Analysis of Microbial Components Using LC− IR

Anal. Chem. 2003, 75, 4606-4611

Analysis of Microbial Components Using LC-IR Scott W. Huffman,†,§ Kara Lukasiewicz,‡,⊥ Susan Geldart,† Susan Elliott,‡ Jay F. Sperry,‡ and Chris W. Brown*,†

Department of Chemistry and Department of Cell and Molecular Biology, University of Rhode Island, Kingston, Rhode Island 02881

Characterization of bacteria is currently an important research area in the medical, military, food, and agricultural sciences. In recent years, FT-IR has found an application as a microbiological detection method and as a general research tool. When coupled with a liquid chromatographic system, a new facet of research has evolved. By utilizing the separation ability of typical liquid chromatography systems, matrix elimination is possible, therefore allowing for clean spectra of cellular components. Information about the compositional makeup of various bacteria enhances the overall understanding of biology at the cellular level, provides a quantification of the chemistry of cellular processes, and can be used as a general identification tool. Both whole cells and lysed Escherichia coli cells were investigated in the present study. The cellular components consisting of proteins, glycoproteins, phospholipids, fatty amides and acids, and genomic materials were separated, isolated, and identified by FT-IR. Throughout human history, bacteria have caused unwanted complications, such as food spoilage, water contamination, disease, and death. To reduce the harmful effects of bacteria, it is necessary to understand the mechanisms that promote and sustain bacterial infections. By understanding the chemical nature of bacteria, it may be possible to prevent food spoilage, illness, and other related problems. The chemical information obtained from bacteria is large and spans the range from structures of proteins to the length and degree of saturation of lipids within the cell membrane. To retrieve this information, many tools have been developed and applied to bacteria. The scope of these tools is vast, ranging from basic wet biochemical techniques to instrumental methods such as Curie point pyrolysis mass spectrometery1 and ultraviolet (UV) resonance Raman spectroscopy.2 Infrared (IR) spectroscopy also has been used extensively to increase the knowledge base on bacteria.1,3-10 * Corresponding author. E-mail: [email protected]. † Department of Chemistry. ‡ Department of Cell and Molecular Biology. § Present address: National Institutes of Health, Bethesda, MD. ⊥ Present address: Mayo Clinic, Rochester, MN. (1) Goodacre, R.; Shann, B.; Gilbert, R. J.; Timmins, E. M.; McGovern, A. C.; Alsberg, B. K.; Kell, D. B.; Logan, N. A. Anal. Chem. 2000, 72, 119-127. (2) Nelson, W. H.; Manoharan, R.; Sperry, J. F. Appl. Spectrosc. Rev. 1992, 27, 67-124.

4606 Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

When applied to complete cells, infrared spectroscopy3,5 can classify bacteria into species and subspecies because it provides a complex whole-cell fingerprint, as shown in Figure 1. Functional groups for some of the major cell components are recognizable in this infrared spectrum; specifically, the bands around 3000 cm-1 are attributed to all of the C-H stretching vibrations in the cell membrane, proteins, nucleic acids, and other minor components; the two bands at 1650 and 1550 cm-1 are assigned to the amide I and amide II vibrations of proteins, and the broad band centered at 1050 cm-1 has many contributors, including the phospholipids and the C-O stretching vibration in sugars. One advantage of this complexity is that it can be used in combination with pattern recognition methods and library searching for identification of bacteria.3,5,6,10 In addition, by changing the chemical environment of bacteria, this whole-cell chemistry can be monitored in vivo. The disadvantage of the complexity of whole-cell spectra is that they are difficult to interpret because the entire composition of the cell contributes to the spectra.5,6 The complexity of the chemical mixture makes structural assignment of bands within whole-cell spectra of bacteria a challenging task. However, spectra of purified components of cells may provide insight toward interpretation of whole cell spectra. Surprisingly, with all of the research utilizing IR spectroscopy for identification and classification of microorganisms, little work has been done to correlate IR bands with bacterial cell components. Liquid chromatography (LC) has been used extensively to separate biochemical molecules;11-13 however, a limited amount of structural information can be obtained with traditional LC detectors. By combining the separation ability of LC and the (3) Goodacre, R.; Timmins, E. M.; Rooney, P. J.; Rowland, J. J.; Kell, D. B. FEMS Microbiol. Lett. 1996, 140, 233-239. (4) Schultz, C.; Naumann, D. FEBS Microbiol. Lett. 1991, 294, 46-46. (5) van der Mei, H. C.; Naumann, D.; Busscher, H. J. Arch. Oral Biol. 1993, 38, 1013-1019. (6) Beattie, S. H.; Holt, C.; Hirst, D.; Williams, A. G. FEMS Microbiol. Lett. 1998, 164, 201-206. (7) Zeroua, W.; Choisy, C.; Doglia, S. M.; Bobichon, H.; Angiboust, J. F.; Manfait, M. Biochem. Biophys. Acta 1994, 1222, 171-178. (8) Lefier, D.; Hirst, D.; Holt, C.; Willians, A. G. FEMS Microbiol. Lett. 1997, 147, 45-50. (9) Brown, C. W.; Li, Y.; Seelenbinder, J. A.; Pivarnik, P.; Rand, R. G.; Letcher, S. V.; Gregory, O. J.; Platek, M. J. Anal. Chem. 1998, 70, 2991-2996. (10) Naumann, D.; Helm, D.; Labischinski, H.; Giesbrecht, P. In Modern Techniques for Rapid Microbiological Analysis; New York: VCH Publishers, 1991; pp 43-96. (11) Brouwers, J. F. H. M.; Gadella, B. M.; Tielens, A. G. M. J. Lipid Res. 1998, 39, 344-353. (12) Nunez, P. E.; Scoging, A. C. Int. J. Food Microbiol. 1997, 36, 39-48. (13) Angelov, D.; Spassky, A.; Cadet, J. J. Am. Chem. Soc. 1997, 119, 1137311380. 10.1021/ac034571w CCC: $25.00

© 2003 American Chemical Society Published on Web 07/30/2003

Figure 1. Whole-cell infrared spectrum of a colony of E. coli cells smeared on a KBr window.

chemical fingerprinting ability of IR, individual components of bacterial cells can be isolated, collected, and identified. It is thought that infrared spectra of cellular components will help to elucidate bands in whole-cell spectra. The main goal of the present research is to lyse the cells, separate individual components, and identify the components with infrared spectroscopy; therefore, developing a technique by which the biochemistry of bacteria including pathogenic bacteria can be investigated at the subcellular level. EXPERIMENTAL SECTION The procedures utilized in this investigation consist of two microbiological preparations and two similar chromatographic/ spectroscopic analyses. The first microbiological preparation was designed to extract the cellular components from the entire cell. The second preparation was designed to extract cellular components on the basis of the cellular structure from which they originate, specifically, the cell membrane and the cytosol. In the latter case, the cells were enzymatically lysed. Bacterial Preparation and Analysis: Entire Cell. The bacteria used in these experiments were grown from in-house stock colonies. Escherichia coli ATCC 25922 is considered a nonpathogenic variety of E. coli. This subtype was used because of its availability and for safety concerns during method development. The bacteria were cultured in Difco (Detroit, MI) nutrient broth, which was prepared as per the manufacturer’s instructions, for 24-48 h at 37 °C. To harvest the cells from the broth, they were centrifuged for 10 min at 1200g using an IEC Spinette centrifuge from Damon (Needham Heights, MA). The supernatant liquid was removed, and the cells were resuspended in deionized water and centrifuged for 10 min. The supernatant was removed, rinsed, and centrifuged a second time. The remaining cell pellet was placed in a liquid nitrogen-filled mortar. The frozen pellet was ground into a powder with the mortar and pestle. This cellular powder was dissolved with 2.0 mL of 0.1% (w/v) sodium dodecyl sulfate (SDS) solution. The SDS was obtained from Fisher Scientific (Suwanee, GA 30024). The sample was stored under refrigeration until analysis. Prior to injection into the chromatographic system, the sample was filtered with a Whatman (Tewksbury, MA 01876-0962) 4.5µm-pore-diameter nylon filter. The chromatographic system consisted of a ThermoQuest Constumetric 4100 pump, (Piscataway, NJ) with a 100-µL-loop Rheodyne (Rohnert Park, CA 949271909) injector interfaced with a Bourne Scientific IRC system

(Acton, MA 01720) and a NovaPack C18 4.6 × 100 mm column. The IRC system consists of a nebulizer for removing the solvent. The solvent-free eluents were deposited onto a zinc selenide (ZnSe) window, which was translated through an infrared sample beam from a Midac (Irvine, California) Illuminator FTIR module. The spectra were measured with a narrow range MCT detector at a resolution of 8 cm-1 with 20 spectral coadds; nine coadded spectra were acquired every minute. The separation method was modified from a LC/MS method.14 In our method, a linear gradient with a rate of 1 mL/min was used. The solvent system ranged from 98% H2O with 0.1% (v/v) formic acid and 2% acetonitrile (v/v) to 100% acetonitrile for 100 min. The eluent was split 20:1 before entering the IRC. All chemicals used in the LC method were HPLC grade from Fisher Scientific (Suwanee, GA). Bacterial Preparation and Analysis: Lysed Cells. Two flasks, each containing 50 mL of nutrient broth, were inoculated with E. coli and allowed to grow overnight at 37 °C in an incubator. The number of colony forming units (CFUs) for the two flasks was determined to be 1.68 × 109 and 2.03 × 109 CFU/mL using a serial dilution, spot-plating method, which was repeated three times. The cells from both flasks were harvested by centrifugation. The cells were washed twice with a phosphate buffer solution (PBS), pH of 6.24, to ensure that the majority of the nutrient broth had been removed from the cells. A lysozyme solution was prepared by dissolving 1 mg of crystalline lysozyme (Sigma Chemical Co., St. Louis, MO) in 1 mL of PBS. Each pellet of bacteria was then suspended in 800 µL of PBS and 200 µL of the lysozyme/PBS solution. The suspended E. coli cells were allowed to sit overnight in the lysozyme/PBS solution. Lysozyme is an enzyme that hydrolyzes the β(1,4) linkage between the N-acetylmuramic acid and N-acetylglucosamine of the peptidoglycan wall of a bacterial cell.15 To remove lysozyme from the mixture, the cells were washed once with PBS, followed by sterile, distilled, deionized (DDI) water. The cells were suspended in 1 mL of DDI water for sonication in a sonifier cell disruptor (Heat Systems Co., Melville, NY). The ultrasonic vibrations of the sonifier break apart the hydrolyzed cell wall structure. Each tube of cells was suspended in packed ice and sonicated using 100-W pulses for 10 s at six repetitions each. The two tubes of cells were alternated in the sonicator so that the samples would not overheat and denature the proteins within the cells. After the cells were sonicated, each tube was transferred into an ultracentrifuge tube. The cells were centrifuged at 100000g for 60 min using a TY65 rotor (Beckman-Coulter, Fullerton, CA). The final product of this ultracentrifugation was a separation in which the pellet contained the ribosomes and membrane molecules, while the supernatant contained the cytosolic molecules. The cytosolic supernatants were pipeted off and preserved. The membrane pellets were each suspended in 500 µL of DDI water. The chromatographic setup for this preparation was slightly different from the previous entire cell setup. Specifically, a Vydac 218TP54 C18 column, 5 µm × 4.6 mm × 25 mm, was used for separation. Unfiltered 250-µL aliquots of the membrane suspension (14) Covey, T. R.; Huang, E. C.; Henion, J. D. Anal. Chem. 1991, 63, 11931200. (15) Voet, D.; Voet, J. G. Biochemistry, 2nd ed.; New York: John Wiley & Sons, 1995.

Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

4607

Figure 2. (a) The infrared spectrum at 2 min from the entire cell chromatogram. This is a spectrum of SDS. (b) The infrared spectrum at 23 min from the entire cell chromatogram. Spectrum of protein extracted from E. coli. (c) The infrared spectrum at 76 min from the entire cell chromatogram. This is a spectrum of a long chain fatty amide. (d) Total chromatogram of E. coli whole cell extract.

and cytosol solution were injected into the chromatographic system. The solvent system also differed from the entire cell run in that a mobile phase modifier (formic acid) was not used. RESULTS AND DISCUSSION In one set of experiments, E. coli cells were frozen with liquid nitrogen, ground with a mortar and pestle, extracted with an SDS solution, and injected into the LC-IR system. In the second set of experiments, the cells were enzymatically lysed, and the cellular materials were separated into membrane and cytosol components and injected as separate samples into the LC-IR system. Whole Cells. The chromatogram of whole bacterial cells is shown in Figure 2d. This chromatogram was created by plotting the largest absorbance value in every infrared spectrum versus time. This chromatogram is referred to as the total chromatogram. In Figure 2, the arrows indicate chromatographic peaks whose infrared spectra are identified. The spectrum of the first chromatographic peak in Figure 2d is shown in 2a and appears to be a complex whose major 4608 Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

component is sodium formate from the mobile phases with a small amount of SDS as a minor component. The infrared spectrum of the compound eluting at 23 min is shown in Figure 2b. The spectrum is identified as a protein from the amide I vibration at 1642 cm-1, amide II at 1550 cm-1, and the amide III vibration at 1268 cm-1. Other indicators that the spectrum is protein are the N-H stretching vibration at 3300 cm-1 and the first overtone of the C-N-H bend at 3086 cm-1, which are normally masked by a broad O-H stretching band of water centered at ∼3400 cm-1. The absence of the water band at 3400 cm-1 suggests that nearly all of the water associated with the protein has been eliminated, which exemplifies the efficiency of the solvent elimination and deposition mechanism of the LC-IR instrument. This elimination of water may or may not include water embedded within the interfoldings of the protein. A consequence of this extreme purification is that the protein many have lost its tertiary structure. The last prominent peak in the chromatogram at 76 min produced the spectrum shown in Figure 2c. The infrared spectrum measured at the apex of this peak is assigned as a long-chain fatty

Figure 3. (a) The infrared spectrum of glycoproteins from the chromatographic peak eluting at 5 min in the membrane LC run of E. coli cell lysate. (b) Representative infrared spectrum of phospholipid from the middle of the large broad peak with elution times ranging from 34 to 63 min from the membrane suspension of the cell lysate of E. coli. (c) CH2 band chromatogram from the membrane portion of lysed E. coli.

amide, which most likely originated in the cell membrane. The doublet around 1630 cm-1 corresponds to a carbonyl and an NH2 deformation. Specifically, the band at 1633 cm-1 is assigned to an NH2 deformation, and the band at 1658 cm-1 is assigned to the CdO stretch of an amide. The absorptions at 3358 and 3192 cm-1 are attributed to the antisymmetric and symmetric NH2 stretches, respectively. The band at 718 cm-1 is assigned to the CH2 rocking vibration, whereas the set of weak bands between 1341 and 1197 cm-1 are attributed to the CH2 waging vibrations of the long hydrocarbon chain. The band at 1411 cm-1 is assigned to the C-N stretching vibration, and the one at 1467 cm-1 is assigned to the CH2 bending vibration. By comparison with library spectra, the length of the hydrocarbon chain was estimated from the ratio of the intensities of the CH2 to CH3 antisymmetric stretches at 2922 and 2958 cm-1 to contain ∼20 CH2 units. The remaining chromatographic peaks in Figure 2d are assigned to SDS complexes with cellular components. These complexes are otherwise unidentifiable and are formed when excess SDS is mixed with cellular material.16 The lack of other biochemically relevant compounds in these experiments suggested that improvements to the extraction method were neces(16) Scopes, R. Protien Purification: Principles and Practice; Springer-Verlag: New York, 1982.

sary. There are two reasons for this inefficiency. One is that the mechanical breaking method under liquid nitrogen does not sufficiently break the cells. The second reason is that much of the cellular material was retained on the filter prior to injection into the chromatographic system. Lysed Cells. The band chromatogram formed by plotting the absorbance of the 2923 cm-1 antisymmetric CH2 stretching vibration of the membrane suspension from the lysed cells is shown in Figure 3c. This chromatogram was created after subtracting the spectrum of sodium formate, which is a common mobile phase modifier and contaminate of LC-IR. The absorbance of the antisymmetric CH2 stretch was chosen for generating this band chromatogram because most biologically relevant materials are organic in nature and contain some aliphatic functionality. The arrows indicate chromatographic peaks whose infrared spectra are identified. The IR spectrum measured at 5 min shown in Figure 3a is assigned to glycoprotein separated from the membrane suspension of the cell lysate. Glycoproteins are common components of the exterior portion of the outer membrane of bacterial cells. As the name implies, the molecule is composed of both sugar and protein. The broad centered band at 1085 cm-1 is assigned to the Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

4609

Figure 4. (a) Representative infrared spectrum of isolated nucleotides from the cytosolic portion of the E. coli. These compounds eluted around 4 min. (b, c) Representative infrared spectra of lipid molecules isolated in the cytosol with elution times of 45-55 min. (d) The CH2 chromatogram of the cytosolic solution of lysed E. coli.

C-O vibrations of the oligosaccharide side chains of the proteins.17 The bands at 1650 and 1538 cm-1 are due to amide I and amide II, respectively. The very strong, broad band in the 2500-3500 cm-1 region is due to the extensive hydrogen bonded O-H stretching vibrations of sugar moieties and water molecule. These proteins, in contrast to those previously shown in Figure 2, have retained their tertiary structure associated water molecules. The sharpness of the band around 3290 cm-1 corresponds to the antisymmetric NH2 stretching vibration of the protein. The IR spectrum of a representative phospholipid eluting at 54 min is shown in Figure 3b. The entire unresolved portion of the band chromatogram from 34 min to 63 min contains phospholipids. Phospholipids constitute the lipid bilayers within the cell membrane. The band at 1070 cm-1 is the P-O-C vibration in the phosphate of the phospholipid. The carboxylic acid CdO stretch of the phospholipids is evident at 1736 cm-1. The C-H stretching region from 3000 to 2840 cm-1 is characteristic of long hydrocarbon chains that are components of fatty acids of phos(17) Parker, F. S. Applications of Infrared Spectroscopy in Biochemistry Biology and Medicine; Plenum Press: New York, 1971. (18) Wilmshurst, J. K. J. Chem. Phys. 1957, 26, 426-427.

4610 Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

pholipids. The relative absorbance ratio of CH2 stretching to CH3 differs throughout the large, unresolved chromatographic peak, suggesting the presence of different chain lengths and branching of chains in the phospholipids. The general trend of this ratio indicates an initial increase in the ratio of CH2 to CH3 followed by a decrease. This behavior can be explained if short, straight chain phospholipids elute first, followed by longer chain phospholipids, and finally, branched phospholipids. The band chromatogram due to the antisymmetric CH2 stretch at 2923 cm-1 of the cytosolic solution is shown in Figure 4d. The spectrum corresponding to the peak at 4 min is shown in Figure 4a and is assigned to nucleotides. It is currently unclear whether these nucleotides are present as monomers or as polymers, such as DNA or RNA. The band located at 1089 cm-1 is attributed to the P-O-C vibrations of the phosphate groups in the nucleotides. The shoulder at 1021 cm-1 along with the broad absorption between 2500 and 3500 cm-1 are attributed to the sugar portion of the nucleotides, either ribose or deoxyribose. The carbonyl band centered around 1690 cm-1 is broader than is usual for a single CdO stretch. This is consistent with the different carbonyls within the nucleotides.

The broad chromatographic peak in Figure 4d between 45 and 55 min is again due to a group of long chain hydrocarbons. Two representative spectra measured at 45 and 48 min are shown in Figure 4b and 4c, respectively. These spectra are assigned to molecular complexes containing long chain hydrocarbons, such as phospholipids, which are identifiable by the prominent C-H stretching region just belong 3000 cm-1. Unfortunately, further identification is hindered by other substances in the completes, such as nitrate, phosphate, and formate salts. Finding fatty acid material in the cytosol chromatographic run was surprising. The design of the centrifugation separation method was supposed to keep this material in the membrane portion. We attribute the presence of lipid molecules in the cytosol to the formation of vesicles of the lipids as they broke away from the membrane during ultrasonic disruption. Some of these vesicles must have been too small to precipitate during centrifugation and, therefore, remained suspended in the supernatant of cytosol portion of the centrifuge tube. CONCLUSIONS These experiments demonstrate the utility of LC-IR for analyzing biological molecules. Two different methods for extracting E. coli cell components were explored giving different results. The simple mechanical grinding of frozen cells and prefiltering liberated only a small amount of cellular material. In contrast, the enzymatic lysing of cells without prefiltering generated copious amounts of biochemically relevant materials. Proteins, lipids, and nucleic acids are the three most abundant compounds, other than water, present in bacterial cells and were all isolated and

measured. Phospholipids and glycoproteins were isolated from the membrane suspension. These two molecules are normally found in the membranes of bacteria cells. Nucleotides were found in the cytosol of the cell, which could originate from the genome of E. coli or energy storage molecules, such as adenosine triphosphate (ATP). By applying the separation and structural identification abilities of LC-IR to bacterial samples, the assignment of bands from whole cell spectra will become clearer. For example, in the whole cell spectrum in Figure 1, the band centered around 1050 cm-1 is assigned to the phosphate groups of the phospholipids. However, from this work, we are reminded that the nucleic acid material (DNA, RNA, and nucleotides), which contains phosphate, will contribute significantly to this band in the whole-cell spectrum. In addition, sugars within the bacterial cells such as those attached to the glycoprotein will also contribute to this band. LC-IR is a powerful analysis tool for complex mixtures whose composition ranges are often unknown. The information gained from these experiments can be applied to the more complex whole cell spectra. ACKNOWLEDGMENT This work was supported in part by the U.S. Department of Agriculture (NRICGP Grant no. 93-37201-9197) and by the URI Partnership for Sensors and Surface Technology. Received for review May 28, 2003. Accepted June 15, 2003. AC034571W

Analytical Chemistry, Vol. 75, No. 17, September 1, 2003

4611