Because components in high concentration may suppress detection of minor components, and peptides and proteins with similar mass-to-charge (m/z) ratios may result in overlapping peaks, a direct mass spectrometric analysis of serum samples often produces unsatisfactory spectra quality. Therefore, a selective enrichment of specific peptides according to their biological, chemical or physical properties can improve spectra quality significantly using magnetic beads with different functionalized surfaces [13].

In this study, we used control and 12 h infected serum samples to screen for suitable magnetic beads. They were enriched using MB-HIC8, MB-HIC18, IMAC-Cu and MB-WCX. MS analysis indicated that there were 53, 44, 63 and 47 different peaks with signal-to-noise ratios >5 between m/z of 1000 and 10,000 in serum samples between the two groups, with 84.85%, 85.86%, 96.34% and 98.61% cross validation, respectively. There was a minor overlapping area in serum samples treated with MB-WCX. Mass spectra and classified figures are shown in Figures 1 and 2. Therefore, MB-WCX was used to enrich serum peptides in this study.

In order to generate a model that can detect ricin infection in mice, we explored serum peptide profiling of subgroups I, II, IV and V, which were also optimized by genetic algorithms and verified by cross validation.

As shown in Table 1, compared to the control group, spectra of ricin infected groups I, II, III, IV and V had 67, 69, 77, 79 and 82 different peaks, respectively, with 93.81%, 94.44%, 97.73%, 97.22%, and 94.45% cross validation. Then, we took the acquired data of each one subgroup to detect the other four groups to screen the optimal detection model. Finally, subgroup V was found to be the most suitable, with detection rates of 100%, 99.2%, 99.6% and 100% with respect to subgroups I, II, III and IV.

For verifying the accuracy of this model, we took another 33 serum samples, which contained 20 samples by peritoneal injection of ricin for 1 h, and 13 samples by aerosol exposure of ricin for one day (aerosol exposure samples were supplied by our laboratory) (Table 2). Results demonstrated that this model’s detection rate almost 100%. So it strongly illuminates that the model had high accuracy and could be used for early diagnosis or detection of ricin intoxication.

Meanwhile, between subgroup V and control, 82 peaks in the mass range 1000–10,000 m/z were detected by ClinProTools software, and five peaks with m/z of 4982.49, 1333.25, 1537.86, 4285.05 and 2738.88 had significant contribution to the accuracy and sensitivity of this model. They may therefore provide biomarkers for ricin intoxication, but we cannot obtain detailed information of these five components and will continue to explore this in the future. In a word, this study may provide new insights into the mechanism of ricin intoxication and develop a new direction for setting up the toxins’ diagnosis.

As a highly toxic protein [16], trace amounts of ricin can lead to the death of animals or humans. So it is relatively difficult to detect ricin intoxication. Most existing technologies appear to be targeted for ricin’s laboratory testing. For now, immune detection is the common technology of ricin detection [17], as Radioimmunoassay, (RIA), Enzyme-linked immunosorbent assay (ELISA) and so on. In 2002, Rong [17] applied an immune colloidal gold technique for ricin detection. It was a simple and fast method, but it belonged to semi-quantitative technology and could not get a high throughput effect. With the use of mass spectrometry, accurate analysis of proteins or peptides makes high-throughput sample preparation possible. Serum peptide levels can be altered in disease conditions, implying that these molecules may be used as potential biomarkers and in the construction of diagnostic methods [8]. Analysis of peptide or protein profiling in biological fluids becomes a new potential tool for the diagnosis of human disease with the advantages of speed, ease of use and accuracy [11–14]. Combined with ClinProt™ microparticle beads launched by Bruker, ricin detection in animals, even in humans, becomes possible.

Six-week-old BALB/c female mice, purchased from the Center of Experimental Animals, Jilin University, Changchun, China, were randomly divided into 2 groups: infected and healthy control. All mice in the infected group were injected intraperitoneal injection (i.p.) with 200 μL ricin (1 μg/mL) diluted with PBS (0.02 M, pH 7.4). Control mice were given PBS without ricin by the same route. All animal experiments were approved by the Center of Laboratory Animals in Jilin Province, China. Ricin toxin was supplied by our laboratory.

Blood samples from the infected and control groups were collected at 3, 6, 12, 24 and 36 h post-infection from femoral artery, 20 mice per sample time, and designated as subgroup I, II, III, IV, V and C, respectively. After allowing to clot at room temperature for 1 h, the samples were centrifuged at 4000g for 15 min to separate out the blood serum. Recovered sera were divided into aliquots immediately and frozen at −80 °C until used. Freeze thawing was not permitted before mass spectrometry analysis.

All serum samples were treated using the ClinProt™ microparticle beads (Bruker Daltonik GmbH, Bremen, Germany) with four different surface functionalities, including hydrophobic interaction (MB-HIC8), MB-HIC18, weak cation ion exchange (MB-WCX) and immobilized metal-affinity chromatography containing copper ions (IMAC-Cu) according to the manufacturer’s instructions.

Spectra derived from infected and control groups were acquired using FlexControl software in an Autoflex III smartbeam-MALDI-TOF/MS with the Clin-Port method. The instrument was operated in a positive ion linear mode. High voltage of the Ion Source 1, Ion Source 2 and Lens were set to 20.05, 18.95 and 6.52 kV, respectively, and the Pulsed ion Extraction time was 200 ns; 200 shots were automatically accumulated for each spectrum. The mass spectrometer was calibrated using the Bruker Protein calibration Standard I Calibration kit after collection of every 10 spectra.

ClinProTools software version 2.2 for Biomarker Detection and Evaluation (Bruker Daltonics) was used for normalization of a set of spectra obtained from the ricin-infected and control groups. The classification models were generated by Genetic Algorithms (GA) within the software suite, and cross validation was performed to evaluate the reliability of a classification model. A p-value of <0.05 was considered statistically significant.

In order to verify accuracy of the model, other serum samples of infected mice, by i.p. and aerosol exposure infection, were collected and treated with magnetic beads in the same way. After data acquisition, we analyzed it with this model.