Thin-Layer Matrix Sublimation with Vapor-Sorption Induced Co-Crystallization for Sensitive and Reproducible SAMDI-TOF MS Analysis of Protein Biosensors
Coupling immunoassays on self-assembled monolayers (SAMs) to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) provides improved assay selectivity compared with traditional photometric detection techniques. We show that thin-layer-transfer (TLT) of α-cyano-4-hydroxycinnaminic acid (CHCA) MALDI matrix via vacuum sublimation followed by organic solvent-based vapor-sorption induced co-crystallization (VIC) results in unique matrix/analyte co-crystallization tendencies that optimizes assay reproducibility and sensitivity. Unique matrix crystal morphologies resulted from VIC solvent vapors, indicating nucleation and crystal growth characteristics depend upon VIC parameters. We observed that CHCA microcrystals generated by methanol VIC resulted in >10× better sensitivity, increased analyte charging, and improved precision compared with dried droplet measurements. The uniformity of matrix/analyte co-crystallization across planar immunoassays directed at intact proteins yielded low spectral variation for single shot replicates (18.5 % relative standard deviation, RSD) and signal averaged spectra (<10 % RSD). We envision that TLT and VIC for MALDI-TOF will enable high-throughput, reproducible array-based immunoassays for protein molecular diagnostic assays in diverse biochemical and clinical applications.
Key wordsMALDI SAMDI Protein analysis Reproducibility
Immunoassays are used in a variety of biochemical and clinical applications, including protein expression assays, biomarker validation analyses, and enzyme kinetics [1, 2, 3]. Detection methods include colorimetry, radioactivity, fluorescence, chemiluminescence, electrochemical, quartz crystal microbalance (QCM), and surface plasmon resonance (SPR) [4, 5, 6]. These methods are sensitive and scalable, often enabling quantitative measurements in parallel and with excellent reproducibility. However, since assay selectivity is dependent upon an antibody, these methods are susceptible to interferences that can lead to false positive or negative readings.
Immunoassays adapted to MS benefit from the enhanced selectivity provided by molecular mass determination, improving specificity and enabling unintended, unknown, or multiplexed components to be identified [7, 8]. Nelson and coworkers have pioneered MALDI-TOF immunoassays using antibody-bound microcolumns to isolate antigens from samples followed by elution to a MALDI plate . Others have performed MALDI-TOF directly on antigen-bound media (e.g., surface-enhanced laser desorption ionization (SELDI-TOF) MS and immuno-MALDI (iMALDI)) [10, 11]. The work highlights that immunoassays with MALDI-TOF necessitate specialized solid-phase substrates (e.g., porous frits, affinity beads, planar surfaces) and chemistries (e.g., self-assembled monolayers (SAMs) , nitrocellulose films ) to optimize speed and reliability of target enrichment and MS detection sensitivity.
While MALDI-immunoassays demonstrate robust detection of target antigens, the methods typically suffer from poor quantitative reproducibility using conventional dried droplet (DD) matrix deposition. DD leads to non-uniform analyte/matrix co-crystallization forming “sweet spots” which reduce inter- and intra-assay reproducibility [14, 15, 16]. Alternative matrix deposition methods have been reported to enhance homogeneity of matrix/analyte co-crystals [17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28], although MALDI-TOF measurements typically yield precisions above 20 % relative standard deviation (RSD). Performing relative quantitation against an internal standard at similar molecular mass to the target analyte improves MALDI precision to <10 % RSD [29, 30].
Monolayers provide a unique platform for diverse biochemical assays, support multiplexed processing via high-density arrays, and are amenable to MALDI-TOF MS (denoted SAMDI-TOF MS) [31, 32, 33]. Previous SAMDI-TOF immunoassays incorporated SAMs composed of alkanethiolates on gold functionalized with polyethylene glycol groups and metal binding functional groups to facilitate antibody immobilization . These previous immunosensors were prepared by immobilizing his-tagged protein G onto Ni2+-bound SAMs, followed by binding of IgG antibodies targeted to the analyte of interest. The work demonstrated analyses of a variety of intact proteins, protein complexes, and PTMs in biofluids, with minimal non-specific binding owing to the surface chemistries used . Compared with traditional adsorption methods that randomly coat antibodies onto substrates, SAM chemistries can be tailored for improved control of antibody orientation and surface density to maximize fractional activity of antibody-antigen interactions and enhance loading capacity . Such precise control over surface characteristics enables reproducible, quantitative measurements with SPR or QCM immunoassays; however, internal standards were still necessary to improve reliability of MALDI measurements .
This work focuses on reducing MALDI-TOF spectral variability for protein immunoassays without the use of internal standards (i.e., “label-free” processing). Recent work suggests that MALDI crystals with a narrow size distribution yield improved reproducibility, resolution, sensitivity and signal-to-noise ratio (S/N) [35, 36]. Based on this work, we postulated that controlled distribution of immobilized proteins on SAMs with uniform MALDI matrix crystallization would improve SAMDI-TOF spectral reproducibility and sensitivity. We adapted common strategies used in materials science to control MALDI matrix deposition and matrix/analyte co-crystallization. Here we show that “thin-layer transfer” (TLT) of matrix via vacuum matrix sublimation combined with organic solvent-based “vapor-sorption induced co-crystallization” (VIC) enables generation of homogeneous matrix/analyte co-crystals, resulting in improved sensitivity and reproducibility compared with DD matrix deposition.
2.1 Materials and Reagents
Detergents, MALDI matrix, and protein standards were purchased from Sigma-Aldrich (St. Louis, MO, USA). All organic solvents used were of HPLC grade, and ethyl alcohol to store and wash SAM plates was absolute anhydrous 200 proof. Water was purified to a resistance of 18 MΩ. The 6× His-tagged protein G was purchased from Biovision Inc. (Mountain View, CA, USA). IgG antibodies to human serum albumin, cystatin C, myoglobin, and transferrin were purchased from Bethyl Laboratories Inc. (Montgomery, TX, USA). Human pooled normal plasma was purchased from Precision BioLogic (Dartmouth, Nova Scotia, Canada). HEPES buffer was purchased from Cellgro (Herndon, VA, USA) and phosphate buffered saline (PBS) was purchased from Gibco (Grand Island, NY, USA). Microscope cover glasses (24 × 60 × 0.15 mm) were purchased from VWR Scientific Inc. (Media, PA, USA).
2.2 Immunoassay Procedure
Nitrilotriacetic acid (NTA) SAMs were generally prepared as described previously . Briefly, 24 × 60 mm microscope cover glasses were cleaned with a residue-free, phosphate-free detergent (Fisher Scientific, Pittsburgh, PA, USA) prior to gold coating. Glass slides were metalized with 30 Å of chromium followed by 300 Å of gold using a CHA Mark 50 E-gun evaporator. Monolayers were generated at room temperature overnight by immersion of gold slides in an ethanolic solution containing triethylene glycol (EG3)-OH and EG3-NTA terminated alkanethiolates. Surface density of NTA was estimated by the EG3-OH to EG3-NTA solution ratio. The SAM-covered gold slides were cut into rectangular pieces measuring 4 × 5 mm. Immunoassays were generated by treating the surface with 10 mM NiSO4 (H2O, 0.01 % Tween-20) for 30 min to generate NTA/Ni2+ complex, followed by immobilization of 0.77 μM His-tagged recombinant protein G (PBS, pH 7.2, 0.01 % Tween) for 30 min and then 0.33 μM IgG antibody (PBS, pH 7.2, 0.01 % Tween-20) for 30 min. Plasma was diluted with PBS or HEPES buffer with 0.01 % Tween-20 and incubated at room temperature for 1–3 h. All incubation steps (except SAM generation) were performed by spotting solution onto the surface. Chips were rinsed with a stream of 18 MΩ water after each incubation step to remove residual solution and decrease non-specific interactions, followed by drying with a gentle stream of N2 gas. For DD matrix deposition, a thin layer of CHCA solution in the solvents described in the results (0.3 μL, 10 mg/mL in organic solvent) was applied to each chip and dried prior to MALDI mass analysis.
2.3 Thin Layer Transfer of MALDI Matrix
A custom device was used for thin-layer transfer (TLT) of matrix via vacuum sublimation. A hotplate (VWR International, West Chester, PA, USA) was placed inside of a transparent vacuum chamber (Thermo Fisher Scientific, Rochester, NY, USA). A 2 mg/mL solution containing MALDI matrix in acetone was uniformly spread on a “TLT” glass slide (47 × 57 × 0.2 mm) using a pipette. The use of CHCA as a matrix for sublimation is common due to the good thermal properties  and mass spectral performance, although other matrix molecules may also be deposited by sublimation [23, 28]. The TLT slide with matrix was placed on top of the hotplate. SAM biochips were attached to an aluminum cooling mantle (170 × 115 × 33 mm) positioned 1 cm away from the TLT slide. The cooling mantel was maintained at 10 °C by a recirculating water cooler (Thermo Fisher Scientific, Rochester, NY, USA). The vacuum chamber was evacuated to ~125 Torr and the hotplate was gradually heated to allow the matrix on the TLT slide to sublime. Typically, matrix sublimation initialized at ~140 °C and completed at ~160 °C (total TLT time: ~3.5 min). Under these conditions, matrix transferred uniformly to the SAM biochip with minimal spatial dispersion. The surface density of sublimed matrix was controlled by the amount of matrix deposited onto the TLT glass slide. Upon completion of TLT the hotplate was turned off, vacuum pressure was released, and the SAM biochips were removed and allowed to cool to RT within 1 min.
2.4 Vapor-Sorption Induced Co-Crystallization
Biochips with CHCA deposited by TLT were treated with a variety of solvent vapors including water, methanol, ethanol, acetone, acetonitrile, isopropanol, dimethyl sulfoxide (DMSO), and N,N-dimethylformamide (DMF). Solvent properties relevant to the VIC process are summarized in Supplementary Table S1. To achieve matrix/analyte co-crystallization, a solvent chamber (80 mm o.d. × 40 mm) was filled with 30 mL solvent. The biochips were placed at the top of the solvent chamber and exposed to vapors at controlled temperatures and time (see Supplementary Table S2). The VIC completion time was empirically determined and coincided with an optical transition (determined visually) of the sublimed amorphous matrix to a glass state. Upon completion of the VIC process the SAM biochips were allowed to air dry at RT to enable matrix/analyte co-crystallization. The typical drying times were <1 s for all solvents except DMSO (~3 min) and DMF (~5 s). After drying, the SAM biochips were attached to a membrane sample plate, #V700698 (AB SCIEX, Foster City, CA, USA) with vacuum grease for MALDI-TOF analysis.
2.5 Mass Spectrometry
All mass spectra in this study were obtained using a Voyager DE-PRO MALDI-TOF mass spectrometer (Applied Biosystems, Framingham, MA, USA) in linear mode. The instrumental parameters were as follows: accelerating voltage (AV): 25 kV, grid voltage: 93 % AV, guide wire: 0.15 %, and extraction delay time: 510 ns. A 337 nm nitrogen UV laser (pulse width: 3 ns, pulse energy: 120 μJ max at 10 Hz) was used at a repetition rate of 3 Hz for sample ionization. The laser energy was tuned for each matrix deposition method. The laser irradiation area was estimated to be ~80–100 μm2 with a substrate incidence angle of ~30°. Mass spectra were generated from 100 laser shots and were stored as raw data without further manipulation unless otherwise stated. To generate calibration curves for albumin and spiked myoglobin, advanced baseline correction in the Data Explorer software (ver. 4.9) was used for peak height measurement. Mass spectra were externally calibrated with equine myoglobin, bovine ubiquitin, and bovine serum albumin. For all deconvoluted plots, data points represent the average of three spectra, unless otherwise stated.
2.6 Light Microscopy, SEM, and AFM
The matrix crystal morphologies of the TLT/VIC surfaces were measured by light microscopy, scanning electron microscopy (SEM), and atomic force microscopy (AFM) to determine matrix thickness, crystal size, and shape. A Nikon MM-40 microscope (Nikon Instruments Inc., Melville, NY, USA) with a Sony CCD-IRIS color video camera (Sony Electronics Inc., New York, NY, USA) was used with 5×, 10×, 20×, 50×, and 150× objective lenses. For SEM, MS immunoassays with sublimed matrix were pre-coated with approximately 10 nm of gold for 60 s in a sputter coater (Cressington 108auto, Watford, Hertfordshire, UK) at an argon gas pressure of 0.2 Torr and an applied current of 20 mA. The matrix crystals were investigated at 10–15 kV with an FEI XL30 environmental SEM (Hillsboro, OR, USA). For AFM, a VEECO Dimension 3100 atomic force microscope (Plainview, NY, USA) was used to characterize matrix film thicknesses.
3 Results and Discussion
3.1 Thin Layer Transfer (TLT)
Matrix deposition by sublimation has proven reliable for a variety of applications in the MALDI imaging field [19, 22, 23, 28, 38, 39]. Conventional sublimation apparatuses for bulk chemical purification provide limited control over thickness and in-plane uniformity of the matrix deposited. To overcome this, we developed a sublimation strategy that controls the matrix surface density on immunoassays with common laboratory equipment. The approach was designated as “thin layer (matrix) transfer” (TLT) to differentiate this specialized application from traditional bulk sublimation deposition; with TLT providing matrix surface densities of tens of ng/mm2 to maximize sensitivity for detection of proteins in a monolayer. Model immunoassays against plasma proteins were used to compare TLT and DD matrix deposition . Unlike DD, where inhomogeneous matrix crystallization was observed (e.g., Figure S1b), deposition of CHCA by TLT provides uniform matrix over the SAM substrate (Figure S1d), although no MS peaks were observed (Figure S1c) unless the surface rested for extended periods in humid environments (Figure S1e). The results suggested that TLT renders CHCA on SAMs in a state that does not facilitate ultra violet (UV) laser energy absorption and/or co-crystallization with the immobilized analyte.
3.2 Vapor-Sorption Induced Co-Crystallization (VIC)
Crystallization and Crystal Morphologies
Organic materials deposited by vacuum sublimation are typically amorphous or exist in polymorphic crystalline states that are less thermodynamically stable than large monodisperse crystals . We hypothesized that vapors from organic solvents would promote the crystalline transformation similar to methods used in materials research [40, 41]. To test this, five polar solvents with varying CHCA solubility were evaluated for suitability for VIC and downstream SAMDI MS (Table S1). Each solvent vapor system was empirically optimized for duration of vapor exposure and solvent temperature during VIC (Table S2) using distinct visual cues such as change of surface color and/or transparency.
Analyte discrimination is well known in MALDI, with previous work showing that matrix solution conditions and crystal morphology influenced analyte incorporation into the matrix crystal [42, 43]. MALDI imaging investigations have shown that hydration of tissue samples prior to matrix deposition improved analyte incorporation into crystals and that analyte signal from tissues may be further tuned by application of solvent vapor post-crystallization . Similarly, with TLT/VIC, discriminating mass spectral features were noted for the different crystal morphologies. Protein G, antibody, or antigen MS signals varied between TLT/VIC preparations; DMSO, and methanol VIC represented the extremes (Figure 2b and c). For methanol VIC, protonated ions derived mostly from the albumin while DMSO VIC produced spectra similar to the DD approach with protonated species of all surface proteins present. Similar results have been observed from other endogenous plasma protein immunoassays (unpublished).
The results also show the presence of multiply charged albumin ions differed for CHCA microcrystals versus films as described in Table S3. The data indicate that singly and doubly charged albumin species dominated spectra from surfaces with crystal films, while microcrystals provided spectra featuring albumin signals with higher charge. Decreased protonation from DMSO- and DMF-treated surfaces may have derived from charge neutralization by solvents with higher pKa or by increased competition between ablated surface components in the gas-phase plume . Interestingly, similar experiments performed in the negative ion mode showed that decreased protonation for DMSO- and DMF-treated surfaces correlated with a higher proportion of negatively charged ions (Figure 2d). This suggests that protonated ion yields in CHCA films may have been reduced by the presence of photoelectrons or electrons that derived from metal surfaces .
The results thus far provide insight into solvent dependent nucleation and crystal growth characteristics that affected endpoint CHCA products (e.g., microcrystals versus planar spherulites). Previous investigations have shown that crystalline transitions of vapor penetrated organic materials were highly dependent upon solubility, initial crystallinity of the material, and temperature . Despite having low vapor pressures, the onset of VIC occurred by exposure to solvent vapor at room temperature for 60 s and 40 s for DMSO and DMF, respectively (Table S2), likely due to high CHCA solubility. At the onset of VIC the surfaces turned glassy and crystallization was completed by drying in the absence of the vapor at ambient temperature and pressure. CHCA crystal formation as a function of time using DMSO VIC was captured via light microscope (Figure S4). The results may indicate that DMSO vapor plasticized CHCA into a system that mimicked a supercooled liquid environment near its glass transition state . Larger crystals separated by less defined films in DMF VIC (Figure 1) suggest that high heat loss at early nucleation sites promoted bulk matrix transport to the growing crystal where 3D crystal growth occurred by Ostwald ripening. The presence of all surface bound components in the mass spectra of DMF and DMSO treated surfaces suggests that vapor penetrants with high solubility easily diffused through the TLT layer to the surface to incorporate surface analytes into the growing crystals.
Methanol, ethanol, and acetone solvents have higher vapor pressures but lower CHCA solubility compared with DMSO and DMF. To increase the VIC transformation rate, these solvents required heating, presumably to better solvate the TLT surface. If complete dissolution of the CHCA particles from the TLT surface occurred in these solvents, then rapid drying (<1 s) supports a high nucleation rate that limited the size of crystals. The selectivity of albumin in MS spectra with methanol VIC might suggest that complete dissolution of the surface proteins did not occur and the solvent served to change the molecular packing of already present crystals. In this case, the extent of analyte incorporation may be dependent on factors such as analyte/solvent compatibility and the location of the antigen near the matrix layer.
3.3 Assay Optimization
Matrix Surface Density
For DMSO VIC the signal decay beyond the optimal value linearly correlated with increased matrix density. Microscope images show that the spherulite planar dimensions were relatively consistent at all matrix surface densities (Figure 3c). Thickness measurements with AFM show that spherulites at 37 ng/mm2 were ~28 ± 3 nm (Figure 3d, left) in agreement with 26 nm thickness based on CHCA density calculations. The results indicate that TLT matrix deposition allowed quantitative transfer of matrix and M/A control in a manner that was consistent across the entire biochip surface.
Laser Irradiation Optimization
Immunoassay Surface Density
Previous investigations have shown that antibody fractional activity can be maximized by orienting the antigenic binding region toward the solution interface . The monolayer used here was composed of alkanethiolates terminated with either NTA or EG3 ligands. The EG3 group served as a built-in blocking reagent to minimize non-specific interactions, while the NTA ligand provides an anchor for histidine-tagged protein G which enables proper antibody orientation through the non-antigenic Fc domain . Investigations have shown that controlled surface density of immobilized binding motifs enhanced surface interactions by minimizing steric exclusion . A series of albumin MS-immunoassays were performed with biochips of varying NTA surface density (1 %–30 % mol/mol). A plot of the deconvoluted albumin signal (Figure 4c) indicates maximum intensity at ~15 % NTA:EG3, consistent with previous work on enzyme-linked immunosorbent assay (ELISA)s that suggest <25 % of randomly immobilized antibodies are surface active . At 15 % NTA density we estimate there is ~1.16 × 10–10 NTA mol/cm2 . With an estimated laser spot size of ~100 μm2 and a 2:1 albumin:NTA stoichiometry, we estimate that a maximum of 23 amol albumin were consumed per laser pulse highlighting the sensitivity of MALDI-TOF measurements of intact proteins from SAM immunoassays.
3.4 Assay Performance
Overall, we envision that our SAMDI immunoassays with TLT/VIC support reproducible protein analysis with the precision necessary for routine molecular diagnostic environments. Similar to other MALDI-immunoassay formats [7, 9, 10, 11, 12, 13], we anticipate diverse applications for this technology; including those that help to rationalize diverse sources of biological variation (e.g., alternative splicing, polymorphisms, and post-translational modifications). Future work will continue to develop this approach for quantitation of monolayer-bound proteins without the use of internal standards. Furthermore, we will evaluate the unique crystallization tendencies of other matrices and determine how matrix selection may improve detection of different molecular species in both the positive and negative ion modes.
We report the combination of antibody-coupled SAM biochips with a facile and rapid method for matrix deposition which enables sensitive and reproducible MS analysis of immobilized antigens. Robust label-free processing with MALDI-TOF is demonstrated using rationally designed substrates and methods that control matrix/analyte co-crystallization. The TLT method provides uniform matrix deposition over large surface areas without disturbing the immobilized analyte, while requiring only small (50–100 μg) quantities of matrix. Coupling the TLT method to organic solvent VIC demonstrated improved MS signal intensities and intra- and inter-assay reproducibilities for proteins bound to SAM biochips compared with DD. The work also highlights how controlling solvent, matrix, and analyte interactions during crystallization impacts analyte charging and S/N. The data presented here suggest that TLT matrix deposition followed by organic solvent VIC may also be employed in diverse diagnostic SAMDI applications for sensitive, reproducible, and label-free protein molecular diagnostics.
The authors acknowledge support for this work by the University of Texas Southwestern Medical Center (UTSW), University of Texas at Dallas (UTD), and the John L. Roach Scholarship in Biomedical Research. The authors thank the Cleanroom Research Laboratory at UTD and the Molecular and Cellular Imaging Facility at UTSW for assistance with imaging studies.
- 7.Cornelius, R.M., Shankar, S.P., Brash, J.L., Babensee, J.E.: Immunoblot analysis of proteins associated with self-assembled monolayer surfaces of defined chemistries. J. Biomed. Mater. Res. A 98, 7–18 (2011)Google Scholar
- 19.Bouschen, W., Schulz, O., Eikel, D., Spengler, B.: Matrix vapor deposition/recrystallization and dedicated spray preparation for high-resolution scanning microprobe matrix-assisted laser desorption/ionization imaging mass spectrometry (SMALDI-MS) of tissue and single cells. Rapid Commun. Mass Spectrom. 24, 355–364 (2010)CrossRefGoogle Scholar
- 20.Chen, Y., Allegood, J., Liu, Y., Wang, E., Cachon-Gonzalez, B., Cox, T.M., Merrill Jr., A.H., Sullards, M.C.: Imaging MALDI mass spectrometry using an oscillating capillary nebulizer matrix coating system and its application to analysis of lipids in brain from a mouse model of Tay-Sachs/Sandhoff disease. Anal. Chem. 80, 2780–2788 (2008)CrossRefGoogle Scholar
- 27.Wei, H., Nolkrantz, K., Powell, D.H., Woods, J.H., Ko, M.C., Kennedy, R.T.: Electrospray sample deposition for matrix-assisted laser desorption/ionization (MALDI) and atmospheric pressure MALDI mass spectrometry with attomole detection limits. Rapid Commun. Mass Spectrom. 18, 1193–1200 (2004)CrossRefGoogle Scholar
- 37.Tarzi, O.I., Nonami, H., Erra-Balsells, R.: The effect of temperature on the stability of compounds used as UV-MALDI-MS matrix: 2,5-dihydroxybenzoic acid, 2,4,6-trihydroxyacetophenone, ⟼-cyano-4-hydroxycinnamic acid, 3,5-dimethoxy-4-hydroxycinnamic acid, nor-harmane, and harmane. J. Mass Spectrom. 44, 260–277 (2009)CrossRefGoogle Scholar
- 39.Poetsch, A., Schlusener, D., Florizone, C., Eltis, L., Menzel, C., Rogner, M., Steinert, K., Roth, U.: Improved identification of membrane proteins by MALDI-TOF MS/MS using vacuum sublimated matrix spots on an ultraphobic chip surface. J. Biomol. Tech. 19, 129–138 (2008)Google Scholar
- 49.Peluso, P., Wilson, D.S., Do, D., Tran, H., Venkatasubbaiah, M., Quincy, D., Heidecker, B., Poindexter, K., Tolani, N., Phelan, M., Witte, K., Jung, L.S., Wagner, P., Nock, S.: Optimizing antibody immobilization strategies for the construction of protein microarrays. Anal. Biochem. 312, 113–124 (2003)CrossRefGoogle Scholar
- 50.Mrksich, M., Whitesides, G.M.: Using self-assembled monolayers that present oligo(ethylene glycol) groups to control the interactions of proteins with surfaces. Poly (Ethylene Glycol) 680, 361–373 (1997)Google Scholar
- 54.Sakai, S., Matsuda, R., Adachi, R., Akiyama, H., Maitani, T., Ohno, Y., Oka, M., Abe, A., Seiki, K., Oda, H., Shiomi, K., Urisu, A.: Interlaboratory evaluation of two enzyme-linked immunosorbent assay kits for the determination of crustacean protein in processed foods. J. AOAC Int. 91, 123–129 (2008)Google Scholar