Near-infrared II fluorescence imaging of PD-L1 and VEGF for assessing and early prediction of “T+A” therapeutic responsiveness in hepatocellular carcinoma Authors: Jukai Feng, Yuting Zhang, Wenhan Liu, Pu Zhang, Jianbin Xiao, Qiang Feng, Zhiyu Wang, Jinyu Zhang, Yang Zhong, Zhongying Shi, Meiqi Pan, Xiaolong Zhang, Jingfeng Liu, Jianmin Wang Published: 2026 ABSTRACT The combination of atezolizumab (T, anti-PD-L1) and bevacizumab (A, anti-VEGF) (“T + A”) are recommended as first line treatment in patients with unresectable hepatocellular carcinoma (HCC). However, there is a lack of reliable methods to accurately screen beneficiary populations and predict therapeutic outcomes for “T + A” treatment. In this study, we developed two antibody-labeled quantum dot (QD)-based nanoprobes with non-overlapping near-infrared II (NIR-II) fluorescence for simultaneous in vivo imaging of VEGF and PD-L1 levels within tumors. These nanoprobes utilized NIR-II QDs (emission: ∼950 nm and 1550 nm) as cores, while coating with 4-Arm-PEG-NH2 as a linker to accomplish highly efficient conjugation of anti-VEGF or anti-PD-L1 antibodies via amide condensation. By using tumor-bearing mouse model, we confirmed positive correlation between the in vivo fluorescence signal and the corresponding protein expression level of tumors as determined by immunohistochemistry. Importantly, our results demonstrated that high fluorescence signal of VEGF and PD-L1 levels within tumors indicated more immunosuppressive microenvironment and was associated with faster tumor progression. At meanwhile, fluorescent imaging for VEGF and PD-L1 levels before treatment could predict the endpoint of therapeutic response to “T + A” regimen. Indeed, high fluorescence signal suggested a better therapeutic response versus that with low fluorescence. Overall, our findings established an in vivo dual-targets fluorescent molecular imaging technique as a promising clinical decision-making tool that can predict response to “T + A” regimen in a minimally invasive manner prior to treatment initiation, therapy enabling personalized therapy selection for patients with advanced HCC. Introduction Hepatocellular carcinoma (HCC) remains a critical global health challenge, characterized by high mortality rates and frequent unresectability in advanced stages [1,2]. In recent years, the atezolizumab (“T”, a programmed death-ligand 1 [PD-L1] antibody) plus bevacizumab (“A”, a vascular endothelial growth factor [VEGF] antibody) (“T + A”) regimen has revolutionized the first-line treatment landscape for unresectable HCC [3,4]. The success of this strategy stems from reversal of two important and intrinsic immune-evasion pathways. The anti-PD-L1 blocks the PD-1/PD-L1 checkpoint to reverse tumor-mediated T-cell dysfunction and reinvigorate antitumor immunity, while the anti-VEGF promotes vascular normalization and alleviates immunosuppression, thereby eliciting a robust synergetic anti-tumor response [[5], [6], [7]]. Nevertheless, the objective response rate (ORR) remains observed in less than 30%, underscoring the necessity of early and effective differentiation between responders and non-responders to guide timely treatment decisions [8]. Conventionally, the assessment of PD-L1/VEGF expression in both clinical and preclinical studies relies primarily on immunohistochemistry (IHC) or genomic testing performed on ex vivo biopsy samples. However, these approaches are invasive and suffer from limitations such as tissue heterogeneity and sampling bias, which preclude monitoring the dynamic spatiotemporal expression changes in situ and in vivo [9,10]. Consequently, there is an urgent demand to develop minimally invasive strategies to accurately quantify intratumoral PD-L1 and VEGF expressions in vivo. Such strategies would enable precise patient stratification and reliable prediction of “T + A” therapeutic responsiveness at early treatment process. Fluorescence molecular imaging offers essential support for real-time, precise and minimally invasive visualization of important biomolecules, physiological activities or structural information in vivo [[11], [12], [13], [14]]. Especially, the development of fluorescence imaging in second near-infrared (NIR-II) window endows a suppressed tissue photon scattering and autofluorescence, which brings the improved spatial resolution and penetration depth for effective in vivo imaging [[15], [16], [17], [18], [19], [20], [21]]. In contrast to single-timepoint biopsies, fluorescence probes circumvent sampling bias by imaging the entire tumor and capturing intratumoral heterogeneity, and allow serial, minimally invasive monitoring of dynamic biomarker changes impossible with repeat biopsies [22,23]. Furthermore, fluorescence imaging with abundant probes and non-ionizing radiation, and enables multiparametric and multiplexing simultaneously of multiple interesting targets in vivo imaging [24,25]. So far, various PD-L1- or VEGF-targeted fluorescence probes (such as BAP+αPD-L1TCO [26], NIR830-bevacizumab-IONPs [27], etc) have been reported to visualize and quantify their intratumoral expression. However, recent advancements have predominantly been confined to the isolated monitoring of a single target, either PD-L1 or VEGF, thereby limiting the accuracy of predictive models for “T + A” regimen. Therefore, developing a novel strategy capable of simultaneously imaging intratumoral PD-L1 and VEGF levels in vivo may provide more information on tumor microenvironment (TME) and predict the efficacy of immunotherapy. Quantum dots (QDs) with NIR-II emission represent an effective fluorophore for in vivo molecular imaging. They exhibit significant advantages such as high quantum yield, superior photo-stability and size-tunable emission [23,25,28]. Their emission can be finely tuned on demand across NIR-II region by adjusting the crystallite size, generating non-overlapping signals for simultaneous multiplexed imaging of different biomarkers [25,29,30]. Moreover, the surface functionalization of NIR-II QDs can be used for targeting molecular imaging [31,32]. Thus, emission-tunable QDs allow for multicolor imaging of PD-L1 and VEGF levels with high accuracy and reliability, further guiding and optimizing “T + A” therapy. Herein, we developed a dual-channel NIR-II imaging strategy based on lead sulfide (core)@lead cadmium (shell) (PbS@CdS) QDs with non-overlapping emissions (Scheme 1). Specifically, two kinds of NIR-IIa PbS@CdS QDs (emission peak ∼950 nm, denoted as IIa QD) and NIR-IIb PbS@CdS QDs (emission peak ∼1550 nm, denoted as IIb QD) were conjugated with VEGF and PD-L1 monoclonal antibodies, respectively. The obtained nanoprobes (Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1) with distinguishable emissions were subsequently administered as an equimolar mixture. In HCC mouse models, the mixed nanoprobe formulation demonstrated minimal spectral crosstalk and enabled high-contrast dual-color imaging of intratumoral PD-L1 and VEGF levels. Crucially, the fluorescence intensities of the nanoprobes exhibited a significant correlation with IHC-validated protein expression levels of PD-L1 and VEGF. Notably, our findings revealed that high fluorescence intensity indicated a more immunosuppressive microenvironment and a more promising response to “T + A” treatment in HCC. This strategy allows early imaging-based prediction of response to “T + A” therapy, providing a minimally invasive methodological advance for stratifying individuals and predicting therapeutic efficacy.Scheme 1Illustration of in vivo two-plex NIR- II fluorescence molecular imaging for intratumoral PD-L1 and VEGF using QD-based nanoprobes with non-overlapping emissions. (a) Illustration of the preparation of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. (b) Prediction of “T + A” therapeutic efficacy in an HCC mouse model through in vivo imaging of the intratumoral PD-L1 and VEGF levels. LP, long-pass; SP, short-pass. Illustration of in vivo two-plex NIR- II fluorescence molecular imaging for intratumoral PD-L1 and VEGF using QD-based nanoprobes with non-overlapping emissions. (a) Illustration of the preparation of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. (b) Prediction of “T + A” therapeutic efficacy in an HCC mouse model through in vivo imaging of the intratumoral PD-L1 and VEGF levels. LP, long-pass; SP, short-pass. Materials and methods Materials Sulfur powder was procured from Alfa Aesar (Shanghai, China). Lead chloride (PbCl2), cadmium oxide (CdO), oleylamine, oleic acid (OA), and 1-octadecene (ODE) were purchased from Sigma-Aldrich (StLouis, USA). DSPE-PEG 2000 and DSPE-PEG-COOH (MW: 2000) were obtained from Ruixi Biological Technology (Xi'an, China) and Pengshuo Biological Technology (Shanghai, China), respectively. Four-arm polyethylene glycol amine (4-Arm-PEG-NH2, MW 5000) was provided by Aladdin (California, USA). Anti-mouse VEGF-A (clone 2G11-2A05) and anti-mouse PD-L1 antibody (clone 10 F.9G2) were purchased from Bio X Cell (New Hampshire, USA) and Leinco Technologies (StLouis, USA), respectively. All other chemicals were analytical grade and used as received. Synthesis of NIR-IIa/IIb PbS@CdS QDs Establishing an argon (Ar) atmosphere: After assembling the reaction setup, three vacuum-refill cycles were conducted, each consisting of evacuation (≥ 10 min) followed by argon refilling (≥5 min). During the final cycle, a steady, slow bubbles in the bubbler was maintained to ensure complete removal of air from the system (Supplementary Fig. 1a). Sulfur precursor: The sulfur powder (0.08 g) and oleylamine (7.5 mL) were mixed in a two-neck flask and heated to 120 °C under Ar for 30 min. After cooling to room temperature (RT), the transparent, dark brown sulfur precursor solution was obtained (Supplementary Fig. 1b). Lead precursor: The PbCl2 (0.834 g) and oleylamine (7.5 mL) were mixed in a three-neck flask and heated to 120 °C under Ar, yielding a white turbid mixture. Subsequently, the OA (2.5 mL) was quickly introduced, which transformed the mixture into a clear, transparent yellow solution. After reacting for 30 min, the pale-yellow lead precursor was obtained (Supplementary Fig. 1c). Synthesis of NIR-IIa/IIb PbS QDs (Core): The PbS QDs were synthesized with a modified hot-injection procedure [29,[33], [34], [35]]. For the synthesis of NIR-IIa PbS QDs, 2.25 mL of sulfur precursor was injected into the lead precursor at 45 °C and allowed to react for 30 min. The reaction was then quenched using pre-cooled hexane/ethanol (10 mL/20 mL) in an ice bath. The products were collected by centrifugation at 10,000 rpm for 10 min, resuspended in hexane/oleic acid (10 mL/20 mL) and shaken for 30 min to remove excess sulfur. After shaking, the mixture was centrifuged again, and this washing procedure was repeated 3 times until the supernatant became colorless (Supplementary Fig. 1d). Finally, the purified PbS QDs were dispersed in 10 mL of ODE for subsequent use. For NIR-IIb PbS QDs, the synthesis followed the same procedural steps, except that the reaction was carried out at an elevated temperature of 160 °C for 30 min. Synthesis of NIR-IIa/IIb PbS@CdS QDs (Core-shell): The PbS/CdS QDs were synthesized via a modified cation-exchange procedure [29,[33], [34], [35]]. For NIR-IIa PbS@CdS QDs (IIa QD), the clear, transparent yellow Cd precursor was prepared by heating a mixture of CdO (0.6 g), OA (4 mL), and ODE (15 mL) at 200 °C under Ar for 30 min and then cooled down to 100 °C (Supplementary Fig. 1e). The previously prepared IIa PbS QDs (5 mL) were injected into the Cd precursor and maintained at 100 °C for 100 min to facilitate shell growth. The reaction was quenched by adding 5 mL of pre-cooled hexane. The resulting QDs were purified via centrifugation and repeated ethanol washes, and finally dispersed in 10 mL of hexane for storage (Supplementary Fig. 1f). For NIR-IIb PbS@CdS QDs (IIb QD), Cd precursor was prepared by mixing CdO (1.2 g), OA (8 mL), and ODE (20 mL) at 200 °C for 90 min. And the reaction time of IIb PbS QDs and Cd precursor was 60 min. All other procedures were the same as above. Synthesis of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes Synthesis of aqueous-dispersible IIa/IIb QD: To transfer hydrophobic QD into the aqueous phase, DSPE-PEG 2000 (2.5 mg) and DSPE-PEG-COOH (MW: 2000, 2.5 mg) were dissolved in anhydrous tetrahydrofuran (THF, 20 μL). IIa QD (or IIb) QD (1 mg) were dissolved in 60 μL of THF via ultrasonication. The above two solutions were then mixed and ultrasonicated for 15 min, then was rapidly injected into 15 mL of deionized water (ddH2O) under ultrasonication for another 15 min. The aqueous-dispersible IIa QD (or IIb) QD were obtained by slowly evaporating THF in a fume cupboard, followed by purification through ultrafiltration (3500 rpm, 10 min). Conjugation of IIa/IIb QD and antibody: A mixture of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC, 2 mg) and N-hydroxysuccinimide (NHS, 2 mg) in ddH2O (0.2 mL) was added to the aqueous-dispersible IIa (or IIb) QDs and incubated at RT for 30 min to activate the carboxyl groups. Subsequently, 4-Arm-PEG-NH2 (5 mg) was introduced and the reaction proceeded under stirring for 3 h at RT to allow amidation. The resulting 4-Arm-PEG-NH2-modified IIa (or IIb) QDs were purified by ultrafiltration. In parallel, the monoclonal antibody (αVEGF for IIa QDs or αPD-L1 for IIb QDs, 80 μg) was activated in a mixture of EDC (2 mg) and NHS (2 mg) in ddH2O (0.2 mL) for 3 h at RT. The activated antibody was then incubated with the corresponding 4-Arm-PEG-NH2-modified QDs for 3 h to enable covalent conjugation via amide bond formation. Final purification via ultrafiltration yielded the targeted bioconjugates IIa QD@αVEGF or IIb QD@αPD-L1. Physicochemical characterization of nanoprobes Transmission electron microscope (TEM) HT7700 (Hitachi, Japan) was employed to analyze the size and morphology of QDs. The elemental analysis of Pb, S, and Cd was analyzed by Talos F200X G2 Scanning Transmission Electron Microscope (Thermofisher Scientific, USA). Dynamic light scattering (DLS) measurements were executed on a Zetasizer Nano ZS90 equipment (Malvern Instruments, Southborough, MA). Optical properties were characterized using an FLS 980 spectrofluorometer (Edinburgh Instruments, U.K.), a SpectraMax M5 microplate reader (Molecular Devices, USA), a Lambda 1050 spectrophotometer (PerkinElmer, USA) and a UV-3600 spectrophotometer (Shimadzu, Japan). To assess the stability of nanoprobes, the particles were dispersed in water or RPMI-1640 culture medium supplemented with 10% FBS stored at 4 °C. The fluorescence intensity, hydrodynamic diameter and polydispersity dispersion index (PDI) were monitored daily for 7 days. Additionally, nanoprobes were irradiated with a continuous 808 nm laser exposure (50 mW/cm2) with fluorescence imaging recorded at 0, 5, 10, 20, 30, 40, 50, and 60 min, to evaluate their photostability. In vitro and in vivo NIR-II imaging of nanoprobes All in vitro and in vivo NIR-II imaging was performed using an NIR-II in vivo imaging system (Artemis Intelligent Imaging, Shanghai, China) equipped with an 808 nm laser. A 900 nm long-pass filter (LP) and a 1200 nm short-pass filter (SP) were used for NIR-IIa channel (IIa QD@αVEGF), while a 1500 nm LP was applied for NIR-IIb channel (IIb QD@αPD-L1). The laser power was set at 50 mW/cm2. The fluorescence intensity was measured by ImageJ software. For in vivo NIR-II imaging, mice were intravenously injected an equimolar mixture of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes (total dose: 1 mg/mouse in 200 μL PBS). All animal procedures were approved by Animal Ethics Committee of Mengchao Hepatobiliary Hospital of Fujian Medical University (MCHH-AEC-2023-13) and conducted in accordance with relevant guidelines. Ascites propagation and establishment of tumor model of H22 cells in mice Hepatoma 22 (H22) cells were first expanded in vitro to logarithmic growth phase, and then prepared as a cell suspension in phosphate-buffered saline (PBS). The female BALB/c mice (Shanghai SLAC Laboratory Animal Co., Ltd.) were intraperitoneally injected with 200 μL of cell suspension containing 1.0 × 107 to 2.0 × 107 cells. After 5–7 days, mice were euthanized upon observation of significant ascites accumulation. The abdominal surface was disinfected, and ascitic fluid was collected aseptically. The harvested fluid was centrifuged at 800 ×g for 3 min at 4 °C, and the supernatant was discarded. The pellet was resuspended in 1–3 mL red blood cell lysis buffer for 10–15 min, and lysis was terminated by adding excess PBS. After repeated centrifugation (800 ×g, 3 min, 4 °C), the resulting cells were resuspended in PBS at a concentration of 2 × 107 cells/mL. To establish an H22 tumor model, 50 μL of H22 cell suspension was injected subcutaneously into the right dorsal flank of BALB/c mice. Palpable tumors developed within 5–10 days post-inoculation, confirming successful establishment of H22 tumor model. In vivo biodistribution and metabolism The in vivo visualization of blood vessels was evaluated in anesthetized BALB/c nude mice following tail intravenous injection of the mixed nanoprobes. At 5 min post-injection, NIR-II imaging was performed with exposure times of 300 ms for IIa QD@αVEGF and 900 ms for IIb QD@αPD-L1. To assess the biodistribution and metabolism of the nanoprobes, healthy BALB/c mice received an identical injection of the mixed nanoprobes. Whole-body NIR-II fluorescence imaging was conducted at designated time points with the exposure time settings specified above. Mice were euthanized via cervical dislocation under anesthesia 72 h post-injection, and major organs (heart, liver, spleen, lungs, kidneys) were harvested for ex vivo NIR-II fluorescence imaging to quantify nanoprobe biodistribution. Meanwhile, the in vivo metabolic profiles of the nanoprobes were evaluated by collecting urine and feces from mice housed in metabolic cages for 0.5–3 h at designated time points over 25 days. Mice were then returned to conventional cages between collections to minimize physiological stress. Feces samples were ground into powder, and both powdered feces and urine were digested in an acid mixture (nitric acid: hydrofluoric acid: hydrogen peroxide: hydrochloric acid = 62.5: 12.5: 12.5: 12.5) at 180 °C for 8 h. After cooling, Pb and Cd contents in the digested solutions were quantified by inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7700, USA) or inductively coupled plasma optical emission spectrometry (ICP-OES, Agilent ICP OES730, USA). The excretion rate was calculated as: Excretion rate (%ID/hour) = [metal content in sample/total injected metal dose/collection times (h)] × 100%. Cytotoxicity assessment The cytotoxicity of a mixed nanoprobes (IIa QD@αVEGF: IIb QD@αPD-L1 = 1:1) was evaluated using the Cell Counting Kit-8 (CCK-8) assay. Briefly, cells (Hepa 1-6 and CL2: 1 × 104 cells/well; H22: 3 × 104 cells/well) were seeded into 96-well plates. After 24 h, the medium was replaced with fresh medium (Hepa 1-6 and CL2: DMEM medium, H22: 1640 medium) containing different concentrations of mixed nanoprobes (0, 2.5, 5, 10, 20, 40, and 80 μg/mL) for another 12 h. Afterward, the medium was removed, and 100 μL of fresh medium containing 10% CCK-8 solution (v/v) was added to each well. The optical density (OD) at 450 nm was measured using a SpectraMax M5 microplate reader (Molecular Devices, USA). Cell viability was calculated as: Cell viability (%) = [(ODtreated-ODblank)/(ODcontrol-ODblank)] × 100%, where the ODblank was the absorbance value of CCK-8 medium solution, ODtreated is the absorbance value of nanoprobes-treated cells, and ODcontrol is the absorbance value of cells without treatment. Untreated cells were designated as 100% viability. Experiments were performed in triplicate for Hepa 1-6 and H22 cells, and in quintuplicate for CL2 cells. Hemolysis assay To assess the blood compatibility of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes, a hemolysis assay was performed. Fresh mouse blood was collected by retro-orbital bleeding, and red blood cells (RBCs) were then isolated by centrifugation at 1000 ×g for 10 min. Next, 200 μL of RBCs were incubated with 200 μL of an equimolar mixture of the two nanoprobes at different concentrations (0.025, 0.05, 0.125, 0.25, 0.5, 1.25, 2.5 mg/mL) for 4 h at 37 °C. After incubation, the samples were centrifuged, and the absorbance of supernatant was measured at 576 nm. For controls, 200 μL of RBCs were mixed with 200 μL of ddH2O (positive) or 200 μL of saline (negative). Three replicates were set for each concentration gradient. Isolation and culture of CD8+ T cells CD8+ T cells were isolated from mice using the Mouse CD8+ T Cell Isolation Kit (Vazyme, China) following the manufacturer's instructions. Freshly isolated CD8+ T cells were co-cultured with H22 cells at a 10:1 ratio under different antibody treatments (isotype IgG, anti-VEGFA, anti-PD-L1, or anti-VEGFA plus anti-PD-L1). Cultures were maintained in RPMI-1640 medium supplemented with 10% FBS, 1% penicillin-streptomycin, 20 ng/mL IL-2 (575,402, Biolegend), 2 μg/mL anti-CD3 (A2104, Selleck) and 2 μg/mL anti-CD28 (102,116, Biolegend). After 72 h, supernatant levels of IL-10, TGF-β and IFN-γ was quantified by ELISA (IL-10: SLBA056Mu, Lablead; TGF-β: EK0515, Boster Biological Technology; IFN-γ: HSEK0375, Boster Biological Technology). For proliferation assay, isolated CD8+ T cells were labeled with CFSE (423,801, BioLegend) before co-culture and analyzed by flow cytometric. For activation assessment, cells were co-cultured for 48 h, then stained with anti-CD8 and anti-CD69, and analyzed by flow cytometry. In vivo NIR-II fluorescence imaging on H22-tumor model H22 tumor-bearing mice (n = 3) received intravenous injections of an equimolar of mixture nanoprobes (1 mg/mouse). Whole-body NIR-II fluorescence imaging was performed at 1, 3, 6, 9 and 12 h post-injection. Fluorescence colocalization and signal intensity kinetics were analyzed across all time points by ImageJ software. For immunohistochemistry (IHC) analysis, small portions of tumors were resected at 1 h post-injection, and the collected tissues were processed for IHC staining of VEGF and PD-L1. Quantitative analysis of IHC images was performed using ImageJ software. Briefly, images acquired under consistent imaging parameters were imported, and the DAB-positive signal channel was separated via color deconvolution. A threshold was applied to identify tissue regions and exclude blank background. The mean optical density of the DAB signal within tissue areas was then determined and used as the quantitative index to objectively assess target protein expression levels. In vivo biosafety evaluation Female BALB/c mice (6–8 weeks old) were intravenously injected with PBS (control group, n = 5) or the equimolar mixture of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes (total dose: 1 mg/mouse in 200 μL PBS; Mixed QD group, n = 5). Body weight of mice was recorded every 2 days. On day 14, their blood was collected to analyze the serum biochemistry. The analysis included the following parameters: total bilirubin (TBIL), alanine aminotransferase (ALT), albumin (ALB), alkaline phosphatase (ALP), gamma-glutamyl transferase (γ-GT), total bile acids (TBA), urea (UREA), creatinine (CREA), and uric acid (UA). Additionally, major organs (heart, liver, spleen, lungs, and kidneys) were harvested, fixed in 4% paraformaldehyde, processed into paraffin sections, and subjected to hematoxylin-eosin (H&E) staining for histological examination. Dual-channel NIR-II imaging-guided stratification and evaluation of T + A therapy Female BALB/c mice (6–8 weeks old) were subcutaneously inoculated with 1 × 106 H22 cells to establish an H22 tumor-bearing model. On day 8 post-inoculation, mice were intravenously injected with an equimolar mixture of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes (total dose: 1 mg/mouse in 200 μL PBS), and dual-channel NIR-II imaging was performed 1 h post-injection to quantify mean fluorescence intensity (MFI) in tumors. Subsequently, mice were stratified into High MFI and Low MFI groups using the median intratumoral MFI of IIa QD@αVEGF and IIb QD@αPD-L1 as the threshold. Then, each group was further randomized into four subgroups: High MFIControl (n = 7, no treatment), High MFIT + A (n = 7, receiving “T + A” therapy), Low MFIControl (n = 5, no treatment), and Low MFIT + A (n = 5, receiving “T + A” therapy). The “T + A” regimen consisted of anti-PD-L1 (T, 10 mg/kg) and anti-VEGFA (A, 10 mg/kg) administered intraperitoneally every 3 days for a total of 5 doses. The control mice did not receive any treatment. Throughout the treatment period, tumor volume and body weight were recorded every 3 days. Tumor volume (V) was calculated using the formula: V = a × b2/2, where “a” is the longest diameter and “b” is the shortest diameter of the tumor. When tumor volume reached 1500 mm3 in any of the mice, the experiment was terminated, and the mouse was euthanized. The tumors were then excised for immunofluorescence analysis and flow cytometry quantification of infiltrating CD8+ T cells. Statistics All data are expressed as the mean ± standard deviation (SD) with individual data points indicated. Statistical analysis was performed using two-way/one-way ANOVA for comparisons among multiple groups, or the two-tailed unpaired Student's t-test for comparisons between two groups. All statistical analyses were conducted using GraphPad Prism 10. Statistical significance was defined as p value: ∗ <0.05, ∗∗ <0.01, ∗∗∗ <0.001, and ∗∗∗∗ <0.0001. Pearson's correlation coefficient was used to assess the correlation matrices. All flow cytometry data were analyzed using FlowJo 10.9.0. For immunofluorescence imaging, CaseViewer 2.4 and ImageJ was used to observe and analyze the images. Results Synthesis and characterization of NIR-IIa/IIb PbS@CdS QDs NIR-IIa/IIb PbS@CdS QDs were constructed via a modified hot-injection and cation exchange approach [29,[33], [34], [35]], entailing the stages of monomer generation, nucleation, and nanocrystal growth (Fig. 1a). Emission peaks of the PbS QDs were tuned to NIR-IIa (∼1100 nm) and NIR-IIb (∼1600 nm) ranges, respectively, by modulating the reaction temperature and nucleation time to control Ostwald ripening-mediated particle growth (Supplementary Fig. 2). Subsequently, a CdS shell was grown onto the PbS core by cation exchange. By varying the reaction time, we obtained two types of PbS@CdS QDs (denoted “IIa QD” and “IIb QD”). Transmission electron microscopy (TEM) images revealed that both the IIa QD and IIb QD exhibited monodisperse spherical morphologies. Statistical quantification further confirmed their narrow size distribution. Notably, the average diameter of IIb QD (6.79 ± 0.84 nm) was larger than that of IIa QD (4.69 ± 1.18 nm), resulting in a longer emission peak for IIb QD [29] (Fig. 1b). Meanwhile, energy-dispersive X-ray spectroscopy (EDS) and elemental mapping confirmed the homogeneous distributions of Pb, Cd, and S throughout the nanostructures (Fig. 1c), validating the successful formation of the core-shell structure.Fig. 1Synthesis and characterization of IIa and IIb QD. (a) Schematic illustration of the synthesis of IIa and IIb QD, and the modification of IIa and IIb QD with carboxyl-terminated polyethylene glycol. (b) TEM images and corresponding size distribution histograms of IIa and IIb QD dispersed in n-Hexane. The particle size was statistically analyzed using ImageJ software from TEM images. Scale bar: 100 nm. (c) High-angle annular dark field TEM (HAADF-TEM) images and energy dispersive spectroscopy (EDS) elemental mapping of IIa and IIb QD. Experiment conditions: 120 kV. Scale bar: 10 nm. (d) TEM images of IIa and IIb QD dispersed in aqueous solution. The insets show photographs of the corresponding QD solution. Scale bar: 100 nm. (e) The absorption spectrum and emission spectrum of IIa and IIb QD measured in n-Hexane or aqueous solution, respectively. The absorption spectrum was measured in organic phase because the serious absorption band of water at 1400–1500 nm influenced the measurement of absorbance in water. Synthesis and characterization of IIa and IIb QD. (a) Schematic illustration of the synthesis of IIa and IIb QD, and the modification of IIa and IIb QD with carboxyl-terminated polyethylene glycol. (b) TEM images and corresponding size distribution histograms of IIa and IIb QD dispersed in n-Hexane. The particle size was statistically analyzed using ImageJ software from TEM images. Scale bar: 100 nm. (c) High-angle annular dark field TEM (HAADF-TEM) images and energy dispersive spectroscopy (EDS) elemental mapping of IIa and IIb QD. Experiment conditions: 120 kV. Scale bar: 10 nm. (d) TEM images of IIa and IIb QD dispersed in aqueous solution. The insets show photographs of the corresponding QD solution. Scale bar: 100 nm. (e) The absorption spectrum and emission spectrum of IIa and IIb QD measured in n-Hexane or aqueous solution, respectively. The absorption spectrum was measured in organic phase because the serious absorption band of water at 1400–1500 nm influenced the measurement of absorbance in water. To improve the QDs for further in vivo application, amphiphilic DSPE-PEG and DSPE-PEG-COOH polymers were coated onto the QDs’ surfaces to grant water dispersibility and facilitate subsequent conjugation with antibodies [36](Fig. 1a). The functionalized QDs maintained good colloidal stability in aqueous media without significant aggregation (Fig. 1d), and retained strong fluorescence emission, accompanied by a shift in emission peaks to either ∼950 nm (IIa QD) or ∼1550 nm (IIb QD) compared to the original PbS QDs (Fig. 1e). Of note, a pronounced spectral separation (>600 nm) was observed between two kinds of QDs, suggesting minimal inter-channel crosstalk and highlighting their suitability for dual-channel NIR-II imaging applications. Preparation and characterization of VEGF- or PD-L1-targeted nanoprobes To achieve quantification and localization of VEGF and PD-L1, mouse anti-VEGF and anti-PD-L1 antibodies were conjugated to these two kinds of QDs, respectively, through a carbodiimide-mediated amidation reaction between carboxyl and amine groups [37], thus yielding Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. To improve the coupling efficiency, DSPE-PEG-COOH on QDs was first conjugated to tetra-branched 4-Arm-PEG-NH2, followed by conjugation of antibody carboxyl groups to amine groups on 4-Arm-PEG-NH2 (Fig. 2a and Supplementary Fig. 3a). This spacer-arm link strategy could increase reactive sites and reduce steric hindrance between QDs and antibodies to enhance conjugation yield [25]. As measured by BCA assay, the conjugation efficiencies of Ⅱa QD@αVEGF reached 68.9% with 4-arm-PEG-NH2 cross-linking, a marked increase from 32.6% without it. In parallel, Ⅱb QD@αPD-L1 exhibited an efficiency of 81.9% with cross-linking, compared to 20.5% without it, demonstrating a substantial enhancement afforded by 4-arm-PEG-NH2 (Supplementary Fig. 3b).Fig. 2Synthesis and characterization of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. (a) Schematic illustration of nanoprobes. (b) Hydrodynamic size distribution of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 in aqueous solution by DLS. (c) Zeta potential of Ⅱa QD, Ⅱa QD@αVEGF, Ⅱb QD, and Ⅱb QD@αPD-L1. (d) UV-vis absorption spectra of anti-VEGF, hydrophilic IIa QD, IIa QD@αVEGF, anti-PD-L1, hydrophilic IIb QD and IIb QD@αPD-L1. (e) Correlation between the fluorescence intensity of IIa QD@αVEGF or IIb QD@αPD-L1 and their corresponding protein concentration via a simple linear regression model (n = 6). (f) NIR-II fluorescence imaging of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes with 900 nm LP+1200 nm SP and 1500 nm LP filters under 808 nm excitation. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 50 ms; IIb QD@αPD-L1, 150 ms. (g) In vitro photostability assessment (photobleaching resistance): fluorescence intensity and images of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes in water under continuous 808 nm laser irradiation. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 50 ms; IIb QD@αPD-L1, 150 ms. (h) Fluorescence intensity stability of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes during storage at 4 °C. (i) Hydrodynamic diameter stability of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes during storage in water or RPMI-1640 culture medium supplemented with 10% FBS at 4 °C over 7 days. Data denote mean ± SD (n = 3). The solid line corresponds to Size, and the dashed line indicates PDI. Statistical analysis in (g) and (h) was performed using two-way ANOVA with Dunnett's multiple comparisons test. Ns, non-significant. Synthesis and characterization of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. (a) Schematic illustration of nanoprobes. (b) Hydrodynamic size distribution of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 in aqueous solution by DLS. (c) Zeta potential of Ⅱa QD, Ⅱa QD@αVEGF, Ⅱb QD, and Ⅱb QD@αPD-L1. (d) UV-vis absorption spectra of anti-VEGF, hydrophilic IIa QD, IIa QD@αVEGF, anti-PD-L1, hydrophilic IIb QD and IIb QD@αPD-L1. (e) Correlation between the fluorescence intensity of IIa QD@αVEGF or IIb QD@αPD-L1 and their corresponding protein concentration via a simple linear regression model (n = 6). (f) NIR-II fluorescence imaging of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes with 900 nm LP+1200 nm SP and 1500 nm LP filters under 808 nm excitation. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 50 ms; IIb QD@αPD-L1, 150 ms. (g) In vitro photostability assessment (photobleaching resistance): fluorescence intensity and images of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes in water under continuous 808 nm laser irradiation. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 50 ms; IIb QD@αPD-L1, 150 ms. (h) Fluorescence intensity stability of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes during storage at 4 °C. (i) Hydrodynamic diameter stability of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes during storage in water or RPMI-1640 culture medium supplemented with 10% FBS at 4 °C over 7 days. Data denote mean ± SD (n = 3). The solid line corresponds to Size, and the dashed line indicates PDI. Statistical analysis in (g) and (h) was performed using two-way ANOVA with Dunnett's multiple comparisons test. Ns, non-significant. After conjugation with antibodies, the hydrodynamic diameters of Ⅱa QD@αVEGF detected by dynamic light scattering (DLS) increased to 135.59 ± 2.75 nm, and their zeta potential changed from −14.9 ± 0.56 mV (Ⅱa QD) to −3.94 ± 0.36 mV. Similarly, Ⅱb QD@αPD-L1 exhibited an increase in hydrodynamic size to 114.43 ± 2.23 nm, with a corresponding alteration in zeta potential from −20.37 ± 0.91 mV (Ⅱb QD) to −4.35 ± 0.41 mV (Fig. 2b and c). The UV-vis spectra of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 exhibited a characteristic absorption peak at 260 nm, due to a blue shift in the absorption band of aromatic amino acids (Trp/Tyr) in the anti-VEGF and anti-PD-L1 antibodies [38] (Fig. 2d). Agarose gel electrophoresis revealed that the electrophoretic mobilities of both nanoprobes decreased markedly after antibody conjugation, along with retained fluorescence intensity (Supplementary Fig. 3c). Moreover, a strong positive correlation was observed between fluorescence intensity of nanoprobe and conjugated antibody concentration (Fig. 2e). These results together confirmed the successful conjugation of antibodies to QDs. Furthermore, we investigated optical images of nanoprobes using 808 nm laser excitation in combination with long-pass (LP) plus short-pass (SP) emission filters. As shown in Fig. 2f, Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 possessed excellent NIR fluorescence emission, with distinct signals under 900 nm LP+1200 nm SP (NIR-IIa) and 1500 nm LP (NIR-IIb) filters, respectively. Their distinct emission profiles demonstrated the significant spectral separation between these nanoprobes, thereby facilitating two-plex imaging in vivo with minimal crosstalk and enhancing detection accuracy in complex biological environments. Next, we evaluated the stability of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. Under continuous 808 nm laser irradiation for 1 h, both nanoprobes kept their strong brightness in aqueous solution without obvious decay, demonstrating their superior photostability attributable to synergistic stabilization between the robust QDs’ cores and protective antibody-PEG corona [39] (Fig. 2g and Supplementary Fig. 3d). Moreover, no significant variations were observed in fluorescence signal, average DLS size, and polydispersity index (PDI) of both nanoprobes storing in aqueous solution or RPMI-1640 + 10%FBS over 7 days, confirming their optical and colloidal stabilities in physiological environments (Fig. 2h and i). Together, these results underscored the excellent potential of these two nanoprobes for long-term in vivo NIR-II fluorescence imaging. Biosafety evaluation of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes in vivo Prior to in vivo imaging application, the biosafety of nanoprobes was first evaluated. Cell Counting Kit-8 (CCK-8) assays demonstrated that mixed IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes, even at concentrations up to 80 μg/mL, did not significantly affect the proliferation of HCC cells (Hepa1-6, H22) or normal cells (CL2), indicating minimal cytotoxicity (Supplementary Fig. 4a). Furthermore, no significant hemolysis was observed for the mixed nanoprobes, even at a concentration as high as 2.5 mg/mL, verifying their favorable blood biocompatibility (Supplementary Fig. 4b). Next, in vivo biosafety assessments were conducted in healthy BALB/c mice by intravenous injection of mixed nanoprobes (500 μg of each nanoprobe per mouse). During the initial 4 days post-injection, a slight loss in appetite and a mean weight reduction of 3.22 g were observed in nanoprobe-treated mice (Fig. 3a). However, body weights recovered to levels comparable to the control group by day 6, indicating transient and well-tolerated systemic toxicity. Additionally, serum biochemical analysis on day 14 revealed no significant abnormalities of renal or hepatic functional markers in nanoprobe-treated mice compared to the control group (Fig. 3b and c). Histological evaluation (H&E staining) of major organs (heart, liver, spleen, lungs, kidneys) further indicated that nanoprobes did not induce obvious inflammation, necrosis, or other pathological lesions (Fig. 3d). Collectively, these findings suggested that the mixed nanoprobes possessed favorable in vivo biosafety, with minimal long-term adverse effects after efficient clearance.Fig. 3Systemic biosafety assessment of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. (a) The body weight changes of mice over 14 days after intravenous administration of PBS or mixed nanoprobes. (b) Serum renal function markers (UREA, UA, CREA) and (c) Serum hepatic function markers (ALT, ALP, ALB, γ-GT, TBIL, TBA) at day 14 post-injection (n = 5 mice pre group). (d) Representative images of hematoxylin and eosin (H&E) of major organs harvested at day 14 post-injection. Scale bar: 200 μm or 400 μm. Full names: urea, UREA; uric acid, UA; creatinine, CREA; alanine aminotransferase, ALT; alkaline phosphatase, ALP; albumin, ALB; gamma-glutamyl transferase, γ-GT; total bilirubin, TBIL; total bile acids, TBA. Data denote mean ± SD (n = 4–5). Statistical analysis in (b) and (c) was performed using a two-tailed unpaired Student's t-test. Ns, non-significant. Systemic biosafety assessment of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 nanoprobes. (a) The body weight changes of mice over 14 days after intravenous administration of PBS or mixed nanoprobes. (b) Serum renal function markers (UREA, UA, CREA) and (c) Serum hepatic function markers (ALT, ALP, ALB, γ-GT, TBIL, TBA) at day 14 post-injection (n = 5 mice pre group). (d) Representative images of hematoxylin and eosin (H&E) of major organs harvested at day 14 post-injection. Scale bar: 200 μm or 400 μm. Full names: urea, UREA; uric acid, UA; creatinine, CREA; alanine aminotransferase, ALT; alkaline phosphatase, ALP; albumin, ALB; gamma-glutamyl transferase, γ-GT; total bilirubin, TBIL; total bile acids, TBA. Data denote mean ± SD (n = 4–5). Statistical analysis in (b) and (c) was performed using a two-tailed unpaired Student's t-test. Ns, non-significant. In vivo fate and tumor-targeting of IIa QD@αVEGF and IIb QD@αPD-L1 Next, the mixture of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes at a 1:1 mass ratio was intravenously injected into healthy mice (500 μg of each nanoprobe per mouse). Using the NIR-Ⅱ imaging setup, clear visualization of blood vessels in BALB/c nude mice was observed within 5 min post-injection of IIb QD@αPD-L1, whereas IIa QD@αVEGF produced no detectable vascular signal (Supplementary Fig. 5), in accordance with previous studies [40]. This distinction further demonstrated the non-overlapping fluorescence emission spectra of both nanoprobes. In addition, both NIR-IIa and NIR-IIb fluorescence signals in the liver of BALB/c mice rapidly accumulated within 1 h, and then gradually declined over time, indicating a hepatic clearance pathway for the nanoprobes [41,42] (Fig. 4a and b). 72 h post-injection, luminescence signals (only ∼35% FL intensity) from IIa QD@αVEGF had nearly vanished throughout the body, with only faint residual signals detected in the liver, indicating near-complete systemic clearance (Fig. 4c and d). Conversely, IIb QD@αPD-L1 exhibited pronounced retention in the liver and spleen. Notably, the signal intensity in the spleen was 1.5-fold greater than that in the liver, likely because the spleen serves as an important immune organ rich in immune cells and as a blood filter, suggesting a potential metabolic pathway mediated by splenic venous circulation [43,44].Fig. 4In vivo fate and tumor targeting of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 in healthy and H22 tumor-bearing mice. (a, b) Biodistribution assessment of the mixed nanoprobes (1:1 wt ratio of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1) following intravenous injection in healthy BALB/c mice. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. (c, d) Ex vivo imaging and corresponding fluorescence quantification of major organs harvested at 72 h post-injection in healthy BALB/c mice. (e, f) The excretion rates of Pb and Cd in mouse feces and urine at different time points after administration determined by ICP-MS and ICP-OES. (g, h) Tumor-targeting evaluation of the mixed nanoprobes following intravenous administration in H22 tumor-bearing mice. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. Data denote mean ± SD (n = 3). In vivo fate and tumor targeting of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1 in healthy and H22 tumor-bearing mice. (a, b) Biodistribution assessment of the mixed nanoprobes (1:1 wt ratio of Ⅱa QD@αVEGF and Ⅱb QD@αPD-L1) following intravenous injection in healthy BALB/c mice. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. (c, d) Ex vivo imaging and corresponding fluorescence quantification of major organs harvested at 72 h post-injection in healthy BALB/c mice. (e, f) The excretion rates of Pb and Cd in mouse feces and urine at different time points after administration determined by ICP-MS and ICP-OES. (g, h) Tumor-targeting evaluation of the mixed nanoprobes following intravenous administration in H22 tumor-bearing mice. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. Data denote mean ± SD (n = 3). To further verify the in vivo excretion pathways of the nanoprobes, Pb and Cd contents in mouse feces and urine collected over a 25-day period post-injection were quantified by ICP-MS and ICP-OES. Both Pb and Cd were predominantly excreted via the feces with a rapid peak at 24 h (Pb, 0.72 %ID/h; Cd, 0.60 %ID/h) and slow clearance thereafter (Fig. 4e and f). In contrast, only trace amounts of Pb and negligible Cd were detected in the urine, revealing that hepatobiliary-intestinal excretion serves as the primary metabolic pathway for the nanoprobes, with renal excretion being secondary. Moreover, the majority of Pb and Cd were eliminated within 7 days of injection, and their excretion in both feces and urine approached zero by day 25. These results demonstrated that the nanoprobes exhibit negligible long-term in vivo accumulation and minimal hepatotoxicity and nephrotoxicity. Subsequently, given the co-expression PD-L1 and VEGF in H22 tumors, we established a subcutaneous H22 tumor model in BALB/c mice to evaluate the NIR-II fluorescence imaging performance and tumor retention ability of NIR-II nanoprobes. As shown in Fig. 4g and h, two distinct fluorescence signals were clearly observed within the tumor region, enabling high-contrast differentiation of tumors from normal tissues after injection of the mixed nanoprobes. Notably, both signals exhibited significant co-localization in the tumors and displayed similar metabolic kinetics, with accumulation peaking within 1 h and remaining clearly detectable (≥67% FL intensity) for ∼12 h after injection. The rapid synchronous accumulation and extended retention time in the tumor could likely be attributed to the passive enhanced permeability and retention (EPR) effect and the active transport and retention (ATR) effect mediated by nanoparticles functionalized with the PD-L1- or VEGF-targeting antibodies [45,46]. Additionally, at 72 h post-injection, fluorescence signals from both nanoprobes were mainly distributed in tumor, liver and spleen, with slightly higher signal intensity in tumor than in normal tissues (Supplementary Fig. 6). These results highlighted the nanoprobes’ exceptional potential for the simultaneous visualization of PD-L1 and VEGF in vivo. In vivo imaging by IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes demonstrated fine association with VEGF/PD-L1 expression in tumors Tumor-bearing BALB/c mice received intravenous injections of the mixed nanoprobes and underwent NIR-Ⅱ fluorescence imaging at 1 h post-injection. Following imaging, tumors were resected for immunohistochemistry (IHC) analysis. Linear regression analysis revealed positive correlations between the fluorescence intensity and the mean optical density (OD) of IHC staining. The correlation coefficient (r) for IIa QD@αVEGF fluorescence with VEGF-specific IHC staining was 0.6793 (P = 0.0053) (Fig. 5a and b), while that between IIb QD@αPD-L1 fluorescence and PD-L1 IHC optical density was 0.7941 (P = 0.0004) (Fig. 5c and d). These results demonstrated that the nanoprobes enabled accurate, minimally invasive visualization of VEGF and PD-L1 expression using NIR-Ⅱ imaging. In addition, the fluorescence intensities of IIa QD@αVEGF and IIb QD@αPD-L1 showed a strong positive correlation (r = 0.9356, P < 0.0001), indicating synergistic expression of VEGF and PD-L1 (Fig. 5e).Fig. 5NIR-II fluorescence imaging validated VEGF/PD-L1 expression and its predicting for tumor progression. (a) Representative fluorescence images of IIa QD@αVEGF and the matched VEGF IHC staining in H22 tumor sections at 1 h post-injection. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. Scale bar: 200 μm. (b) Correlation between the fluorescence intensity of IIa QD@αVEGF and the mean optical density of VEGF IHC staining via a simple linear regression model (n = 15). (c) Representative fluorescence images of IIb QD@αPD-L1 and the matched PD-L1 IHC staining in H22 tumor sections at 1 h post-injection. Scale bar: 200 μm. (d) Correlation between the fluorescence intensity of IIb QD@αPD-L1 and the mean optical density of PD-L1 IHC staining via a simple linear regression model (n = 15). (e) Correlation between the fluorescence intensity of IIa QD@αVEGF and IIb QD@αPD-L1 (n = 20). The 95% confidence intervals were obtained by two-tailed analysis. (f) Tumor growth curve (n = 5–7mice/group). (g) The quantification results of CD3+CD8+ T cells within tumors (n = 3). Data denote mean ± SD. Statistical analysis in (f) and (g) was performed using a two-way ANOVA followed by Bonferroni multiple comparisons test and a two-tailed unpaired Student's t-test, respectively. ∗ <0.05, ∗∗∗∗ <0.0001. NIR-II fluorescence imaging validated VEGF/PD-L1 expression and its predicting for tumor progression. (a) Representative fluorescence images of IIa QD@αVEGF and the matched VEGF IHC staining in H22 tumor sections at 1 h post-injection. Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. Scale bar: 200 μm. (b) Correlation between the fluorescence intensity of IIa QD@αVEGF and the mean optical density of VEGF IHC staining via a simple linear regression model (n = 15). (c) Representative fluorescence images of IIb QD@αPD-L1 and the matched PD-L1 IHC staining in H22 tumor sections at 1 h post-injection. Scale bar: 200 μm. (d) Correlation between the fluorescence intensity of IIb QD@αPD-L1 and the mean optical density of PD-L1 IHC staining via a simple linear regression model (n = 15). (e) Correlation between the fluorescence intensity of IIa QD@αVEGF and IIb QD@αPD-L1 (n = 20). The 95% confidence intervals were obtained by two-tailed analysis. (f) Tumor growth curve (n = 5–7mice/group). (g) The quantification results of CD3+CD8+ T cells within tumors (n = 3). Data denote mean ± SD. Statistical analysis in (f) and (g) was performed using a two-way ANOVA followed by Bonferroni multiple comparisons test and a two-tailed unpaired Student's t-test, respectively. ∗ <0.05, ∗∗∗∗ <0.0001. Next, H22 tumor-bearing mice were grouped into high and low mean fluorescence intensity (MFI) groups according to the median intratumoral MFI values (IIa QD@αVEGF: 210.57; IIb QD@αPD-L1: 115.34) as thresholds. As expected, tumors with low MFI exhibited markedly slower progression, showing a 28.2% smaller final volume than those in the high-MFI group (Fig. 5f). Tumor-infiltrating CD8+ T cells were detected by flow cytometry. As shown in Fig. 5g, CD8+ T cells infiltration was substantially lower in high-MFI tumors, suggesting that high expression level of VEGF and PD-L1 collectively induced the immunosuppressive TME in HCC and accelerate tumor progression. In addition, functional assays in vitro further confirmed that VEGF and PD-L1 synergistically suppressed the activation (e.g., CD69 expression, IFN-γ secretion) and proliferation of CD8+ T cells, while promoting the secretion of immunosuppressive factors (e.g., IL-10, TGF-β), thereby jointly contributing to a highly immunosuppressive TME (Supplementary Fig. 7). Taken together, these findings demonstrated that in vivo imaging using IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes enabled reliable visualization of VEGF and PD-L1 expression patterns, and that higher fluorescence intensity indicated a more immunosuppressive microenvironment in HCC. Molecular imaging by IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes enables early prediction of “T + A” therapeutic responsiveness To test whether intratumoral biodistribution of VEGF and PD-L1 could predict response to “T + A” regimen, H22 tumor-bearing mice were dichotomized into high- and low-MFI groups according to the cutoff value of 210.57 and 115.34 (the median intratumoral MFI of IIa QD@αVEGF and IIb QD@αPD-L1, respectively) (Fig. 6a and Supplementary Fig. 8). Next, mice were randomized into two subgroups, each receiving either “T + A” treatment (10 mg/kg anti-PD-L1 + 10 mg/kg anti-VEGFA) or no treatment (Control), and therapeutic efficacy was evaluated by monitoring tumor growth (Fig. 6a and b). By day 15, significant tumor volume suppression was evident in both “T + A”-treated subgroups compared with control subgroups (Fig. 6c and Supplementary Fig. 9), with no significant difference in body weight changes during the entire treatment period (Supplementary Fig. 10). However, a dramatically higher tumor inhibition rate was observed in the High MFIT + A subgroup (75.4%) than in the Low MFIT + A subgroup (41.8%) (Fig. 6d–f). This confirmed a markedly superior response to “T + A” therapy, which was attributable to high VEGF and PD-L1 co-expression. Notably, although tumor inhibition was confirmed in all “T + A”-treated mice, the High MFIT + A subgroup exhibited an earlier rapid response and stronger inhibitory effect compared with the Low MFIT + A subgroup (Fig. 6g).Fig. 6Prediction of “T + A” treatment efficacy in H22 subcutaneous tumor model using IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes. (a) Schematic illustration of the timeline of “T + A” treatment. (b) NIR-Ⅱ fluorescence quantification and representative images (High MFI, n = 14; Low MFI, n = 10). Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. (c) Average tumor growth kinetics of mice. (d) Terminal tumor volumes of high-MFI groups: control vs. “T + A” treatment groups. (e) Terminal tumor volumes of low-MFI groups: control vs. “T + A” treatment groups. (f) Photographs of tumors excised from all treated mice at the experimental endpoint. (g) Relative tumor volume change in High MFIT + A and Low MFIT + A subgroups compared to their respective control subgroups. %Volumes = (1-VT + A/Vcontrol) × 100. (h) and (i) Typical FCM plots and corresponding quantification results of CD3+CD8+ T cells within tumors after different treatments. (j) and (k) Representative immunofluorescence images staining of CD8+ (red) and nuclei (blue) in tumor tissues after different treatments. Scale bar: 50 μm. Data denote mean ± SD. Statistical analysis in (b, c) and (d, e, g, h, i) was performed using a two-way ANOVA followed by Bonferroni multiple comparisons test and a two-tailed unpaired Student's t-test, respectively. Ns means no significant difference. ∗ <0.05, ∗∗ <0.01, and ∗∗∗∗ <0.0001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Prediction of “T + A” treatment efficacy in H22 subcutaneous tumor model using IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes. (a) Schematic illustration of the timeline of “T + A” treatment. (b) NIR-Ⅱ fluorescence quantification and representative images (High MFI, n = 14; Low MFI, n = 10). Laser power: 50 mW/cm2. Exposure time: IIa QD@αVEGF, 300 m s; IIb QD@αPD-L1, 900 m s. (c) Average tumor growth kinetics of mice. (d) Terminal tumor volumes of high-MFI groups: control vs. “T + A” treatment groups. (e) Terminal tumor volumes of low-MFI groups: control vs. “T + A” treatment groups. (f) Photographs of tumors excised from all treated mice at the experimental endpoint. (g) Relative tumor volume change in High MFIT + A and Low MFIT + A subgroups compared to their respective control subgroups. %Volumes = (1-VT + A/Vcontrol) × 100. (h) and (i) Typical FCM plots and corresponding quantification results of CD3+CD8+ T cells within tumors after different treatments. (j) and (k) Representative immunofluorescence images staining of CD8+ (red) and nuclei (blue) in tumor tissues after different treatments. Scale bar: 50 μm. Data denote mean ± SD. Statistical analysis in (b, c) and (d, e, g, h, i) was performed using a two-way ANOVA followed by Bonferroni multiple comparisons test and a two-tailed unpaired Student's t-test, respectively. Ns means no significant difference. ∗ <0.05, ∗∗ <0.01, and ∗∗∗∗ <0.0001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Furthermore, the infiltration of CD8+ T cells in the TME was detected by flow cytometry to assess the activation of anti-tumor immune response after “T + A” therapy. The High MFIT + A subgroup exhibited a marked 2.29-fold increase in intratumoral CD3+CD8+ T cells infiltration compared with High MFIControl subgroup (Fig. 6h and Supplementary Fig. 11). In contrast, only a slight, non-significant increase was observed in the Low MFIT + A subgroup upon T + A treatment (Fig. 6i and Supplementary Fig. 11). A similar result was seen by immunofluorescence staining of tumor tissues, in which the High MFIT + A subgroup showed more noticeable activation of anti-tumor immune response as determined by CD8+ T cells infiltration (Fig. 6j and k). These findings demonstrated that in vivo NIR-Ⅱ fluorescence imaging using IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes facilitated stratification of high- and low-responders prior to “T + A” treatment. This capability underscored its potential as a minimally invasive strategy to guide personalized therapy scheduling, particularly in low-responder population. Discussion As an effective predictive biomarker, PD-L1 mediates tumor immune escape by inhibiting the function of CD8+ T cells [47]. However, growing evidence indicates that PD-L1 expression alone in tumors provided limited prediction for immunotherapy responses in HCC, whereas the immunosuppressive status of the TME offers greater insight into potential treatment outcomes [6,48]. Within the multifactorial regulation of the TME, VEGF promotes abnormal angiogenesis and upregulates checkpoint molecules (e.g., PD-1 and CTLA-4) on CD8+ T cells, thereby suppressing their infiltration and activation [49,50]. Therefore, simultaneous detection of the expression and distribution of PDL1 and VEGF in tumors may facilitate accurate profiling of the TME and predictive evaluation of immunotherapy response. Currently, although various fluorescent probes have been reported for in vivo imaging of PD-L1 or VEGF, they were generally confined to single target detection, failing to comprehensively reflect the potential response to immunotherapy [26,27,51]. In this study, we developed the first dual-target NIR-II molecular imaging platform for VEGF and PD-L1. Two nanoprobes targeting VEGF and PD-L1 were constructed using two NIR-II QDs with emission wavelengths at 950 nm and 1550 nm, respectively. Benefiting from their non-overlapping fluorescence spectra, the nanoprobes achieved simultaneous imaging of VEGF and PD-L1 with minimal signal interference. Moreover, compared to organic dyes, the excellent brightness, photostability, and tissue penetration of QDs afford superior the signal-to-noise ratio and detection sensitivity in vivo imaging possess [52,53]. These properties facilitated the accuracy of predicting the efficacy of “T + A” regimen in HCC. The dual-target NIR-II imaging strategy proposed in this work exhibits distinct advantages over existing clinical evaluation methods and single-target imaging approaches, as follows. (1) Minimal invasiveness: this technique requires only intravenous nanoprobe injection, thereby avoiding the tissue damage, infection risks, and patient discomfort caused by invasive (puncture or surgical) sampling for clinical IHC and genomic testing. (2) Short detection time: it enables rapid evaluation of target expression levels within hours before treatment, circumventing the multi-day processing and analysis required for IHC and genomic assays. (3) Enhanced accuracy: in this study, IHC validation confirmed that intratumoral fluorescence signals positively reflect the protein expression levels of both VEGF and PD-L1, with a consistency expression trend. These simultaneous dual-target imaging provides a more comprehensive view of the tumor immunosuppressive status and reduces missed detection or misjudgment caused by single-target imaging. In addition, IHC and genomic results are severely susceptible to tumor heterogeneity and sampling bias. (4) Potential cost-effectiveness: the QD-based nanoprobes used in this study could be mass-produced under standardized protocols to reduce costs, and could accomplish dual-target detection in a single imaging session, probably offering a clear economic advantage over IHC and genomic testing. However, this study still has several limitations. First, the long-term biosafety of the nanoprobes requires further validation. Although serum biochemistry, histopathology, and in vivo metabolism assays indicated that negligible long-term in vivo accumulation and favorable hepatic and renal safety within 25 days post-injection, extended animal studies and clinical trials remain necessary for further verification. The development of heavy-metal-free QDs may represent a crucial breakthrough direction in the future. Second, the clinical validity of the detection results needs to be strengthened. While animal experiments verified that H22 tumor-bearing mice with higher fluorescence signals of both nanoprobes responded better to the “T + A” regimen, the sample size and tumor models need to be further expanded. Further work incorporating clinical samples is essential to validate the correlation between imaging signals and therapeutic outcomes, and to standardize imaging protocols and signal interpretation criteria. Third, the application cost of the nanoprobes is constrained by imaging equipment. Most high-performance NIR-II imaging devices are designed for research, featuring high costs and complicated operation, while standardized clinical-grade systems are lacking [54]. Therefore, the development of low-cost, user-friendly clinical NIR-II imaging systems is critical to promoting its clinical translation. Conclusion In summary, we have developed a two-plex NIR-II fluorescence imaging platform utilizing PbS@CdS QDs functionalized with VEGF- and PD-L1-targeting antibodies for minimally invasive visualization of VEGF and PD-L1 expression in the HCC microenvironment. The designed nanoprobes exhibited excellent optical properties, colloidal stability, and well-resolved emission peaks at 950 nm and 1550 nm, enabling simultaneous dual-channel imaging of different targets with minimal crosstalk. Importantly, the intratumoral fluorescence signals of IIa QD@αVEGF and IIb QD@αPD-L1 nanoprobes showed strong positive correlations with the corresponding VEGF and PD-L1 protein levels, as validated by IHC. Furthermore, the early imaging signal robustly predicted therapeutic response in mouse xenograft models receiving “T + A” treatment, effectively stratifying tumors into high- and low-responders at the beginning of treatment. This predictive ability provided a valuable framework for assessing “T + A” therapy efficacy and optimizing personalized treatment regimens, thereby offering promising prospects for novel immunotherapy strategies in HCC. CRediT authorship contribution statement Jukai Feng: Data curation, Methodology, Visualization, Writing – original draft. Yuting Zhang: Data curation, Methodology, Validation, Visualization, Writing – original draft. Wenhan Liu: Data curation, Investigation, Validation. Pu Zhang: Methodology, Validation. Jianbin Xiao: Investigation, Validation. Qiang Feng: Investigation, Validation. Zhiyu Wang: Methodology, Validation. Jinyu Zhang: Validation. Yang Zhong: Validation. Zhongying Shi: Validation. Meiqi Pan: Validation. Xiaolong Zhang: Investigation, Supervision, Writing – review & editing. Jingfeng Liu: Supervision, Writing – review & editing. Jianmin Wang: Funding acquisition, Supervision, Writing – review & editing. Declaration of competing interest Jingfeng Liu, Jianmin Wang, Xiaolong Zhang, Jukai Feng and Yuting Zhang are inventors on a patent application based on technology presented in this manuscript. All other authors declare no competing interests.