Subscriber access provided by MT ROYAL COLLEGE
Article
Multispectral Atomic Force Microscopy-Infrared Nano-Imaging of Malaria Infected Red Blood Cells David Perez-Guaita, Kamila Kochan, Mitchell Batty, Christian Doerig, Jose Garcia-Bustos, Shirley Josefina Espinoza Herrera, Don McNaughton, Philip Heraud, and Bayden R. Wood Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b04318 • Publication Date (Web): 12 Jan 2018 Downloaded from http://pubs.acs.org on February 6, 2018
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
1 2
Hyperspectral Multispectral Atomic Force Microscopy-Infrared Nano-Imaging of Malaria Infected Red Blood Cells
3 4
David Perez-Guaitaa, Kamila Kochana, Mitchell Batty b, Christian Doerigb, Jose Garcia-Bustosb, Shirley Josefina Espinoza Herrera c, Don McNaughtona, Phil Heraud*a,b, Bayden R. Wood*a
5 6 7 8 9 10 11
a
Centre for Biospectroscopy, Monash University, Clayton, 3800, Victoria, Australia b Department of Microbiology and Infection & Immunity Program Monash Biomedicine Discovery Institute, Monash University, Clayton, 3800, Victoria, Australia c ELI Beamlines. Institute of Physics. Czech Academy of Science. Na Slovance 2, 18221 Prague, Czech Republic *
[email protected],
[email protected] 12
Abstract
13
Atomic Force Microscopy-Infrared (AFM-IR) spectroscopy is a powerful new technique that can be
14
applied to study molecular composition of cells and tissues at the nano-scale. One limitation of the
15
technology is that the maps generated are univariate in natureAFM-IR maps are generally acquired
16
using over one single wavenumber value: they show either the absorbance plotted against a single
17
wavenumber value, or a ratio of two absorbance values. Here we implement for the first time
18
hyperspectral multivariate image analysis to generate multicomponent multivariate AFM-IR maps,
19
and use this approach to resolve subcellular structural information in red blood cells infected with
20
Plasmodium falciparum at different stages of development. This was achieved by converting the
21
discrete spectral points into a hyperspectral multispectral line spectrum prior to multivariate image
22
reconstruction. The approach was used to generate compositional maps of subcellular structures in
23
the parasites, including the food vacuole, lipid inclusions and the nucleus, based on the intensity of
24
haemozoin, haemoglobin, lipid and DNA IR marker bands, respectively. Confocal Raman
25
sSpectroscopy was used to validated the presence of haHemozoin in the regions identified by the
26
AFM-IR technique. The high spatial resolution of AFM-IR combined with hyperspectral modelling
27
enables the direct detection of subcellular components, without the need for cell sectioning or
28
immunological/biochemical staining. HyperspectralMultispectral-AFM-IR thus has the capacity to
29
probe the phenotype of the malaria parasite during its intra-erythrocytic development. This enables
1
ACS Paragon Plus Environment
Formatted
Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
1
novel approaches to studying the mode of action of antimalarial drugs and the phenotypes of drug-
2
resistant parasites, thus contributing to the development of diagnostic and control measures.
3
Keywords: AFM-IR, RBC, Plasmodium, nanoimaging, HyperspectralMultispectral analysis.
4
5
2
ACS Paragon Plus Environment
Page 2 of 30
Page 3 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
1
INTRODUCTION
2
The study of intracellular structures is an important aspect of biology that requires the use of
3
techniques with submicron resolution. To date, there are only a fewis not a wide range of available
4
techniques that can achieve this level of resolution have been limited tosuch as fluorescence and
5
transmission electron microscopy (TEM). The former requires the use of fluorescence stains
6
(sometimes coupled to antibodies) and can be used with fixed or more rarely with live cells, while
7
the latter can only be applied to cell sections. Other emerging techniques include scanning Near-
8
field Optical Microscopy (s-SNOM) (s-SNOM), Atomic Force Microscopy (AFM) or nNanoscale
9
Ssecondary Iion Mmass Sspectroscopy (nano-SIMS). Fourier transform Infrared (FTIR)
10
hyperspectral imaging has been developed over the recent decades to investigate the composition of
11
biological samples
12
each IR spectrum (a set of 15-25 discrete wavenumber valuess in the case of multispectral imaging)
13
has a corresponding unique “x, y” coordinate, providing spatial and chemical information
14
simultaneously. Sophisticated data treatment procedures have been introduced to generate chemical
15
maps based on spectral variance, enabling compositional studies of biological tissues3. The main
16
advantages of FTIR hyperspectral imaging is that it provides unbiased chemical information,
17
including: i) quantitative distribution of biological molecules such as DNA, carbohydrates, proteins,
18
lipids and, in the case of Plasmodium falciparum, haemozoin (Hz)5; ii) higher order data such as
19
protein secondary structure 6, nucleic acid conformations
20
main drawback is that the long wavelength of the IR photons (3-8 µm) results in diffraction-limited
21
images with coarse spatial resolution, unsuitable for the study of features smaller than a few
22
microns9.
23
utilising a ×25 objective with high Numerical Aperture (NA=0.81) and a focal plane array IR
24
imaging microscope, we were previously able to image the food vacuole (FV) in a single malaria
25
infected red blood cell (RBC) based on the lower concentration of organic matter in the FV but no
26
additional detail within the cell was distinguishable.10
1–4
. In an FTIR hyperspectral image captured using a focal plane array detector,
7
or the oxidation state of lipids8. The
This makes resolution of all but the largest sub-cellular structures impossible. By
3
ACS Paragon Plus Environment
Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
1
More recently, IR has been coupled with atomic force microscopy, leading to AFM-FTIR
2
microscopes able to acquire IR spectra with submicron resolution11. Scanning near-field optical
3
microscopy (s-SNOM) s-SNOM and photo-thermal induced resonance (PTIR) have also been
4
coupled to AFM systems12. The latter uses pulses from a tuneable IR laser, with either an Optical
5
Parametric Oscillator (OPO) system or a Quantum Cascade Laser (QCL). Laser energy at
6
wavenumber values corresponding to absorbance bands in the sample is absorbed and causes rapid
7
thermal expansion, much faster than the AFM feedback, exciting resonant oscillations in the AFM
8
cantilever13. A spectrum can thus be obtained by measuring the amplitude cantilever oscillation,
9
which is proportional to the absorption coefficient of the sample at different wavenumber values.
10
This technique has opened the door to nano-analysis of biological samples using mid-IR
11
spectroscopy with resolutions below the diffraction limit 14. The high spatial resolution of AFM-IR
12
has allowed the technique to be applied to bacteria15–17, as well as to larger size samples such as
13
human hair18 and skin19 . Studies on eukaryotic cells have focused so far on mapping nanoparticles
14
within cells. For example, AFM-IR has been applied to map the localization of gold nanoparticles
15
within adenocarcinoma cells20 and to study nano-diamond-induced alterations in cellular proteins21.
16
To date, no AFM-IR studies have investigated malaria parasites. We have previously applied AFM-
17
Raman, known as Tip Enhanced Raman Scattering (TERS), to investigate Hz crystals protruding
18
from the surface of sectioned red blood cells infected with Plasmodium falciparum trophozoites22.
19
For that application the cell had to be embedded in resin, sectioned, and spectra could only be
20
recorded from crystals protruding from the sectioned cell, consequently the approach did not reveal
21
any additional structural information besides the haemozoin crystal location.
22
Previous AFM-IR studies have mainly focussed on using measuring the absorbance at one
23
given wavenumber value, or on obtaining the absorbance ratio from two wavenumber values of
24
interesta combination of point spectral measurements with single wavelength imaging. Other s-
25
SNOM based studies have used multispectral and hyperspectral imaging in other applications such
26
as intramolecular dynamics23 and polymer blends and hair compositional studies24. In this study we 4
ACS Paragon Plus Environment
Page 4 of 30
Page 5 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
1
apply IR hyperspectral imagingmultivariate imaging applied to cells, which is normally applied to
2
spectra spanning the full mid-IR range (4000-800 cm-1) and mainly used when analysing focal plane
3
array data from tissue sections3. Figure 1 shows a schematic diagram of our approach to generate
4
the images. First, we use the combination of AFM and IR for measuring the IR signal maps at
5
different
6
hyperspectralmultispectral images (1b). Finally, the image is analysed using multivariate data
7
analysis. The results show that the lateral resolution of the technique along with the multivariate
8
analytical approach enable chemical identification of a range of sub-cellular Plasmodium sp.
9
structures, such as the FV, lipid inclusions and the nucleus, providing information about the
10
chemical composition of the parasite at submicron resolution without staining or modification of
11
cellular chemistry beyond the initial fixation.
12
METHODS
13
P. falciparum culturing and staining
14
P. falciparum 3D7 parasites were cultured in O+ erythrocytes at 4% haematocrit in RPMI 1640 - HEPES
15
medium (Thermo Fisher), supplemented with 0.5% Albumax II (Thermo Fisher), 50mg/L hypoxanthine
16
(Sigma), 10mg/L gentamycin (Pfizer) and 25mM sodium bicarbonate (Sigma) as previously described23.
17
Cultures were gassed with 1% O2, 5% CO2 and 94% N2. Ring stage parasites were synchronised by
18
incubating with 10ml of 5% D-sorbitol (Sigma) for 10 minutes at 37°C and lifecycle stage determined for
19
each experiment using Giemsa staining. After AFM-IR and Raman analysis, parasites were stained for
20
observation by light microscopy by fixing thin blood smears in 100% methanol for 10 seconds, followed by
21
staining with 10% (v/v) Giemsa solution for 5 minutes and imaged using an Olympus BX51 microscope
22
coupled with an Olympus DP70 digital camera.P. falciparum 3D7 parasites were cultured in O+ erythrocytes
23
at 4% haematocrit in RPMI 1640 - HEPES medium (Thermo Fisher), supplemented with 0.5% Albumax II
24
(Thermo Fisher), 50mg/L hypoxanthine (Sigma), 10mg/L gentamycin (Pfizer) and 25mM sodium
25
bicarbonate (Sigma) as previously described23. Cultures were gassed with 1% O2, 4% CO2 and 95% N2. After
26
AFM-IR and Raman analysis, parasites were stained for observation by light microscopy by fixing thin
wavenumber
values
(1a).
Then
the
maps
are
registered
to
5
ACS Paragon Plus Environment
generate
the
Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
1
blood smears in 100% methanol for 10 seconds, followed by staining with 10% (v/v) Giemsa solution for 5
2
minutes and imaged using an Olympus BX51 microscope coupled with an Olympus DP70 digital camera.
3
Page 6 of 30
Formatted: Line spacing: Double
4
Sample Preparation
5
1 microliter of packed RBC, which contained 5% parasitemia was deposited in a 10x2 mm2 window
6
purchased from Crystan Limted (Dorset, UK). Because we needed to examine the samples using Raman
7
spectroscopy after the AFM-IR analysis, only windows with Raman grade purity were used. The blood
8
sample was smeared using a standard glass coverslip and the resulting monolayer of RBCs was dried and
9
fixed using 100% methanol for 10 seconds. The CaF2 window was then mounted to a steel holder.
10 11
Calibration and optimization of the AFM-IR signal
12
In order to daily calibrate the IR signal and compensate the contribution of atmospheric gases, a background
13
was acquired averaging 3 spectra obtained in the spectral ranges 3050-2750 and 1800-800 cm-1, using a
14
spectral resolution of 4 cm-1 and co-averaging 2048 values per point. Prior to each measurement, the second
15
cantilever oscillation mode was used for the optimization of the cantilever frequency, which was normally
16
located at 17000 Hz ± 60. Then, for maximizing the IR signal, the tip was placed in contact with an
17
uninfected RBC and the IR laser position on the sample was optimised. The signal was monitored as a
18
function of the laser position, establishing the optimum position for a representative set of wavenumber
19
values across the spectra (e.g. 2914, 1640 and 1208 cm-1). A function that relates the position of the laser to
20
the wavenumber value was computed using the measured optimum position of the representative
21
wavenumber values. The function was used for interpolating the optimal position of the other wavenumbers.
22 23
Analysis of RBCs infected with P. Falciparum with AFM-IR
24
AFM-IR analysis was carried out using a nanoIR2 instrument from Anasys (Santa Barbara, USA) controlled
25
by the Analysis Studio software also from Anasys. An OPO laser was used as the IR emission source and
26
standard contact mode NIR2 probes provided by Anasys. First, topography images were recorded at 1Hz rate 6
ACS Paragon Plus Environment
Page 7 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
1
revealing spatcial features corresponding to the presence of the parasite. Then, the tip was placed in the
2
regions of interest and spectra were recorded within the regions 3050-2750 and 1800-800 cm-1 by co-
3
averaging 2048 scans at a spectral resolution of 4 cm-1. The laser power was selected by considering a trade-
4
off value between a reasonable signal-to-noise ratio and a safe power for not burning the sample. Because
5
each spectral region absorbed a variable amount of light, the laser power was tuned to the different spectral
6
regions always ranging between 5% and 45%. Wavenumber values with a large variance were selected and a
7
sequence of maps using the chosen wavenumber values was recorded. IR maps were created measuring 200
8
horizontal lines of 200 points each giving simultaneously 4 × 200 pixel by 200 pixel maps containing the IR
9
peak, IR amplitude, deflection and height values at each pixel. Considering the 1000 Hz frequency of the
10
OPO laser, the scanning rate was adjusted to 0.07 Hz, which allowed the accumulation of 64 scans for each
11
point. The measurement time for each map was approximatelyround 47 minutes. Further information about
12
calibration and optimization of the signal is available in the Section B of Supporting Material.
13
Analysis of standards with AFM-IR
14
Haemozoin was obtained from Invivogen (San Diego, USA). Human haemoglobin (98%) and DNA from
15
Calf thymus was obtained from Sigma Aldrich (St Louis, USA). 5 microliters of a suspension of each
16
standard in MiliQ water was spread onto a CaF2 window and allowed to dry. Spectra were acquired in the
17
same conditions as the RBCs.
18 19
Data Analysis
20
Data analysis was performed using Matlab from Mathworks (Natick, USA), using in-house written scripts
21
based on functions available in the Imaging processing toolbox from Matlab and in PLS-toolbox from
Formatted: Font: Italic
22
eigenvector Eigenvector Researchresearch (Manson, USA). The set of maps were registered to one of the
Formatted: Font: (Default) Times New Roman, 11 pt, Font color: Text 1, Pattern: Clear
23
maps obtained in the middle of the sequence in order to minimize the effect of the drift. This map was
24
considered as the reference map. In short, the drift was calculated comparing the deflection maps
25
corresponding to the IR map to register to the deflection map of the reference IR map and then this drift was
26
used to co-register the IR map. The Matlab function imregtform, available in at the image processing
27
toolbox was employed for the registering the maps. All the registered images were integrated in a 3D matrix. 7
ACS Paragon Plus Environment
Formatted: Font: (Default) Times New Roman, 11 pt, Font color: Text 1, Pattern: Clear Formatted: Font: (Default) Times New Roman, 11 pt, Font color: Text 1, Pattern: Clear Formatted: Font: (Default) Times New Roman, 11 pt, Font color: Text 1, Pattern: Clear
Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
1
The hierarchical cluster analysis was performed using a Standard Normal Variate (by dividing each sample
2
vector by the standard deviation of the sample) and autoscaling. For figures 3e, 4eh and 5f, the average
3
variables of each class where computed after normalizing the signal of each pixel (Sum=1). For each variable,
4
the average value of each class was divided by the average value of all the intensity values.
5
Raman Spectroscopy
6
Raman maps were acquired with the use of Confocal Raman Microscope (WITec alpha300 R, Melbourne,
7
Australia). The system is equipped with a dry Olympus MPLAN (10×/0.90NA) objective, an air–cooled
8
solid–state laser, 600 grooves/mm grating (BLZ = 500 nm) and a back – illuminated CCD camera, cooled to
9
–60°C. The laser was operating at 532 nm and was coupled to the spectrometer by an optical fibre with a
10
diameter of 50 µ m. The laser power was adjusted individually to each sample and did not exceed 6 mW. A
11
standard single point calibration procedure based on Raman scattering line of silica plate (520.5 cm-1) was
12
performed before data collection. Spectra were collected with spectral resolution of from 3 cm-1 to 5 cm-1, in -1
13
the range of 0 – 3705 cm . The integration time of single spectrum was 0.05 – 0.1 s and the size of imaged
14
areas was 10 × 10 µm (30 × 30 points, sampling density: 0.33 µm). All measurements were controlled by
15
Witec Control Software.
16
17
RESULTS AND DISCUSSION
18
AFM-IR location of Hz inside RBCs infected with P. falciparum.
19
We imaged human RBCs infected with P. falciparum parasites at different stages (rings,
20
trophozoites and schizonts) fixed in methanol. These are the blood stages relevant for malaria
21
pathology, named for their morphology and listed in the temporal order of development inside red
22
blood cells. Immediately after cell invasion the parasite adopts a characteristic ring morphology.
23
This is followed hours later by the irregularly shaped trophozoite stage, responsible for
24
haemoglobin (Hb) degradation and the deposition of liberated haem groups as crystalline
25
haemozoin (Hz). Trophozoites are the most biochemically active stage and they completely remodel
26
the red blood cell architecture. Finally, the single-celled parasite undergoes multiple rounds of 8
ACS Paragon Plus Environment
Page 8 of 30
Formatted: Not Superscript/ Subscript
Page 9 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
1
nuclear divisions transforming itself into a schizont, which precedes an atypical form of cell
2
division yielding 8 to 32 daughter parasites that rupture the cell to reinitiate the cycle. Figure 2a
3
shows the topography of a fixed RBC infected with a trophozoite on a CaF2 window, obtained using
4
AFM. The centre of the infected RBC presents an irregular topography compared with the
5
homogenous cell membrane observed in uninfected RBCs. These irregularities have already been
6
characterised by AFM in a previous study2624. A semi-spherical protrusion in the centre of the
7
infected red blood cell is shown in Figure 1a and is marked with an arrow. The protrusion measured
8
300 nm in height and 650 nm in radius. In order to study the chemical composition of this
9
protrusion, 15 spectra in the 1800-800 cm-1 and 3050-2750 cm-1 regions were obtained from inside
10
the protrusion and compared with 15 spectra obtained in parts of the cell not showing evidence of
11
the parasite along with spectra of uninfected RBCs adjacent to the infected cell (see points marked
12
blue and red in the deflection AFM image shown in Figure 2b). Figures 2c and 2d show AFM-IR
13
spectra acquired at different cellular locations along with AFM-IR spectra obtained from crystals of
14
Haemoglobin (Hb), Hz and DNA standards. The different band assignments used in this study are
15
summarized in Table 1. Clear spectral differences are apparent, with bands at 1708 cm-1 and 1207
16
cm-1 present in the protrusion (blue class) but not in the non-infected cellular locations (red class).
17
These bands are prominent also in the Hz reference spectrum and have been assigned as IR markers
18
of Hz, a by-product of Hb digestion by the parasite27–2925–27. The 1207 cm-1 band is assigned to the
19
C-O stretching mode of the propionate side chain of Hz, whilst the 1708 cm-1 band is assigned to a
20
carbonyl group2826. Hence the protrusion appears to be mainly Hz deposited inside the FV of the
Field Code Changed
21
parasite3028. The band at 1540 cm-1 has higher absorbance in the uninfected cell spectra and is also
Field Code Changed
22
prominent in the Hb standard. It is assigned to the amide II mode, indicating a higher concentration
23
of Hb outside the protrusion. An image of the absorbance in the 1207 cm-1 band (See Figure 2e)
24
indicates that Hz is present only in the protrusion. This is in contrast to the absorbance intensity
25
map at 1660 cm-1 (Figure 2f), which has contributions from both the carboxylate stretch of Hz and
26
the Amide I mode of the protein peptide group of Hb. The image in Figure 2f shows that uninfected 9
ACS Paragon Plus Environment
Field Code Changed
Field Code Changed
Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
1
cells have a homogenous distribution of protein based on the absorbance of the amide II mode. It
2
has to be noted that the intensity of the Amide II band is abnormally low compared with the
3
intensity of the Amide I band
4
observation was consistent in all the spectra acquired during the study and may be caused by
5
changes in the focus of the laser during the optimization process.
(see Figure 2d) in both, the spectra of RBCs and Hb. This
6
Principal Component Analysis (PCA) was performed on the CH stretching (3050-2800 cm-1)
7
region to investigate lipid contributions, which have been previously shown to be associated with
8
Hz formation3129. Scores and loadings of the first PC explaining 47.34% of the variance are
9
depicted in Figure 2g and 2h, respectively. The box plot (Figure 2g) indicates that the score values
10
from spectra obtained inside the FV were significantly higher than the ones obtained from regions
11
in the uninfected RBCs. Loading values describe the relationship between the PC and the IR
12
wavenumber variables (Figure 2h). Loading values show positive bands at 2854 cm-1 and 2914 cm-1,
13
which are assigned to νsCH2 and νasCH2 stretching vibrational modes. The negative bands at 2870
14
cm-1 and 2958 cm-1 are assigned to νsCH3 and νasCH3 stretching vibrational modes. This indicates
15
that the ratio CH2 to CH3 is higher inside the FV, which can be interpreted as a higher concentration
16
of aliphatic chains around Hz, presumably reflecting a lipid-rich FV environment when compared to
17
the uninfected RBCs and the cytoplasm surrounding the parasite in infected cells.
18
The spatial resolution of the measurement was carefully determined by acquiring spectra over
19
a line across the Hz deposit (Figure 3a). The line comprises of 10 spectra spanning a 2.2 µ m
20
distance, resulting in a separation of 220 nm between points. Progressive changes are observed in
21
the spectra along the line across the FV. Bands at 1207 cm-1 and 1300 cm-1 increased as the AFM
22
tip approached the FV and decreased as spectra were recorded further away. This demonstrates
23
submicron resolution of subcellular structures, because spectral differences could be clearly
24
observed between points located only 220 nm apart.
25
Page 10 of 30
HyperspectralMultispectral AFM-IR imaging of P. falciparum infected RBC. 10
ACS Paragon Plus Environment
Field Code Changed
Page 11 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
1
The chemical characterization of the FV using AFM-IR shows that more information can be
2
obtained compared to topographical AFM images of infected RBCs. Figure 4 shows AFM-IR
3
images of a trophozoite-infected RBC using 23 wavenumber absorbance variables. Absorption
4
maps at different wavenumber values were sequentially collected during several hours18 hours of
5
measurement. During this time a 2 µm drift was observed, precluding direct co-registration of the
6
IR maps. AFM topographic images collected simultaneously with the IR intensity images were used
7
to calculate the drift for each IR image, and subsequently used for co-registration. In this image,
8
each pixel contained a “spectrum” with 23 intensity values from non-continuously spaced
9
wavenumber values (See Supplementary Information SI-1). The wavenumber values of the
10
absorbances to be measured were selected according to two criteria: either they showed large
11
changes in preliminary spectra obtained across the cell; and/or they were well described markers of
12
specific biological constituents.
13
The AFM topographical map showed that the RBC (see Figure 4b and the superposition
14
image in 4d) was thicker along its perimeter (a well-known property of erythrocytes), at 400 nm in
15
height. The rest of the cell was only 200 nm in height except for the centre of cell, which contained
16
a mass of Hz with a height of 400 nm and two circular holes approximately 600 nm in diameter. In
17
order to study the relationship between the spectral variables and the pixels in the image, k-means
18
cluster analysis (KMCA) was employed. This is a standard procedure in the analysis of
19
hyperspectralmultispectral images that groups pixels according to their distance apart in the spectral
20
space32. The KMCA identified 5 classes related to the topographical features (see cluster image in
21
Figure 4c and the superposition image in Figure 4d). Classes coded orange and green are located at
22
the edge of the cell, a class coded violet is associated with the Hz crystal mass, a blue-coded class
23
appears in the centre of the cell, and pixels coded yellow are mainly found in or scattered around
24
the cavities.
25
The average class intensities of selected wavenumber values shown in figure 4e provides
26
information on the chemical composition of each cluster. Boxplots of the distribution of the 23 11
ACS Paragon Plus Environment
Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
1
variables measured for each class are available in the SI-2. Green and orange coded classes show
2
significant high intensity bands (p