Anal. Chem. 2001, 73, 1393-1397
Raman Sensitivity Enhancement for Aqueous Protein Samples Using a Liquid-Core Optical-Fiber Cell M. J. Pelletier*,† and Robert Altkorn*,‡
Kaiser Optical Systems, Inc., 371 Parkland Plaza, P.O. Box 983, Ann Arbor, Michigan 48106, Department of Physics and Astronomy, Northwestern University, 1801 Maple Avenue, Evanston, Illinois 60201
We have demonstrated a sensitivity enhancement factor of 500 in aqueous solutions using a liquid core optical fiber (LCOF) Raman cell made from Teflon-AF. We were able to collect a spectrum of 54 µM lysozyme with a signalto-noise ratio of 31 in the LCOF Raman cell using 24 mW of laser power and 3 min of integration time. The lysozyme Raman intensity was only 1% of the background Raman intensity from water, but the water-subtracted lysozyme spectrum was still shot-noise-limited and essentially free of nonrandom noise. The lack of nonrandom noise indicates that it should be possible to collect good quality Raman spectra of proteins such as lysozyme at even lower concentrations. The 2.4-µL sample volume of the LCOF Raman cell is an added benefit when limited quantities of sample are available. This volume of a 54 µM lysozyme solution corresponds to only 13 nanomoles or 1.9 µg of lysozyme. The sensitivity of optical methods such as Raman, fluorescence, or UV/vis absorption spectroscopies can be greatly enhanced by confining the sample inside an optical waveguide, such as the Teflon-AF liquid core optical fiber (LCOF) Raman cell used in this work. A waveguide uses multiple reflections to trap light in the sample for an extended distance, making the optical path length through the sample greater than it would normally be. In the case of Raman or fluorescence spectroscopy, light emitted by the sample is also trapped by the waveguide, making collection more efficient. Ideally, sensitivity is proportional to the effective optical path length. In this work, we demonstrate large sensitivity enhancements in Raman spectroscopy using an LCOF. We show the effects of this sensitivity enhancement on the Raman spectra of aqueous lysozyme solutions. We also compare our Raman spectra of aqueous lysozyme solutions to those collected in a traditional 90° scattering geometry. Enhanced sensitivity does not necessarily lead to reduced detection limits or improved signal-to-noise ratios, however. Sensitivity is defined as the change in analytical signal divided by the change in sample concentration. Increasing sensitivity may increase the noise intensity just as much as the signal intensity. * E-mail:
[email protected] (M.J.P.);
[email protected] (R.A.). † Kaiser Optical Systems, Inc.. ‡ Northwestern University. 10.1021/ac001220y CCC: $20.00 Published on Web 02/13/2001
© 2001 American Chemical Society
As a result, the signal-to-noise ratio and detection limit can sometimes be changed only slowly, if at all, by increasing sensitivity. This report explores the real benefits of the enhanced sensitivity from an LCOF, as well as the current obstacles to the full utilization of these benefits. EXPERIMENTAL SECTION Raman Instrument. Raman spectra were collected with a HoloProbe (Kaiser Optical Systems, Inc., Ann Arbor, MI) using 532-nm excitation. The HoloProbe1 and its fiber-optic probehead1,2 have been described previously. Briefly, a single-mode optical fiber delivered laser light from a 100 mW CW 532-nm laser (Coherent Inc.) to the fiber-optic probehead. A 10×, 0.25 NA Olympus microscope objective lens focused the collimated laser light from the probehead into the LCOF Raman cell. It also returned collimated Raman light from the LCOF Raman cell back into the probehead. A precision three-axis translator aligned the LCOF Raman cell with the objective lens. The probehead coupled the Raman light into a 0.27 NA, 100µm core-diameter graded-index optical fiber. This fiber delivered the Raman light to the axial transmissive spectrograph3 inside the HoloProbe. The spectrograph used a 50-µm entrance slit width, a holoplex transmission grating,4 and an 85-mm focal length lens to focus the Raman spectrum onto a charge-coupled device (CCD) camera. The CCD camera (Princeton Instruments, Trenton, NJ) used a 1024 × 256 pixel front-illuminated EEV detector chip (EEV Ltd., Essex, U.K.) operated at -40 °C. The pixel size was 27 µm × 27 µm. The axial transmissive spectrograph split the Raman spectrum into two halves and stacked the two halves at the CCD detector. As a result, the entire Raman spectrum from 100 to 4400 cm-1 was collected simultaneously. The holoprobe was calibrated with a HoloLab series calibration accessory (Kaiser Optical Systems, Inc.) using a previously described method.5,6 Briefly, the calibration procedure consists of calibrating the spectrograph wavelength, the holoprobe intensity (1) Owen, H.; Battey, D. E.; Pelletier, M. J.; Slater, J. B. Proc. SPIE 1995, 2406, 260-267. (2) Owen, H.; Tedesco, J. M.; Slater, J. B. U.S. Patent 5,377, 004, 1994. (3) Battey, D. E.; Slater, J. B.; Wludyka, R.; Owen, H.; Pallister, D. M.; Morris, M. D. Appl. Spectrosc. 1993, 47, 1913-1919. (4) Battey, D. E.; Owen, H.; Tedesco, J. M. U.S. Patent 5,559, 597, 1996. (5) Tedesco, J. M.; Davis, K. L. Proc. SPIE 1998, 3537, 200-212. (6) Davis, K. L.; Tedesco, J. M.; Shaver, J. M. Proc. SPIE 1999, 3608, 148156.
Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 1393
throughput as a function of wavelength, and the laser wavelength. Wavelength calibration was done by fitting selected atomic lines from a diffused neon-glow discharge to a third order polynomial. Intensity throughput was calibrated in units of photons per wavenumber using a diffused NIST-traceable incandescent source. Then the laser wavelength was accurately determined from the measured wavelength of the cyclohexane 801 cm-1 band. LCOF Raman Cell. The LCOF Raman cell has been described previously.7 Briefly, a Teflon-AF capillary 121.5 cm long, having an inside diameter of 50 µm (2.4 µL total volume), was glued into a stainless steel plug. Liquid exiting the Teflon-AF capillary at the plug was guided into the sample exit tube, also glued into the plug, by a 150-µm-thick stainless steel flow channel covered by a borosilicate glass window. The flow channel volume was estimated to be 0.08 µL. The borosilicate glass window also provided optical access to the fluid exit end of the Teflon-AF capillary. A threeaxis translator kinematically held the plug. This translator and a 10×, 0.25 NA Olympus microscope objective (focal length ,18 mm) formed an assembly that attached directly to the fiber-optic probehead. Translation of the capillary with respect to the objective was used to optimize the optical coupling of the capillary to the probehead. The fluid entrance end of the capillary (opposite the laser entrance end) was glued into a large bore Luer-lock syringe needle. Liquids were injected into the capillary using disposable syringes. Chemicals. Triple-crystallized, dialyzed, and lyophilized chicken egg white lysozyme was obtained from Sigma Chemical Company (St. Louis, MO). It was approximately 95% protein and 5% buffer salts (sodium acetate and chloride). Aqueous lysozyme solutions were prepared by dissolving the lyophilized lysozyme in water. No buffer was used, and the solutions were not filtered. The water used in the work was locally produced deionized water. Spectrophotometric grade acetonitrile was obtained from Aldrich Chemical Co., Inc. (Milwaukee, WI). Method for Water Subtraction from Lysozyme Spectra. Over- or undersubtraction of the water Raman spectrum from the aqueous lysozyme is difficult to avoid at low lysozyme concentrations. Plausible but distorted spectra are easily created using either manual or automatic subtraction procedures. We used a reproducible water-subtraction criterion to minimize operator subjectivity in the subtraction process. Acetonitrile was added to a single sample of aqueous lysozyme as an internal standard. After solvent subtraction, the peak height ratio of the amide I band from near 1660 cm-1 to the aromatic band near 1552 cm-1 was determined for this sample. We then used that ratio as the goal for all water subtractions from aqueous lysozyme solutions containing no acetonitrile. We expect this ratio to be very sensitive to the accuracy of solvent subtraction, because the water-bending vibration near 1640 cm-1 strongly overlaps the 1660 cm-1 band but not the 1552 cm-1 band. Unfortunately, the peak-height-ratio criterion used in this work may not be generally applicable for the subtraction of water contributions from Raman spectra of aqueous protein solutions. The peak height of the conformationally sensitive amide I band depends on protein secondary structure,8-10 which can, in turn, (7) (8) (9) (10)
Pelletier, M. J.; Altkorn, R. Appl. Spectrosc. 2000, 54. Williams, R. W. J. Mol. Biol. 1983, 166, 581-603. Berjot, M.; Marx, J.; Alix, A. J. P. J. Raman Spectrosc. 1987, 18, 289-300. Alix, A. J. P.; Pedanou, G.; Berjot, M. J. Mol. Struct. 1988, 174, 159-164.
1394
Analytical Chemistry, Vol. 73, No. 6, March 15, 2001
depend on numerous conditions, including protein concentration. The secondary structure of lysozyme is expected to be the same in all of the aqueous solutions analyzed in the present work, however. Lysozyme partially and reversibly unfolds in acetonitrile,11 but 11% acetonitrile caused no shift or peak height change in the lysozyme amide I band. RESULTS AND DISCUSSION Sensitivity and Noise. The LCOF Raman cell increases the measured Raman intensity by a factor called the sensitivity enhancement factor. The sensitivity enhancement factor is defined as the ratio of Raman intensity from an LCOF to that from a 1-cm cuvette when all other experimental factors are the same. The sensitivity enhancement factor is essentially the factor by which laser power or integration time can be reduced in an LCOF and still get the same Raman intensity as measured in a 1-cm cuvette. We used a 1-cm cuvette because this path length is greater than the depth-of-field of the fiber-optic probehead. The Raman intensity we measured from the 1-cm cuvette is, therefore, the maximum unenhanced Raman intensity we can obtain from the clear liquid using the fiber-optic probehead in a 180° backscattering geometry. Dispersive Raman measurements are usually shot-noiselimited. The water Raman background intensity, however, is much stronger than the analyte Raman signal for dilute aqueous solutions. The signal-to-noise ratio for the shot-noise-limited measurement of an analyte signal in the presence of a much larger background signal is described by eq 1:
S/N )
k[A]PtE
xk[A]PtE + xk′[W]PtE
≈
k[A]PtE
xk′[W]PtE
)
k[A]
xk′[W]
xPtE (1)
where S/N is the analyte signal-to-noise ratio; k, k′ are constants; [A] is the analyte concentration; P is the laser power; t is the photon detection time; E is the sensitivity enhancement factor; and [W] is the water concentration (55.6 molar). Eq 1 shows that the signal-to-noise ratio increases with the square root of laser power, integration time, and sensitivity enhancement factor. The signal-to-noise ratio increases linearly with analyte concentration, however. If the analyte signal were much larger than the water background, the signal-to-noise ratio would increase with the square root of analyte concentration. The signal-to-noise ratio of Raman spectra becomes independent of the number of detected photons as the dominant noise becomes the proportional nonrandom type. The LCOF enhancement factor does not improve signal-to-noise ratios, detection limits, and lower concentration limits for quantitative analysis when the measurement is already limited primarily by proportional nonrandom noise. In this case, the enhancement factor does reduce the time or laser power required to make random noise much smaller than proportional nonrandom noise. With sufficient sensitivity enhancement, the LCOF Raman cell could, therefore, allow an inexpensive Raman instrument using a low-power laser and a noisy CCD detector to reach a performance level typical of some state-of-the-art research Raman instruments. (11) McNay, J. L.; Fernandez, E. J. J. Chromatogr. A 1999, 849, 135-148.
The LCOF Raman cell design and operation minimize some of the major sources of nonrandom noise. Imperfect intensity calibration causes variation in sensitivity among CCD pixels, which leads to proportional nonrandom noise in Raman spectra. Subtraction of the solvent spectrum from the sample spectrum effectively eliminates this source of proportional nonrandom noise. Proportional nonrandom noise can also result from variations in coupling between the fiber-optic probe and the sample cell. Differences in coupling can be caused by variations in sample cell placement or by variation in the energy distribution among the modes of the LCOF and the probehead silica optical fiber. The LCOF Raman cell homogenizes both the laser intensity in the sample and the Raman intensity from the sample at the focal plane of the collection optics, which stabilizes the coupling between the fiber-optic probe and the sample cell. The LCOF Raman cell is a flow cell, so there is no sample cell position change between sample measurements. The sample and the solvent usually have nearly identical refractive indices, which eliminates refractive-index-induced nonrandom noise. Subtraction of sequentially collected spectra are, therefore, nearly free of proportional nonrandom noise, even after extensive signal averaging. The LCOF Raman cell is well-suited for highprecision Raman difference spectroscopy. Water Spectra. The Raman enhancement factor of the LCOF is strongly affected by scattering losses due to imperfections on the inside surface of the Teflon-AF capillary. In a previous publication, we reported an enhancement factor of 120 for water using 532-nm excitation.12 A capillary 1 m long and having a 50µm i.d. transmitted about 20% of the 532-nm laser light coupled into it when filled with water. Since that time, the Teflon-AF capillary manufacturing process has been improved. The 1.21-m Teflon-AF capillary used in this work produced an enhancement of 500 for water. Its transmission was 43%. Teflon-AF fluoresces strongly when excited with 532-nm light. It also has Raman bands at 831, 713, and 331 cm-1.12 Both the excitation and the collection of optical emission from the TeflonAF are very inefficient in the LCOF optical configuration, however. We compared the Raman spectrum of water collected in the LCOF to that collected from a cuvette. The background intensity from the LCOF near 1400 cm-1 is 0.87 times as strong as the background-corrected peak height of the water-bending vibrational band near 1640 cm-1. This is similar to the factor of 0.65 we calculated from a published Raman spectrum of water,13 indicating that our background intensity is primarily due to the water itself, rather than to the LCOF. This factor is 1.08 for our Raman spectrum that was collected in the cuvette. The somewhat greater background intensity from the cuvette is largely due to fluorescence from the holographic beam combiner inside the fiber-optic probehead. This fluorescence was not detected in the LCOF Raman spectrum because of the much shorter data acquisition time. Teflon-AF Raman bands from the LCOF also tend to be near or below the detection limit. For example, the 832 cm-1 Teflon band intensity was about 0.5% of the water background intensity in that spectral region. These comparisons demonstrate that the background intensity from a water-filled LCOF is mainly from the Raman spectrum of water, not luminescence from the LCOF. In (12) Altkorn, R.; Koev, I.; Pelletier, M. J. Appl. Spectrosc. 1999, 53, 1169-176. (13) Schrader, B. Raman/Infrared Atlas of Organic Compounds, 2nd ed.; John Wiley & Sons: New York, 1989.
fact, the LCOF sensitivity enhancement discriminates against luminescence from the fiber-optic probehead. The intensity of the water Raman spectrum places an upper limit on the CCD integration time for aqueous solutions. The CCD detector in the HoloProbe Raman analyzer was nearly saturated by the Raman intensity from the OH-stretching vibration of water (centered near 3450 cm-1) in only 20 ms when 24 mW of laser power was injected into the LCOF. When the spectral region above 2500 cm-1 was ignored, the weak water-bending vibration near 1640 cm-1 nearly saturated the CCD detector in 600 ms under similar conditions. To signal-average for longer periods of time, multiple spectra had to be collected and then added together in software. Unfortunately, the software required 2.3 s to transfer data from the CCD into the computer memory.14 Therefore, a 20 ms exposure required 2.3 s to complete. The observed sensitivity enhancement is degraded by a factor of 115 in this case if total acquisition time, rather than simply CCD exposure time, is considered. If the spectral region above 2500 cm-1 were not needed, a 600-ms exposure could be used and the 2.3-s data transfer time would only degrade the apparent sensitivity enhancement by a factor of (2.3 + 0.6)/0.6 ) 4.8. Shorter datatransfer times are possible and would greatly increase the utility of the LCOF Raman cell. The 100 kHz A/D converter speed of the CCD detector used in this work should allow full-spectrum transfer times of 106 photons/binned CCD column. This would produce a shot noise level of
