High-Throughput ScreeningDOI: 10.1002/anie.201408558

Direct Surface-Enhanced Raman Scattering Analysis of DNADuplexes**Luca Guerrini,* Zeljka Krpetic, Danny van Lierop, Ramon A. Alvarez-Puebla, andDuncan Graham

Abstract: The exploration of the genetic information carriedby DNA has become a major scientific challenge. RoutineDNA analysis, such as PCR, still suffers from importantintrinsic limitations. Surface-enhanced Raman spectroscopy(SERS) has emerged as an outstanding opportunity for thedevelopment of DNA analysis, but its application to duplexes(dsDNA) has been largely hampered by reproducibility and/orsensitivity issues. A simple strategy is presented to performultrasensitive direct label-free analysis of unmodified dsDNAwith the means of SERS by using positively charged silvercolloids. Electrostatic adhesion of DNA promotes nanoparticleaggregation into stable clusters yielding intense and reprodu-cible SERS spectra at nanogram level. As potential applica-tions, we report the quantitative recognition of hybridizationevents as well as the first examples of SERS recognition ofsingle base mismatches and base methylations (5-methylatedcytosine and N6-methylated Adenine) in duplexes.

DNA is the custodian of the memory of life. Although verystable, it is subject to modifications that are associated withevolution but more frequently with genetic diseases orcancer.[1] Owing to the small content of DNA in cells and itsstructural complexity, its analysis require denaturation (heat),fragmentation (restriction endonucleases), separation (elec-trophoresis), amplification (polymerase chain reaction, PCR)and detection (quantitative PCR or microarray techniques).[2]

Besides the cost both in time and money, during theseprocesses samples are subject to alterations[3] and losses ofepigenetic information.[4] The advent of nanophotonics offersa unique opportunity for the development of direct, fast, andultrasensitive methods for DNA analysis.[5] In the broad fieldof plasmonics, surface-enhanced Raman scattering (SERS)

spectroscopy has arisen as a powerful analytical tool fordetection and structural characterization of biomolecules.[6]

The large majority of SERS-based DNA detection strategiesrely on the selective recognition between two complementarystrands to identify a specific sequence, and the use of extrinsicRaman labels for SERS readout.[7] However, labeling of theDNA strands requires complex and costly chemistry, and doesnot provide any chemical-specific information.[8] A secondapproach is based on the direct detection of the distinctiveSERS signal from DNA strands directly adsorbed onto thenanostructured surface. This strategy shows outstandingpotential in terms of sensitivity, selectivity, and chemical-specific information.[8, 9] To date, however, direct SERSanalysis of unmodified DNA is mainly restricted to single-stranded DNA sequences (ssDNA), as the direct contact ofdouble-stranded sequences (dsDNA) with nanostructuredmetallic surfaces (negative at the physiological pH) is largelyhindered by the negative charge of the phosphate backbone,unless relatively high DNA concentrations are employed.[9b,c]

As a result, dsDNA SERS spectra traditionally suffer frominherent poor spectral reproducibility[8,9b] and/or limitedsensitivity, hampering the extended application of label-freeSERS strategies for the study of unmodified duplexes. Thus,successful translation of the analytical potential of SERS tothe direct study of dsDNA would be a great leap forward forthe full exploration of the genetic information.

To overcome this major challenge, in this study we usepositively charged silver nanoparticles coated with sperminemolecules (AgNP@Sp). Spermine bound to AgNPs acts asstabilizer, and upon addition of negatively-charged DNA, itpromotes NP aggregation into stable particle clusters.[10] Thishas important consequences for the direct SERS analysis of

[*] Dr. L. Guerrini, Prof. R. A. Alvarez-PueblaDepartamento de Qu�mica Fisica e InorganicaUniversitat Rovira i Virgili and Centro de Tecnologia Qu�micaCarrer de Marcel l� Domingo s/n, 43007 Tarragona (Spain)andMedcom Advance SA, Viladecans Bussines Park, Edificio BrasilC/Bertran i Musitu, 83–85, 08840 Viladecans (Barcelona) (Spain)E-mail: luca.guerrini@ctqc.org

Dr. Z. KrpeticSchool of Chemistry and Chemical BiologyCentre for BioNano Interactions, University College DublinBelfield, Dublin 4 (Ireland)

Dr. Z. Krpetic, Dr. D. van Lierop, Prof. D. GrahamCentre for Molecular Nanometrology, Pure and Applied Chemistry,University of Strathclyde (UK)

Prof. R. A. Alvarez-PueblaICREA, Passeig Llu�s Companys 23, 08010 Barcelona (Spain)

[**] This work was funded by the Spanish Ministerio de Economiay Competitividad (CTQ2011-23167), the European Research Council(CrossSERS, FP7/2013 329131, PrioSERS FP7/2014 623527), andMedcom Advance SA. D.G. acknowledges support from the RoyalSociety through a Wolfson Research Merit Award. We are grateful toDr. Matteo Masetti (Department of Pharmacy and Biotechnology,Universit� di Bologna, Bologna, Italy) for the dsDNA drawings.

Supporting information for this article (experimental, AgNP@Spcharacterization (Section S1), DNA-driven aggregation ofAgNP@Sp and spectral reproducibility (Section S2), vibrationalassignment of dsDNA SERS spectra (Section S3), SERS of ss and dssequences on silver hydroxylamine colloids (Section S4), ds1 SERSspectrum and the pondered sum of the individual spectra of ss1 andssc (Section S5), and SERS spectra of ds2, ds3 and ds4 (Sec-tion S6)) is available on the WWW under http://dx.doi.org/10.1002/anie.201408558.

.AngewandteCommunications

1144 � 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2015, 54, 1144 –1148

dsDNA. Firstly, the DNA adsorption occurs by non-specificelectrostatic interaction of the phosphate groups rather thanby base-specific nucleoside–metal binding. Secondly, DNApromotes the nanoparticle aggregation removing the need ofexternal aggregating agents. On the one hand, this signifi-cantly simplifies the sensing scheme, reducing the number ofvariables to be controlled (type and concentration of theaggregating agent, time and order of addition, and so on). Onthe other hand, DNA sequences are selectively sandwichedbetween interparticle junctions (that is, hot-spots[11]), thusmaximizing the intensification of the Raman signal (Fig-ure 1a). Thirdly, the nanoparticle aggregation occurs fast andindependently of the base composition, generating long-termstable clusters in suspension, as demonstrated by the extinc-

tion spectra illustrated in Figure 1b. This also allows us toperform the SERS measurements after several hours uponthe analyte addition (that is, when the process of DNAadhesion has reached its equilibrium). Finally, SERS spectraare acquired in colloidal suspensions with a long workingdistance objective. In this experimental set-up, the recordedSERS spectra are the statistically-averaged result of a largeensemble of nanoparticle/DNA clusters in continuous Brow-nian motions within the scattering volume. Such averagedbulk SERS regime yields good-quality spectra with well-defined average band centers, bandwidths, and relativeintensities.[12] As a result, dsDNA electrostatic adhesiononto the AgNP@Sp nanoparticles allows for the ultrasensitiveanalysis of duplexes and their modifications at the nanogramregime with high batch-to-batch reproducibility, yieldingSERS spectra that are truly spectral representations of thenucleoside composition and organization within the helix. Toconfirm this, we show an ample set of examples of direct label-free SERS detection of unmodified duplexes, includingquantification of hybridization events as well as recognitionof single-base mismatches and base methylations (5-methy-lated cytosine and N6-methylated adenine), which to datewere restricted to single-stranded sequences.[9a,d]

The average SERS spectra of two complementary single-stranded sequences (ss1 and ssc) and their correspondingdouble-helix structure (ds1) on AgNP@Sp colloids are shownin Figure 1c. The overall dsDNA concentration was keptconstant throughout the entire study at ca. 630 ng mL�1

(corresponding to less than 3.0 ng in the probed volume; seethe Supporting Information, Section S2) and appropriatelyselected to yield long-term stable clusters in suspension withreproducible and unvaried SERS spectral profiles (Sec-tion S2). Vibrational assignment of dsDNA was based onliterature references and comparative spectral analysis withhomo- and bi-polymeric sequences (Section S3). In contrastwith previous results on negatively charged colloids (Sec-tion S4),[13] when the individual ss units hybridize into the dsa large reshaping of their SERS spectra is observed (Fig-ure 1c), influencing peak position, bandwidths, and relativeintensity. Notably, if ss1 and ssc spectra show small changes inrelative intensity that are associated with the base composi-tion, the stacking and pairing of the nucleobases into thehybridized native structure is revealed by several character-istic features of the Raman melting profiles.[14] Among others,we highlight the marked intensity decrease and peak-shiftingof the ring breathing modes (700–800 cm�1) and the carbonylstretching modes (1653 cm�1), the latter being extremelysensitive to the disruption of Watson–Crick hydrogen bond-ing.[14] We also observe a general narrowing of the Ramanfeatures in the 1150–1600 cm�1 spectral region, containingmany overlapping bands mainly arising from in-plane vibra-tions of base residues.[15] Such spectral modification can beascribed to the adoption of a more rigid and well-definedconformation by the two individual ss when they self-organizeinto the rigid duplex geometry. The investigation of a set ofsolutions containing different ds1 and ssc molar ratios, Rhybr =

[ds1]/([ds1] + [ssc]) (Figure 1 d) shows a progressive shift ofthe ring breathing bands of adenine (from 734 cm�1 to730 cm�1) when the DNA population is enriched with ds1.

Figure 1. a) Diagram of dsDNA sandwiched between two positivelycharged AgNP@Sp. b) Extinction spectra of pure AgNP@Sp colloidsand in the presence of ds1 (final conc. 630 ng mL�1) acquired 2 h and24 h upon DNA addition (the AgNP@Sp + ds1 mixture was sonicatedfor few seconds before the spectral acquisition). c) SERS spectra ofss1 and ssc (3.15 mgmL�1), and ds1 (630 ng mL�1) on AgNP@Sp.SERS spectra were normalized to the PO2

� stretching at 1090 cm�1.The ds1 SERS spectrum in the 2700–3050 cm�1 region was divided bytwo. d) Detail of the 680–890 cm�1 spectral region for the SERS spectraof ssc + ds1 mixtures at different molar ratiosRhybr = [ds1]/([ds1]+ [ssc]) (from top to bottom: 0.011, 0.024, 0.041,0.063, 0.091, 0.130, 0.189, 0.286, 0.474, and 1). The ds1 concentrationwas progressively increased from 0 to 630 ng mL�1 while the sscconcentration was simultaneously decreased from 3.15 mgmL�1 ng tozero. The dotted line is the difference spectrum ssc-ds1 obtained bydigitally subtracting the SERS spectra shown in (c). e) Ratiometricpeak intensities I724/I738 vs. Rhybr = [ds1]/([ds1]+ [ssc]) molar ratio.

AngewandteChemie

1145Angew. Chem. Int. Ed. 2015, 54, 1144 –1148 � 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org

The spectral subtraction of the SERS spectra allows us to fullydisclose the spectral alterations associated with changes in theDNA structure. In particular, the difference spectrum ds1-ssc(dotted line, Figure 1d) reveals a minimum at 724 cm�1 anda maximum at 738 cm�1. The corresponding ratiometric peakintensities I724/I738 were then selected to monitor the relativeds1/ssc populations in the samples, and plotted against Rhybr

(Figure 1e). Data shows that ds sequences can be clearlyidentified within the mixture even when present at values< 1%, whereas SERS spectra are fully dominated by ds1contributions for duplex populations � 20 %. Such result isconsistent with the higher affinity of dsDNA to AgNP@Spowing to the larger availability of negatively charged phos-phate groups as compared to ssDNA. Linear correlation (r2>

0.99) is observed in the 0.01–0.19 Rhybr range.To test the possibility of detecting single-base mismatches

in DNA duplexes, SERS spectra of the full-complementaryds1 and the heteroduplexes ds2, ds3, and ds4 were acquiredand compared (Figure 2). These heteroduplexes contain one

adenine base, A, in place of: (ds2) one guanine, G, (ds3) onecytosine, C, (terminal position), and (ds4) one cytosine(internal position). Subtraction of the SERS spectra of ds1from the other samples generates difference spectra contain-ing vibrational signatures associated with the additional(positive features) and removed nucleobase (negative fea-tures). Positive A bands emerge in all difference spectra at730 and 1507 cm�1, whereas negative G features are observedat ca. 620, 675, and 1354 cm�1 in the ds2�ds1. In this case,negative bands also appear at 1487 and 1577 cm�1, ascribed topurine modes (mainly G contribution) while the other purinefeature at 1325 cm�1 (mainly A contribution) does not suffer

from a marked change in the relative intensity. These results,together with the lack of significant changes in SERS intensityof the pyrimidine bases, clearly indicate that the spectralalterations are ascribed to the A!G base-mismatch in theds2. On the other hand, ds3�ds1 and ds4�ds1 differencespectra show very similar spectral patterns where, in additionto the positive A (730 and 1507 cm�1) and the negative C(1250 and 1528 cm�1) contributions, a consistent intensityincrease of the purine bands (1325, 1487, and 1577 cm�1) isobserved. The pyrimidine ring breathing (787 cm�1, C + T)also undergoes a drastic intensity decrease whereas thyminemarker bands do not reveal significant alterations. Thesespectral changes can therefore be associated with the A!Cbase-mismatch in ds3 and ds4. Minor differences between ds3and ds4 difference spectra (Figure 2) can be ascribed to thedifferent position of the base mismatch within thesequence.[15]

The potential application of AgNP@Sp in SERS analysisof 5-methylated cytosine bases (mC) and N6-methylatedadenine bases (mA) within ds structures was examined byacquiring the SERS spectra of dsmC and dsmA and comparingto their non-methylated analogous ds1. In dsmC and dsmA, theC or A nucleobases of one of the strands were all replaced bytheir respective 5-methylated and N6-methylated counter-parts (see sequence structures in Figure 3 and Figure 4). All

the spectral changes are in good agreement with the normalRaman studies of the substitution of the methyl group into the5C and N6 atom in cytosine[16] and adenosine[17] nucleosides,respectively. The major spectral markers of the cytosinemethylation can be easily recognized through the wholedifference spectrum (Figure 3a). Among others, we observea red-shift and relative intensity decrease of the pyrimidinering breathing band at 787 cm�1, together with the appearanceof a new weak feature at ca. 755 cm�1. In the 1200–1450 cm�1

Figure 2. SERS spectra of ds1 and ds2; and digitally subtracted SERSspectra ds2�ds1, ds3�ds1, and ds4�ds1. For the sake of clarity, thedigitally subtracted spectra were multiplied by 3 (for ds2�ds1) and 4(for ds3�ds1 and ds4�ds1). Frequency positions of characteristicnucleotide bands are highlighted in: blue (adenine), yellow (cytosine),orange (guanine), and green (thymine). The overall ds concentrationwas kept constant at 630 ng mL�1.

Figure 3. a) Molecular structure of 5-methyl cytosine (mC). SERS spec-tra of ds1 and dsmC, and the difference spectrum dsmC-ds1. For thesake of clarity, the digitally subtracted spectrum was multiplied bya factor of two. b) Detail of the 700–840 cm�1 region for the SERSspectra of ds1 + dsmC at different cytosine base ratios RmC = [mC]/([C] + [mC]) (from the top to the bottom, RmC = 0, 0.045, 0.091, 0.182,0.273, 0.364, and 0.455). The overall ds concentration was keptconstant at 630 ngmL�1. c) Ratiometric peak intensities I732/I788 vs. themolar ratio RmC = [mC]/([C]+ [mC]).

.AngewandteCommunications

1146 www.angewandte.org � 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2015, 54, 1144 –1148

spectral region, the normal Raman of unmodified cytidineaqueous solution contains two main bands at 1242 and1292 cm�1, whereas the 5-methyl substituted shows a verydifferent spectral signature with three discernible bands at1224, 1258, and 1302 cm�1 and a new very intense feature at1362 cm�1.[16] Accordingly, the difference spectrum in Fig-ure 3a reveals two negative bands at ca. 1244 and 1287 cm�1;three positive contributions broadly centered at 1218, 1268,and 1315 cm�1, and a very strong and sharp band at 1362 cm�1.Moreover, the C ring band at about 1510 cm�1 also undergoesa marked red-shift and intensity decrease while the broadcarbonyl stretching band centered at 1653 cm�1 shows a largered-shift (well highlighted in the difference spectrum by thenegative feature at ca. 1622 cm�1 and the positive one at ca.1662 cm�1).

The spectral modifications of the SERS spectral profile ofthe normal DNA upon methylation of six of their eleven theA bases are also striking (Figure 4a, difference spectrum).The A ring breathing band at 731 cm�1 undergoes a drastic11 nm up-shift together with a notable intensity decrease,whereas the pyrimidine stretching of the adenine at 1509 cm�1

almost disappears from the spectrum. Furthermore, weobserve up-shifts of the phosphate stretching mode at1090 cm�1 (together with a marked intensity increase); thev(CN) imidazole feature at 1487 cm�1 and the 1577 cm�1 bandascribed to base ring modes (mainly G + A).

Quantification of relative methylated base populationswithin the sample was investigated by monitoring theratiometric peak intensities I732/I788 for mC and I743/I730 formA, at different molar ratios (RmC = [mC]/([C] + [mC]) andRmA = [mA]/([A] + [mA])). Details of the ring breathing spec-tral regions are illustrated in Figure 3b and Figure 4b. In thecase of dsmC (Figure 3b), we highlight the progressiveintensity decrease of the C + T ring breathing band at787 cm�1 upon increasing of the relative mC populations.Differently, for dsmA (Figure 4b), the A ring breathing bandat 730 cm�1 suffers a dramatic intensity decrease and large

peak red-shift as the RmA value becomes larger. Linearcorrelations were obtained in both cases (r2> 0.99 for mA and> 0.94 for mC) with detection limits of one mC per 22 totalcytosine bases and one mA per 10 total adenine bases in thesample (Figure 3c and Figure 4c, respectively).

In summary, we have demonstrated the potential appli-cation of SERS for the development of a fast and affordablehigh-throughput screening method for gaining detailedgenomic information on DNA duplexes. Importantly, owingto the low amount of DNA required (comparable to only 10–100 cells), the analysis can be performed without the necessityof amplification, thus providing realistic direct information ofthe DNA in its native state. Owing to the direct nature of thissensitive and selective assay, we anticipate these findings ofkey importance in a number of different fields includingdiagnosis, genetic engineering, biotechnology, drug discovery(antibiogram in microbiology), agriculture, and forensicscience.[1, 18]

Received: August 26, 2014Revised: September 30, 2014Published online: November 24, 2014

.Keywords: DNA · hybridization · methylated bases ·single-base mismatch · surface-enhanced Raman spectroscopy

[1] a) L. Pikor, K. Thu, E. Vucic, W. Lam, Cancer Metastasis Rev.2013, 32, 341 – 352; b) S. P. Jackson, J. Bartek, Nature 2009, 461,1071 – 1078.

[2] a) S. Yang, R. E. Rothman, Lancet Infect. Dis. 2004, 4, 337 – 348;b) A. Git, H. Dvinge, M. Salmon-Divon, M. Osborne, C. Kutter,J. Hadfield, P. Bertone, C. Caldas, RNA-Publ. RNA Soc. 2010,16, 991 – 1006.

[3] X. Qiu, L. Wu, H. Huang, P. E. McDonel, A. V. Palumbo, J. M.Tiedje, J. Zhou, Appl. Environ. Microbiol. 2001, 67, 880 – 887.

[4] E. E. Schadt, S. Turner, A. Kasarskis, Human Molec. Genetics2010, 19, R227 – R240.

[5] a) J. N. Anker, W. P. Hall, O. Lyandres, N. C. Shah, J. Zhao, R. P.Van Duyne, Nat. Mater. 2008, 7, 442 – 453; b) K. Saha, S. S.Agasti, C. Kim, X. N. Li, V. M. Rotello, Chem. Rev. 2012, 112,2739 – 2779.

[6] a) R. A. Alvarez-Puebla, L. M. Liz-Marzan, Small 2010, 6, 604 –610; b) W. Xie, S. Schl�cker, Phys. Chem. Chem. Phys. 2013, 15,5329 – 5344.

[7] D. Graham, K. Faulds, Chem. Soc. Rev. 2008, 37, 1042 – 1051.[8] A. Barhoumi, D. Zhang, F. Tam, N. J. Halas, J. Am. Chem. Soc.

2008, 130, 5523 – 5529.[9] a) A. Barhoumi, N. J. Halas, J. Phys. Chem. Lett. 2011, 2, 3118 –

3123; b) S. R. Panikkanvalappil, M. A. Mackey, M. A. El-Sayed,J. Am. Chem. Soc. 2013, 135, 4815 – 4821; c) S. R. Panikkanva-lappil, M. A. Mahmoud, M. A. Mackey, M. A. El-Sayed, ACSNano 2013, 7, 7524 – 7533; d) E. Papadopoulou, S. E. J. Bell,Angew. Chem. Int. Ed. 2011, 50, 9058 – 9061; Angew. Chem. 2011,123, 9224 – 9227; e) Z. Krpetic, I. Singh, W. Su, L. Guerrini, K.Faulds, G. A. Burley, D. Graham, J. Am. Chem. Soc. 2012, 134,8356 – 8359.

[10] D. van Lierop, Z. Krpetic, L. Guerrini, I. A. Larmour, J. A.Dougan, K. Faulds, D. Graham, Chem. Commun. 2012, 48,8192 – 8194.

[11] F. J. Garc�a de Abajo, Rev. Mod. Phys. 2007, 79, 1267 – 1290.[12] R. Aroca, Surface-enhanced Vibrational Spectroscopy, Wiley,

Chichester, 2006.

Figure 4. a) Molecular structure of N6-methyladenine (mA). SERS spec-tra of ds1 and dsmA, and the difference spectrum dsmA�ds1. For thesake of clarity, the digitally subtracted spectrum was multiplied by two.b) Detail of the 700–760 cm�1 region for the SERS spectra ofds1 + dsmA at different adenine base ratios RmA = [mA]/([A]+ [mA])(from the top to the bottom RmA = 0, 0.1, 0.2, 0.3, 0.4, 0.5, and 0.6).The overall ds concentration was kept constant at 630 ng mL�1.c) Ratiometric peak intensities I743/I730 vs. molar ratio RmA.

AngewandteChemie

1147Angew. Chem. Int. Ed. 2015, 54, 1144 –1148 � 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org

[13] a) E. Papadopoulou, S. E. J. Bell, Chem. Eur. J. 2012, 18, 5394 –5400; b) N. E. Marotta, K. R. Beavers, L. A. Bottomley, Anal.Chem. 2013, 85, 1440 – 1446.

[14] J. G. Duguid, V. A. Bloomfield, J. M. Benevides, G. J. Thomas,Biophys. J. 1996, 71, 3350 – 3360.

[15] H. Deng, V. A. Bloomfield, J. M. Benevides, G. J. Thomas,Biopolymers 1999, 50, 656 – 666.

[16] A. Gfrçrer, M. E. Schnetter, J. Wolfrum, K. O. Greulich, Ber.Bunsen-Ges. 1991, 95, 824 – 833.

[17] S. Mansy, W. L. Peticolas, R. S. Tobias, Spectrochim. Acta Part A1979, 35, 315 – 329.

[18] a) G. H. Clever, M. Shionoya, Coord. Chem. Rev. 2010, 254,2391 – 2402; b) C. J. Lord, A. Ashworth, Nature 2012, 481, 287 –294.

.AngewandteCommunications

1148 www.angewandte.org � 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2015, 54, 1144 –1148

Supporting Information

� Wiley-VCH 2015

69451 Weinheim, Germany

Direct Surface-Enhanced Raman Scattering Analysis of DNADuplexes**Luca Guerrini,* Zeljka Krpetic, Danny van Lierop, Ramon A. Alvarez-Puebla, andDuncan Graham

anie_201408558_sm_miscellaneous_information.pdf

Abbreviations

ssDNA – single stranded DNA

dsDNA – double stranded DNA

AgNP – silver nanoparticles

AgNP@Sp – Spermine-coated silver nanoparticles

AgHX – hydroxylamine-reduced silver nanoparticles

A – adenine

C – cytosine

G – guanine

T – thymine

mC – 5-methylated cytosine

mA – N6-methylated adenine

2

Experimental Section

Materials. All materials were of highest purity available and obtained from Sigma Aldrich, unless

stated otherwise. DNA oligonucleotides were purchased from Eurofins MWG Operon. The

oligonucleotide base sequences are listed in Table 1. Stock solutions of each oligonucleotide were

prepared in Milli-Q water (final concentration ca. 4x10-4 M). Annealing was conducted by heating to

95°C for 10 minutes equimolar solutions of oligonucleotides ss1, ss2, ss3, ss4, ss5, ssmC and ssmA;

and their complementary strand ssc in PBS (0.3 M). This yielded the corresponding double-stranded

DNA solutions ds1, ds2, ds3, ds4, ds5, dsmC and dsmA (final concentration 10-5 M) which were

stored at -20 °C until required. As reference samples, we selected 21 base homopolymeric sequences

(pA, pC, pT and pG) as well as 22 base self-complementary oligonucleotides (ssCG and ssAT). The

reference oligonucleotide base sequences are listed in Table 2. Annealing of ssCG and ssAT was

conducted as indicated before and the resulting dsCG and dsAT samples (final concentration 10-5 M)

were stored at -20 °C.

Table 1. Single-stranded DNA sequences (ssDNA) used in the study.

Single-strand Sequences

ss1

ss2

ss3

ss4

ssmC

ssmA

ssc

CAT CGC AGG TAC CTG TAA GAG

CAT CGC AGG TAC CTG TAA GAA

AAT CGC AGG TAC CTG TAA GAG

CAT CGC AGG TAA CTG TAA GAG

mCAT mCGmC AGG TAmC mCTG TAA GAG

CmAT CGC mAGG TmAC CTG TmAmA GmAG

GTA GCG TCC ATG GAC ATT CTC

mC and mA indicate 5-methyl Cytosine and N6-methyl Adenine residues, respectively (see Fig. S5a for

nucleobase structures).

Table 2. Reference single-stranded DNA sequences (ssDNA) used in the study.

Single-strand Sequences

pA

pC

pT

pG

ssCG

ssAT

AAA AAA AAA AAA AAA AAA AAA

CCC CCC CCC CCC CCC CCC CCC

TTT TTT TTT TTT TTT TTT TTT

GGG GGG GGG GGG GGG GGG GGG

CCG CGC CGC GCG CGC GGC GCGG

AAT ATA ATA TAT ATA TTA TATT

3

Synthesis of Silver Colloids. Synthesis of Positively-Charged Silver Nanoparticles (AgNP@Sp).

Spermine coated-silver nanoparticles (AgNP@Sp) were prepared as previously reported.[1] Briefly,

20 μL of a 0.5 M AgNO3 solution were added to 10 mL of Milli-Q water, followed by addition of 7

μL of a 0.1 M spermine tetrahydrochloride. Subsequently, under vigorous stirring, 250 μL of a

freshly prepared NaBH4 (0.01 M aqueous solution) were added drop wise to the mixture under

stirring. Finally, the solution was gently stirred for 20 min. Glass vials were previously coated with

polyethylene immine (PEI, average mw ca. 25,000 by LS) by an overnight immersion into an

aqueous 0.2% v/v PEI solution, followed by extensive rinsing with Milli-Q water and N2 dried. See

section S1 for AgNP@Sp characterization.

Synthesis of Negatively-Charged Silver Hydroxylamine Nanoparticles (AgHX). Silver nanoparticles

obtained by chemical reduction of Ag+ using hydroxylamine hydrochloride (AgHX) were

synthesized as described by Leopold et al.[2] Briefly, 300 μL of a NaOH 0.1 M solution were added

to 90 mL of a hydroxylamine hydrochloride 1.66 mM solution. Subsequently, 10 mL of a AgNO3

9.06 mM solution were added drop by drop to the mixture under vigorous stirring.

SERS Experiments. For SERS studies, 1 μL of PBS (0.3 M) solutions of dsDNA (10-5 M) or 1 μL

of PBS (0.3 M) ssDNA solutions (10-4 M) were mixed with 200 μL of AgNP@Sp. The final analyte

concentration in the sample was 5x10-8 M for dsDNAs and 5x10-7 M for ssDNAs, which corresponds

to approximately 630 ng/mL and 3.15 μg/mL, respectively. After the addition of the ds/ss DNA, the

colloids were left to equilibrate for 2 h and redispersed by quick sonication before running the SERS

measurements.

DNA Hybridization Study. Mixed ss and ds DNA samples were prepared by combining different

volumes of 10-5 M ssc and ds1 solutions (the final ssc+ds1 concentration in the mixtures was kept

constant to 10-5 M) corresponding to the following molar ratios, R= [ds1]/([ds1]+[ss5]) = 0.011,

0.024, 0.041, 0.063, 0.091, 0.130, 0.189, 0.286, 0.474 and 1.

Detection of methylated bases in dsDNA. Mixed ds samples were prepared by combining different

volumes of 10-5 M solutions of ds1 and dsm (either mC or mA). The final ds concentration in the

mixtures was kept constant to 10-5 M. Since the DNA spectra depend only on base composition, and

the interaction of DNA with the metal surface takes place via the external phosphate groups, we

could expect that having unmodified and methylated Adenine on different sequences should be

similar that having them on the same ds structure.[3]

Equipment and Instrumental Settings. CPS disc centrifuge DC24000 (CPS Instruments Inc.) was

used to measure the particle size distribution. A gradient fluid, 8-24 wt % sucrose solution in Milli-Q

4

water, was freshly prepared and filled successively in nine steps into the disc, rotating at a speed of

24 000 rpm starting with the solution of highest density. Calibration was performed using

poly(vinylchloride) particles (0.476 μm, Analytik Ltd.) as calibration standard before each

measurement. For calculations, particle density of 10.49 g/cm3 was used and a density of 1.385

g/cm3 of the calibration standard.

FEI Tecnai G2 20 Twin TEM operating at accelerating voltage of 200 kV was used for imaging.

UV-vis spectra were recorded using a Thermo Scientific Evolution 201 UV-visible

spectrophotometer.

The nanoparticle hydrodynamic diameter distribution was measured on a Malvern Nanosizer

(ZSeries) using a low-volume cuvette, averaging 11 runs at 25ºC.

The ζ (Zeta) potential measurements were carried out using a Malvern 2000 Zetasizer, using the

default method protocol and a minimum sample volume of 3 ml. Before each measurement a

standard solution of -68.0 ± 6.8 mV was measured.

SERS experiments were conducted using a Renishaw InVia Reflex confocal microscope equipped

with a high-resolution grating consisting of 1800 grooves/cm for visible wavelengths, additional

band-pass filter optics, and a CCD camera. A 532 nm laser was focused onto the sample by a long-

working distance objective (0.17 NA, working distance 30 mm) and the spectra were typically

acquired with an exposure time of 5x10 s. All SERS spectra reported in the manuscript were

obtained by averaging the SERS response of 4 different sample replications (N=4) obtained by

combining the same colloids (prepared the day before the measurement and left aging overnight)

with the specific DNA buffered solution. Baseline correction was applied to all spectra, unless stated

otherwise. Importantly, difference SERS spectra were calculated from the original no baseline-

corrected spectra to avoid any generation of spectral artefacts.

5

Section S1. AgNP@Sp characterization

Differential Centrifugal Sedimentation (DCS) measurements of spermine coated-silver nanoparticles

(AgNP@Sp) yielded a centered value at 31.5 nm (weight distribution, Fig. S1a). Additionally,

AgNP@Sp colloids show a hydrodynamic diameter of 48.0 nm, with a ζ-potential of +35.3 ± 7.6

mV. The final bulk pH is 6.1.

Nanoparticle concentration in the AgNP@Sp colloids (ca. 0.3 nM) was calculated by Beers law

using the extinction coefficient for silver nanoparticles of 1.85x1010 M-1 cm-1, derived from

literature.[4]

Background SERS spectrum of AgNP@Sp was obtained by aggregating the colloids with MgSO4 0.1

M. MgSO4 is a “passive” electrolyte that induces nanoparticle aggregation by simply increasing the

ionic concentration without firmly adsorbing onto the silver surface (as it occurs when halide salts

such as NaCl are used to aggregate the nanoparticles)[5]. AgNP@Sp colloids are characterized by a

low/null SERS background signal (Fig. S1c) which, in contrast to what occurs for citrate-reduced

colloids, does not obligate to the digital subtraction of intense spurious bands, thus removing an

additional factor that may affect the spectra-to-spectra reproducibility.

Figure S1. (a) Characterization of spermine coated-silver nanoparticles (AgNP@Sp) by Differential Centrifugal

Sedimentation (DCS). (b) Representative TEM micrograph of AgNP@Sp colloids prepared by evaporating 10 μL of

diluted colloidal dispersion onto formvar-coated copper grids. (c) Background SERS spectrum of AgNP@Sp aggregated

with MgSO4 0.1 M.

6

Section 2. DNA-driven aggregation of AgNP@Sp into stable clusters in

suspension and spectral reproducibility.

Spectral reproducibility is a key factor in determining the successful application of label-free SERS

in the DNA analysis. Aggregated nanoparticle systems are known to generate clusters with high

SERS activity but with poor control over the final geometry, size and distribution of the aggregates.

This lack of control may severely affect the reproducibility of the overall SERS intensity but, when

analyzed in the averaged bulk SERS regime, “the good-quality signal from the statistical average

SERS of an ensemble of scatterers normally produces stable and reproducible spectra with well-

defined average band centers, FWHMs and relative intensities”.[6] In this study, we are not

interested in the overall SERS intensity but we do analyze changes in the spectral profile (i.e.

average band centers, FWHMs and relative intensities). For this reason, SERS spectra are acquired in

colloidal suspensions with a long working distance objective adopting an experimental set-up where

a large ensemble of nanoparticle/DNA clusters, in continuous Brownian motions within the

scattering volume of the objective, are simultaneously investigated by the laser. The averaging effect

of these numerous contributions is a critical factor for obtaining ensemble signals which are stable

and reproducible.

Another key feature that contributes to the high spectral reproducibility relies on the fact that DNA

both acts as an aggregating agent of individual nanoparticles as well as a very effective stabilizing

agent of the so-formed clusters. This is particularly important since it allows us to measure the

sample after several hours upon the DNA addition and not during the aggregation process when the

conformational equilibrium of the DNA onto the metal is likely not achieved yet.

The DNA concentration employed throughout the whole study was carefully selected to yield long-

term stable clusters in suspension and unvaried SERS spectra. Fig. S2 illustrates the extinction and

SERS spectra of AgNP@Sp (nanoparticle concentration ca. 0.3 nM) in the presence of different

concentration of ds1, acquired 2 hours after the addition of the analyte. The spectra are stacked for

visual comparison. As can be seen in Fig. S2a, the LSPR of the colloids at 391 nm undergoes a

progressively weakening with the simultaneous appearance of a broad contribution at longer

wavelength when the ds1 concentration is raised from 0 to ca. 0.126 μg/mL. The samples, however,

do not show long-term stability and the Eppendorf walls, where the samples are stored, are

progressively coated with a layer of silver nanoparticles, indicating that the DNA coverage onto the

metal surface is not sufficient to “quench” the spermine positively charges and provide protection

against the wall-adhesion. Differently, when the ds1 concentration is further increased, we clearly

observe a change in the aggregation pattern with the progressive blue-shift of the broad plasmon

7

contribution associated with the cluster formation, from ca. 600 nm to ca. 500 nm. In this second

aggregation regime, the silver nanoparticles are fully stabilized by the DNA adsorption and the

extent of the aggregation decreases as the ds1 concentration increases (i.e. as much as the

nanoparticle surfaces are saturated with the analyte). The samples corresponding to ds1

Figure S2. (a) Extinction spectra and (b) SERS spectra of AgNP@Sp colloids in the presence of ds1 at different

concentration, 2 hours after the initial DNA addition (the samples were quickly sonicated prior to the measurements).

Top: representative TEM images corresponding to AgNP@Sp and AgNP@Sp+ds1 (0.63 μg/mL). These images provide

a qualitative visualization of the DNA-promoted aggregation. In this case, in the attempt to minimize drying-induced

aggregation effects, 30 μL of each colloidal sample were deposited onto formvar-coated copper grids in a humidity

chamber and left to settle for 20 min before being removed. The substrate was finally dried by N2 flow.

8

concentration above 0.315 μg/mL have shown long-term stability with null/minimal changes in the

extinction spectra even after several days.

The original non-baselined SERS spectra of the same samples are illustrated in Fig. S2b.

First of all, the results show that the Raman features of the DNA can be detected down to 63 ng/mL,

which corresponds to 12.6 ng in the 200 μL of colloids employed in the sample preparation.

However, the actual volume investigated by the laser is much smaller. The illuminated volume (V),

when using a macrosampling objective, is defined by the focal length, f, and the lens diameter, D. [7]

In our case, f= 30 mm and D= 10 mm. Therefore, for a 532 nm laser, V is approximately 40 μm3.

However, several experimental factors, often impossible to evaluate, contribute to make the

scattering volume much larger than the theoretical calculations do. For instance, it is known that

resolution criteria that are perfectly adequate for imaging are not sufficient for Raman spectroscopy

and the observed signal originates not only within the focal volume but tails off with distance (and

the extension of this spatially remote contribution is very case specific and difficult to estimate).[8] In

addition, the continuous diffusion of the NP-DNA clusters in-and-out the scattering volume increases

the actual number of investigated sequences and particles, and is strictly dependent of the cluster size

distribution, laser exposure time, number of accumulations etc.[9] To safely include all these

experimental factors and avoid unreliable claims on the levels of sensitivity, we conservatively

assume the laser investigated volume as a cylinder of 1 mm diameter and the depth of the sample (in

this case, 6 mm well). This corresponds to a volume of 4.7 μL and, therefore, the limit of detection

for ds1 analysis can be re-estimated as less than 0.3 ng in the probed volume.

However, more importantly than the limit of detection itself, the analysis of the SERS

spectral profiles indicates that the Raman features of the DNA remain unaltered when the ds1

concentration lies within the second aggregation regime (> 0.315 μg/mL) whereas, for lower analyte

amount, the SERS bands significantly broaden, weaken and, in some cases, undergo spectral shifts.

Based on these results, we selected 0.63 μg/mL as the optimum dsDNA concentration to perform our

studies (corresponding to less than 3.0 ng in the scattering volume), since it provides long-term

stable clusters in suspension and a SERS signal which is not affected by fluctuations of the

[DNA]/[NP] ratio within a large ds1 interval. It is important to note that potential minor fluctuations

of the SERS spectra that may arise from colloidal batch-to-batch dissimilarities (such as NP

concentration and size distribution) are “washed away” by the direct comparison of SERS spectra of

the “original” dsDNA and the “modified” one (both acquired by using the same colloidal batch).

9

A further example of the spectral reproducibility achieved under this experimental conditions is

illustrated in Fig. S3a, where 8 SERS spectra of 8 different repetitions of ds1 (0.63 μg/mL) onto

AgNP@Sp ([NP]~0.3 nM) are reported as acquired. The repetitions were performed as follows: 200

μL of AgNP@Sp from the same colloidal batch were placed in 8 different vials, then mixed with the

same volume of a ds1 buffered solution and finally investigated by SERS. As can be seen in Fig.

S3b, only minimal fluctuations of the background are observed whereas the spectral profile remains

unperturbed, showing a consistent sample-to-sample reproducibility.

Figure S3. Original no-baselined SERS spectra of 8 different repetitions of AgNP@Sp + ds1 (0.63 μg/mL) visualized as

(a) stacked and (b) overlapped. (c) Corresponding average SERS spectrum.

10

Section S3. Vibrational Assignment of the Main Bands of the dsDNA SERS Spectra

Vibrational assignment of ds1 sequence was based on both the data reported in the literature[10] and

by measuring the SERS spectra of 21-mer homopolymeric sequences of the four bases (pA, pC, pT

and pG) and 22 base self-complementary dsCG and dsAT double-stranded sequences (Fig. S5b).

The spectral interval 600-800 cm-1 contains Raman features associated with vibrations involving

concerted ring stretching motions (ring breathing) of purine or pyrimidine residues, often in

combination with stretching of the glycosidic bond and possibly also stretching of bonds within the

linked deoxyribose ring.[10e] Thus, these features are very informative about the helical geometry. In

this region, we identify bands mainly ascribable to the guanine residues at 502, 621, and 665/677 cm-

1, whereas strong bands at 730 and 787 cm-1 are mainly due to ring breathing vibrations of adenine

and the combination of cytosine + thymine residues, respectively. Analysis of the purine breathing

modes coupled to the deoxyribose in the SERS spectrum of ds1 reveals a large predominance of the

Z tertiary structure (bands at 621 and 665 cm-1) whereas the B-form (band at 686 cm-1) was the

dominant geometry in the SERS spectra of single-stranded oligonucleotides[11]. Interestingly, in

contrast to what observed for negatively charged colloids,[5, 12] (Fig. S6) the characteristic phosphate

band at 1090 cm-1, originating from the localized symmetric stretching vibration of the

phosphodioxy moiety, is rather intense in the SERS spectra of both single- and double-stranded

DNA structures on AgNP@Sp. This data clearly indicates that the phosphate groups are in close

proximity to the metal surface, providing a further evidence of the electrostatic-driven adsorption of

the DNA sequences onto the spermine-coated silver nanoparticles. The spectral region 1150–1600

cm-1 contains many overlapping bands, mainly arising from in-plane vibrations of base residues[10e].

Based on a spectral comparison with the SERS spectra of the experimental controls and on the

analysis of the vibrational studies reported in the literature, we tentatively assign the broad multiple

peak bands centered at 1246 cm-1 mainly to cytosine residues (with guanine contribution); the sharp

intense feature at 1325 cm-1 mainly to adenine residues (with guanine and thymine contributions);

the weaker features at 1354 and 1376 cm-1 to guanine and thymine residues, respectively; the 1421

cm-1 band to the DNA backbone vibrations; the base ring modes at 1487 cm-1 mainly to guanine

residues with some adenine contribution; the two weaker features at 1507 and 1528 cm-1, to adenine

and cytosine ring modes, respectively; and the intense band at 1577 cm-1 to ring modes mainly

ascribable to similar contributions of guanine and adenine residues. Finally, in the 1600-1800 cm-1

region overlap carbonyl group stretching vibrations of thymine (C2=O and C4=O), guanine (C6=O),

and cytosine (C2=O)[10e]. Thymine residues are known to generate intense Raman features around

1652 and 1672 cm-1, whereas the guanine mode at ca. 1720 cm-1 is only moderately intense.

11

Conversely, the cytosine C2=O mode near 1680 cm-1 is usually extremely weak[10e]. It is worth

noticing, however, that Raman signature associated with each DNA base sequence cannot be

approximated by a simple weighted sum of the Raman spectra of alternating AT/TA or GC/CG

repeats[10e] since base sequences play a key role in determining the final vibrational signature.

12

Figure S5. (a) Molecular structures of the 4 DNA nucleobases and the two methylated cytosine and adenine derivatives

(at the position C5 and N6, respectively). (b) Normalized SERS spectra of homopolymeric single stranded sequences

(pA, pG, pC and pT), and dsCG and dsAT, and ds1 duplexes. ssDNA concentrations was kept fixed at 5x10-7 M and

dsDNA at 5x10-8 M, which correspond to a final amount of analyte in the probed volume equals to 3.15 μg/mL (pA),

2.94 μg/mL (pC), 3.36 μg/mL (pG), 3.01 μg/mL (pT) and 630 ng/mL (dsCG, dsAT and ds1). Table (c) lists the main

Raman shifts and the tentative vibrational assignment of the main SERS bands of ds1 on AgNP@Sp colloids.

13

Section S4. SERS of single and double-stranded DNA sequences on negatively

charged silver colloids.

Based on the excellent results recently reported by Bell and co-workers[5, 12a] in the direct SERS

investigations of ssDNA in solution, we selected hydroxylamine-reduced silver nanoparticles

(AgHX) aggregated with MgSO4 as a reference colloidal substrate. As it can be seen from Fig. S6,

no significant band shifts can be observed when ss spectra are compared to ds spectra, whereas we

do observe a certain degree of sample-to-sample irreproducibility in terms of relative intensity in the

case of the duplex sample (see for instance the nucleobase ring breathing band in the 700-800 cm-1

spectral region and the bands above 1500 cm-1, highlighted in yellow). The lack of differentiation

between the SERS spectra of ss vs. ds (unmodified sequences) using negatively-charged colloids is

consistent with the previous findings reported by Marotta et al.[13] and Papadopoulou et al.[12a] Thus,

it may be speculated that the observed SERS signal for ds1 could be due either to the partial

denaturation of the dsDNA or to the small fraction of non-hybridised sequences in the ds1 solution

resulting from uncontrollable experimental errors in the preparation of equimolar mixtures of the two

complementary strands.

14

Figure S6. Representative SERS spectra of four different repetitions of hydroxylamine-reduced silver nanoparticles

(AgHX) aggregated with MgSO4 in the presence of a constant concentration of (a) ss1 and (b) ds1 (3.15 and 0.63 μg/mL

of analyte, respectively).

15

Section S5. Reshaping of the SERS spectrum of the single-stranded

complementary sequences upon hybridization.

The calculation of the pondered sum of the individual SERS spectra of ss1 and ssc relies on the

experimental evidence that the peak intensity of the symmetric phosphodioxy stretching vibration

band at 1090 cm-1 is largely independent to the DNA hybridisation [14], i.e. the peak height at 1090

cm-1 is approximately proportional to the number of phosphate groups in the sequence, either single

or double-stranded. Thus, ss1+ssc can be approximated as the combined spectral result of the

hypothetical equimolar mixture of non hybridised ss1 and ssc sequences.

Figure S7. SERS spectrum of ds1 (630 ng/mL) and digitally calculated SERS spectrum of ss1+ssc, both normalized to

the phosphate band at 1090 cm-1. The ss1ssc pondered spectrum was obtained by summing the two separated SERS

spectra of ss1 and ssc (3.15 ng/mL), both normalized to the 1090 cm-1 band.

16

Section S6. Single-base Mismatch

DNA damages or biosynthetic errors of DNA polymerases during the DNA replication can result in

heteroduplex sequences containing base-base mismatches. Deficiency in the cell mismatch repair

machinery leads to the accumulation of such mutations which, in turn, results in a high level of

microsatellite instability, a characteristic feature of a number of cancers, including gastric,

endometrial, ovarian, lung, and colorectal cancer.[15] Nowadays, the application of SERS

spectroscopy to the identification of single-nucleotide polymorphisms in unmodified sequences is

limited to single-stranded structures.[5]

Figure S8. SERS spectra of ds1, ds2, ds3 and ds4 (each 630 ng/mL).

17

Section S7. 5C and N6 methylation of cytosine and adenine.

In mammals, cytosine methylation at position C5 is the most abundant epigenetic modification and

mainly occurs at the CpG regions of DNA, where the rate of methylation usually lays in between the

70% and 80% of all cytosine bases[16]. The chemical instability of 5-methylcytosine (mC) renders the

CpG islands mutational hot spots as demonstrated by the evidence that a large number of human

genetic disease are found to be strictly associated with such transitions within CpG sequences[17]. If

the 5-methylcytosine is considered as the 5th DNA base, recent data point to the biological

importance of N6-methyladenine (mA) as the “other” methylated base[18]. mA is mainly found in the

genomes of bacteria, but accumulating evidence indicate its presence in some archaea and eukaryotic

cells, where its role remain largely unknown[18-19]. Indirect evidence seems to suggest that mA might

be present also in mammalian DNA[18]. In bacterial DNA, the function of mA has been associated

with genome defense, DNA replication and repair, nucleoid segregation, regulation of gene

expression, control of transposition, and host–pathogen interactions[18-19]. In particular, DNA-adenine

methyltransferases has been recognised as a key factor in determining bacterial viability and

virulence[19]. To date, the application of SERS to the identification of methylated bases in

unmodified sequences is limited to qualitative studies on single-stranded structures[3].

18

References

[1] D. van Lierop, Z. Krpetic, L. Guerrini, I. A. Larmour, J. A. Dougan, K. Faulds, D. Graham,

Chemical Communications 2012, 48, 8192-8194.

[2] N. Leopold, B. Lendl, J. Phys. Chem. B 2003, 107, 5723-5727.

[3] A. Barhoumi, N. J. Halas, J. Phys. Chem. Lett. 2011, 2, 3118-3123.

[4] a) J. Yguerabide, E. E. Yguerabide, Anal. Biochem. 1998, 262, 137-156; b) J. Yguerabide, E.

E. Yguerabide, Anal. Biochem. 1998, 262, 157-176.

[5] E. Papadopoulou, S. E. J. Bell, Angewandte Chemie (International ed. in English) 2011, 50,

9058-9061.

[6] R. Aroca, Surface-enhanced Vibrational Spectroscopy, John Wiley & Sons, Chichester,

2006.

[7] R. A. Alvarez-Puebla, J. Phys. Chem. Lett. 2012, 3, 857-866.

[8] N. J. Everall, Analyst 2010, 135, 2512-2522.

[9] E. C. Le Ru, P. G. Etchegoin, Principles of Surface-Enhanced Raman Spectroscopy, Elsevier,

Amsterdam, 2009.

[10] a) E. Taillandier, J. Liquier, M. Ghomi, J. Mol. Struct. 1989, 214, 185-211; b) J. M.

Benevides, S. A. Overman, G. J. Thomas, J. Raman Spectrosc. 2005, 36, 279-299; c) J. M.

Benevides, A. H. J. Wang, G. A. Van der Marel, J. H. Van Boom, G. J. Thomas,

Biochemistry 1989, 28, 304-310; d) L. Movileanu, J. M. Benevides, G. J. Thomas,

Biopolymers 2002, 63, 181-194; e) H. Deng, V. A. Bloomfield, J. M. Benevides, G. J.

Thomas, Biopolymers 1999, 50, 656-666; f) P. Y. Turpin, L. Chinsky, A. Laigle, B. Jolles, J.

Mol. Struct. 1989, 214, 43-70.

[11] A. Toyama, H. Takeuchi, I. Harada, J. Mol. Struct. 1991, 242, 87-98.

[12] a) E. Papadopoulou, S. E. J. Bell, Chem.-Eur. J. 2012, 18, 5394-5400; b) E. Papadopoulou, S.

E. J. Bell, Chemical communications 2011, 47, 10966-10968.

19

[13] N. E. Marotta, K. R. Beavers, L. A. Bottomley, Anal. Chem. 2013, 85, 1440-1446.

[14] J. G. Duguid, V. A. Bloomfield, J. M. Benevides, G. J. Thomas, Biophys. J. 1996, 71, 3350-

3360.

[15] a) L. Pikor, K. Thu, E. Vucic, W. Lam, Cancer Metastasis Rev. 2013, 32, 341-352; b) C.

Lengauer, K. W. Kinzler, B. Vogelstein, Nature 1997, 386, 623-627.

[16] M. Esteller, Lancet Oncol. 2003, 4, 351-358.

[17] a) G. P. Pfeifer, Curr.Top.Microbiol.Immunol. 2006, 301, 259-281; b) K. D. Robertson, Nat.

Rev. Genet. 2005, 6, 597-610.

[18] D. Ratel, J. L. Ravanat, F. Berger, D. Wion, Bioessays 2006, 28, 309-315.

[19] D. Wion, J. Casadesus, Nat. Rev. Microbiol. 2006, 4, 183-192.