suggest that both probes can generate sufficient contrast through thin overlying tissue layers, though sensitivity will depend on the optical properties and thickness of the intervening tissue in clinical settings. This dual‐channel approach allowed us to simultaneously visualize cartilage and bone within one imaging session—an ability that is highly valuable in contexts like osteoarthritis research, where evaluating the cartilage‐bone interface is crucial.

The dual imaging showcases the power of spectral multiplexing in photoacoustic imaging: by leveraging two tailored contrast agents, we can distinguish and map multiple tissue types in situ. Previous multiplexed PA studies have often combined nanoparticle agents or dye with endogenous hemoglobin, but here, we present two compatible small‐molecule PA probes that are each tissue‐specific and jointly applied. This represents, to our knowledge, the first example of simultaneous cartilage and bone photoacoustic imaging with molecular specificity. The concept is analogous to multi‐color fluorescence imaging but executed in the PA modality, which affords greater imaging depth. We anticipate that such dual‐channel PA imaging can provide new insights into musculoskeletal disease models—for instance, monitoring cartilage degradation alongside subchondral bone remodeling in arthritis.

Beyond photoacoustic detection, we note that the strong visible absorption of these dyes also produces macroscopic color changes upon tissue staining that are apparent by eye. Cart‐670 imparts a visible blue‐green tint to cartilage, while Osteo‐750 darkens bone surfaces. This raises the intriguing possibility that these probes could serve a dual role: providing molecular contrast for photoacoustic imaging while simultaneously enabling gross visual assessment of tissue boundaries during surgery, analogous to proposed strategies that use chromophores to render tumor margins visible to the surgeon. Additionally, conventional B‐mode ultrasound—which is inherently co‐registered with photoacoustic data in our imaging system—offers complementary structural information, including delineation of bone surfaces for anatomical context and Doppler capability for assessing regional blood flow. The combination of molecular photoacoustic contrast, visual tissue staining, and co‐registered ultrasonographic data could provide a comprehensive intraoperative toolkit for musculoskeletal procedures.

3. Conclusion

In summary, we have extended the use of non‐fluorescent dyes in photoacoustic imaging by converting them into tissue‐selective contrast agents for musculoskeletal applications. Through an initial screen of NIR‐absorbing quenchers, QSY21 was identified for its exceptional PA signal generation and low detection limit and was used to construct the cartilage‐targeted probe Cart‐670 via NHS ester chemistry. In parallel, the bone‐targeted probe Osteo‐750 was generated by conjugating a bisphosphonate targeting moiety to Alexa Fluor 750, a NIR‐absorbing dye compatible with photoacoustic imaging. Both probes preserve strong NIR absorption while gaining selective accumulation in their intended tissues. Cart‐670 leverages electrostatic affinity for glycosaminoglycans to concentrate on cartilage, whereas Osteo‐750 exploits bisphosphonate‐mineral binding for bone targeting—strategies grounded in established targeting motifs from prior fluorescence imaging studies. Importantly, the probes operate at distinct wavelengths, enabling two‐color photoacoustic imaging without spectral interference. The dual‐channel imaging results clearly delineate cartilage and bone simultaneously, demonstrating the feasibility of multiplexed optoacoustic visualization of complex skeletal tissues.

Several limitations should be noted. First, the absorption maximum of Cart‐670 (∼660 nm) is blueshifted relative to the minimum excitation wavelength of the Vevo LAZR system (680 nm), meaning Cart‐670 is not excited at its optimal wavelength; instruments with broader tuning ranges would improve performance. Second, the ex vivo model (chicken joints) does not fully recapitulate human tissue composition and in vivo validation with systemic delivery will be needed to assess pharmacokinetics and off‐target accumulation. Third, the selectivity demonstrated here reflects topical application conditions; the extent to which these probes maintain tissue selectivity under more complex in vivo conditions remains to be determined.

Conflicts of Interest

The authors declare no conflicts to declare.

Supporting information

Supporting File: The authors have cited additional references within the Supporting Information [49–51].

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.