Introduction

Pathogenic microorganisms are a group of bacteria that can invade the human body and cause infections or even infectious diseases, currently, including Escherichia coli O157:H7, Acinetobacter baumannii, and Staphylococcus aureus, which pose a serious risk to public health [1,2,3]. Prompt detection of these pathogens is important for effective infection control and clinical diagnosis. The common means of detecting pathogens include culture, PCR and mass spectrometry [4,5,6], but all these methods face some problems. For instance, the widespread application of the conventional microbial culture identification approach is hampered by its length of time and complexity. PCR often yields false positives and non-specific amplification [7, 8]. Although mass spectrometry offers rapid results, it necessitates an expensive instruments and skilled workers, making it a challenging process [9]. Therefore, there is an urgent need for faster and more accurate methods of pathogen detection is critical.

Surface-enhanced Raman spectroscopy (SERS) is a technique that capitalizes on the surface enhancement effect to amplify Raman scattering signals of samples, enabling highly sensitive detection of low concentration samples [10,11,12,13,14]. This technique boasts advantages such as high selectivity, sensitivity, and non-destructiveness, making it a powerful tool for analyzing proteins and biomolecules [15]. SERS analyses are bifurcated based on the interaction between signaling molecules and active substrates into labeling methods and “label-free” methods [16]. The label-free approach allows direct sample analysis to obtain a more complete structural spectrum of the substance [17, 18], whereas the label-based method, utilizing specific Au nanoparticles (AuNPs) as Raman tags, ensures amplified and quantifiable signals, thereby exhibiting superior reproducibility and sensitivity. For instance, Duan’s group presented a label-based SERS technique for detecting traces of S. typhimurium [19,20,21]. The detection of Listeria monocytogenes has been achieved using multiple SERS tags that employ distinct Raman reporter molecules and specialized identification components [Full size image

Performance of the CDNA-based SERS platform

To optimize the assay’s performance, several experimental parameters, including the concentration of Bst DNA polymerase, incubation time, and concentrations of Mg2+ and MB, were fine-tuned (Supplementary Fig. S7). Subsequently, different concentrations of target bacteria were used to test the detection sensitivity of the constructed platform. The concentration of E. coli O157:H7 was determined by the plate counting method (Supplementary Fig. S8). With increasing E. coli O157:H7 concentrations, a consistent rise in SERS intensity was observed (Fig. 4a). The linear connection was established using the typical Raman spectra of SERS tags at 1621 cm− 1. The signal indicated good linearity of E. coli O157:H7 concentrations ranging from 5 CFU/mL to 5 × 105 CFU/mL, with a limit of detection (LOD) of 1.91 cells/mL (R2 = 0.997) (Fig. 4b). Notably, the CDNA-based SERS platform exhibited superior sensitivity when comapared with conventional SERS tags (Fig. 4c). Furthermore, when compared to other SERS tags for detecting E. coli O157:H7, our proposed platform demonstrated superior performance (Table 1) [37,38,39,40,41].

Table 1 Comparison with other SERS tags for E. coli O157:H7 detection
Full size table

The specificity of the developed biosensor was assessed against other pathogenic species (5 × 103 cells/mL), including P.aeruginosa, S.aureus, K. pneumoniae, and A. baumannii. As exhibited as in Fig. 4d, the target bacteria E. coli O157:H7 showed a high intensity at 1621 cm− 1, which was over 9-fold higher than that of other nontarget groups. Mixtures of E. coli O157:H7 and other bacteria at a 1:10 molar ratio were used to further explore the specificity of this platform in complicated samples. There was no substantial variation in SERS intensity between E. coli O157:H7 and other complicated bacterial samples (Fig. 4e). The good stability and reproducibility of the proposed CDNA-based SERS platform was verified through 20 random measurements (Supplementary Fig. S9), and the RSD of the 20 random Raman intensities was 2.22% (Fig. 4f).

Fig. 4
figure 4

a Raman spectra of E. coli O157:H7 at different concentrations (5 ~ 5 × 105 CFU mL− 1); b The linear relationship between E. coli O157:H7 concentration and Raman intensity at 1621 cm− 1; c Linear analysis of E. coli O157:H7 detection by traditional SERS tag (blue) and CDNA SERS tags (orange); d Specificity evaluation of the SERS platform, only E. coli O157:H7 generated obvious Raman signal; e SERS intensity between E. coli O157:H7 and other complicated bacteria samples, the amount of other pathogens was ten times that of E. coli O157:H7; f The Raman intensity of 20 randomly selected SERS spectra acquired from the measurements for 5 × 105 CFU/mL of E. coli O157:H7.

Full size image

Measurement of E. coli O157:H7 in real samples

Three sample types were selected to validate the accuracy of the CDNA-based SERS platform in real samples. As shown in Fig. 5a, the signals from spiked 5 × 105 and 5 × 103 CFU/mL of E. coli O157:H7 were recovered from 87.86 to 98.72% in three real samples. Even with a spike of just 5 CFU/mL E. coli O157:H7, recovery rates were recorded at 107.97% in water, 91.62% in milk, and 103.77% in human serum.

To underscore the platform’s practical applicability, we assembled a cohort of real samples consisting of pasteurized milk samples (n = 20) and contaminated milk samples (n = 20). As shown in Fig. 5b, the SERS intensity of E. coli O157:H7 was able to differentiate between contaminated milk and pasteurized milk, with the former exhibiting greater intensity. While qPCR measurements also identified differential E. coli O157:H7 signals between pasteurized milk samples and infected milk samples, there was an observable signal overlap in the signal percentages between the groups (Fig. 5c). Moreover, our proposed platform achieved an AUC of 0.996 (sensitivity: 95%, specificity: 100%) in contaminated milk samples, outperforming qPCR (AUC of 0.874, with both sensitivity and specificity at 81%) (Fig. 5d). The CDNA-based SERS platform has higher sensitivity than qPCR, possibly due to the following reason: sample preparation process for qPCR, such as nucleic acid extraction, might induce false negatives in low-load E. coli O157:H7 samples, whereas our nucleic acid-free extraction SERS platform, by directly detecting a cascade of amplified signals, potentially circumvents this issue.

Fig. 5
figure 5

Measurement of E. coli O157:H7 in real samples. a Raman intensity of pasteurized milk versus contaminated milk; b Raman signal recovery of E. coli O157:H7 spiked in water, milk and human serum; c Ct value for genomic DNA of E. coli O157:H7 in pasteurized milk versus contaminated milk. (paired two-tailed Student’s t test, ****P < 0.0001); d ROC curve analysis of the CDNA-based SERS platform and qPCR.

Full size image

Measurement of E. coli O157:H7 in mouse serum

Given the excellent performance of our proposed platform in complex samples, we extended its application to noninvasive monitoring of infection in vivo using mouse models after studying the stability and biocompatibility of WMRs (Fig. S10 and S11). We orally administered E. coli O157:H7 preparations to 8 to 10 -week-old mice and subsequently monitored the presence of E. coli O157:H7 presence in their serum (Fig. 6a). Our platform detected E. coli O157:H7 in each mouse hourly for the initial 6 h, followed by 12-hour intervals until 120 h. Clinical symptoms appeared 1 h post-administration of the E. coli O157:H7 preparation. Figure 6b and c illustrate the in vivo proliferation of E. coli O157:H7 in the infected mice, as evidenced by a marked increase in SERS intensity compared to uninfected mice. A minor decline in SERS intensity from 3 to 5 h might be attributed to immune-mediated bacterial clearance, as demonstrated by the previous study from Shen et al. [42]. The capacity to detect E. coli O157:H7 in mouse serum enabled the early identification of infection. This platform had a high AUC of 0.969 (sensitivity: 90%, specificity: 94%) (Supporting information Table S2) for infection diagnostics, indicating its significant potential for early infection analysis (Fig. 6d).

Fig. 6
figure 6

Measurement of E. coli O157:H7 in mouse serum. a Schematic representation of the methodology employed to establish an E. coli O157:H7-infected mouse model and subsequent blood collection; b Raman intensity profiles of six mice infected with E. coli O157:H7 and six healthy mice (represented by the broken line) are shown, along with the average trace (indicated by the red and black lines). The error bars represent standard deviations; c Raman intensity of E. coli O157:H7 expressed in serum of infected mice and healthy mice at the indicated times (paired two-tailed Student’s test, *P < 0.05, **P < 0.01, ***P < 0.001); d ROC curve analysis was performed to compare healthy mice (n = 6) with E. coli O157:H7-infected mice (n = 6)

Full size image