Article

www.acsnano.org

Fast Optical Chemical and Structural

Classification of RNA
Judit Morla-Folch,†,‡ Hai-nan Xie,† Ramon A. Alvarez-Puebla,*,†,‡,§ and Luca Guerrini*,†
†
Medcom Advance, Viladecans Business Park, Edificio Brasil, Bertran i Musitu 83-85, 08840 Viladecans, Barcelona, Spain
‡
Universitat Rovira i Virgili and Centro Tecnológico de la Química de Catalunya, Carrer de Marcel·lí Domingo s/n, 43007 Tarragona,
Spain
§
ICREA, Passeig Lluís Companys 23, 08010 Barcelona, Spain
*
S Supporting Information

ABSTRACT: As more biological activities of ribonucleic

acids continue to emerge, the development of efficient
analytical tools for RNA identification and characterization
is necessary to acquire an in-depth understanding of their
functions and chemical properties. Herein, we demonstrate
the capacity of label-free direct surface-enhanced Raman
scattering (SERS) analysis to access highly specific
structural information on RNAs at the ultrasensitive level.
This includes the recognition of distinctive vibrational
features of RNAs organized into a variety of conformations
(micro-, fully complementary duplex-, small interfering- and
short hairpin-RNAs) or characterized by subtle chemical differences (single-base variances, nucleobase modifications and
backbone composition). This method represents a key advance in the ribonucleic acid analysis and will have a direct
impact in a wide range of different fields, including medical diagnosis, drug design, and biotechnology, by enabling the
rapid, high-throughput, simple, and low-cost identification and classification of structurally similar RNAs.
KEYWORDS: RNA, surface-enhanced Raman spectroscopy, nanoparticles, plasmonic, label-free

T he role of ribonucleic acid (RNA) in cellular processes

is one of the most dynamic and fast growing fields in
biology.1,2 New biological activities are continuously
emerging besides RNA key functions in the gene regulation,
protein synthesis or catalysis of biological reactions, all essential
therapeutic strategy, where the highly specific mechanisms of
sequence-specific gene silencing are exploited as efficient tools
to design novel drugs that selectively interfere with disease-
causing or disease-promoting genes.9,10 Additionally, aside from
the four canonical bases, many more chemically modified
for the advance of molecular diagnostics. Among others, small nucleotides exist in multiple types of RNA increasing their
RNAs are proved to play a major function in gene regulation, chemical diversity (i.e., epigenetic modifications) to fulfill all the
mainly in the form of microRNAs (miRNA), and double- different RNA functions within the biological systems.11
stranded RNAs such as small interfering RNAs (siRNA) and However, the lack of robust analytical methods has hindered,
short-hairpin RNAs (shRNA).3 miRNA are a class of non- for decades, the in-depth understanding of their distribution
protein-coding, small single-stranded (about 19−23 nucleo- and biochemical roles,12 and only very recent developments of
tides) ribonucleic acids that play critical regulatory roles in new genomic technologies allowed to progress in this
almost every biological processes.4,5 Mounting evidence direction.13 Thus, the development of fast, sensitive, and
associates aberrant expressions of miRNA with a number of efficient detection and characterization methods becomes more
diseases and genetic disorders as well as highlighting the and more imperative.
importance of miRNA as both oncogenes and tumor The analysis of small RNAs poses several challenges due to
suppressors.4,5 Thus, miRNA not only can be used as their intrinsic characteristics such as the small size of their
biomarkers for cancer diagnostics and monitoring, but also sequences, which greatly complicates the application of PCR;
exploited as new tools for cancer treatment via modulation of their low abundance; and their sequence similarity.4 Thus,
the miRNA pathways and activities4,6 and in gene therapy.3 On emerging technologies, such as the optical sensing through
the other hand, 21−23-mer siRNA and shRNA can silence gene
expression in the RNA interference (RNAi) process.7,8 Received: December 17, 2015
Growing understanding of genetic and the mechanisms of Accepted: February 1, 2016
RNAi pathways fueled the development of the RNAi-based Published: February 1, 2016

© 2016 American Chemical Society 2834 DOI: 10.1021/acsnano.5b07966

ACS Nano 2016, 10, 2834−2842
ACS Nano Article

Figure 1. Hybridization of single-stranded miRNAs sequences into fully complementary duplexes. Averaged SERS spectra of miRNAs ss1 and
ss2, and double-stranded RNA (dsRNA). The (ss1 + ss2) spectrum was digitally calculated by summing the two separated SERS spectra of ss1
and ss2, both previously normalized to the 1090 cm−1 band. Raman shifts and the tentative vibrational assignment of the main SERS bands of
single-stranded RNA on AgNP@Sp colloids are listed in the right side table.

plasmons, have the potential of reshaping the detection and technique to (i) recognize the distinctive vibrational features of
characterization scenario of nucleic acids14−16 by offering the structural characteristics of RNA including those of fully
simple, fast, and highly sensitive platforms that overcome the complementary double-stranded, small interfering and short
limitations of PCR and Northern blot based microarrays. hairpin RNA; and (ii) classify highly similar RNA sequences
Within the field of plasmonics, surface-enhanced Raman with subtle differences such as a single-base variance (including
scattering (SERS) spectroscopy has developed into a mature, that of uracil and thymine that defines RNA or DNA) or even a
reliable technique combining the rich molecular information nucleobase modification (i.e., N6-methyladenine and pseudour-
contained in the vibrational spectra with outstanding ultra- idine) in both single strands and duplexes.
sensitive, multiplexing, and quantification capabilities.17−21
However, until today, most of the application of SERS to RESULTS AND DISCUSSION
nucleic acid analysis largely relied on indirect strategies, mainly The SERS analysis of DNA can be efficiently and reproducibly
based on SERS encoded nanoparticles.15,22,23 Although this achieved in solution by using positively charged colloids.28−30
approach has been demonstrated very effective for pure This is mainly due to the intrinsic electrostatic affinity in
identification purposes, it completely dismisses the exquisite between the polyamine ligands adsorbed on the colloidal
chemical and structural information provided by the direct surfaces and the eminently negative charge of the DNA
acquisition of the vibrational fingerprint of the biomolecule. In phosphate backbones. Assuming a similar behavior for RNA, we
this regard, label-free direct SERS analysis of DNA has enjoyed then prepared silver spheroidal nanoparticles coated with
in the past few years a renewed interest,24−30 whereas a spermine (AgNP@Sp). These nanostructures are characterized
surprisingly scarce number of studies are reported for by a mean diameter of 25 nm, a ζ-potential of +43 mV, and a
RNA.31−35 Further, direct label-free SERS investigation of localized surface plasmon resonance (LSPR) centered at 392
RNA has been limited to short single strands (microRNA), nm (Figure S1). Addition of the genetic material into the
often in the form of simple homopolymeric or bipolymeric colloids promotes an immediate change in the color from the
sequences, and mostly performed by simple drying the sample typical greenish-yellow to red/orange, as also revealed by the
directly onto plasmonic substrates with the consequent lack of red-shift of the LSPR to ∼525 nm (Figure S1). RNA-driven
spectral reproducibility. nanoparticle aggregation yields stable colloidal clusters in
Herein, we demonstrate the potential of direct SERS analysis solution that entrap the target biomolecules at interparticle gaps
to screen composition and conformations of RNAs at the where highly efficient electromagnetic hot spots are generated.
ultrasensitive level. Specifically, we prove the capability of this SERS measurements were performed using a green laser (532
2835 DOI: 10.1021/acsnano.5b07966
ACS Nano 2016, 10, 2834−2842
ACS Nano Article

nm) to maximize the optical response of the RNA-mediated thermal treatment promotes homogeneous chain extension42
aggregates by illuminating the cluster suspension with a long which drastically improves the sample-to-sample reproducibility
working distance objective. Under such experimental con- (Figure S4).
ditions, a large ensemble of stable clusters in solution is In addition to fully complementary duplexes, double-
investigated during the acquisition time, providing highly stranded RNAs exist in several other variants, such as small
averaged SERS spectra with well-defined features, i.e., average interfering (siRNA) and short hairpin (shRNA). siRNA are
SERS regime (Figure S2). short double-stranded segments with two overhanging
Vibrational assignment of the main bands appearing in the nucleotides, whereas shRNA also include in their structure an
SERS spectrum of RNA (Figure 1) is required for accessing and additional loop.9 In our case, the investigated siRNA is formed
interpreting the rich structural information contained in it. The by the hybridization between ss1 and the partially comple-
reported vibrational assignment for single-stranded RNA was mentary ss5 strand (Figure 2A and Figure S6) resulting in a
based on normal Raman studies reported in the literature36,37 as duplex structure of 19 base pairs and two overhanging bases at
well as on the direct analysis of homopolymeric sequences both ends. The ss5 sequence has identical base composition as
(Figure S3). For an appropriate comparison, all SERS spectra ss2; thus, siRNA structurally differs from dsRNA only in the
were normalized to the phosphate stretching band at 1090 unpaired condition of the two extremities. On the other hand, a
cm−1, which is commonly selected as the internal standard in partially self-complementary single-stranded RNA of 43
the normal Raman and direct SERS investigations of nucleobases was selected to yield a shRNA structure containing
DNAs27−29,38 due to its limited sensitivity to structural changes a stem of 17 base pairs, a loop of 7 nucleotides, and a
and its constant 1:1 ratio with the overall nucleobase content. dinucleotide overhang at the 3′-end (Figure 2A and Figure S6).
Changes in RNA Conformation. To investigate the Except for the central loop, the base sequences of the duplex
impact of the Watson−Crick hydrogen bonding and base stem and the overhang region are identical to those contained
stacking formation on the vibrational profiles of single-stranded in the siRNA, while the base composition differs only by an
RNAs, the fully complementary miRNAs ss1 and ss2 were additional extra cytosine (Figure S6). Thus, it may be expected
hybridized into the resulting duplex, dsRNA, by thermal that the spectral reshaping of the SERS profile should be
annealing at 90 °C. The SERS spectra of these structures are dominated by the different structural conformation of the RNA
illustrated in Figure 1, together with the digital sum of the rather than its base composition. This is corroborated by the
individual single strands spectra (ss1 + ss2). As a first analysis of the SERS spectra of dsRNA and shRNA (Figure 2A).
approximation, ss1 + ss2 can be considered as the hypothetical At first glance, these spectra show little deviations. Never-
equimolar mixture of nonhybridized ss1 and ss2 sequences and theless, once subtracted, the corresponding SERS difference
used as a reference to properly disclose the spectral reshaping spectrum reveals a spectral profile which is remarkably similar
associated with the structural rearrangement into the duplex to (ss1 + ss2) − dsRNA throughout the whole spectral range
conformation. (Figure 2A) with few exceptions such as relatively less extended
Phosphate bands at 814 and 1090 cm−1, sensitive to the intensity changes of the composite broad CO stretching and
ribose-phosphate chain backbone structure and helix con- the ring breathing modes at 726 and 783 cm−1. In particular, in
formation,36 do not undergo significant changes both in the shRNA − dsRNA difference spectrum, the A band at 726
frequency position and relative intensity upon strand hybrid- cm−1 undergoes a minor spectral shift as compared to the 783
ization, suggesting that ss1, ss2, and dsRNA adopt a similar cm−1 ring breathing mode. By assuming that the extension of
conformation when electrostatically adsorbed onto the such shifts is correlated with the relative number of mismatched
plasmonic surface. On the contrary, a large set of spectral nucleobases, we can then ascribe this observation to the larger
differences clearly emerges for the nucleotide bands, which can percentage of unpaired C bases in shRNA (4 out of 10), with an
be summarized as follows. First, as a result of the base stacking additional contribution of an extra unpaired U base, as
interactions, the ring breathing modes of cytosine (792 cm−1) compared to A, which has only 1 unpaired base out of the
and adenine (730 cm−1) suffer a frequency-shift in dsRNA (−9 11 contained in the whole hairpin sequence. As previously
and −4 cm−1, respectively). This is also consistent with the described, the small interfering RNA is structurally analogous to
relative intensity increase of the U ring vibration at 631 cm−1. dsRNA, except for the unpaired dinucleotide ends (GA and
Second, the hydrogen bonding promotes an intensity decrease CU). Consistently, their SERS spectra only show subtle but still
and a ∼7 cm−1 red-shift of the carbonyl stretching (mainly U + detectable differences, well-highlighted in the corresponding
C) and the rise of a new shoulder (∼1690 cm−1) ascribed to difference spectrum siRNA − dsRNA (multiplied by a factor of
the CO stretching of paired guanine (G).37,39 Third, changes 4 in Figure 2A). In this case, the spectral markers of H-bonding
in the molecular conformation are also evident in the relative (such as the purine bands at 1244, 1487, and 1575 cm−1 and
intensity increase of both U and C ring stretching (∼1231 and the broad carbonyl stretching feature) do not exhibit noticeable
1244 cm−1), attributed to the RNA folding,40 and the alterations and the detectable spectral changes appear to be
characteristic prominent hyperchromicity associated with H- mainly due to the base unpairing, as indicated by the red-shift
bonding of A + G bands (1487 and 1575 cm−1).41 and intensity decrease of the C + U ring breathing mode, the
It is worth noticing that single-stranded RNAs were intensity increase at ∼1330 cm−1 (ascribed to the disordering of
thermally treated before the SERS analysis to enhance the the purine A−G bases43) and the weakening of the A feature at
spectral reproducibility (see Supporting Information, p S8). 1507 cm−1.
The lack of the methyl group in uracil (U) as compared with A clear visual representation of the different degrees of
thymine (T) decreases its hydrophobicity. As a result, although similarities between shRNA, siRNA and dsRNA is depicted in
U preferably pairs with adenine (A), it can also weakly associate Figure 2B using partial least-squares discriminate analysis (PLS-
with almost any other bases. Thus, one may expect that single- DA) to classify the SERS spectra of the different RNAs. As is
stranded RNAs in solutions would exist in a broad set of evident, the spectral profile of shRNA significantly diverges
different configurations for loosely interacting strands. Gentle from that of siRNA and dsRNA, which are on the contrary very
2836 DOI: 10.1021/acsnano.5b07966
ACS Nano 2016, 10, 2834−2842
ACS Nano Article

oligonucleotides with specular base content (6A, 5U, 3C and

7G for ss1; and 5A, 6U, 7C and 3G for ss2), is plainly reflected
in their SERS spectra. The decrement of G and A bands and
the intensity increase of the C and U features can be
straightforwardly visualized in Figure 1, when going from the
SERS spectrum of ss1 to that of ss2. On the other hand, more
intriguing is the possibility to differentiate single-base
mismatches in highly similar miRNA sequences. To test the
ability of our method to achieve such result, we acquired the
SERS spectra of highly similar miRNAs (ss2, ss3 and ss4) with
the identical base sequence except for one single-base
mismatch. Specifically, if compared to ss2, one C was replaced
with an A in the case of ss3, whereas for ss4, the extra A was
included to the detriment of one U (see base sequences in
Figure 3). Subtraction of the SERS spectrum of ss2 from those

Figure 3. Averaged SERS spectra of single-stranded miRNAs ss2,

ss3 and ss4. The spectra were baseline-corrected and normalized to
the ν(PO2−) band at ∼1090 cm−1. Difference spectra (ss3 − ss2)
Figure 2. (A) Averaged SERS spectra of fully complementary and (ss4 − ss2) were obtained by digital subtraction of the SERS
double-stranded RNA (dsRNA), small interfering RNA (siRNA) spectrum of ss2 from the SERS spectra of ss3 and ss4. For a better
and short hairpin RNA (shRNA). Difference spectra (ss1 + ss2) − comparison, the difference spectra were both multiplied by a factor
dsRNA, shRNA − dsRNA, and siRNA − dsRNA obtained by digital of 4. miRNA structures and sequences are outlined at the top of
subtraction of the SERS spectrum of dsRNA from the SERS spectra figure (red shapes correspond to Adenine, yellow to Uracil, green
of shRNA and siRNA; and the pondered SERS spectrum (ss1 + ss2) to Guanine, and blue to Cytosine).
illustrated in Figure 1. For a better comparison, the difference
spectra shRNA − dsRNA and siRNA − dsRNA were both multiplied
by a factor of 2 and 4, respectively. (B) Partial least-squares of ss3 and ss4 was performed to better show the spectral
discriminant analysis (PLS-DA) of SERS spectra from dsRNA,
shRNA and siRNA. Score plot of loading vectors (LV) 1, 2, and 3
changes associated with the nucleobase substitution. In both
with 95% confidence ellipses, in which each spectrum is difference spectra, intense positive bands associated with A
represented by a data point. The position of each point reflects motions (730, ∼1327, and 1507 cm−1) emerge, whereas no
characteristic features of the spectrum. RNA structures and apparent changes occur in the bands with major G
sequences are outlined at the top of panels A (red shapes contribution, such as the broad ring breathing band at ∼660
correspond to Adenine, yellow to Uracil, green to Guanine, and cm−1. On the other hand, a general intensity decrease of the C
blue to Cytosine). features, such as those at 599, 792, and 1029 cm−1, is only
observed in the case of ss3, whereas the difference spectrum of
similar even though they can still be differentiated with 100% ss4 shows a decrease of the U ring stretching at ∼1231 cm−1
sensitivity and specificity. and a much milder intensity decrease of the ring breathing band
Changes in RNA Composition. The diversity in at 792 cm−1. The intense band at 792 cm−1 is indeed the result
nucleobase composition of ss1 and ss2, complementary of different contributions mainly from ring breathing modes of
2837 DOI: 10.1021/acsnano.5b07966
ACS Nano 2016, 10, 2834−2842
ACS Nano Article

C (larger contribution40) and U,43 with an additional

overlapping with a weaker A feature (790 cm−1, Figure S3).
Consistently, when one C is replaced with one A in ss3, the 792
cm−1 band appears in the difference SERS spectrum as an
intense negative feature, whereas when one U is removed in
favor of A, as for ss4, the overall intensity is left largely
unperturbed (i.e., the positive A contribution and the negative
U contribution to the 792 cm−1 band approximately balance
themselves out). Finally, both difference spectra reveal a relative
intensity decrease of the CO stretching, mainly ascribed to U
+ C. Overall, these spectral changes are consistent with a C →
A single-base mismatch in ss3, and U → A single-base mismatch
in ss4. Direct SERS single-base mismatch detection in miRNAs
was limited so far to the simple differentiation via partial least-
squares discriminate analysis of miRNA members of the let-7
family achieved by investigating a relatively large amount of
samples (1 μg) dried onto silver nanorods arrays.31
In addition to the different content of the four canonical
nucleobases, structural differences within RNAs can arise from
chemical modifications, such as epigenetic alterations. Among
all post-transcriptional modifications that have been identified
in RNAs of all organisms, pseudouridine (ψ) is found to be the
most abundant one and it is recognized as the “fifth” nucleoside
in RNA.44,45 The second most common RNA modification is
nucleobase methylation.46 In particular, N6-methyladenine is
the most abundant internal mRNA modification in eukaryotes,
as well as in viruses.12,13 To demonstrate the reliability of our
direct method to recognize different types of nucleobase
modifications in the SERS spectra of both miRNA and dsRNA,
we acquired the characteristic vibrational changes imposed on
the spectral profile of RNA by replacing A and U with their
epigenetic variants N6-methyladenine (mA) and pseudouridine Figure 4. N6-methyladenine in miRNA and dsRNA: identification
(ψ), respectively. and quantification. (A) Averaged SERS spectra of ss2 and dsRNA,
N6-methylation of adenine (Figure 4) dictates characteristic and the analogous ssmA and dsmA samples where the 5 adenines
were replaced with the N6-methyladenine variant (thus, dsmA
spectral changes in the nucleotide spectral fingerprint47 which
contains 5 mA from ssmA and 6 A from ss1). Top: molecular
are plainly resolved in the corresponding SERS profile of both structures of adenosine (A) and N6-methyladenosine (mA). (B and
single strand and duplex. In Figure 4A, we represent the SERS C) Detail of the A ring breathing spectral region for the SERS
spectra of ss2 and dsRNA and their corresponding ssmA and spectra of ss2 + ssmA and dsRNA + dsmA mixtures at different molar
dsmA analogues, where the 5 adenine of the ss2 were fully ratios, R ([ssmA] vs total [ssmA] + [ss2] content, and [dsmA] vs total
replaced by the mA variant. It is important to note that the [dsmA]+[dsRNA] content, respectively). Molar ratios decrease
methyl group at the N6 position of adenosine does not prevent going from the darkest spectrum to clearest one as follows: R =
the Watson−Crick base-paring.12,48 The spectra changes 0, 0.1, 0.2, 0.3, 0.4, 0.5, and 0.6. (D) Ratiometric peak intensities
imposed by the methylation on both ss- and dsRNA are I743/I729 and I743/I725 for miRNA and dsRNA mixtures, respectively,
striking. First, we highlight the remarkable spectral red-shifts against the corresponding molar ratios. RNA structures and
sequences are outlined at the top of panel A (red shapes
(>10 cm−1) of the A ring breathing modes and their relative correspond to Adenine, yellow to Uracil, green to Guanine, and
intensity decrease. In the case of the duplex, this mode appears blue to Cytosine).
as a doublet at 726 cm−1, ascribed to the unmodified A bases in
the complementary ss1 sequence, and at 740 cm−1, assigned to
the methylated adenines of the ssmA strand. Furthermore, 3−4 U base pairs in the double helix leading to different
cm−1 red-shifts are observed in several other A related bands, perturbations of the U electronic structure depending on the
such as those at 1487 and 1575 cm−1, while the very intense A structural properties of the Watson−Crick counterpart (A vs
+ G feature at ∼1327 cm−1 undergoes a notable intensity drop m
A).
and the 1507 cm−1 adenine band almost disappears from the Pseudouridine (ψ) is derived from uridine via base-specific
spectra. These spectral changes nicely match those described in isomerization.44 This process introduces an additional hydro-
the normal Raman studies of the N6-methylation of the gen bond donor, N−H, in the nucleobase (Figure 5) which lays
individual adenine nucleoside.47 Interestingly, while the the foundation for the special properties of ψ relative to U such
alterations of the vibrational profile are largely limited to A as the enhanced rigidity conferred to the phosphodiester
related bands in the miRNA spectra, for the corresponding backbone and the structural stabilization of the neighboring
duplex, we also register additional differences such as the nucleosides, via increased base stacking, for both single- and
broadening and red-shift of the carbonyl stretching vibration double-stranded RNAs.44,45 As for ssmA, the analysis of the
(mainly U) from 1635 to 1648 cm−1, as well as the more subtle SERS spectra of pseudouridine-containing miRNA, ssψ, (Figure
2 cm−1 blue-shift of the C + U ring breathing mode. Once 5A) reveals spectral changes to the uracil bands that
more, these results are consistent with the formation of the A− qualitatively agree with normal Raman studies of the individual
2838 DOI: 10.1021/acsnano.5b07966
ACS Nano 2016, 10, 2834−2842
ACS Nano Article

investigate the possibility to quantitatively correlate the spectral

changes in the SERS profiles with the relative contents of
modified nucleobases in the biomolecule. To do so, we
prepared mixtures of modified and unmodified miRNAs (ss2 +
ssmA and ss2 + ssψ) as well as duplexes (dsRNA + dsmA and
dsRNA + dsψ) at different molar ratios, R (where R is equal to
the concentration of the modified sample versus the overall
RNA content). The corresponding SERS spectra are illustrated
in Figures S7−S10, whereas the characteristic spectral regions
containing the A ring breathing and CO stretching markers
are detailed in Figure 4B,C and Figure 5B,C, respectively. The
relative SERS intensities I743/I729 for ssmA, and I743/I725 for dsmA
are plotted against R in Figure 4D. Similarly, the values I1650/
I1625 for ssψ and I1655/I1630 for dsψ were monitored at different
molar ratios as illustrated in Figure 5D. The results show
outstanding linear correlations between the content of modified
bases and the extent of spectral perturbations introduced in the
SERS spectra, with sensitivity below the single-base discrim-
ination for N6-methyladenine in both ss- and dsRNA, and for
pseudouridine in single-stranded sequences.
The reliability of direct SERS to fully extract the rich
structural information contained in the Raman spectra of
nucleic acids was also exploited to differentiate and structurally
characterize RNA and DNA analogues. To do so, we
investigated single and double-stranded DNA and RNA with
identical sequence, except of course for the thymine bases
replaced by their uracil counterparts and the deoxyribose sugar
moieties substituted by the ribose. The direct comparison of
SERS spectra is illustrated in Figure 6 and unravels a set of
spectral differences that can be broadly classified into the
following classes: spectral changes associated with (i) different
Figure 5. Pseudouridine in miRNA and dsRNA: identification and backbone conformations, (ii) nucleobase modifications (i.e.,
quantification. (A) Averaged SERS spectra of ss2 and dsRNA, and removal of the methyl group from the T structure yielding the
the analogous ssψ and dsψ samples where the 6 uracils of ss3 were U base), and (iii) indirect perturbations on the remaining A, G,
replaced with the pseudouridine variant (thus, dsψ contains 6 ψ and C nucleobases imposed by (i) and (ii).
from ssψ and 5 U from ss1). Top: molecular structures of uridine Regarding the comparison of the unpaired strands, the
(U) and pseudouridine (ψ). (B and C) Detail of the carbonyl analysis of backbone conformation markers indicates that
spectral region for the SERS spectra of ssψ + ss2 and dsRNA + dsψ miRNAs exist mainly in the A-form secondary structure, as
mixtures at different molar ratios, R ([ssψ] vs total [ssψ] + [ss2]
content, and [dsψ] vs total [dsψ] + [dsRNA] content, respectively).
revealed by the intense spectral markers at ∼660 cm−1 (G band
Molar ratios decrease going from the darkest spectrum to clearest diagnostic of the C3′-endo/anti ribofuranose puckering) and
one as follows: R = 0, 0.1, 0.2, 0.3, 0.4, 0.5, and 0.6. (D) the presence of a shoulder at ∼814 cm−1, ascribed to the
Ratiometric peak intensities I1650/I1625 and I1655/I1630 for miRNAs sugar−phosphate backbone vibration.39,50 However, although
and dsRNA mixtures, respectively, against the corresponding molar the A-form helicity is found to be predominant, the coexistence
ratios. RNA structures and sequences are outlined at the top of of the G ring breathing band at ∼684 cm−1, characteristic of the
panel A (red shapes correspond to Adenine, yellow to Uracil, green C2′-endo/anti conformation of the B-type backbone, suggests
to Guanine, and blue to Cytosine). that some nucleotide residues can adopt additional conforma-
tions.39,50 Differently, single-stranded DNA (ssDNA) SERS
nucleosides.49 These include (i) the drastic red-shift of the spectra only display a narrow G band at 684 cm−1 with no
carbonyl stretching band; (ii) the intensity perturbation and 2 apparent phosphate contribution at 814 cm−1, which is
cm−1 shift of the C + U ring breathing mode; and (iii) the large consistent with a B-like conformation.36 It is also worth
reshaping of the spectral contour of the broad band in the noticing the different spectral contour of the weak bands in the
1220−1250 cm−1 range, containing U ring vibrations.49 In this 850−1050 cm−1 region, originating mainly from vibrations of
case, however, we do also register a general intensity increase of the sugars, which are also sensitive to the backbone
purine bands, mostly ascribed to A (e.g., 730, 1327, and 1575 conformation.36 On the other hand, the replacement of T
cm−1), even though with no significant spectral shifts. with U in the RNA structure is mainly revealed in the SERS
Noteworthy, this base-neighboring effect of pseudouridine is spectra by the relative intensity increase and dramatic blue-shift
far less explicit in the SERS spectrum of the corresponding of the broad CO stretching as well as for the 2 cm−1 shift of
duplex, dsψ, which however retains the main pseudouridine the C + U ring breathing mode (see Figure S11 for details of
spectral marker (i.e., the large red-shift of the carbonyl the 600−950 cm−1 spectral region). This agrees with what was
stretching band). observed by directly comparing the SERS spectra of
Once the A ring breathing band and the carbonyl stretching homopolymerics uracil RNA, pU, and thymine DNA, pT
are identified as the major spectral markers of N 6 - (Figure 6). Furthermore, the analysis of the homopolymers
methyladenine and pseudouridine variants, respectively, we suggests that the relative intensity increase of the feature at
2839 DOI: 10.1021/acsnano.5b07966
ACS Nano 2016, 10, 2834−2842
ACS Nano Article

are thermally treated (Figure S4) and uridine replaces

pseudouridine in single-stranded sequences (Figure 5). More-
over, as clearly visible in Figure 6, SERS spectra of dsRNA and
dsDNA only differ from spectral alterations associated with
nucleobase modification (T vs U) and backbone conformation
with very minor deviations with regard to the general purine vs
pyrimidine relative intensities. In contrast, for single-stranded
sequences, the difference is remarkable (see for instance the
peak height ratios between the A and C + U breathing modes,
Figure S12B). All experimental evidence converges to indicate
that dsRNA and dsDNA share similar adsorption conformations
on silver surfaces (i.e., the distribution of SERS enhancements
profited by the ensemble of nucleobases is similar for RNA/
DNA duplexes of identical base sequence). On the other hand,
when individual miRNA strands are not forced to pair into the
rigid double helix, they are likely free to adopt a different range
of possible conformations which can be restricted by an
appropriate thermal treatment or modified upon inclusion of a
nucleobase variant conferring, such as pseudouridine, an
enhanced rigidity to the phosphodiester backbone. Controlling
the nucleic acid conformation onto the metallic surface is a key
feature in determining the overall spectral reproducibility as
well as affecting the final spectral profile, and must be kept in
careful consideration for the direct SERS analysis of single-
stranded RNAs.

CONCLUSIONS
In summary, we demonstrated the potential of label-free direct
SERS analysis to fully disclose the unique vibrational
Figure 6. Vibrational comparison between analogous RNA and information on a variety of different small RNAs with high
DNA. Averaged SERS spectra of uracil RNA and thymine DNA sensitivity. As vibrational spectra and molecular features are
homopolymers (pU and pT, respectively); the two complementary intrinsically linked, it was possible to identify and classify highly
microRNAs ss1 and ss2, and the two complementary single- similar RNA structures based on both their conformation and
stranded ssDNA1 and ssDNA2. The averaged SERS spectra of the
corresponding dsRNA and dsDNA duplexes are also shown. RNA
composition. Among others, we reported the first example of
structures and sequences are outlined at the top of the figure (red SERS recognition of fully complementary duplexes, as well as
shapes correspond to Adenine, yellow to Uracil, green to Guanine, short hairpin and small interfering RNAs. We diversified
and blue to Cytosine). microRNAs sequences with single-base mismatches. Nucleo-
base variants such as pseudouridine and N6-methyladenine
were recognized and quantified with single-base sensitivity
∼1029 cm−1 observed for miRNA may also be due to the U within both single strands and duplexes. Finally, direct
presence in the structure. comparison of structurally analogous RNA and DNA molecules
Similar considerations in terms of helicity form and U/T also permitted to characterize distinctive signatures of their
nucleobase modification can be derived when comparing the corresponding sugar moieties as well as their secondary
SERS spectra of RNA and DNA duplexes (Figure 6). structure. This work represents a step forward in the
Additionally, we also observe a significant 4 cm−1 down-shift ribonucleic acid analysis by SERS, paving the way for the
of the A ring breathing mode in dsRNA which does not occur final translation of its analytical potential to the in-depth
for miRNA (better visualized in Figure S11). This experimental investigation of these intriguing biomolecules. Owing to the
result is congruent with the different electronic perturbation extremely rich structural information provided by this simple,
that U and T impose on the A structure upon duplex formation highly sensitive and low-cost sensing approach, we believe that
via base stacking and H-bonding, as also reported in the normal this method will have an important impact in basic and
Raman studies of bipolymeric A−U and A−T nucleic acids.51 therapeutic research, drug design, and diagnosis applications.
Interestingly, in the case of duplexes, the remaining features of
the spectra show minor or null differences, whereas for miRNA METHODS
and ssDNA, some remarkable changes are observed in several All materials were of highest purity available and obtained from Sigma-
bands which are not strictly related to either the backbone Aldrich. RNA sequences were purchased from Eurofins Genomics
conformation or T/U nucleobases. Specifically, these spectral (Germany) except for shRNA, purchased from IDT Integrated DNA
differences can be mainly described as the variation of the Technologies (Denmark). The 200 μM stock solutions were prepared
intensity ratio between purine and pyrimidine bands, such as by dissolving into Milli-Q water. The microRNAs and single-stranded
DNAs selected for this study are listed in Table S1. Duplex structures
the A ring breathing band and the intense A+G ring band were prepared in 0.3 M PBS by the annealing at 90 °C for 10 min of
centered at 1327 cm−1 (purine), and the C+U ring breathing equimolar solutions of ss1 and ss2; ss1 and ss5; ssDNA1 and ssDNA2
feature (pyrimidine). yielding the corresponding dsRNA, siRNA, and dsDNA. Short hairpin
Overall, the data showed that the relative intensity of these RNA structure (shRNA) was also self-hybridized in 0.3 M PBS by
purine bands decreases in a very similar fashion when miRNAs annealing at 90 °C for 10 min. The final concentration of the duplex

2840 DOI: 10.1021/acsnano.5b07966

ACS Nano 2016, 10, 2834−2842
ACS Nano Article

structures in the stock solutions was 20 μM. The solutions were stored ACKNOWLEDGMENTS
at −20 °C until required.
Positively charged spermine-coated silver nanoparticles (AgNP@ The work was funded by European Research Council
Sp) were synthesized following the previously reported protocol.28,29 (PrioSERS FP7/2014 623527), Ministerio de Economia y
Nanoparticle concentration (∼0.3 nM) was calculated by Lambert− Competitividad (CTQ2014-59808R), Generalitat de Catalunya
Beer’s law using the extinction coefficient for silver nanoparticles of (2014-SGR-480), and Medcom Advance SA.
1.85 × 1010 M−1 cm−1, derived from literature.52 The UV−vis
spectrum of AgNP@Sp colloids is shown in Figure S1. AgNP@Sp
colloids are characterized by a low/null SERS background signal ABBREVIATIONS
(Figure S1C). Samples for SERS measurements were prepared as mi, micro; ss, single-stranded; ds, double-stranded; sh, short
follows: 100 μL of AgNP@Sp ([NP] ∼ 0.3 nM) was mixed with 4 μL hairpin; m, messenger; Sp, spermine; AgNP, silver nano-
of RNA or DNA aqueous solutions (10 μM for single-stranded particles; AgNP@Sp, positively charged spermine-coated silver
sequences and 5 μM for duplexes, respectively). The colloidal
suspension immediately turned its color from yellow to orange, nanoparticles; LSPR, localized surface plasmon resonance; A,
revealing the nanoparticle aggregation. The samples were left to adenine; C, cytosine; G, guanine; T, thymine; U, uracil; mA, N6-
equilibrate for 3 h and redispersed by quick sonication before methyladenine; ψ, pseudouridine; pA, poly adenine; pC, poly
acquiring the SERS spectra. It is worthy of note that the sensitivity of cytosine; pG, poly guanine; pU, poly uracil; pT, poly thymine;
the method can be easily improved down to ∼30 nM of miRNA and rib, ribose; let-7, lethal-7 (the let-7 family is one of the key
∼10 nM of dsRNA (corresponding respectively to ∼21 and ∼14 ng miRNA regulators in development and cancer)
per 100 μL of investigated colloidal sample) by diluting the
nanoparticle suspension 4 times. Further, the overall amount of
sample required to perform the SERS study can be drastically reduced, REFERENCES
below the picogram level of RNA, via straightforward implementation (1) Sharp, P. A. The Centrality of RNA. Cell 2009, 136, 577−580.
of the SERS method with microfluidics devices, as described in our (2) Dumelin, C. E.; Chen, Y. Y.; Leconte, A. M.; Chen, Y. G.; Liu, D.
previous study.30 R. Discovery and Biological Characterization of Geranylated RNA in
SERS experiments were conducted using a Renishaw InVia Reflex Bacteria. Nat. Chem. Biol. 2012, 8, 913−919.
confocal microscope and a 532 nm laser. A long-working distance (3) He, L.; Hannon, G. J. MicroRNAs: Small RNAs with a Big Role
objective (0.17 NA, working distance 30 mm) was used to focus the in Gene Regulation. Nat. Rev. Genet. 2004, 5, 522−531.
laser on the sample (laser power at the sample of 6.9 mW). All SERS (4) Dong, H. F.; Lei, J. P.; Ding, L.; Wen, Y. Q.; Ju, H. X.; Zhang, X.
spectra illustrated in the article were obtained by averaging the SERS J. MicroRNA: Function, Detection, and Bioanalysis. Chem. Rev. 2013,
responses of 5 different replications per each sample (15 113, 6207−6233.
accumulations, 10 s exposure time). These replications were obtained (5) Bartel, D. P. MicroRNAs: Genomics, Biogenesis, Mechanism, and
by combining the same colloids (prepared the day before the Function. Cell 2004, 116, 281−297.
measurement and left aging overnight) with the specific RNA/DNA (6) Garzon, R.; Calin, G. A.; Croce, C. M. MicroRNAs in Cancer.
solution. UV−vis spectra were recorded using a Thermo Scientific Annu. Rev. Med. 2009, 60, 167−179.
Evolution 201 UV−visible spectrophotometer. TEM was performed (7) Rao, D. D.; Vorhies, J. S.; Senzer, N.; Nemunaitis, J. siRNA vs.
with a JEOL JEM-1011 transmission electron microscope. The ζ shRNA: Similarities and Differences. Adv. Drug Delivery Rev. 2009, 61,
(zeta) potential measurements were carried out using a Malvern Nano
746−759.
Zetasizer.
(8) Fire, A. Z. Gene Silencing by Double-Stranded RNA. Cell Death
Partial Least-Squares Discriminant Analysis (PLS-DA): PLS-DA is a
Differ. 2007, 14, 1998−2012.
linear classification method for highlighting differences between
(9) Kim, D. H.; Rossi, J. J. Strategies for Silencing Human Disease
multivariate data sets of different classes. It encompasses the partial
least-squares regression properties to fine-tune a principal component Using RNA Interference. Nat. Rev. Genet. 2007, 8, 173−184.
analysis model. Acquired SERS spectra were imported to the PLS tool (10) De Fougerolles, A.; Vornlocher, H.-P.; Maraganore, J.;
box (Eigenvector Research, Inc.) in Matlab for PLS-DA modeling. The Lieberman, J. Interfering with Disease: A Progress Report on
spectra were normalized using Multiplicative Scatter Correction53 and siRNA-Based Therapeutics. Nat. Rev. Drug Discovery 2007, 6, 443−
mean centered before being imported for the construction of the 453.
model. (11) Cantara, W. A.; Crain, P. F.; Rozenski, J.; McCloskey, J. A.;
Harris, K. A.; Zhang, X.; Vendeix, F. A. P.; Fabris, D.; Agris, P. F. The
RNA Modification Database, RNAMDB: 2011 Update. Nucleic Acids
ASSOCIATED CONTENT Res. 2011, 39, D195−D201.
*
S Supporting Information (12) Dominissini, D.; Moshitch-Moshkovitz, S.; Schwartz, S.;
Salmon-Divon, M.; Ungar, L.; Osenberg, S.; Cesarkas, K.; Jacob-
The Supporting Information is available free of charge on the Hirsch, J.; Amariglio, N.; Kupiec, M.; Sorek, R.; Rechavi, G. Topology
ACS Publications website at DOI: 10.1021/acsnano.5b07966. of the Human and Mouse M6A RNAMethylomes Revealed by M6A-
Optical and TEM characterization of AgNP@Sp colloids; Seq. Nature 2012, 485, 201−206.
sample-to-sample spectral reproducibility; RNA and (13) Lee, M.; Kim, B.; Kim, V. N. Emerging Roles of RNA
DNA base sequences; thermal treatment of single- Modification: M(6)A and U-Tail. Cell 2014, 158, 980−987.
(14) Lu, C. H.; Willner, B.; Willner, I. DNA Nanotechnology: From
stranded RNAs; additional SERS characterization Sensing and DNA Machines to Drug-Delivery Systems. ACS Nano
(PDF) 2013, 7, 8320−8332.
(15) Peng, H. I.; Miller, B. L. Recent Advancements in Optical DNA
Biosensors: Exploiting the Plasmonic Effects of Metal Nanoparticles.
AUTHOR INFORMATION
Analyst 2011, 136, 436−447.
Corresponding Authors (16) Li, D.; Song, S. P.; Fan, C. H. Target-Responsive Structural
*E-mail: luca.guerrini@ctqc.org. Switching for Nucleic Acid-Based Sensors. Acc. Chem. Res. 2010, 43,
*E-mail: ramon.alvarez@urv.cat. 631−641.
(17) Schlücker, S. Surface-Enhanced Raman Spectroscopy: Concepts
Notes and Chemical Applications. Angew. Chem., Int. Ed. 2014, 53, 4756−
The authors declare no competing financial interest. 4795.

2841 DOI: 10.1021/acsnano.5b07966

ACS Nano 2016, 10, 2834−2842
ACS Nano Article

(18) Rodriguez-Lorenzo, L.; Fabris, L.; Alvarez-Puebla, R. A. (36) Hobro, A. J.; Standley, D. M.; Ahmad, S.; Smith, N. I.
Multiplex Optical Sensing with Surface-Enhanced Raman Scattering: Deconstructing RNA: Optical Measurement of Composition and
A Critical Review. Anal. Chim. Acta 2012, 745, 10−23. Structure. Phys. Chem. Chem. Phys. 2013, 15, 13199−13208.
(19) Guarrotxena, N.; Bazan, G. C. Antitags: SERS-Encoded (37) Benevides, J. M.; Tsuboi, M.; Bamford, J. K. H.; Thomas, G. J.
Nanoparticle Assemblies That Enable Single-Spot Multiplex Protein Polarized Raman Spectroscopy of Double-Stranded RNA from
Detection. Adv. Mater. 2014, 26, 1941−1946. Bacteriophage Phi6: Local Raman Tensors of Base and Backbone
(20) Kang, B.; Austin, L. A.; El-Sayed, M. A. Observing Real-Time Vibrations. Biophys. J. 1997, 72, 2748−2762.
Molecular Event Dynamics of Apoptosis in Living Cancer Cells Using (38) Duguid, J. G.; Bloomfield, V. A.; Benevides, J. M.; Thomas, G. J.
Nuclear-Targeted Plasmonically Enhanced Raman Nanoprobes. ACS DNA Melting Investigated by Differential Scanning Calorimetry and
Nano 2014, 8, 4883−4892. Raman Spectroscopy. Biophys. J. 1996, 71, 3350−3360.
(21) Phan-Quang, G. C.; Lee, H. K.; Phang, I. Y.; Ling, X. Y. (39) Carmona, P.; Molina, M.; Rodriguez-Casado, A. Raman Spectra
Plasmonic Colloidosomes as Three-Dimensional SERS Platforms with and Structure of a 25mer HCV RNA. J. Raman Spectrosc. 2009, 40,
893−897.
Enhanced Surface Area for Multiphase Sub-Microliter Toxin Sensing.
(40) Reipa, V.; Niaura, G.; Atha, D. H. Conformational Analysis of
Angew. Chem. 2015, 127, 9827−9831.
the Telomerase RNA Pseudoknot Hairpin by Raman Spectroscopy.
(22) Dougan, J. A.; Faulds, K. Surface Enhanced Raman Scattering
RNA 2006, 13, 108−115.
for Multiplexed Detection. Analyst 2012, 137, 545−554. (41) Fujimoto, N.; Toyama, A.; Takeuchi, H. Effects of Hydrogen
(23) Fabris, L.; Dante, M.; Braun, G.; Lee, S. J.; Reich, N. O.; Bonding on the Uv Resonance Raman Bands of the Adenine Ring and
Moskovits, M.; Nguyen, T. Q.; Bazan, G. C. A Heterogeneous PNA- Its C8-Deuterated Analog. J. Mol. Struct. 1998, 447, 61−69.
Based SERS Method for DNA Detection. J. Am. Chem. Soc. 2007, 129, (42) Barhoumi, A.; Zhang, D.; Tam, F.; Halas, N. J. Surface-
6086−6087. Enhanced Raman Spectroscopy of DNA. J. Am. Chem. Soc. 2008, 130,
(24) Panikkanvalappil, S. R.; Mackey, M. A.; El-Sayed, M. A. Probing 5523−5529.
the Unique Dehydration-Induced Structural Modifications in Cancer (43) Thomas, G. J., Jr; Chen, M. C.; Hartman, K. A. Raman Studies
Cell DNA Using Surface-Enhanced Raman Spectroscopy. J. Am. Chem. of Nucleic Acids X. Conformational Structures of Escherichia Coli
Soc. 2013, 135, 4815−4821. Transfer RNAs in Aqueous Solution. Biochim. Biophys. Acta, Nucleic
(25) Panikkanvalappil, S. R.; Mahmoud, M. A.; Mackey, M. A.; El- Acids Protein Synth. 1973, 324, 37−49.
Sayed, M. A. Surface-Enhanced Raman Spectroscopy for Real-Time (44) Ge, J. H.; Yu, Y. T. RNA Pseudouridylation: New Insights into
Monitoring of Reactive Oxygen Species-Induced DNA Damage and Its an Old Modification. Trends Biochem. Sci. 2013, 38, 210−218.
Prevention by Platinum Nanoparticles. ACS Nano 2013, 7, 7524− (45) Charette, M.; Gray, M. W. Pseudouridine in RNA: What,
7533. Where, How, and Why. IUBMB Life 2000, 49, 341−351.
(26) Papadopoulou, E.; Bell, S. E. J. Label-Free Detection of Single- (46) Liu, J.; Jia, G. F. Methylation Modifications in Eukaryotic
Base Mismatches in DNA by Surface-Enhanced Raman Spectroscopy. Messenger RNA. J. Genet. Genomics 2014, 41, 21−33.
Angew. Chem., Int. Ed. 2011, 50, 9058−9061. (47) Mansy, S.; Peticolas, W. L.; Tobias, R. S. Raman Spectra of
(27) Xu, L.-J.; Lei, Z.-C.; Li, J.; Zong, C.; Yang, C. J.; Ren, B. Label- Methyl Derivatives of 5′-Adenosine Monophosphate, Tubercidin,
Free Surface-Enhanced Raman Spectroscopy Detection of DNA with Inosine, Uridine and Cytidine. Perturbation of Nucleoside Vibrations
Single-Base Sensitivity. J. Am. Chem. Soc. 2015, 137, 5149−5154. by Electrophilic Attack at Different Sites. Spectrochim. Acta A Mol.
(28) Guerrini, L.; Krpetić, Ž .; van Lierop, D.; Alvarez-Puebla, R. A.; Biomol. Spectrosc. 1979, 35, 315−329.
Graham, D. Direct Surface-Enhanced Raman Scattering Analysis of (48) Niu, Y.; Zhao, X.; Wu, Y.-S.; Li, M.-M.; Wang, X.-J.; Yang, Y.-G.
DNA Duplexes. Angew. Chem., Int. Ed. 2015, 54, 1144−1148. N6-Methyl-Adenosine (M6A) in RNA: An Old Modification with a
(29) Masetti, M.; Xie, H.-n.; Krpetić, Ž .; Recanatini, M.; Alvarez- Novel Epigenetic Function. Genomics, Proteomics Bioinf. 2013, 11, 8−
17.
Puebla, R. A.; Guerrini, L. Revealing DNA Interactions with
(49) Ueda, T.; Shinozaki, K.; Ushizawa, K.; Tsuboi, M. Raman
Exogenous Agents by Surface-Enhanced Raman Scattering. J. Am.
Scattering Tensors of Pseudouridine. J. Raman Spectrosc. 1997, 28,
Chem. Soc. 2015, 137, 469−476. 947−952.
(30) Morla-Folch, J.; Xie, H.-n.; Gisbert-Quilis, P.; Gómez-de Pedro, (50) Benevides, J. M.; Overman, S. A.; Thomas, G. J. Raman,
S.; Pazos-Perez, N.; Alvarez-Puebla, R. A.; Guerrini, L. Ultrasensitive Polarized Raman and Ultraviolet Resonance Raman Spectroscopy of
Direct Quantification of Nucleobase Modifications in DNA by Nucleic Acids and Their Complexes. J. Raman Spectrosc. 2005, 36,
Surface-Enhanced Raman Scattering: The Case of Cytosine. Angew. 279−299.
Chem., Int. Ed. 2015, 54, 13650−13654. (51) Grygon, C. A.; Spiro, T. G. Uv-Resonance Raman Spectroscopy
(31) Driskell, J. D.; Seto, A. G.; Jones, L. P.; Jokela, S.; Dluhy, R. A.; of Nucleic Acid Duplexes Containing A-U and A-T Base Pairs.
Zhao, Y. P.; Tripp, R. A. Rapid MicroRNA (miRNA) Detection and Biopolymers 1990, 29, 707−715.
Classification Via Surface-Enhanced Raman Spectroscopy (SERS). (52) Yguerabide, J.; Yguerabide, E. E. Light-Scattering Submicro-
Biosens. Bioelectron. 2008, 24, 917−922. scopic Particles as Highly Fluorescent Analogs and Their Use as
(32) Muniz-Miranda, M.; Gellini, C.; Pagliai, M.; Innocenti, M.; Salvi, Tracer Labels in Clinical and Biological Applications - I. Theory. Anal.
P. R.; Schettino, V. SERS and Computational Studies on MicroRNA Biochem. 1998, 262, 137−156.
Chains Adsorbed on Silver Surfaces. J. Phys. Chem. C 2010, 114, (53) Geladi, P.; Macdougall, D.; Martens, H. Linearization and
13730−13735. Scatter-Correction for near-Infrared Reflectance Spectra of Meat. Appl.
(33) Prado, E.; Daugey, N.; Plumet, S.; Servant, L.; Lecomte, S. Spectrosc. 1985, 39, 491−500.
Quantitative Label-Free RNA Detection Using Surface-Enhanced
Raman Spectroscopy. Chem. Commun. 2011, 47, 7425−7427.
(34) Abell, J. L.; Garren, J. M.; Driskell, J. D.; Tripp, R. A.; Zhao, Y.
P. Label-Free Detection of Micro-RNA Hybridization Using Surface-
Enhanced Raman Spectroscopy and Least-Squares Analysis. J. Am.
Chem. Soc. 2012, 134, 12889−12892.
(35) Chen, Y. P.; Chen, G.; Zheng, X. W.; He, C.; Feng, S. Y.; Chen,
Y.; Lin, X. Q.; Chen, R.; Zeng, H. S. Discrimination of Gastric Cancer
from Normal by Serum RNA Based on Surface-Enhanced Raman
Spectroscopy (SERS) and Multivariate Analysis. Med. Phys. 2012, 39,
5664−5668.

2842 DOI: 10.1021/acsnano.5b07966

ACS Nano 2016, 10, 2834−2842