Multispectral Atomic Force Microscopy-Infrared Nano-Imaging of

Jan 12, 2018 - Centre for Biospectroscopy, Monash University , Clayton , Victoria 3800 , Australia. ‡ Department of Microbiology and Infection & Imm...
1 downloads 17 Views 2MB Size
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