Infrared spectroscopy as a tool for discrimination between sensitive and multiresistant K562 cells Anthoula Gaigneaux, Jean-Marie Ruysschaert and Erik Goormaghtigh Laboratory of Structure and Function of Biological Membranes, Free University of Brussels, Belgium Fourier transform infrar ed spectroscopy was p erformed on human leukemic daunorubicin-sensitive K 562 cells and their multiresistant counterpart derived by selection. Statistical analysis, including variable reduction and linear discrimi- nant analysis was performed on sensitive and multiresistant cells spectra in order to establish a diagnostic tool for multiresistant pattern. For each of t he two methods of data reduction tested [genetic algorith m or principal component analysis (PCA)] discrimination between the two cell lines was found to be possible. The best results, obtained with PCA-reduction, showed an accuracy of 93% on a distinct test set of spectra. These results d emonstrate the efficiency of Fourier transform infrar ed spectroscopy for c lassification. Further analysis o f t he spectral differences indicated that discrimination between r esistant and sensitive cells was based o n variations in all cellular contents. Lipid and nucleic acid decreased, relatively, while the protein content increased. Keywords: multiresistance; infrared spectroscopy; multivar- iate statistics; K562; leukemia. In recent years, infrared spectros copy has b een a powerful tool for biodiagnostics [1]. A major advantage of infrared spectroscopy over more classical t echniques of investigation is that neither s taining of t he samples nor chemical reagent additions are necessary. Just a few minutes and a few lLof a cell suspension are sufficient to obtain a spectrum representative of all cell constituents. This technique is based on absorption of infrared light by the vibrational transitions in covalent bond s. Intensities provide quantitative information, while frequencies give qualitative information about the nature of these bonds, their structure, and their molecular environment. In complex systems such as ce lls, the main absorptions arise from N–H, C ¼O, C–H and P¼O bonds from the proteins, lipids, and nucleic acids present in the cells. An infrared spectrum of cells is the sum of all these contribu - tions. A classical group frequency approach c an be used to interpret c hanges in one of t he cell component, as p reviously done on leukemic cell lines [2]. Another w ay to analyse infrared spectra is to use the s pectral signature to correlate spectral patterns with biological properties. Rigas [3] p roved that IR spectroscopy was a ble to detect features of human normal or malignant cultured colonocytes. Multivariate statistics known a s Ôpattern r ecognition techniques Õ have been used to classify spectra in intr insic g roups when they are unsupervised (clus ter analysis, o r principal componen t analysis). Naumann et al. [4] successfully used cluster analysis to characterize hundreds of bacterial cell lines. T he same approach was also used to clearly distinguish between normal and chronic lymphocytic leukemia cells [5]. Super- vised multivariate methods, such a s linear discriminant analysis (LDA) or partial least squares regression are powerful tools t o build rules o f discrimination t hat are used later t o i dentify n ew samples. This method was successfully applied t o s kin tumours [6] and to l ymph c ells and tissues [7]. The multiresistant phenotype is an s ignificant problem in cancer chemoth erapy. It i s c haracterized by cell resistance to multiple and s tructurally unrelated drugs [8]. I t m ay be expressed by cells selected for resistanc e to a single agent. Many of these m ultiresistant cells differ f rom their sensitive counterpart by overexpression of a membranous protein of 170 k Da, named P-glycoprotein (P-gp) [9]. Although the sole presence of P -gp has proven in s ome cell lines to con fer multidrug resistance phenotype [10], previous studies have shown that molecular c hanges in lipid an d nucle ic acid fractions of the cells accompany P -gp overexpression [11,12]. In this study, w e worked with sensitive (K562/DNS) and multiresistant (K 562/DNR) human chronic myelogenous leukemia K562 cells. First, we examined whether infrared spectroscopy, a ssociated w ith d ata reduction techniques a nd multivariate statistics, is a ble t o identify multidrug resistant phenotype in t hese cells with a high accuracy. S econd, we tried to learn more about biological origin of the spectral differences that exist b etween the K562-multiresistant cell line and its sensitive counterpart. MATERIALS AND METHODS Cell culture K562 is a human chronic myelogenous leukemia cell line. In this study, two different K 562 lines were used. The first cell line ( cell line A) has been described p reviously [13]. A second Correspondence to G. Erik, Laboratory of Structure and Function of Biological Membranes, Free University of Brussels, CP 206/2, Boulevard du Triomphe, B-1050 Brussels, Belgium. Fax: + 32 2 650 5382, Tel.: + 32 2 650 5386, E-mail: egoor@ulb.ac.be Abbreviations: P-gp, P-glycoprotein; K562/DNS, sensitive K562 cells; K562/DNR, daunorubicin resistant K562 cells; PCA, principal com- ponent analysis; LD A, linea r discriminant analysis; MDR, multidrug resistant. (Received 4 January 2002, acc epted 21 J anuary 2002) Eur. J. Biochem. 269, 1968–1973 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.02841.x cell line ( cell line B ) was obtained from A. Delforge (Bordet Hospital, Bruxelles). F rom each cell line (K562/DNS), a multiresistant subline (K562/DNR) was derived by s election on daunorubicin. All cell lines were kept in exponential growth in RPMI 1640 medium, s upplemented w ith 10% fetal bovine serum, L -glutamine (2%), and 1% antibiotic/ antimycotic solution, at 37 °C, in an humidified atmosphere of 5% CO 2 . All growing media and supplement w ere purchased at Life Technologies (Paisley, Scotland). To maintain resistance phenotype, K562/DNR was selected in a m edium c ontaining 1 l M daunorubicin or doxorubicin for 1 w eek every 2 months. A ll infrared measurements were carried out at least one week after the interruption of culture with selection agent. The cell lines were maintained at the same density of cells and then harvested in the same phase of culture growth (exponential) for IR measurement. For h arvesting, cells were centrifuged 3 min at 300 g and the pellet washed twice in a solution 0 .9% N aCl t o r emove all growing medium. FTIR spectroscopy An aliquot of cell p ellet was deposited on a g ermanium crystal (  2–5 · 10 5 cells per smear). The sample was rapidly e vaporated in N 2 flux to obtain a homogenous film of entire cells. IR m easurements were recorded be tween 4000 a nd 8 00 cm )1 by a Bruker E quinox spectrophotometer (Bruker, Karlsruhe, Germany) containing a liquid N 2 -refrigerated Mercury Cadmium Telluride detector. Each spectrum w as obtained by averaging of 256 scans at a resolution of 4 cm )1 . T he spectra were baseline c orrected and nor malized f or equal area b etween 1711 and 1485 cm )1 . Spectra were encoded every 1 cm )1 . Data analysis All spectra were treated with i n-house s oftware w orking in a MATLAB environment ( MATLAB 6, Mathworks Inc., Natick, USA). Spectra were separated in a training set constituted of 48 samples o f the cell line A and a test set c omposed of 30 spectra of cell line A and cell line B. The trainin g set was the only one used for model calculations (PCA, genetic algorithm and LDA). Data reduction by principal component analysis (PCA). IR spectra are samples defined by  3000 variables. To reduce this number, PCA was performed. PCA is a method of variable reductio n that builds linear combinations between variables (wavenumbers) varying together. The first linear combination is called the first principal c ompo- nent, a nd contains almost 98% of the variance. The second principal c omponent is a linear combination of wavenum- bers, which explains t he maximum of residual variance and is perpendicular to the first one. The following principal components obey the sam e rules. This method allows the reduction of a spectrum to 1 0 v ariables (th e first 10 principal components) that explain almost 100% of v ariance. Selection of wavenumbers with genetic algorithm.The genetic a lgorithm is a supervised method that uses muta- tion/selection principles to solve problems [14]. Many parameters can be adjusted for increasing the efficiency of the algorithm. The data were analysed with a window of five wavenumbers, assuming that adjacent wavenumbers are highly corr elated. A population o f 3 2 solutions was built at each generation, and e valuated. The algorithm stopped a t generation 100 or when 50% of convergence was reached between all the solutions. The mutation rate was 0.005, with double c rossing-over, and data were divided in nin e subsets to cross-validate the models. As the solutions proposed by this method are not deterministic, running the algorithm several times allows a more precise solution to be obtained. Only the wavenum- bers selected in more than 80% of all m odels built wer e kept in the final model. Linear discriminant analysis (LDA).Thisstatistical multivariate method is supervised. It searches for the variables containing the greatest interclass variance and the smallest intraclass variance, and constructs a linear combination of the variables to discriminate between the classes. The rule i s constructed with training set of samples, and further tested with the test set. We performed LDA in standard method, i.e. including all the variables in the model. RESULTS Spectral information contained in a cell IR spectrum Figure 1 (line A) shows a representative spectrum of K 562/ DNS cells, whic h can be divided in three regions. The absorption between 300 0 and 2800 cm )1 is dominated by the stretching vibration of CH 2 and CH 3 groups mainly contained in fatty acids of the cell. The band at 2963 cm )1 can be a ssigned to the asymmetric stretching of CH 3 ,and the band at 2 873 cm )1 to its symmetric mode. T he bands at 2926 and 2853 cm )1 can be a ssigned to an asymmetric and symmetric stretching mode of CH 2 , r espectively [1]. The peak shoulder present at 1740 cm )1 canbeassignedtothe ester C¼O s tretching of phospholipids [15,16], not present in DNA and proteins. Between 1700 and 1300 cm )1 , contributions are primarily due to proteins, with some Fig. 1. K5 62 cell s pectrum and spectral areas se lected by ge netic a lgo- rithm. A smear of about 2 · 10 5 cells was dried on an area o f 2 cm 2 on the germanium surface as explained in Materials and methods. Wavenumbers selected by g enetic algorithm a re in shaded. Ó FEBS 2002 K562 cell classification by FT1R (Eur. J. Biochem. 269) 1969 absorptions from lipids. The stretching o f protein amide C¼O b onds arises at 1650 cm )1 (amide I). The deformation of protein amide N–H bond appears at 1 540 cm )1 (amide II) [15]. The 1450 and 1400 cm )1 bands arise from the side chain o f p roteins [ 15], but the C –H bend ing vibration of fatty acids at 1467 and 1450 cm )1 [3] and the carboxylate vibration o f fatty acids at 1400 cm )1 [17] are s uperimposed. Absorptions between 1300 and 900 cm )1 arise mainly from phosphate associated with nucleic acids, i.e. DNA and RNA. The a bsorption bands a t 1245 a nd 1087 cm )1 are c haracteristic o f asymmetric and symmetric pho spho- diester vibration of nucleic acids [15]. In glycogen-poor cells such as lymphocytes, B enedetti et al. assigned the shoulders present at 1117 and 1020 cm )1 to RNA and DNA, respectively [1 8]. Classification by LDA LDA was applied t o discriminate t he two c ell lines. T he large number of variables ( 3000) of an infrared spectrum is a p roblem for t his a pproach that needs more observations than variables. W e attempted to reduce this number b y two distinct methods: genetic algorithm (supervised method), and PCA (unsupervised method). Classification by LDA on spectra restricted by genetic algorithm Genetic algorithm was performed on the training set composed of K562/DNR (22 spectra) and K562/DNS (26 s pectra) cells. Each spectrum was obtained for another cell culture. The 48 spectra were accumulated over a period of eight months. The region between 2800 and 1800 cm )1 , which does not contain any chemical infor- mation excepted from atmospheric CO 2 , was discarded. After 16 runs of th e algorithm, w e selected wavenumbers present i n more t han 80% of the 185 models built. They were distributed in 10 regions of the spectra (Fig. 1), including several areas in lipid and in nucleic acid regions, and one area associated with proteins (amide II). T raining set spectra were used for mo del building in L DA. The model was tested with the 30 test s pectra (not inc luded in the training set) on which the global accuracy was 73% (Table 1). About the half of the resistant spectra were classified i n the sensitive class. Classification by LDA on spectra reduced by PCA PCA was performe d on t he training set. At this stage, only two or three principal c omponents were sufficient to obtain a partial separation between the two cell lines; Fig. 2 shows the s pectra reduced with PCA proje cted on vector 2 and vector 4. Each one of these two vectors (Fig. 3) h as features at characteristic wavenumbers of nucleic acids, lipids, and protein. I t is i nteresting to note that, in the second vector, a negative influence of 1625 cm )1 (attribut- ed to a b eta sheet secondary structure of proteins) is associated with a positive value of 1 667 cm )1 (a helix secondary structure). This may reflect a m odification i n t he global secondary structure c omposition in the c ells. Reduced training set spectra was used for model building in LDA. The results obtained show 100% of co rrect classification f or the t raining set. F or the test sets ( Table 2), the global accuracy was 93%. Table 1 . Results of LDA f or spectra of t he test s e t w he n spectra were reduced b y genetic algorithm. Overall a ccuracy on the training set w as 100% and overall accuracy on the t est set is 73%. Actual assi gnments in columns, LDA p redicted assignments i n rows. K562/DNS K562/DNR Accuracy K562/DNS 7 0 100% A line K562/DNS 6 0 100% B line K562/DNR 6 5 45% A line K562/DNR 2 4 67% B line Fig. 2. Two -dimensional plot of PCA-reduced s pectra of K5 62 cells. Resistant K562 cells (39 spectra, black stars), sensitive K562 cells (39 sp ectra, circles) o f the training (full) and test s et (empty). The percentage of variance represe nted by ea ch co mpo nent is ind icated on the axes. Fig. 3. Princ ipal components w hich allow a partial separation between resistant and sensitive K562 cells. (A) Second ve ctor. (B) Fo urth vector. All components are on the same scale. 1970 A. Gaigneaux et al. (Eur. J. Biochem. 269) Ó FEBS 2002 Assignment of the spectral discrimination to the resistant phenotype As suggested by a r eviewer, it remains a possibility that the difference observed between sensitive and resistant K562 cells is the result of a natural divergence and i s not related to the multiresistant phen otype. In o rder to test this hypoth- esis, we obtained another K562 cell line (Ôcell line BÕ,as described in Materials and methods) with sensitive and resistant sublines. Spectra of these new cell lines were included in the test spectra set. Tables 1 and 2 show that spectra of the B cell line were correctly classified by the discriminant vector constructed on the A cell line in the same proportion as spectr a of the A cell line. This result indicates that t he discriminant vector constructed by LDA reflects the multiresistant phenotype in K562 cells instead of a natural variation occurring between cell lines. Biochemical origins of the difference between K562/DNS and K562/DNR spectra Figure 4 reports the m ean spectra of K562/DNS (curve a), K562/DNR (curve b), a nd the spectrum obtained b y difference between these two cells lines (sensitive cells spectrum minus its r esistant counterpart; curve c). Regions where the spectra of the two cell lines are significantly different were determined by Student t-test (shaded areas). In the CH region (3000–2800 cm )1 ), spectra display significant differences, indicating that the resistant cells have a p rotein/lipid ratio higher t han s en sitive cells (Table 3). The decrease of intensities at 1740 cm )1 (C¼O bonds of lipids) is also consistent with a decrease of lipid content in multire- sistant cells. In addition , t he ratio of CH 3 /CH 2 ,calculatedas the ratio of absorbance at 2871 cm )1 over the a bsorbance a t 2853 c m )1 , is also significantly higher in K562/DNR cells than in K562/DNS cells (Table 3). T hese res ults s uggest that a lipid/protein ratio m odification occurs in the resistant phenotype. Because proteins contain, on the average, an equal amount of methyl and methylene groups, a protein change alone would have modified the CH 3 stretching as well as the CH 2 stretching to the s ame extent. As indicated by the Student’s t-test, the difference between K 562/DNR a nd K 562/DNS cells is not significant in the amide protein. This is not surprising as spectra were normalized for equal content of proteins. But a qualitative change of the proteins, for example i n t heir global secondary structure, would have led to a modification in the shape of amide I peak (1 700–1600 cm )1 ). The K 562/DNR cells exhibit a significant decrease of absorption in the 1300–900 cm )1 region, characteristic of the protein/nucleic acid ratio. The DNA/RNA ratio, calculated as the r atio of the intensity at 1020 cm )1 (representative of DNA) by the intensity a t 1121 cm )1 (representative o f RNA), is not significantly different between K562/DNS and K562/DNR cells. The discriminant vector constr ucted by L DA with the spectra reduced by PCA is shown o n t he Fig. 4 (curve d). This vector is quite similar to the difference spectra. DISCUSSION Many studies have provided evidence that infrared spec- troscopy is a useful a nd powerful tool to screen cell and tissue evolution occurring during c ancer progression. The aim of this study was to show that among similar cancerous cell lines, a subtle change such as expression o f MDR phenotype can be identified by infrared spectrosc opy. Fig. 4 . Compariso n of spectra and s pec tral features responsible for cell type determination. (A) Mean spectrum of sensitive K562 cells. (B) Mean spectrum of resistant K562 cells. (C) Difference spectrum (sen- sitive/resistant) magnified by five. (D) Significant differences are shaded (a ¼ 1% ). Discriminant vector built by LDA. Table 2. R esults of LDA for spectra of the tes t se t w hen s pectra were reduced by PCA. Overall a ccuracy o n t he training set was 10 0% and overall accuracy on t h e test set i s 93%. Actual assignmen ts in columns, LDA predicted assignments in rows. K562/DNS K562/DNR Accuracy K562/DNS 7 0 100% A line K562/DNS 4 2 67% B line K562/DNR 0 11 100% A line K562/DNR 0 6 100% B line Table 3. R esults of Student t-tests for equality of m ean between resistant (22 spectra) a nd sensitive (26 spectra) K562 cells. Spectra were nor- malized for e qual s urface b etween 1711 and 1585 cm )1 . Lipid/P rotein: area between 3000 a nd 2800 cm )1 /area of Amide I. CH 3 /CH 2 : 2871/ 2853 cm )1 absorbance ratio. N ucleic acids/protein: 1085/1653 cm )1 absorbance r atio. DNA/RNA: 1020/1121 c m )1 absorbance ratio. K562/DNR (mean ± SD) K562/DNS (mean ± SD) P-value Lipid/protein 10.42 ± 0.71 11.48 ± 0.69 3.2 · 10 )6 CH3/CH2 0.842 ± 0.11 0.732 ± 0.071 2.8 · 10 )5 Nucleic acids/protein 0.25 ± 0.025 0.28 ± 0.02 2.8 · 10 )5 DNA/RNA 0.29 ± 0.063 0.278 ± 0.062 0.46 Ó FEBS 2002 K562 cell classification by FT1R (Eur. J. Biochem. 269) 1971 To reduce t he number of variables, we used t wo different methods, genetic algorithm and PCA. In our case, PCA seemed to be the best method. Moreover, PCA is a nonsupervised m ethod and avoids d iscarding t he greatest part of the original data information. In the second part of our work, we focused on the biochemical information available i n infrared s pectra, and we found that infrared spectra of resistant K562 cells are significantly modified relative t o their sensitive counterpart at all molecular levels. The protein/nucleic acid ratio is significantly higher in resistant cells. As discussed b efore, this can be caused either by a real decrease in n ucleic acid content o r by a n increase of the compaction state of DNA possibly related to the cell cycle stage. Concerning the latter h ypothesis, Boydston-White et al. reported that among nucleic acid absorption bonds, RNA contribution can be prominent when cells are i n G1 an d G2 ph ases, because h ighly compacted DNA is opaque to infrared light [19]. In fact, less condensed chromatin in K562 MDR cells in G1 phase (com pared to their sensitive c ounterpart in G1 cell phase) has been previously reported [20], suggesting that our experimental results can not be explained b y c hrom- atin condensation. In favour of the former hypothesis, flow cytometry studies on K562 MDR cell lines have shown a relative de crease in DNA co ntent at con stant cell phase distribution o f cells [12]. Our results a re therefore consistent with a r eal decrease i n nucleic acid content i n resistant K562 cells. At the lipid level, we observed in r esistant cells a d ecrease of the absorptions assigned to fatty a cids and phospholipids relative to their protein content. This qua ntitative change was accompanied by qualitative modification, as the methyl/methylene CH 3 /CH 2 ratio w as fou nd to increase significantly in these cells. A nuclear magnetic resonance study of these cells had s hown previously that multiresistant cells had a higher fatty acid methyl/methylene ratio when cultured in the absence of drug [11]. This feature was partially reversed after about 50 passages w ithout selection. As our c ells were se lected eve ry t wo mo nths in a medium containing doxorubicin, we did not observed such a reversion. The precise origin of the methyl/methylene ratio increase remains to be determined; it may arise only from a decrease of lipid content i n resistant cells, but can also be associated with a modification of the membrane composi- tion. The m ain modification r eported i n m ultiresistant cells is the p resence o f P -gp in the cell membrane [21]. This protein was found to have an evident c onnection with its membrane environment [22]. So far, it not clear whether the presence of P-gp modifies the membrane composition and ultrastructure as reported by Arsenault et al. on CHO cells [23], or if the s election process independently induces bo th P-gp overexpression and membrane modifications [11] (reviewedin[24]). A particularity of the ATR approach is that the cells are sampled with the evan escent fi eld, the i ntensity of which decays exponentially as a function of the distance from t he reflecting interface. In turn, spectral d ifferences may a rise from either a difference in the chemical composition or from a different distribution of the chemicals along an axis perpendicular to the reflecting interface. This feature precludes definitive conclusions on the overall chemical composition of the different cell lines. In this study, w e conclude th at infrared spectroscopy is a useful tool to identify K562 multiresistant cell lines. W e also demonstrated that infrared spectroscopy is a powerful tool for investigating the global biochemical modifications related to the multiresistant phenotype. ACKNOWLEDGEMENTS Anthoula Gaigneaux is recip ient of a Televie Grant from the Fonds National de l a Recherche Scientifique (Belgium). We thank A. Delforge and C. D orval for giving us the second K562 ce ll line. REFERENCES 1. Jackson, M., S owa, M.G. & Mantsch, H.H. ( 1997) Infrared spectroscopy: a new frontier in med icine. Biophys. C hem 68, 109–125. 2. Le Gal, J.M., Morjani, H. & Manfait, M. (1993) Ultrastructural appraisal of the multidrug r esistance in K562 and L R73 cell lines from Fourier transform infrared spectroscopy. Cancer Res. 53, 3681–3686. 3. Rigas, B. & Wong, P.T. ( 1992) Human c olon adenocarcinoma cell lines display in frared sp ectrosco pic fe atures o f m alignant c olon tissues. Cancer Res. 52, 84–88. 4. Na um ann, D., Helm, D. & Labischinski, H. (1991) Micro- biological characterization s by FT-IR spec troscopy. Nature 351 , 81–82. 5. Schultz, C.P., Liu, K., J ohnston, J.B. & Mantsch, H.H. (1996) Study of chr onic lymphocytic l euke mia cells b y FT-IR spectro- scopy and cluster a nalysis. Leuk. Res. 20, 649– 655. 6. McIntosh, L.M., Jackson, M., Mantsch, H.H., Stranc, M.F., Pilavdzic, D. & C rowson, A.N. (1999) I nfrared spectra of basal cell carcinomas are distinct from non-tumor-bearing s kin c om- ponents. J. Invest. Dermatol. 112, 951–956. 7. Haaland, D.M., Howland, D.T.J. & Thom as, E.V. ( 1997) Multi- variate classification of the infrar ed spectra o f cell and tissue samples . Appl. Spectrosc.51, 340– 345. 8. Pastan, I . & Gottesman, M. (1987) Multiple-drug resistance in human cancer. N.Engl.J.Med.316, 1388–1393. 9. Kartner,N.,Evernden-Porelle,D.,Bradley,G.&Ling,V.(1985) Detection of P-glycoprotein in multidrug-resistant cell lines by monoclonal antibodies. Nature 316, 820–823. 10. Ueda, K., Cardarelli, C., Gottesman, M.M. & Pastan, I. (1987) Expression of a f ull-length cD NA fo r the hu man M DR1 ge ne confers r esistan ce to colchicine, doxorubicin, and vinblastine. Proc. Natl Acad. Sci. USA 84 , 3004–3008. 11. LeMoyec,L.,Tatoud,R.,Degeorges,A.,Calabresse,C.,Bauza, G., Eugene, M. & C alvo, F. ( 1996) Proton nuclear m agnetic resonance spectroscopy reveals cellular lipids involved in resistance to adriamycin and t axol by t he K5 62 leukemia c ell line. Cancer Res. 56, 3461–3467. 12. Palissot, V., Liautaud-Roger, F., Carpentier, Y. & Dufer, J. (1996) Analysis of DNA c ontent in multidrug-resistant cells by image and flow cytometry. Cell Prolif. 29, 5 49–559. 13. Praet, M., Stryckmans, P. & Ruysschaert, J.M. (1996) Cellular uptake, cytotoxicity, and transport kin etics of anthrac yclines in human sensi tive and multidrug-resistant K562 cells. Biochem. Pharmacol. 51, 1341–1348. 14. Forrest, S. ( 1993) Genetic a lgorithms: principles of natural s el ec- tion applied to c om putat ion. Science 261, 872–878. 15. Diem, M., Boydston-White, S. & Chiriboga, L. (1999) infrared spectroscopy of cells and tissues: shinin g light on a n ovel subject. Appl. Spectrosc. 53, 148A –16 1A . 16. Wong, P .T., Wong, R .K., Caputo, T .A., Godwin, T .A. & Ri gas, B . (1991) Infrared spectroscopy of exfoliated human cervical c ells: evidence of extensive structural changes during carcinogenesis. Proc. Natl Acad. Sci. USA 88 , 10988–10992. 1972 A. Gaigneaux et al. (Eur. J. Biochem. 269) Ó FEBS 2002 17. J ackson, M., Choo, L.P., Watson, P.H., Halliday, W.C. & Mantsch, H.H. (1995) Beware of connective tissue proteins: assignment and implications of collagen absorptions in infrared spectra of h uman tissues. Biochim. Biophys. Acta 1270,1–6. 18. B enedetti, E ., Bramanti, E ., Papineschi, F . & Rossi, I . ( 1996) Determination of the relative amount of nucleic aion of t he relative amount of nucleic acids in leukemic and normal lymphocytes by means of FT-IR microspectroscopy. Appl. Spec- trosc. 51, 792–797. 19. B oydston-White , S., Gopen, T., Houser, S., Bargonetti, J . & Diem, M. ( 1 999) In frared sp ec troscop y of human tissue. V . Infrared spectroscopic s tudies of my eloid leukemia ( ML-1 ) cells at different phases of t he cell c ycle. Biospectroscopy 5, 2 19–227. 20. D ufer, J., Millot-Broglio, C., Oum’Hamed, Z., Liautaud-Roger, F., Joly, P., Desplaces, A. & Jardillier, J.C. (1995) Nuclear DNA content a nd chromatin texture in multidrug-resistant human l eu- kemic c ell lines. I nt. J. Cancer 60, 108 –114. 21. Nielsen, D. & Skovsgaard, T . (1992) P-glycoprotein as multidrug transporter: a critical review of current multidrug resistant cell lines. Biochim. Biophys. Acta 1139, 169–183. 22. W adkins, R .M. & Ro ep e, P.D. (1997) Biophysical a spects of P-glycoprotein-mediated m ultidrug resistance. Int. Rev. Cytol. 171, 121–165. 23. Arsenault, A.L., Ling, V. & Kartner, N. (1988) Altered plasma membrane ultrastructure in multidrug-resistant cells. Bioc him. Biophys. Ac ta 938, 315–321. 24. Ferte, J. (2000) Analysis of the t angled relationships between P-glycoprotein-mediated multidrug re sistance and the lipid phase of the cell membrane. Eu r J. Biochem. 267, 2 77–294. Ó FEBS 2002 K562 cell classification by FT1R (Eur. J. Biochem. 269) 1973 . Infrared spectroscopy as a tool for discrimination between sensitive and multiresistant K562 cells Anthoula Gaigneaux, Jean-Marie Ruysschaert and Erik. indicated that discrimination between r esistant and sensitive cells was based o n variations in all cellular contents. Lipid and nucleic acid decreased,