On the Difference Between Surface-Enhanced Raman Scattering(SERS) Spectra of Cell Growth Media and Whole Bacterial Cells

W. RANJITH PREMASIRI, YOSEPH GEBREGZIABHER, and LAWRENCE D. ZIEGLER*Department of Chemistry and The Photonics Center, Boston University, Boston, Massachusetts 02215

It has been recently suggested [N. E. Marotta and L. A. Bottomley, Appl.

Spectrosc. 64, 601–606 (2010)] that previously reported surface-enhanced

Raman scattering (SERS) spectra of vegetative bacterial cells are due to

residual cell growth media that were not properly removed from samples

of the lab-cultured microorganism suspensions. SERS spectra of several

commonly used cell growth media are similar to those of bacterial cells, as

shown here and reported elsewhere. However, a multivariate data analysis

approach shows that SERS spectra of different bacterial species grown in

the same growth media exhibit different characteristic vibrational

spectra, SERS spectra of the same organism grown in different media

display the same SERS spectrum, and SERS spectra of growth media do

not cluster near the SERS spectra of washed bacteria. Furthermore, a

bacterial SERS spectrum grown in a minimal medium, which uses

inorganics for a nitrogen source and displays virtually no SERS features,

exhibits a characteristic bacterial SERS spectrum. We use multivariate

analysis to show how successive water washing and centrifugation cycles

remove cell growth media and result in a robust bacterial SERS spectrum

in contrast to the previous study attributing bacterial SERS signals to

growth media.

Index Headings: Surface-enhanced Raman scattering spectroscopy; SERS;

Principle component analysis; PCA; Bacteria identification; Diagnostics;

Growth media.

INTRODUCTION

The potential for using the optical signature resulting from asurface-enhanced Raman spectrum for the detection andidentification of whole vegetative bacterial cells has beendemonstrated by several research groups over the past fewyears.1–22 The successful development of a surface-enhancedRaman spectroscopy (SERS) based methodology for thespecies- and strain-specific identification of bacterial cellscould have enormous impact for the development of a rapid,easy-to-use, point-of-care pathogen diagnostic platform. SERSoffers some unique advantages with respect to other optical andnon-optical techniques as a method for bacterial infectiousdisease identification. The enormous Raman signal enhance-ment permits the acquisition of SERS spectra at the near-singlebacterial cell limit on a time scale of seconds without the needfor sample amplification by growth or polymerase chainreaction (PCR) methods and without the need for extrinsiclabeling. SERS-based bacterial identification can be achievedrapidly by minimally trained personnel and is less subject tocontamination than PCR approaches, which rely on sampleamplification. Because low incident laser power (1–5 mW) isrequired for SERS data acquisition due to the large Ramanenhancement effect, relatively low cost, portable SERSplatforms for point-of-care diagnostics may be developed.Furthermore, the near-diffraction-limited spatial resolution

afforded by microscopy potentially allows SERS specificityto be applied to the analysis of bacterial mixtures.

In previous reports we have shown that SERS spectra ofbacterial cells grown in a variety of cell growth media exhibitunique SERS signatures that allow their identification whencombined with multivariate data analysis techniques.9,18 Whilea number of SERS substrates have been demonstrated toprovide bacterial SERS spectra, the largest bacterial SERSsignals we have observed were obtained on SiO2 substratescovered with Au nanoparticle resulting from an in situ metal-ion-doped sol-gel process.8 Sample signal data acquisitiontimes are ;10 seconds with about 1 mW of 785 nm incidentlaser power focused and collected by a 503 objective. Single-cell-level sensitivity has been achieved on these substrates.8

Reproducible SERS spectra have been shown to allow thedevelopment of libraries that allow diagnostic identification ofbacterial species and strains.18 Furthermore, we have previouslydemonstrated that media-cultured bacteria that have been spikedinto human blood, enriched following a centrifugation lysisprocedure, also result in species-specific SERS signatures.9

The chemical/molecular identity of these bacterial SERSvibrational bands has not yet been established. Due to thedistance dependence of the SERS enhancement mechanism, ithas been generally assumed that the observed vibrational bandsappearing in the SERS spectra of whole vegetative bacterialcells is dominated by the contributions of cell wall compo-nents, such as peptidoglycan, lipids, lipopolysaccharides, ormembrane proteins of the outer cell layers. However, thishypothesis has not been proven. As an example of the currentunderstanding of these bacterial SERS spectra, the generallystrongest band observed in red-excited bacterial SERS spectraat ;730 cm�1 has been variously assigned to the ring-stretching mode of adenine3,12 or adenosine,19 the flavineadenine dinucleotide,17 the glycosidic ring of polysaccharidessuch as N-acetyl-D-glucosamine (NAG) and N-acetylmuramicacid (NAM), components of peptidoglycan,6,7,15 or a symmet-ric O–P–O vibrational mode of the phosphate group.20

In a recent study by Marotta and Bottomley23 the SERSspectra of bacterial cell growth media and bacterial cellscultured in these same media on Ag nanorod substrates werereported. The observed SERS spectra of the vegetativebacterial cells are similar to those reported previously by othergroups, but they also suspiciously resemble SERS spectra ofseveral cell growth media alone once they have been diluted inwater. These authors, however, do not follow the standardsample bacterial preparation protocol of repeated cycles ofwashing with pure water and centrifugation but instead acquireSERS spectra of diluted bacterial cultures. Marotta andBottomley23 attribute the expected bacterial SERS signatureto contributions from the cell wall structures, as describedabove. Based on (1) the similarity between growth media andbacterial SERS spectra, (2) the similarity of Gram-positive and

Received 28 October 2010; accepted 17 February 2011.

* Author towhomcorrespondence shouldbe sent.E-mail: lziegler@bu.edu.

DOI: 10.1366/10-06173

Volume 65, Number 5, 2011 APPLIED SPECTROSCOPY 4930003-7028/11/6505-0493$2.00/0

� 2011 Society for Applied Spectroscopy

Gram-negative bacterial SERS signatures, (3) the spatialdistribution of SERS signals on the SERS substrate, and (4)dilution dependence of SERS signals, these authors23 arguethat the SERS spectra previously reported for bacteria areprobably not due to inherent chemical features of the bacterialcells themselves but, in all likelihood, trivially result fromresidual cell growth media not fully removed from the bacterialsample solutions or cell growth media that remains adsorbed tothe outer walls of these bacterial cells despite multiple cellwashing cycles.23 Thus, the authors suggest that the SERSmethodology for identification of bacterial species/strains mayhave been flawed in the previous studies and that any observedspectral differences reported for the SERS spectra of a givenbacterial species could simply result directly from contributionsof different types or amounts of cell media components.Marotta and Bottomley23 thus conclude that if this hypothesiswere correct, the best strategy would be to develop onlyRaman-based diagnostics that are media free, such as confocalnon-SERS Raman or optical tweezer based Raman techniquesbecause SERS spectra of vegetative bacterial cells appear to bedominated by cell growth media.

The purpose of this report is to provide data to quantitativelytest the hypothesis that SERS bacterial spectra are due to residualgrowth media that has not been completely removed frombacterial solutions or surfaces. More specifically, we report onthe SERS spectra of typical cell growth media and usemultivariate data analysis methods to compare SERS spectra oflaboratory-cultured bacteria with the SERS spectra of the mediain which they are grown. Furthermore, the limited datarepresented by these data also serve to address the question ofthe more general dependence of bacterial cell SERS spectra oncell growth media. Although not the main subject of this report,the analysis presented here calls into question the assumedexplanation for the origins of the bacterial SERS spectra. A reporton the molecular origins of the SERS spectra of bacteria will bethe subject of a subsequent manuscript from this laboratory.

The unequivocal conclusion supported by the data presentedhere is that although the SERS spectra of some cell growthmedia indeed resemble the SERS spectra of well-washedbacterial cells, the bacterial SERS signatures are not due tocontamination from the medium and result from inherentspecies- and strain-specific biochemical activity. The SERSspectra of bacteria are readily distinguished from that of cellgrowth media and proper water-washing protocols are shownto effectively remove cell growth media as evidenced bymultivariate data analysis. The similarity in the SERSsignatures arises from the molecular origins of the bacterialand some growth media SERS spectra.

EXPERIMENTAL PROCEDURES

Bacterial Strains, Growth Media, and Sample Prepara-tions. The bacterial strains used in these studies and theirsources are summarized in Table I. Bacterial strains weregrown in ;15 mL of growth media broth (typical OD ; 1) andharvested during the log growth phase by centrifugation andwashed 3 to 4 times with deionized (DI) Millipore water. Theresulting cell pellet was resuspended in 0.25 mL of water and 1lL of the resulting ;109/mL bacterial suspension was pipetteddirectly onto the SERS substrate for the purpose of the dataanalysis described here. SERS measurements were made,usually, after about two minutes when nearly all the water hadevaporated.

SERS Substrates. Substrates resulting from an in situ Auion doped SiO2 sol-gel procedure previously developed in ourlaboratory8 were used to acquire all the SERS spectra reportedhere. This procedure results in small aggregates (2–15particles) of monodispersed ;80 nm Au nanoparticlescovering the outer layer of ;1 mm2 SiO2 substrate. Detailsconcerning the production of these SERS active chips and thecharacterization of their performance for providing reproduc-ible and strongly enhanced (single cell level) SERS spectra ofbacteria when excited by a few milliwatts of 785 nm radiationhave been previously described.8 The shelf life of these sol-gel-based substrates is currently ;5 months.

Data Acquisition. All of the spectra shown here wereacquired with the RM-2000 Renishaw Raman microscopeemploying a 503 objective and excited at 785 nm. SERSspectra were obtained with ;2 mW of incident laser power in;10 seconds of illumination time. The 520 cm�1 band of asilicon wafer was used for frequency calibration.

Multivariate Data Analysis. As shown previously,18 asecond-derivative-based barcode protocol results in enhancedprinciple component analysis (PCA) specificity and reproduc-ibility as compared to PCA analysis based on spectral intensityor first or second derivatives. In this procedure, after Fourierfiltering high-frequency noise from the normalized spectra, abarcode or binary vector is created based on the sign of thesecond derivative of the observed SERS spectrum. The inputvector of 0’s or 1’s is then used to perform the PCA analysis.More details and demonstrations of the efficacy of this barcodeprocedure can be found elsewhere.18 This barcode multivariatedata analysis procedure will be used here to quantitativelyassess the relationship between the SERS spectra of bacterialcells and those of their growth culture media.

RESULTS AND DISCUSSION

SERS Spectra of Washed Bacteria and Culture Media.SERS spectra of five bacterial species, K. pneumonia, E. coli,P. aeruginosa, E. faecalis, and two strains of S. aureus, areshown in Fig. 1. The growth media used to culture theindicated strains of these representative bacterial species areindicated in this figure as well. All except the K. pneumoniahave been grown in tryptic soy broth (TSB). The K. pneumoniawas cultured in nutrient broth (NB) media. E. coli, K.pneumonia, and P. aeruginosa are Gram-negative and S.aureus and E. faecalis are Gram-positive bacteria. This classicGram staining characterization delineates bacterial types on thebasis of outer wall structure.24 As evident in this figure, therepresentative bacterial SERS spectra share a number of similarspectral features. For example bands near 620 cm�1, 730 cm�1,960 cm�1, 1350 cm�1, 1450 cm�1, and 1580 cm�1 can beidentified in the SERS spectra shown in Fig. 1, although theexact band maximum frequency may vary. It should be

TABLE I. Summary of bacterial strains used in these SERS studies.

Bacterial species Strain ID Source

K. pneumoniae 11998 BD TechnologiesS. aureus USA300 ATCCS. aureus NCTC 8325 Microbiotix, Inc.E. coli 29425 ATCCE. coli 11996 BD TechnologiesP. aeruginosa P14 Microbiotix, Inc.E. faecalis 29212 ATCC

494 Volume 65, Number 5, 2011

emphasized that the sample preparation procedure we follow toacquire the displayed bacterial SERS spectra, as well as allpreviously published bacterial spectra from this lab,8,9,18,21

includes 3 to 5 cycles of washing with DI Millipore water andcentrifugation followed by re-suspension in water before thebacterial sample is applied to the SERS substrate (in situ grownAu nanoparticle-covered SiO2 matrix).

Surface-enhanced Raman scattering spectra of six differentbacterial cell growth media obtained on these same Aunanoparticle-covered SiO2 substrates are shown in Fig. 2. Inaddition, the SERS spectrum of one of the key components ofColumbia and LB broth, yeast extract, is also displayed in thisfigure. Comparison of the SERS spectra shown in Figs. 1 and 2reveals that several of the spectral features seen in the SERSspectra of bacteria are similar to those observed in some of thevarious growth media. This observation is in agreement withthe previous report on the SERS of bacteria and growthmedia.23 For example, the strong band in the ;730 cm�1

region is evident in the SERS spectra of all the bacterial speciesas well as appearing in SERS spectra of NB, CB, LB, TSB, andtyptone growth media. As indicated above, there is currentlylittle agreement as to the molecular origin of these vibrationalbands.

Multivariate Analysis of Bacterial SERS Spectra andGrowth Media Dependence. However, despite the similaritybetween the SERS spectra of some growth media and somevegetative bacterial cells, when SERS spectra of bacteria that

have been prepared by repeated washing in water are comparedwith the SERS spectra of growth media in which they aregrown, clearly distinct and reproducible differences arediscerned. We will use a multivariate data analysis approachthat has been previously developed for bacterial identificationto provide a quantitative measure of the degree of similarity/difference of these SERS spectra.18

In Fig. 3 the results of a PCA procedure based on the second-derivative barcode approach for SERS spectra of E. faecalisgrown in LB and TSB, P. aeruginosa and S. aureus grown inTSB and LB broth, and TSB and LB media alone are shown.Representative spectra for these different samples are given inFigs. 1 and 2. In this two-dimensional (2D) PCA contour, PC2versus PC3 is plotted because these are the principle componentsof greatest significance that show the largest differences betweendistinct groups. Two-dimensional standard deviation ringscentered on the mean of each group are also displayed in thisfigure. In other words, spectra identified by PC2, PC3 values thatfall within the smallest contour are within one standard deviationfrom the mean PC2 and PC3 values for a given resulting cluster.The second largest ring corresponds to two standard deviationsfrom the mean for the cluster, etc. As evident in the PC2 vs. PC3plot (Fig. 3), the same organism, E. faecalis, grown in twodifferent media, LB and TSB, cluster together. Furthermore,barcodes based on SERS spectra of three different organisms, P.aeruginosa, S. aureus, and E. faecalis, grown in the samemedium (TSB) form widely separated PCA clusters, as also seen

FIG. 1. SERS spectra of five bacterial species including two strains of S.aureus on in situ grown Au nanoparticle-cluster-covered SiO2 chips. The strainand growth media (TSB or NB) are indicated.

FIG. 2. SERS spectra of bacterial growth media. The ingredients of the mediaare listed in Table II. Yeast extract is the main component of LB and CB.

APPLIED SPECTROSCOPY 495

in Fig. 3. Finally, the PCA clusters corresponding to the SERS ofLB, probably the spectrum that most resembles that of bacteria,and TSB alone are likewise included in this PCA analysis. Asseen in Fig. 3, the two growth media and the corresponding E.faecalis grown in these media are far apart on this two-dimensional PCA plot.

The PCA clustering evident in Fig. 3 emphaticallycontradicts the hypothesis23 that the SERS spectra of properlywashed bacteria are due to residual growth culture media. Thesecond-derivative barcode based PCA used here enhancesspecificity; thus, the media independence for the SERSsignature of E. faecalis indicates that whatever the molecularorigins of this signal, it is not dependent on cell growth mediatype, at least for this species and growth media. Theconsistently different SERS signatures for the three speciesgrown in the same broth, TSB, similarly argues that what isgiving rise to these SERS signals is not directly the mediacomponents in which the bacteria were cultured via adsorptionto the cell surface or incomplete removal, as suggested byMarotta and Bottomley.23 Finally the large and consistentseparation between the SERS based clusters of the LB and TSBgrowth media and the water-washed bacteria grown in thesesame media make it clear that these are distinct vibrationalsignatures.

In order to ensure that the results in Fig. 3 are not strictly aconsequence of the specific bacterium and growth mediaselected, we carried out an analogous multivariate analysis ofthe growth media dependence of another bacterium. In Fig. 4,SERS spectra of S. aureus USA300 cultured in three differentgrowth media, TSB, BHI, and CB, and three additionalbacteria, E. faecalis, E. coli, and P. aeruginosa, are analyzedtogether via the barcode PCA treatment. Consistent with thePCA results displayed in Fig. 3, the PCA clusters given by thisenhanced specificity approach for S. aureus USA300 grown inthree different media all lie on top of one another in this PCAplot (Fig. 4). In other words, no SERS spectral distinctions canbe discerned for this single strain that has been cultured in threedifferent media. However, for the four different bacteria allgrown in the same media included in this analysis (Fig. 4), four

distinct and widely separated clusters are observed. Figures 3and 4 unequivocally argue that the growth media of properlywater-washed vegetative bacterial cells cannot be attributed tothe SERS signatures of the growth media in which they aregrown despite the close resemblance of some of the media andbacterial SERS spectra.

SERS Spectra of Bacteria Grown in Minimal Media. Inanother attempt to assess the relationship between the SERSspectra of bacteria and culture media, E. coli (11996) weregrown in M9 minimal media. As indicated in Table II, M9media offers a nearly inorganic diet for bacterial growth. Thenitrogen source is NH4Cl and the carbon source is glucose inlow concentration. Although bacterial growth is slower in thisminimal medium, results for bacteria grown in M9 were ofparticular interest because its ingredients are so markedlydifferent from those of the other media employed here, whichcontain far richer mixtures of organic nutrients to support cellgrowth. One consequence of this simple inorganic mixture isthat the corresponding SERS spectrum of the M9 medium isalso minimal, as seen in Fig. 5. Except for a low frequencyband at ;400 cm�1, the SERS spectrum of the M9 medium atthe concentration level used to grow bacteria is virtually devoidof any other vibrational features. In reference to the dilutionobservations reported by Marotta and Bottomley,23 furtherdilution of this aqueous solution only diminishes the observedsignal of the weak 400 cm�1 feature. The lack of any strongvibrational signatures in the 500 cm�1 to 1800 cm�1 range is incontrast to the SERS spectra of the more commonly usedbacterial growth culture media, as is evident from comparisonwith the spectra displayed in Fig. 2. Most notably there isclearly nothing in the ;730 cm�1 region of the M9 SERSspectrum.

The SERS spectrum of a washed E. coli strain (11996)grown in M9 minimal medium is also shown in Fig. 5. As isevident in this figure, a characteristic bacterial SERS spectrum(see Fig. 1) is obtained following cell culturing in this medium.For example, the strong feature at ;730 cm�1 is observed inthis E. coli SERS spectrum although there is no such band inthe M9 medium in which it was grown. The SERS spectrum of

FIG. 3. Barcode-based PCA analysis of SERS spectra of E. faecalis grown inLB and TSB, P. aeruginosa and S. aureus grown in TSB, and growth mediaTSB and LB alone. The SERS spectra of E. faecalis grown in LB and TSB arefound to be identical.

FIG. 4. Barcode-based PCA analysis of SERS spectra of S. aureus grown inCB, TSB, and BHI and P. aeruginosa, E. coli, and E. faecalis grown in TSB.The SERS spectra of S. aureus grown in CB, TSB, and BHI are found to beidentical.

496 Volume 65, Number 5, 2011

this E. coli strain is very similar to that obtained for growth inTSB (Fig. 1), although some small differences are evident forthe SERS spectra of this strain grown in these extremelydifferent growth media. Of most direct significance to thisstudy, the molecular species giving rise to the observedbacterial SERS spectrum are not those molecular species foundin the growth media. It is unequivocally an inherentbiochemical mechanism of these bacterial cells that is givingrise to the molecular species that contribute to the observedSERS bacterial signatures.

Sample Preparation Procedure Analysis. In the priorreport suggesting SERS bacterial spectra result from dilutedgrowth media components,23 bacterial samples were preparedwithout washing and centrifugation cycles, the standardprocedure for enriching bacterial cells from their culturingmedium. Cell growth media suspensions were simply dilutedby factors of typically 102 prior to SERS data acquisition onthe nanostructured (Ag nanorods) substrates used in thosestudies.23 The bacterial sample preparation procedure followedin our laboratory8,18 and followed by other groups working inthis area11,15,17 is quite different than the simple dilutionpreparation procedure employed by Marotta and Bottomley.23

As summarized above and elsewhere,8,18 we wash and thencentrifuge the cultured bacteria 4 to 5 times before weresuspend in water and place this diluted sample on the SERSsubstrate.

In another attempt to more quantitatively investigate thehypothesis23 that the observed bacterial SERS spectra aredominated by culture media, we performed a multivariate dataanalysis of the SERS spectra of bacterial suspensions as afunction of wash cycle and the growth media alone. The resultsof a PCA barcode treatment of SERS spectra of diluted NB

growth medium, the initial B. cereus culture (NB growth mediaand bacteria), the re-suspended pellet resulting from the spun orcentrifuged initial culture media, and the re-suspended pelletfollowing four successive water-wash/centrifugation cycles,i.e., our normal sample preparation protocol, are given in Fig.6a. Several relevant conclusions may be drawn from the PCAanalysis represented by the PC2 vs. PC3 results in this figure.First, the SERS based NB and washed bacteria clusters do notoverlap and thus are, once again, not identical. The bacterial

FIG. 5. SERS spectra of M9 minimal media and E. coli (strain 11996) grownin M9 minimal media.

FIG. 6. Barcode-based PCA analysis of wash cycles of (a) B. cereus grown inNB and (b) K. pneumonia grown in NB. SERS spectra of NB media alone, theinitial liquid culture, and the suspended pellet of the spun initial culture are alsoincluded in this PCA analysis.

TABLE II. Summary of bacterial growth media used in this study.

Growth media Main ingredients Source

Difco Columbia Broth (CB) Pancreatic digest of casein, yeast extract, proteose peptone,dextrose, sodium chloride, tryptic digest of beef heart, tris

(hydroxymethyl) aminomethane HCl, sodium carbonate

BD Technologies

LB Broth (LB) Typtone, yeast extract, NaCl SigmaDifco Nutrient Broth (NB) Peptone, beef extract, yeast extract, sodium chloride BD TechnologiesTryptic soy broth (TSB) Casein peptone (pancreatic), soya peptone (papainic), NaCl SigmaBrain heart infusion (BHI) Calf brains, beef heart, peptone, dextrose, sodium chloride, disodium phosphate BD TechnologiesM9 minimal media (M9) Na2HPO4�7H2O, KH2PO4, NaCl, NH4Cl, MgSO4, CaCl2, glucose Sigma

APPLIED SPECTROSCOPY 497

culture, corresponding to a mix of the bacterial cells, eitherbefore or after centrifugation, exhibits some higher orderdifferences relative to the NB-media-only spectra as evidencedby the same PC2 score but different PC3 values for these setsof spectra. However, of greatest significance for the subject ofthis study, one can view the evolution of the B. cereus SERSspectral signature as a function of wash cycle from no wash,i.e., in liquid culturing medium, to four wash cycles. As isevident in the PC plot resulting from this barcode multivariatedata analysis result, significant changes in the SERS spectra areobserved from the no-wash to the second-wash sample. TheSERS bacterial spectra, however, exhibit virtually no changeafter the second water wash. The B. cereus SERS spectraclusters following 3 or 4 water washes are essentiallyoverlapping that of the second wash spectra.

This experiment was repeated with a second bacterialsample, K. pneumoniae cultured in NB broth. The correspond-ing PCA clustering results are given in Fig. 6b. The sametrends found for the evolution of the B. cereus SERS spectra asa function of washing cycle are observed for the K. pneumo-niae data. The clusters corresponding to the twice- and three-times-washed bacterial samples overlap and thus indicate nofurther SERS spectral changes after the second wash cycle. Thecluster of SERS spectra of the NB medium is well separatedfrom those of the washed K. pneumoniae. The initial cultureand re-suspended spun culture pellet are overlapped and againshare nearly the same PC2 value as the NB broth alone,consistent at least with growth media contributions to the un-washed bacterial culture spectra. With washing, the PCAclusters systematically move further away from the NB-media-only location on this two-dimensional PC contour.

The PCA clustering analysis represented by Figs. 6a and 6billustrates the following points. (1) The water-washing samplepreparation procedure is essential for the acquisition ofrepresentative reproducible SERS spectra of bacteria in orderto remove all vestiges of the culture media. This is usuallyaccomplished after two washing cycles as evidenced in Figs. 6aand 6b. Work from this laboratory is characterized by SERS dataacquisition after 3 to 4 water-washing cycles.8,9,18,21 (2) Thisadditional multivariate analysis also argues that the bacterialSERS spectra are clearly not due to residual growth media simplyadsorbed to the outer layer cell walls or still contaminating thesuspension of properly washed bacterial samples placed on theSERS substrates. (3) The robust signature of the SERS spectra ofbacteria is seen to systematically emerge from that of the growthmedium–whole cell mixture (bacterial culture) with successivecell wash cycles. The data in this figure make it clear how SERSspectra of diluted mixtures of growth media and bacteria could infact exhibit more similarity to one another when comparingdifferent bacterial species if only water dilution is employed forthe sample preparation,23 as compared to employing the standardbiochemical protocols of water washing and centrifugation.

Molecular Origin of Bacterial SERS Signals. Two otherarguments are offered by Marotta and Bottomley23 as supportfor their hypothesis that previously published SERS spectralfeatures of bacteria are predominantly attributable to growthmedia contamination. Both of these issues are based on thelogical, yet unproven, assumption about expectations of theorigins of the vibrational bands that constitute the reportedSERS spectra of bacteria. It is assumed that these spectrashould be dominated by vibrational features associated with theouter bacterial cell wall structure. It is noted that SERS spectra

of some Gram-positive and Gram-negative bacteria shareseveral common features and do not appear to have verydifferent spectral signatures23 despite significant differences inthe character of their outer cell walls.24 For example, the SERSspectra of Gram-positive S. aureus and E. faecalis do notexhibit consistently significant differences from the Gram-negative E. coli, K. pneumonia, and P. aeruginosa, as seen inFig. 1. In particular, the band at ;730 cm�1 is the most intensevibrational transition observed in all of the bacterial SERSspectra shown in Fig. 1. Because the outer cell wall structure isdifferent for Gram-positive and Gram-negative bacteria and yetthe observed SERS spectra are similar, the argument is madethat the published SERS spectra could result from cell growthmedia. However, other possibilities exist for the molecularorigins of these signals that may not be strictly correlated withstructural differences of the cell wall. For example, these SERSsignals may arise from other molecules that are involved withcell activities such as cell signaling or cell metabolism, whichcan offer chemical signatures not directly correlated with outercell wall structure, but which still offer species/strainspecificity. Cell lysis may also contribute to these signals,although this seems less likely given that the signal intensity islargest when Raman excitation is focused on the intact cellsand does not change with time in DI water or with additionalcentrifugation cycles. Furthermore, sonication of washed cellsdoes not result in increased SERS intensity.

A second additional argument offered in this earlier work23

was that the SERS spectrum appeared continuously distributedacross the SERS substrate, albeit with different intensities, ascompared to being localized only at random but concentratedregions where bacterial cells were located. However, againdepending on the nature of these bacterial signals, e.g., if theywere due to metabolites or signaling molecules, they could bepresent in the sample cell suspension. Furthermore, if thesesignals were due to the cell growth media from the diluted cellgrowth media suspension sample preparation, then of coursethese signals, which can resemble bacterial SERS signals (seeFigs. 1 and 2), will be observed over the SERS substrate surfaceand not just near the local concentrations of aggregated bacterialcells evident in scanning electron microscopy (SEM) images.

CONCLUSION

The purpose of this paper is to investigate and evaluate thehypothesis offered by Marotta and Bottomley23 that previouslyreported SERS spectra of bacteria were due to cell growthmedia that had not been completely removed from culturedbacterial cell samples that were placed on the SERS substrates.This hypothesis was based on (1) the resemblance between theSERS spectra of bacteria and of several cell growth media, (2)the appearance of a SERS spectrum of the media upon dilution,(3) the similarity between SERS spectra of Gram-positive andGram-negative bacterial species, and (4) the spatial distribu-tions of the bacteria-like spectrum on the SERS substrate.However, the multivariate data analysis of the SERS spectra ofbacteria prepared by the standard multiple washing/centrifuga-tion protocol shows that different bacterial species grown in thesame media exhibit different SERS spectra, the same bacterialspecies grown in different media show the same SERS spectra,and, finally, SERS spectra of bacteria and the culture media inwhich they are grown are reproducibly distinct. The compar-ison of the M9 minimal media and the E. coli strain cultured inM9 SERS spectra is an extreme example of this difference

498 Volume 65, Number 5, 2011

between the inherent SERS signature of (washed) bacterialcells and that of the cell growth medium.

The apparent similarity in appearance between the SERSspectra of the bacterial cells and the growth media undoubtedlyreflects the fact that there are several components or moleculesstructurally similar to some components of these cell growthmedia that also contribute to the SERS spectra of washedbacterial cells. This appears to be particularly true for mediacontaining extracts from other microorganisms such as yeastextract, as seen in Fig. 1. The molecular origins of thesebacterial SERS signals more specifically will be the subject of asubsequent paper from this laboratory.

The appearance of a SERS spectrum of some media uponwater dilution, as reported by Marotta and Bottomley, isprobably just due to the optical or physical thickness of thenon-diluted medium sample. This interpretation is supported bythe reported SERS spectrum of undiluted growth media, whichis not a SERS spectrum since a large fluorescence backgroundis observed.23 Spontaneous fluorescence is effectivelyquenched when SERS can be observed due to the efficientenergy transfer to the metal when the analyte is close to themetal surface.8 As discussed above, the expected significantdifferences between SERS spectra of Gram-positive and Gram-negative bacteria and the spatial distribution of the SERSbacterial signal itself on the SERS substrate presumes that thebacterial SERS signals predominantly arise from structuralcomponents of the outer cell wall, which may not be accurate.Furthermore, as shown here, removal of cell growth culturemedia can be accomplished by following standard water-washing/centrifugation cycles (;3). If the bacterial cells aresimply diluted in water, the observed SERS spectrum mayindeed be contaminated by residual growth media (Fig. 6). Inconclusion, the results of this study argue that the previouslyreported bacterial SERS signals do not arise from residualgrowth media, as suggested by Marotta and Bottomley,23 andthe use of SERS to achieve rapid, portable bacterial diagnosticsremains a possibility as long as proper protocols are used toprepare these vegetative cells.

ACKNOWLEDGMENTS

The support of the National Institute of Health (Grant # 1R01AI090815-01)and BD Technologies, Research Triangle Park, NC, are gratefully acknowl-edged. We thank Dr. Jean Lee, Harvard Medical School, Boston, MA, forhelpful discussions about M9 minimal media, Dr. Don Moyer at Microbiotix,

Inc., Worcester, MA, and Dr. Kristin Weidemaier at BD Technologies,Research Triangle Park, NC, for supplying some of the bacterial strains usedhere.

1. S. Efrima, B. V. Bronk, and J. Czege, Proc. SPIE-Int. Soc. Opt. Eng. 3602,164 (1999).

2. L. Zeiri, B. V. Bronk, Y. Shabtai, J. Czege, and S. Efrima, Colloid Surf. A208, 357 (2002).

3. A. A. Guzelian, J. M. Sylvia, J. Janni, S. L. Clauson, and K. M. Spenser,Proc. SPIE-Int. Soc. Opt. Eng. 4577, 182 (2002).

4. N. F. Fell, Jr., A. G. B. Smith, M. Vellone, and A. W. Fountain III, Proc.SPIE-Int. Soc. Opt. Eng. 4577, 174 (2002).

5. R. M. Jarvis, A. Brooker, and R. Goodacre, Anal. Chem. 76, 5198 (2004).

6. R. M. Jarvis and R. Goodacre, Anal. Chem. 76, 40 (2004).

7. R. M. Jarvis, A. Brooker, and R. Goodacre, Faraday Discuss. 132, 281(2006).

8. W. R. Premasiri, D. T. Moir, M. S. Klempner, N. Krieger, G. Jones II, andL. D. Ziegler, J. Phys. Chem. B 109, 312 (2005).

9. W. R. Premasiri, D. T. Moir, M. S. Klempner, and L. D. Ziegler, in NewApproaches in Biomedical Spectroscopy, K. Kneipp, R. Aroca, H. Kneipp,and E. Wentrup-Byrne, Eds. (Oxford University Press, New York, 2007),p. 164.

10. A. Sengupta, M. L. Laucks, and E. J. Davis, Appl. Spectrosc. 59, 1016(2005).

11. M. Kahraman, M. M. Yazici, F. Sahin, O. F. Bayrak, and M. Culha, Appl.Spectrosc. 61, 479 (2007).

12. M. Kahraman, M. M. Yazici, F. Sahin, and M. Culha, Langmuir 24, 894(2008).

13. M. Culha, M. Kahraman, D. Cam, I. Sayin, and K. Keseroglu, Surf.Interface Anal. 42, 462 (2010).

14. H. Chu, Y. Huang, and Y. Zhao, Appl. Spectrosc. 62, 922 (2008).

15. T.-T. Liu, Y.-H. Lin, C.-H. Hung, T.-J. Liu, Y. Chen, Y.-C. Huang, T.-H.Tsai, H.-H. Wang, D.-W. Wang, J.-K. Wang, Y.-L. Wang, and C.-H. Lin,PLoS ONE 4, e5470 (2009).

16. L. Zeiri and S. Efrima, J. Raman Spectrosc. 36, 667 (2005).

17. L. Zeiri, B. V. Bronk, Y. Shabtai, J. Eichler, and S. Efrima, Appl.Spectrosc. 58, 33 (2004).

18. I. S. Patel, W. R. Premasiri, D. T. Moir, and L. D. Ziegler, J. RamanSpectrosc. 39, 1660 (2008).

19. J. Guicheteau and S. D. Christesen, Proc. SPIE-Int. Soc. Opt. Eng. 6218,62180G-1-80G-8 (2006).

20. Y. Liu, Y.-R. Chen, X. Nou, and K. Chao, Appl. Spectrosc. 61, 824(2007).

21. W. R. Premasiri, D. T. Moir, and L. D. Ziegler, Proc. SPIE-Int. Soc. Opt.Eng. 5795, 19 (2005).

22. J. Guicheteau, S. D. Christesen, D. Emge, and A. Tripathi, J. RamanSpectrosc. 41, 1632 (2010).

23. N. E. Marotta and L. A. Bottomley, Appl. Spectrosc. 64, 601 (2010).

24. J. W. Lengeler, G. Drew, and H. G. Schlegel, Biology of the Prokaryotes(Blackwell Science, Thieme, 1999).

APPLIED SPECTROSCOPY 499