Development of a Rare Earth Nanoprobe Enables In Vivo Real-Time Detection of Sentinel Lymph Node Metastasis of Breast Cancer Using NIR-IIb Imaging

A sentinel lymph node (SLN) is the first in a chain of lymph nodes (LN) draining a primary tumor site (1). SLN biopsy (SLNB) has now become the standard method for axillary staging of clinically node-negative breast cancer (2). SLN localization and targeted excision are usually guided by injectable dye [methylene blue (MB) or indocyanine green (ICG)], radiolabeled tracer [technetium ⁹⁹m sulfur colloid (⁹⁹mTc)], or a combination thereof, novel tracers such as carbon nanoparticle suspension (CNS), ⁹⁹mTC-Rituximab and ICG-Rituximab. There are drawbacks to the above, such as insufficient SLN accumulation, poor imaging signal contrast, or radiation-induced tissue damage.

The radiation hazards of ⁹⁹mTc radioisotope, for example, pose logistic and legal issues that curtail its broader use; and although MB and patent blue are the dyes most widely used to identify SLNs in this setting, the learning curve for acquired proficiency by beginners is relatively lengthy (e.g., 30–40 cases exposures; refs. 3, 4). ICG [an FDA-approved near-infrared (NIR) dye] is also commonly used for SLN mapping in many clinical trials. Several retrospective studies have proven ICG fluorescence imaging to be a reliable and safe alternative to a radiotracer approach, allowing real-time transcutaneous and intraoperative visualization of lymphatic vessels and SLNs (5–8). Again, the disadvantages of ICG, such as severe photobleaching and poor imaging detection depth or signal contrast, limit its surgical applicability and procedural efficiency. In some instances, surgeons must perform operations under dim light or frequently switch the room lighting to avoid interference and bleaching of dyes during repeated light exposures (3). Novel tracers such as CNS (9), ⁹⁹mTc-Rituximab (10), and ICG-rituximab (11) have been applied to detect SLNs, even though CNS have shown some limitations, for example, showing prolonged operation time, with radioactive issues (⁹⁹mTc-Rituximab) and limited penetration depth (ICG-rituximab). Moreover, none of them can visualize the metastatic status of LNs.

Determining the presence or absence of SLN metastasis is pivotal for therapeutic planning and clinical decision-making. Unfortunately, ICG, MB, and ⁹⁹mTc are simply tools for SLN mapping, providing no indication of whether or not tumor is present (12). Once SLNs are excised, histologic examinations take place to make such determinations (12). A substantial time savings, benefitting both patients and surgeons, would result from accurately establishing SLN involvement intraoperatively. More importantly, detecting metastasis with precision at an early-stage stands to enhance patient outcomes and quality of life (13).

Intraoperative frozen section or cytoprinting is the customary means of identifying SLN metastasis. Although evaluations may be completed within 30 minutes, it is nearly impossible to evaluate an entire SLN due to technical limitations. Furthermore, van de Vrande and colleagues have shown that both techniques have rather low sensitivities for detecting metastases (69.7% and 76.2%, respectively), which fall to dismal levels for micrometastases (23.5% and 35.4%, respectively; ref. 14).

As the gold standard, paraffin-embedded tissue sections are routinely processed to evaluate LN status and render a final diagnosis. However, this is a time-consuming practice, usually requiring 2 to 5 days from time of surgery for completed reports to be issued. Medical costs and patient anxiety are ultimately heightened, because up to 40% of patients must thereafter undergo axillary LN dissections (15). One-step nucleic acid amplification (OSNA) analysis was developed to facilitate metastasis screening in this setting, offering a means of detecting nodal involvement intraoperatively through expression levels of cytokeratin 19 (CK19) mRNA. The OSNA approach has shown sufficient predictive value in LNs to distinguish between macro- or micrometastases and tumor-free states, but false-negative results are possible if CK19 expression is lacking (15, 16). Still, the demand for fast and sensitive methods of accurately identifying SLN metastases intraoperatively is strong.

Because discernible SLNs are tumor-free in >70% of patients subjected to SLNB, there is a great deal of unnecessary surgery entailed, with lymphedema, shoulder or arm pain, and numbness as complications (17–19). A noninvasive mode of intraoperatively verifying SLN metastasis in vivo is therefore urgently needed. As described above, ICG fluorescence imaging is suitable for SLN mapping, given its high sensitivity, nonradioactive nature, and low cost. Fluorescence imaging in the NIR-II window (1,000–1,700 nm) achieves high tissue penetration and spatial resolution, which is particularly appealing (20). To date, a number of nanoprobes capable of NIR-II luminescence have emerged for LN imaging, including semiconducting quantum dots (QD; refs. 21–24), single-walled carbon nanotubes (SWNT; refs. 25–27), rare earth-based nanomaterials (28), and small organic molecules (29, 30). Rare earth-based nanomaterials especially (vs. existing fluorescence probes) have distinct physical and chemical properties, including extensive spectral tunable range, low photobleaching, low cytotoxicity, and narrow emission bandwidths (29). Such properties are of probable use for highly sensitive, noninvasive detection of SLN metastatic status in real time.

CXCR4 is a chemokine receptor and the focus of much scientific interest in recent years. Some preclinical studies have shown a close relation between expression of CXCR4 and invasion or distant metastasis of breast cancer cells (31–35). It is likely that CXCR4 is overexpressed in nodal metastases, thus indicating a higher risk of spread to distant organs (31, 32, 36). Many molecular reagents (including small molecules, peptides, and proteins) have been developed for CXCR4-based imaging in various tumor models, although most have addressed primary tumors rather than metastatic disease (35, 37–39). The feasibility of a CXCR4-targeted probe for monitoring SLN metastasis in real-time (noninvasively and in vivo) remains barely explored at this juncture, prompting the present investigation. Herein, we have devised a straightforward and adaptable fluorescent probe (ErNPs@POL6326) using erbium ion-doped nanoprobes (ErNP) covalently conjugated with balixafortide (POL6326). The latter is a CXCR4 antagonist tested in multiple clinical trials for treatment of metastatic breast cancer (40). This construct has been successfully configured for NIR-IIb fluorescence bioimaging (Fig. 1), allowing precision assessment of tumor presence/absence in SLN and secondary LNs with high sensitivity and specificity. The fluorescence signals emitted positively correlate with LN metastasis.

Going forward, our goal is avoidance of unnecessary SNLBs through noninvasive in vivo visualization of nodal metastasis. The method we propose holds considerable promise as a reliable, real-time diagnostic tool for clinical use, one that may be important in axillary staging of breast cancer and improve patient prognosis overall.

Purchased chemicals were sourced as follows: sodium hydroxide (NaOH, 99%) and ammonium fluoride (NH₄F, 98%) from Shanghai Aladdin Biochemical Technology; oleic acid (OA, 98%), 1-octadecene (ODE, 95%), erbium acetate hydrate (Er[CH₃COO]₃·xH₂O, 99.9%), yttrium acetate hydrate (Y[CH₃COO]₃·xH₂O, 99.9%), and polyacrylic acid (PAA) from Sigma-Aldrich Corp.; balixafortide from MedChemExpress; anti-CXCR4 antibodies from Abcam Trading Co.; and anti-GAPDH from Proteintech Group.

Mouse breast cancer cells (4T1-Luc), human breast cancer cells (MDA-MB-231-Luc, MCF7, BT-549, T-47D, SK-BR-3, BT-474), and a noncancerous mammary epithelial cell line (MCF10A) were purchased from Procell Life Science & Technology Co Ltd. Only those with <25 passages were utilized herein. All cell cultures were incubated in 5% CO₂ at 37°C and 5% humidity, according to vendor recommendations.

CXCR4 knockdown lentivirus (target sequence: 5′-CCATATCATCTACACTGTCAA-3′) was purchased from HANBIO Biological. Lentivirus were added to infect 4T1-Luc cells according to the recommended MOI value. Cell screening was performed 24 hours after infection with puromycin (8 μg/mL). Finally, the efficiency of knockdown construction was verified by Western blot detection of CXCR4 expression level (anti-CXCR4, Santa Cruz Biotechnology).

The animal experiments were approved by the Institutional Animal Care and Use Committee of Xiamen University (Ethics Approval No. XMULAC20180037). Female BALB/c nude mice and BALB/c mice (4–6 weeks) were obtained from the Xiamen University Experimental Animal Center.

Typically, Er(CH₃COO)₃. xH₂O (1 mmol) was mixed with 5 mL OA and ODE at three times the volume of OA in a three-neck round-bottom flask. The mixture was then heated to 120°C in a vacuum, with constant stirring for 30 minutes. Once cooled to 50°C (and the solution had become clear), 4 mmol NH₄F and 5 mmol NaOH were dissolved in methanol, added to the flask, and stirred for 30 minutes at 50°C. To evaporate the methanol, the reaction vessel was heated to 100°C for 30 minutes and further heated to 300°C under vigorous nitrogen (N₂) flow for 1.5 hours. Core nanocrystals were then collected by centrifugation at 8,000 rpm for 10 minutes at room temperature, washed three times with a mixture of cyclohexane and absolute ethanol, and finally dispersed in cyclohexane (5 mL).

To generate NaErF₄@NaYF₄ nanoparticles (ErNPs), Y(CH₃COO)₃^. xH2O (0.5 mmol), ODE (10.5 mL), and OA (4.5 mL) were mixed in a three-neck round-bottom flask. Subsequent steps were essentially identical, adding 2 mmol NH₄F and 2.5 mmol NaOH instead.

Having been collected as above (centrifugation at 8,000 rpm for 10 minutes), we added hydrochloric acid (pH 1) to the ErNPs under intense ultrasound to redisperse the precipitate and stirred at room temperature for ∼50 minutes. The resultant ligand-free ErNPs were again collected by centrifugation at 13,000 rpm for 15 minutes, washed three times with ultrapure water, and redispersed in it for further use. Thereafter, ligand-free ErNPs (5 mL) were mixed with PAA aqueous solution (100 mg, 5 mL) and vigorously sonicated for 6 hours at room temperature. The PAA-modified ErNPs (ErNPs@PAA) were subsequently retrieved by centrifugation and washed several times to remove any unreacted PAA. To activate the product, we used N-(3-dimethylaminopropyl)-N'-ethylcarbodiimide hydrochloride (EDC) and N-hydroxysuccinimide (NHS). Upon dissolving 0.5 mmol ErNPs@PAA in 10 mL of 2-(N-morpholino) ethanesulfonic acid (MES) buffer (0.1 M, pH 5), 50 mg EDC was added while stirring, followed by 50 mg NHS; and the combination was stirred for 2 hours at room temperature. This solution was then treated with 5 mg of POL6326 and vigorously stirred for 24 hours at 4°C. Centrifugation was used to recover the POL6326-functionalized ErNPs@PAA (ErNPs@POL6326), which were subsequently washed several times to remove excess POL6326. Finally, 0.9% sodium chloride injection was used for ErNPs@POL6326 dispersal.

Images were captured using a transmission electron microscope (Talos F200S; Thermo Fisher Scientific) at an accelerating voltage of 100 kV. At 25°C, zeta potentials of ErNPs before and after linkage to POL6326 were measured (NanoBrook Omni analyzer; Brookhaven Instruments Corp). XRD patterns were acquired using a powder X-ray diffractometer (Miniflex 600; Rigaku Corp.). FTIR spectroscopy was performed using a Thermo Nicolet iS50 FTIR spectrometer (Thermo Fisher Co., Ltd.). A fluorescence spectrophotometer [Cary Eclipse; Varian (Agilent Technologies)] served to measure UV absorption spectra at 720-nm excitation wavelength. Elemental content was determined using an ICP optical emission spectrometer (ULTIMA 2; HORIBA Scientific).

To investigate fluorescence stability, the probe was dissolved in 0.9% sodium chloride injection, 10% fetal bovine serum (FBS), and water and stored at 25°C for specified times (0, 1, 3, 7 days) under natural light conditions. In addition, the probe and ICG, respectively were dissolved in 0.9% sodium chloride injection, 10% FBS, and water, recording fluorescence intensities under continuous irradiation (808 nm) for 30 min (Suzhou Yingrui Sensing Technology Co Ltd, Suzhou, China).

For comparing depth of penetration, capillary glass tubes containing ErNPs were placed in a transparent petri dish, adding various volumes of 1% intralipid to cover the tubes at varying heights (0–7 mm). Fluorescence signals from the tubes were recorded by NIR I/II spectrometer (Series III 900/1700; Suzhou Yingrui Sensing Technology Co. Ltd.).

A CCK-8 Test Kit was used to assess the biocompatibility of ErNPs@POL6326 (nanoparticles). MCF10A, MDA-MB-231-Luc, and 4T1-Luc cells were seeded in 96-well plates (5,000 cells per well) for incubation overnight at 37°C. The medium was replaced with fresh aliquots containing varying concentrations of ErNPs@POL6326, and the cells were cultured continuously for 24 hours. The CCK-8 assay was used to determine cell viability.

Using 96-well plates, 4T1-Luc cells (5,000 cells per well) were seeded and cultured for 24 hours. The cells were then incubated with ErNPs@POL6326 for various times (1, 2, 4, or 6 hours) before fixation in 4% paraformaldehyde (PFA) solution for 20 minutes. Fluorescence signals of 4T1-Luc cells at respective time points were recorded by spectrometer (Series III 900/1700; Suzhou Yingrui Sensing Technology Co. Ltd.). For the concentration gradient experiment, 4T1-Luc cells were incubated in medium containing ErNPs@POL6326 (Er³⁺: 0, 1, 2.5, 5, 10, μg/mL) for 6 hours. To investigate the effect of POL6326 on ErNPs uptake, 4TI cells were treated for 6 hours with medium containing 10 μg ErNPs@POL6326, alone or with excess POL6326 (100 nmol/L), or ErNPs@PAA, followed by fluorescence detection. MDA-MB-231-Luc and MCF7 (5,000 cells per well) were seeded in 96-well plates and incubated with 10 μg ErNPs@POL6326 for 6 hours to investigate the tumor-targeting ability of ErNPs@POL6326 across breast cancer cell lines.

Solutions ErNPs and ErNPs@POL6326 (4 mL, 1 mg/mL) were treated with rhodamine (50 μL, 5 mg/mL) and centrifuged several times (13,000 rpm, 15 min) to remove free or unstable rhodamine (until supernatant became colorless and transparent), yielding rhodamine-labeled ErNPs (Rho@ErNPs@PAA) and ErNPs@POL6326 (Rho@ErNPs@POL6326).

To test the cellular uptake efficiency of Rho@ErNPs@POL6326, Rho@ErNPs@PAA, or Rho@ErNPs@POL6326 (rhodamine, 8 μg/mL) with or without free POL6026 (100 nmol/L) were added to the culture medium after 6 hours. The cells were then fixed in 4% PFA and counterstained with 4′,6-diamidino-2-phenylindole (DAPI), capturing their images by fluorescence microscope (DM2700; Leica Microsystems).

To test the safety of ErNPs@POL6326 in vivo, healthy BALB/c mice (18–22 g) were randomly divided into five groups. On days 1, 3, 7, and 28, mice in the four test groups were injected with ErNPs@POL6326 through tail veins. The control group received normal saline instead. Blood samples were collected from all mice prior to sacrifice (Day 28) to evaluate hematopoietic, liver, and kidney functions, looking for signs of acute or chronic toxicity. Important organs (i.e., liver, kidney) were rinsed and fixed with cold saline and then stained with hematoxylin and eosin (H&E), again checking for pathologic changes. The tissue sections were viewed under a microscope (BX51; Olympus) equipped with digital imaging (DP72; Olympus) and photographed.

We inoculated 4T1-Luc cells (1 × 10⁵) subcutaneously in left flanks of BALB/c mice. When tumor volume reached 100 mm³, these tumor-bearing mice were subjected to in vivo NIR fluorescence imaging and biodistribution studies.

To evaluate targeting specificity, BALB/c mice bearing 4T1-Luc tumors were randomly grouped (n = 3 each) as ErNPs@POL6326 (with or without 2 mg/kg of free POL6326) or ErNPs@PAA (controls), administering intravenous injections of ErNPs@POL6326 or ErNPs@PAA (10 mg/kg) accordingly and acquiring images at specified time points (0, 6, 12, 24, 36, and 48 hours). The captured fluorescence images were analyzed [tumor-to-background ratios (41), TBR = (tumor fluorescence signal − imaging background signal)/(mouse background signal − imaging background signal)].

We injected 4T1-Luc cells (8 × 10⁵) or MDA-MB-231-Luc cells (3.5 × 10⁶) transfected with firefly luciferase subcutaneously into left hind footpads of 6-week-old female BALB/c or BALB/c nude mice to create a SLN metastasis tumor model. Recipients of 4T1-Luc cells were evaluated by in vivo imaging system (IVIS) at 7, 14, 21, and 28 days, once injected with luciferase substrate for 5 minutes into left SLN (inoculated side). The same steps were applied to recipients of MDA-MB-231-Luc cells, examining SLNs on the left (inoculated side) at 2, 4, 6, 7, and 8 weeks post-injection.

We injected 20 μL of ErNPs@POL6326 solution (10 mg/kg) subcutaneously into both hind footpads of BALB/c mice with lymphatic metastases. Tert-Amyl alcohol was used for sedation. At various time points after injection (1, 3, 6, 12, 24, 36, and 48 hours), the mice were imaged via IVIS Lumina apparatus (PerkinElmer Inc.) equipped with an 808-nm laser. Signal-to-background tissue ratios (SBR) = tumor fluorescence signal /mouse background signal (42).

We injected 4T1-Luc cells (8 × 10⁵) transfected with firefly luciferase subcutaneously into left hind footpads of 6- to 8-week-old female BALB/c mice, creating a SLN micrometastasis tumor model. The animals were injected on Days 3, 6, 10, 14, and 21, undergoing IVIS evaluations 5 minutes after luciferase substrate injections.

The procedure was consistent with the LN metastasis detection by fluorescence imaging described above.

GraphPad Prism 7.0 (GraphPad Software Inc.) was engaged for statistical analysis. To compare changes in fluorescence signals, two-tailed unpaired t test was applied. At least three independent readings (n ≥ 3) were recorded for measurements obtained. All data were expressed as mean ± SD values.

All raw data generated in this study are available upon request from the corresponding author.

Using a layer-by-layer epitaxial growth strategy, we succeeded in generating NaErF₄@NaYF₄ NIR-IIb nanoprobes (ErNP) for testing (43). Transmission electron microscopy (TEM) and dynamic light scattering (DLS) results confirm the merits of both core (NaErF₄) and enshelled (NaErF₄@NaYF₄) nanoparticles in terms of dispersibility and size uniformity (Fig. 2A and B; Supplementary Fig. S1). In addition, the number of regions analyzed for TEM microscopic analysis of NaErF₄ and NaErF₄@NaYF₄ nanoparticles was 150 and 98, respectively. Nanoagents with hydrodynamic diameters of 10 to 50 nm are widely known for rapid uptake into the lymphatics, serving to effectively identify SLNs (44). Furthermore, the high-resolution TEM (HRTEM, Supplementary Fig. S2A) image shows that the distance between the lattice fringes is 0.52 nm, which corresponds to the lattice spacing of the (100) plane of hexagonal NaErF₄ (JCPDF 27–0689). The diffraction rings of the SAED pattern from ErNPs in Supplementary Fig. S2B can be assigned to the (100), (110), (200), (111), (201), (210), and (300) planes of the standard hexagonal NaErF₄ structure. Polyacrylic acid (PAA) introduced into water-soluble ErNPs (to provide carboxyl groups; Supplementary Fig. S3) gradually diminishes the zeta potential of ErNPs after conjugation (ErNPs@PAA) from 44.7 ± 1 mV to 33 ± 5 mV (Supplementary Fig. S4A). Successful conjugation of POL6326 and ErNPs@PAA by amidation reaction further reduces the zeta potential to −23.4 ± 1 mV (Supplementary Fig. S4A). DLS (Supplementary Fig. S1) and TEM (Fig. 2C) also provide evidence of good ErNPs@POL6326 dispersal and size uniformity. Elemental mapping of ErNPs@POL6326 indicates that the Er fraction is concentrated within the core, whereas ytterbium (Y) is distributed throughout the shell (Fig. 2D). The dashed box and circle on the plot represented the characteristic peaks at 280 nm of the UV absorption spectra of POL6326 and ErNPs after successful coupling of POL6326 on the surface, respectively (Fig. 2E), and the disappearance of the characteristic bands at 1,700 cm⁻¹ (–COOH stretching vibration of ErNPs@PAA) in conjunction with the appearance of the characteristic peaks at 1,650 cm⁻¹ (stretching vibration of –CONH–) in the FTIR spectrum of ErNPs@POL6326 also confirm the successful conjugation of POL6326 onto the ErNPs@PAA (Supplementary Fig. S4B). In powder X-ray diffraction (XRD) analysis, ErNPs showed a high degree of crystallinity, retaining the hexagonal phase structure of pure NaErF₄ before and after further modification by POL6326 (JCPDS no. 27–0689, Supplementary Fig. S4C). Under laser excitation (808 nm), all nanoparticulates (ErNPs, ErNPs@PAA, and ErNPs@POL6326) exhibited strong NIR-IIb luminescence (Fig. 2F); and similar to nonmodified ErNPs, the emission spectra of ErNPs@PAA and ErNPs@POL6326 peaked at 1,556 nm, indicating that optical properties of ErNPs are unimpaired by PAA and POL6326. Also, the fluorescence intensity of ErNPs remained stable upon dispersal in various solutions, including deionized water, physiologic saline, and serum, for at least 7 days (Supplementary Fig. S5A and S5B). To ensure future biologic applications, optical stability of the probe was tested under continuous laser irradiation (808 nm) for 30 minutes. Compared with ICG, ErNPs proved more optically stable (Supplementary Fig. S5C and S5E), and fluorescence images of capillaries filled with ErNPs or ICG in solution illustrated the comparative superiority in penetration depth and signal-to-noise ratio of ErNPs (5 mm vs. 7 mm; Fig. 2G and H). Fast body vascular imaging of BALB/c nude mice (under 808-nm laser excitation), after injecting ICG or ErNPs, likewise has confirmed that NIR-IIb imaging at 1556 nm provides much deeper tissue penetration depths on a centimeter scale (Supplementary Fig. S6). These results suggest that ErNPs enable highly sensitive imaging of deep tissue.

As a feasibility exercise in applying ErNPs@POL6326 (nanoprobes) to living organisms, we evaluated their biosafety and targetability at a cellular level. First, CCK-8 assays were performed on nontumorigenic (MCF10A) and breast cancer (4T1-Luc and MDA-MB-231-Luc) cells to assess potential cytotoxicity of ErNPs@PAA or ErNPs@POL6326. Cell viability was unaltered by various concentrations of either nanoparticulate, implying that ErNPs@PAA and ErNPs@POL6326 pose no overt cytotoxic threat to cells (Fig. 3A–C). To determine the IC₅₀ and working concentrations of the probe, 4T1-Luc cells were incubated with different concentrations of ErNPs@POL6326 for 24 hours (Supplementary Fig. S7), indicating that the IC₅₀ was about 3000–4000 μg/mL in 4T1-Luc cells.

We then verified the targetability of ErNPs@POL6326 (nanoprobes) and observed the cell uptake of such probes more intuitively, conjugating rhodamine with ErNPs@POL6326 or ErNPs@PAA via intermolecular cross-linking and electron attraction. According to quantitative confocal imaging results, the fluorescence intensity of 4T1-Luc cells increased in time- and concentration-dependent manners after incubation with Rho@ErNPs@POL6326 (Supplementary Figs. S8–S10). A rhodamine concentration of 8 μg/mL (for Rho@ErNPs@POL6326) and a 6-hour time point were then selected for later experiments. Once incubated with Rho@ErNPs@POL6326, fluorescence signaling from 4T1-Luc and MDA-MB-231-Luc cells exceeded levels conferred by Rho@ErNPs@PAA incubation, with amply expressed CXCR4 on Western blot analysis (Supplementary Fig. S11). However, when co-incubated with excess POL6326, signal intensities of 4T1-Luc and MDA-MB-231-Luc cells became weaker (Fig. 3D–F). Hence, the tumor specificity of Rho@ErNPs@POL6326 in these cell lines is likely mediated by CXCR4. NIR-IIb fluorescence imaging also revealed that fluorescence intensity in 4T1-Luc cells incubated with ErNPs@POL6326 increased over time and in accord with concentration (Supplementary Figs. S12 and S13). An Er³⁺ concentration of 10 μg/mL (for ErNPs@POL6326) and a 6-hour time point were thus selected for further experiments. In quantitative analysis of fluorescence signals from 4T1-Luc cells incubated with ErNPs@PAA or ErNPs@POL6326 (with or without POL6326), ErNPs@POL6326 produced the strongest fluorescence signals (Supplementary Fig. S14). The low fluorescence observed after ErNPs@PAA incubation implies that ErNPs@POL6326 uptake in cancer cells is mediated by CXCR4–POL6326 interaction. Intracellular fluorescence of 4T1-Luc cells that were pretreated in free POL6326 and then incubated with ErNPs@POL6326 was also significantly reduced (Supplementary Fig. S14). The targeting of 4T1-Luc cells by ErNPs@POL6326 is therefore attributable to binding of POL6326 with CXCR4. In order to investigate the targeting ability of the probes, various breast cancer cells including 4T1-Luc, MDA-MB-231-Luc, BT-549, BT-474, MCF7, SK-BR-3 cells, and MCF10A normal breast epithelial cells were used to incubate with ErNPs@POL6326. The probes were specifically enriched in all CXCR4-expressing cells, but only a small amount of probes was taken up in MCF10A cells (Supplementary Fig. S15). To further illustrate the targeting specificity of this probe, we also co-incubated it with 4T1-Luc shCXCR4 cells (CXCR4 knocked down by lentivirus shRNA) and parental 4T1-Luc cells. The results showed that the fluorescence signal intensity of 4T1-Luc cells was much higher than that of 4T1-shCXCR4 cells, suggesting its higher specificity with CXCR4 expressing cells (108.1 ± 5.6, vs. 39.0 ± 1.5, P < 0.0001, Supplementary Fig. S16).

To ascertain any acute or chronic toxic effects of ErNPs@POL6326 (nanoprobes), we conducted pertinent studies in mice that were administered the probes. Serum biochemical analysis, including liver and renal function indices (Fig. 4A; Supplementary Table S1) and blood cell counts (Fig. 4B; Supplementary Table S2), showed no significant differences in ErNPs@POL6326-exposed and control animals, suggesting that ErNPs@POL6326 nanoprobes pose no overt hepatotoxic or nephrotoxic threats and do not endanger hematopoietic function. Within a 28-day period, there were no significant group-wise weight differences (Fig. 4C), and no significant adverse events presented in these mice. H&E-stained tissue sections (Fig. 4D) revealed no obvious short-term (3 days post-injection) or long-term (4 weeks post-injection) damage or inflammatory changes in major organs. To observe the retention time/clearance duration of these nanoprobes from the murine system, the vital organs of mice injected with ErNPs@POL6326 1, 2, 3, and 4 weeks later or without probes were nitrified and the [Er³⁺] in sample was detected using ICP. The results showed that the probe was only enriched in the LNs of the injection side, but it had been cleared about 4 weeks after local injection, only 0.02% of the total amount of Er³⁺ injected remained in the injection side's LNs. However, there was almost no accumulation of Er³⁺ in other organs like liver, lung, heart and kidneys, and the difference in the Er³⁺ concentration before and after injection was not statistically significant (Supplementary Fig. S17).

To investigate the tumor targetability of ErNPs@POL6326 in living mice, ErNPs@POL6326 (nanoprobes) were injected intravenously into 4T1-Luc tumor-bearing mice, administering ErNPs@PAA as control. Fluorescence signaling in vivo at tumor sites was higher after intravenous administration of ErNPs@POL6326 than after ErNPs@PAA delivery (Fig. 5A and B). Notably, ErNPs@POL6326 recipients exhibited significantly higher tumor-to-background ratios (TBR, tumor vs. surrounding normal tissue), compared with controls (ErNPs@PAA), the highest TBR in test mice being 4.22-fold greater (8.403 vs. 1.093) and outpacing control TBRs (1.989 vs. 0.1834) at 36 hours (Fig. 5C). A blocking study was done to confirm that in vivo tumor accumulation reflected specific ligand/receptor interaction. The addition of free POL6326 significantly reduced probe uptake after injection of ErNPs@POL6326 (Fig. 5A and C). These findings affirm that enrichment of ErNPs@POL6326 in cancerous tissues is due to specific targeting of POL6326 peptide, rendering the boundary between tumor and normal tissue clearer. The biodistribution of ErNPs@POL6326 also disclosed that nanoparticles largely accumulated in liver, spleen, lung, and tumor. Both fluorescence images and fluorescence intensities of isolated tissues support the premise that ErNPs@POL6326 (vs. ErNPs@PAA) is more effectively aggregated in tumor (Supplementary Fig. S18). Hence, ErNPs@POL6326 (nanoprobes) are potentially useful for detection of breast cancer.

After successfully establishing and verifying that ErNPs@POL6326 specifically target CXCR4-overexpressing breast cancer cells in subcutaneous transplants, we then subjected our mouse model of SLN breast cancer metastasis to NIR-IIb imaging. This model was created by injecting luciferase-transfected 4T1 (4T1-Luc) cells subcutaneously into left hind paws and monitoring thereafter (via bioluminescence imaging) until tumor migration to SLNs was evident (Supplementary Fig. S19). In vivo NIR-IIb fluorescence imaging served to document the targeting specificity of ErNPs@POL6326 in mice with SLN metastasis. To do so, each animal was placed prone (positioned for SLN imaging), injecting both footpads with ErNPs@POL6326 or ErNPs@PAA. Compared with ErNPs@PAA, enrichment of SLN metastases (left paws) with ErNPs@POL6326 was far greater, emitting fluorescence more intensely (Fig. 6A–C). The specific targeting of tumor cells by ErNPs@POL6326 (nanoprobes) was thereby demonstrated. Likewise, fluorescence images of SLNs (Fig. 6A, D, and E) confirmed higher signal intensities in nodal metastases 72 hours after ErNPs@POL6326 injection, compared with tumor-free SLNs. These results suggest that ErNPs@POL6326 (nanoprobes) do enter SLNs and specifically target metastatic 4T1-Luc tumor cells. In living mice, the strong fluorescence intensities emitted helped distinguish tumor-bearing and normal SLNs. SLN metastasis was additionally corroborated in H&E-stained tissue sections (Fig. 6D and E). We also performed NIR-IIb fluorescence LN imaging in other bodily positions, allowing visualization of sacral (SaLN; Fig. 6D) and inguinal (InLN; Fig. 6E) nodal metastases. Corresponding SBR (Fig. 6F and G) were much higher than those of normal LNs. We determined the mean SBR (M ± SD, 1.21 ± 0.79) value in nonmetastatic LNs (control lower limb) after injecting ErNPs@POL6326 through foot-pad. The SBR above 2 was used as a threshold value for “detectable” signal and this arbitrarily selected for following experiments. In analyzing 40 LNs from BALB/c mice bearing 4T1-Luc cells (including 13 true positives, 1 false positive, 25 true negatives, and 1 false negative), sensitivity and specificity were 92.86% and 96.15%, respectively.

To further establish that ErNPs@POL6326 (nanoprobes) will effectively identify SLN metastases of human breast cancer, a model incorporating MDA-MB-231-Luc tumor cells was separately tested. ErNPs@POL6326 proved capable of detecting SLN metastases, as well as serial metastases from breast tumors to LNs (Supplementary Figs. S20 and S21). We ultimately harvested 66 LNs, including 14 true positives, 2 false positives nodes, 49 true negatives, and 1 false negative. Sensitivity and specificity were 93.33% and 96.08%, respectively. These results imply that the ErNPs@POL6326 nanoprobes may be used to image LNs bearing human breast cancer metastases with high sensitivity and specificity.

To further confirm that the probe can specifically recognize metastatic LNs by binding to the CXCR4 receptor on the tumor surface. Mouse models of breast cancer LN metastasis with or without CXCR4 knockdown were constructed. The results (Supplementary Figs. S22–S24) showed that the fluorescence signal of the normal group was significantly stronger than that of the knockdown group, which was consistent with the results of IHC, indicating that the fluorescence intensity of the probe was positively correlated with the expression of CXCR4.

Compelled by the overwhelmingly positive observations on nodal metastases, the capability of ErNPs@POL6326 nanoprobes to detect early-phase LN metastasis was then evaluated at 3, 6, 10, 14, and 21 days in BALB/c mice, whose footpads were injected with 4T1-Luc cells. NIR-IIb fluorescence imaging of these animals showed obvious signaling at 6 days (Fig. 7A), although H&E-stained sections proved difficult to thoroughly assess for nodal spread (Fig. 7B). Surprisingly, results of NIR-IIb fluorescence imaging matched well with findings of pan-CK immunostaining, the most commonly used ancillary technique in this context (Fig. 7A and C; Supplementary Fig. S25A–S25E). Using IHC results as a standard, we discovered that ErNPs@POL6326 (nanoprobes) successfully identify both macro- and micrometastases in LNs, achieving an overall accuracy (macro- and micrometastases included) of 96.18% (126/131). Results of ex vivo fluorescence imaging (Supplementary Fig. S25A) also matched in vivo fluorescence imaging results (Fig. 7A); and it was readily apparent that micrometastases as small as 0.5 mm are detectable this way, extending the screening limit of current clinical imaging modalities (Supplementary Fig. S25F and S25G). Under the same conditions, we carried out a parallel experiment, again confirming that ErNPs@POL6326 do identify macro- and micrometastases in SLNs and in other LNs (Supplementary Fig. S26; Supplementary Table S3). In aggregate, there is ample evidence of the excellent optical properties inherent in ErNPs@POL6326 (nanoprobes), affording high-resolution NIR-IIb fluorescence images of nodal metastases.

Owing to recent (and dramatic) technologic advances, molecular imaging techniques are now used to visualize breast cancer LN metastases in vivo, before and during surgery. Fluorescence imaging is the prevailing method (45), with FDA-approved dyes, such as ICG and methylene blue, being widely applied in SLN mapping (12). However, there are a number of issues, such as serious photobleaching and low penetration depth, that limit the clinical usage of NIR-I imaging dyes, particularly in nodal assessments (29).

ICG, a small molecule dye with a hydrodynamic of novel nm (46, 47) flow out LNs very quickly within nearly 3 hours. Thus, ICG is suitable for identifying LNs themselves, and could not determine whether LNs were metastatic or not. In comparison, the ErNPs@POL6326 particle is about 20 nm in diameter, and could remain in LNs with metastases for more than 48 hours. Thus, ErNPs@POL6326 shows longer staying in LNs especially when they were metastasized.

Rare earth nanoparticles, on the other hand, have narrow-band emissions and high photostability, producing high-quality imaging due to superior penetration and immutability (48, 49).

For this investigation, we successfully developed ErNP-based NIR-IIb nanoprobes showing deep tissue penetration and optical stability. Compared with ICG NIR-I counterparts, the images generated are of much higher resolution when visualizing blood vessels, given their smaller apparent diameters/widths and higher SBRs. LNs are usually the first line of distant spread, and such involvement bodes poorly for patient survival (50, 51). Precise detection and proper characterization of nodal metastases are critical steps in patients with breast cancer, essential for accurate pathologic staging, recurrence risk assessment, treatment planning, and prognostic optimization (52). Although more commonly used imaging methods, including magnetic resonance studies, ultrasound, and computed tomography, have been invoked for LN assessments, their sensitivities and spatial resolutions have proven inadequate for this purpose (45).

At present, SLNB is the gold standard for clinical axillary staging of early breast cancers, the labeling of SLNs often achieved by blue dye or a radioactive nanocolloid (53). However, surgeons are still deprived of outcomes in real time, having to rely on histologic examinations (12, 54). In previous studies re-evaluating initially determined node-negative breast cancer status (based on H&E slide preparations), 9% to 32% of cases actually harbored micrometastases and isolated tumor cells (ITC). In the American Joint Committee on Cancer (AJCC) Staging Manual (8th edition; ref. 55), micrometastases (MM) are defined as tumor deposits from 0.2 to 2.0 mm in size. Unlike micrometastases, ITCs are defined as single tumor cells or small clusters of cells that are below 0.2 mm in size and usually show no histologic evidence of malignant activity. Micrometastases and ITC were ultimately documented using additional histologic methods (e.g., CK19 immunostaining; ref. 56). Such discrepancy underscores the potential for misjudgment of axillary node status through current practices, precluding the option of postoperative adjuvant therapy. Improved histologic accuracy is clearly welcomed in this setting, but the capacity for accurate and real-time intraoperative identification of nodal metastasis, regardless of scale, is an even greater asset.

To this end, the prospect of a probe that will target metastatic disease at the molecular level is particularly meaningful. The C-X-C motif chemokine receptor 4 (CXCR4) has been shown to be upregulated in a number of cancerous growths, making it an excellent target for molecular imaging (57). Indeed, preclinical studies have shown a close relation between CXCR4 expression level and breast cancer invasion or distant spread (31–35). There is a greater risk of distant metastasis attached to CXCR4 overexpression (31, 32). Involvement of regional LNs, lungs, and bone marrow is also typical of breast cancer cells; and because CXCR4 expression is detected at high levels in nodal metastases (36), this is a legitimate marker for screening procedures. Herein, we conjugated ErNPs with POL6326 (a CXCR4 antagonist) to generate novel NIR-IIb ErNPs as fluorescent probes. Compared with nanoprobes that lack such modification, POL6326-functionalized ErNPs achieve much higher TBRs for in vivo tumor imaging, demonstrating superb tumor targeting specificity.

As a unique property, we have further proven that ErNPs@POL6326 not only establish LN status (tumor +/−) with great sensitivity and specificity, they also enable highly sensitive detection of macro- or micrometastases. Using these innovative functionalized rare earth-based ErNPs@POL6326 nanoprobes, accurate visualization of LN metastasis may be entirely feasible in instances of early-stage breast cancer, avoiding second axillary operations. This approach thereby promises to maximize benefits of early tumor staging and treatment decision-making.

Although the capacity to identify LN metastasis as above is generally quite reasonable, there are obstacles to clinical implementation of rare-earth nanomaterials. To date, we have demonstrated the safety of ErNPs@POL6326 (nanoprobes) in vitro and in vivo. Local confinement in LN tracing predictably may curb potential toxicity as well. Nevertheless, further research on the safety/toxicity of ErNPs@POL6326 use in primates is needed to ensure eventual mainstream medical implementation. Moreover, the quantum yield of the probe is still not high enough compared with that of quantum dots, and further optimization of the probe is needed. It is perhaps also possible to reduce toxicity by improving quantum yields of ErNPs@POL6326. For now, our efforts remain prototypic, with the intent of clinically integrating ErNPs@POL6326 to achieve in vivo LN screening in the near future.

G. Zhang reports grants from National Natural Science Foundation Committee and Fujian Major Scientific and Technological Special Project for Social Development during the conduct of the study. G. Zhang also has a patent for A molecular targeted probe for non-invasive evaluation of metastatic SLNs in breast cancer pending. No disclosures were reported by the other authors.

Y.-Y. Zhu: Data curation, methodology, writing–original draft. L. Song: Resources, methodology. Y.-Q. Zhang: Data curation, investigation, writing–original draft. W.-L. Liu: Formal analysis, investigation. W.-L. Chen: Formal analysis, writing–original draft. W.-L. Gao: Formal analysis. L.-X. Zhang: Data curation. J.-Z. Wang: Investigation. Z. Ming: Investigation. Y. Zhang: Conceptualization, project administration. G.-J. Zhang: Conceptualization, writing–original draft, project administration, writing–review and editing.

Thanks to Dr. Wen-he Huang, Department of Breast-Thyroid-Surgery and Cancer Center, Xiang'an Hospital of Xiamen University, School of Medicine, Xiamen University, 2000 East Xiang'an Road, Xiamen 361100, China, for assisting animal model building.Thanks to Dr. Rong-hui Li, Department of Medical Oncology, Xiang'an Hospital of Xiamen, No. 2000 East Xiang'an Road, Xiamen, 361100, China, for assisting lentiviral transduction. Thanks to Prof. Xiao-long Wei and Dr. Yun-Zhu Zeng, Department of Pathology, Shantou University Cancer Hospital, for assisting pathological analysis. This work was supported by the National Natural Science Foundation of China (No. 32171363, 62105333); Fujian Major Scientific and Technological Special Project for Social Development (No.2020YZ016002); Xiamen's Key Laboratory of Precision Medicine for Endocrine-Related Cancers; Fujian Key Laboratory of Precision Diagnosis and Treatment in Breast Cancer; and the Key Research Program of the Chinese Academy of Sciences (No. ZDRW-CN-2021–3).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.