FTIR and Raman microspectroscopy of normal ... - Semantic Scholar

1 downloads 0 Views 304KB Size Report
In: Clark. RJH, Hester RE (eds) Biomolecular spectroscopy, Part A (Advances in spectroscopy series, vol 20). Wiley, Chichester, UK, ch 3, p 129. 1656.
Anal Bioanal Chem (2007) 387:1649–1656 DOI 10.1007/s00216-006-0827-1

ORIGINAL PAPER

FTIR and Raman microspectroscopy of normal, benign, and malignant formalin-fixed ovarian tissues C. Murali Krishna & G. D. Sockalingum & Rani A. Bhat & L. Venteo & Pralhad Kushtagi & M. Pluot & M. Manfait

Received: 5 July 2006 / Revised: 30 August 2006 / Accepted: 1 September 2006 / Published online: 17 October 2006 # Springer-Verlag 2006

Abstract Ovarian cancer is the sixth most common cancer among women worldwide, and mortality rates from this cancer are higher than for other gynecological cancers. This is attributed to a lack of reliable screening methods and the inadequacy of treatment modalities for the advanced stages of the disease. FTIR and Raman spectroscopic studies of formalin-fixed normal, benign, and malignant ovarian tissues have been undertaken in order to investigate and attempt to understand the underlying biochemical changes associated with the disease, and to explore the feasibility of discriminating between these different tissue types. Raman spectra of normal tissues indicate the dominance of proteins and lower contents of DNA and lipids compared to malignant tissues. Among the pathological tissues studied, spectra from benign tissues seem to contain more proteins and less DNA and lipids compared to malignant tissue C. M. Krishna (*) Center for Laser Spectroscopy, Manipal Academy of Higher Education, Manipal 576104, India e-mail: [email protected] C. M. Krishna : G. D. Sockalingum : M. Manfait Unité MéDIAN, CNRS UMR 6142, UFR Pharmacie, IFR 53, Université de Reims Champagne-Ardenne, 51 rue Cognacq-Jay, 51096 Reims Cedex, France R. A. Bhat : P. Kushtagi Department of Obstetrics and Gynecology, Kasturba Medical College, Manipal Academy of Higher Education, Manipal 576104, India L. Venteo : M. Pluot Laboratoire Central d’Anatomie et Cytologie Pathologiques, CHU de Reims, Hôpital Robert Debré, Avenue du Général Koenig, 51092 Reims Cedex, France

spectra. FTIR studies corroborate these findings. FTIR and Raman spectra of both normal and benign tissues showed more similarities than those of malignant tissues. Cluster analysis of first-derivative Raman spectra in the 700– 1700 cm−1 range gave two clear groups, one corresponding to malignant and the other to normal+benign tissues. At a lower heterogeneity level, the normal+benign cluster gave three nonoverlapping subclusters, one corresponding to normal and two for benign tissues. Cluster analysis of second-derivative FTIR spectra in the combined spectral regions of 1540–1680 and 1720–1780 cm−1 resulted into two clear clusters corresponding to malignant and normal+ benign tissues. The cluster corresponding to normal+benign tissues produced nonoverlapping subclusters for normal and benign tissues at a lower heterogeneity level. The findings of this study demonstrate the feasibility of Raman and FTIR microspectroscopic discrimination of formalin-fixed normal, benign, and malignant ovarian tissues. Keywords Ovarian cancers . Optical diagnosis . Raman . FTIR . Formalin-fixed tissues . Cluster analysis Introduction Cancer of the ovary is the sixth most frequent cancer among women worldwide. Mortality rates of this cancer are the highest among other gynecological cancers, with a fiveyear survival rate of 30%. It has been shown that early diagnosis of the disease can lead to better survival. Present treatment modalities can cure 90% of stage IA and 70% of stage II ovarian cancers. Due to lack of reliable screening methods, the disease is most often diagnosed at a very advanced stage (III or IV). Early menarche, late menopause, low parity, high-fat diet, and use of estrogen for relief of menopausal symptoms are implicated as risk

1650

factors. [1–3]. The development of new methods for the early detection and screening of ovarian cancer is urgently required due to the lack of reliable population screening methods for the early detection of curable stages of the disease and prognosis [2, 3]. The detection of neoplastic changes by optical spectroscopy techniques such as FTIR, Raman, and fluorescence spectroscopy, has been one of the most active areas of recent research into the discrimination of oral, cervical, breast and other cancers [4–10]. These methods are more objective, less time-consuming, and have the ability to be applied in vivo. A review of the literature shows that there are very few reports of studies on ovarian cancer diagnosis by optical spectroscopy [11–13]. A recent study has demonstrated that ovarian cancers can be discriminated based on the tissue autofluorescence pattern [12]. Diagnosis of ovarian cancers by proteomics-based methods has also been reported [13]. In a recent study, we have demonstrated the discrimination of normal and malignant ovarian tissues handled ex vivo by Raman microspectroscopy [14]. In the present study, we have carried out a combined approach utilizing both Raman and FTIR spectroscopy in order to analyze normal, malignant, and benign formalinfixed ovarian tissues. The aims of the study are to evaluate the efficacy of the two complementary technologies for discriminating tissue types and also to investigate and attempt to understand the underlying biochemical variations in normal, benign, and malignant ovarian tissues. Hierarchical cluster analysis (HCA) was used to classify tissue types.

Materials and methods Tissue samples Ovarian tissue specimens in 10% formaldehyde (formalin) were collected from routine biopsies or surgical resections submitted for histopathological examinations. A total of 24 specimens comprising eight normal, ten benign and six malignant tissues were recruited in these exploratory investigations. Fourteen micron-thick sections, obtained by freezing microtome, were investigated by Raman as well as FTIR microspectroscopy. Samples were dried under mild vacuum conditions for ten minutes, before being investigated by microspectroscopy. In each case another adjacent section was stained by H and E. The stained sections were used for pathological verification and also to identify regions of interest, i.e., epithelial sections. Each tissue specimen employed in this study was subjected to two independent pathological examinations in order to ascertain the pathological status. Only clear malignant, benign and normal specimens were recruited for spectroscopic investigations and data analysis.

Anal Bioanal Chem (2007) 387:1649–1656

Raman microspectroscopy Micro-Raman spectra were recorded on a commercial Raman microspectrometer (LabRam, Jobin Yvon Horiba, Longjumeau, France), which is described in detail elsewhere [15]. In this set-up, the 833 nm radiation from an Ar+ laser-pumped titanium:sapphire tunable laser (Spectra Physics, Mountain View, CA, USA) was used for sample excitation. The laser power at the sample was ∼60 mW. A combination of a 900 gr/mm grating and a CCD detector (1024×256 pixels) was used for data collection. An average of three spectral acquisitions with integration time of 30 s was used for each spectral point. These parameters were maintained throughout the study. The spectrograph was calibrated using the Raman signal from a silicon wafer. Spectra were recorded in the 700–1700 cm−1 region from epithelial sections of the tissues. Probing areas in tissue sections were selected under pathological guidance. It is well-known that point spectra recorded by a microspectrometer, due to the high spatial resolution of the instrument, depend very much on the probing area, for example a nucleus or the extracellular matrix. Therefore, even spectra recorded from two adjacent sites can be very different [16]. This can be a minor inconvenience if one is interested in studying overall variations (the cells and microenvironment in a tissue section), as in this study. To overcome this, the mean spectrum from several spectra that are recorded from several sites within the area of interest can be computed and used to represent the area [14, 17]. Raman spectra were collected from 100×100 μm2 sites at intervals of 10 μm. The coadded spectrum calculated from 100 spectra was taken as being representative of the area of interest. Figure 1 displays a typical white-light image of normal ovarian tissue. The area probed is indicated by a square. The mean spectra obtained from several areas in epithelial regions were used for data analysis.

Fig. 1 Typical white light image of normal ovarian tissue section. The area probed is indicated by a square

Anal Bioanal Chem (2007) 387:1649–1656

1449

a

1297

1443

1.0

0.5

808

Formalin fixation of tissue samples is the most widely used procedure for storing samples in histopathology. Vibrational spectroscopic studies of formalin-fixed samples are welldocumented [14, 17–20]. Spectral contamination from bands at 1041 and 1492 cm−1 due to formalin was observed during conventional Raman spectroscopy of whole tissues. This contamination was reduced considerably by thoroughly washing the specimens in saline [18, 19]. Raman microspectroscopy of thoroughly washed tissue sections did not show any contribution from formalin [14, 17]. Mean micro-Raman spectra of normal, benign, and malignant ovarian tissues are shown in Fig. 1. The spectra exhibit significant differences between their profiles. The

Intensity (a.u.)

Results and discussion

1662

718 760 818

1.5

1241 1296 1338

Baseline subtraction of Raman spectra was performed by fitting and subtracting a third-order polynomial function, and spectra were vector-normalized using the whole spectral region. FTIR spectra were corrected using an elastic scattering correction algorithm and then vector-normalized. Corrected mean spectra were subjected to hierarchical cluster analysis (HCA) using Ward’s algorithms of OPUS NT (Bruker, Reinstetten, Germany). In this analysis, data points from the spectra were compared based on the Euclidean distances between them. Using these distances, the spectra were classified and a dendogram was created. Mean spectra were computed from the clusters corresponding to each tissue type and used as representative spectra in order to attempt to understand the biochemical variations between tissues.

1205

Data preprocessing and analysis

851 919 935 1002 1031 1057 1098 1126

The same sites of the same tissue sections as those investigated with Raman microspectroscopy were also analyzed by FT-IR microspectroscopy. Spectral images were collected in transmission mode by an FTIR imaging system (Spotlight, Perkin-Elmer, Courtaboeuf, France) coupled to a spectrometer (Spectrum 300, PerkinElmer). The imaging system was equipped with a liquid N2-cooled MCT linear detector comprising 16 pixel elements. In this study, FTIR images were collected in a raster-scan mode from selected tissue sites in the 4000–700 cm−1 range with a spectral resolution of 4 cm−1 and spatial resolution of 6.25 μm/pixel. As in the micro-Raman investigations, regions of the interest in the tissue sections were selected under the pathologist’s guidance. Spectra were recorded from several sites on the tissue sections and an average spectrum from the spectra recorded at each site was computed and used to represent the given area for further data analysis.

normal tissue spectrum, shown in Fig. 2a, is characterized by strong protein contributions, as revealed by strong amide I and amide III bands, vibrational modes of C–C backbones, and amino acid side-chains. Further, the spectrum also suggests the presence of helical (919 and 933 cm−1) and disordered (1242 cm−1) structures. For benign tissue, the mean spectrum (shown in Fig. 2b) exhibits similarities to that of the normal tissue spectrum. The mean malignant microRaman spectrum, Fig. 2c, shows several differences from that of normal tissue. The prominent differences are: a decrease in the intensity of the amide I band, a minor blue shift of the δCH2 band, and changes in the relative intensities of several bands in the 800–1200 cm−1 region and the amide III region. In order to interpret these spectral differences and to understand the biochemical variations in malignant and benign tissues with respect to normal tissue, difference spectra were computed by subtracting the mean normal tissue spectrum from those of benign and malignant tissue spectra, as shown in Fig. 3a and b, respectively. The intensity of the difference spectrum between benign and normal tissues, Fig. 3a, is about twofold weaker than that between malignant and normal tissues, Fig. 3b. The presence of higher protein content in normal tissues with respect to benign and malignant tissues is indicated by the negative peaks centered around 760, 876, 937, 1001, 1034, 1240, 1329, 1458 and 1655 cm−1 in the difference spectrum of benign and normal tissues (Fig. 3a) and around 761, 873, 937, 1001, 1031, 1242, 1456, 1638 and 1655 cm−1 in the difference spectrum between malignant and normal tissues (Fig. 3b). The positive peaks at 1080, 1136, 1300, and 1433 cm−1 in the difference spectrum between benign and normal tissues, Fig. 3a, suggest the presence of excess

b

889 920

FTIR microspectroscopy

1651

c 800

1000

1200

1400

1600

Wavenumber (cm-1) Fig. 2 Mean micro-Raman spectra of ovarian tissues: a normal, b malignant, and c benign

1652

Anal Bioanal Chem (2007) 387:1649–1656

Fig. 3a–c Difference microRaman spectra of ovarian tissues: a benign–normal, b malignant–normal, and c malignant–benign

lipids in benign tissues. The other weak positive peaks in the spectrum at 1340 and 1477 cm−1 can be attributed to excess DNA content in benign tissues. The positive peaks in the difference spectrum (Fig. 3b) between malignant and normal tissues may indicate the presence of an excess of biomolecules such as DNA (features around 805, 1356 and 1480 cm−1) and lipids (bands at 1085, 1135, 1295 and 1433 cm−1). Conformational differences in the protein contents of malignant and normal tissues are also indicated by a decrease in the intensity of bands around 937, 1638 and 1666 cm−1, and the appearance of a new band at 893 cm−1. Similarly, the biochemical differences between benign and malignant tissues are also shown by the difference spectrum (Fig. 3c). The negative peaks at 726, 761, 873, 938, 1001, 1030, 1238, 1454, 1637 and 1668 cm−1 suggest higher protein content in benign tissues. Positive features at 797, 1330 and

1480 cm−1 could be due to DNA, and other positive peaks at 1080, 1125, 1299, and 1430 cm−1 can be assigned to lipids. Conformational changes in proteins of benign and malignant tissues are indicated by negative peaks at 938 cm−1,1238, 1637 and 1668 cm−1, and a positive band at 893 cm−1. The various vibrational modes in ovarian tissue spectra are assigned based on available literature data [21, 22]. Typical mean and normalized FTIR spectra of normal, benign, and malignant tissues are shown in Fig. 4a–c, respectively. The spectra exhibit pronounced differences between normal and malignant tissues, whereas differences between spectral profiles of normal and benign spectra are less marked. Prominent differences between normal and benign tissues can be seen in the polysaccharide (1000– 1200 cm−1) and protein (1500–1700 cm−1) absorbing regions. The differences between the spectra of malignant

1734

1655

1544

1375 1400 1452

1653

a

0.5

1373

1091

1.0

b

828 857

Intensity (a.u.)

1.5

1204 1238 1281

1033 1095

Anal Bioanal Chem (2007) 387:1649–1656

c

0.0 800

1000

1200

1400

1600

Wavenumber (cm-1) Fig. 4 Mean FTIR spectra of ovarian tissues: a normal, b malignant, and c benign

tissue and normal tissue are: induction of new bands at 828 and 857 cm−1, increase in the relative intensities of DNA bands in the 1000–1200 cm−1 region and the lipid band at 1734 cm−1, and decreases in the intensities of amide I and amide II bands. We have also characterized the differences in FTIR spectra between normal and pathological tissues. As in the case of Raman spectra, we subtracted the normal spectrum from the benign and malignant spectra, Fig. 5a and b. The differences between benign and normal were twofold weaker than those between malignant and normal tissues, as was observed for corresponding Raman spectra. The difference spectra, Fig. 5a and b, suggest low protein content in pathological tissues with respect to normal tissue, as indicated by the negative amide I and amide II peaks at 1652 cm−1 and 1544 cm−1 in the difference spectrum between benign and normal tissues (Fig. 5a) and at 1657 cm−1 and 1553 cm−1 in the difference spectrum between mlignant and normal tissues (Fig. 5b), respectively. High DNA and lipid contents in benign tissue also observed with positive peaks at 1035, 1086 and 1131 cm−1 (Fig. 5a). The positive peaks in the difference spectrum between malignant and normal tissues, shown in Fig. 5b, indicate higher contents of several molecular species in malignant tissue, such as lipids (1738 cm−1), DNA (1092, 1248 cm−1), and glycogen (1045 and 1074 cm−1), and in the fingerprint region (828 and 845 cm−1). The difference spectrum between malignant and benign tissues, shown in Fig. 5c, exhibits a similar tendency as seen in Fig. 5b (same spectral profile), except for the redshift of amide I (1663 cm−1) and the increase in DNA content. It is well-known that Raman and FTIR spectroscopy are complementary techniques. Therefore, the spectral features are expected to be complementary as well as corroborative. This is evident from Figs. 2, 3, 4, 5. As can be seen from

these figures, FTIR and Raman spectra of normal and benign tissues seem to have more similarities, whereas malignant spectra are very different from those for both of these tissues. Further, normal tissue spectra show higher protein content than spectra for pathological tissues. Malignant tissues seem to contain higher levels of nucleic acids and lipids with respect to both benign and normal tissues. A difference in content of secondary structures of proteins between malignant and normal tissues as well as between benign and malignant tissues is also apparent. Thus, the findings of this study indicate the potential of the combined FTIR and Raman approach to aid in our understanding of the biochemical variations between different tissue conditions: in our case normal, benign, and malignant tissues. Objective diagnosis is one of the most attractive aspects of optical pathology, as conventional pathology can often be highly subjective. Furthermore, the data are amenable to multivariate statistical analysis for pattern recognition and clustering. Unsupervised and supervised methods such as PCA, HCA, LDA, and ANN have been widely used in optical pathology. In the present study, HCA was used for data analysis. HCA was tested under different conditions (over the entire spectral range, over selected regions, for first- and second-derivative spectra), in order to achieve a better discrimination between normal, benign, and malignant tissues. In our analysis, first-derivative Raman spectra in the 700–1700 cm−1 region have resulted into two very clear clusters corresponding to malignant and to benign plus normal tissues, respectively, at a heterogeneity level of 7, as shown in Fig. 6. Further, at a heterogeneity level of around 2, three different clusters can be observed, one corresponding to normal tissue and the other two subclusters corresponding to benign tissues. All the clusters were clear, with no overlap among spectra corresponding to different tissues. In the case of FTIR data, secondderivative spectra based on the combined spectral regions of 1540–1680 plus 1720–1780 cm−1 gave two clear clusters corresponding to malignant and normal plus benign tissues at a heterogeneity level of 4, (Fig. 7). Further subclustering among benign and normal spectra can be seen at a lower heterogeneity level. However, there was no overlap of spectra among subclusters corresponding to benign and normal tissues. For both Raman and FTIR data, the observed subclustering of normal and benign tissues could be due to minor spectral variations. Thus, the results indicate the feasibility of employing formalin fixed tissues in optical pathology of ovarian cancers.

Conclusions The Raman and FTIR spectra obtained from normal and benign tissue show similarities, whereas spectra from

1654

Anal Bioanal Chem (2007) 387:1649–1656

Fig. 5a–c Difference FTIR spectra of ovarian tissues: a benign-normal, b malignant-normal, and c malignant-benign

Anal Bioanal Chem (2007) 387:1649–1656

1655

Fig. 6 HCA of first-derivative mean Raman spectra of ovarian tissues calculated in the 700–1700 cm−1 region

malignant tissues are very different to these. Normal tissue spectra are characterized by higher protein contents, while the more DNA and lipid signals are exhibited by malignant tissues. Among pathological tissues, malignant tissues seem to contain higher levels of lipids and DNA, and lower levels of proteins compared to benign tissues. HCA of firstderivative Raman spectra and second-derivative FTIR spectra gave good delineation of malignant from normal and benign tissues. Normal and benign spectra could be further clustered into subgroups at the lower heterogeneity levels. Thus, the findings of this study indicate the Fig. 7 HCA of second-derivative mean FTIR spectra of ovarian tissues in combined regions of 1540–1680 and 1720–1780 cm−1

feasibility of employing formalin-fixed tissues in optical pathology of ovarian cancers. The findings of this study may have considerable significance, since there are currently no reliable screening procedures for early diagnosis of ovarian cancers, and present procedures used to treat advanced stages of the disease are inadequate. Prospectively, the results obtained by Raman spectroscopy are of considerable significance from an in vivo diagnosis/screening point of view, as this type of pathological site can be accessed by an endoscope.

1656

Anal Bioanal Chem (2007) 387:1649–1656

Acknowledgement One of the authors (CMK) is thankful to the University of Reims for a visiting scientist position. 11. 12.

References 1. Notani PN (2001) Curr Sci 81:465–474 2. Hoskins WJ (1995) J Cell Biochem Suppl 23:189–199 3. Brewer MA, Johnosn K, Follen M, Gershenson D, Bast R Jr (2003) Clinical Cancer Res 9:20–30 4. Hanlon EB, Manoharan R, Koo TW, Motz JT, Fitzmurice M, Kramer JR, Itzkan I, Dasari RR, Feld MS (2000) Phys Med Biol 45:R1–R59 5. Min YK, Yamato T, Kahada E, Ito T, Hamaguchi H (2005) J Raman Spectrosc 36:73–76 6. Koljenovic S, Schut TCB, Van Meerbeeck JP, Maat APWM, Burgers SA, Zondevvan PE, Kros JM, Puppels GJ (2004) J Biomed Opt 9:1187–1197 7. Diem M, Romeo M, Boydston-White S, Miljkovic M, Matthaus C (2004) Analyst 129:880–885 8. Mantsch HH, Choo Smith LP, Shaw RA (2002) Vib Spectrosc 30:31–41 9. Wagnieres G, Star WM, Wilson B (1998) Photochem Photobiol 68:603–632 10. Kartha VB, Kurien J, Lavanya Rai, Mahato KK, Murali Krishna C, Santhosh C (2005) Diagnosis at the molecular level: Analytical laser spectroscopy for clinical applications. In: Koneco S, Viswanathan B, Funasaka K (eds) Photoelectrochemistry and

13.

14. 15. 16. 17.

18.

19. 20. 21. 22.

photobiology in environment, energy and fuel. Research Signpost, Trivendrum, India, pp 153–221 Brewer M, Utzinger, Silva E, Gershenson D, Bast RC Jr, Follen M, Wharton JT, Richard-Kortum R (2001) Laser Surg Med 29:128–135 Malins DC, Polissar NL, Schaefer S, Su Y, Vinson M (1998) Proc Natl Acad Sci USA 95:7637–7642 Petricoin EF, Ardekani AM, Hitt BA, Levine PJ, Fusaro VA, Steinberg SM, Mills GB, Simone C, Fishaman DA, Kohn EC, Loitta LA (2002) Lancet 359:572–577 Murali Krishna C, Sockalingum GD, Veneto L, Bhat RA, Pralhad Kustagi, Pluot M, Manfait M (2005) Biopolymers 79:269–276 Piot O, Autran JC, Manfait M (2002) Appl Spectrosc 56:1132–1138 Puppels GJ, de Mul FFM, C Otto, Greve J, Robert-Nicoud M, Arndt-Jovin DJ, Jane TM (1990) Nature 347:301–303 Murali Krishna C, Sockalingum GD, Jacob Kurien, Lakshmi Rao, Venteo L, Pluot M, Manfait M, Kartha VB (2004) Appl Spectrosc 58:107–114 Podshyvalov A, Sahu RK, Mark S, Kantarovich K, Guterman H, Goldstein J, Jagannathan R, Argov S, Mordechai S (2005) Appl Opt 44:3725–3734 Huang Z, McWilliams A, Lam S, English J, McLean DI, Lui H, Zeng H (2003) Int J Oncol 23:649–655 Kaminaka S, Ito T, Yamazaki H, Khoda E, Hamaguchi H (2002) J Raman Spectrosc 33:498–502 Parker ES (1983) Applications of infrared and Raman and resonance Raman spectroscopy in biochemistry. Plenum, New York Tonge PJ, Carey PR (1993) Raman, resonance Raman and FTIR spectroscopic studies of enzyme–substrate complexes. In: Clark RJH, Hester RE (eds) Biomolecular spectroscopy, Part A (Advances in spectroscopy series, vol 20). Wiley, Chichester, UK, ch 3, p 129