Experiment and Data Analysis
7
II-1. Confocal Raman micro-spectrometer
All Raman measurements presented in this thesis, including space- and time-resolved measurements and imaging experiments, were performed using a laboratory-built confocal Raman microspectrometer [3], [4] and its schematic illustration is shown in Figure II-1. The 632.8-nm output of a He-Ne laser (Thorlabs) with a beam diameter of 0.98 mm was used as the Raman excitation source. Our scheme is similar to that of any contemporary Raman spectrometer. First, excitation laser light passed through a laser line filter to clean up any spontaneous emission. Then, in order to reduce the spot size at the focal point, the laser beam was magnified by a factor of ~2.7 using a lens pair to effectively cover the exit pupil of the objective. Collimation of the laser beam was also achieved in this step. The expanded beam was then introduced to an inverted microscope (Nikon) by a pair of an edge filter (Semrock;
LP02-633RU-25) and a hot mirror (Thorlabs; FM02). The microscope was custom-made in collaboration with Nikon engineers by modifying a TE2000-U microscope. The beam was focused onto the sample by an oil-immersion objective (CFI Plan Fluor; 100×, NA = 1.3) placed on the microscope stage, and backward scattered light was collected by the same objective. The backward scattered light was guided along the opposite direction to the incoming path. Rayleigh and anti-stokes Raman scattered light was rejected at the edge filter and only Stokes Raman scattered light was transmitted. The Raman scattered light was then focused onto a 100-μm pinhole by a 150-mm lens and collimated by another 150-mm lens.
With this confocal configuration, a spatial resolution of about 0.3 m in lateral direction and 2.7 m in axial direction were achieved (see below). The Raman scattered light was dispersed by an imaging spectrometer (HORIBA Scientific; iHR320) and detected by a back-illuminated, deep-depletion, liquid-nitrogen cooled CCD detector (Princeton Instruments;
8
Spec-10:100B) with 100 × 1340 pixels operating at −120 oC. A 600 grooves/mm grating was used. The resulting spectral resolution was 7 cm−1, which was high enough in this work because the Raman spectra of biological samples usually exhibit relatively broad Raman bands. In addition, this grating can cover a wide spectral range over the fingerprint region (>2000 cm−1).
For bright-field observation, the sample was illuminated by a halogen lamp (or a mercury lamp) and optical micrographs were acquired by a digital camera (Nikon; DS-Ri1) mounted on a side port of the microscope.
For imaging experiments, the laboratory-built confocal Raman microspectrometer was also equipped with a high-precision piezoelectric nanopositioning stage (PI; P-563.3CD). A LabVIEW (National Instruments) program was run to automatically control the piezoelectric stage such that Raman imaging experiments were performed by translating the sample both horizontally and vertically. In the present work, the sample was translated with a 0.5-μm step
Figure II-1. Schematic representation of our laboratory-built confocal Raman micro-spectrometer.
9
in both X and Y directions. Because these steps were greater than the estimated lateral resolution (0.32 μm), they determined the effective spatial resolution in our experiments.
II-2. Estimation of lateral and axial resolutions
In order to evaluate the performance of our confocal Raman micro-spectrometer, it is crucial to know the resolution in both lateral (XY) and axial (Z) directions. First, to estimate lateral resolution, the intensity of the first-order phonon band of silicon at 520 cm-1 was measured while the laser spot was scanned horizontally, with 100 nm step size using the piezostage, through a sharp edge of a silicon wafer (a few µm inside the Si wafer through the edge until a few µm away from the wafer). The intensity of the silicon band will be strong when the laser hits inside the Si wafer, but as the laser spot approaches the edge, the intensity will drop rapidly and become negligible outside. Assuming that the edge of the Si wafer is infinitely sharp [i.e., described by a step function H(x)] and that the laser beam profile is a Gaussian function g(x) with its full width at half-maximum (FWHM) being the lateral resolution, the measured intensity profile can be approximated by a convolution of H(x) and g(x). Here x is the scanned distance in X direction. By fitting the experimentally obtained intensity profile with Equation II-1 [5], the lateral resolution can be estimated.
𝑓(𝑥, 𝑁, 𝑐, σ) = 𝑁 ∫ 𝐻(𝑎′)𝑔(𝑎′− 𝑎)𝑑𝑎′ =𝑁
2(1 + 𝑒𝑟𝑓 (𝑥 − 𝑎
√2𝜎 )) + 𝑐 (II − 1)
∞
−∞
Here, N is a normalization constant, a is the onset of the step function, is the width of the Gaussian function, c is a constant, and erf denotes the error function. The lateral resolution is equal to 2√2 ln 2 σ. Similarly, to estimate axial resolution, the intensity profile of a Raman band of cyclohexane at 801 cm-1 measured at a solvent glass interface was used. Raman intensity profiles as a function of the scanned distance in the X and Z directions are shown in
10
Figure II-2 (a) for lateral resolution and in Figure II-2 (b) for axial resolution, respectively.
The fitted results show that the lateral resolution is 0.32 (± 0.04) m and axial resolution is 2.7 (± 0.1) m in the axial direction. The errors here represent fitting precision.
II-3. Singular value decomposition analysis
From our own experience and literature [6], we understand that cell viability decreases with increasing excitation laser power and exposure time. Since we want to see the dynamics in single cells and other biological samples that often take long time (e.g., days or even longer), it is very important not to disturb the regular functioning of the cell. So, in order to provide a better growing environment for the cells, we need to use sufficiently low laser power (1 mW) with short exposure time (typically 1 to 2 s). But the resulting disadvantage is that the Raman signal will be reduced drastically under such low experimental conditions, resulting in poor signal-to-noise (S/N) ratio spectra and low Raman image quality.
To circumvent this practical problem, we make use of a numerical post-treatment Figure II-2. Evaluation of (a) lateral and (b) axial resolution of our laboratory built confocal Raman micro-spectrometer. Observed Raman intensity in red and the best fit with the model function (Equation II-1 ) in blue.
(a) (b)
FWHM = 0.32±0.04 µm FWHM = 2.7±0.08 µm
11
method based on singular value decomposition (SVD). This approach has been used successfully by us [3], [4], [7] and several other researchers [8]-[10] as a spectral de-noising technique.
SVD is a mathematical technique that factorizes an arbitrary 𝑚 × 𝑛 matrix 𝑨 into the product of three matrices as 𝑨 = 𝑼𝑾𝑽𝑇.Geometrically, SVD decomposes 𝑨 into three simple
Figure II-3. Principle of SVD. The upper left shows the unit disc together with the two canonical unit vectors. The upper right shows the action of A on the unit disc: it distorts the circle to an ellipse. The SVD decomposes A into three simple transformations: a rotation VT, a scaling W along the coordinate axes and a second rotation U. The SVD gives σ1 and σ2 which are just the singular values that occur as diagonal elements of the scaling W.
12
transformations: a rotation 𝑽𝑇, a scaling 𝑾 along the rotated coordinate axes and a second rotation 𝑼 as shown in Figure II-3. The SVD reveals the lengths σ1 and σ2 of the semi-major and semi-minor axes of the ellipse which are just the singular values that occur as diagonal elements of the scaling W. The rotation of the ellipse with respect to the coordinate axes is given by U.
Mathematically, 𝑼 is an 𝑚 × 𝑛 column-orthonormal matrix (also called left singular values), 𝑾 an 𝑛 × 𝑛 diagonal matrix of positive singular values, and 𝑽 an 𝑛 × 𝑛 orthonormal
matrix (right singular matrix). 𝑼 and 𝑽 represent the spectral and positional matrices, respectively as shown in Figure II-4. Only components of 𝑼 and 𝑽 having significantly large singular values were retained to reproduce matrix 𝑨, because other components with much smaller singular values contributed to the original data negligibly and can be regarded as noises.
The matrix 𝑨 was then reconstructed by using the components of 𝑼 and 𝑽 associated with large singular values. The number of singular values retained in this reconstruction was typically in
Figure II-4. Graphical representation of SVD. Raw data matrix A can be factorized into a product of three matrices U (contains spectral information), W (contains singular values) and VT(spatial information of each spectral component)
13
between 10-15. The main criterion for determining how many components are taken into account was whether or not the spectral component of 𝑼 corresponding to a particular singular value shows definite Raman features. The SVD was computed in Igor Pro (WaveMetrics) using LAPACK routines. Figure II-5 illustrates how well SVD denoising works. Space-resolved Raman spectra of a fission yeast cell are compared before and after the SVD. It is clear that the SVD reduces noises dramatically. As compared in Figure II-5 (b) and (c), the Raman image constructed for the 1602 cm-1 band from the raw data (without SVD) is featureless due to noisy spectra, but the image constructed from the SVD-analyzed data shows high contrast.
14
(a) (b) (c)
(d) before SVD
after SVD
Figure II-5. (a) Optical image of a typical fission yeast cell; (b,c) Raman images constructed for 1602 cm-1 with raw data before SVD (b) and after SVD analysis (c); and (d) representative space-resolved Raman spectra from the same data which were acquired with an exposure time of 1.5 s and 1 mW laser power. The SVD-treated spectra (blue) exhibit much higher S/N than the corresponding raw data (red).
15
Chapter III
Detection of Leucine Pools in
Escherischia coli Biofilm
16
III-1. Introduction
Ever since Anton Van Leeuwenhoek reported the discovery of microorganisms in 1676, they have been characterized primarily as planktonic, freely floating cells. Following the discovery, their growth characteristics were studied extensively in nutritionally rich culture media to describe them. Undoubtedly, these studies have been highly informative and have advanced our knowledge on the biology of bacteria. However, in the last four decades, the discovery of many microbiological phenomena that surface-associated microbes exhibit distinct phenotypes from that of their planktonic counterparts aroused significant attention among the scientific community to understand the microbial life in the naturally occurring forms known as biofilms.
There were some observations as early as 1940 that introducing a surface enhances bacterial growth and activity, but the first detailed study of biofilm was reported by Jones et.al. in 1969 [11]. They found that biofilms contain a variety of microorganisms by examining biofilms on trickling filters in a wastewater treatment plant by electron microscopy. Additionally they also confirmed for the first time that the extracellular matrix was mostly composed of polysaccharides. Later in 1973, studies on microbial slimes from industrial water systems by Characklis showed that biofilms are resistant to disinfectants such as chlorine and are tightly bound to the surface [12]. In 1978, from the studies on microorganisms in dental plaque and attached communities in mountain streams, Costerton et.al. proposed that microbes adherent to the living or non-living surfaces gain special characters depending on the environmental conditions [13]. Since then biofilms became a topic of interest among researchers across various disciplines.
17
III-2. What are biofilms?
Bacteria in nature usually live most of their life as members of a ‘socialized’ community called a biofilm rather than as free-living cells. A biofilm is a structured consortium of sessile bacterial cells irreversibly associated on a biotic or an abiotic surface. In a biofilm, cells are encased in a matrix of self-secreted extracellular polymeric substances (EPS) primarily composed of polysaccharides along with proteins and DNA. Although traditional microbiology has focused on planktonic bacteria, it is bacterial cells adopting the biofilm lifestyle that play an essential role in various processes such as bacterial infections [14], [15], wastewater treatment [16], and bioremediation [17]. Indeed, a public announcement by the US National Institutes of Health reads ‘‘Biofilms are medically important, accounting for over 80% of microbial infections in the body”.
Biofilm bacteria are known to show exceedingly high resistance to antimicrobial agents.
Furthermore some bacterial species in their biofilm mode of growth can communicate with one another via chemical signalling in the way that the cell density inside microcolonies (cell clusters) in the biofilm regulates cell’s transcriptional processes (quorum sensing). It has been shown that hormone-like signalling molecules termed autoinducers are responsible for gene regulation and resultant biofilm formation. In fact, bacterial biofilms can function as if they were tissues of a higher organism. In all these processes, various biomolecules serve as chemical signals, regulators, and structural components [18]. For example, in many species including Pseudomonas aeruginosa [19], [20] and Staphylococcus aureus [21], extracellular DNA has been shown to be required for initial biofilm formation.
Recently it has been reported that some D-amino acids trigger biofilm disassembly in Bacillus subtilis and other bacteria [22]. All those characteristics, which make bacterial cells in the biofilm milieu unique and quite different from their planktonic counterparts, are closely
18
associated with the structure of biofilms and molecular distributions therein.
III-3. Developmental stages in biofilms
Biofilm development is a complex process involving several stages, each with unique characteristics. It can be divided into five stages [23], [24] (Figure III-1).
1. Initial attachment: reversible attachment of cells to the surface
2. Irreversible attachment: production of EPS, resulting in more firmly adhered irreversible attachment of cells
3. Maturation I: early development of biofilm architecture in which active recruitment of cells from the bulk fluid takes place, and formation of microcolonies
4. Maturation II: steady increase in size, which leads to formation of mature cell clusters 5. Dispersion: dispersal of cells from mature colonies to form new microcolonies
Figure III-1. Developmental stages of a biofilm
Source: Monroe D. Looking for Chinks in the Armor of Bacterial Biofilms.
PLoS Biol 5(11), 2007
19
III-4. General methods to study biofilms
In the early days, most of the research on biofilms relied on techniques such as electron microscopy and standard biological characterization protocols. Later, to understand the biofilm architecture and distributions of biofilm constituents in a non-destructive manner, a variety of optical microscopic methods have been developed and applied. Most commonly used technique for biofilm studies is confocal laser scanning microscopy (CLSM). This method is based on detecting fluorescence from a dye probe introduced to the sample, allowing for three-dimensional (3D) in situ visualization of the biofilm structure with high sensitivity and sub-µm resolution. CLSM studies have revealed that, despite the name of biofilm, the biofilm structure is far more complicated than what one would imagine from a film: a biofilm often consists of microcolonies and channels through which water and other fluids can flow [25]. In CLSM, however, staining cells in a biofilm with an appropriate fluorescence dye is a prerequisite for imaging. Thus, as yet unknown substances can never be explored by this method, because relevant fluorescent probes for the target are not known as well. CLSM (fluorescence imaging) can provide space-resolved information that the conventional biochemical assays often lacks, but it has only limited access to the information on molecular structures and microenvironments in biofilms.
As a first step to fully understand how the biomolecules in biofilms, regardless of whether they are already identified or not, fulfill their advanced functions, we use a label-free Raman imaging method to study the constituents of an Escherichia coli (E.coli) biofilm in situ and to visualize their distributions in the biofilm.
20
III-5. Escherichia coli
E. coli was the bacterium of choice for our demonstration, a typical bacterium suited for
a first step of systematic Raman studies, since it is the most widely studied prokaryotic model organism. It is a gram-negative, facultative anaerobic, rod shaped bacterium. Most of the strains
are harmless and are part of normal flora of the gut preventing the establishment of pathogenic bacteria within the intestine. Typically they are 2 µm in length and 0.5 µm in diameter as can be seen from Figure III-2.
III-6. Raman spectrum of a single planktonic E. coli
A laboratory strain of E.coli (XL1-blue) was routinely cultured in LB-Miller medium Figure III-2. Scanning electron micrograph of E.coli
grown in culture medium and adhered to a cover slip Source: Rocky Mountain Laboratories, NIAID, NIH
Table III-1. Composition of LB-Miller medium
21
at 37 ºC, whose composition is given in Table III-1. As a preliminary step, the Raman spectrum of a single planktonic E. coli was measured by laser trapping (632.8 nm for both trapping and excitation) with a minimal laser power of 3 mW at the sample point. The planktonic Raman
spectrum (Figure III-3) shows several bands characteristic of lipids, proteins and nucleic acids.
Main features are the amide I (mainly C=O stretch of amide bonds) band at 1660 cm−1, the CH bending of the aliphatic chain at 1342 and 1457 cm−1, and the breathing mode of the phenylalanine residue at 1004 cm−1. Nucleic acid bands at 784 and 1096 cm−1 attributable to the (deoxy) ribose-linked phosphodiester backbone vibrations of RNA (DNA) were observed.
Cytosine could contribute to the 784 cm−1 band as well. The doublet observed at 815 and 857 cm−1 is known as the tyrosine doublet arising from the Fermi resonance of a ring-breathing vibration and the overtone of an out-of-plane ring-bending vibration of tyrosyl residues [26].
The bands at 670 and 725 cm−1 are due to nucleobases guanine and adenine, respectively. The assignments [27]-[30] are summarized in Table III-2.
Figure III-3. Typical Raman spectrum of a single E. coli bacterium, with 632.8 nm excitation, optically trapped at the focal point. Raman bands characteristic of proteins P, lipids L and nucleic acids NA are observed. (Laser power, 3 mW; exposure, 100s)
22
Table III-2. Tentative assignments of Raman bands observed in planktonic E.coli spectrum
23
III-7. Space-resolved Raman spectra of E. coli biofilms
E.coli biofilms were grown on a glass-bottomed dish at 37 ºC under static conditions
for 30 hours. The static condition here meant that, in the biofilm growth and subsequent Raman measurements, we did not employ any continuous-flow system in which the biofilm would always be subject to hydrodynamic shear force. Since E. coli biofilms are generally known to be fragile, we did not wash the biofilm sample. Instead, to remove the planktonic population and suppress fluorescence, we gently pipetted excess LB medium out from the edges of the dish. Neither cell staining nor immersing the biofilm in a suspension of metal nanoparticles [31], [32] to enhance Raman signals was required. Thus, our method can be said to look at the target with minimum external perturbation.
The sample was then transferred directly to the microscope stage for Raman measurements without any further pre-treatment. To obtain space-resolved Raman spectra of the 30-hour old E. coli biofilm, we fixed the focal plane of the excitation laser beam at approximately 3 µm above the substrate and collected Raman scattered light from selected locations inside/outside the microcolony on the plane. To avoid possible photo bleaching and heat accumulation, we used laser power of 3 mW at the sample point throughout the present study. For reference, the Raman spectrum of a planktonic E. coli cell was also measured using optical trapping. A 60 s exposure time for the Raman spectra of the biofilm and 100 s for that of the planktonic cell were used.
Almost everywhere in the unwashed E. coli biofilm, a number of microcolonies were found as shown in Figure III-4. Their sizes were diverse, ranging from ~5 µm to several tens
24
of micrometers wide. The morphology varied as well from one microcolony to another. Figure III-5 (a) shows one of the colonies on which further measurements were done, while Figure III-5 (b)–(d) compares the space-resolved Raman spectra, for the spectral range 400–1800 cm−1, recorded inside and outside the microcolony at ∼ 3 μm above the substrate with that of a planktonic E. coli cell measured independently using optical trapping. The extracolonial spectrum was almost identical to the planktonic Raman spectrum with few exceptions including the Raman band at 1408 cm−1. This band is due probably to the COO− symmetric stretch of polysaccharides such as alginates which are a major constituent of the extracellular matrix[25], [31].
25
The extracolonial and planktonic spectra represent a typical cellular Raman spectrum dominated by proteins and DNA/RNA bands. These results, along with the optical micrograph [Figure III-5. (a)], show that the intercolonial space of the biofilm was filled with a dense population of E. coli cells held together by the extracellular matrix.
Figure III-4. Optical micrograph of a 30-hour old biofilm.
Red arrows indicate microcolonies formed in the biofilm.
Scale bar = 10 µm.
26
Figure III-5. (a) Optical micrograph of a single microcolony in 30 h old biofilm on which further experiments were done. Space-resolved Raman spectra of the 30-h old E.coli biofilm (b) intracolonial, (c) extracolonial and (d) a planktonic E.coli cell.
Stars denote the positions at which (b) and (c) are recorded.
(a)
*
*
27
The intracolonial spectrum differs substantially from the extracolonial spectrum, indicating distinct chemical composition of the microcolony. By scrutinizing Raman spectra of biomolecules that could potentially occur in the microcolony, we found that the intracolonial
The intracolonial spectrum differs substantially from the extracolonial spectrum, indicating distinct chemical composition of the microcolony. By scrutinizing Raman spectra of biomolecules that could potentially occur in the microcolony, we found that the intracolonial