Anal. Chem. 2005, 77, 2895-2900
Classification of Nerve Cells from Substantia Nigra of Patients with Parkinson’s Disease and Amyotrophic Lateral Sclerosis with the Use of X-ray Fluorescence Microscopy and Multivariate Methods Joanna Chwiej,*,† Katarzyna Fik-Mazgaj,† Magdalena Szczerbowska-Boruchowska,† Marek Lankosz,† Jerzy Ostachowicz,† Dariusz Adamek,‡ Alexandre Simionovici,§ and Sylvain Bohic|
Department of Radiometry, Faculty of Physics and Applied Computer Science, AGHsUniversity of Science and Technology, Al. Mickiewicza 30, 30-059 Krakow, Poland, Institute of Neurology, Collegium Medicum, Jagiellonian University, ul. Botaniczna 3, 31-503 Krakow, Poland, Laboratoire de Sciences de la Terre, ENS-Lyon, 69007 Lyon, 46 Alee d′Italie, France, and European Synchrotron Radiation Facility, BP 220, 38043 Grenoble Cedex 9, France
The causes of Parkinson’s disease and amyotrophic lateral sclerosis are still not known, but there is evidence that metal ions can be involved in processes leading to degeneration and atrophy of neurons in the case of these two neurodegenerative disorders. A synchrotron microbeam X-ray fluorescence technique was applied for topographic and quantitative analyses of selected elements on central nervous system tissue. The thin slices of brain were measured on the undulator beamline ID 22 at the European Synchrotron Radiation Facility in Grenoble, France. The polychromatic beam with the dimension of 5 µm × 2 µm (horizontal × vertical) was used in measurements. Tissues of substantia nigra representing Parkinson’s disease, amyotrophic lateral sclerosis, and the control case were scanned. The results obtained indicated that accumulation of some elements depends on the case that the substantia nigra represents. Some variability in the elemental distribution for a given case was noticed as well. To investigate if present differences in the elemental accumulation between analyzed cases are statistically significant, multivariate methods were used. Cluster and discriminant analyses confirmed the significance of the differences in elemental accumulation in biological structures representing the examined cases. The methods used let us classify these structures in separate groups and determine elements, which play the greatest role in the differentiation of the biological structures for each case. Parkinson’s disease (PD) and amyotrophic lateral sclerosis (ALS) are neurodegenerative disorders that are characterized by * Corresponding author. Phone: +48 12 617 44 24. Fax: +48 12 634 00 10. E-mail:
[email protected]. † AGHsUniversity of Science and Technology. ‡ Jagiellonian University. § ENS-Lyon. | European Synchrotron Radiation Facility. 10.1021/ac048173k CCC: $30.25 Published on Web 03/26/2005
© 2005 American Chemical Society
the loss of particular subsets of neurons. ALS is a motor neuron degenerative disease for which the progressive loss of nerve cells is observed in the spinal cord, brainstem, and motor cortex. PD is a slowly progressive neurodegenerative disorder for which degeneration of neuromelanin-rich neurons in the substantia nigra (SN) and the frequent deposition of Lewy bodies are noticed.1 The pathogenesis of these disorders is still not known, but oxidation stress, excitotoxicity, protein aggregation, and mitochondrion dysfunction are suspected to lead to degeneration and atrophy of nerve cells.1-5 There is strong evidence that these upsets can be promoted by the abnormal biochemical reactions catalyzed by selected metal ions.1,4 The quantitative and topographic analysis of selected elements in the central nervous system tissue may be performed with the use of the synchrotron microbeam X-ray fluorescence technique (µ-SXRF). This technique was previously used by Ektessabi et al. for analysis of iron distribution in single neurons from PD and control sample. However, it is necessary to mention that the analyzed SN samples were fixed in formalin and embedded in paraffin.6 The purpose of our research was to investigate the role of metals in processes leading to degeneration and atrophy of nerve cells in case of the two aforementioned disorders (PD and ALS). The elements P, S, Cl, K, Ca, Fe, Cu, Zn, Se, Br, Rb, and Sr were found in the central nervous system tissues. The µ-SXRF technique enabled us to obtain two-dimensional maps of elemental distributions. The comparison of these maps with the histopathological view of the tissue showed that the higher intensities of selected elements X-ray lines are correlated with the positions of nerve (1) Bains, J. S.; Shaw, C. A. Brain Res. Rev. 1997, 25, 335-358. (2) Robberecht W. J Neurol. 2000, 247, I/1-I/6. (3) Mattson, M. P.; LaFerla, F. M.; Chan, S. L.; Leissring, M. A.; Shelpel, P. N.; Geiger, J. D. Trends Neurosci. 2000, 23, 222-229. (4) Bush, A. I. Curr. Opin. Chem. Biol. 2000, 4, 184-191. (5) Cassarino, D. S.; Bennet, J. P. Brain Res. Rev. 1999, 29, 1-25. (6) Ektessabi, A.; Yoshida S.; Takada, K. X-ray Spectrometry 1999, 28, 456460.
Analytical Chemistry, Vol. 77, No. 9, May 1, 2005 2895
cells. The differences in the accumulation of elements in the examined areas of substantia nigra depending on the analyzed case (PD, ALS and control case) were observed.7-9 However, some variability in these areas for a given case was noticed as well. To confirm the significance of the differences between areas representing the three examined cases, multivariate methods were applied. Multivariate methods are a perfect tool for the multidimensional data analyses. These statistical techniques make it possible to investigate the dependence and connections among variables, helping to reduce and to simplify data and also, more importantly, to classify objects into groups. The literature indicates that these methods are playing an important role in technical, social, and medical sciences. In medical research, multivariate techniques were successfully applied for diagnosis of lung cancer based on the metal contents in serum and hair,10 distinction of cervical precancers from normal tissue based on their near-infrared Raman spectra, 11 and distinction of drug-free subjects from drug abusers based on concentration of trace metals in their hair.12 Multivariate methods, such as cluster analysis and principal component analysis, are often applied in order to compare quantitative results obtained by means of different independent analytical methods.13-15 There are also some examples of the use of multivariate methods for the statistical analysis of the data obtained using X-ray fluorescence techniques.16-18 The most important statistical analyses that represent multivariate methods are cluster analysis, discriminant analysis, canonical analysis, and factor analysis. Two of the aforementioned statistical techniquesscluster and discriminant analysisswere applied for classification of nerve cells, white matters, and areas surrounding neurons for PD, ALS, and the control cases. Cluster analysis was used to identify similarities among observations and to classify them into groups that contain objects similar to each other and simultaneously different from those belong to the other groups. The hierarchical Ward’s method of agglomeration and squared Euclidean as a measure of distance between observations (7) Szczerbowska-Boruchowska, M.; Lankosz, M.; Ostachowicz, J.; Adamek, D.; Krygowska-Wajs, A.; Tomik, B.; Szczudlik, A.; Simionovici, A.; Bohic, S. ESRF Highlights 2002, p 87-88. (8) Szczerbowska-Boruchowska, M.; Lankosz, M.; Ostachowicz, J.; Adamek, D.; Krygowska-Wajs, A.; Tomik, B. Szczudlik, A.; Simionovici, A.; Bohic, S. J Phys. IV 2003, 104, 325-328. (9) Szczerbowska-Boruchowska, M.; Lankosz, M.; Ostachowicz, J.; Adamek, D.; Krygowska-Wajs, A.; Tomik, B. Szczudlik, A.; Simionovici, A.; Bohic, S. X-ray Spectrom. 2004, 33 (1), 3-11. (10) Ren, Y. L.; Zhang, Z. Y.; Ren, Y. Q.; Li, W.; Wang, M. C.; Xu, G. Talanta 1997, 44, 1823-1831. (11) Mahadevan-Jansen, A.; Mitchell, M. F.; Ramanujam, N.; Malpica, A.; Thomsen, S.; Utzinger, U.; Richards-Kortum, R. Photochem. Photobiol. 1998, 68, 123-132. (12) Bermejo-Barrera, P.; Moreda-Pineiro, A.; Bermejo-Barrera, A.; BermejoBarrera, A. M. Anal. Chim. Acta 2002, 455, 253-265. (13) Cariati, F.; Fermo, P.; Gilardoni, S.; Galli, A.; Milazzo, M. Spectrochim. Acta B 2003, 58, 177-184. (14) Ernst, T.; Popp, R.; Van Eldik, R. Talanta 2000, 53, 347-357. (15) Danzer, K.; Florian, K.; Hassler, J.; Matherny, M.; Schron, W.; Zaray, G. J Anal. Atom. Spectrom. 1998, 13, 371-375. (16) Hida, M.; Sato, H.; Sugawara, H.; Mitsui, T. Forensic Sci. Int. 2001, 115, 129-134. (17) Kessler, T.; Hoffmann, P.; Greve, T.; Ortner, H. M. X-ray Spectrom. 2002, 31, 383-390. (18) Kierzek, J.; Kunicki-Goldfinger, J.; Kasprzak, A. J.; Malozewska-Bucko, B. Book of Abstracts, European Conference on Energy Dispersive X-ray Spectrometry, June 16-21, 2002, Berlin.
2896
Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
were applied. The purpose of the discriminant analysis was not only the classification of objects but also the separation of all the variables, which strongly discriminate between the biological structures representing the examined cases. EXPERIMENTAL SECTION 1. Samples. The samples of the substantia nigra were taken during autopsy from patients deceased with Parkinson’s disease and amyotrophic lateral sclerosis and from patient who died due to nonneurological conditions. The specimens were frozen and cut into 20-µm-thick slices with the use of a cryomicrotome. From each section one slice was taken for routine histopathological investigation and the other one for µ-SXRF analysis. The slice designed for elemental analysis was immediately placed onto AP1 foil and freeze-dried. The commercially available AP1 foil is ultrathin (0.15 µm), ultrapure (contains no trace metals), and transparent for X-rays. From each sample of the substantia nigra two areas were selected for scanning. The first one contained pericarial parts of neurons and the other one was without nerve cell bodies. These two areas represent gray and white matter, respectively. 2. Measurements. For elemental analysis of SN samples, the µ-SXRF technique was applied. The measurements were carried out at the European Synchrotron Radiation Facility on the undulator beamline ID 22.19 The polychromatic beam mode excitation was applied for all the measurements. The microbeam with the dimension of 5 µm × 2 µm (horizontal × vertical) was used in measurements. Samples were positioned at an angle of 45° with respect to the incident beam, while the exit angle was 45° as well. The characteristic X-ray lines were measured by a Si (Li) detector. The energy resolution of the detector was 150 eV at 5.9 keV. Typical areas selected for scanning were 500 µm × 500 µm. Step size applied for mapping was 10 µm × 5 µm (horizontally × vertically). The acquisition time was equal to 3 s per pixel. The collected spectra were deconvoluted with the use of the AXIL program. All the computed X-ray line intensities were normalized to the value of the incident photon flux. RESULTS The higher intensities of the characteristic X-ray lines of selected elements in the µ-SXRF images reflected positions of the neurons in the tissue slices. Two-dimensional maps of elemental distribution in comparison with the microscopic view of the SN tissue for the control and PD case respectively are presented in Figures 1 and 2.9 The ranges of masses per unit area of elements for nerve cell bodies of the three examined cases are shown in Table 1. The masses per unit area of elements were calculated according to expression 1
MT )
YT tTITpinS
(1)
where YT is the net peak area of the measured element for the tissue sample, tT is the measurement time for the tissue sample, and ITpin is the incident photon flux for the tissue sample. (19) http://www.esrf.fr/exp_facilities/ID22/.
Figure 1. Substantia nigra of the control case: (A) microscopic view of scanned area of the tissue and (B) distribution of selected elements. The neurons are seen on the optical image as dark points.
Figure 2. Substantia nigra of the PD case: (A) microscopic view of scanned area of the tissue and (B) distribution of selected elements. The arrow shows the localization of the neuron in the microscopic view.
S is the sensitivity for the measured element and is given by the following formula
S)
Ys tsISpinMs
(2)
where Ys is the net peak area of measured element for the standard sample, ts is the measurement time for the standard sample, ISpin is the incident photon flux for the standard sample, and Ms is the mass per unit area of the measured element in the standard sample. Estimated uncertainties of masses per unit area of elements were not higher than 5%. Additionally, the detection limits of elements were calculated for the SN tissue, at the typical measurement conditions. It was done according to the formula given by Currie.20 The detection limits of elements and their uncertainties at the 95% confidence level are presented in the Table 1 as well. Higher levels of S, Ca, Fe, Cu, Zn, Se, and Br were observed for SN neurons in the PD sample in comparison with the control case. For the same part of the brain, an increased content of Cl, Ca, Zn, and Br was noticed for the ALS case.9 To examine if (20) Currie, L. A. Anal. Chem. 1968, 40, 586-593.
present differences are statistically significant, multivariate methods such as cluster and discriminant analyses were applied. These methods are exploratory tools for solving classification problems, such as, for instance, the different populations collected in an experiment. Three biological structures of substantia nigrasnerve cell bodies, areas surrounding neurons, and white mattersswere examined with the use of these statistical techniques. Cluster analysis attempts to group populations based on their properties into several distinct ensembles. The resulting groups, called clusters, can be classified on the basis of the distances separating them or their intracluster variabilities. The distances are actually generalized measures of the separations of the points in a N-dimensional parameter space. In this study we used two such measures: the squared Euclidean and Ward’s method. The former is based on the actual geometric distances between groups and places increasingly greater weights on objects that are further apart. The latter uses a variance approach to evaluate the distances between clusters by minimizing the sum of squares of any two candidate clusters formed at each step. The final result of a cluster analysis is a taxonomy tree, also called a dendrogram, which represents graphically all groups identified. Discriminant analysis is used to select the variables that best discriminate between two or more naturally occurring groups. Specifically, one determines whether groups differ with regard Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
2897
Table 1. Ranges of Masses Per Unit Area of Elements for Nerve Cell Bodies of the Three Examined Cases (PD, ALS, and Control) and the Detection Limits of Elements Calculated for the Tissue Sample (for typical measurement conditions) range of Ma (µg/cm2)
a
element
control
PD
ALS
DLb
P S Cl K Ca Fe Cu Zn Se Br Rb Sr
13-17 5.0-5.6 1.0-3.4 8.5-8.9 0.15-0.19 0.38-0.61 0.05-0.06 0.06-0.08 0.001-0.004
