3.1 Material Characterization

The prepared COFs and Au@COFs were characterized by Fourier transform infrared spectroscopy (FTIR). As shown in Fig. 1a, peaks at 1705, 3460, and 3337 cm−1 appeared, which were ascribed to the C = O and N–H stretching vibrations of TAPP and TFPA, respectively. The characteristic peak at 1617 cm−1 was assigned to C = N of COFs, indicating that the COFs were formed by Schiff base reaction between amino and aldehyde groups [18]. The FTIR spectrum of Au@COFs revealed a blue shift in the characteristic peak of C = N stretching compared with that of pure COFs (Fig. 1b), indicating the specific interaction between COFs and Au NPs [20]. FTIR spectroscopy was also applied to prove the proper functionalization of the Fe3O4 with -NH2 and FA moieties. The COFs and Au@COFs were characterized via X-ray photoelectron spectroscopy (XPS). As shown in Fig. 1c, the characteristic peaks of elements C, N, and O were obtained in COFs and Au@COFs. Figure 1d shows the narrow spectrum of Au in Au@COFs. Two peaks at a binding energy of 83.6 and 87.3 eV corresponded to Au 4f7/2 and Au 4f5/2, respectively. These results indicated that Au NPs were loaded on COFs. As shown in Fig. 1e, the characteristic peaks of the Fe3O4 could be observed at 586 and 3420 cm−1, which were assigned to the stretching vibrations of Fe–O and –OH, respectively. The peaks at 2973, 2885, and 1050 cm−1 were attributed to the stretching vibration and asymmetric bending of C–H and Si–O–Si from APTES. The peak at 1706 cm−1 was assigned to the C = O of FA, which indicated that FA was successfully conjugated with Fe3O4. Furthermore, as shown in Fig. 1f, the obtained X-ray diffraction (XRD) patterns of COFs and Au@COFs proved that the crystal structures of COFs and Au NPs (JCPDS no.04-0784) remained unchanged after composition. Fig. S1 indicates the XRD pattern of Au NPs. Peaks at 38.94°, 44.47°, 65.44°, and 76.63° corresponding to the crystal facets (111), (200), (220), and (311) confirm the formation of Au NPs.

Fig. 1

a, b Infrared spectra of COFs and Au@COFs; c, d XPS images of COFs and Au@COFs; e Infrared spectra of Fe3O4, Fe3O4-NH2, Fe3O4-NH2, and FA; f XRD patterns of COFs and Au@COFs

Subsequently, the morphologies of COFs and Au-COF nanozyme were characterized by TEM. As shown in Fig. 2a and b, COFs were spherical with an average particle diameter of ~ 200 nm. Au NPs were well dispersed on COFs with a diameter of ~ 21 nm due to the coordination of phosphine and unsaturated amino groups in COFs (Fig. 2c and d). As shown in the inset of Fig. 2d, the Au NPs on COFs were larger than those in a previous work [18], suggesting that the Au NPs were likely located outside the framework of COFs constructure. The size of Au NPs on COFs was expected to have good catalytic activity for glucose oxidation. The morphologies of COFs and AuNPs@COFs were characterized by High resolution transmission electron microscope (HRTEM). As shown in Fig. S2a, the prepared COFs are approximately spherical with an average diameter of about 200 nm. As shown in Fig. S2b, AuNPs were uniformly distributed on the COFs after being loaded, indicating that the AuNPs adsorbed onto the COFs. The crystal plane spacing is measured to be 0.236 nm, close to the lattice parameter of Au cubic structure (0.2355 nm for (111) plane) shown in Fig. S2c. And the homogenously distribution of element Au on the COFs surface was further verified by the energy-dispersive X-ray spectroscopy (EDS) elemental mapping (Fig. S2d). Figure 2e and f show the morphologies of Al-MOFs and Fe3O4 magnetic nanoparticles, respectively. The rod-like structure of MOFs and the spherical structure of Fe3O4 with a diameter ~ 300 nm could be clearly observed.

Fig. 2

TEM images of a, b COFs, c, d Au@COFs (The illustration is the size distribution of Au NPs), e MIL-53(Al) and f Fe3O4

3.2 Catalytic Activity of Au@COFs

The Michaelis–Menten kinetics parameters, including the Michaelis constant (Km) and maximum velocity (Vmax), are indicators of enzyme–substrate kinetics and are determined by slope analysis of the enzymatic reaction. Figure 3a and b show the typical Michaelis–Menten curve and Lineweaver–Burk double-reciprocal plot (1/V max [V0] vs 1/4-NP concentration [S]), respectively. The kinetic parameters of Au@COFs for reducing 4-NP were determined using the Lineweaver–Burk double-reciprocal plot. As shown in Fig. S3a and b, the Km and Vmax of COFs were investigated. The Vmax of COFs and Au@COFs were 2.73 × 10–5 M/min and 3.88 × 10–5 M/min, respectively. The Km of Au@COFs was 0.0166 mM, much lower than that of the COFs (0.364 mM). Km represents the affinity of the enzyme to the substrate. A low Km value indicates a strong affinity of the enzyme to the substrate. Therefore, the values of Km and Vmax indicated that Au@COFs had a higher catalytic activity for 4-NP reduction than COFs. The catalytic activity of Au@COFs for glucose oxidation was also evaluated (Fig. 3c and d). Compared to COFs (with Km = 1.1 mM, Vmax = 1.9 × 10–4 M/min), the Au@COFs exhibited a modest increase in catalytic activity for glucose oxidation, as evidenced by their values of Km and Vmax, which were measured as 0.98 and 2.1 × 10–4 M/min, respectively.

Fig. 3

a, c Steady-state kinetics of 4-NP reduction and glucose oxidation in the presence of different concentrations of substrate by Au@COFs nanozyme and rate relationship curves; b, d Lineweaver–Burk linear fitting

3.3 Characterization of MIL-53(Al)

The elements of MIL-53(Al) were characterized by XPS. As shown in Fig. S4b, the characteristic element peaks of C, O, and Al are observed in MIL-53(Al). The FTIR spectrum of MIL-53(Al) is shown in Fig. S4c. The peak at 3436 cm−1 is attributed to the tensile vibration of –OH. The bands at 1607–1418 cm−1, belonging to the C = C of benzene ring, appear. The peak at 1007 cm−1 is observed and attributed to the Al–O bond, indicating successful coordination between Al and O atoms. The XRD pattern of the product and the simulated pattern of MIL-53(Al) are shown in Fig. S4d. The good match between experimental and simulated XRD patterns demonstrated the formation of MIL-53(Al). The morphology of MIL-53(Al) was further characterized by SEM and TEM. As shown in Fig. S5a–d, MIL-53(Al) was composed of uniform particles with a diameter of approximately 70 nm. These results indicated the successful preparation of MIL-53(Al), which agreed with those in a previous work [22]. Figure S6 shows that the optimal time of ion exchange is 15 min.

3.4 Cell Capture

To evaluate the performance of designed nanomaterials for capturing cancer cells, equal amount of MCF-7 cancer cells and MCF-10A normal cells were incubated with Fe3O4-FA and EGFR aptamer-based COFs nanozymes at 37 °C for 1 h. Then, the cells captured by the material were separated using a magnet and seeded into a 96-well plate for microscopic observation. As shown in Fig. 4a, a few MCF-10A cells (normal cells) were observed on the surface of the capture materials in the light field. However, a large number of MCF-7 cells were observed on the surface of Fe3O4-FA and EGFR aptamer-based COFs (Fig. 4b). As shown in Fig. 4c and d, no blue fluorescence was observed under the microscopic field of view, indicating that the material was unable to capture the MCF-10A cells. In contrast, as illustrated in Fig. 4e and f, blue fluorescence was clearly observed demonstrating that the material was able to effectively capture the MCF-7 cells. These results indicated that the prepared materials can selectively capture cancer cells and hardly bind normal cells, suggesting a good specificity. Thus, the prepared materials can be used in the enrichment and separation of cancer cells in the subsequent experiments.

Fig. 4

Microscopic images of Au@COFs-Apt and magnetic nanocomposites incubated with a MCF-10A cells and b MCF-7 cells. Bright-field and fluorescence-field microscope images of c, d MCF-10A cells or e, f MCF-7 cells after incubation with the Au@COFs-Apt magnetic nanocomposite

3.5 Colorimetric and Fluorescent Dual Mode Detection of CTCs by Cytosensor

The MCF-7 cells are cancer cells with a high expression of FR and EGFR on the surface, and were selected as model cells to test constructed cytosensor, which consist of signal and capture units. The EGFR aptamer was first immobilized on the surface of Au@COFs through Au–S bonds to form the signal unit. FA functionalized magnetic nanoparticles act as the capture unit. The UV–visible detection of CTCs was based on the catalytic activity of the Au@COFs for 4-NP oxidation. The absorbance of 4-NP was measured by time-dependent UV–visible spectroscopy at 406 nm in the presence of cells with various concentrations ranging from 50 to 10,000 cells/mL. A linear correlation was observed between ln(A/A0) and the final absorbance intensity at 406 nm. Here, A0 is the original absorbance of 4-NP at 406 nm, and A is the final absorbance at 406 nm. The catalytic reaction rate constant (k) was obtained from the change in the linear slope ln(A/A0) at different times. As shown in Fig. 5a, k increased with the increase of the concentration of MCF-7 cells from 50 to 10,000 cells/mL. As shown in Fig. 5b, a linear correlation range was obtained from calibration curves between k and the log[cells] (cells is the concentration of MCF-7 cells). The regression equation was k = 12.33 log[cells]  − 20.46 (R2 = 0.9980), with a limit of detection (LOD) of 17 cells/mL (S/N = 3). On the other hand, the fluorescent detection of CTCs was based on the catalytic activity of the Au@COFs for glucose oxidation. Au@COFs is a metal–organic framework material, in which Au NPs are located on COFs equably without aggregation. This special structure endows it excellent catalytic activity, which can catalyze the production of hydrogen peroxide under specific reaction conditions. In this paper, hydrogen peroxide is produced by Au@COFs catalyzing glucose, and the specific reaction equation is presented as follows:

Fig. 5

a The reaction rate constant k value increases as the concentration of MCF-7 cells increases. b The regression equation for the calibration curve of logarithmic concentration of k with MCF-7 cells. c The fluorescence intensity of MIL(Al)-MOF decreases as the cell concentration increases. d The regression equation for the calibration curve of F/F0 versus the logarithmic concentration of MCF-7 cells

$$\text+}_ \xrightarrow}\text+}_}_$$

It is well-known that the hydrogen peroxide was produced by Au NPs and identified by ultraviolet–visible spectrum. As shown in Fig. S7, the production of H2O2 was confirmed in the presence of horseradish peroxidase (HRP) and 3,3′, 5,5′-tetramethylbenzidine (TMB). In the presence of glucose, the characteristic absorption peak of oxidized TMB increased at 650 nm, indicated Au@COFs catalyze the conversion of glucose to hydrogen peroxide, and the hydrogen peroxide produced was then catalyzed by horseradish peroxidase to generate hydroxyl radicals. This catalytic process causes the oxidation of TMB, where the color of TMB solution turned from colorlessness to blue, indicating the successful production of H2O2 by the Au@COFs. The reactions generate hydrogen peroxide to oxidize Fe2+ to Fe3+, which in turn converts the MIL(Al)-MOF to MIL(Fe)-MOF through ion exchange, and thus quenches the fluorescence of MIL(Al)-MOF [23,24,25]. As shown in Fig. 5c, the fluorescence intensity of MIL(Al)-MOF decreases with the increase of cells concentration. As shown in Fig. 5d, the regression equation obtained from the calibration curves between F/F0(F0 is the original fluorescence intensity, and F is the final fluorescence intensity) and the logcells (cells is concentration of the MCF-7 cells) is F/F0 = 1.256 logcells − 0.16 (R2 = 0.9921), with the LOD of 17 cells/mL (S/N = 3).

Compared with other cytosensors, our assay showed a lower LOD for cancer cell detection, which may meet the detection demand of low CTCs concentration in cancer patients. The lower LOD of the cytosensor may be due to the substantial redox catalytic activity of Au@COFs. In our work, Au NPs were well dispersed on the surface of COFs by coordination with phosphine and unreacted amino groups of COFs. It is worthy to note that such Au NPs in phosphine-COFs are not easy to be aggregated to big particles and may expose more catalytic sites than normal COFs [26]. For the separation strategy, due to the presence of numerous blood red cells and the instability of the CTCs in blood, the simple and fast separation is needed. In this work, specific magnetic nanoparticles for cancer cells were used, which can enrich and separate the CTCs simply by a magnet. Furthermore, the nanomaterial used in our study can catalyze oxidation and reduction toward different substrates, which enables both colorimetric and fluorescent detection for CTCs. The dual signal mode may be beneficial to avoid false positive often associating with cancer cell detection.

3.6 Detection of Cancer Cells

A standard addition method was adopted to quantify various concentrations of MCF-7 cells to evaluate the application of cell sensor. Different concentrations of MCF-7 cells were dispersed in human serum and detected by the assay. As shown in Table S1, the recovery rates of the sensor are 98.1%–106.0%, and the relative standard deviations are 3.1%–7.2%, indicating that the sensor has potential applications in clinical detection and diagnosis. Notably, FR and EGFR proteins are overexpressed on the cell membrane surface of most tumor cells. To further identity the applicability of the assay, 10 blood samples (5 normal samples and 5 breast cancer samples) were obtained and applied to the proposed assay for CTCs detection. As shown in Table 1, for the colorimetric detection mode, the absorbance of the normal samples was low, while those from breast cancer patients showed a high value, suggesting the successful detection of CTCs. Similar results were obtained for fluorescent detection mode. Thus, our sensor is expected to be able to detect other cancer cells. Further investigation of the large-scale samples using the cytosensor is needed.

Table 1 Comparison of analytical performance of the proposed assay with the others