Anal. Chem. 2003, 75, 4801-4810
Total Protein Determinations by Particle Beam/ Hollow Cathode Optical Emission Spectroscopy Fuxia Jin,† Keith Lenghaus,‡ James Hickman,*,‡ and R. Kenneth Marcus*,†
Department of Chemistry, Howard L. Hunter Chemical Laboratories, Clemson University, Clemson, South Carolina 29634-0973, and Department of Bioengineering, Hybrid Neuronal Systems Laboratory, Clemson University, Clemson, South Carolina 29634-0905
A novel method for quantitative total protein determinations is presented. Total protein content is determined by particle beam/hollow cathode optical emission spectroscopy (PB/HC-OES) through monitoring of carbon atomic emission. The PB/HC-OES offers such advantages as ease of operation, exclusion of labor-intensive sample pretreatment processes, rapid analysis, high sensitivity, and low detection limit. The method could also be adapted to be integrated to current microfluidics devices. Parametric optimization for sample introduction, nebulization, desolvation, and hollow cathode source conditions is performed for the analysis of aqueous bovine serum albumin solutions. Response curves of C (I) 193.0-nm emission were obtained under the optimized conditions with both 10% HCl and 100 µg/mL KCl added to the sample matrix as potential carriers. The detection limit for triplicate injections of bovine serum albumin standards was found to be on the single-nanogram level with 200-µL injections. The addition of KCl significantly improved the sensitivity, supporting the proposed “carrier effect” of chloride salts in the particle transport process. Results obtained here suggest a range of applications for the use of the PB/HCOES source for total protein determinations; emphasis here is future use in assessing protein quantification in microfluidic systems. The interaction of native proteins with materials of various forms has long been an area of investigation, particularly in the field of prosthetic implants. As new analytical devices are being developed specifically for biomedical applications, there are new challenges arising concerning basic protein adsorption mechanisms. A novel approach to performing total protein analyses in flowing systems is described. Microfabrication is an important aspect of modern science and technology.1,2 Microfluidic systems have gained increasing interest over the past decade, in the area of biochemical analysis in particular;3-10 however, only a few †
Department of Chemistry. Department of Bioengineering. (1) Heitmann, D.; Kotthaus, J. P. Phys. Today 1993, 46, 56-63. (2) Bryzek, J. Sens. Actuators, A 1996, 56, 1-9. (3) Madou, M. Fundamentals of Microfabrication; CRC Press: Boca Raton, FL, 1997. (4) Regnier, F. E.; He, B.; Lin, S.; Busse, J. Trends Biotechnol. 1999, 17, 101106. ‡
10.1021/ac034109b CCC: $25.00 Published on Web 08/19/2003
© 2003 American Chemical Society
prototypes for the direct analysis of biological fluids in the clinical and forensic fields have realized commercialization. Perhaps the largest limitation in the further implementation of microfluidic systems in bioanalyses is the tendency for polypeptides and proteins to adsorb to many of the materials from which MEMS devices are constructed.11 While passing through the flow channels in a microfluidic system, proteins in samples can adsorb to the surface materials both through biologically specific and nonspecific interactions.12,13 The extent of adsorption will vary due to the nature of the protein, the material, and the transport solution composition.14 As a first step, it is necessary that protein interactions with proposed MEMS materials be evaluated and quantified during the development and fabrication of candidate systems. One method to assess the extent of the interaction between proteins and potential bioanalysis or implant materials is a “total protein” determination of test solutions before and after exposure to the relevant surface. The practicalities of this sort of determination require high sensitivity for protein content in volume-limited solutions that are complicated by the presence of buffers and perhaps high salt content. To a first approximation, the ability to distinguish between different proteins is not required, as proteinspecific experiments can be used in this evaluation stage. This is certainly easier than the case of whole serum determinations, for example. The ability to perform the determinations with a minimum of chemical manipulations in a short period of time would be desirable. In the ideal case, this method of protein determination would be directly coupled to a fluidic test bed for real-time monitoring. However, most protein detection methodologies are time-consuming and plagued by interferences present in most protein-containing matrixes. (5) Van den Berg, A.; Lammerink, T. S. J. Top. Curr. Chem. 1998, 194, 2149. (6) Qin, D.; Xia, Y.; Rogers, J. A.; Jackman, R. J.; Zhao, X.; Whitesides, G. M. Top. Curr. Chem. 1998, 194, 1-20. (7) Delamarche, E.; Bernard, A.; Schmid, H.; Bietsch, A.; Michel, B.; Biebuyck, H. J. Am. Chem. Soc. 1998, 120, 500-508. (8) Fang, Z.; Fang, Q. Fresenius J. Anal. Chem. 2001, 370, 978-983. (9) Judy, J. W. Smart Mater. Struct. 2001, 1115-1134. (10) Verpoorte, E. Electrophoresis 2002, 23, 677-712. (11) Xiong, L.; Regnier, F. E. J. Chromatogr., A 2001, 924, 165-176. (12) Leckband, D. E.; Sivasankar, S. Colloids Surf., B 1999, 14, 83-97. (13) Butler, J. E. In Structure of Antigens; van Regenmortel, M. H. V., Ed.; CRC Press: Boca Raton, FL, 1992; Vol. 1, pp 209-259. (14) Slack, S. M.; Horbett, T. A. In Proteins at Interfaces; Horbett, T. A., Brash, J. L. Eds.; ACS Symposium Series 602; American Chemical Society: Washington, DC, 1995; Vol. 2, 112-128.
Analytical Chemistry, Vol. 75, No. 18, September 15, 2003 4801
Over the past 20 years, a wide range of colorimetric assays have gained increased attention for the identification and quantification of proteins.15 The Biuret method uses the reaction between cupric ions and proteins in alkaline solution to form purple protein-cuprous ion complexes that are subsequently measured by spectrophotometry.16 Although unaffected by the composition of the protein, the Biuret method is somewhat insensitive due to its dependence on different types of buffers and concentrations of alkali metals as well as the interference from other reducing agents and copper chelates.17,18 The procedure to eliminate these interferences is not only labor-intensive but also time-consuming. The Biuret method typically needs a sample size of 1 mL containing 1-10 mg of proteins with a detection limit of 1 mg of protein. The overall analysis time is over 30 min; it could be longer, though, depending on the sample incubation period. The Biuret method can provide satisfactory results that are comparable to results obtained from the Kjeldahl method,19 which is considered a benchmark method for crude protein determinations. The Kjeldahl method is composed of three distinct steps, digestion, distillation, and titration, which tend to be timeconsuming and have large sample consumption requirements. The proteins are digested with concentrated sulfuric acid in the presence of catalysts. The nitrogen present in the protein molecules is converted into ammonium sulfate, which is quantified by an acid/base titration or colorimetry. The typical analysis time for Kjeldahl method is 1 h to several hours, and about 0.5-5 g of crude protein samples is needed depending upon the nitrogen content.20 The detection limit of the Kjeldahl method is ∼0.1 mg of N/L. As an improvement to the Biuret reaction, the Lowry method uses a two-step reaction in which the Cu2+ ions in alkaline solution first react with proteins to produce Cu1+.21 Addition and subsequent reduction of Folin-Ciocalteu reagents (phosphomolybdate and phosphotungstate) is completed after the alkaline solution reaction. The blue color of the reduced Folin-Ciocalteu reagent is readily detected spectrophotometrically in the range of 500750 nm. The major disadvantage of the Lowry method is the narrow pH range of ∼10-10.5 where accurate results can be obtained. Although the Lowry method is more sensitive than the Biuret reaction method, it takes more time and is susceptible to many interfering compounds such as detergents, carbohydrates, glycerol, EDTA, tricine, and tris(hydroxymethyl)aminomethane (Tris) buffers, potassium compounds, disulfide compounds, magnesium, and calcium. Most of these species are commonly used in buffers for preparing protein solutions. In addition, the composition of the protein plays an important role in the determination of protein concentration as specific amino acids effect the color development in the Lowry reaction.21 The Lowry method usually takes over 1 h to complete, requiring a sample volume of 0.1 mL
of protein solutions containing 50-500 µg of protein. The detection limit of the Lowry method is 20-50 µg of protein. The bicinchoninic acid (BCA) assay was developed as a variation of the Lowry assay.22 In the BCA assay, the first step is still a Biuret reaction, which reduces Cu2+ to Cu+. In the second step, BCA forms a complex with Cu+ that is purple and is detectable via absorbance at 562 nm. The use of BCA instead of the Folin-Ciocalteu reagent makes the analysis less susceptible to interferences from common buffer substances. The BCA assay has very little variation in response to different proteins and also has a broad linear working range for protein concentrations. However, the reaction does not proceed to completion when performed at room temperature or human body temperature of 37 °C. Although the BCA assay can be used for protein measurements in very dilute solutions,23 it still suffers from interference from a wide variety of substances including EDTA, H2O2, and some biogenic amines. The typical sample size for the BCA assay is 0.1 mL of samples containing ∼5-300 µg/mL proteins. The BCA assay process takes more than 30 min and achieves a detection limit of 5 µg of protein/mL. The Bradford assay24,25 is based on the equilibrium between three forms of Coomassie Blue G dye. Under strongly acidic conditions, the dye is most stable in its doubly protonated red form. Upon binding to proteins, however, it is most stable in its unprotonated blue form. The Bradford assay is faster, involves fewer mixing steps, does not require heating, and gives a more stable colorimetric response than the assays described above. The assay is prone to influence from nonprotein sources, particularly detergents, and becomes progressively more nonlinear at the high end of its useful protein concentration range. The response is also protein-dependent, varying with the amino acid composition of the protein. The Bradford assay typically goes to completion within 30 min and needs a sample size of ∼5 mL of solution containing 20-140 µg of proteins. The detection limit of the Bradford assay is also at the single microgram of protein per milliliter level. Another common strategy for protein detection is the silver staining technique.26 The technique involves the saturation of gels with silver ions and subsequent formation of metallic silver through reduction of the protein-bound metal ions. This method has several disadvantages including high complexity, poor gelto-gel reproducibility, narrow linear dynamic range (typically only a factor of 10), and poor sensitivity.27 On the other hand, the actual detection limit of the silver staining assay is very low, such that even single nanograms of protein can be detected. The silver staining technique takes tens of hours to finish. The sample size of this method is hundreds of milliliters, and the method has a low detection limit of single nanogram of protein. Mertens and co-workers28 evaluated the total protein concentration in enzymes via the determination of sulfur by total reflection X-ray fluorescence spectrometry (TXRF). This approach is based
(15) Sapan, C. V.; Lundblad, R. L.; Price, N. C. Biotechnol. Appl. Biochem. 1999, 29, 99-108. (16) Cotton, F.; DeLobbe, E.; Gulbis, B. Clin. Biochem. 1997, 30, 313-314. (17) Shrivastaw, K. P.; Singh, S.; Sharma, S. B.; Sokhey, J. Biologicals 1995, 23, 299-300. (18) Reichardt, W.; Eckert, B. Nahrung 1991, 35, 731-738. (19) Lof, A. L.; Gustafson, G.; Novak, V.; Engman, L.; Mickaelsson, M. Vox. Sang. 1992, 63, 172-177. (20) Casal, J. A.; Vermaat, J. E.; Wiegman, F. Aquatic Botany 2000, 67, 61-67. (21) Viner, R. I.; Huhmer, A. F. R.; Bigelow, D. J.; Schoneich, C. Free Radical Res. 1996, 24. 243-259.
(22) Brenner, A. J.; Harris, E. D. Anal. Biochem. 1995, 226, 80-84. (23) Schoel, B.; Welzel, M.; Kaufmann, S. H. F. J. Biochem. Biophys. Methods 1995, 30, 199-206. (24) Zor, T.; Seliger, Z. Anal. Biochem. 1996, 236, 302-308. (25) Atherton, B. A.; Cunningham, E. L.; Splittgerber, A. G. Anal. Biochem. 1996, 233, 160-168. (26) Sinha, P.; Poland, J.; Schnolzer, M.; Rabilloud, T. Proteomics 2001, 1, 835840. (27) Patton, W. F. J. Chromatogr., B 2002, 771, 3-31. (28) Mertens, M.; Rittmeyer, C.; Kolbesen, B. O. Spectrochim. Acta, Part B 2001, 56, 2157-2164.
4802
Analytical Chemistry, Vol. 75, No. 18, September 15, 2003
on the fact that sulfur is a component of two of the amino acids, methionine and cysteine. The advantage of this technique is the exclusion of sample pretreatment. A determination of multiple elements allows a cross-check regarding conventional quantitative determination of protein concentration. However, proteins that have high molecular weights or are stored in concentrated buffer solutions yield poor results. According to Mertens and coworkers,28 the TXRF method only consumes 4 µL of samples containing ∼10 µmol of protein/L, with an overall measuring time of ∼20 min. Detection limits of 0.5-5 ng of sulfur were calculated for the measured enzymes. In addition to the aforementioned spectrophotometric methods of protein determination, other spectroscopic methods also find their way into this interesting arena. Tsenkova and co-workers29 studied the potential of near-infrared spectroscopy for total protein determinations in nonhomogenized milk. In their study, the influence of the spectral region, sample thickness, and spectral data treatment on the accuracy of determination was investigated. Yao and co-workers30 first reported the determination of proteins based on the interaction with carboxyarsenazo (CAA) with detection by Rayleigh light scattering (RLS). They found that the weak RLS of CAA was enhanced greatly by the addition of proteins resulting in three characteristic peaks. Based on the RLS enhancement, nine proteins were evaluated and the method was proven to be very sensitive (0.10-15.5 µg/mL) for bovine serum albumin (BSA) samples, with the advantages of rapid determinations (240 °C), both solvents experience appreciable in-capillary vaporization and unstable operation. The final nebulizer parameter is the concentric helium gas flow rate. The nebulizer gas serves to provide indirect heating of the solution in the silica capillary as well as to break up the solution stream exiting the capillary. The nebulizer gas flow also plays a role in terms of aerosol transport through the spray chamber and momentum separator. Across the operable range of He flow rates, 200-800 mL/min, the C (I) 193.0-nm emission response for BSA sample increases only slightly, but overall, the intensity and stability are best at the highest flow rate. Below 200 mL/min, stable nebulization could not be achieved. Based on the results of these nebulizer optimization studies, conditions involving a solution flow rate of 0.7 mL/min, a nebulizer tip temperature of 220 °C, and a nebulizer gas glow rate of 800 mL/min are employed throughout the remainder of these studies. Effect of Desolvation Conditions on Analyte Response. The particle beam interface achieves particle separation from the solvent in two steps: desolvation and momentum separation. Through the combination of heating and sheath gas dispersion, aerosols are dried in the high-pressure (100-600 Torr) desolvation chamber with the analyte particles carried in the gas flow through the differentially pumped momentum separator. The degree of aerosol enrichment (i.e., the skimming of the He and solvent vapors) is based on the pressure differential. Gas dynamics also play a role in transporting the aerosol across the length of the spray chamber. The C (I) response was monitored as a function of the auxiliary He gas flow rate across a range of nebulizer flow rates(a plot of which appears as Figure S3 in the Supporting Information). Two interesting effects are seen. First, the addition Analytical Chemistry, Vol. 75, No. 18, September 15, 2003
4807
of the auxiliary flow improves the relative analytical responses by up to ∼20% in each case when only the nebulizer flow provides gas to the system. Second, there is an optimum flow rate that is dependent on the nebulizer flow. This suggests that there is a fixed range of total gas input that yields the best particle transport through the interface. It is also interesting to note that the highest flow rates result in the greatest degree of imprecision, which is likely reflective of aerodynamic turbulence within the spray chamber. Effect of HC Source Block Temperature on Analyte Response. The hollow cathode temperature has an important effect on the mode of entrance of analyte atoms/molecules into the plasma of the hollow cathode.46 According to Slevin and Harrision,47 three atomization modes exist in hollow cathode glow discharges depending on the surface temperatures of the HC. If the temperature of the HC source is sufficiently low, cathodic sputtering is the major mechanism for atomization. When the HC source has a modest temperature (likely ∼100 °C), the mechanisms include both cathodic sputtering and selective vaporization of volatile compounds. At an elevated HC temperature (i.e., in a heated HC source), thermal volatilization contributes substantially to the atomization processes. Early PB/GD studies clearly indicated the need to provide a thermal component for sample atomization, beyond simple cathodic sputtering. Collision of the analyte particles with the heated HC walls provides sufficient energy for thermal vaporization/atomization followed by excitation and ionization in the gas phase. Dempster and co-workers46 demonstrated the presence of species-specific threshold temperatures for a number of copper salts, above which the emission intensity increased owing to increased vaporization/atomization efficiency. However, at excessively high HC temperatures, the vaporization efficiency decreased due to the deposition of the analyte on the heated HC wall or pyrolysis of the analytes.46 Certainly, the latter process is of high concern in the PB/HCOES analysis of proteins. The effect of source block temperature on C (I) emission response for BSA solutions was determined across the temperature range of 130-280 °C for solvent compositions of 20 and 100% methanol (a plot of which appears as Figure S4 in the Supporting Information). There is an optimized C (I) emission response at a HC temperature of ∼160 °C for both solvent systems, much lower than values of 250-350 °C found for copper salt solutions in the previous study.46 To a first approximation, this may be due to the lower melting points of protein samples versus inorganic salts. When heated, BSA forms soluble aggregates through disulfide and noncovalent bonds.48 At higher temperatures, it is possible that the gelation of BSA enables the formation of larger molecular aggregates, which deposit on the HC walls and eventually undergo pyrolysis. Once pyrolyzed, the proteins have greatly reduced volatility and in fact would be much more difficult to dissociate if they were to enter the gas phase. The occurrence of pyrolysis at high block temperatures was indicated by black deposits on the HC wall and the boron nitride support housing. (46) Dempster, M. A.; Davis, W. C.; Marcus, R. K.; Cable-Dunlap, P. R. J. Anal. At. Spectrom. 2001, 16, 1080-1086. (47) Slevin, P. J.; Harrison, W. W. Appl. Spectrosc. Rev. 1975, 10, 201-255. (48) Matsudomi, N.; Oshita, T.; Kobayashi, K.; Kinsella, J. E. J. Agric. Food Chem. 1993, 41, 1053-1057.
4808 Analytical Chemistry, Vol. 75, No. 18, September 15, 2003
Figure 5. Analyte response curves for C (I) (193.0 nm) emission intensity in bovine serum albumin standards in 10% HCl. Solvent flow rate 0.7 mL/min; nebulizer tip temperature 220 °C; HC block temperature 160 °C; nebulizer gas flow rate ∼800 mL/min; desolvation chamber flow rate ∼85 mL/min; source pressure 1 Torr He; discharge current 60 mA. (a) linear response curve, (b) logarithmic-scale plot.
Figures of Merit for Protein Analysis Based on Carbon Emission. Having optimized the various aspects of protein sample nebulization, particle transport, vaporization, and optical emission response, general figures of merit were determined for bovine serum albumin. The analytical response curve (linear least squares) for C (I) 193.0-nm emission of a sample of bovine serum albumin dissolved in 10% HCl is shown in Figure 5a. The solvent used is 100% methanol. The standards used to prepare the response curve ranged from 0.025 to 100 µg/mL BSA. In comparison to results obtained prior to this systematic optimization study, the slope of this calibration curve exhibits a 15-fold increase in sensitivity. Error bars represent the range of the transient signal intensities for triplicate injections, with the variability for most of the triplicate injections being
