Gold nanorods with iron oxide dual-modal bioprobes in SERS-MRI enable accurate programmed cell death ligand-1 expression detection in triple-negative breast cancer

The efficiency of immunotherapy for triple-negative breast cancer (TNBC) is relatively low due to the difficulty in accurately detecting immune checkpoints. The detection of TNBC-related programmed cell death ligand-1 (PD-L1) expression is important to guide immunotherapy and improve treatment efficiency. Surface-enhanced Raman spectroscopy (SERS) and magnetic resonance (MR) imaging exhibit great potential for early TNBC diagnosis. SERS, an optical imaging mode, has the advantages of high detection sensitivity, good spatial resolution, and “fingerprint” spectral characteristics; however, the shallow detection penetration of SERS bioprobes limits its application in vivo. MR has the advantages of allowing deep penetration with no radiation; however, its spatial resolution needs to be improved. SERS and MR have complementary imaging features for tumor marker detection. In this study, gold nanorod and ultrasmall iron oxide nanoparticle composites were developed as dual-modal bioprobes for SERS-MRI to detect PD-L1 expression. Anti-PD-L1 (aPD-L1) was utilized to improve the targeting ability and specificity of PD-L1 expression detection. TNBC cells expressing PD-L1 were accurately detected via the SERS imaging mode in vitro, which can image at the single-cell level. In addition, bioprobe accumulation in PD-L1 expression-related tumor-bearing mice was simply and dynamically monitored and analyzed in vivo using MR and SERS. To the best of our knowledge, this is the first time a SERS-MRI dual-modal bioprobe combined with a PD-L1 antibody has been successfully used to detect PD-L1 expression in TNBC. This work paves the way for the design of high-performance bioprobe-based contrast agents for the clinical immunotherapy of TNBC.


INTRODUCTION
The triple-negative breast cancer (TNBC) is one of most aggressive forms and has a very poor prognosis. 1,2 Nonetheless, the treatment for TNBC has some promising outcomes attributed to the programmed cell death protein 1 (PD-1) and programmed cell death ligand-1 (PD-L1) immune checkpoint inhibitors (ICI). 3 The detection and analysis of PD-1/PD-L1 checkpoints could help direct individual treatment strategies, which could greatly affect the efficacy and responsiveness of ICI. 4 The conventional methods for immune checkpoint detection include immunohistochemistry (IHC), western blot (WB), enzyme linked immunosorbent assay (ELISA), and so on. 5 However, all of aforementioned technologies require tissue sample collection, which can be time consuming while introducing pain and risk to patients. It is most important to develop new noninvasive, dynamic, and quantitative technologies for whole-body imaging to map the spatiotemporal heterogeneity of PD-L1 expression. 6 Positron emission tomography (PET), a noninvasive and high-sensitive imaging method, plays a vital role in the management of ICI therapy by using different nuclide molecular probes and metabolic parameters in preclinical studies and helps diagnose and treat disease in clinics. 7 Bensch et al. reported the feasibility of zirconium-89-labeled anti-PD-L1 for PD-L1 imaging using PET imaging for the first time in the year of 2018. 8 However, the zirconium-89 labeled antibody showed a long biological half-life in vivo, which would cause damage via exposure to radionuclide probes and x-ray emission sources. 6 Therefore, it is necessary to develop safe and noninvasive molecular imaging methods to monitor PD-1/PD-L1 checkpoint expression before, during, and after immunotherapy.
Magnetic resonance (MR) imaging, a nonradiative imaging technology, can offer abundant physiological and anatomic information about tumor tissue and is widely utilized for cancer diagnosis. Although gadolinium-based agents are common T 1 -enhanced contrast agents for clinical MR imaging (MRI) application, serious nephrotoxicity side effects and allergic reactions can occur. 9 Thus, iron oxide nanoparticle (IO NP)-based T 2 -enhanced contrast agents with better biocompatibility have attracted increasing attention in recent years. It is noteworthy that ultrasmall IO NPs, with a size distribution of 1-10 nm, have both T 1 -and T 2 -enhanced capabilities. [10][11][12][13] IO NPbased MRI bioprobes have the advantage of deep penetration, but their limited detection sensitivity restricts their application with lowconcentration biomarkers. Surface-enhanced Raman scattering (SERS) technology is based on the molecular and lattice vibration of light inelastic scattering. 14-16 SERS spectra is regarded as a promising molecular imaging method due to its significant advantages including high detection sensitivity, non-destructiveness, fast response time, and unique information supply. [17][18][19] For SERS substrates, noble metal materials play a vital role in boosting target molecular Raman signal owing to the surface plasmon resonance (SPR) effect. 20,21 Although silver NPs produce the largest SERS enhancement factor, 17 a certain amount biotoxicity limits silver SERS substrate application in the field of biological detection. Gold NPs, another high-performance SERS signal amplifier, have been reported to have important applications in the biosensing field, such as circulating tumor cell detection, tumor cell bioimaging, and tumor cell differentiation. [22][23][24] In particular, gold nanorods (GRs) have been reported as good SERS bioprobes in biosensing applications because of their near-infrared absorption wavelength, which is determined by the nanorod morphology-produced large longitudinal SPR effect. [25][26][27] GR-based SERS bioprobes exhibit great application potential in tumor PD-L1 expression detection in vitro. However, the relatively shallow penetration of SERS technology is an inevitable weakness of in vivo biodetection and bioimaging.
SERS and MRI modalities can complement tumor diagnostic modes and offer complementary bioimaging features with high sensitivity and deep penetration. The development of a SERS-MRI double-mode imaging bioprobe could be a promising method for accurately detecting TNBC-related PD-L1 expression. In this study, GRs with ultrasmall iron oxide (IO) composite bioprobes (GR@IO) were designed to integrate SERS and MR imaging properties, as detailed in Scheme 1. 4-mercaptobenzoic acid (4MBA) molecules were adsorbed on the synthesized GR@IO SERS samples (GRm@IO); the uniform SERS substrate morphology was beneficial for acquiring highly reproducible SERS signals. The GRm@IO composite was modified with aPD-L1 to improve the detection specificity and targeting ability for tumor PD-L1 expression detection. Almost no toxicity was observed for the GRm@IO composite SERS bioprobes based on a Cell Counting Kit-8 (CCK-8) assay. High PD-L1 expression in single TNBC cells, compared to that of MCF-7 cells, was accurately detected by the SERS detection/imaging mode using the GRm@IO composite bioprobes in vitro with detection sensitivity at the single-cell level. Moreover, PD-L1 expression in positive and negative cancer cells was accurately distinguished through the MR mode of GRm@IO composite bioprobes, and dynamic monitoring of bioprobe aggregation in tumor-bearing mice was realized in vivo via MR. To the best of our knowledge, this is the first time a GR@IO-based bioprobe combined with a PD-L1 antibody has been developed to qualitatively distinguish PD-L1 expression in breast cancer cells based on SERS-MRI dualmodal bioprobes in vitro and in vivo. This study will pave the way for the design of high-performance SERS-MRI bioprobes for tumor PD-L1 expression detection, guide TNBC clinical immunotherapy, and achieve tumor precision medicine.

RESULTS AND DISCUSSION Synthesis and characterization of aPD-L1-conjugated GRm@IO
The design route for the GRm@IO based-bioprobe, which was synthesized via seed growth and nanoprecipitation, is shown in Scheme 1. GRs were prepared using the seed generation method, as previously described, 28 and 4MBA molecules were absorbed on the surface of the GRs for SERS detecting/imaging. To obtain a substrate with high SERS activity, GRs with a length-to-diameter ratio of three were successfully synthesized by regulating the sodium hydroxide (NaOH) concentration [22.5 ll, 27.5 ll, 32.5 ll (1 mol/l)]. Broad view transmission electron microscopy (TEM) images of the GR samples are shown in Fig. S1(a). UV-visible absorption spectra and SERS spectra results show that 4MBA molecules adsorbed on GRs (GRm), with the length-to-diameter ratio of 3.72 (GRs-3.72) exhibiting the best SERS enhancement under 785 nm laser illumination [ Fig. S1(b) and S1(c)]. The high SERS activity of the GRs-3.72 sample can be explained by the fact that its absorption peak is closer to 785 nm than those of the other two samples. The TEM image of a single GRs-3.72 sample is shown in Fig. 1(a). The length was $60 nm, the diameter was $16 nm, and the size of the synthesized IO NPs was $6 nm [ Fig.  1(b)]. The zeta potentials of GRm and IO NPs were positive and negative, respectively, as shown in Fig. S2. This was attributed to the modification of SH-PEG-NH 2 in the GRs and polyacrylic acid (PAA) on the surfaces of the IO NPs. GRm@IO was prepared based on the amidation reaction between the -COOH of PAA and the -NH 2 of SH-PEG-NH 2 , 29  To increase the biocompatibility of the GRm@IO composite in solution, a layer of polydopamine (PDA) was wrapped on the surface of the material. aPD-L1 was modified with GRm@IO@PDA to endow the GRm@IO-based SERS bioprobe with high targeting efficiency to PD-L1 expression-related tumor sites. The PDA layer on the GRm@IO composite was observed by TEM ( Fig. S3) and N/Au/Fe elemental mapping images were observed in the HAADF-STEM images shown in Fig. S4, which indicated the presence of a PDA layer on the SERS bioprobe. In addition, UV-vis absorption spectra [ Fig. 1(i)] and dynamic light scattering ( Fig. S5) further verified the presence of the dopamine layer and aPD-L1 antibodies. As the GRm@IO composite was modified with the dopamine layer and aPD-L1 antibodies, the absorption spectra were shifted apparently toward red, and the particle sizes of hydration gradually increased. aPD-L1 antibodies modified on dopamine layer through the Schiff base reaction was affirmed by the zeta potential results (Fig. S2). 32 Electrical properties of the bioprobes became positive as the hydroxyl group on dopamine was consumed in the amido bonding formation process. 33 In addition, green fluorescence images of Alexa Fluor 488-labeled aPD-L1 were observed on the GRm@IO-based SERS bioprobes (Fig. S6), demonstrating that aPD-L1 antibodies were effectively linked to the GRm@IO@PDA surface, and GRm@IO@PDA-aPD-L1 SERS bioprobes were successfully synthesized. Importantly, although PDA and aPD-L1 were modified on the GRm@IO surface, the 4MBA SERS signals were clearly visible, and the signal-to-noise ratio of the designed SERS bioprobes was sufficient for tumor PD-L1 expression detection and imaging [ Fig. 1(j)]. The MR contrast agent performance of the GRm@IObased bioprobes was evaluated using MR equipment (0.5 T). T 1and T 2 -weighted images of IO and GRm@IO@PDA-aPD-L1 with different Fe concentrations (0-0.2 lg/ml) are shown in Fig. 1(k). Enhanced GRm@IO@PDA-aPD-L1 nanoparticle abilities were observed in T 1 and T 2 MR images, indicating they could be used as bimodal MR contrast agents in tumor imaging. 9 Therefore, these bioprobes exhibit SERS-MRI dual-modal detection/imaging capabilities for targeted tumors with PD-L1 expression.
To assess the capability of detecting PD-L1 expression with our SERS-MRI dual-modal bioprobes, human breast cancer cell lines of MDA-MB-231 and MCF-7, which have different PD-L1 expression levels, were selected as the research model. 1 First, biocompatibility was evaluated to ensure safety. The cytotoxicity of the GRm@IO-based bioprobes in MDA-MB-231 and MCF-7 cancer cells was studied using the CCK-8 assay [Figs. 2(a) and S7]. The CCK-8 assay results illustrate that the GRm@IO@PDA@aPD-L1 bioprobe displays almost no cytotoxicity, and its good biocompatibility supports the subsequent detection of PD-L1 expression in tumor cells. The SERS detection mode of the GRm@IO@PDA-aPD-L1 bioprobe was used to differentiate between MDA-MB-231 and MCF-7 cancer cells with different PD-L1 expression levels [ Fig. 2(b)]. We found that MDA-MB-231 tumor cells presented higher SERS signal intensities than MCF-7 tumor cells, and  and S10(c), respectively. GRm@IO@PDA-aPD-L1 groups showed more obvious T 1 and T 2 signals in TNBC cells than that for MCF-7 cells due to the high uptake and PD-L1 expression level in MDA-MB-231 cells. This is consistent with the results shown in Figs. 2(b), 2(c), and 3(a). The T 1 MR images of MDA-MB-231 cells incubated with GRm@IO@PDA-aPD-L1 for 8 h were brighter than those incubated for 24 h, which may due to the degradation of IO over time in these cells. 34 The aforementioned SERS mapping and MRI imaging results demonstrate that GRm@IO@PDA-aPD-L1 bioprobes can efficiently target TNBC cells, and that TNBC PD-L1 expression can be accurately detected by these SERS-MRI dual-modal bioprobes.

In vivo SERS-MRI dual-modal imaging of PD-L1 expression
To further evaluate the potential of our biprobes for PD-L1 expression detection in tumors, the in vivo biocompatibility of GRm@IO@PDA-aPD-L1 was investigated in Institute of Cancer Research (ICR) mice. Fifteen days after tail vein injection of PBS or GRm@IO@PDA-aPD-L1, the blood and major organs of the mice were collected for hematology analysis and hematoxylin-eosin (H&E) staining to study organ pathology (Figs. S11 and S12). GRm@IO@PDA-aPD-L1 administration had no influence on all indices examined in routine blood tests. 35 Furthermore, there were no appreciable pathological lesions in the heart, liver, spleen, lung, or kidney of mice treated with GRm@IO@PDA-aPD-L1. This indicated that our bioprobes possessed good biocompatibility for application in PD-L1 expression detection in vivo.
GRm@IO-based bioprobes were injected into tumor-bearing mice via the tail vein for in vivo PD-L1 expression detection by SERS-MRI. To explore the optimum time for maximum bioprobe accumulation, the amount of bioprobes accumulated in TNBC tumor-bearing mice based on PD-L1 expression was directly and dynamically monitored using MRI, as shown in Fig. 4(a). MR revealed that the optimum time for maximum GRm@IO@PDA-aPD-L1 bioprobe accumulation at TNBC tumor sites was 10 h. This is when the MRI signal was strongest in In order to validate the detection specificity of GRm@IO@PDA-aPD-L1 bioprobes, it could be found that there is no obvious cumulative effect of the GRm@IO@PDA bioprobes (200 ll, 0.98 mg/ml) in tumor during 24 h post injection (Fig. S13). These results demonstrated that GRm@IO@PDA-aPD-L1 bioprobes could greatly increase recognition and accumulation effects in tumors. The optimum time for bioprobe accumulation at the tumor site was confirmed by fluorescence imaging; in vivo fluorescence imaging measurements from 0 to 96 h were performed as a contrast experiment (Fig. S14). Fluorescent dye IR-808 (200 ll, 0.98 mg/ml) was coated on the GRm@IO@PDA-aPD-L1 bioprobes (200 ll, 0.2 lg/ml) and injected into MDA-MB-231 tumor bearing mice via the tail vein. Fluorescence imaging indicated that the largest number of bioprobes in TNBC tumors was observed at 10 h, which is consistent with the MRI results. Importantly, owing to the relatively deep laser penetration and significant SERS enhancement of the GR substrate with 785 nm illumination, SERS signals of the bioprobes in TNBC tumor-bearing mice (10 h) were directly acquired. Raman signals (1076 and 1580 cm À1 ) of 4MBA molecules collected from 10 laser points of the tumor site were stable and distinct [ Fig.  4(d)]. This indicated that bioprobes with high SERS activity could be a quick and simple strategy for PD-L1 expression detection in TNBC cells. These results provide a new approach for in vivo PD-L1 expression detection based on the strength of the SERS-MRI dualmodal bioprobes and offers a new method to analyze PD-L1 expression in breast tumors.

CONCLUSION
In order to effectively detect TNBC-related PD-L1 expression, GRm@IO@PDA-aPD-L1 bioprobes with SERS-MRI dual-modal imaging functions were developed. GRs and IO NPs served as the SERS and MRI platforms, respectively, to detect PD-L1 expression. SERS and MRI imaging exhibit complementary features for PD-L1 expression detection in vitro and in vivo. aPD-L1 was selected as an antibody to boost the detection specificity and improve the ability of bioprobes to target tumor sites with PD-L1 expression. PD-L1 expression in TNBC cells was accurately imaged/detected via in vitro SERS imaging/signal modes owing to the single-cell-level detection sensitivity of SRES; SERS signals were stable and evident. Importantly, MRI and SERS signal dual-detection modes were successfully utilized to directly and dynamically monitor bioprobe accumulation in TNBC tumor-bearing mice with high PD-L1 expression levels. SERS detection signals under 785 nm illumination were consistent with the MRI imaging results, indicating that the largest amount of PD-L1 expression-related bioprobes aggregated in TNBC tumors after 10 h. This work demonstrates that SERS-MRI dual-modal bioprobes exhibit great application potential in PD-L1 expression detection and will pave the way for designing functional bioprobes for TNBC clinical immunotherapy.

Synthesis of GRm (Gold nanorods with 4-mercaptobenzoic acid, 4-MBA)
Gold nanorods were prepared using the seeds growth method. 28 For the first gold seed preparation, 20 ll of 50 mM HAuCl 4 , 5 ml of 0.1 M CTAB, and 24 ll of 0.1 M NaBH 4 were mixed at a high stirring rate for few minutes to synthetize seeds. Next, the growth solution was prepared by mixing 10 ml of 0.1 M CTAB, 100 ll of 50 mM HAuCl 4 , 8 ll of 0.1 M AgNO 3 , and 40 ll of 1 M NaOH for 5 min. Next, 56 ll of H 2 O 2 (30%) was added into the growth solution until the color of the solution faded. Finally, 30 ll gold seeds was added to the resulting solution, stirred for 30 s, and kept in the dark for 8 h. Gold nanorods (GR) were obtained by centrifugation at 10 000 rpm for 10 min, three times. To prepare 4-MBA-coated gold nanorods (GRm), 5 ml of 1.25 mM GR solution was mixed with 1 ml of 10 À4 M 4-MBA signal molecules and stirred for 30 min. The GRm products were collected by centrifugation at 10,000 rpm for 10 min three times. The final GRm products were used for the Raman test (laser wavelength: 785 nm; laser power: 26.7 mW, 25%; lens: 50 Â objective).

Synthesis of iron oxide nanoparticles (IO NPs)
Ultra-small IO NPs were prepared via a one-step water-dispersible method. 36 50 ml of 0.21 g/mL PAA solution was passed through N 2 for 40 min to remove O 2 , heated to 80 C, and then, 0.145 g FeCl 3 Á6H 2 O and 0.0554 g FeCl 2 Á4H 2 O were rapidly added to the PAA solution while stirring intensely. Next, 15 ml of 28% NH 3 ÁH 2 O was injected to adjust the pH value of the solution to 9-10. When the color of the solution turned black, mixing and heating were continued for 1 h during cooling. Using a dialysis bag of 14 000 Da, 48 h of dialysis was performed; ultra-small IO NPs with carboxyl-capped ends were obtained.
Synthesis of PD-L1 antibody-conjugated GRm@IO (GRm@IO@PDA-aPD-L1) A quantity of 12 mg of NH 2 -PEG-SH 2 was added into 5 ml of a 1.25 mM GR solution and mixed for 8 h. Following this, 2 ml of 8 mg/ ml EDCÁHCl and 2 ml of 6 mg/ml NHS were mixed with 2 ml of a 1.25 mM IO NPs solution for 30 min with stirring to activate the -COOH of IO. Additionally, 2 ml of a 1.25 mM activated IO NP solution was added to 5 ml of a 1.25 mM GRm solution. These two treated samples were stirred at 600 rpm for 12 h, then centrifuged at 10 000 rpm for 20 min three times to wash the probes and obtain GRm@IO NPs. Under an alkaline environment, 600 ll of 28% NH 3 ÁH 2 O, 8 ml of ethanol, 13 ml of H 2 O, and 5 ml of 0.2 lg/ml GRm@IO NPs were wrapped with 2 ml of 10 mg/mL dopamine by stirring for 5 h. Then, GRm@IO@PDA NPs were isolated by centrifugation (10 000 rpm, 10 min) using Tris-HCl buffer (pH 8.5), washed three times, and redispersed in Tris-HCl of the same volume. Next, 10 ll of 0.98 lg/ml aPD-L1 was added into 1 ml of GRm@IO@PDA NPs, and the reaction was shaken at 600 rpm for 12 h at 25 C in the dark. Finally, the products were centrifuged at 10 000 rpm for 10 min three times to obtain the GRm@IO@PDA-aPD-L1 bioprobes.

Animal experiments
Female ICR (London, UK) mice (7-8 weeks old) and female BALB/c-nu mice (5-6 weeks old) were purchased from Hangzhou Ziyuan Biotechnology Co. Ltd. (Hangzhou, China). All animal experiments were reviewed and approved by the Regional Ethics Committee for Animal Experiments of Ningbo University [permit no. SYXK (Zhe) 2019-0005].
The biocompatibility of bioprobes in vivo was evaluated in female ICR mice. Mice were divided into three groups and injected with 100 ll PBS or the same volume containing bioprobes with different Fe concentrations (0.1 or 0.2 lg/ml) via the tail vein. After 14 days, mice were killed after cervical dislocation under anesthesia, and blood and major organs were collected for hematological analysis and hematoxylin-eosin (H&E) section staining.
BALB/c-nu mice were subcutaneously implanted with 1 Â 10 7 MDA-MB-231 tumor cells on the front of the right leg. After two weeks, the tumor volume was increased to 100 mm 3 for subsequent MRI and SERS detection. SERS and MRI experiments were performed on tumor-bearing mice using a LabRAM Odyssey spectrometer (Horiba Scientific) with Labspec 6 software and a MAGNETOM Vida syngo MR XA10 MRI scanner (Siemens Healthineers, Erlangen, Germany). All mice were then injected with 100 ll of 0.2lg/ml GRm@IO@PDA-aPD-L1, and the accumulation was observed using MR equipment. MRI images were acquired at different time points(0, 2, 4, 10, 12, 24, and 48 h). The MRI field of view was selected according to the breast cancer type. According to the MRI results, the SERS spectra of in vivo tumors were acquired 10 h after tail vein injection using LabRAM microscopy.