Anal. Chem. 2003, 75, 5703-5709
Raman Detection of Proteomic Analytes Dongmao Zhang,† Yong Xie,† Melissa F. Mrozek,† Corasi Ortiz,† V. Jo Davisson,‡ and Dor Ben-Amotz*,†
Department of Chemistry and Department of Medicinal Chemistry and Molecular Pharmacology, Purdue University, West Lafayette, Indianapolis, Indiana 47907
The compatibility of nonenhanced Raman spectroscopy with chromatographic and mass spectroscopic proteomic sensing is demonstrated for the first time. High-quality normal Raman spectra are derived from protein solutions with concentrations down to 1 µM and 1 fmol of protein nondestructively probed within the excitation laser beam. These results are obtained using a drop coating deposition Raman (DCDR) method in which the solution of interest is microdeposited (or microprinted) on a compatible substrate, followed by solvent evaporation and backscattering detection. Representative applications include the DCDR detection of insulin derived from an HPLC fraction, nondestructive DCDR followed by MALDI-TOF of lysozyme, the DCDR detection of protein spots deposited using an ink-jet microprinter, and the identification of spectral differences between glycan isomers of equal mass (such as those derived from posttranslationally modified proteins). Raman spectroscopy can be a powerful probe of structure and function in biochemical systems.1 However, no previous studies have succeeded in demonstrating the nonenhanced Raman detection of proteins under conditions that are compatible with current chromatographic and mass spectroscopic proteomic separation and sensing methods. In particular, the best previously reported normal Raman protein detection limit is ∼1 nmol in a lysozyme solution of 50 µM concentration.2 In this work, we report a detection limit improvement of over 1000 times for lysozyme derived from solutions down to 1 µM concentration. These results are obtained using a drop coating deposition Raman (DCDR) method in which a microvolume of solution is deposited on a suitable substrate, followed by solvent evaporation and nondestructive Raman detection. We demonstrate the compatibility of DCDR sensing with HPLC separation and MALDI-TOF detection, as well as the identification of glycan isomers (of equal mass), such as those cleaved from posttranslationally modified proteins.3 Furthermore, although most of the DCDR measurements reported * Corresponding author. E-mail:
[email protected]. † Department of Chemistry. ‡ Department of Medicinal Chemistry and Molecular Pharmacology. (1) For example, see: (a) Carter, E. A.; Edwards, H. G. M. Practical Spectrosc. 2001, 24, 421-475. (b) Mulvaney, S. P.; Keating, C. D. Anal. Chem. 2000, 72, 145R-157R. (c) Lyon, L. A.; Keating, C. D.; Fox, A. P.; Baker, B. E.; He, L.; Nicewarner, S. R.; Mulvaney, S. P.; Natan, M. J. Anal. Chem. 1998, 70, 341R-361R. (d) Carey, P. R. Biochemical Applications of Raman and Resonance Raman Spectroscopies; Academic Press: New York, 1982. (2) Pelletier, M. J.; Altkorn, R. Anal. Chem. 2001, 73, 1393-1397. (3) Mechref, Y.; Novotny, M. V. Anal. Chem. 1998, 70, 455-463. 10.1021/ac0345087 CCC: $25.00 Published on Web 09/20/2003
© 2003 American Chemical Society
in this work have been obtained from manual micropipet depositions of 1-10-µL solution volumes to produce protein spots of 0.55-mm diameter, preliminary measurements performed using inkjet microprinting of an 8-nL solution volume, to produce protein spots of 15-µm diameter containing less than 25 fmol of protein, demonstrate the compatibility of DCDR sensing with high-density protein microarrays. The previous Raman studies of very low concentration protein solutions were performed using methods that are quite different from DCDR. Carey et al.4 obtained Raman spectra from a 100 µM aqueous lysozyme solution (of 40-µL volume) and Raman difference spectra of protein-ligand complexes in the 200-400 µM range using an axial transmissive (AT) spectrograph. More recently, Pelletier and Altkorn2 demonstrated a multipass enhancement factor of 500 in aqueous solutions using a liquid-core optical fiber Raman cell coupled to an AT spectrograph, enabling the detection of a 54 µM (24 µL) lysozyme solution. The much lower detection limits demonstrated in the present work result from the fact that DCDR preconcentrates and localizes the protein derived from very small volumes of very dilute solutions. Thus, DCDR may be viewed as a rapid protein preconcentration method.5 The total time required for DCDR deposition and solvent evaporation (∼10 min for microliter depositions and ∼1 min for nanoliter depositions) is comparable to the time required for Raman signal collection. Further improvement in throughput may be achieved by depositing smaller (faster drying) spots and using higher laser powers. Although the DCDR preconcentration (solvent evaporation) process produces protein deposits that are in a solidlike state, they appear to remain well hydrated. More specifically, the DCDR spectrum of lysozyme is essentially identical to previously reported Raman spectra of dilute lysozyme solutions.2,6 In addition, the DCDR spectrum of insulin is the same as that attributed to insulin in its native state and clearly differs from the spectrum attributed to denatured insulin.7 For smaller molecules, our results suggest that DCDR deposits may be in either a solution-like or a crystallike state. For example, the DCDR spectrum of glucose looks like the Raman spectrum of a saturated glucose solution and quite (4) Dong, J.; Dinakarpandian, D.; Carey, P. R. Appl. Spectrosc. 1998, 52, 11171122. (5) Walker, P. A., III; Kowalchyk, W. K.; Morris, M. D. Anal. Chem. 1995, 67, 4255-4260. (6) Dong, J.; Dinakarpandian, D.; Carey, P. R. Appl. Spectrosc. 1998, 52, 11171122. (7) (a) Yu, N.-T.; Liu, C. S. J. Am. Chem. Soc. 1972, 94, 3250-3251. (b).Yu, N.-T.; Liu, C. S.; O’Shea, D. C. J. Mol. Biol. 1972, 70, 117-132. (c). Jacob, J.; Krafft, C.; Welfle, H.; Welfle, H.; Saenger, W. Acta Crystallogr. 1998, D54, 74-808.
Analytical Chemistry, Vol. 75, No. 21, November 1, 2003 5703
different from that of glucose crystals. On the other hand, the DCDR spectrum of fructose more closely resembles that of crystalline fructose than that of a saturated aqueous fructose solution. It is also instructive to compare our DCDR results with previous pioneering work by McCreery that demonstrated nonenhanced Raman detection of monolayer or submonolayer quantities of small aromatic molecules electrochemically adsorbed on glassy carbon and ordered graphite.8 The latter method required liquid nitrogen cooling of the substrate and several hours of integration time (using a 488-nm, 18-mW laser). The DCDR method described in this work, on the other hand, is capable of detecting biomolecules passively adsorbed on a noncooled substrate with ∼100 times shorter integration times (using a 632nm, 12-mW laser). To achieve these results we have found that it is critically important to optimize the properties of the substrate. In particular, an ideal DCDR substrate should have (i) low optical absorbance, (ii) high optical reflectance, (iii) little or no interfering background signals, and most importantly (iv) a nonwetting interaction with the analyte solution (i.e., low solvent affinity). The latter requirement is critical because a wetting substrate will spread the deposited solution over a larger area while a nonwetting substrate concentrates the deposited analyte in a smaller area, significantly increasing the analyte surface density, and thus improving the Raman detection limit. Furthermore, protein deposits on nonwetting substrates often accumulate in a circular ring formed during the evaporation process once the protein reaches its saturation concentration. This well-known “coffee ring” effect9 further localizes the deposited protein and thus facilitates additional detection limit improvement. In this work, we describe results obtained using several DCDR substrates, including Tefloncoated stainless steel, nominally flat gold foil, and gold-coated glass, either with or without an organic self-assembled monolayer overcoat. Our best DCDR results are obtained using very thin (10b 5 5
a HPLC solvent with 48:52 volume ratio. b Spots of greater than 10 mm were also produced when depositing a 5-µL volume.
background (as has been observed with other polymers),20 while thicker Teflon coatings (>1 µm thick) produce Raman features with prominent peaks around 1381 and 733 cm-1. When sample solutions of the same volume are deposited on different substrates, the diameter of the droplet on the surface depends on the degree of affinity (wetting) between the solution and the substrate. Table 1 lists the average initial (before drying) spot sizes produced when 10-µL volumes of either H2O, acetonitrile/water (48:52 v/v, HPLC solvent), or methanol are deposited onto quartz, gold foil, Teflon-coated stainless steel, and SAMcoated gold. Note that the smallest spot sizes are very close to the theoretical limit expected on a nonwetting substrate if the initially deposited drop has the shape of a half sphere, in which case a 10-µL volume should produce a spot diameter of 3.4 mm. A suitable substrate should be resistant to chemical and thermal degradation. No evidence of damage to either the substrates or DCDR-deposited analytes was observed when the nominally flat gold foil, gold-coated glass, and Teflon-coated gold or stainless steel substrates were used. The only evidence for possible chemical degradation on any of our substrates was observed with glycans (6-NADL, 3-NADL) deposited onto a silver foil substrate, whose Raman spectra were found to change with time. Laser power-dependent studies reveal no evidence for optical/thermal damage of either the analytes or substrates used in this study (as discussed below). Raman Spectra. Figure 2 shows a white light image of a region of the Anchorchip substrate where a 5-µL volume of 1 mM aqueous glucose solution was deposited, after complete evaporation of the solvent. The bar represents 50 µm. Note that although the Anchorchip contains gold-coated islands, our DCDR measurements were performed only on the Teflon-coated area of the Anchorchip. Following solvent evaporation, the glucose appears to uniformly cover a surface area of ∼0.3 mm2. Raman spectra a and b in Figure 2 were obtained from this deposited sample. For comparison purposes, corresponding spectra of glucose obtained from saturated solution and microcrystals are shown as spectra c and d, respectively, which are also in good agreement with previously reported spectra.21,22 It is evident that spectra a and b, obtained from the deposited glucose sample, match that from a saturated solution (c) rather than crystalline (20) Mcanally, G. D.; Everall, N. J.; Chalmers, J. M.; Smith, W. E. Appl. Spectrosc. 2003, 57, 44-50. (21) For example: (a) Wells, H.; Atalla, R. H. J. Mol. Struct. 1990, 224, 385424. (b) Soderholm, S.; Roos, Y. H.; Meinander, N.; Hotokka, M. J. Raman Spectrosc. 1999, 30, 1009-1018. (22) For example: Mathlouthi, M.; Luu, D. V. Carbohydr. Res. 1980, 81, 203212.
5706 Analytical Chemistry, Vol. 75, No. 21, November 1, 2003
Figure 2. White light image obtained after deposition (and evaporation) of 5 µL of 1 mM glucose solution on a Teflon-coated stainless steel surface. Spectrum a was taken immediately after drying, while spectrum b was acquired after exposing the sample to atmosphere for 7 days. Spectra c and d were obtained from a saturated aqueous solution and a microcrytal of glucose, respectively. All spectra were collected with an integration time of 50 s (and the bar in the photograph represents 50 µm).
glucose (d). The stability of the deposited sample is illustrated by comparing spectra a and b, as spectrum a was obtained immediately following solvent evaporation, while spectrum b was obtained after storing the sample in atmosphere for 1 week. A previous study employed SERS for the detection and identification of monosaccharides such as glucose.14b The signalto-noise ratios of the normal Raman spectra obtained from the deposited samples shown in Figure 2 are comparable to those reported by Mrozek and Weaver14a on roughened silver electrodes, yet the sample deposited on the Anchorchip substrate here is 10 times lower in concentration. This phenomenon may be due in part to the low SERS enhancement of monosaccharides,14b,23 or a decreased sample density on the roughened silver surface at the laser probing area. The lowest concentration of glucose from which Raman spectra have be obtained using DCDR is ∼100 µM. At lower concentrations, the best Raman signal is obtained from “hot spots”, which appear to contain solid glucose deposits of
