ArticlePDF Available

Development and Automation of a Bacterial Biosensor to the Targeting of the Pollutants Toxic Effects by Portable Raman Spectrometer


Abstract and Figures

Water quality monitoring requires a rapid and sensitive method that can detect multiple hazardous pollutants at trace levels. This study aims to develop a new generation of biosensors using a low-cost fiber-optic Raman device. An automatic measurement system was thus conceived, built and successfully tested with toxic substances of three different types: antibiotics, heavy metals and herbicides. Raman spectroscopy provides a multiparametric view of metabolic responses of biological organisms to these toxic agents through their spectral fingerprints. Spectral analysis identified the most susceptible macromolecules in an E. coli model strain, providing a way to determine specific toxic effects in microorganisms. The automation of Raman analysis reduces the number of spectra required per sample and the measurement time: for four samples, time was cut from 3 h to 35 min by using a multi-well sample holder without intervention from an operator. The correct classifications were, respectively, 99%, 82% and 93% for the different concentrations of norfloxacin, while the results were 85%, 93% and 81% for copper and 92%, 90% and 96% for 3,5-dichlorophenol at the three tested concentrations. The work initiated here advances the technology needed to use Raman spectroscopy coupled with bioassays so that together, they can advance field toxicological testing.
Content may be subject to copyright.
Citation: Bandeliuk, O.; Assaf, A.;
Bittel, M.; Durand, M.-J.; Thouand, G.
Development and Automation of a
Bacterial Biosensor to the Targeting of
the Pollutants Toxic Effects by
Portable Raman Spectrometer.
Sensors 2022,22, 4352. https://
Academic Editors: Simone Morais,
Álvaro Miguel Carneiro Torrinha and
Iria Bravo
Received: 2 May 2022
Accepted: 26 May 2022
Published: 8 June 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
Development and Automation of a Bacterial Biosensor
to the Targeting of the Pollutants Toxic Effects by Portable
Raman Spectrometer
Oleksandra Bandeliuk 1,2, Ali Assaf 1, Marine Bittel 2, Marie-Jose Durand 1and Gérald Thouand 1, *
1Nantes Université, ONIRIS, CNRS, GEPEA, UMR 6144, 85000 La Roche-sur-Yon, France; (O.B.); (A.A.); (M.-J.D.)
Tronico-Tame-Water, 26 Rue du Bocage, 85660 Saint-Philbert-de-Bouaine, France;
Water quality monitoring requires a rapid and sensitive method that can detect multiple
hazardous pollutants at trace levels. This study aims to develop a new generation of biosensors using
a low-cost fiber-optic Raman device. An automatic measurement system was thus conceived, built
and successfully tested with toxic substances of three different types: antibiotics, heavy metals and
herbicides. Raman spectroscopy provides a multiparametric view of metabolic responses of biological
organisms to these toxic agents through their spectral fingerprints. Spectral analysis identified the
most susceptible macromolecules in an E. coli model strain, providing a way to determine specific
toxic effects in microorganisms. The automation of Raman analysis reduces the number of spectra
required per sample and the measurement time: for four samples, time was cut from 3 h to 35 min by
using a multi-well sample holder without intervention from an operator. The correct classifications
were, respectively, 99%, 82% and 93% for the different concentrations of norfloxacin, while the results
were 85%, 93% and 81% for copper and 92%, 90% and 96% for 3,5-dichlorophenol at the three tested
concentrations. The work initiated here advances the technology needed to use Raman spectroscopy
coupled with bioassays so that together, they can advance field toxicological testing.
Keywords: biosensor; Raman spectroscopy; toxicity; microorganism; pollutant
1. Introduction
Access to safe drinking water and the preservation of existing water resources are
highly dependent on the ability to monitor water pollutants and ensure their detection in
situ. To meet environmental regulatory objectives, the multitude and diversity of substances
to be monitored, often at trace levels, represents a major constraint. This has led to the
development of a large number of specific and sensitive analytical methods, a hundred
of which are European standards [
]. In addition, complementary methods to assess
bioavailability and toxicity are critical for determining the true environmental impact of
contaminants [
]. Unlike when using physicochemical methods, it is not necessary to
presuppose the nature of the pollutants present in the samples. If the sample is toxic or
exerts any effect, a bioassay will respond even if the substance responsible was previously
unknown. The design of techniques suitable for toxicity assessment of wastewater along
with the implementation of basic ecotoxicity tests dates back to the 1940s [
]. These tests
were succeeded by several microbial bioluminescent bioassays and biosensors [
]. A
test of this kind forms a central part of the technique that was developed in the present
study. For a better assessment of environmental toxicity, several research teams have
proposed combining the responses of several microorganisms with single parameters such
as bioluminescence, respiration, etc. [
]. Thus, covering all metabolic events that may
be caused by a toxic phenomenon requires the involvement of as many bioelements and,
Sensors 2022,22, 4352.
Sensors 2022,22, 4352 2 of 17
where appropriate, the multiplication of complex genetic manipulations [
]. This was the
case when using a fluorescent library of 1870 transcriptomic reporters from Escherichia coli
K12-MG1655 to report the ecotoxic status of environmental samples [
]. However, their
routine use or integration into a biosensor remains very difficult for field applications.
To overcome these difficulties, the number of examined bioelements can be multiplied.
In addition, the number of parameters observed in each of them can be increased, and in
this case, there is a move toward a multiparametric approach rather than single parameter
measurement. Raman spectroscopy provides a multiparametric tool to help understand the
physiological changes induced by toxic substances [
]. This technology has been widely
exploited as an alternative optical method for fast microbial detection and identification
over the past two decades [
]. Raman spectra of microorganisms are typically complex,
with bands (i.e., peaks) deriving from the thousands (or more) of molecules that make up a
sample, each with its own unique Raman signature. In other words, Raman spectra can be
understood as molecular fingerprints of the samples under study.
In this context, recent advances in Raman spectroscopy offer new research opportuni-
ties by providing a non-destructive approach for monitoring metabolic responses to toxic
substances [
]. Work by Bittel et al. [
] on a set of four microbial strains (two bacteria, one
yeast and one microalga) and four pollutant models (antibiotics, heavy metals, herbicides
and phenol compounds), demonstrated that it is possible to determine the place in the cell
where the pollutant acts (DNA, proteins, etc.). A statistical analysis strategy was developed
on the basis of independent component analysis (ICA). Advantages of the multiparametric
aspect of Raman spectroscopy measurements include the ability to split up the overall
signal according to the contribution of different cell components and identify the Raman
bands, which are decisive for toxic effect detection with statistical calculations. This enables
an operator to identify physiological changes and to characterize the bacterial responses
to toxic exposure. The step where ’good’ spectra are selected, designed in a previous
study [
], permitted a significant improvement in spectral classification according to
toxicant concentration.
Current Raman equipment, although well developed for these purposes in the lab-
oratory, is not suitable for use in the harsher conditions of the field. Indeed, through the
years, although new Raman spectroscopy techniques have been developed, most of them
rely on large microscope systems. These include using metal-coated glass slides [
automated imaging microscopy [
], laser tweezers Raman spectroscopy (LTRS) [
LTRS performed on microorganisms directly in the aqueous suspension [
], using a
surface-enhanced Raman scattering (SERS) effect [3941] or filtration [4246].
All of the spectroscopic techniques need to be optimized in further development, to
create an automatic approach of measurement and simplifying the cell preparation step.
This will generate large amounts of data (continuous, simultaneous measurements of many
different samples). The present study proposes, firstly, a new cell biosensor, considering
both the way the cells should be prepared and the tested environment, so as to avoid any
physiological shifts. Secondly, the present study proposes a dedicated automatic Raman
platform for toxicity measurements on living microorganisms to simplify the procedure
and reduce analysis time.
2. Materials and Methods
2.1. Bacterial Culture
The bacterial strain Escherichia coli K12-MG1655 (ATCC 700926; genotype, F-lambda-
ilvG-rfb-50 rph-1) was used as a model organism. Bacteria were cultivated according to
Bittel et al. [
]. Luria-Bertani (LB) medium was used, which was prepared as follows:
1 L of distilled water was supplemented with 10 g tryptone (Biokar Diagnostics, Allonne,
France, ref A1401HA), 5 g yeast extract (Biokar Diagnostics, ref A1202HA) and 5 g NaCl
(Carlo Erba Reagents, Milan, Italy, ref 479687). Sterilization was performed by autoclaving
at 120
C for 20 min. Starting from
10 mL
overnight precultures, 50 mL cultures were
generated with an optical density (
) of = 0.1, with both cultures shaken at 250 rpm
Sensors 2022,22, 4352 3 of 17
at 30
C (Eppendorf Innova
42 Benchtop incubator shaker, Eppendorf, France). Growth
was monitored over time on bacterial culture diluted 1/10 in LB media measured by a
spectrophotometer (SAFAS UVmc2) at 620 nm in disposable cuvettes (Brand GmBH, ref
759015) until it reached
= 0.4, which corresponds to the middle of the exponential
growth phase.
2.2. Toxic Substances
Three toxic substances, norfloxacin, copper and 3.5-dichlorophenol, were used to
represent three families of pollutants present in the environment (antibiotics, heavy metals
and herbicides, respectively). A 250 mg.L
solution of antibiotic was prepared by adding
25 mg norfloxacin (Sigma-Aldrich, St. Louis, MI, USA, ref N9890-5G) to 100 mL distilled
water. A 2 g.L
metal solution was prepared by adding 62.6 mg CuSO
O (Fisher
Scientific, Hampton, NH, USA, ref A778701) to 20 mL distilled water. A 10 g.L
solution was prepared by adding 0.25 g 3.5-dichlorophenol (Sigma-Aldrich, ref D7060-0) to
25 mL distilled water. All concentrated stock solutions were prepared in advance, sterilized
by filtration using a 0.22
m porosity cellulose acetate membrane (Dutsher, Brumath,
France, ref 146560) in a UV biological safety cabinet (JOUAN MSC 12, France), and then
aliquoted and stored at 20 C in a light-protected place.
2.3. Sample Preparation for Toxicant Exposure
STEP 1. The protocol used was adapted from the work of Bittel et al. [
]. After
bacteria reached
= 0.4, the microorganisms were washed to remove any residues of
medium or substance that could affect measurement. Washing was done by centrifugation
for 5 min at 4800 g and 4
C (Awel MF20-R), and the pellet was then resuspended in
20 mL
of 10
solution (MgSO
O Sigma, ref M1880) prepared in ultrapure water
and then sterilized by autoclaving. For each toxicity test, four Erlenmeyer flasks with
biomass production were cultivated separately and then mixed during the washing phases.
The cycle of washing was repeated three times.
STEP 2. After washing, the biomass was placed in different Erlenmeyer flasks con-
taining 50 mL of MgSO
solution to which the desired toxicant concentrations were added.
For each tested pollutant, three concentrations were tested in triplicate. The concentra-
tions were chosen to represent, where possible, subtoxic, toxic and lethal concentrations
(Table 1). A control sample without toxicant was included in parallel with each test. Initial
optical density was 0.4, and incubation was carried out at 30
C under constant agitation at
250 rpm for 40 min under the same conditions as during the biomass production phase.
Table 1. Tested concentrations of the three pollutants in the study.
Subtoxic Toxic Lethal
Norfloxacin 0.25 mg.L12.5 mg.L125 mg.L1
Copper 0.25 mg.L11 mg.L12.5 mg.L1
3.5-Dichlorophenol 2.5 mg.L125 mg.L1250 mg.L1
STEP 3. After exposure, the microorganisms were washed twice with MgSO
in the same way as in STEP 1, using centrifugation to remove residues of substances that
might bias the measurement. Then, 1 mL of MgSO
solution was added to the biomass
pellet, and the samples were centrifuged in 2 mL Eppendorf tubes for 2 min at 10,000 rpm
(MinSpin plus, Eppendorf). All liquid was then separated from the bacterial pellet and,
finally, 25
L of MgSO
solution was added to each sample to give a final volume of 60
STEP 4. This volume of washed cells was then deposited on a quartz fiber filter
(Dutscher, grade 293 with a 0.2
m pore size) using a micropipette. The separated liquid
was absorbed using cotton rolls (Teqler, 258 Praxisdient, 12-mm diameter). The filtration of
four samples took around 10 min (Figure 1).
Sensors 2022,22, 4352 4 of 17
Figure 1.
Schema of the toxicity test procedure with the automated spectroscopic system: (
) Bacterial
preparation and toxicant exposure followed by washing steps to prepare sample for filtration. (
Set-up of multi-well support system for filtration with absorbing cotton rolls and quartz fiber filters.
) Principle of microorganism suspension filtration and liquid absorption by cotton rolls with a
function for disconnecting the filters from the wet cotton and fixing them on the motorized 3-axis
platform. Demonstration of the ability of quartz fiber filter (diameter 16 mm) to absorb different
volumes of Lugol’s solution when in contact with cotton rolls (diameter 12 mm, height 32 mm).
) Automatic spectroscopic measurements with fixed Raman fiber-optic probe head connected
to a spectrometer and motorized 3-axis programmed platform simultaneously controlled from a
connected computer using a specifically designed API.
2.4. Raman Measurements
A confocal Raman microscope (SENTERRA, Bruker Optics, Germany) was used for
the measurements of the different filters. This device was equipped with two gratings (400
and 1200 lines/mm), a CCD camera cooled to
C and a BX51 Olympus microscope
with multiple objectives (LMPLFLN 100x/0.8 objective, laser spot = 1.12
m). The analyses
were performed at 785 nm, and the laser power was approximately 25 mW on the sample.
The spectral resolution was approximately 8 cm
, and integration time was 30 s with two
co-additions. Raman spectra were acquired using OPUS software (Bruker Optics, Ettlingen,
For field work, we used a portable Raman device (QE Pro-Raman spectrometer, Ocean
Optics, Netherlands), hereafter referred to as the ‘Portable Fiber-Optic system’. OmniDriver
and SPAM (spectral processing and maths, respectively) libraries were used for the Portable
Fiber-Optic system. The spectrometer has low-noise electronics (dynamic range: 85,000:1
and System SNR: 1000:1) with typical back-thinned CCD array miniature spectrometers
cooling to
C below ambient air. The spectral resolution was approximately equal
to 13 cm
in the measured wavelength range (200–4000 cm
). The fiber-optic probe
(InPhotonics RPB785) consists of a permanently aligned combination of two single fibers
m excitation fiber and a 200-
m collection fiber) with filtering and steering micro-
optics, in a rugged polyurethane jacket. The stainless-steel probe tip is 38 mm long with
a working distance of 7.5 mm. Analyses were performed with laser excitation at 785 nm
wavelength and 239 mW power on the control sample run before every experiment with a
power-meter (PM100D Thorlabs). The integration time of spectral acquisition was 30 s.
Sensors 2022,22, 4352 5 of 17
A fiber-optic probe head with a small numeric aperture (NA = 0.22) was selected due
to its greater depth of field, thus allowing less precise focusing in the z-direction for a more
homogeneous signal from multiple samples.
2.5. Design and Fabrication of a Cell Biosensor
The automated spectroscopic system developed in this work is able to measure up
to nine biological samples. It consists of a multi-well filtration support system for sample
deposition and fixation, a motorized 3-axis platform (Standa, 8-0026) with a fixation base
for a sample holder placement, and Raman spectrometer with flexible enhancement and col-
lection arms for easy access to the sample surface. The measuring system is in an enclosed
space with an external light-blocking function that allows spectroscopic measurements to
be performed in the dark. The system is controlled from a pilot computer with a specially
developed application programming interface (API) for simultaneous motorized platform
displacement and spectroscopic measurements (Figure 1).
Part 1. Filtration system design
A new optimized filtration method was developed in which the cotton rolls are brought
into contact with the filter in order to optimize the absorption of the aqueous medium
during filtration (Figure 1B). Nine filters and cotton rolls are fixed in one custom made
multi-well support system. The cotton rolls can absorb up to 4 mL of liquid (Figure 1C).
The filtration support was designed in the SOLIDWORKS environment and produced from
acetal plates using a programmable milling machine (Charly4U); clips to avoid the use of
screws for fixing and positioning were produced on a 3D printer (DAGOMA, Neva magis).
Part 2. Set-up of the measurement chamber
The fixation support with the filters is separated from the wet cotton rolls and placed on
a custom-made base so they can be positioned on a motorized 3-axis platform (
Figure 1B,C
The entire system is secured in a lockable box, coated on the inside with black opaque
adhesive film to block out light from the outside and reduce possible reflections of laser
radiation on the inside. The Raman probe is fixed in a specially made cage, which is fixed
to the ’ceiling’ of the enclosure and has a height adjustment function (Figure 1D).
Part 3. Programming automated spectroscopic measurements
An API was specifically developed to simultaneously operate the motorised 3-axis
platform and the Portable Fiber-Optic system. The Libximc cross-platform library was used
to control the motorized 3-axis platform.
The API developed makes it possible to select up to nine wells containing samples for
measurements and their sequence. The position focusing on the z-axis is first performed
manually on the first well. The spectrometer is calibrated with the dark current level and
set with the spectroscopic measurement parameters such as the integration time, number
of acquisitions and number of runs per well (laser power is set separately with the Oxxius
program, LaserBoxx HPE series). Spectroscopic measurements are then run automatically,
one spectrum from each sample in sequence, with programmable parameters for the
specified number of runs. The API is programmed to measure spectra in seven different
spots on the filter (one in the center and six around this point) to homogenize the spectral
response from the sample.
Sensors 2022,22, 4352 6 of 17
Part 4. Data mining for toxicity analysis
The raw data had a spectral range from 181 to 4045 cm
. The first pre-processing
step was to cut the raw spectra in the defined spectral zone of interest: between 550 and
1780 cm
for E. coli MG1655. The baseline correction was processed by an elastic concave
method (64
and 10 iterations) using OPUS software. Data processing was then performed
with MATLAB (version 2019) using the SAISIR package [
]. The spectra were normalized
using the probabilistic quotient normalization (PQN) method with respect to the median
spectrum of each sample group.
Statistical analysis
The statistics were also performed with MATLAB (version 2019) using the SAISIR
package [
]. The statistical analysis strategy was based on independent component
analysis (ICA), which was performed using the JADE algorithm [
]. The significance of
differences between groups was tested by ANOVA with a statistical significance threshold
of p-value < 0.05. To quantify the classification results of the spectra, stepwise factorial
discriminant analysis (sFDA) procedures were then performed on scores from the ICA
selected after the ANOVA analysis. Each discriminant model was calculated using a cross-
validation procedure (random selection of 2/3 of scores for the calibration model and 1/3
of scores for the validation test). The final classification rates corresponded to an average of
600 sFDA iterations.
3. Results and Discussion
3.1. Selection of a Suitable Filter for Bacterial Raman Analysis
Commonly used filter types (glass fiber, cellulose, PTFE, isotropic aluminium and
quartz fiber) were tested in order to select the best one for bacterial analysis by Raman
spectroscopy. In Figure 2(left), Raman signatures of the five analyzed filters are shown
beside the Raman signal released from E. coli deposited on a gold surface. This allows to
compare the filter backgrounds, which can overlap with Raman bands of bacteria. Bulk
gold is known to be Raman inactive due to its pure face-centred cubic crystal structure [
All of the spectra were measured using the same acquisition parameters. Glass, cellulose
and PTFE filters have a significant Raman signal at different Raman shifts, making these
filters unusable for analyzing bacteria. Isotropic aluminum filters have a weaker Raman
response but are very expensive, which is not practical for daily routine applications.
Raman signal intensity and band density are smallest for spectra measured on a quartz
fiber filter (Figure 2, left). Raman signals recorded from a gold surface and quartz fiber filter
are very similar (Figure 2, right). The quartz fiber filter was selected as the best developed
Raman biosensor for many reasons: it has a low Raman signal at 785 nm wavelength and
is produced with a variety of pore sizes, which ensures an effective retention of bacteria
from a filtrated suspension. Moreover, its price is low (0.13
for 1 cm
), which is important
for future field applications. Figure 2(right) also shows that Raman spectra of bacteria
measured on a gold surface and those measured on quartz filter have no visible differences.
Sensors 2022,22, 4352 7 of 17
Figure 2.
Filter selection based on spectra: (
) Table: Raman signatures of the five tested filters
compared with that of E. coli MG1655. Blue spectra: filter signatures; red spectra: E. coli bacterial
signature on a gold surface. All spectra presented were measured at a 785 nm laser excitation
wavelength and the same acquisition parameters, using the microscope objective
100 NA = 0.8.
) Spectral comparison of the five filters, a clean gold surface and E. coli bacteria on a gold
surface and on a quartz fiber filter.
3.2. Impact of Drying Time on the Quality of Raman Spectra of Bacteria
Bacterial cultures were deposited on three quartz filters and measured over six dif-
ferent drying times (5, 45, 90, 135, 180, 220 min). Every spectrum in Figure 3A represents
an average of 21 spectra (seven spectra from three samples of filtrated bacteria with iden-
tical drying times). Visual inspection of these spectra allows to identify 11 Raman bands
impacted by drying time, which were labeled from A to K (Figure 3A,B). It can be seen
that the RNA band (808 cm
, band C) decreases with drying time, which signifies a
morphological change in the bacteria. After 220 min, the bacterial RNA band has almost
disappeared, which indicates the death of the cell [
]. The best time to extract infor-
mation from a microorganism about its viability and toxicity response is when a control
sample (one not exposed to the toxic substance) has a DNA/RNA ratio of the order of 1
(780 cm1/808 cm1), as shown in Figure 3A (bands at positions B and C).
Sensors 2022,22, 4352 8 of 17
Figure 3. Validation of the best drying time for a bacterial suspension after filtration for subsequent
measurement on the semi-automatic system: (
) Average of 21 Raman spectra of E. coli MG1655
bacteria performed after the same drying time on three quartz fiber filters (seven spectra per filter).
) Average correlation level of each Raman band per drying time with the corresponding Raman
band of reference spectrum (E. coli deposited on gold surface). (
) Repeatability or homogeneity of
the measurements represented by the autocorrelation level between seven spectra from each sample
for every time interval. (
) Reproducibility of measurements represented by the correlation level
between all three samples for every time period (21 spectra per time period).
The correlations of the 11 Raman bands with the reference spectrum were calculated to
determine at what moment the spectra from filtrated bacteria could be qualified for toxicity
evaluation. The reference spectrum is the Raman signal issued from E. coli deposited on
a gold surface. The average correlation values of each Raman band per drying time with
the corresponding Raman band of the reference spectrum are presented in the table of
Figure 3B.
All the Raman bands presented were evaluated with >90% or >98% correlation confi-
dence. As can be seen in Figure 3B, during the first hour of measurements, the spectrum of
a wet bacterium on a quartz filter is quite noisy, which is reflected by the lower correlation
values and, as a result, only two Raman bands have a correlation of >98% with the reference
spectrum. The correlation values increase with time and, as can be seen for a drying time of
135 min, eight of the Raman bands have a correlation >98% and all 11 Raman bands have a
good correlation, with a reliability >90%. Nevertheless, after 180–220 min of drying, the
bacteria begin to degrade, resulting in reductions in the correlations and richness of the
spectral information.
Sensors 2022,22, 4352 9 of 17
To confirm these observations, the repeatability and reproducibility of measurements
were calculated to show whether the Raman signal from a sample was stable over time and
between samples. The repeatability of measurements is represented by the autocorrelation
level between spectra from each sample for every time interval (Figure 3C). The dispersion
of values from three samples taken after 5 and 45 min of drying is about 5%, which is very
significant in terms of accuracy requirements for toxicity measurements (it should be of the
order of 1%). Starting from 90 min of drying time, the dispersion in autocorrelation values
between spectra is at an acceptable level of 98%. The reproducibility of measurements
illustrates the correlation level between all three samples for every time period (21 spectra
per period), which was also calculated (Figure 3D). This value is greater than 98% at 90
and 135 min of drying time. Based on these results, the best time for spectroscopic toxicity
measurements on the filtrated bacteria was, therefore, set at between 90 and 135 min after
deposition without any impact on the quality of spectra.
3.3. Evaluation of the Molecular Targets of Chemical Pollutants in Bacterial Cells
The assignment of Raman bands in the reference bacterial spectrum makes it possible
to highlight the bands of molecules that may be impacted by the pollutants (Figure 4A).
Most biological molecules in a bacterial cell are visible in this spectrum, e.g., adenine
(720 cm
), DNA/RNA (785/850 cm
), tyrosine (830 cm
), phenylalanine (1000 cm
phosphate groups (1100 cm
), group III amides (amides III, 1240 cm
group II amides (amides II, 1330 cm
), proteins and fatty acids (1450 cm
), guanine
adenine and uracil (1570 cm
), lipids and group I amides (amides I, 1650–1680 cm
) [
Figure 4. Cont.
Sensors 2022,22, 4352 10 of 17
Figure 4.
Effects of norfloxacin on the Raman spectra of E. coli MG1655: (
) Assignment of some
characteristic bands in the bacterial spectrum. (
) Averages of seven Raman spectra obtained
following exposure of the bacteria to different concentrations of norfloxacin. The highlighted bands
are those that allow the spectra to be classified according to the different concentrations of toxicant.
) Three-dimensional (3D) representation of the spectral distribution according to the three most
significant components from the ICA. (
) Loadings of the most significant ICs from the analysis of
Raman spectra of E. coli MG1655 exposed to norfloxacin. The spectra were decomposed by ICA, and
the most significant ICs were then selected. (
) ANOVA analysis of the distribution of the spectra
according to the most significant component (p-value < 0.05). (
) Classification results of the sFDA
performed after the pre-processing steps of spectrum selection (size of sample: 24 spectra).
3.3.1. Molecular Targets of Norfloxacin in E. coli Cells
The analysis of the spectral signature of E. coli cells exposed to the different concen-
trations of norfloxacin shows Raman bands impacted by this toxicant (Figure 4B). The
effects concern the DNA and RNA bands at 785 and 850 cm
, the DNA
phate groups at 1070 and 1150 cm
, amides II (band at 1330 cm
), amides I and lipids
(band at 1650 cm
). These differences in the molecular fingerprint of the bacteria result
from physiological changes provoked by reactions to the antibiotic. These bands, high-
lighted in Figure 4B, allow the best discrimination of the spectra according to the different
concentrations of antibiotic.
The loadings of the three independent components (ICs) show that the variability in
the spectra is a function of antibiotic concentration (Figure 4D). It can first be seen that the
distribution of the spectra according to these components makes it possible to distinguish
the different concentration groups (3D representations, Figure 4C). This selection was
made computationally by observing the distribution of spectra for each of the ICs. ICs
were selected for which the variability of the inter-group spectra (i.e., as a function of
concentrations) was the lowest possible, while maximizing the mean difference with the
other groups. The variability of the spectra according to these IC components was also
analyzed by ANOVA (Figure 4E). The results show a distribution of groups consistent with
a dose–response effect of the substance, which is well underlined by the ANOVA results on
the IC6 specific to lipids and amides I (band at 1650–1680 cm1).
These observed variations are related to known mechanisms of the functioning of this
antibiotic [
]. The spectra of bacteria exposed to increasing concentrations of norfloxacin
show a decrease in the intensity of the bands corresponding to DNA and RNA (Loading
IC5, Figure 4D). Norfloxacin belongs to a family of second-generation quinolones and
acts, in particular, by inhibiting DNA gyrase and type IV topoisomerases at the DNA
Sensors 2022,22, 4352 11 of 17
segmentation stage. Inhibition of these enzymes disrupts the DNA segmentation phases,
preventing the re-pairing of the strands. This disrupts DNA replication mechanisms,
leading to the inhibition of DNA synthesis, which may explain the decrease in the intensity
of the corresponding Raman bands. After a period of bacteriostasis, cell death is associated
with the appearance of double-strand breaks, causing chromosome fragmentation.
Norfloxacin is known to induce an increase in the production of fatty acids in E. coli
and a decrease in glycerophospholipid production [
]. Because fatty acids are used by the
cell as raw materials for more complex lipids, an increase in fatty acid synthesis combined
with a decrease in phospholipid synthesis would be consistent with the activation of cell’s
resistance mechanism to compensate for the deterioration of its membrane.
Stepwise factorial discriminant analysis (sFDA) was performed on the scores of inde-
pendent components from the ICA procedure to assess the level of correct prediction in
assigning spectra to a particular group (Figure 4F). Correct classification percentages for
the control and the different norfloxacin concentrations (0.25, 2.5 and 25 mg.L
) were 97,
99, 82 and 93%, respectively. These results, obtained by a portable fiber-optic spectrometer,
are comparable to those from a previous study obtained with a benchtop confocal Raman
spectrometer [28].
3.3.2. Molecular Targets of Copper in E. coli Cells
The analysis of the spectral signature of E. coli MG1655 cells exposed to copper shows
Raman bands impacted by this toxicant (Figure 5). The most significant spectral changes
are in the DNA and RNA bands at 785, 810 cm
, and bands located between 1050 and
1150 cm
associated with the DNA
phosphate groups. The intensity of all these
bands decreases with increasing copper concentration (Figure 5B). Nevertheless, two
subgroups may be observed: the spectra for low copper concentrations (0 and 0.25 mg.L
have the same Raman profiles, in particular for the band at 810 cm
. For the higher
concentrations (1 and 2.5 mg.L
), the Raman band at 810 cm
has disappeared. These
very interesting results show that copper heavily impacts nucleic acid bands and causes
the death of cells at higher concentrations. Copper toxicity also impacted the bands located
between 1200 and 1500 cm
associated with amides and proteins. Highlighted by the
most discriminating component of the ICA results, these bands are the most significant for
characterizing the effects of copper on E. coli MG1655 (Figure 5C,D). These ICs were also
analyzed by ANOVA (Figure 5E), which confirmed the existence of two sub-groups: one
for low concentrations (0 and 0.25 mg.L
) and another for higher concentrations (1 and
2.5 mg.L
). The ANOVA done on the IC8 (specific to the DNA
phosphate group
band at 1100 cm1) underlines this result.
It is interesting to note that these results are very similar to those obtained in a previous
study showing the impact of arsenic on E. coli cells [
]. The corresponding spectral changes
may be the result of oxidative phenomena. Indeed, both copper and arsenic induce large
numbers of free radicals, which are responsible for the irreversible denaturation of DNA
and RNA [
]. In addition, although less significant, other variations also appear in a
similar way for these metals. An increased intensity can be observed in the bands located
between 1200 and 1400 cm
, which are associated with amides III, which are characteristic
of proteins and lipids (Figure 5B). These changes can be attributed to the denaturation
phenomena [
]. Similar spectral changes can be seen for the band corresponding to
phenylalanine (1000 cm
) and the band corresponding to lipids and amides I (between
1650 and 1690 cm
). In both cases, there is observed not only an increase in the intensity of
the latter band with increasing toxicant concentrations but also a slight shift of a few cm
to the right. Both copper and arsenic act at the membrane level, notably through chelation
mechanisms [
]. The resulting depolarization may explain the slight shift observed in the
Raman bands. Copper and arsenic, which have similar toxicity mechanisms, thus seem to
induce equivalent variations in the Raman spectra of E. coli.
Sensors 2022,22, 4352 12 of 17
Figure 5.
Effects of copper on the Raman spectra of E. coli MG1655: (
) Averages of seven Raman
spectra obtained following exposure of the bacteria to different concentrations of copper. The high-
lighted bands are those that allow the spectra to be classified according to the different concentrations
of toxicant. (
) Three-dimensional (3D) representation of the spectral distribution according to the
three most significant components from the ICA. (
) Loadings of the most significant ICs from the
analysis of Raman spectra of E. coli MG1655 exposed to copper. The spectra were decomposed by
ICA, and the most significant ICs were then selected. (
) ANOVA analysis of the distribution of the
spectra according to the most significant component (p-value < 0.05). (
) Classification results of the
sFDA performed after the pre-processing steps of spectrum selection (size of sample: 24 spectra).
Stepwise factorial discriminant analysis (sFDA) shows good classification scores for
the control and the different copper concentrations (0.25, 1 and 2.5 mg.L
), correct at 97,
85, 93 and 81%, respectively (Figure 5E). Most of the misclassified spectra can be attributed
to neighbouring groups. For example, 18.81% of the 2.5 mg.L
concentration spectra
were attributed to the 1 mg.L
concentration group. Similarly, the misclassified spectra of
the 0.25 mg.L
group were reciprocally attributed to the control (12.43%) and 1 mg.L
Sensors 2022,22, 4352 13 of 17
(2.1%) groups. This classification confirms the visual observations and allows classifying
the impact of copper into two modes. For concentrations lower than 0.25 mg.L
, copper
does not impact the bacteria greatly. However, from a concentration of 1 mg.L
above, copper becomes very toxic and causes cell death (as was also observed for the two
high concentrations).
3.3.3. Molecular Targets of 3,5-Dichlorophenol in E. coli Cells
The analysis of the spectral signature of E. coli cells exposed to different concentra-
tions of 3,5-dichlorophenol shows Raman bands impacted by this toxicant (Figure 6). The
toxicity signature of this pollutant specifically concerns the bands associated with pheny-
lalanine (1000 cm
) and the DNA
phosphate groups (1100 cm
). The intensity
of these bands diminishes with increasing concentrations of the toxicant (Figure 6B). A
band associated with amides II (1330 cm
) is slightly higher for the highest concentration
of toxicant, although it can be noted that in general, the intensity of this band does not
change very much. Highlighted by the most discriminating component of the ICA results,
these bands are the most significant for characterizing the effects of 3,5-dichlorophenol on
E. coli MG1655 (Figure 6C,D). These three IC loadings can be seen to vary, and ANOVA
reveals a significant difference between the control and 3,5-dichlorophenol-treated bacteria
(p-value < 0.05) (Figure 6E).
Phenolic compounds are lipophilic compounds whose toxic effects are mainly driven
by their action at the microbial membrane level. Their effect leads to variations in the
lipid/protein ratio as well as to the dysfunction of certain membrane proteins. A study by
Keweloh et al. [
] on the growth of E. coli in the presence or absence of phenols, showed
a decrease in the lipid/protein ratio under the effect of the toxicant, in particular due to
a decrease in the quantity of phospholipids. Thus, the relative variations of these two
bands, recurrently identified in the loadings of the significant ICs, tend to show a relative
decrease in the bands associated with lipids compared with those associated with proteins
(Figure 6D). Furthermore, the action of phenols would cause an increase in membrane
permeability, and the increase in the quantity of membrane proteins would then result in
adaptation to limit the leakage of cellular constituents.
Figure 6. Cont.
Sensors 2022,22, 4352 14 of 17
Figure 6.
Effects of 3,5-dichlorophenol on the Raman spectra of E. coli MG1655: (
) Averages of
seven Raman spectra obtained following exposure of the bacteria to different concentrations of
3,5-dichlorophenol. The highlighted bands are those that allow the spectra to be classified according
to the different concentrations of toxicant. (
) Three-dimensional (3D) representation of the spectral
distribution according to the three most significant components from the ICA. (
) Loadings of
the most significant ICs from the analysis of Raman spectra of E. coli MG1655 exposed to 3,5-
dichlorophenol. The spectra were decomposed by ICA, and the most significant ICs were then
selected. (
) ANOVA analysis of the distribution of the spectra according to the most significant
component (p-value < 0.05). (
) Classification results of the sFDA performed after the pre-processing
steps of spectrum selection (size of sample: 24 spectra).
Stepwise factorial discriminant analysis (sFDA) demonstrates good classification scores
for the control and different 3,5-dichlorophenol concentrations (2.5, 25, and 250 mg.L
correct at 100, 92, 90, and 96%, respectively (Figure 6E).
4. Conclusions
A newly designed filtration technique for microorganism suspensions coupled to an
automated spectroscopic measurement device for up to nine samples opens broad research
perspectives for a Raman biosensor development using a Portable Fiber-Optic system. We
determined the best parameters for toxicity measurements on filtered bacteria to ensure a
favorable repeatability and reproducibility of the tests. These findings demonstrate that it
is possible to reduce the number of spectra needed from each sample from 40 to only seven
for toxicity evaluation, which reduces the measurement time for four samples from 3 h to
35 min in automatic mode without intervention from an operator. Toxicity tests to validate
the new method with the automated spectroscopic system showed good agreement with
previous results [28] and the potential for future biosensor development.
Nonetheless, the system developed here remains a device for laboratory use. Further-
more, the systematic analyses of the results are still an integral part of the work required to
make a useful interpretation of the information collected. To this end, the development of a
model allowing the simultaneous consideration of signatures obtained for all the microor-
ganisms in a given water sample or environment will make it possible to automatically
evaluate the information provided by the different cellular components, which can be
observed through the combination of their Raman spectra. This future work will require
transversal collaboration of different specialities in biology and chemometrics.
Sensors 2022,22, 4352 15 of 17
Author Contributions:
Conceptualization, G.T.; Methodology, A.A.; Project administration, M.B.;
Resources, M.B.; Supervision, A.A. and G.T.; Validation, O.B. and M.-J.D.; Visualization, M.-J.D.;
Writing—review & editing, O.B. and A.A. All authors have read and agreed to the published version
of the manuscript.
This research was funded by Association Nationale de la Recherche et de la Technologie,
grant number 2017/1221.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
The authors thank the Association Nationale de la Recherche et de la Technologie
(ANRT) and the Tronico company for their support and financial assistance for this work.
Conflicts of Interest:
The authors declare no competing financial interest and respect the ethics
of references.
Hasan, J.; Goldbloom-Helzner, D.; Ichida, A.; Rouse, T.; Gibson, M. Technologies and Techniques for Early Warning Systems to Monitor
and Evaluate Drinking Water Quality: A State-of-the-Art Review; Environmental Protection Agency Washington DC Office of Water:
Washington, DC, USA, 2005.
Su, X.; Sutarlie, L.; Loh, X.J. Sensors, biosensors, and analytical technologies for aquaculture water quality. Research
8272705. [CrossRef] [PubMed]
Abdullah, A.M. Water Quality Testing Protocol 2015: A Guideline for Water Quality Monitoring; WaterAid Bangladesh: Dhaka,
Bangladesh, 2005. [CrossRef]
Kokkali, V.; van Delft, W. Overview of commercially available bioassays for assessing chemical toxicity in aqueous samples. TrAC
Trends Anal. Chem. 2014,61, 133–155. [CrossRef]
Bacon, J.R.; Butler, O.T.; Cairns, W.R.; Cavoura, O.; Cook, J.M.; Davidson, C.M.; Mertz-Kraus, R. Atomic spectrometry update—A
review of advances in environmental analysis. J. Anal. At. Spectrom. 2021,36, 10–55. [CrossRef]
ISO 13.060.50. Détermination des Substances Chimiques de L’Eau. Available online:
(accessed on 27 July 2021).
ISO 13.060.60. Détermination des Propriétés Physiques de L’Eau. Available online:
(accessed on 27 July 2021).
Brack, W.; Aissa, S.A.; Backhaus, T.; Dulio, V.; Escher, B.I.; Faust, M.; Altenburger, R. Effect-based methods are key. The European
Collaborative Project SOLUTIONS recommends integrating effect-based methods for diagnosis and monitoring of water quality.
Environ. Sci. Eur. 2019,31, 10. [CrossRef]
Anderson, B.G. The toxicity thresholds of various sodium salts determined by the use of Daphnia magna. Sew. Work. J.
Fernández-Piñas, F.; Rodea-Palomares, I.; Leganés, F.; González-Pleiter, M.; Muñoz-Martín, M.A. Evaluation of the ecotoxicity of
pollutants with bioluminescent microorganisms. In Bioluminescence: Fundamentals and Applications in Biotechnology—Volume 2;
Springer: Berlin/Heidelberg, Germany, 2014; pp. 65–135. [CrossRef]
Thouand, G. Introduction to Cell Biosensors Through 55 Years of Scientific Production. In Handbook of Cell Biosensors; Springer:
Berlin/Heidelberg, Germany, 2022; pp. 3–40. [CrossRef]
Nakamura, H. Current status of water environment and their microbial biosensor techniques–Part II: Recent trends in microbial
biosensor development. Anal. Bioanal. Chem. 2018,410, 3967–3989. [CrossRef] [PubMed]
13. Martin-Yken, H. Yeast-based biosensors: Current applications and new developments. Biosensors 2020,10, 51. [CrossRef]
Kabessa, Y.; Korouma, V.; Yagur-Kroll, S.; Belkin, S.; Agranat, A.J. Simultaneous quantification of the fluorescent responses of an
ensemble of bacterial sensors. Biosens. Bioelectr. 2013,49, 394–398. [CrossRef]
Jouanneau, S.; Durand, M.J.; Courcoux, P.; Blusseau, T.; Thouand, G. Improvement of the identification of four heavy metals in
environmental samples by using predictive decision tree models coupled with a set of five bioluminescent bacteria. Environ. Sci.
Technol. 2011,45, 2925–2931. [CrossRef]
Elad, T.; Benovich, E.; Magrisso, S.; Belkin, S. Toxicant identification by a luminescent bacterial bioreporter panel: Application of
pattern classification algorithms. Environ. Sci. Technol. 2008,42, 8486–8491. [CrossRef]
Song, Y.; Jiang, B.; Tian, S.; Tang, H.; Liu, Z.; Li, C.; Jia, J.; Huang, W.; Zhang, X.; Li, G. A whole-cell bioreporter approach for the
genotoxicity assessment of bioavailability of toxic compounds in contaminated soil in China. Environ. Poll.
,195, 178–184.
Song, Y.; Li, G.; Thornton, S.F.; Thompson, I.P.; Banwart, S.A.; Lerner, D.N.; Huang, W.E. Optimization of bacterial whole cell
bioreporters for toxicity assay of environmental samples. Environ. Sci. Technol. 2009,43, 7931–7938. [CrossRef]
Sensors 2022,22, 4352 16 of 17
Cregut, M.; Hua, A.; Jouanneau, S.; Assaf, A.; Cordella, C.B.; Thouand, G.; Durand, M.J. Screening of metallic pollution in complex
environmental samples through a transcriptomic fingerprint method. Environ. Sci. Poll. Res. 2022, 29, 1037–1050. [CrossRef]
Moritz, T.J.; Polage, C.R.; Taylor, D.S.; Krol, D.M.; Lane, S.M.; Chan, J.W. Evaluation of Escherichia coli cell response to antibiotic
treatment by use of Raman spectroscopy with laser tweezers. J. Clin. Microbiol. 2010,48, 4287–4290. [CrossRef]
Neugebauer, U.; Schmid, U.; Baumann, K.; Holzgrabe, U.; Ziebuhr, W.; Kozitskaya, S.; Popp, J. Characterization of bacterial
growth and the influence of antibiotics by means of UV resonance Raman spectroscopys. Biopolym. Orig. Res. Biomol.
306–311. [CrossRef]
Zu, T.N.; Athamneh, A.I.; Wallace, R.S.; Collakova, E.; Senger, R.S. Near-real-time analysis of the phenotypic responses of
Escherichia coli to 1-butanol exposure using Raman spectroscopy. J. Bacteriol. 2014,196, 3983–3991. [CrossRef]
Athamneh, A.I.M.; Alajlouni, R.A.; Wallace, R.S.; Seleem, M.N.; Senger, R.S. Phenotypic profiling of antibiotic response signatures
in Escherichia coli using Raman spectroscopy. Antimicrob. Agents Chemother. 2014,58, 1302–1314. [CrossRef]
Daniel, P.; Picart, P.; Bendriaa, L.; Sockalingum, G.D.; Adt, I.; Charrier, T.; Thouand, G. Effects of Toxic Organotin Compounds on
Bacteria Investigated by Micro-Raman Spectroscopy. Spectrosc. Lett. 2008,41, 19–28. [CrossRef]
Song, Y.; Cui, L.; López, J.Á.S.; Xu, J.; Zhu, Y.G.; Thompson, I.P.; Huang, W.E. Raman-Deuterium Isotope Probing for in-situ
identification of antimicrobial resistant bacteria in Thames River. Sci. Rep. 2017,7, 16648. [CrossRef]
Stöckel, S.; Kirchhoff, J.; Neugebauer, U.; Rösch, P.; Popp, J. The application of Raman spectroscopy for the detection and
identification of microorganisms. J. Raman Spectrosc. 2016,47, 89–109. [CrossRef]
27. Yi, X.; Song, Y.; Xu, X.; Peng, D.; Wang, J.; Qie, X.; Huang, W.E. Development of a Fast Raman-Assisted Antibiotic Susceptibility
Test (FRAST) for the Antibiotic Resistance Analysis of Clinical Urine and Blood Samples. Anal. Chem.
,93, 5098–5106.
Bittel, M. Détection de Polluants Chimiques par Biocapteurs Bactériens Couplés à la Spectroscopie Raman. Ph.D. Thesis,
University of Nantes, Nantes, France, 2017.
Bittel, M.; Cordella, C.B.Y.; Assaf, A.; Jouanneau, S.; Durand, M.J.; Thouand, G. Potential of Raman spectroscopy to monitor
arsenic toxicity on bacteria: Insights toward multiparametric bioassays. Environ. Sci. Technol. 2015,49, 12324–12332. [CrossRef]
Butler, H.J.; Ashton, L.; Bird, B.; Cinque, G.; Curtis, K.; Dorney, J.; Martin, F.L. Using Raman spectroscopy to characterize biological
materials. Nat. Protoc. 2016,11, 664. [CrossRef]
Assaf, A.; Cordella, C.B.; Thouand, G. Raman spectroscopy applied to the horizontal methods ISO 6579: 2002 to identify
Salmonella spp. in the food industry. Anal. Bioanal. Chem. 2014,406, 4899–4910. [CrossRef]
Cui, L.; Butler, H.J.; Martin-Hirsch, P.L.; Martin, F.L. Aluminium foil as a potential substrate for ATR-FTIR, transflection FTIR or
Raman spectrochemical analysis of biological specimens. Analyt. Meth. 2014,8, 481–48. [CrossRef]
Rüger, J.; Schie, I.W.; Mondol, A.S.; Ramoji, A.; Weber, K.; Popp, J. Markierungsfreies Hochdurchsatzscreening mit Raman-
Spektroskopie BIOspektrum 2018,24, 499–503. [CrossRef]
34. Koch, M.D.; Shaevitz, J.W. Introduction to optical tweezers. Methods Mol. Biol. 2017,1486, 3–24. [CrossRef] [PubMed]
Li, Y.; Wang, G.; Yao, H.L.; Liu, J.; Li, Y.Q. Dual-trap Raman tweezers for probing dynamics and heterogeneity of interacting
microbial cells. J. Biomed. Opt. 2010,15. [CrossRef]
Zhang, P.; Kong, L.; Setlow, P.; Li, Y.Q. Multiple-trap laser tweezers Raman spectroscopy for simultaneous monitoring of the
biological dynamics of multiple individual cells. Opt. Lett. 2010,35, 3321–3323. [CrossRef]
Lau, A.Y.; Lee, L.P.; Chan, J.W. An integrated optofluidic platform for Raman-activated cell sorting. Lab Chip
,8, 1116–1120.
Kastanos, E.K.; Kyriakides, A.; Hadjigeorgiou, K.; Pitris, C. A novel method for urinary tract infection diagnosis and antibiogram
using Raman spectroscopy. J. Raman Spectrosc. 2010,41, 958–963. [CrossRef]
39. Mosier-Boss, P.A. Review on SERS of Bacteria. Biosensors 2017,7, 51. [CrossRef] [PubMed]
Koł ˛ataj, K.; Krajczewski, J.; Kudelski, A.A. Plasmonic nanoparticles for environmental analysis. Environ. Chem. Lett.
529–542. [CrossRef]
Weidemaier, K.; Carruthers, E.; Curry, A.; Kuroda, M.; Fallows, E.; Thomas, J.; Muldoon, M. Real-time pathogen monitoring
during enrichment: A novel nanotechnology-based approach to food safety testing. Int. J. Food Microbiol.
,198, 19–27.
Wu, X.; Han, C.; Chen, J.; Huang, Y.W.; Zhao, Y. Rapid detection of pathogenic bacteria from fresh produce by filtration and
surface-enhanced Raman spectroscopy. JOM 2016,68, 1156–1162. [CrossRef]
Gao, S.; Pearson, B.; He, L. Mapping bacteria on filter membranes, an innovative SERS approach. J. Microbiol. Methods
69–75. [CrossRef]
Cho, I.H.; Bhandari, P.; Patel, P.; Irudayaraj, J. Membrane filter-assisted surface enhanced Raman spectroscopyfor the rapid
detection of E. coli O157: H7 in ground beef. Biosens. Bioelectron. 2015,64, 171–176. [CrossRef]
Chen, J.; Wu, X.; Huang, Y.W.; Zhao, Y. Detection of E. coli using SERS active filters with silver nanorod array. Sens. Actuators
Chem. 2016,191, 485–490. [CrossRef]
Escoriza, M.F.; VanBriesen, J.M.; Stewart, S.; Maier, J.; Treado, P.J. Raman spectroscopy and chemical imaging for quantifcication
of filtered waterborne bacteria. J. Microbiol. Methods 2006,66, 63–72. [CrossRef]
47. Cordella, C.B.; Bertrand, D. SAISIR: A new general chemometric toolbox. TrAC Trends Anal. Chem. 2014,54, 75–82. [CrossRef]
Sensors 2022,22, 4352 17 of 17
JADE. Available online:
(accessed on 28 July 2021).
Devine, T.M.; Adar, F. Raman Spectroscopy of Solids. In Characterization of Materials; John Wiley & Sons, Inc.: Hoboken, NJ, USA,
2002; pp. 1–38. [CrossRef]
Escoriza, M.F.; Vanbriesen, J.M.; Stewart, S.; Maier, J. Studying bacterial metabolic states using Raman spectroscopy. Appl.
Spectrosc. 2006,60, 971–976. [CrossRef]
Schenk, E.R.; Nau, F.; Thompson, C.J.; Tse-Dinh, Y.C.; Fernandez-Lima, F. Changes in lipid distribution in E. coli strains in
response to norfloxacin. J. Mass Spectrosc. 2015,50, 88–94. [CrossRef]
Macomber, L.; Imlay, J.A. The iron-sulfur clusters of dehydratases are primary intracellular targets of copper toxicity. Proc. Natl.
Acad. Sci. USA 2009,106, 8344–8349. [CrossRef]
Escoriza, M.F.; VanBriesen, J.M.; Stewart, S.; Maier, J. Raman spectroscopic discrimination of cell response to chemical and
physical inactivation. Appl. Spectrosc. 2007,61, 812–823. [CrossRef]
Keweloh, H.; Weyrauch, G.; Rehm, H.J. Phenol-induced membrane changes in free and immobilized Escherichia coli. Appl.
Microbiol. Biotechnol. 1990,33, 66–71. [CrossRef]
... A spectrometer is a basic optical detection instrument that can obtain the spectral information of the measured object. At present, various spectrometers have been widely used in color measurement, gas composition analysis, agricultural medicine, food safety and other fields [1][2][3][4][5][6], and have become among the most important optical detection instruments. Among them, the Raman spectrometer [7][8][9], which is composed of a probe and a spectrometer, is a rapidly developed instrument in recent years. ...
Full-text available
Raman spectroscopy, measured by a Raman spectrometer, is usually disturbed by the instrument response function and noise, which leads to certain measurement error and further affects the accuracy of substance identification. In this paper, we propose a spectral reconstruction method which combines the existing maximum a posteriori (MAP) method and deep learning (DL) to recover the degraded Raman spectrum. The proposed method first employs the MAP method to reconstruct the measured Raman spectra, so as to obtain preliminary estimated Raman spectra. Then, a convolutional neural network (CNN) is trained by using the preliminary estimated Raman spectra and the real Raman spectra to learn the mapping from the preliminary estimated Raman spectra to the real Raman spectra, so as to achieve a better spectral reconstruction effect than merely using the MAP method or a CNN. To prove the effectiveness of the proposed spectral reconstruction method, we employed the proposed method and some traditional spectral reconstruction methods to reconstruct the simulated and measured Raman spectra, respectively. The experimental results show that compared with traditional methods, the estimated Raman spectra reconstructed by the proposed method are closer to the real Raman spectra.
... A transducer converts a biochemical signal, resulting from the interaction of a biological component, into a measurable signal. Thus, when the interaction between the analyte and the bioreceptor occurs, a quantifiable signal is generated, which can be optical, electrochemical, thermometric, piezoelectric, magnetic, or micromechanical [4][5][6]. ...
Full-text available
The coronavirus pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has accelerated the development of biosensors based on new materials and techniques. Here, we present our effort to develop a fast and affordable optical biosensor using photoluminescence spectroscopy for anti-SARS-CoV-2 antibody detection. The biosensor was fabricated with a thin layer of the semiconductor polymer Poly[(9,9-di-n-octylfluorenyl-2,7-diyl)-alt-2,2′-bithiophene-5,5′-diyl)] (F8T2) as a signal transducer material. We mounted the biosensors by depositing a layer of F8T2 and an engineered version of RBD from the SARS-CoV-2 spike protein with a tag to promote hydrophobic interaction between the protein and the polymeric surface. We validated the biosensor sensitivity with decreasing anti-RBD polyclonal IgG concentrations and challenged the biosensor specificity with human serum samples from both COVID-19 negative and positive individuals. The antibody binding to the immobilized antigen shifted the F8T2 photoluminescence spectrum even at the low concentration of 0.0125 µg/mL. A volume as small as one drop of serum (100 µL) was sufficient to distinguish a positive from a negative sample without requiring multiple washing steps and secondary antibody reactions.
Full-text available
Characterizing waste ecotoxicity is laborious because of both the undefined nature of environmental samples and the diversity of contaminants that can be present. With regard to these limitations, traditional approaches do not provide information about the nature of the pollution encountered. To improve such assessments, a fluorescent library of 1870 transcriptomic reporters from Escherichia coli K12 MG1655 was used to report the ecotoxic status of environmental samples. The reliability of the approach was evaluated with 6 metallic pollutants (As, Cu, Cd, Hg, Pb, Zn) used alone and in mixture in pure and complex matrices. A total of 18 synthetic samples were used to characterize the specificity of the resulting metallic contamination fingerprints. Metallic contamination impacted 4.5 to 10.2% of the whole transcriptomic fingerprint of E. coli. The analysis revealed that a subset of 175 transcriptomic reporters is sufficient to characterize metallic contamination, regardless of the nature of the sample. A statistical model distinguished patterns due to metallic contamination and provided information about the level of toxicity with 93 to 98% confidence. The use of the transcriptomic assessment was validated for 17 complex matrices with various toxicities and metal contaminants, such as activated sludge, wastewater effluent, soil, wood and river water. The presence of metals and their associated toxicity, which seems linked to their bioavailabilities, were thereby determined. This method constitutes a possible tool to screen unknown complex samples for their metallic status and identify those for which a deeper characterization must be achieved by the use of traditional biosensors and analytical methods.
Full-text available
Human health is at great risk due to the spreading of antimicrobial resistance (AMR). The lengthy procedure of conventional antimicrobial susceptibility testing (AST) usually requires a few days. We developed a fast Raman-assisted antibiotic susceptibility test (FRAST), which detects single bacterial metabolic activity in the presence of antibiotics, using Raman single-cell spectroscopy. It was found that single-cell Raman spectra (SCRS) would show a clear and distinguishable Raman band at the “silent zone” (2000–2300 cm–1), due to the active incorporation of deuterium from heavy water (D2O) by antibiotic-resistant bacteria. This pilot study has compared the FRAST and the conventional AST for six clinical standard quality controls (four Gram-negative and two Gram-positive bacteria strains) in response to 38 antibiotics. In total, 3200 treatments have been carried out and approximately 64 000 SCRS have been acquired for FRAST analysis. The result showed an overall agreement of 88.0% between the FRAST and the conventional AST assay. The gram-staining classification based on the linear discriminant analysis (LDA) model of SCRS was developed, seamlessly coupling with the FRAST to further reduce the turnaround time. We applied the FRAST to real clinical analysis for nine urinary infectious samples and three sepsis samples. The results were consistent with MALDI-TOF identification and the conventional AST. Under the optimal conditions, the “sample to report” of the FRAST could be reduced to 3 h for urine samples and 21 h for sepsis samples. The FRAST provides fast and reliable susceptibility tests, which could speed up microbiological analysis for clinical practice and facilitate antibiotic stewardship.
Full-text available
Biosensors are regarded as a powerful tool to detect and monitor environmental contaminants, toxins, and, more generally, organic or chemical markers of potential threats to human health. They are basically composed of a sensor part made up of either live cells or biological active molecules coupled to a transducer/reporter technological element. Whole-cells biosensors may be based on animal tissues, bacteria, or eukaryotic microorganisms such as yeasts and microalgae. Although very resistant to adverse environmental conditions, yeasts can sense and respond to a wide variety of stimuli. As eukaryotes, they also constitute excellent cellular models to detect chemicals and organic contaminants that are harmful to animals. For these reasons, combined with their ease of culture and genetic modification, yeasts have been commonly used as biological elements of biosensors since the 1970s. This review aims first at giving a survey on the different types of yeast-based biosensors developed for the environmental and medical domains. We then present the technological developments currently undertaken by academic and corporate scientists to further drive yeasts biosensors into a new era where the biological element is optimized in a tailor-made fashion by in silico design and where the output signals can be recorded or followed on a smartphone.
Full-text available
In aquaculture industry, fish, shellfish, and aquatic plants are cultivated in fresh, salt, or brackish waters. The increasing demand of aquatic products has stimulated the rapid growth of aquaculture industries. How to effectively monitor and control water quality is one of the key concerns for aquaculture industry to ensure high productivity and high quality. There are four major categories of water quality concerns that affect aquaculture cultivations, namely, (1) physical parameters, e.g., pH, temperature, dissolved oxygen, and salinity, (2) organic contaminants, (3) biochemical hazards, e.g., cyanotoxins, and (4) biological contaminants, i.e., pathogens. While the physical parameters are affected by climate changes, the latter three are considered as environmental factors. In this review, we provide a comprehensive summary of sensors, biosensors, and analytical technologies available for monitoring aquaculture water quality. They include low-cost commercial sensors and sensor network setups for physical parameters. They also include chromatography, mass spectrometry, biochemistry, and molecular methods (e.g., immunoassays and polymerase chain reaction assays), culture-based method, and biophysical technologies (e.g., biosensors and nanosensors) for environmental contamination factors. According to the different levels of sophistication of various analytical techniques and the information they can provide (either fine fingerprint, highly accurate quantification, semiquantification, qualitative detection, or fast screening), we will comment on how they may be used as complementary tools, as well as their potential and gaps toward current demand of real-time, online, and/or onsite detection.
Full-text available
Abstract The present monitoring and assessment of the chemical status of water bodies fail to characterize the likelihood that complex mixtures of chemicals affect water quality. The European Collaborative Project SOLUTIONS suggests that this likelihood can be estimated with effect-based methods (EBMs) complemented by chemical screening and/or impact modeling. These methods should be used to identify the causes of impacted water quality and to develop programs of measures to improve water quality. Along this line of reasoning, effect-based methods are recommended for Water Framework Directive (WFD) monitoring to cover the major modes of action in the universe of environmentally relevant chemicals so as to evaluate improvements of water quality upon implementing the measures. To this end, a minimum battery of bioassays has been recommended including short-term toxicity to algae, Daphnia and fish embryos complemented with in vitro and short-term in vivo tests on mode-of-action specific effects as proxies for long-term toxicity. The likelihood of adverse impacts can be established with effect-based trigger values, which differentiate good from poor water quality in close alignment with Environmental Quality Standards for individual chemicals, while taking into account mixture toxicity. The use of EBMs is suggested in the WFD as one avenue to establish the likelihood of adverse effects due to chemical pollution in European water systems. The present paper has been written as one component of a series of policy briefs to support decisions on water quality monitoring and management under the WFD.
Full-text available
Raman microspectroscopy displays a label-free and non-destructive modality providing highly specific information on single-cell level. Thus, it bears great potential as a standard tool in biomedical studies. We have developed a high-throughput screening Raman spectroscopy platform combining both automated imaging microscopy and spectra acquisition. The system allows for rapid screening of whole cell populations enabling differential white blood cell count and circulating tumor cell detection. © 2018, Springer-Verlag GmbH Deutschland, ein Teil von Springer Nature.
This handbook is an interdisciplinary and comprehensive reference covering all aspects of cell biosensors. It is divided into four main sections which are led and organized by numerous international experts. The scope of coverage includes: • Fundamentals and genetics for biosensor applications • Transducers, Materials and Systems • Markets, innovation and education • Application of biosensors in business Biosensor research is an exciting hybrid world where biologists, chemists, physicists, engineers and computer engineers come together. This handbook will serve as an invaluable living resource for all researchers in academia and industry working with cell biosensors.
This review covers advances in the analysis of air, water, plants, soils and geological materials by a range of atomic spectrometric techniques including atomic emission, absorption, fluorescence and mass spectrometry.
Gold and silver nanoparticles have unique optical properties. For instance, the intense colour of nanoparticle suspensions results from the excitation of a collective oscillation of surface conduction electrons, named surface plasmons. This excitation is done using an electromagnetic radiation that interacts with nanoparticles having a negative real and small positive imaginary dielectric constant, such as nanoparticles of gold or silver. The plasmonic optical properties of metal nanostructures are dependent on their shape and size, the dielectric properties of the metal and the surroundings and on the possible electromagnetic coupling with the localized surface plasmons in nearby other plasmonic objects. The other important consequence of the excitation of surface plasmons is a local significant enhancement of the electromagnetic field at some places of the illuminated nanoparticles. Specific plasmonic properties of gold and silver nanoparticles have allowed the development of many sensors for chemical analysis, including sensors dedicated for environmental analysis. Some of these sensors are so sensitive that recording of the reliable analytical signal of a single molecule is possible. Here, we review analytical techniques based on plasmonic properties of metallic nanoparticles for environmental analysis. We present the theory and mechanism of interaction of the electromagnetic radiation with the plasmonic nanoparticles. We detail analytical techniques including methods utilizing local enhancement of the intensity of the electromagnetic field induced by plasmons, and hence increase in the efficiency of some optical processes in the proximity of the plasmonic nanoparticles. Those techniques are surface-enhanced Raman scattering, surface-enhanced infrared absorption and metal-enhanced fluorescence and methods based on the change in the optical properties of plasmonic nanoparticles caused by the analyte-induced aggregation or by analyte-influenced growth or etching of plasmonic nanostructures. Environmental compounds include heavy metal cations, metallo-organic compounds, polycyclic aromatic hydrocarbons, pesticides, nitrite ions, bacterial cells and bacterial pathogens.
In Part I of the present review series, I presented the current state of the water environment by focusing on Japanese cases and discussed the need to further develop microbial biosensor technologies for the actual water environment. I comprehensively present trends after approximately 2010 in microbial biosensor development for the water environment. In the first section, after briefly summarizing historical studies, recent studies on microbial biosensor principles are introduced. In the second section, recent application studies for the water environment are also introduced. Finally, I conclude the present review series by describing the need to further develop microbial biosensor technologies. Current water pollution indirectly occurs by anthropogenic eutrophication (Part I). Recent trends in microbial biosensor development for water environment are described in part II of the present review series.