Previous bimetallic MOF-based biosensors targeting E. coli have typically exhibited higher detection limits depending on the used transducer and bioreceptor, which relied on broad-spectrum anti-E. coli antibodies38,39,40,41,42,43,44,45,46. Our earlier study disclosed that incorporating Mn outstandingly boosts electron transfer as supported by electrochemical cyclic voltammetry (CV) measurements47. Building upon this, we present a Co/Mn-based ZIF-67 (Co/Mn ZIF) composite, leveraging the synergistic action of two transition metals to enhance sensor sensitivity. This study systematically investigates how varying Mn content affects electron transfer and key structural features, including crystallinity, surface area, porosity, and morphology, to determine the optimal configuration for sensing applications. To date, no prior study has reported the development of a bimetallic ZIF-based MOF specifically tailored for the detection of E. coli. Although the modifications of MOFs have been broadly investigated, the study of Mn - especially within the Co ZIF-67 framework for sensing purposes - has been partially overlooked. This study conjugates anti-O-specific antibodies that bind selectively to the O-polysaccharide region of E. coli, thereby optimizing the selectivity of the biosensor. By leveraging the inherent advantages of electrochemical biosensors (e.g., high sensitivity, stability, and cost-effectiveness)48,49,50, we introduce a sensing platform that integrates bimetallic MOF with selective bioreceptors. The goal is to develop a highly sensitive and selective electrochemical biosensor for E. coli detection, achieving a low detection limit while maintaining reliable performance in real sample testing.

The physicochemical properties of the fabricated Co/Mn ZIF materials were characterized prior to their application in E. coli sensing. First, X-ray diffraction (XRD) spectroscopy was employed to evaluate the crystallinity. X-ray diffractograms (Fig. 1a) revealed distinct peaks at 2θ of 7.44°, 10.5°, 12.82°, 14.76°, 16.48°, 18.1°, 24.26°, 24.62°, 26.66°, 29.66°, 30.58°, and 32.34°, corresponding to (011), (002), (112), (022), (013), (222), (114), (233), (134), (044), (244), and (235) crystal planes for both pristine ZIF-67 and Co/Mn ZIF of the selected Co/Mn ratios of 10:1, 5:1, 2:1, and 1:1. The patterns exhibit a topology consistent with the sodalite crystal framework (crystallography open database (COD) ID: 7222297, Supplementary Data 1) of ZIF-67 indicating that Mn incorporation does not introduce new peaks in the host. However, the outset graph of Fig. 1a reveals a shift of the (011) and (112) peaks, indicating an Mn-driven reconstruction with Co. This shift occurs towards higher 2θ angles as the Mn concentration increases to Co/Mn = 5:1. Further increasing Mn content (Co/Mn ZIF 2:1 and 1:1) resulted in a peak shift back to lower angles, attributed to lattice expansion due to a larger atomic radius of Mn compared to Co. Following Bragg's law (Eq. (1)), the right shift of the (011) peak from 7.19° (ZIF-67) to 7.27° (Co/Mn ZIF 10:1), and 7.40° (Co/Mn ZIF 5:1) corresponds to a decrease in the interplanar d spacing from 12.27 Å to 12.14 Å, and 11.92 Å, respectively. Conversely, a left shift of the peak to 7.31° (Co/Mn ZIF 2:1) and 7.26° (Co/Mn ZIF 1:1) indicates an increase in d spacing to 12.07 Å and 12.14 Å (see Table 1). These observations suggest partial Mn incorporation into the lattice, as the shifts reflect competing effects of initial lattice contraction at lower Mn concentration, followed by expansion at higher Mn ratios, consistent with its larger ionic radius compared to Co.

Second, following XRD analysis, Fourier-transform infrared (FTIR) spectroscopy was employed to identify functional groups relevant to surface modification. As shown in Fig. 1b, FTIR spectra of bare ZIF-67 and Co/Mn ZIF exhibit a band at ~426 cm⁻¹ corresponding to the Co-N vibration (420-450 cm⁻¹ range), while the ~1143 cm⁻¹, ~1304 cm⁻¹, and ~1422 cm⁻¹ bands are attributed to the vibrations of C-N and C=C stretching, consistent with the aromatic amine range (1266-1440 cm⁻¹). The region between 600 and 1500 cm⁻¹ arises from imidazole ring modes, while the bands at 2840-3100 cm⁻¹ are attributed to C-H stretching. The abundance of aromatic chains creates a hydrophobic environment, facilitating van der Waals interactions. Moreover, these aromatic groups offer surface modification sites that enhance adsorption capacity and increase the accessible surface area. Notably, the FTIR spectra of Co/Mn ZIF remain nearly identical across all concentration ratios, confirming that Mn incorporation does not introduce a new functional group, further supporting the XRD findings of partial lattice integration rather than phase formation.

Third, to investigate the surface properties of the modified Co/Mn ZIF, N adsorption-desorption measurements and the specific surface area was calculated using the Brunauer-Emmett-Teller (BET) method. Figure 1c demonstrates the type I isotherm, which is typical for microporous material with a large N adsorbed volume at low relative pressure, revealing that modified Co/Mn ZIF has a higher specific surface area than the pristine ZIF-67 (S = 1583 m² g). The Co/Mn ZIF 1:1 sample exhibits the largest surface area (S = 2025 m² g), followed by the Co/Mn ZIF 2:1, 10:1, and 5:1 samples with S = 1724, 1702, and 1647 m² g, respectively. Interestingly, the modified Co/Mn ZIF 5:1 (1647 m² g) exhibits the lowest increase in surface area, corresponding to the smallest d spacing value (see Table 1). A reduced d spacing likely results from the more efficient packing of atoms or ions within the crystal lattice, leading to denser packing with fewer exposed sites on the surface, reducing the amount of accessible surface area. These values exceed other cobalt-based bimetallic materials.

Furthermore, the total pore volume reaches its maximum value of 0.86 cm g in the Co/Mn ZIF 1:1 sample, corresponding to the highest specific surface area. In comparison, pristine ZIF-67 has a total pore volume of 0.70 cm g. As Mn concentration increases, the total pore volumes for the Co/Mn ZIF 10:1, 5:1, and 2:1 samples are 0.75, 0.73, and 0.76 cm g, respectively. The total pore volume is the cumulative pore volume per unit mass, calculated from adsorption data at a relative pressure close to 1 (i.e., p/p = 0.99). It corresponds to the integrated area under the dV/dd curve across all pore diameters, as determined using BJH plot analysis in Fig. 1d. Figure 1d (inset) highlights the micropore size distribution of the modified Co/Mn ZIF as determined by micropore plot analysis. The pore diameters range from 0.42 to 2.7 nm, in which most of the pores are between 1 and 2 nm large. Pores having a diameter of <2 nm are classified as micropores according to the International Union of Pure and Applied Chemistry (IUPAC) standards. However, this microporosity significantly affects the surface area of the material, as it provides abundant adsorption sites for molecular attachment. Micropores predominate the adsorption process, as revealed by the significant adsorption capacity observed at comparatively low pressure (p/p < 0.1).

In these physicochemical structural attributes, the fourth characterization is scanning electron microscopy (SEM) analysis (see Fig. 2a-e), revealing that all Co/Mn ZIF samples having different concentration ratios retain the rhombic dodecahedral morphology of pristine ZIF-67 (see Fig. 2a). The particle size distribution was obtained by the Gaussian fitting according to the SEM images shown in Supplementary Fig. 1 on two spots taken at a magnification of 5000× (n = 100, each spot). The values are given as mean ± standard error (SE) of the fitted parameter obtained from the covariance matrix in Fig. 2a-e, and standard deviation (SD) from n = 200 are mentioned in this discussion. Figure 2a-e shows that the average particle size was determined from the peak center, ranging from 214 to 672 nm for Co/Mn ZIFs, consistent with the reported ZIF-67 dimensions (228 nm to 5.2 μm). The Co/Mn ZIF 10:1 sample showed slightly reduced particle size (468.92 ± 6.90 nm) (SD = 60.31 nm) compared to pristine ZIF-67 (472.83 ± 2.54 nm) (SD = 64.99 nm) (see Fig. 2a, b), along with rougher surfaces and sharper edges, indicating initial pore formation even at low Mn concentration (see Fig. 2b). Progressive morphological changes became evident at higher Mn ratios (Co/Mn ZIF 5:1 and 2:1) samples with the particle sizes of (436.25 ± 1.70 nm) (SD = 54.24 nm) and (480.51 ± 1.71 nm) (SD = 67.20 nm) (see Fig. 2c, d) displaying increased surface roughness, suggesting framework distortions that enhanced porosity. The structure exhibits high disorder at the highest Co/Mn ZIF 1:1 ratio, with high surface roughness and significant structural changes (442.04 ± 1.65 nm) (SD = 62.25 nm) (see Fig. 2e). The incorporation of Mn²⁺ improves pore formation due to its higher ionic radius compared to Co²⁺. The valence state dictates the spatial arrangement of the organic ligands, which in turn defines the pore shape and the framework's size. Charge balance during Mn²⁺ substitution into ZIF-67 may also introduce defects or vacancies, further modifying porosity. The observed structural evolution correlates well with the BET results, where increasing Mn²⁺ content leads to greater surface area and pore volume.

Mn content was quantitatively evaluated by X-ray fluorescence (XRF) spectroscopy. Table 2 indicates that the corresponding Mn/(Co+Mn) molar ratio continuously ascended from 0.0011 (Co/Mn ZIF 10:1) to 0.0098 (Co/Mn ZIF 1:1), respectively. This trend reflects a controlled introduction of Mn into the synthesis process of the ZIF-67 framework, further supported by an inverse correlation in the elemental weight percentages of Co and Mn. The observed trend follows the intended stoichiometric ratios, despite the Mn content being found at a comparatively low rate. Mn preferentially utilizes defect sites instead of uniformly substituting Co within the lattice.

In addition to physicochemical characterization, the electrochemical analysis revealed enhanced electron transfer properties of the modified screen-printed carbon electrode (SPCE/Co/Mn ZIF). Before conducting detailed CV measurements, two electrolyte systems-0.01 M phosphate buffer saline (PBS) and mixed 5 mM potassium ferricyanide (K[Fe(CN)]) with 0.1 M potassium chloride (KCl) in 0.01 M PBS-were tested to check whether the SPCE/Co/Mn ZIF electrode was electrochemically active or passivated through a scan rate (V) of 20 mV s and potential range from -0.5 to 1.0 V. As shown in Supplementary Fig. 2a, a distinct and reversible peak current (I) corresponding to the hexacyanoferrate (III)/(II) ([Fe(CN)]/) redox couple was observed in the mixed electrolyte, while a broad peak was present in PBS alone. It is plausible that phosphate anions in PBS could interact with the metal centers, potentially facilitating redox activity through a synergistic effect between two metals, and the phosphate as a typical solid-state electrocatalyst. Nevertheless, this broad peak remains less pronounced than the well-defined [Fe(CN)]/ peaks. The presence of a broad peak demonstrates that the Co/Mn ZIF electrode facilitates efficient electron transfer when an external redox probe, like [KFe(CN)], is introduced.

Following this evaluation, various concentrations of mixed K[Fe(CN)] were prepared with 0.1 M KCl in 0.01 M PBS and the SPCE/Co/Mn ZIF electrode was evaluated through a scan rate of 20 mV s and potential range from -0.5 to 1.0 V. Supplementary Fig. 2b demonstrates an increasing I of the redox reaction occurred by increasing the concentration of K[Fe(CN)] electrolyte. The electroactive surface area was calculated based on the Randles-Ševčík equation (Eq. (2)), with the highest value at a concentration of 5 mM (0.19 cm). In comparison, lower and higher concentrations yielded reduced electroactive surface areas, i.e., 1 mM (0.04 cm), 10 mM (0.13 cm), and 15 mM (0.12 cm). A high electroactive surface area improves the electrochemical efficiency by providing more available sites for redox reaction, which boosts the current response. The drop in the other concentration's electroactive surface area suggests a deviation from the optimal diffusion-controlled behavior, brought by mass transport limitations or capacitive distributions. Based on this observation, 5 mM K[Fe(CN)] electrolyte was selected as the optimal concentration for subsequent electrochemical measurements. Although concerns over the toxicity of ferricyanide exist, the application of 5 mM of K[Fe(CN)] electrolyte is well-supported by several previous studies.

Figure 3a provides the CV curve of [Fe(CN)]/ redox reaction, demonstrating that the modified SPCE/Co/Mn ZIF 2:1 electrode performance is enhanced through anodic peak current (I) (124 µA), which is higher than measured with the pristine SPCE/ZIF-67 electrode (13.6 µA). Its performance exceeds other modified SPCE/Co/Mn ZIF and other bimetallic MOF-based SPCE electrodes (e.g., Fe-HMOF-5, NiCo MOF, and CoNi MOF) (see Supplementary Table 1). According to the Randles-Ševčík equation (Eq. (2)) this optimal ratio also demonstrated a larger electroactive surface area of 0.24 cm vs. 0.03 cm for ZIF-67, suggesting efficient Mn incorporation that enhances electron transfer pathways and increased surface roughness at the electrode interface while maintaining structural integrity, consistent with the SEM image (see Fig. 2d). The electrochemical result indicates the controlled Mn incorporation at the 2:1 ratio creates an ideal balance between defect generation and charge transport properties for biosensing applications.

Figure 3b demonstrates the scan rate dependence in the modified SPCE/Co/Mn ZIF 2:1 electrode from 2 mV s to 1000 mV s, performing a quasi-reversible behavior with symmetric anodic/cathodic peak current (I/I) ≈ 1 (see Supplementary Table 2). The observed potential peak separation (ΔE = 180-510 mV) is significantly larger than its theoretical value of 59 mV, representing the thermodynamic and kinetic relationship for a single-electron transfer reaction under ideal conditions at room temperature. This suggests a slow electron transfer process that is beneficial in biosensing applications since biosensors should enable stable detection across a wide range of analyte concentrations, leading to a gradual system response with reduced susceptibility to interference, thereby enhancing selectivity. Furthermore, a linear relationship of I vs. square root (sqrt) scan rate (V) confirms electron transfer is diffusion-controlled across all scan rate regions-low (2-20 mV s), medium (20-100 mV s), and high (100-1000 mV s)-with only slight variations in R value fit close to ≈1 (0.96, 0.97, 0.98 for I and 0.97, 0.99, 0.98 for I, respectively) confirm the assumption of linearity (see Fig. 3c). To further evaluate the nature reaction process, we constructed a log-log plot of I vs. V (Fig. 3d). In this approach, a slope near ≈1 indicates adsorption-controlled, while a slope close to ≈0.5 indicates diffusion-controlled. Figure 3d shows a slope value of 0.43 at the lowest scan rate region (2-20 mV s). In contrast, medium and high regions performed lower slope values of 0.28 and 0.34, respectively, indicating a deviation from ideal diffusion behavior. The decrease in slope with increasing scan rate is likely due to rising kinetic constraints from limited ion accessibility or slower surface reactions. Additionally, the reduced response at high scan rate may be associated with partial saturation of active sites, where the rate of ion transport cannot keep up with the scan rate, and the ions do not have adequate time to diffuse deeply into the structure. This limitation may lead to the appearance of shoulder peaks or peak broadening at high scan rate, as observed in Fig. 3b. This observation led to the selection of the lowest region of scan rate (2-20 mV s) for the subsequent differential pulse voltammetry (DPV) measurement.

Electrochemical impedance spectroscopy (EIS) characterization was performed to quantify charge transfer kinetics and determine the ideal Co/Mn ratio. Figure 3e displays the Nyquist plots revealing distinct semicircular regions corresponding to the charge transfer resistance (R), with an enlarged view presented in Fig. 3f to better visualize the R differences among Co/Mn ZIF electrodes (the fitted values detailed in Supplementary Table 3). The R was determined by subtracting the solution resistance (R), obtained from the high-frequency intercept of the semicircle, from the total resistance at low frequency (R). The lowest R value of 322 Ω was obtained for Co/Mn ZIF 2:1 electrode, indicating enhanced kinetics and the most efficient electron transfer among the selected Co/Mn ratios, and supporting the CV test results. The R demonstrates a non-linear trend as a trade-off between the conductivity initiated by Mn-related defects and the structural integrity of the composite. This is reinforced by the elemental mapping using energy-dispersive X-ray spectroscopy (EDS), indicating the presence of Mn in the Co/Mn ZIF 2:1 sample, though the distribution appears sparse due to its low content (see Table 2). Nevertheless, the corresponding EDS spectrum confirms that the Mn signal is distinguishable above the background (see Fig. 4). The enhancements observed in both CV and EIS measurements, along with the material's high surface area (1724 m g) and retained polyhedral morphology, underscore the Co/Mn ZIF 2:1 composite as the optimal platform for E. coli biosensor development, offering simultaneous enhancements in electron transfer efficiency and mass transport capabilities.

To ensure the aqueous endurance, the hydrostability of Co/Mn ZIF 2:1 was examined after seven days of water incubation. A gradual color change from bright to dark purple was observed, suggesting a marginal change in the coordination of the metal centers, potentially related to the ligand local symmetry (see Supplementary Fig. 3). XRD analysis results (Supplementary Fig. 4 and Supplementary Table 4) show that the main crystal plane before seven days of water incubation (Fig. 1a) remained detectable, and its corresponding (011) d spacing was largely unchanged, from 12.07 to 12.04 Å. However, the peak area under the (011) plane was also lowered (from 1456.18 to 574.82), providing a quantitative measure of the significant loss of crystallinity. This confirms that a substantial portion of the material has undergone partial degradation and may transform to the amorphous phase. These findings suggest minor reordering in the core crystal structure, while a significant portion of the material loses crystallinity, as is often observed in ZIFs exposed to an aqueous environment.

After the optimized Co/Mn ZIF 2:1 material composition had been determined, further surface modifications were conducted to improve the physicochemical performance and selectivity of the Co/Mn ZIF 2:1 electrode to detect E. coli. These include immobilizing anti-O antibodies through physical adsorption and bovine serum albumin (BSA) to block non-specific areas, after depositing Co/Mn ZIF 2:1 on the working electrode (see Fig. 5). Several evaluation methods (i.e., water contact angle (WCA), attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy, and CV measurements) were used to verify the physicochemical and electrochemical changes.

WCA measurements shown in Supplementary Fig. 5a-d reveal distinct wettability changes across the functionalization steps, representing values of mean ± SD, n = 5. The working electrode (bare SPCE) that initially has a WCA of (94.50 ± 0.40)° has become more hydrophobic after Co/Mn ZIF 2:1 deposition (142.30 ± 0.70)° (see Supplementary Fig. 5a, b), attributed to the outward orientation of 2-methylimidazole hydrophobic CH groups and air pocket formation in micropores, which enhances the lotus effect. Anti-O antibody immobilization reduces hydrophobicity (WCA of (121.10 ± 1.20)°) (see Supplementary Fig. 5c), performing competing interfacial effects that the Fab region of the antibody exposes polar residues for antigen binding, while the Fc domain CH/CH of the antibody maintains hydrophobic anchoring to the ZIF via van der Waals interactions. The WCA decreases significantly to (97.80 ± 0.90)° upon BSA blocking, demonstrating a shift to hydrophilic behavior (see Supplementary Fig. 5d). It is ascribed to hydrophilic amino acid residues in BSA that form structured hydration layers. This improved wettability ensures optimal biosensor performance in aqueous environments.

ATR-FTIR spectra (Fig. 6a) confirm the stepwise biofunctionalization of electrode surfaces. The SPCE/Co/Mn ZIF 2:1 spectrum maintains its characteristic fingerprint region (1000-1200 cm⁻¹, C-N stretching from 2-methylimidazole) after the deposition onto the working electrode of SPCE, matching the previous investigation on its powder phase (see Fig. 1b). Anti-O antibody immobilization induces two spectral changes. The first change is the broadening of C-N stretching (1000-1200 cm⁻¹), suggesting distortion of imidazole ring vibrations due to anti-O antibody physical adsorption within ZIF pores. The second alteration is the emergence of amide II (1506-1541 cm⁻¹, N-H bending and C-N stretching) and amide I (1602 cm⁻¹, C=O stretch) bands, verifying protein presence. Subsequent BSA blocking amplifies these amide signals, further confirming the immobilization via hydrophobic interactions (Fc region anchoring) without covalent bond formation. The existence of this amide is also related to the hydrophilic state of the surface material, orienting the protein by exposing the amide group to the environment. As a result, hydrogen bonding with water molecules on the surface may be improved.

DPV analysis was performed to evaluate the surface modification steps using an electrolyte containing 5 mM of K[Fe(CN)] with 0.1 M KCl in 0.01 M PBS at a potential range of -0.5 to 1.0 V and a scan rate of 20 mV s. The attenuation of the reduction peak currents (see Fig. 6b) quantitatively validated successful antibody immobilization, revealing reduced electron transfer kinetics due to surface blocking. For E. coli detection, DPV responses exhibited concentration-dependent decrease in peak current across the range of 10-10 CFU mL (see Fig. 6c). The E. coli was detected through its cell wall containing lipopolysaccharides with the O-polysaccharides as the specific identity of E. coli occupying about 75% of the surface area of the bacteria. Figure 6d presents the calibration curve, revealing a linear relationship (reduction peak current vs. log E. coli; , R = 0.99), with a calculated limit of detection (LoD) of 1 CFU mL (Eq. (3), SD of blank responses (S) = 19.67, n = 3). The developed biosensor outperforms MOF-based optical-electrochemical counterparts (see Table 3) in both detection range (10-10 CFU mL) and sensitivity, while avoiding the complexity and cost of an optical setup.

Further evaluations of biosensor performance were applied to verify the selectivity, stability, reproducibility, and recovery of SPCE/Co/Mn ZIF 2:1/anti-O/BSA electrode. These evaluations were conducted through DPV measurements using an electrolyte containing 5 mM of K[Fe(CN)] with 0.1 M KCl in 0.01 M PBS at a potential range of -0.5 to 1.0 V and a scan rate of 20 mV s. A selectivity test was conducted against 10 CFU mL (in 0.01 M PBS) Salmonella, Pseudomonas, Staphylococcus, and the blank containing only 0.01 M PBS. All non-target bacteria exhibit minimal deviation from the baseline current in the selectivity test, remaining below 35% of the E. coli current response (SD, n = 3) (see Fig. 7a). Additionally, the blank sample showed a low current response (29.15 ± 4.83)% (see Supplementary Table 5), indicating that the biosensor was unaffected by external interference. These findings suggest that SPCE/Co/Mn ZIF 2:1/anti-O/BSA electrode can selectively detect E. coli over non-selective bacteria. For the long-term stability test, Fig. 7b shows the results of the stability assay based on a retention calculation (Eq. (4)). The SPCE/Co/Mn ZIF 2:1/anti-O/BSA-based biosensors for E. coli detection demonstrated good stability, maintaining over 80% functionality by week 5. This implies the SPCE/Co/Mn ZIF 2:1/anti-O/BSA electrode is reliable and stable for E. coli detection for up to one month, with only a gradual decrease in sensitivity. By averaging n = 3, the SD of the measured response was applied to compute the error bars complementing the data points (see Supplementary Table 6). It suggests consistent and reproducible data, with minimal fluctuation in stability over the testing period.

Furthermore, reproducibility tests were performed on five SPCE/Co/Mn ZIF 2:1/anti-O/BSA electrodes that were fabricated in the same batch and tested against 10 CFU mL of E. coli. Figure 7c depicts the excellent reproducibility of the five fabricated biosensors with a relative standard deviation (RSD) of 1.78% (n = 3 per electrode), encompassing all data points at all fabricated electrode batches (see Supplementary Table 7). In addition, a recovery test was performed by measuring the observed current response on the specific concentration (10, 10, and 10 CFU mL) of E. coli spiked in tap water, in comparison to the obtained current on the use of standard 0.01 M PBS (Eq. (5)). The percentage recovery shows satisfaction with the values close to 100% from (93.10 ± 2.61)% to (107.52 ± 5.90)% (n = 5), and the RSD was found to be <7% (see Fig. 7d and Supplementary Table 8). This combination of batch-to-batch reproducibility and reliable recovery in spiked samples validates the platform's readiness for environmental monitoring applications where precision and accuracy are critical.