The synthetic scheme for the antibody-conjugated MNPs is detailed in Figure 1A. The iron oxide nanoparticles (Fe₃O₄) were prepared by mixing FeCl₂ and FeCl₃ under basic conditions according to the previously reported procedure 40 . Through treatment with 3-aminopropyltrimethoxysilane (APS), the aminosilane coated MNPs (AS@Fe₃O₄ MNPs) exhibited an increased stability and contained amino groups for surface functionalization. The direct cross-linking of antibody with the AS@Fe₃O₄ MNPs was achieved through activation with a bifunctional linker, suberic acid bis(N-hydroxysuccinimide) ester (DSS), followed by incubation with the H5N2-specific monoclonal antibodies 41 . Compared with conventional approaches using protein G or protein A for antibody immobilization, direct conjugation was chosen to avoid non-specific association arising from protein A-conjugated MNPs. Because of the presence of other abundant non-antigenic proteins in the allantoic fluid, we noted that proper surface blocking of the MNPs was essential to avoid nonspecific interactions, which seriously compete with specific nanoprobe-virus recognition. Thus, further surface capping with an optimized concentration of methoxy-terminated ethylene-glycol amine (MEGA) was conducted to give the desired antibody-conjugated MNPs (designated as Ab_H5N2@Fe₃O₄ MNPs). The MEGA-capped Ab_H5N2@Fe₃O₄ MNPs were washed with phosphate buffered saline (PBS, pH 7.4) and stored at 4°C for months without loss of activity.

The synthesized AS@Fe₃O₄ MNPs exhibited a spherical shape with an average diameter of ~90 ± 30 nm, as seen in transmission electron microscopy (Figure 2A). The spinel structure of AS@Fe₃O₄ MNP was revealed by powder X-ray diffraction (Figure 2B), which showed the characteristic pattern of diffraction peaks at (220), (311), (400), (422), (511) and (440) 42 . The AS@Fe₃O₄ MNP was superparamagnetic at room temperature (Figure 2C), as evidence by its hysteresis loop in superconducting quantum interference device magnetometer (SQUID) (i.e., the magnetization reached saturation in an external magnetic field, but the magnetic character diminished in the absence of an external magnetic field). The unique superparamagnetic property of AS@Fe₃O₄ MNPs allowed easy magnetic separation from a complex mixture during synthesis or incubation.

Our strategy for isolation and identification of HA proteins by Ab_H5N2@Fe₃O₄ MNP is depicted in Figure 1B. Briefly, the Ab_H5N2@Fe₃O₄ MNPs were incubated with virus lysate in a chosen buffer. After incubation, the newly formed complexes containing Ab_H5N2@Fe₃O₄ MNPs and extracted HA (designated as HA-MNP complexes) were readily isolated by applying a magnet. The nonspecifically bound proteins were washed away, and the HA-MNP complexes were directly analyzed by either SDS-PAGE or MALDI-TOF MS. It was noted that both assays could be performed in an efficient manner without commonly required elution and desalting; our method thus reduced the time required for assay.

It was also noted that the incubation buffer was crucial to this experiment. Virus lysate in allantoic fluid has a complex composition, containing not only the H5N2 virus, but also other proteins abundant in the culture medium. Non-specific adsorption from these abundant proteins often interferes with immunoassays. Furthermore, HA is a trimeric transmembrane protein; thus detergent or salts are required to properly solubilize the hydrophobic protein and to avoid protein-protein aggregation. The effect of using different buffers on the specificity of affinity extraction was therefore evaluated. RIPA buffer, which is commonly used as a lysis buffer for immunoprecipitation assays, showed the best removal of abundant proteins capable of nonspecific interactions with Ab_H5N2@Fe₃O₄ MNPs. Consequently, isolation of the viral HA protein by Ab_H5N2@Fe₃O₄ MNPs was more efficient in RIPA buffer than in water or PBS buffer.

MALDI-MS can be combined with a biologically active probe to rapidly and specifically detect proteins of interest. To explore the capability of on-MNP readout by MALDI-MS, a protein pool mimicking a complex biological medium was prepared in 60 μL of PBS comprising various amounts of the recombinant HA protein of antigenic H5 type (200 ng, 100 ng and 50 ng) and a significant excess of other "nonantigenic" proteins such as transferrin (5 μg), fetuin (5 μg) and ribonuclease (1 μg). The abundance of the HA protein (1.8-0.5%, w/w) was purposely low to test the extraction efficiency of the Ab_H5N2@Fe₃O₄ MNPs. After enrichment of HA proteins, the HA-MNP complex was mixed with sinapinic acid (SA), a MALDI matrix, and directly subjected to MALDI-MS analysis.

Although the antibody-antigen complexes have strong interactions with dissociation constants (K _d) ranging from 10^-7 to 10^-11 M, most antibody-antigen complexes can still be dissociated at extreme pH (i.e., pH < 2 or pH > 12). The SA matrix solution used for MALDI-MS analysis has a low pH (< 2) and thus may directly elute the antigen bound to the antibody-conjugated MNPs on the MALDI sample plate.

Prior to affinity extraction (Figure 3A), the MALDI spectrum of the protein mixture was complex and the targeted HA was not observed due to its low abundance and ion suppression. After affinity extraction (Figure 3B), two distinct signals appeared at m/z 75238 and 37607 Da, which were derived from the glycosylated HA protein with one and two charges ([M + H]⁺ and [M + 2H]²⁺, respectively) 25 . The characteristic broad peak shape of the glycosylated protein 43 44 45 could not be resolved under the mass resolution of our instrument. The shift of several kDa compared with the theoretical molecular weight of HA (theoretical average mass ≈ 64 kDa) 43 44 45 might be attributable to the extensive glycosylation of HA 25 . When the solution containing less HA (100 ng) was analyzed, the signal intensity significantly decreased and partially overlapped with the antibody signals to form a doublet mass peak (Figure 3D). After subtraction of the mass spectrum of the antibody (Figure 3B), the mass peak occurring at m/z ~75400 Da clearly showed the presence of HA protein.

The affinity extraction of HA protein from a complex mixture was very efficient using the antibody-conjugated MNPs, as even a minute amount (50 ng) of HA protein was visible (Figure 3E). Assuming full recovery of all the HA protein (50 ng) present in the solution, the absolute detection limit was estimated to be 0.7 pmol (9 nM).

To evaluate the efficiency of our detection method, we investigated the effect of incubation time on HA protein and antibody-conjugated MNP recognition. After incubation of antibody-conjugated MNPs with HA containing solution, the amount of remaining HA was concentrated and measured by MALDI-MS. The time course of such affinity extraction indicated that > 99% of HA protein in solution was captured by MNPs within 1 min (Figure 4). Thus, using antibody-conjugated MNPs for rapid and specific extraction, followed by direct mass spectrometric analysis for ambiguous readout, provides an efficient and accurate assay for HA protein.

The affinity extraction of H5N2 viruses in allantoic fluid was also realized using Ab_H5N2@Fe₃O₄ MNPs. As shown in SDS-PAGE analysis, without concentration of the virus lysate from allantoic fluid, only the abundant protein NP was observed in H5N2 virus (Figure 5A, lane 1). Using 2 × RIPA buffer, the Ab_H5N2@Fe₃O₄ MNPs selectively isolated the viral HA protein from allantoic fluid (Figure 5A, lanes 2 and 3). As influenza HA protein (H₀) is composed of two disulfide-linked polypeptides, HA₁ and HA₂, the use of a reducing reagent, like β-mercaptoethanol, in SDS-PAGE analysis will dissociate the HA protein into the two subunits HA₁ and HA₂. After treatment with β-mercaptoethanol, an electrophoresis band occurring at ~45-50 kDa was attributable to the glycosylated HA₁ protein (Figure 5A, lanes 2 and 3) 25 43 44 45 . The band with molecular mass in the range of 50-60 kDa was ascribed to the proteins of antibody from Ab_H5N2@Fe₃O₄ MNPs (Figure 5A, lane 5). As a negative control, we also synthesized MEG@Fe₃O₄ MNPs from aminosilane coated AS@Fe₃O₄ MNPs and MEGA without antibody conjugation. As expected, SDS-PAGE analysis showed that no viral protein was trapped by MEG@Fe₃O₄ MNPs. These results demonstrated the low background and the lack of false positives (Figure 5A, lane 4) in the presented method.

To evaluate the assay sensitivity, Ab_H5N2@Fe₃O₄ MNPs (10 μg) were incubated with different amounts (0.05-1 μL) of the 1000-fold diluted H5N2 virus lysate (~10^7.8 EID₅₀/mL, Figure 5B). As low as 0.1 μL of virus lysate, corresponding to ~10^3.8 EID₅₀ of H5N2 viruses, was detected by SDS-PAGE using silver staining for visualization (Figure 5B, line 4).

To further confirm the identity of the captured protein, the electrophoresis band at ~45 kDa was digested with trypsin and subjected to LC-MS/MS analysis. Database searching with the Mascot engine confidently identified eight peptide sequences corresponding to the viral HA₁ protein (Figure 6B). An example MS/MS spectrum is shown in Figure 6A for the peptide SELEYGNCNTR at m/z 1341.4837, which corresponds to amino acids 282-292 of the HA protein from the H5N2 virus.

Finally, we evaluated whether the high specificity of Ab_H5N2@Fe₃O₄ MNP could be used to unambiguously differentiate virus subtypes. Besides the H5N2 virus (A/Duck/Taiwan/3233/04), three recombinant H5N1 viruses (RG5, RG23, and NIBRG14) were investigated. All of these influenza viruses belong to the H5 category of influenza A, but it was expected that the Ab_H5N2@Fe₃O₄ MNP incorporating the monoclonal antibody specifically against the H5N2 virus would not have cross-reactivity with the H5N1 virus subtype. To validate such detection specificity, the lysates of H5N2 and H5N1 viruses were incubated separately with Ab_H5N2@Fe₃O₄ MNPs in 2 × RIPA buffer at 25°C. After magnetic separation, the pellet was washed with 1 × RIPA buffer and subjected to MALDI-MS analysis. As shown in Figure 7A, the MALDI mass spectrum obtained from the on-MNP detection of extracted H5N2 virus (A/Duck/Taiwan/3233/04) revealed strong signals from viral proteins bound on Ab_H5N2@Fe₃O₄ MNP. After subtraction of the antibody spectrum shown in Figure 7C, the mass spectrum clearly showed signals for glycosylated HA together with the relatively abundant NP (56002 ± 10 Da) and M1 (28000 ± 10 Da) proteins from H5N2 virus (Figure 7B). Gratifyingly, none of the recombinant H5N1 viruses were trapped by the Ab_H5N2@Fe₃O₄ MNP. A representative mass spectrum obtained from a screening of H5N1 (RG5) is shown in Figure 7D, and only antibody signals were observed, confirming highly specific affinity isolation of H5N2. These data demonstrated the promising application of Ab_H5N2@Fe₃O₄ MNPs to distinguish different subtypes in influenza virus surveillance.

All of the chemicals were of reagent grade unless indicated otherwise. Ferrous chloride tetrahydrate (FeCl₂ ^.4H₂O), ferric chloride (FeCl₃), tetraethyl orthosilicate (TEOS), 3-aminopropyltrimethoxysilane (APS), 1-propanol, ammonia solution (28%), and dimethyl sulfide were purchased from Acros. Suberic acid bis-N-hydroxysuccinimide ester (DSS), 2-bromoethylamine hydrobromide (BEI), sodium carbonate, silver nitrate, α-cyano-4-hydroxycinnamic acid (CHCA), and sinapinic acid (SA) were purchased from Sigma-Aldrich. Formalin, methanol, acetonitrile (HPLC grade), and trifluoroacetic acid (TFA) were purchased from Merck. Acetic acid and sodium thiosulfate were purchased from J. T. Baker. All the chemicals were used as received without further purification. A homemade magnet was used in the separation of magnetic nanoparticles.

The HA proteins of H5 type were prepared using the H5 consensus sequence previously described 46 . The 1 × PBS buffer (pH 7.4, GIBCO) contained NaCl (137 mM), KCl (2.7 mM), KH₂PO₄ (1.5 mM) and Na₂HPO₄ (8.1 mM). The 1 × RIPA buffer (pH 8.0) contained Tris-HCl (50 mM, Amersham Bioscience), NaCl (150 mM), NP-40 (1%, Calbiochem), sodium deoxycholate (0.5%, Sigma) and sodium dodecyl sulfate (0.1%, USB). The 2 × RIPA buffer (pH 8.0) contained Tris-HCl (50 mM), NaCl (150 mM), NP-40 (2%), sodium deoxycholate (1%) and sodium dodecyl sulfate (0.1%). The TEN buffer (pH 7.5) contained Tris-HCl (50 mM, pH 7.5), EDTA (pH 8.0, 1.3 mM) and NaCl (100 mM).

Transmission electron microscopy (TEM) images were obtained on a Hitachi H-7100 Transmission Electron Microscope. Powder X-Ray Diffraction (XRD) was recorded by PANalytical X' Pert PRO. The magnetic effect was measured on a superconducting quantum interference device magnetometer (SQUID, Quantum Design MPMS7). LC-MS/MS spectra were recorded on a Waters Q-TOF™ Premier mass spectrometer (Waters Corp, Milford, MA). MALDI-TOF MS analyses were performed using an Applied Biosystems 4800 mass spectrometer (Applied Biosystems, Foster City, USA) equipped with a Nd-YAG laser (355 nm).