Glioblastoma

Diagnostic Potential of Autofluorescence for an Assisted Intraoperative Delineation of Glioblastoma Resection Margins (para)

ABSTRACT

The intrinsic autofluorescence properties of biological tissues can be affected by the occurrence of histological and biochemical alterations induced by pathological processes. In this study the potential of autofluorescence to distinguish tumor from normal tissues was investigated with the view of a real-time diagnostic application in neurosurgery to delineate glioblastoma resection margins. The autofluorescence properties of nonneoplastic and neoplastic tissues were analyzed on tissue sections and homogenates by means of a microspectrofluorometer, and directly on patients affected by glioblastoma multiforme, during surgery, with a fiber-optic probe. Scan-- microspectrofluorometric analysis on tissue sections evidenced a reduction of emission intensity and a broadening of the main emission band, along with a redshift of the peak position, from peritumoral nonneoplastic to neoplastic tissues. Differences in both spectral shape and signal amplitude were found in patients when the glioblastoma lesion autofluorescence was compared with those of cortex and white matter taken as healthy tissues. Both biochemical composition and histological organization contribute to modify the autofluorescence emission of neoplastic, with respect to nonneoplastic, brain tissues. The differences found in the in vivo analysis confirm the prospects for improving the efficacy of tumor resection margin delineation in neurosurgery.

Abbreviations: FAD, flavin adenine dinucleotide; FWHM, full width at half intensity maximum; GBL, glioblastoma lesion: GBM, glioblastoma multiforme; GMG, half-gaussian modified gaussian; NA, numerical aperture; NAD(P)H, reduced nicotinamide adenine dinucleotide (phosphate); PBS, phosphate-buffered saline; SD, standard deviation; WHO, World Health Organization.

INTRODUCTION

Several biomolecules involved both in functional and metabolic processes (coenzymes, flavins, lipopigments, porphyrins) and in histological tissue organization (constitutive proteins) act as endogenous fluorophores (1,2), giving rise to a fluorescence emission (autofluorescence) covering the visible range upon excitation in the UV-blue spectral region. Because autofluorescence emission is related to the nature, relative amount, spatial distribution and microenvironment of the endogenous fluorophores, the occurrence of pathological conditions affecting histological and histochemical features of tissues is expected to result in an alteration of the autofluorescence emission properties. On this assumption tissue autofluorescence is currently considered as a possible parameter for in situ cancer diagnosis in different tissues and organs, such as colon, bronchi, cervix and esophagus (3-13).

An interesting, recently considered field of application concerns brain tumors. Of these tumors glioblastoma represents the most important group, constituting about 50% of all gliomas. This tumor is formed from the glial (supportive) tissue of the brain. The common site of its occurrence is the deep white matter in the frontal lobes of the cerebral hemisphere. The tumor is morphologically and biologically heterogeneous and is often characterized by a multifocal localization (14). Quality of life and time of tumor recurrence in patients with malignant glioblastoma have been shown to improve with accuracy of the surgical resection. Unfortunately, most of these tumors do not exhibit an easily distinguishable boundary on the basis of their gross anatomy, thus making a complete resection difficult to obtain. This can result either in an incomplete resection of the tumor mass, the main cause of early recurrence of the disease, or in an excessive resection of the healthy tissue, causing additional neurological deficits.

Owing to the dependence of its properties on the biochemical and histological characteristics of biological tissues, autofluorescence can represent an important approach to improve the accuracy and therefore the efficacy of the surgical resection.

Fluorescence spectroscopy has already been proposed as a diagnostic technique for the intraoperative delineation of tumor resection margins based on the emission properties of both exogenous (15-18) and endogenous (19-22) fluorophores.

As for human brain autofluorescence, very few studies have been conducted on the feasibility of optical spectroscopy for tumor brain demarcation. Chung et al. (20) reported that the autofluorescence signals were lower in gliomas than in normal brain tissue, in ex vivo samples under excitation in the range 360-490 run. An alteration of the metabolic activity in the tumor with respect to normal tissue was proposed to justify the signal reduction. Lin et al. (21) reported an in vivo study that showed that the absolute autofluorescence intensity alone could not be used to differentiate between normal and tumoral brain tissue at 337 nm excitation, although some differences were observed. Under this excitation condition, a discrimination algorithm based on the ratio of the fluorescence and the reflectance intensity at 625 nm is to be used to differentiate tumor from normal tissues with good accuracy. Preliminary results obtained by our group on a limited number of patients during surgery (22) suggested that normal and tumor tissue could be distinguished on the basis of their autofluorescence properties by exciting at 366 nm.

The aim of this work was to characterize the spectroscopic properties of brain tissues on ex vivo samples and in vivo, in patients with glioblastoma multiforme (GBM) during surgery, to (1) define the biochemical basis of a different autofluorescence response between normal and tumor tissues by means of a curvefitting procedure on spectrofluorometric data from tissue homogenates and sections; and (2) verify the potential of autofluorescence to differentiate brain tumor from normal brain tissue in vivo.

MATERIALS AND METHODS

Ex vivo study. This study was performed on brain neoplastic lesions histopathologically diagnosed as glioblastoma Grade IV according to the guidelines of the World Health Organization (WHO) (23). Biopsy samples obtained intraoperatively from a total of 20 patients were considered in this study. At least three biopsy samples were obtained from each patient.

Specimens (dimensions ranging roughly from 3 x 3 x 3 mm to 7 x 7 x 7 mm) were frozen at -180 deg C (liquid nitrogen) soon after surgical resection and stored at -80 deg C until they were processed for homogenization or mounted in a cryostat for section cutting.

Tissue homogenates. Biopsy samples both of nonneoplastic tissue and neoplastic tissue from three patients (two biopsies for each tissue type from each patient) were used to obtain pure homogenates of the two tissue types (see Table 1). Before homogenization, histological analysis was performed on haematoxylin-eosin-stained sections obtained from different parts of the samples to verify neoplastic or nonneoplastic tissue types.

Parts of the biopsy samples were homogenized at moderate ionic strength in 0.005 M phosphate buffer (pH 7.4; 200 mg of tissue/1 mL of solution) containing an antioxidant (dl-dithiothreitol, 20 mM), a proteolytic inhibitor (phenylmethane sulphonyl fluoride, 1 mM), a chelating agent (ethylenediaminetetraacetic acid, 5 mM) and a detergent (Triton X-100, 0.01% wt/vol) for protein solubilization, according to a previously described procedure (4). Tissue homogenates were centrifuged at 5000 g for 15 min, and supernatants were submitted to spectral analysis. The interference of the buffer components with fluorescence measurements was previously verified to be negligible (4).

To validate spectrofluorometric data on reduced nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) content, an NAD(P)H biochemical assay was performed. To this end, parts of the biopsy samples from the same three patients were submitted to extraction at 85 deg C in 1 N NaOH, according to the procedure proposed by Glock and McLean (24).

Tissue sections. Sections of tumor tissue (25 lan thickness) with surrounding areas of peritumoral, nonneoplastic tissue were obtained at cryostat from frozen biopsy samples from 15 of the 20 patients studied (see Table 1). Histology of the areas measured on unstained sections were identified by referring to images of serial 8 pm thick tissue sections stained with haematoxylin-eosin. Tumor boundary was identified based on cell morphology analysis.

Sections, mounted with a small drop of phosphate-buffered saline (PBS), were submitted to microspectrofluorometric analysis immediately after they were cut, in the absence of any further treatment or staining.

Ex vivo spectral analysis. Fluorescence emission spectra of tissue homogenates were recorded from supernatant drops put on a microscope slide by means of a Leitz microspectrograph (Wetzlar, Germany) equipped with an Optical Multichannel Analyzer (OMA, EG&G, Princeton Applied Research, Princeton. NJ), with a Jarrell-Ash Monospec 27 spectrograph (Allied Analytical System, Waltham, MA; mod. 82-499, H150 g/mm grating) and a 512-element intensified diode array detector (mod. 1420/ 512). Fluorescence excitation light, under epi-illumination condition, was provided by a 100 W Hg lamp (Osram, Berlin, Germany) combined with KG1 and BG38 antithermal filters. Excitation wavelengths of 366 and 405 nm were selected, which fitted with the excitation spectra of the fluorophores mainly responsible for brain tissue autofluorescence (reduced nicotinamide adenine dinucleotide (phosphate) (NAD(P)H), flavins and lipopigments). The following excitation conditions were used: 366 nm interference filter (T% = 25) and 390 nm dichroic mirror (T%^sub 366^

Autofluorescence spectra from tissue sections were recorded by means of the apparatus described under the same experimental conditions. The microspectrofluorometric measurements were taken with a Leitz objective 40 x (NA 0.65) on fixed areas (60 (mu)m^sup 2^, selected with an iris diaphragm) for a reliable quantitative comparison of the fluorescence amplitudes measured on different tissue regions.

At least three tissue sections from each of the 49 biopsy samples from the 15 patients considered were analyzed. A minimum of three independent measurements were taken for each type of tissue of each section, giving a total of at least 10 sites per tissue type per biopsy sample.

Each spectral acquisition lasted for 10 sequential scans of 200 ms each, for a total measuring time of 2 s. The time required for handling and measuring of each tissue section did not exceed 17 min. Preliminary control experiments were performed to evaluate the effects of both photobleaching and time after section cutting on autofluorescence intensity and spectral shape. The tissue section was picked up on a slide immediately after cutting, mounted with a small drop of PBS and finally placed under the microscope objective and submitted to measurements within 15 s after cutting.

Photobleaching was evaluated by measuring the emission intensity on the same area exposed to continuous irradiation at 366 run. Fluorescence intensity was monitored at 440 and 520 nm, corresponding to the NAD(P)H and flavin adenine dinucleotide (FAD) emission wavelengths, respectively. For both wavelengths the fluorescence appeared to decay exponentially. The solid and broken lines in Fig. 1 are the best-fit exponentials to the 440 and 520 nm data, respectively. The two curves are described by the functions FI^sub 440nm^(t) = 1028 + 1998 exp(-t/125) (where FI is the fluorescence intensity), with r^sup 2^ = 0.997, and F1520 nm0) = 903 + 1102 exp(-t/111). with r^sup 2^ = 0.993, implying 1/e decay times of 125 and Ill s for NAD(P)H and FAD, respectively. No appreciable differences were found between normal and tumor tissues. The 1/e decay time values calculated indicate that neither an appreciable intensity reduction nor a spectral distortion occurred during measuring time (2 s).

The fluorescence spectra of tissue sections were monitored as a function of time after section cutting in different tissue regions to avoid any photobleaching effect. Because no appreciable change of the autofluorescence spectral shape was detected as a function of time after cutting, the emission intensity integrated in the 440-540 nm spectral range was considered. The solid line of Fig. 2 is the best-fit exponential to the FI^sub 440-540 nm^ data of tumor tissue. The curve is described by the function Faao-540 nm(t) = 108 + 144 exp(-t/156), with r^sup 2^ = 0.994, resulting in a lle decay time of 156 min. Such a decay time value implies a fluorescence intensity reduction not exceeding 10% at the maximum time (17 min) required for analysis of each tissue section. Comparable results were obtained for nonneoplastic tissue.

It is worth noting that in the absence of PBS mounting media, an increase in intensity is observed at a time later than about 30 min, which can be attributed to the fact that the dry condition of the tissue section resulting from exposure to air favors fluorescence efficiency.

Ex vivo fluorescence imaging. Fluorescence images were acquired by means of an Argus VIM 100 processor digital system (Hamamatsu Photonics Deutschland GmbH, Herrsching am Ammersee, Germany), using an ISIT camera (Hamamatsu C2400-09) coupled to a Leitz fluorescence microscope. The excitation was performed with a 366 nm interference filter; images were acquired through a 430 nm barrier filter. For presentation, images were processed with of Scion Image software (Scion Corporation, Frederick, MD).

Spectral fitting analysis. Autofluorescence spectra recorded on both tissue homogenates and tissue sections under excitation at 366 nm were submitted to a curve-fitting analysis to evaluate the relative contribution of each fluorophore involved in the overall emission. Because limited interpatient variations were found-as indicated by the peak position and fluorescence intensity ratio data reported in Tables 2 and 4-the fitting analysis was performed on a single set of averaged spectra. For the tissue homogenate study, an averaged spectrum from three independent measurements was obtained for each homogenate. The spectra of the two homogenates of the same tissue type from the biopsy samples of each patient were averaged to obtain a single spectrum. The spectra corresponding to the same tissue type were then averaged for the three patients studied to obtain a single set of nonneoplastic or tumor homogenate spectra for curve-fitting analysis. For tissue sections the average spectrum from 10 independent measurements for each type of tissue in the section (neoplastic, boundary and nonneoplastic) was calculated for each biopsy sample, then averaged for the number of biopsies per patients and finally averaged for the number of patients considered (see Table 1).

Spectra were corrected for the system response and converted into wave number units for curve fitting, supposing an inhomogenous line broadening in frequency.

Spectra were analyzed by means of an iterative nonlinear curve-fitting procedure (PeakFit, SPSS Science. Chicago, IL) based on the Marquardt-Levenberg algorithm (25), by finding the true absolute minimum value of the sum of squared deviations (chi^sup 2^). Fitting analysis was done using a linear combination of half-gaussian modified gaussian (GMG) spectral functions, each of them representing the spectrum of each fluorophore mainly responsible for brain autofluorescence emission (NAD(P)H bound form, NAD(P)H free form, flavins). Collagen was not considered because, unlike multilayered epithelial tissues, brain tissue does not contain collagen fibers other than those in the meninges and in the walls of large blood vessels.

GMG spectral functions were defined by parameters derived from the real spectra of the pure fluorophores (26) through subsequent adjustments to match the best fit for the line shape of the spectra. The peak center wavelength (lambda) and the full width at half intensity maximum (FWHM) of each spectral function are given in Tables 3 and 6. The parameter values obtained were then used to define the relative contribution of each single spectral component to the whole fluorescence emission. After a first subjective matching of the contribution of individual GMG bands, the fit of the spectral components was adjusted and verified by the iterative process until a satisfying goodness of fit was obtained, according to both correlation r^sup 2^ and residual analysis. A very good fit of the fluorescence spectra-as shown by the r^sup 2^ values (given in the tables)-was obtained by considering a minor additional function (Area

In vivo study. In vivo autofluorescence analysis was performed on 12 patients affected by histopathologically diagnosed glioblastoma Grade IV, according to WHO (23), with frontal-basal or frontal-temporal localization. Two patients presented with both Grade-III and Grade-IV glioblastoma lesions (GBL). In vivo analysis was performed in accordance with the ethical standards of the committee on human experimentation of the institution where the experiments were carried out.

In vivo fluorescence measurements. Autofluorescence emission measurements were taken on patients during surgery by means of a fiber-optic probe (Fiberlan. Milan, Italy). The fiber-optic probe consists of 20 fibers: four fibers (200 (mu)m diameter each) deliver the excitation light to the tissue. 16 fibers (50 (mu)m diameter each) randomly distributed around the excitation fibers collect the fluorescence emission light (Fig. 3). Excitation light was provided by a 100 W high-pressure Hg lamp, combined with KG1 + BG38 antithermal filters and 390 nm dichroic mirror + UGI excitation filter (quartz, 2 mm), and coupled to the excitation fibers by means of microscope optics (objective 10x, NA 0.25). The spectral shape of the light output from the fiber-optic probe tip corresponds to the Hg lamp 366 nm line. This filter combination ensures the same excitation wavelength as that of the 366 nm interference filter with higher output power. The tissue autofluorescence collected by the fibers was delivered to the linear entrance slit of a Jarrell-- Ash Monospec 27 spectrograph, to be analyzed by means of the Optical Multichannel Analyzer.

The tissue depth probed at 366 nm was estimated by measuring the fluorescence intensity as a function of tissue thickness according to Gmitro et al. (27). Preliminary results on ex vivo tumor tissue samples, whose thickness was varied by stacking serial sections (5), gave (mu) = [(mu)^sub eff^(lambda^sub exc^) + (mu)^sub eff^(lambda^sub em^)] = 43-45 cm^sup -1^. where (mu)^sub eff^ is the effective optical attenuation coefficient at the excitation and emission wavelengths and represents a combination of absorption and scattering effects for light propagation in tissue. This indicates that a l/e optical penetration depth of 220-230 (mu)m is obtained with the experimental setup described above.

Before each patient study, a background spectrum was acquired, with the probe dipped in distilled water in the dark, and subtracted from all spectra acquired for that patient. Wavelength calibration was performed with reference to the Hg lamp emission lines. Before each measurement the probe tip was washed with saline and the investigated site was rinsed with saline to remove blood accumulation at the tissue surface. During measurements, the fiber probe tip was kept perpendicularly in contact with the tissue surface, without any pressure, by a neurosurgeon. The room lights were temporarily turned off during spectral acquisition.

For each case at least five measurements were taken on each region, corresponding to cortex, white matter and neoplastic lesion, exposed during the surgery (see Table 1). Each spectral acquisition lasted for 10 sequential scans of 200 ms each for a total of 2 s.

Unless stated otherwise, the ex vivo and in vivo spectra are not corrected for the spectral response of the optical systems used, i.e. microscope and fiber-optic probe.

RESULTS

Tissue homogenates

Fluorescence spectra of supernatants of both tumor and nonneoplastic tissue homogenates by excitation at 366 and 405 nm are shown in Fig. 4A,B. Differences in the spectral shapes are observed for both the excitation conditions used.

Under excitation at 366 nm the normal tissue has a main emission band peaking at about 440 nm, with a minor shoulder at wavelengths longer than 490 nm. In the tumor tissue the main emission band is redshifted and the long-wavelength shoulder becomes more marked. As for the signal amplitude, a higher value is found in normal than in tumor tissue. A comparison of the autofluorescence spectral shape was done in terms of ratios between the fluorescence signals in selected spectral regions that can be approximately ascribed to NAD(P)H, flavins and lipopigments, namely 470 +/- 10 nm, 520 +/- 10 nm and 580 +/- 10 nm, respectively. An increase in both FI^sub (520+/-10 nm)^/FI^sub (470=10 nm)^ and FI^sub (580+/-10nm)^/FI^sub (470+/-10 nm)^ ratio values was found in tumor with respect to nonneoplastic tissues (Table 2).

The spectra obtained under excitation at 366 run were analyzed by means of a fitting procedure (Table 6). Fitting analysis confirms the role of NAD(P)H as the fluorophore mainly responsible for emission under 366 nm excitation, contributing to the total emission for a percentage ranging from 74 (neoplastic tissue) to 82 (nonneoplastic tissue). Bound-free NAD(P)H ratio values of 3.9, 2.9 and 1.4 were obtained for nonneoplastic tissue, tumor boundary and tumor tissue, respectively.

The total amount of NAD(P)H revealed by microfluorometric analysis and calculated through fitting parameters (Table 6) as described for tissue homogenates turned out to be lower in boundary tissue (6041 a.u.) and in tumor tissue (5824 a.u.) than in nonneoplastic tissue (8323 a.u.), with a reduction of 27% and 30%, respectively.

A relative increase in the contribution of oxidized flavins and lipopigments was observed in neoplastic tissue with respect to both tumor boundary and nonneoplastic tissues.

Measurements on patients

In vivo spectrofluorometric analysis was performed by comparing the autofluorescence signals recorded directly on different brain regions of patients during surgery. As healthy tissues, both cortex and white matter were considered, and measurements were taken during surgical exposure in regions sufficiently far from the lesion to be considered free from infiltrating tumor cells.

Differences in both autofluorescence emission intensity and spectral shape are found among tumor tissue, normal cortex and white matter. Typical examples of autofluorescence emission spectra are shown in Fig. 8. The results of the spectrofluorometric analysis performed under excitation at 366 nm are summarized in Table 7. A similar peak position is observed for both white matter and cortex, whereas a redshift is found in the case of tumor. As for the spectral shape, a marked broadening of the emission band occurs for tumor in comparison with white matter, and even more so with cortex, as evaluated in terms of the FI^sub (520+/-10 nm)/ FI^sub (470+/-10 nm)^ ratio. As for the fluorescence intensity, the mean signal value measured inside the tumor mass is less than about half and one-fifth of those measured for cortex and white matter, respectively.

Measurements taken on the tumor bed surface during the last phases of tumor mass removal gave fluorescence signals approaching the values recorded in nonneoplastic tissues when at the surgeon's evaluation the bed surface was assumed to be free of tumor remnants. Histological analysis performed on biopsy specimens from investigation sites corroborated the fluorometric evaluation.

Two among the 12 patients considered in this study were affected by both Grade-III and Grade-IV glioblastomas. Although the data are very preliminary because of the limited number of cases, they are reported because to our knowledge no autofluorescence study on low-grade glioblastoma is at present reported in the literature.

In comparison with nonneoplastic tissue, the Grade-III lesions of both patients exhibited the same kind of alterations recorded for the Grade-IV lesions, although to a lesser extent. The autofluorescence properties of tumor tissues with different degrees of malignancy measured for one of the two patients are listed in Table 8.

DISCUSSION

Spectrofluorometric analysis performed on brain tissue homogenates showed that autofluorescence emissions of nonneoplastic and neoplastic tissues differ in both signal amplitude and spectral shape. The extent of the differences was dependent on the excitation wavelength in the range 366-480 nm, indicating that the two tissues differ in the relative concentration or in the nature (or both) of the endogenous fluorophores involved in the overall autofluorescence emission.

Microspectrofluorometric analysis performed on tissue sections under excitation at 366 and 405 nm confirmed the results obtained on homogenates under the same excitation conditions. The differences among nonneoplastic, boundary and neoplastic tissues in both spectral shape and signal amplitude were evidenced better by excitation at the shorter wavelength.

Fitting analysis performed on spectra recorded on both tissue homogenates and sections under excitation at 366 nm gave quite comparable results. The main emission band (440-480 nm) is attributable to the reduced form of NAD(P)H, in agreement with data reported for single living cells under normal and transformed conditions (1,26). The differences in autofluorescence properties between normal and tumor tissues suggest that changes in the metabolic and functional conditions occur in neoplastic tissue. The decrease of the autofluorescence signal amplitude indicates an alteration of redox equilibrium or a lowering of the total amount of the coenzyme (or both) as a consequence of an impairment of the aerobic or anaerobic metabolic pathway in tumor tissue (29).

In particular, fitting analysis indicates a reduction of the total NAD(P)H tissue contents in neoplastic with respect to nonneoplastic tissue. Differences of 21% and 30% were calculated from the spectra recorded on tissue homogenates and sections, respectively, which are comparable with that evaluated by biochemical analysis (25%). A decrease of NAD(P)H emission in in vitro human brain tumor tissues in comparison with healthy brain tissues was detected already by means of excitation-mission matrix analysis (20).

Further evidence of the altered metabolism in tumor tissue is provided by the enhancement of the relative contribution of oxidized flavins found in tumor with respect to the surrounding normal tissues.

Moreover, fitting analysis shows that the equilibrium between the bound and free forms of NAD(P)H is altered in tumor tissue. For both tissue homogenates and sections, the bound-free NAD(P)H ratio values were markedly lower in tumor than in nonneoplastic tissues, thus indicating that the tumor tissue environment favors the presence of the free NAD(P)H form. The bound-free NAD(P)H ratio has been reported to vary from cell line to cell line, but it shows a constancy within a cell line under standard metabolic conditions (30). Quite low ratio values were found in neoplastic cell lines (28), which were justified by the loss of binding sites for NAD(P)H described for malignant cells (31). A reduction of NAD(P)H binding sites could occur in malignant brain tumors, where a decreased metabolic function of mitochondrial enzymes in comparison with normal tissue has been found (32).

The comparison of the autofluorescence properties of the brain tissue types provides different evidence when the excitation is performed at a longer wavelength (405 nm).

In particular, less marked differences are found between nonneoplastic and neoplastic tissues in both fluorescence intensity and spectral shape. These findings can be interpreted by taking into account that the contribution of NAD(P)H-that is, the fluorophore mainly responsible for the differences-to the overall autofluorescence emission is reduced when excitation is done at wavelengths longer than 366 nm. Moreover, excitation at 405 nm favors the emission of other fluorophores, such as flavins and lipopigments that, although to a small extent, are relatively more abundant in tumor than in nonneoplastic tissue. The increase in the relative contribution of fluorophores emitting at long wavelength-flavins and lipopigments-under excitation at 405 nm accounts for the widening of the emission spectra occurring in both tissue types, as shown by the FI^sub (520+/-10 nm)/FI(470+/-10 nm) and FI(580+/-10 nm)/ FI^sub (470 + 10 nm)^ ratios.

Excitation at 405 nm results in a reduction of the differences in autofluorescence properties between neoplastic and boundary tissues. The presence of infiltrating cells both in surrounding and in tumor tissues can justify, at least in part, this similarity. A role may be played by lipopigments that exhibit variability in their fluorescence excitation and emission properties because of their chemical complexity, with differences in basic structure, degree of polymerization and oxidation, nature and amounts of nonlipid material such as carotenoid, and flavins. These compounds, apart from being associated with the physiological occurrence of agerelated lipofuscins in normal brain tissue, can be associated with alterations in tissue metabolic functions, with lipid peroxidation processes, and with the presence of various types of host-tumor response migrating cells such as macrophages, eosinophils and mast cells, which bear intracellular bright fluorescent granules (22,33-38).

Spectrofluorometric analysis performed in vivo by means of the fiber-optic probe evidenced that neoplastic lesions, like nonneoplastic tissues, could be distinguished from cortex and white matter by their autofluorescence properties. A reduction in fluorescence intensity and a broadening of the emission band along with a redshift of the peak position were observed in tumor in comparison with nonneoplastic tissue, although some variability occurred among the patients. The nature of the changes in autofluorescence in the two tissues suggests that the alterations in the biological conditions occurring in tumor with respect to nonneoplastic tissues-as evidenced by fluorometric analysis on tissue homogenates and sections-can affect the autofluorescence properties of the in vivo tissue. It is to be considered that in vivo spectra, unlike tissue section spectra, are affected by tissue turbidity, so that the differences observed in vivo cannot be related directly to those found on ex vivo samples. A redshift of the autofluorescence emission bands is observed in the in vivo, with respect to the ex vivo, measurements for both tumor and nonneoplastic tissues. As for the in vivo autofluorescence intensity, a reduction was observed in tumor with respect to nonneoplastic tissue, which was more pronounced than that expected on the basis of the values measured in tissue sections. This evidence confirms that other factors, in addition to tissue autofluorescence properties, have to be considered. Changes in the optical properties of the tissues in terms of absorption, reflectance and scatter (7,21), due to alterations of the histological organization and biochemical nature of the lesion, along with different tissue blood supplies (39,40), can influence the migration of light, the yields of excitation at the different tissue depths and the collection of fluorescence.

Our findings seem to be in disagreement with results recently reported by Lin et al. (21), who did not find any evidence of a redshift in the fluorescence spectra of GBM sites as compared with normal tissues. However, we believe that the different excitation wavelengths used (337 nm by Lin, 366 nm by us) can account for the different results. The influence of excitation wavelength in determining the autofluorescence emission properties, in terms of the different degrees of involvement of the fluorophores, was already pointed out by Shomacker et al. (5) and Richards-Kortum et al. (6), who compared the results by different authors on human normal and neoplastic colon tissue. In particular, a very important role could be played by the equilibrium between free and bound NADH forms, taking into account that the two forms differ for both excitation and emission spectra. Under excitation at 366 nm both free and bound forms contribute to the autofluorescence emission. Under this condition the autofluorescence spectral shape was found to be strongly affected by the bound-free NAD(P)H equilibrium (28,30). It was reported that a blueshift of the order of 25 nm occurs in the NAD(P)H absorption-excitation spectrum on binding to a variety of dehydrogenases (41,42). In this view, excitation at a shorter wavelength, namely 337 nm, is assumed to represent a condition quite favorable for the bound NAD(P)H form. As a consequence the emission spectrum shape becomes less sensitive to the bound-- free NAD(P)H equilibrium, and changes in fluorescence intensity are mainly to be expected.

In conclusion, further investigations are necessary for a more exhaustive understanding of the factors causing differences in autofluorescence emission properties between glioma and normal brain tissue in vivo. The data obtained, however, provide promising prospects for developing an autofluorescence-based diagnostic technique that can be applied easily to guide brain tumor resection during surgery.

Acknowledgement-This work was supported by the Italian National Research Council (CNR Special Project "Biotechnology").

(para) Posted on the website on 21 January 2003.

REFERENCES

1. Aubin, J.E. (1979) Autofluorescence of viable cultured mammalian cells. J. Biochem. Cytochem. 27. 36-43.

2. Rost, F. W. D. (1995) Fluorescence microscopy, Vol. 11. Cambridge University Press, Cambridge. UK.

3. Wagnieres, G. A., W. M. Star and B. C. Wilson (1998) In vivo fluorescence spectroscopy and imaging for ontological applications. Photochem. Photobiol. 68, 603-632.

4. Bottiroli, G., A. C. Croce, D. Locatelli, R. Marchesini, E. Pignoli, S. Tomatis, C. Cuzzoni, S. Di Palma, M. Dal Fame and P. Spinelli (1995) Natural fluorescence of normal and neoplastic colon: a comprehensive ex vivo study. Lasers Surg. Med. 16, 48-60.

5. Shomacker, K. T., J. K. Frisoli, C. C. Compton, T. J. Flotte, J. M. Richter, N. S. Nisioka and T. F. I. Deutsch (1992) Ultraviolet laser induced fluorescence of colonic tissue: basic biology and diagnostic potential. Lasers Surg. Med. 12, 63-78.

6. Richards-Kortum, R., R. P. Rava, R. E. Petras, M. Fitzmaurice, M. Sivak and M. S. Feld (1991) Spectroscopic diagnosis of colonic dysplasia. Photochem. Photobiol. 53, 777-786.

7. Marchesini, R., E. Pignoli, S. Tomatis, S. Fumagalli, A. E. Sichirollo, S. Di Palma, M. Dal Fante, P. Spinelli, A. C. Croce and G. Bottiroli (1994) Ex vivo optical properties of human colon tissue. Lasers Surg. Med. 15, 351-357.

8. Haringsma, J., G. N. Tytgat, H. Yano, H. Iishi, M. Tatsuta, T. Ogihara, H. Watanabe, N. Sato, N. Marcon, B. C. Wilson and R. W. Cline (2001) Autofluorescence endoscopy: feasibility of detection of GI neoplasms unapparent to white light endoscopy with an evolving technology. Gastrointest. Endosc. 53, 642-650.

9. Wang, T. D., J. M. Crawford, M. S. Feld, Y. Wang, L. Itzkan and J. Van Dam (1999) In vivo identification of colonic dysplasia using fluorescence endoscopic imaging. Gastrointest. Endosc. 49, 447-455.

10. Hung, J., S. Lam, J. C. LeRiche and B. Palcic (1991) Autofluorescence of normal and malignant bronchial tissue. Lasers Med. Surg. 11, 99-105. 11. Sato, M., A. Sakurada, M. Sagawa, M. Minowa, H. Takahashi, T.

Oyaizu, Y. Okada, Y. Matsumura, T. Tanita and T. Kondo (2001) Diagnostic results before and after introduction of autofluorescence bronchoscopy in patients suspected of having lung cancer detected by sputum cytology in lung cancer mass screening. Lung Cancer 32, 247-253.

12. Ramanujam, N., M. F. Mitchell, A. Mahadevan, S. Thomsen, A. Malpica, T. Wright, N. Atkinson and R. Richards-Kortum (1996) Spectroscopic diagnosis of cervical intraepithelial neoplasia (CIN) in vivo using laser-induced fluorescence spectra at multiple excitation wavelengths. Lasers Surg. Med. 19, 63-74.

13. Mayinger, B., P. Homer, M. Jordan, C. Gerlach, T. Horbach, W. Hohenberger and E. G. Hahn (2001) Light-induced autofluorescence spectroscopy for the endoscopic detection of esophageal cancer. Gastrointest. Endosc. 54, 195-201.

14. Bruner, J. M. (1994) Neuropathology of malignant gliomas. Semin. Oncol. 21, 126-138.

15. Poon, W. S., K. Schomacker, T. F. Deutsch and R. L. Martuza (1992) Laser-induced fluorescence: experimental intraoperative delineation of tumor resection margins. J. Neurosurg. 76, 679-686.

16. Tsai, J. C., M. C. Kao and Y. Y. Hsiao (1993) Fluorospectral study of the rat brain and glioma in vivo. Lasers Med. Surg. 13, 321-331.

17. Stammer, W., A. Novotiny, H. Stepp, C. Goetz, K. Bise and H. J. Reulen (2000) Fluorescence guided resection of glioblastoma multiforme by using 5-aminolevulinic acid-induced porphyrins: a prospective study in 52 consecutive patients. J. Neurosurg. 93, 1003-1013.

18. Zimmermann. A., M. Ritsch-Marte and H. Kostron (2001) mTHPCmediated photodynamic diagnosis of malignant brain tumors. Photochem. Photobiol. 74, 611-616.

19. Montan, S. and L. G. Stromblad (1987) Spectral characterisation of brain tumours utilising laser-induced fluorescence. Lasers Life Sci. 1, 275-285.

20. Chung, Y. G., J. A. Schwartz, C. M. Gardner, R. E. Sawaya and S. L. Jacques (1997) Diagnostic potential of laser-induced autofluorescence emission in brain tissue. J. Korean Med. Sci. 12, 135-142.

21. Lin, W. C., S. A. Toms, M. Johnson, E. D. Jansen and A. MahadevanJansen (2001) In vivo brain tumor demarcation using optical spectroscopy. Photochem. Photobiol. 73, 396-402.

22. Bottiroli, G., A. C. Croce, D. Locatelli, R. Nano, E. Giombelli, A. Messina and E. Benericetti (1998) Brain-tissue autofluorescence: an aid for intraoperative delineation of tumor resection margins. Cancer Detect. Prev. 22, 330-339.

23. Kleihues, P., P. C. Burger and B. W. Scheithauer (1993) Histological Typing of Tumours of the Central Nervous System: World Health

Organization International Histological Classification of Tumours. Springer, Berlin.

24. Glock, G. E. and P. McLean (1955) The determination of oxidized and reduced diphosphopyridine nucleotide and triphosphopyridine nucleotide in animal tissues. Biochem. J. 61, 381-390.

25. Marquardt, D. W. (1963) An algorithm for least-squares estimation of non-linear parameters.J. Soc. Ind. Appl. Math. 11, 431-441.

26. Croce, A. C., A. Spano, D. Locatelli, S. Bami, L. Sciola and G. Bottiroli (1999) Dependence of fibroblast autofluorescence properties on normal and transformed conditions. Role of the metabolic activity. Photochem. Photobiol. 69, 364-374.

27. Gmitro, A. F., F. W. Cutruzzola, M. L. Stetz and L. I. Deckelbaum (1988) Measurements depth of laser-induced tissue fluorescence with application to laser angioplasty. Appl. Optics 27, 1844-1849.

28. Salmon, J. M., E. Kohen, P. Viallet, J. G. Hirscheberg, A. W. Wouters, C. Kohen and B. Thorell (1982) Microspectrofluorometric approach to the study of free/bound NAD(P)H ratio as metabolic indicator in various cell types. Photochem. Photobiol. 36, 585-593.

29. Warburg, 0. (1956) On the origin of the cancer cells. Science 123, 309-314.

30. Viallet, P., E. Kohen, D. 0. Schachtschabel, A. Marty, J. M. Salmon, C. Kohen, H. B. Leising and B. Thorell (1978) The effect of atebrine and acridine analog (BCMA) on the coenzyme fluorescence spectra of cultured melanoma and Ehrlich ascites (EL2) cell. Histochemistry 57, 189-201.

31. Galeotti, T., G. D. V. van Rossum, D. H. Mayer and B. Chance (1970) On the fluorescence of NAD(P)H in whole cell preparations of tumours and normal tissue. Eur. J. Biochem. 17, 485-496.

32. Meixensberger, J., B. Herting, W. Roggendorf and H. Reichmann (1995) Metabolic patterns in malignant gliomas. J. Neuro-Oncol. 24, 153-161.

33. Wolman, M. (1980) Lipid pigments (chromolipids): their origin, nature and significance. Pathobiol. Ann. 10, 253-267.

34. Mayeno, A. N., K. J. Harriman and G. J. Gleich (1992) Granuleassociated flavin adenine dinucleotide (FAD) is responsible for eosinophil autofluorescence. J. Leukoc. Biol. 51, 172-175.

35. Kaiser, H. E. (1989) Stroma, generally a non-neoplastic structure of the tumour. In Influence of Tumour Development on the Host (Edited by L. A. Lotta), pp. 1-8. Kluwer Academic Publishers, Dordrecht, The Netherlands.

36. Weller, P. F., S. J. Ackerman, A. Nicholson-Weller and A. M. Dvorak (1989) Cytoplasmic lipid bodies of human neutrophilic leukocytes. Am. J. Pathol. 135, 947-959.

37. Barnes, D., S. Aggarwal, S. Thomsen, M. Fitzmaurice and R. A. Richards-Kortum (1993) Characterization of the fluorescent properties of circulating human eosinophils. Photochem. Photobiol. 58, 297-303.

38. Tappel, A. L. (1973) Lipid peroxidation damage to cell components. Fed. Proc. 32, 1870-1874.

39. Tang, G. C., A. Pradhan and R. R. Alfano (1989) Spectroscopic differences between human cancer and normal lung and breast tissues. Lasers Med. Surg. 9, 290-295.

40. Lin, W. C., S. A. Toms, E. D. Jansen and A. Mahadevan-Jansen (2001) Intraoperative application of optical spectroscopy in the presence of blood. IEEE J. Sel. Top. Quantum Electron. 7, 996-1003.

41. Theorel, H. and R. Bonnichsen (1951) Studies on liver alcohol dehydrogenase. I. Equilibria and initial reaction velocity. Acta Chem. Scand. 5, 1105-1126.

42. Wolfbeiss, 0. S. (1985) The fluorescence of organic natural products. In Molecular Luminescence Spectroscopy: Methods and Applications (Edited by S. G. Schulman), pp. 167-370. John Wiley & Sons, New York.

Anna C. Croce1, Sabrina Fiorani1, Donata Locatelli,1 Rosanna Nano1, Mauro Ceroni2, Flavio Tancioni3, Ermanno Giombelli4, Eugenio Benericetti4 and Giovanni Bottiroli*1

1Istituto di Genetica Molecolare, Sezione di Istochimica e Citometria, CNR, Dipartimento di Biologia Animate, Universita, Pavia, Italy;

2Dipartimento di Scienze Neurologiche, Fondazione C. Mondino, Universita, Pavia, Italy;

3Istituto Humanitas, Divisione di Neurochirurgia, Rozzano, Milan, Italy and

4Azienda Ospedaliera di Parma, Divisione di Neurochirurgia, Parma, Italy

Received 2 July 2002; accepted 19 December 2002

*To whom correspondence should be addressed at: Istituto di Genetica Molecolare, Sezione di Istochimica e Citometria, CNR, Piazza Botta, 10, 27100 Pavia, Italy. Fax: 39-382-506-430; e-mail: both@dragon.ian.pv.cnr.it

Copyright American Society of Photobiology Mar 2003
Provided by ProQuest Information and Learning Company. All rights Reserved