Study of Tumor Cell Invasion byFourier Transform InfraredMicrospectroscopy

Ying Yang1

Josep Sule-Suso1,2

Ganesh D. Sockalingum3

Gregory Kegelaer3

Michel Manfait3

Alicia J El Haj11 Institute for Science and

Technology in Medicine, KeeleUniversity, Stoke-on-Trent ST4

7QB, UK

2 Staffordshire OncologyCentre, University Hospital of

North Staffordshire,Stoke-on-Trent ST4 7LN, UK

3 Unite Median, CNRS UMR6142, Faculte de Pharmacie,IFR53, Universite de Reims,

Champagne-Ardenne,51 rue Cognacq-Jay, 51096

Reims cedex, France

Received 15 November 2004;revised 25 March 2005;accepted 12 April 2005

Published online 16 May 2005 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/bip.20297

Abstract: Lung cancer is usually fatal once it becomes metastatic. However, in order to developmetastases, a tumor usually invades the basal membrane and enters the vascular or lymphatic system.In this study, a three-dimensional artificial membrane using collagen type I, one of the main compo-nents of basal membranes, was established in order to investigate tumor cell invasion. Lung cancer cellline CALU-1 was seeded on this artificial membrane and cell invasion was studied using the Fouriertransform infrared (FTIR) imaging technique. This approach allowed identification of tumor cellsinvading the collagen type I membrane by means of their infrared spectra and images. The mappingimages obtained with FTIR microspectroscopy were validated with standard histological section analy-sis. The FTIR image produced using a single wavenumber at 1080 cm�1, corresponding to PO2

�

groups in DNA from cells, correlated well with the histological section, which clearly revealed a celllayer and invading cells within the membrane. Furthermore, the peaks corresponding to amide A, I,and II in the spectra of the invading cells shifted compared to the noninvading cells, which may relateto the changes in conformation and/or heterogeneity in the phenotype of the cells. The data presentedin this study demonstrate that FTIR microspectroscopy can be a fast and reliable technique to assesstumor invasion in vitro. # 2005 Wiley Periodicals, Inc. Biopolymers 78: 311–317, 2005

This article was originally published online as an accepted preprint. The ‘‘Published Online’’ datecorresponds to the preprint version. You can request a copy of the preprint by emailing the Biopoly-mers editorial office at biopolymers@wiley.com

Correspondence to: Ying Yang; e-mail: bea00@keele.ac.uk

Biopolymers, Vol. 78, 311–317 (2005)

# 2005 Wiley Periodicals, Inc.

Contract grant sponsor: Franco-British Partnership Programme

311

Keywords: tumor cell invasion; cell imaging; Fourier transform infrared microspectroscopy;three-dimensional collagen model

INTRODUCTION

Morbidity of many tumor types is associated with inva-

sion of tumor cells through the basal membrane and

subsequent metastasis to vital organs. Tumor invasion

is frequently present in many patients with advanced

disease. Therefore, a good knowledge of the cell–

matrix interaction that controls cell invasion is neces-

sary when seeking treatment strategies aimed not only

at cell killing but also at the inhibition of tumor cell

invasion and metastases.1 In our previous study,2 we

monitored lung tumor cell growth in two artificial

membrane systems composed of either collagen type I

or Matrigel by the new nondestructive in situ imaging

technique, optical coherence tomography (OCT).

Tumor cell invasion could be visualized by this tool but

no molecular-level information could be obtained. In

the present study, one step further was taken to reveal

not only the morphology of cell invasion, but also the

chemical compositions of the invading cells in the arti-

ficial membrane model system by using Fourier trans-

form infrared (FTIR) microspectroscopy and imaging.

FTIR has been extensively used to study the struc-

ture and composition of biological substances such as

proteins, lipids, and inorganic compounds.3–6 The

advantage of using FTIR to characterize biological

substances lies in its relatively simple sample prepara-

tion and low operation cost. Not only can the composi-

tions be detected, but the structural configuration,

bonding nature, and conformation can be character-

ized. Recent developments in FTIR microspectroscopy

permit quantitative determination of the relatively

small amount, molecular nature, distribution, and ori-

entation of the analyzed substances at a spatial resolu-

tion of about 10 �m.7,8 The use of FTIR spectroscopy

combined with a microscope provides a direct probe of

the chemical composition for mapping an intact bio-

logical system with a high spatial resolution. In this

article, we present FTIR microspectroscopy data

obtained from lung cancer cells growing and migrating

on an artificial membrane system, a three-dimensional

model, and demonstrate how to explore the potential of

this new approach in the study of cell invasion.

EXPERIMENTAL PROCEDURES

Cell Line

The cells used in this study were the nonsmall cell lung

cancer cell line, CALU-1, purchased from the European

Collection of Cell Cultures (Salisbury, UK). Cells were cul-

tured in complete medium [Dulbecco’s Modified Eagle’s

Medium (DMEM) with 10% fetal calf serum (FCS)] ac-

cording to the provider’s instructions. The culture medium

was changed every 3–4 days and cell cultures maintained in

a 5% CO2 incubator at 378C.

Cell Growth on Artificial Membrane

Collagen type I of rat tail was used to form an artificial

membrane. Although the collagen gel used was made with

varying concentrations from 1.5 to 2.5 mg/mL, cell inva-

sion was observed only in the system with the collagen gel

at 1.5 mg/mL concentration and seeding 0.5 � 106 cells as

in the previous study.2 Therefore, these conditions were

used in the present experiments. The required volume of

collagen solution was placed at the bottom of each well in

24-well plates anchored by a thin ring made from filter

paper to avoid the gels’ shrinking. Gels were allowed to set

firmly by incubating them at 378C for 30 min. After this

incubation period, 0.5 � 106 cells were seeded on the top of

the gel. The cell membrane constructs were cultured for

3 weeks at 378C and 5% CO2, and the culture medium was

changed twice a week.

FTIR Microspectroscopy

FTIR imaging was performed using the Spectrum Spotlight

300 imaging system (Perkin Elmer Life Sciences, UK).

This system has the particular advantage of having a liquid

nitrogen cooled mercury cadmium telluride (MCT) line

detector composed of 16 elements for rapid image analysis.

It is well adapted to analyze samples of varying sizes (small

size < 100 �m to bigger sizes of �75 � 45 mm2). The sam-

ple was continuously scanned and the IR image calculated

in real time either at low (25 �m) or high (6.25 �m) spatial

resolution. A visible image matching exactly the IR image

can also be captured for comparison. Samples were studied

in reflectance mode with a 4 cm�1 resolution and a coaddi-

tion of 32 scans. All data acquisition, pre-processing

(atmospheric and baseline corrections and normalization),

and image construction were performed using the Spotlight

software (Perkin Elmer, UK).

Sample Preparation

After 3 weeks’ culture, the cell membrane constructs were

fixed by 10% physiological buffered saline (PBS) buffered

formalin. The fixed specimens were subjected to routine

paraffin embedding and sectioning. For the light micro-

scopic imaging, the deparaffinized sections were stained by

hematoxylin & eosin (H&E). The optical images were

taken with a Leica digital camera. For the sections used for

FTIR microspectroscopy, aluminium (Al)-coated glass

slides were used and the sections (7 �m thick) from three

312 Yang et al.

different specimens were deparaffinized following the

standard procedure.

Statistics

The amide peak intensity values from nine individual spec-

tra of three different specimen sections were collected, at

cell layer, spot cell, and collagen gel locations respectively.

Analysis of variance (ANOVA) by multicomparison test

(Tukey–Kramer) was applied to the mean values to work

out whether the peak shift among the three locations was

significant or not. Significance values were calculated with

p < 0.05.

RESULTS AND DISCUSSION

Figure 1a shows the histological section image of the

deparaffinized cell membrane construct cultured for 3

weeks. The construct was made of 1.5 mg/mL collagen

gel seeded with 0.5 � 106 cells on the top of the gel.

Since cells were seeded on top of the collagen gel, it is

easy and reasonable to assign the predominant brighter

area as collagen gel and the loose and darker layer at

the top and the spots away from the top as cells indi-

cated by arrows. The spots away from the original

seeding site clearly demonstrate that the tumor cells

did indeed migrate into the membrane. This assign-

ment has been supported by a parallel experiment as in

the image shown in Figure 1b, in which the construct

was made of 2.5 mg/mL collagen gel seeded with 0.5

� 106 cells. It can be seen that the cell layer was on the

top of the gel with a dense and well-defined boundary.

The cells did not migrate into the gel.

FIGURE 2 (a) Visible image of the section for the same

specimen as in Figure 1 taken by the visible camera of the

FTIR imaging system. (b) FTIR spectral image constructed

on the whole frequency range of 4000–750 cm�1 of the

same area as in (a) using a resolution of 6.25 �m/pixel. The

bar represents 85 �m.

FIGURE 1 (a) Light micrograph of a histological section of the CALU-1 cells seeded on colla-

gen membrane with a collagen concentration of 1.5 mg/mL and cell seeding number of 0.5 million

per membrane (H&E staining). (b) Light micrograph of a histological section of the CALU-1 cells

seeded on collagen membrane with a collagen concentration of 2.5 mg/mL and cell seeding number

of 0.5 million per membrane (H&E staining). The solid arrows indicate cells, while the hollow one

denotes collagen. The bar represents 100 �m.

Study of Tumor Cell Invasion 313

Figure 2(a) displays the visible image of a deparaf-

finized section of the same sample block as in Figure

1a placed on an Al-coated slide and recorded by the

visible camera of the FTIR imaging system. Without

staining, a similar image to Figure 1a can be seen,

i.e., the resting cells appearing as a dark layer and the

spots corresponding to migrated cells away from the

layer. The remaining brighter area of the section,

which corresponds to the collagen membrane, is

homogenous.

The visible image provides information on the posi-

tion of the cells but no molecular information could be

obtained. In contrast, FTIR imaging provides in a sin-

gle measurement both spatial and molecular informa-

tion of the cell and its environment. Figure 2(b) depicts

the FTIR spectral image that matches exactly the visi-

ble image obtained in Figure 2(a). The corresponding

false color spectral image is based on the total absorb-

ance of the spectra collected in the spectral range of

4000–750 cm�1. It can be seen that an image very sim-

ilar to that presented in Figure 1a has been obtained.

The darker region becomes brighter in the IR image

due to the higher absorption. In this case, the collagen

matrix appears blue due to a weaker absorption com-

pared to the cells.

However, in order to verify whether the spots

within the membrane corresponded to invading cells

and not simply to a higher signal from the collagen

membrane, it was necessary to reconstruct the

FTIR image on spectral information that is not com-

mon to both cells and collagen matrix. The peak at

1080 cm�1 seems to be an ideal candidate as shown

in Figure 3(a). The choice of this peak lies in the fact

that its presence is indicative of phospholipids or

phosphodiester bond stretches from DNA,9 which

originate from the nucleus and cell membrane—

hence cell components. With this information, which

permits spectral information of cells to be distin-

guished from that of collagen, the IR spectral image

was reconstructed using the band intensity (two-point

baseline corrected) at 1080 cm�1, as shown in Figure

3(b). This false color spectral image now shows, with

sufficient spatial resolution, the position of the super-

ficial layer of cells as well as the migrating cells in

the collagen gel. The color bar indicates the intensity

of the 1080 cm�1 band. The high intensity values

(red color) in the spectral image correspond to the

regions of high cell information, whereas low inten-

sity values (blue color) reflect those without cells or

the weak signal from the collagen matrix. The fact

that the spots exhibit the same intensity as in the cell

layer further demonstrates that the spots within the

membrane correspond to cells that have migrated into

the membrane from the top cell layer.

FTIR spectra extracted from three typical sites in

the image, i.e., the cell layer (seeded cells), the spots

(invading cells), and the collagen membrane site,

indicated by the arrows in Figure 2(b), are compared

in Figure 4. The three spectra exhibited quite similar

features. However, the precise location and intensity

of amide peaks differentiated them from each other,

particularly in the absorption regions of amide A,

amide I, and amide II. Nine independent measure-

ments from three different samples have been carried

out. It was found that the mean peak values of colla-

gen gel at amide A, I, and II peaks were at 3313,

1659, and 1549 cm�1, respectively, while the cell

layer exhibited the corresponding peaks at 3287,

1644, and 1540 cm�1. This clearly distinguished the

molecular information of the resting cells from that

of the collagen gel. It is interesting to note that, when

the FTIR spectra of the superficial layer of cells were

compared to those of the migrated cells, these peaks

FIGURE 3 (a) FTIR spectra showing the 1080 cm�1 band

for the different sites of the image marked by arrows in Fig-

ure 2(b). (b) Reconstruction of the IR spectral image based on

the 1080 cm�1 band intensity (PO2� from DNA) for the same

section as in Figure 2. The bar represents 85 �m.

314 Yang et al.

shifted for spot cells compared to the cell layer,

and were found to be situated at 3301, 1639, and

1534 cm�1, respectively. By performing a statistical

analysis of the mean values, it was also found that the

amide peak shifts were significantly different

between collagen, cell layer, and spot cells, as shown

in Figure 5.

Payne and Veis10 studied cell migration and

assigned the downshift of amide I peak in migrated

cells to the disruption of the local triple-helical struc-

ture of proteins within the macromolecules. In the

present case, we hypothesize that the shifts of the spot

cells at amide peaks might be caused, beside the

change in the secondary structure to the antiparallel

� type of protein structure, by two further potential

factors: the changes in conformation or in phenotype

of the invading cells, and the presence of surrounding

collagen gel molecules and/or proteinases produced

by the cells. In fact, the contribution of collagen to

the spectra of invading cells was negligible. This was

verified by running a line-map experiment across the

invading cells, and collecting and comparing the

spectra from collagen gel over a single cell spot. We

found that, within the limit of the pixel resolution

(6 �m), the spectra were very sensitive to subtle dif-

ferences around the spot cells, as shown in Figure 6.

All points corresponding to cell information (2, 3,

and 4) were centered on the 1080 cm�1 DNA band,

while those taken from outside the cell (1 and 5) had

bands closer to 1085 cm�1, which was from collagen.

Also, since the maxima of these three points did not

shift, they corresponded approximately to the cell

dimension, i.e., to about three pixel sizes (3 � 6.25

�m ¼ about 20 �m), therefore to the cell size (15–20

�m). Further experiments with a synchrotron IR

source are planned to confirm this observation.

It is worth mentioning that the 1080 cm�1 absorp-

tion peak, due to the symmetric stretching of the

PO2� group, is a unique band to distinguish cell and

collagen. The peak at 1240 cm�1 is assigned to the

antisymmetric stretch band for the PO2� group.11,12

However, from our own data and what has been

reported elsewhere, collagen also has an absorption

band that coincides with 1240 cm�1. Therefore, the

1240 cm�1 should be absent in the specimen with

some other protein matrices but not in the present

case where collagen is the matrix protein. The anti-

symmetric stretch band was therefore not appropriate

to be used for reconstructing the FTIR image.

The purpose of this work was to assess not only

whether FTIR microspectroscopic imaging could dif-

ferentiate the cells from the collagen (the artificial

membrane) if the cells invaded/penetrated into the

membrane layer, but also whether the migrating cells

displayed different spectra compared to resting cells.

Cells contain large quantities of proteins. Therefore,

it is expected that the spectra of collagen and cells

would be very similar. However, the different types

and configurations of proteins in cells, in addition to

the presence of phospholipids in the cell membrane

and phosphate in the nucleic acid, might change the

positions and intensity of the amide peaks and give

rise to the presence of additional unique peaks. The

differences exhibited by the amide peaks in the spec-

tra between CALU cells and collagen are clearly visi-

ble (Figure 4). These differences are further enhanced

when the cells have been cultured and invaded within

the collagen gel. Indeed, the spectral differences are

observed between invading and noninvading cells.

Further work will be undertaken to clarify the mecha-

nisms that cause the spectral changes.

In conclusion, we have demonstrated that FTIR

microspectroscopy is a simple, reliable method to

study tumor cell invasion. The chemical fingerprints

of the spectra recorded in imaging mode show not

FIGURE 4 Comparison of FTIR spectra representing (a)

the amide A range and (b) the amide I and II range, taken at

different positions of the image in Figure 2(b) marked by

arrows.

Study of Tumor Cell Invasion 315

FIGURE 5 The ANOVA analysis of mean values of (a) amide A, (b) amide I, and (c) amide II,

for collagen, cell layer, and spot cells based on nine spectra from three different specimens.

only the position of the cells within a membrane, but

also the configuration change of the cells invading

the membrane. In addition, the capability of FTIR

microspectroscopy to detect chemical compositions

in imaging mode should enable us to study the matrix

change due to cells invading a membrane. Three

events are involved in the penetration of tumor cells

into basement membranes: tumor cells attach to the

membrane; secretion of enzymes by the tumor cells

or activation of proteases causes the degradation of

the adjacent basement membrane; and cell migra-

tion.13,14 It has been reported that FTIR can follow

proteolytic processes induced by invasive cancer cells

in a two-dimensional model system.15 Our three-

dimensional model system shows a potential for mon-

itoring the collagen degradation process near cells.

Further experiments are underway to establish the

correlation of membrane compositions and tumor cell

invasion ability and the local composition change

caused by the cells.

This project was partially supported by the Franco-British

Partnership Programme (Alliance 2004).

REFERENCES

1. Ehlers, E. M.; Bakhshandeh, A.; Wiedemann, G.;

Kuhnel, W. Ann Anat 2002, 184, 417–424.

2. Yang, Y.; Sule-Suso, J.; El Haj, A. J.; Hoban, P. R.;

Wang, R. K. Biosens Bioelectro 2004, 20, 442–447.

3. Surewicz, W. K.; Mantsch, H. H.; Chapman, D. Bio-

chemistry 1993, 32, 389–394.

4. Salman, A.; Ramesh, J.; Erukhimovitch, V.; Talyshin-

sky, M.; Mordechai, S.; Huleihel, M. J Biochem Bio-

phys Methods 2003, 55, 141–153.

5. Blazewicz, M.; Paluskiewicz, C. J Mol Struct 2001,

563–564, 147–152.

6. Manoharan, R.; Baraga, J. J.; Rava, R. P.; Dasari, R.

R.; Fitzmaurice, M.; Feld, M. S. Atherosclerosis 1993,

103, 181–193.

7. Camacho, N. P.; West, P.; Torzilli, P. A.; Mendelsohn,

R. Biopolymers 1993, 62, 1–8.

8. Diem, M.; Romeo, M.; Boydston-White, S.; Miljkovic,

M.; Matthaus, C. Analyst 2004, 129, 880–885.

9. Wang, H. P.; Wang, H. C.; Huang, Y. J. Sci Total Envi-

ron 1997, 204, 283–287.

10. Payne, K. J.; Veis, A. Biopolymers 1988, 27, 1749–

1760.

11. Jackson, M.; Mantsch, H. H. Spectrochim Acta Rev

1993, 15, 53–69.

12. Jackson, M.; Mansfield, J. R.; Dolenko, B. D.; Somor-

jai, R. L.; Mantsch, H. H.; Watson, P. H. Cancer Det

Prev 1999, 23, 245–253.

13. Albini, A.; Iwamoto, Y.; Kleinman, H. K.; Martin, G.

R.; Aaronson, S. A.; Kozlowski, J. M.; McEwan, R. N.

Cancer Res 1987, 47, 3239–3245.

14. Terranova, V. P.; Hujanen, E. S.; Martin, G. R. J Natl

Cancer Inst 1986, 77, 311–316.

15. Federman, S.; Miller, L. M.; Sagi, I. Matrix Biol 21,

2002, 567–577.

Reviewing Editor: Laurence A. Nafie

FIGURE 6 (a) Line-map profile across the FTIR image

of one invading cell spot. The bar represents 10 �m. (b)

FTIR spectra corresponding to the numbered locations.

Study of Tumor Cell Invasion 317