Biosensors and Bioelectronics 20 (2004) 442–447

Monitoring of lung tumour cell growth in artificial membranes

Ying Yanga,∗, Josep Sulé-Susoa,c, Alicia J El Haja, Paul R Hobana, RuiKang Wangb

a Institute for Science and Technology in Medicine, School of Medicine, Keele University, Stoke-on-Trent ST4 7QB, UKb Institute of Bioscience & Technology, Cranfield University at Silsoe, Bedfordshire, Silsoe MK45 4DT, UKc Staffordshire Oncology Centre, University Hospital of North Staffordshire, Stoke-on-Trent ST4 7LN, UK

Received 23 November 2003; received in revised form 23 February 2004; accepted 24 February 2004Available online 13 May 2004

Abstract

Morbidity of many tumour types is associated with invasion of tumour cells through the basement membrane and subsequent metastasis tovital organs. Tumour invasion is frequently detected late on as many patients present with advanced disease. The method of detecting invasionis through conventional histological staining techniques, which are time consuming and require processing of the sample. This can affectinterpretation of the results. In this study, a new imaging technique, optical coherence tomography (OCT), was used to monitor lung tumourcell growth in two artificial membranes composed of either collagen type I or Matrigel. In parallel, standard histological section analysis wasperformed to validate the accuracy of the monitoring by OCT. Cross-sectional images from OCT revealed that lung tumour cells infiltrated onlywhen low cell seeding density (5× 105) and low collagen concentration (1.5 mg/ml) were combined. The cells could be easily differentiatedfrom the artificial membranes and appeared as either a brighter layer on the top of the membrane or brighter foci embedded within the darkermembrane. These cell-membrane morphologies matched remarkably to the standard histological section images. Our results suggest that OCThas a great potential to become a useful tool for fast and robust imaging of cell growth in vivo and as a potential assessment of cell invasion.© 2004 Elsevier B.V. All rights reserved.

Keywords: Tumour cell invasion; Lung cancer; Monitoring; Optical coherence tomography

1. Introduction

Lung cancer is a world-wide health problem associatedwith a high death rate (Pisani et al., 2002). A major factorassociated with morbidity of the disease is that tumour cellspossess the ability to invade the basement membrane of theorgan in which they arise and metastasise and invade othervital organs. For example, most lung cancer patients are cur-rently diagnosed with overtly or covertly metastatic disease,which, given the current lack of effective anti-cancer agents,is likely to be fatal. Whilst the early detection of diseaseshould improve cure rates, it is not currently clear at whichstage primary lesions metastasise.

The invasive behaviour of tumour cells in part dependson cellular factors such as enzymes that enable cells to de-grade their matrix components as well as on the characteris-tics of the extracellular matrix surrounding the cell (Ivaska

∗ Corresponding author. Tel.:+44-1782-554606;fax: +44-1782-717079.

E-mail address: bea00@keele.ac.uk (Y. Yang).

and Heino, 2000; Curran and Murray, 1999). Therefore, agood knowledge of the cell–matrix interaction is necessarywhen seeking treatment strategies aimed not only at cellkilling but also at inhibition of tumour cell invasion andmetastases.

In this study we proposed to investigate lung tumourgrowth and invasion. Anatomically, the healthy pleural cav-ity is a closed space lined by a continuous membrane andcontaining a small volume of clear colourless fluid. Its sur-face is lined by a single layer of flattened mesothelial cellsanchored to a basement membrane that lies on layers ofcollagen and elastic tissue. Vascular and lymphatic vesselsare found in the subpleural layer. The extracellular matrixcontains an abundance of hyaluronic acid, types I and IIIcollagen and fibronectin. Elastin fibres are plentiful andbasement membrane proteins, including type IV collagenand laminin, are found at the mesothelial cell–stromal inter-face (Carter et al., 1998). Establishing a three-dimensionalmodel for lung tumour/matrix interaction that mimics natu-ral pleural membrane could enable study of the mechanismsof cell invasion. Furthermore, it would be important to

0956-5663/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.bios.2004.02.030

Y. Yang et al. / Biosensors and Bioelectronics 20 (2004) 442–447 443

develop an imaging tool that provides a rapid and robustassessment and visualization of cell invasion.

Several studies have been carried out to better under-stand the interaction between tumour cells and basementmembrane (Engbring and Kleinman, 2003). Various systemshave been reported for the study of tumour cell invasionin basement membrane (Park et al., 2003; Connolly andMaxwell, 2002; Starkey et al., 1984). The most commonlyused techniques to visualise cell penetration through mem-branes are histological sectioning and transmission electronmicroscopy. They are cumbersome and time consuming, andfurthermore, the processes involved in specimen preparationcould distort images.Albini et al. (1987)developed a rapidassay for quantitating the invasion potential of tumour cellsin which a Matrigel layer was laid as a barrier on the topof a porous filter of a Boyden chamber. The invaded cellswould be viewed in the porous filter with a light microscopeif they penetrated through the Matrigel. This is a simplemonitoring method. However, it is not possible to assess thedynamic process of cell invasion through the thickness ofthe membrane.

In the past decade, a new imaging technology, opticalcoherency tomography (OCT) (Huang et al., 1991), has beendeveloped. OCT is a non-invasive real-time measurementtechnique with high spatial resolution (Wang et al., 2001;Wang and Elder, 2002), which can acquire cross-sectionalimages at high speed, an important feature for monitoringcell behaviours. In this study, a model system composedof lung tumour cells and two types of artificial membraneswas developed. We used two non-small cell lung cancer(NSCLC) cell lines CALU-1 and SK-MES as these cell lineswere already available in our laboratory and we had alreadyexpertise in the in vitro growth of these cells (Sulé-Susoet al., 2004).

The invasiveness of the tumour cells into the artificialmembranes was monitored by OCT and standard histolog-ical analysis. The accuracy and advantages of using OCTare discussed and compared to conventional histologicalmethods. Our hypothesis is that OCT has sufficient reso-lution to detect cell clusters within a gel, which could beused to develop a new and robust method to study cellinvasion.

2. Materials and methods

2.1. Cell lines

The cell lines used in this study were two non-small celllung cancer cell lines (CALU-1 and SK-MES) purchasedfrom the European Collection of Cell Cultures (Salis-bury, UK). Cells were kept in culture in complete media(DMEM with 10% foetal calf serum (FCS)) according tothe provider’s instructions. Culture media was changedevery 3–4 days and maintained in a 5% CO2 incubatorat 37◦C.

2.2. Artificial membranes

Two types of artificial membranes were made up eitherfrom collagen type I from rat tail (BD Bioscience) or Ma-trigel from EHS mouse sarcoma (BD Bioscience). A thinring made from filter paper was placed at the bottom of eachwell in 24-well plates to anchor the gels to avoid shrinking.The gels were placed on these filter paper rings. The col-lagen gel concentration varied from 1.5 to 2.5 mg/ml. TheMatrigel concentration was 13.9 mg/ml. Gels were allowedto set firmly by incubating them at 37◦C for 30 min.

2.3. Cell growth on artificial membranes

Cell growth was carried out in 75 cm2 tissue culture flasks(Sarstedt, UK). Cells were detached before reaching con-fluence by removing the culture media and adding 4 ml oftrypsin/EDTA (Sigma, UK). Cells were then incubated for5 min at 37◦C and 5% CO2. Complete medium was thenadded and cells were spun at 1200 rpm for 7 min to obtaincell pellets. Two concentrations of tumour cells 5×105 and1 × 106 cell/well were used and the cells were seeded onthe surface of the artificial membranes. The cell-membraneconstructs were cultured for 3 weeks at 37◦C and 5% CO2and the culture media was changed twice a week. For a posi-tive control experiment, 1×106 fibroblasts were mixed withcollagen gel at 2.5 mg/ml and following the same cultureconditions as above.

The morphology of the cells growing on the surface of themembranes was observed using an optical microscope. Af-ter fixed culture periods, cell-membrane constructs were cutinto two semicircles. One half of the membrane was fixedby phosphate buffered formalin and then subjected to con-ventional histological paraffin embedding and sectioning.Another half was analysed using OCT without fixation orfurther preparation. In some cases, a specimen was scannedby OCT first, and then the same specimen was subjected tothe process of paraffin embedding and section. All sectionswere stained using haemotoxylin and eosin.

2.4. OCT imaging

The OCT system used in this study was a bench-top in-strument, and its schematic diagram is presented inFig. 1.This system employs a broad band light source, an 1300 nmsuperluminescent diode with a bandwidth of 52 nm, whichis coupled into an optical fibre, and split by a 50/50 opticalfibre coupler. The light source yields 14�m axial resolutionin free space, or 10�m in the tissue if the mean refractiveindex of bulk tissue is assumed to be 1.4; this determinesthe imaging axial resolution of the system. Half of the lightis directed to the reference arm of the interferometer wherea rapid double-pass scanning system (Tearney et al., 1997)is employed to modulate the interference signal and pro-vide the optical path length scanning. The remaining lightis guided towards the sample, with a focusing optics. The

444 Y. Yang et al. / Biosensors and Bioelectronics 20 (2004) 442–447

Fig. 1. A schematic diagram of the OCT system where CL denotes the collimating lens; FC, the fibre coupler; PC, the polarisation controller; OL, theobjective lens and D, the detector.

movement of the mirror is precisely controlled by a highresolution motorised translation stage. Light reflected fromthe sample is combined with light reflected from the mirror.These two beams interfere only if the optical path lengths ofthe two beams are matched to within the coherence lengthof the light. Polarisation controllers are used in both armsto obtain the maximum obtainable interference fringe visi-bility. The system employs a balanced detector scheme tominimise the excess noise arising from the light source. Thetransverse resolution was measured at 16�m, limited bythe numerical aperture of the lens used to deliver the lightonto the sample, and the optical frequency of the incidentlight as in conventional microscopy. The signal-to-noise ra-tio (SNR) of the system was measured at 100 dB by theuse of a 4-OD neutral density filter. OCT measures the in-tensity of backscattering light within the tissue by meansof interference and uses this to represent the tissue opticaldiscontinuity.

The cell-membrane constructs adhered to the filter paperring were removed from the wells and placed on a glassslide for OCT scanning. With the reinforcement of the fil-ter paper ring, the constructs remained integrate throughoutprocessing and the filter paper also acted as an indicator ofcell location. The position of the probe beam on the sam-ple was monitored using a visible-light guiding beam. Thescanning started from the top of the gel where the cells wereseeded through the thickness of the gel. The scanning areawas about 2 mm× 6 mm (depth× length). The time to ac-quire an OCT image was 2 s and the image format in pixelswas 256× 180 (depth× length).

3. Results

The typical light microscopic images of the tumour cellson different artificial membranes are presented inFig. 2. It

was found that the tumour cells grew into a network on thecollagen type I membrane, whereas cells formed dense mul-tiple cell layers on the Matrigel. In both cases, the majorityof cells were visible on the surface of the gel.

To determine the effect of cell seeding density and geldensity on invasion behaviour, we carried out experimentsin which a fixed number of cells were seeded on collagen atdifferent concentrations. Other experiments included seed-ing cells at two different concentrations on a fixed collagenconcentration. After the cell–gel constructs were culturedfor 3 weeks, the cross-sectional images of the membraneswere taken by OCT.

Fig. 3a shows the image of 5× 105 cells seeded on2.5 mg/ml collagen. It can be seen that OCT displayed theconstruct through its whole gel thickness. It is a two-layerimage: a thin, brighter layer on the top and a homogeneous,thick and darker layer on the bottom. The boundary betweenthe layers is very sharply delineated. By the nature of thecell seeding procedure, the brighter top layer was ascribedto the cells, whilst the underlying layer corresponded to thegel, which is confirmed by the scanning of the same gelstarting from the bottom where there was no cell seeding asshown inFig. 3b. This interpretation is confirmed from theresults obtained from histological sectioning of the speci-men (Fig. 3c).

We next determined whether we could detect cell infiltra-tion into the gel by altering cell density and collagen con-centration. The mechanism to generate an image contrast inOCT lies in the different refractive indexes within an object.The fact that the seeded tumour cells on the artificial mem-branes appeared as a brighter layer in OCT images indicatesthat the refractive index difference of the gel and the cellsis sufficient to differentiate between them. The assignmentof the top brighter layer, as a cell layer, is supported by thecell seeding experiments described. We would thus assign anOCT image with bright foci within the gel as cell infiltration.

Y. Yang et al. / Biosensors and Bioelectronics 20 (2004) 442–447 445

Fig. 2. Light micrograph of the tumour cells seeded on: (a) collagen gel (1.5 mg/ml) and (b) Matrigel (13.9 mg/ml). The bar is 55�m.

Fig. 3. Images of 5× 105 tumour cells seeded on collagen gel (2.5 mg/ml): (a) OCT with the cells on the top; (b) OCT with the gel on the top and (c)standard histological section (H and E stain). The bar is 100�m. The inset in (c) is the image of the cell–gel construct through the thickness.

Fig. 4. Images of 1×106 fibroblasts mixed with collagen gel (2.5 mg/ml): (a) OCT and (b) standard histological section (H and E stain). The bar is 100�m.

446 Y. Yang et al. / Biosensors and Bioelectronics 20 (2004) 442–447

Fig. 5. Images of 5× 105 tumour cells seeded on collagen gel (1.5 mg/ml): (a) OCT and (b) standard histological section (H and E stain). The bar is100�m. The inset in (b) is the image of the cell–gel construct through the thickness.

Fig. 6. Images of 1×105 tumour cells seeded on Matrigel (13.9 mg/ml): (a) OCT and (b) standard histological section (H and E stain). The bar is 66�m.

To confirm this one million fibroblasts were mixed with col-lagen gel at 2.5 mg/ml. After the cell–gel construct was cul-tured for 3 days at 37◦C and 5% CO2, an OCT image wastaken and shown inFig. 4a. It is clearly seen that there isno top cell layer but bright foci corresponding to cell clus-ters are distributed all through the gel. This interpretationis supported by the histological imaging of the specimen(Fig. 4b).

We next increased the cell number to 1×106 but keepingthe collagen concentration unchanged (2.5 mg/ml); this ledto a similar OCT and histological image as described, indi-cating that there is no cell penetration through the gel (datanot shown). In contrast, when the collagen concentrationwas reduced to 1.5 mg/ml and seeding with 5× 105 cells, abrighter diffused top cellular layer was observed with addi-

tional similarly bright cellular foci infiltrating within the gellayer (Fig. 5a). This was replicated with histological images(Fig. 5b). The OCT image for 5×105 cells on Matrigel with13.9 mg/ml demonstrated no cell foci within the gel layer,implying no cell infiltration which was supported from thehistological section of the specimen (Fig. 6a and b).

4. Discussion

In this study we have used OCT imaging as method todetect lung tumour cell growth and infiltration in an in vitroartificial membrane model. We have found that OCT is equalto histological staining methods for detecting tumour cellgrowth and infiltration. The relative transparent gel system

Y. Yang et al. / Biosensors and Bioelectronics 20 (2004) 442–447 447

in this study provides extra merit for OCT imaging. A strongbackscattering and deep penetration of the incident light insuch media were easily obtained. As demonstrated by ourcell seeding and mixing experiments, regardless whether thecells are on the top or within the gel, they appear as a brighterimage compared to the gel. Thus, OCT demonstrates itselfas a simple, rapid and reliable technique to assess cell inva-sion. The non-invasive and rapid data acquisition manner ofthis technique enables dynamic studies of cell invasion inartificial membranes as well. Furthermore, although artifi-cial gel could be made similar to tissue matrix, they may besofter and more fragile with higher proportional water con-tent and without tough network formation. In these cases,when the cell–gel construct is subjected to standard paraffinembedding, a significant shrinking and deformation of theconstructs frequently occur during the dehydration stage.The deformed artificial gel can cause distorted images,which can lead to misinterpretation of the data obtained.With OCT technology, non-distorted images can be obtainedeasily.

The invasion behaviour in our study apparently related tothe composition of artificial membrane and its density. Thereis a dose response of gel concentration for tumour cell in-vasion.Ehlers et al. (2002)reported that mesothelioma celllines invaded completely into collagen gel (1.5 mg/ml and5×104 cells) but did not penetrate within Matrigel (10 mg/mland 5× 104). We observed a similar phenomenon but thecells only invaded in certain concentrations of collagen gel.A lower collagen concentration enhanced cell penetration.The penetration of tumour cells into basement membranesrequires three basic events: tumour cells attach to the mem-brane; secretion of enzymes by the tumour cells or activa-tion of protease that cause the degradation of the adjacentbasement membrane; migration of the cells (Albini et al.,1987; Terranova et al., 1986). Therefore, the cell type andthe properties of basement membrane are the main factorsto determine cell invasion behaviour.

The basement membrane is a barrier that blocks the pas-sage of cells and macromolecules. The malignant grade ofa neoplasm is influenced by the invasion and metastatic po-tential as well as the growth rate of the tumour cells. Wewould envisage the future combination of in vitro base-ment membrane models with OCT imaging as a usefultool for analysing the molecular events determining cell in-vasion, in addition to providing an assay for the screen-ing of potential anti-cancer drugs which may inhibit thisprocess.

5. Conclusion

The data from this study demonstrate OCT as a rapidand reliable imaging technique for monitoring tumour cell

growth and invasion, which is comparable to standard his-tological analysis.

Acknowledgements

This work was partially supported by the grant fromEngineering and Physical Science Research Council, UK,GR/N13715.

References

Albini, A., Iwamoto, Y., Kleinman, H.K., Martin, G.R., Aaronson, S.A.,Kozlowski, J.M., McEwan, R.N., 1987. A rapid in vitro assay forquantitating the invasive potential of tumor cells. Cancer Res. 47,3239–3245.

Carter, D., True, L., Otis, C.N., 1998. Serous membranes. In: SternbergS.S. (Ed.), Histology for Pathologists. Lippincott-Raven Publishers,Philadelphia, USA, pp. 223–239.

Connolly, L., Maxwell, P., 2002. Image analysis of Transwell assays inthe assessment of invasion by malignant cell lines. Br. J. Biomed. Sci.59, 11–14.

Curran, S., Murray, G.I., 1999. Matrix metalloproteinases in tumour in-vasion and metastasis. J. Pathol. 189, 300–308.

Ehlers, E.M., Bakhshandeh, A., Wiedemann, G., Kuhnel, W., 2002. Inva-siveness of human pleural mesothelioma cells is influenced in vitroby the three-dimensional structure of the ECM and its composition.Ann. Anat. 184, 417–424.

Engbring, J.A., Kleinman, H.K., 2003. The basement membrane matrixin malignancy. J. Pathol. 200, 465–470.

Huang, D., Swanson, E.A., Lin, C.P., Schuman, J.S., Chang, W., Hee,M.R., Flotte, T., Gregory, K., Puliafito, C.R., Fujimoto, J.G., 1991.Optical coherence tomography. Science 254, 1178–1181.

Ivaska, J., Heino, J., 2000. Adhesion receptors and cell invasion: mecha-nisms of integrin-guided degradation of extracellular matrix. Cell Mol.Life Sci. 57, 16–24.

Park, D.W., Choi, D.S., Ryu, H.S., Kwon, H.C., Joo, H., Min, C.K.,2003. A well-defined in vitro three-dimensional culture of humanendometrium and its applicability to endometrial cancer invasion.Cancer Lett. 195, 185–192.

Pisani, P., Bray, F., Parkin, D.M., 2002. Estimates of the world-wideprevalence of cancer for 25 sites in the adult population. Int. J. Cancer97, 72–81.

Starkey, J.R., Hosick, H.L., Stanford, D.R., Liggitt, H.D., 1984. Interactionof metastatic tumor cells with bovine lens capsule basement membrane.Cancer Res. 44, 1585–1594.

Sulé-Suso, J., Forster, A., Zholobenko, V., Stone, N., El Haj, A., 2004.Effects of CaCl2 and MgCl2 on Fourier transform infrared spectra oflung cancer cells. Appl. Spectrosc. 58, 61–67.

Tearney, G.J., Bouma, B.E., Fujimoto, J.G., 1997. High speed phase andgroup-delay scanning with a grating-based phase control delay line.Opt. Lett. 22, 1811–1813.

Terranova, V.P., Hujanen, E.S., Martin, G.R., 1986. Basement membraneand the invasive activity of metastatic tumor cells. J. Natl. CancerInst. 77, 311–316.

Wang, R.K., Xu, X., Tuchin, V.V., Elder, J.B., 2001. Concurrent enhance-ment of imaging depth and contrast for optical coherence tomographyby hyperosmotic agents. J. Opt. Soc. Am., B. 18, 948–953.

Wang, R.K., Elder, J.B., 2002. High resolution optical tomographic imag-ing of soft biological tissue. Laser Phys. 12, 611–617.