Deeper Understanding of Biological Tissue ... - ACS Publications

Subscriber access provided by UNIV FED DE SAO PAULO UNIFESP BIREME

Article

Deeper understanding of biological tissue: Quantitative correlation of MALDI-TOF and Raman imaging Thomas W Bocklitz, Anna Christina Crecelius, Christian Matthäus, Nicolae Tarcea, Ferdinand Eggeling, Michael Schmitt, Ulrich S. Schubert, and Juergen Popp Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/ac402175c • Publication Date (Web): 15 Oct 2013 Downloaded from http://pubs.acs.org on October 20, 2013

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Deeper understanding of biological tissue: Quantitative correlation of MALDI-TOF and Raman imaging T.W. Bocklitz,†,‡,⊥ A.C. Crecelius,¶,§,⊥ C. Matthäus,†,‡ N. Tarcea,†,‡ F. von Eggeling,k,§ M. Schmitt,†,‡ U.S. Schubert,∗,¶,§,# and J. Popp∗,†,‡,# Institute of Physical Chemistry and Abbe Center of Photonics, Friedrich-Schiller-University Jena, Helmholtzweg 4, Jena, Germany, Institute of Photonic Technology, Albert-Einstein-Str. 9, 07745 Jena, Germany, Laboratory of Organic and Macromolecular Chemistry (IOMC), Friedrich-Schiller-University Jena, Humboldtstrasse 10, 07743 Jena, Germany, Jena Center for Soft Matter (JCSM), Friedrich-Schiller-University Jena, Philosophenweg 7, 07743 Jena, Germany, and Core Unit Chip Application (CUCA), Institut für Humangenetik, Universitätsklinikum Jena, Leutragraben 3, 07743 Jena, Germany E-mail: [email protected]; [email protected]

∗ To

whom correspondence should be addressed

† IPC ‡ IPHT ¶ IOMC § JCSM k CUCA ⊥ T.W.B.

and A.C.C. contributed equally to the presented work. and requests for materials concerning Raman spectroscopy and chemometrics should be addressed to J.P ([email protected]), while the corresponding author for MALDI-TOF imaging is U.S.S ([email protected]). # Correspondence

1 Environment ACS Paragon Plus


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract In order to achieve a comprehensive description of biological tissue, spectral information about proteins, lipids, nucleic acids and other biochemical components needs to be obtained concurrently. Different analytical techniques may be combined to record complementary information of the same sample. Established techniques, which can be utilized to elucidate the biochemistry of tissue samples are, for instance MALDI-TOF-MS and Raman microscopic imaging. With this contribution we combine these two techniques for the first time. The combination of both techniques allows the utilization and interpretation of complementary information, i.e. the information about the protein composition derived from the Raman spectra with data of the lipids analyzed by the MALDI-TOF measurements. Furthermore, we demonstrate how spectral information from MALDI-TOF experiments can be utilized to interpret Raman spectra.

Introduction In the last decade molecular imaging techniques have become applicable for medical diagnosis, histology and pathology, because of their unique ability to extract bio-chemical information. Raman- and MALDI-TOF imaging 1,2 represent such emerging techniques. Both methods could potentially be used as diagnostic tools for the detection and characterization of different diseases, as for instance cancer 3–5 or inflammatory bowel diseases. 6 As a consequence, MALDITOF and Raman imaging can complement histology and pathology, because the diagnosis can be supported by objective criteria, associated with the spectral information. 1,2 However, these techniques display also several disadvantages, which restrict their application. Raman imaging is often not very specific, since a superposition of spectral information from proteins, lipids and nucleic acids is represented in the Raman spectra of biological tissues. On the other hand high molecular specificity can be obtained by MALDI-TOF spectrometry, however, the sample preparation is still rather demanding. 7 The simultaneous analysis of proteins and lipids in a single 2 Environment ACS Paragon Plus

Page 2 of 19

Page 3 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


MALDI experiment setup is so far, only possible using a very specific inkjet printing approach of multiple matrices or by washing of the matrix from the tissue section as shown recently. 8 In addition, MALDI-TOF imaging is invasive and can not be employed in-vivo. Raman imaging, in contrast, requires only a minimal sample preparation and is a non-invasive method, which can be potentially employed in vivo. 9 However, no direct information on molecular masses are obtained. A combination of both techniques could be highly beneficial for the understanding of biological tissue composition as well as for diagnostic purposes; both methods can provide complementary information. If lipids are analyzed in the MALDI-TOF experiment, additional information associated with proteins can be extracted from the Raman data and vice versa. A direct analytical correlation between the spectral information of a MALDI-TOF and a Raman scan offers the possibility perform a Raman experiment under in vivo conditions and concurrently search for the information correlated with the MALDI spectra, preliminarily obtained under ex vivo conditions. Here, we demonstrate the combination of Raman and MALDI-TOF imaging and show, how a combined application can lead to a more comprehensive understanding of the composition of biological tissue. Recently, a review was illustrating the combination of mass spectrometric techniques with vibrational spectroscopic methods 10 and thereby demonstrating the gain of information, which is achieved if these methods are combined. The combination of different mass spectrometric methods with vibrational spectroscopic techniques was carried out by Petit et al. 11 for time-of-flight-secondary ion mass spectrometry, synchrotron-FT-IR, and synchrotron-UV and by Li et al. 12 for Raman micro-spectroscopy and secondary ion mass spectrometry, respectively. Nevertheless, only a qualitative comparison of both measurements was established. 11,12 Here, we present a methodology for a quantitative comparison, which we call ’quantitative correlation’.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Material and Methods Sample Preparation

A cryosection of a 10 µm mouse brain (Mus musculus) sample was

cut on a cryostat and transferred onto a pre-cooled conductive ITO-coated glass slide. Because common fixation protocols alter the chemical composition of tissue, the sample was simply dried on the ITO slide. For comparison with the histology of the sample a Nissl stain was performed on a parallel section.

Raman imaging

Raman spectra were acquired using a Confocal Raman Microscope Model

CRM alpha300R (WITec, Ulm, Germany). Excitation at 633 nm (about 10 mW at the sample) is provided by a HeNe laser (Melles Griot). The exciting laser radiation is coupled into a Zeiss microscope through a wavelength-specific single mode optical fiber. The incident laser beam is collimated via an achromatic lens and passes a holographic band pass filter before it is focused onto the sample through the objective of the microscope. A Zeiss EC Epiplan air objective (50×/0.95 NA) was used in the studies reported here. The sample is scanned through the laser focus in a raster pattern point by point stepping modus. Spectra were collected at a 25 µm grid with a dwell time of 2 s and a pre-bleaching time of 1 s. The spectrometer is equipped with a 300 /mm (BLZ=750 nm) grating and a back illuminated, deep depletion CCD. The recording of the Raman scan was performed before the MALDI measurement, because of the applied Matrix in the case of MALDI imaging.

MALDI-TOF imaging

For the analysis of lipids the common matrix alpha-cyano 4-hydroxy

cinnamic acid (5 mg/mL) in 50% acetonitrile and 0.2% trifluoracetic acid was prepared and was applied onto the tissue section using the ImagePrep station (Bruker Daltonics) following R a standard protocol. The MALDI-TOF imaging analysis was performed on an Ultraflex III

MALDI-TOF/TOF mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with a R ’smartbeam’ laser (λ = 355 nm, repetition rate 200 Hz). The Fleximaging software version


Page 4 of 19

Page 5 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


3.0 (Bruker Daltonics) was used for spectral acquisition and evaluation throughout all experiments. Before the measurement, the instrument was calibrated with an external standard, a peptide calibration mixture (Bruker Daltonics). All spectra were measured in the positive reflectron mode, and the m/z range 500 to 1600 was scanned. Typically, 500 shots were accumulated for a spectrum; 8277 positions were recorded for the entire mouse brain section with a spatial resolution of 75 µm, which was set up at the instrument. The acquisition time was in the range of 14 h.

Computational All of the computations were performed using the statistical programming language R 13 running on a commercial computer system (Intel(R) Core(TM) 2Duo CPU E6750 2.66GHz, 2×1.97 GB RAM). The packages ’readBrukerFlexData’, 14 ’Peaks’, 15 ’akima’ 16 and ’pls’ 17 were used for data analysis. For details about spectral pre-processing, image generation, interpolation and modeling see the ’Supporting Information’ section.

Results The presented methodology for quantitatively correlating MALDI-TOF and Raman imaging consists of two steps and is sketched in Figure 1. First a transformation, which connects the MALDI coordinate system (x0 , y0 ) with the Raman coordinate system (x, y), has to be determined. After the transformation matrix has been evaluated, the Raman spectra can be interpolated to the MALDI coordinate system and for every point of the MALDI coordinate system a MALDI spectrum can be compared with the associated Raman spectrum. This will be referred to as registration workflow and is sketched in the upper part of Figure 1. Furthermore, it is possible to calculate mean Raman spectra for regions, which exhibit certain MALDI peak intensity. Subsequently, a statistical model is build, which can be used in a predictive manner. This step, the correlation analysis itself (lower part of Figure 1), starts with choosing a MALDI peak or



a MALDI fingerprint of interest. Using a multivariate calibration method the MALDI peak intensity or the outcome of the MALDI fingerprint can be modeled from the Raman spectra. This model can afterwards be applied to a Raman data set and, thus, allows the use of information obtained from MALDI-TOF experiments in a non-invasive manner or under in-vivo conditions. MALDI-TOF spectra

Raman spectra

Pre-processing & Image generation

Pre-processing & Image generation

MALDI-TOF image

Raman image Coordinate transformation & Interpolation

Correlation Workflow Construct PLS model

transformed Raman scan

Registration Workflow

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 19

Predict MALDI image using PLS model predicted MALDI-TOF image

Figure 1: The outline of the presented framework is sketched, which consists of a registration workflow and a quantitative correlation workflow. See text for details.

Registration Workflow The registration workflow is a simple version of the methodology suggested by Schaaff et al. 18 and starts with choosing N marker points, represented in both scans. For this purpose, it is posR software (Bruker Daltonics). Utilizing sible, to use the registration points of the Fleximaging

these registration points has a few disadvantages. One of these disadvantages is that the fieldof-view (FOV) of the Raman device is smaller and exhibits a higher precision or reproducibility of position than the FOV from the MALDI imaging device. Apart from these technical aspects, Raman images are usually smaller in dimension, but have a higher spatial resolution. In order to optimize the transformation from the MALDI coordinate system to the Raman coordinate system, three parameters (α, ∆x and ∆y) have to be estimated. These parameters correspond to the rotation angle and the shift in x and y direction, respectively. To generate a robust estimate, it is advisable to choose more points than parameters needed. Therefore, we used six clear tissue 6 Environment ACS Paragon Plus

Page 7 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


features as teaching points within the MALDI image and the Raman image. The registration workflow is sketched in Figure 2. First the MALDI peaks of interest are chosen. By summation of the peak area and application of a logarithm an image is generated. In the example of Figure 2 A&B the image is constructed from the m/z 703 and m/z 799 peaks, which can be attributed to phospholipids. 19 The intensities are visualized by the green and blue color, respectively, and indicate the different composition of phospholipids in both regions. In the same manner an image is generated from the background corrected Raman scan (Figure 2 C&D). In Figure 2 C the impact of the background correction is shown. Image Figure 2 D was generated using the integrated intensities of the CH-stretching region (2790–3106 cm−1 ). Afterwards the Raman image and the Raman spectra are shifted and rotated based on the transformation explained above. The result is plotted in Figure 2 E, which shows the overlay of the generated Raman and MALDI image. From distinct features of both images it is obvious that the transformation is almost optimal. The long line-type feature in the upper right corner is visible in both the MALDI and Raman image and both overlap perfectly. At the end of the registration workflow mean Raman spectra can be calculated for regions, which exhibit distinct MALDI intensities or characteristic MALDI peaks. For the mouse brain sample, the m/z 703 and m/z 799 intensities resulted in the best image contrast. The generated mean Raman spectra clearly indicate that both regions (m/z 703 and m/z 799) are characterized by different Raman signals (data not shown). The blue regions (m/z 799) shows higher Raman band intensities within the CH-stretching, if compared to the green regions (m/z 703) (see Figure 2 E). Although a distinct assignment of the MALDI peaks is challenging, because of the complex composition of the tissue sample, tentative assignments can be suggested. The peak at m/z may origin from H+ adducts of phosphocholine (PC) 34:0 and/or sphingomyelin (SM) 16:0, whereas the m/z 799 peak can be assigned to K+ adduct of PC 34:1. Further imaging experiments need to be performed for reproducible assignments.


Registration


600

m/z

700

800

900

1 mm

m/z 799

Raman Intensity [a.u.] 500

E

1000

2000

1 mm

3000

C

D

Transformation

Registration

Preprocessing

-1

2500

Wavenumber / cm

1500

Raman imaging

Figure 2: Registration Workflow: A. Mean MALDI spectrum of scanned area; B. From two MALDI peaks (m/z 703 and m/z 799) a false color image is generated. C. Mean Raman spectrum of scanned area with and without background correction; D. The CHstretching region is integrated and a false-color image is generated. E. The Raman spectra are rotated and shifted, and every MALDI spectrum can be connected with a Raman spectrum. Here, a shift and a rotation by ≈ 183◦ has to be applied. By the spatial connection from MALDI spectrum and Raman spectrum both information can be used together. See text for further details.

1 mm

B

Signal Intensity [a.u.]

Preprocessing

500

m/z 703

MALDI-TOF imaging 3000 2000

600 1000

200

5000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

A

Analytical Chemistry Page 8 of 19

1170 cm-1 1255 cm-1

0.00 0.05 0.10 0.15

C vertical shifted by 0.1 and scaled by 5

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Normalized Raman Intensity

Page 9 of 19

B 500

A 1000 µm

1000

1500 -1 Wavenumber / cm

2800

3100 D

1000 µm

Figure 3: The Raman scanned region of the MALDI scan in Figure 2 B is visualized in A. The background corrected and vector-normalized mean Raman spectra of the blue and green region of panel A are given in B. The spectral difference of both is visualized in panel C. The difference Raman spectrum is negative in the CH-region, while it is mostly positive in the fingerprint region (between 500 − 1800 cm−1 ). This result corresponds to a higher lipid content in the blue region (m/z 799) as compared to the green region (m/z 703) as the cross section of lipids is higher than the cross section of proteins. The spatial lipid distribution found within the MALDI image can be complemented by the protein information of the Raman spectra. Panel D shows the band integration of the m/z 703 peak, but without thresholding, so it corresponds to the green region in panel A. All intensities are coded from dark-blue (low values) to red (high values). This plot is the starting point for the correlation workflow (see Figure 4).

Correlation Workflow In this paragraph we discuss the possibilities in case MALDI and Raman spectra are obtained from the same sample. First the workflow for using complementary spectral information from both techniques is explained. The workflow is shown in Figure 3. After the above mentioned


A Relative Intensity 200 600 1000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

m/z 703

500

600

700 m/z

800

900

m/z 703

B

Normalized Raman Intensity 0.00 0.05 0.10 0.15


C

Page 10 of 19

D

500

1000

1500

2000

2500

3000

-1

Wavenumber / cm

Raman Fingerprint

1000 µm

1000 µm

Figure 4: Workflow of Quantitative Correlation: A: Mean MALDI spectrum of the (Raman) scanned region. B: Integrating a peak region (m/z 703) leads to a false-color image, which can be interpreted as chemical map. This false-color image is related to the green region of Figure 3 A, except that no thresholding is applied. C: By multivariate calibration methods a model can be constructed, which translates a MALDI peak into a complex Raman signature. Here we used a PLS-model to construct a model for the MALDI marker of grey matter (m/z 703). The Raman signature can then be applied on sections without applying MALDI imaging. This allows the use of information associated with the MALDI marker of grey matter (m/z 703) under non-invasive conditions or together with a contrary MALDI matrix. D: The mean Raman spectrum of the region is given. The color inside the spectrum represents the translated Raman marker of grey matter. The Raman signature corresponds to the weights of the PLS-model, which was used to construct C. Peaks which are red are mostly expressed in regions, where the panel B and C show high values (red). On the other hand the blue area in the CH-wavenumber region of panel D is related to the blue and green region of panel B and C.

ACS Paragon 10 Plus Environment

Page 11 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


registration and transformation the transformed Raman spectra can be interpreted along with the MALDI spectra. Thereby it is for example possible to calculate mean Raman spectra for regions, which exhibit a certain intensity of a MALDI peak or MALDI fingerprint. In the present example we used the logarithm of the sum of the m/z 799 (blue) and m/z 703 (green) peak for the MALDI image generation (Figure 3 A). The mean Raman spectra are calculated with a threshold of 9 for the logarithm of the sum MALDI intensity and are plotted in panel B of Figure 3. The color of both (pre-processed) Raman spectra corresponds to the region associated with the calculation. The comparison of the MALDI image (Figure 2 B) with a Nissl stain of an adjacent section (Figure 5) indicates that the m/z 799 peak (blue) is a marker signal for the white matter. The m/z 703 peak (green) can function in contrast as an indicator for the grey matter. The identification of the two selected lipids could unfortunately not successfully be performed by MALDI-TOF MS/MS, since the signal intensity of the two ions was not sufficiently strong enough and, additionally, the window of the precursor ion selector could only be closed up to m/z 5 without losing the precursor ion signal. As mentioned above an unambiguous assignment of two selected lipid peaks is challenging and could unfortunately not supported by performing MALDI-TOF MS/MS, since the signal intensity of the two ions was not sufficiently strong enough and, additionally, the window of the precursor ion selector could only be closed up to m/z 5 without losing the precursor ion signal. Furthermore, the topography of the sample causes a peak broadening, which hinders a high numerical precision of the peak positions. Because the Raman spectra were normalized, an interpretable difference Raman spectra can be calculated. The difference Raman spectra is shown in Figure 3 C. From this difference it is obvious that the blue spectrum shows enhanced Raman peaks at 2890 cm−1 as compared to the green spectrum, which can be assigned to CH2 methylene of lipids. 20,21 In contrast the green Raman spectrum shows noticeably enhanced band intensities at 1170 and around 1255 cm−1 , which can be assigned to the CH bending of tyrosine and the Amide III vibrations, respectively. 20,21 The mean Raman spectrum of the blue region, which was only background corrected, shows a ACS Paragon 11 Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

stronger absolute CH stretching intensity as compared to the mean Raman spectrum of the green region. These findings correspond to a higher lipid content inside the blue region, while the green region is dominated by proteins. The higher lipid content of white matter (blue region) as compared to grey matter (green region) is in accordance with the literature. 22 Due to the fact that the blue Raman spectrum and the green Raman spectrum differ in the intensity of the 2890 cm−1 band of lipids, 20,21 but show almost the same intensity for the 2936 cm−1 band of lipids (CH3 stretching), both regions are different with respect to their lipid composition. This demonstrates that parts of the MALDI information are also represented in the Raman spectra; therefore a Raman model based on specific spectral information obtained from the MALDI experiment can be developed. Besides using complementary information from the MALDI spectra and Raman spectra, it can also be tested if certain spectral information from the MALDI experiment is represented in the Raman spectra or vice-versa. Principally Raman microscopy has a far better resolution compared to MALDI imaging. For the described experimental conditions the Raman resolution is given by the step size of 25 µm, which results in undersampling. However, since the spatial resolution of the MALDI images is about 75 µm, the undersampling is not critical for a comparison with the spectral information obtained from the MALDI experiment. Here we demonstrate how spectral information from the MALDI scan can be extracted from the Raman spectra. This correlation workflow is shown in Figure 4 for the m/z 703 peak, which apparently correlates with the distribution of grey matter. It starts with the MALDI image generated from the m/z 703 peak intensities (Figure 4 A), but without applying thresholding (see Figure 4 B). Subsequently, a multivariate calibration model can be used to model the m/z 703 peak intensity from the corresponding (transformed) Raman scan. Here we utilized a PLS model 23 for that purpose, which constructs a linear model by maximizing the correlation of the predicted MALDI intensity and the recorded MALDI intensity. We used a PLS model with only one component, thus a purely linear model results. This has the advantage that the spectral behavior of the model can be inACS Paragon 12 Plus Environment

Page 12 of 19

Page 13 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


vestigated and visualized. The resulting modeled m/z 703 peak intensity is plotted in Figure 4 C, while the corresponding Raman fingerprint associated with the MALDI peak is visualized in Figure 4 D. The spectral information is encoded in false-colors inside of the mean Raman spectrum plotted in Figure 4 D, ranging from blue (negative) to red (positive). The Raman fingerprint is a marker for grey matter and can be regarded as transformation of the MALDI-marker (m/z 703). The multivariate calibration model is capable of predicting the MALDI marker for grey matter of Figure 4 B, because important features from Figure 4 B are also visible in Figure 4 C. The advantage of this model, is that the measurement, which leads to Figure 4 C, is non-invasive and requires only minor sample preparation. The PLS model can be utilized for the characterization of the metabolome co-localized with the MALDI marker for grey matter. This is achieved by the Raman fingerprint in Figure 4 D, which reflects the bio-molecular composition of grey matter. In order to illustrate the application of the PLS model and hence the use of the MALDI information of the m/z 703 peak intensity without measuring a MALDI image, we recorded a Raman scan of an adjacent brain section. The result is given in Figure 5. A Nissl stained adjacent brain section is visualized in Figure 5 A together with the original MALDI scan in panel B. The CH-intensity of the independent section is visualized in panel C, while the predicted m/z 703 image (marker for grey matter) of the independent section is given in Figure 5 D. The predicted image of the m/z 703 intensity highly correlates (negatively) with the CH-intensity, which was found in Figure 2 as well. Therefore, the region showing a low m/z 703 intensity is characteristic for a region with high lipid content, e.g. white matter region and vice versa. The corresponding Raman fingerprint can be applied with minimal sample preparation and in a non-invasive manner. It should be noted here, that the connection between the MALDI peak intensity of m/z 703 and the corresponding Raman marker signature (Figure 4 D) is not a general one. The transformation of the MALDI marker for grey matter into a Raman signature for grey matter is only valid for mouse brain tissue, which exhibit almost the same bio-chemical composition as the sample used to generate the correlation. If newly measured spectra exhibit different spectral features not ACS Paragon 13 Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

reflected in the spectra used to construct the correlation model, the MALDI marker and also the transformed Raman marker are not working. However, this is not a crucial problem, as a marker for certain diseases or tissue types can be applied only to the type of data used to generate the marker. But the stability of the marker has to be proven individually.

Conclusion and Outlook In this contribution we present, to the best of our knowledge, the first combination of MALDITOF and Raman imaging, which was shown for a cryosection of a mouse brain. We demonstrated how the registration of the MALDI and Raman scan can be performed in a reliable and robust manner. After the alignment and the transformation of the Raman scan to the MALDI coordinate system, both information (Raman + MALDI) can be used for interpretation. Thus, every point of the MALDI grid can be described by a Raman spectrum and a MALDI spectrum. Raman imaging provides information about all important bio-molecules, like proteins, lipids and nucleic acids. This information can be combined with high specific lipid information obtained from the MALDI experiment, if the sample preparation and mass spectrometric parameters for detecting lipids were used for the MALDI imaging. Additional information about nucleic acid composition can be obtained from the Raman spectra together with spectral information about proteins and lipids, allowing a more comprehensive description of the sample (see Figure 3) as possible with only one of the techniques alone. In the presented example the m/z 799 peak acts as a marker signal for white matter, while the grey matter is represented by the m/z 703 peak. Furthermore, we demonstrated, how a certain MALDI signal can be translated into a complex Raman fingerprint. This was achieved by using a multivariate calibration model for predicting the MALDI peak intensity on basis of the transformed and pre-processed Raman scan (see Figure 4). Especially, the MALDI marker signal (m/z 703) for the grey matter, which has been introduced here, can be utilized by the chemometric model (see Figure 4 and Figure 5) without


Page 14 of 19

Page 15 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


recording a MALDI image. The PLS model can be applied to Raman spectra, which results in similar information as the MALDI marker for grey matter (m/z 703). Thus, a MALDI peak was transformed into a complex Raman spectral signature, which can be applied with less restrictions than a MALDI measurement. This Raman fingerprint can be interpreted as metabolome in the spatial surrounding of a certain MALDI-fingerprint. The presented methodology could be utilized to translate any disease marker into a complex Raman fingerprint, allowing an in-vivo application of the marker in a diagnostic procedure. C D

B

Medulla oblongata

A

4

7 6 5

3

1. Hypocampus 2. Artefact (fold crebrellum) 3. Medulla oblongata 4. Striatum 5. Thalamus 6. Lateral ventricle 7. Cerebral Cortex

1

2

Figure 5: A Nissl stain of an adjacent sagittal brain section is visualized in A. Most important anatomical structures are depicted. The MALDI scan of Figure 2 B in panel B. In Panel C the CH-intensity of another adjacent brain section is plotted, which was used as independent test set. The prediction of the MALDI signal intensity for the m/z 703 could be calculated and plotted without performing a MALDI experiment (Panel D). This shows, that the presented framework can be used to transfer information from one technique (MALDI-TOF-MS) to another technique (Raman imaging), which can be applied under different conditions.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 19

Acknowledgement Financial support of the German Research Foundation (DFG) for the research projects PO 563/13-1, PO 563/16-1, EG 102/5-1 and SCHU 1229/15-1 are gratefully acknowledged. The authors A. C. C. and U. S. S. wish to thank the Thüringer Kultusministerium (grant no. B515-07008) for the financial support of this study. T.W.B. and A.C.C. contributed equally to the presented work. Correspondence and requests for materials concerning Raman spectroscopy and chemometrics should be addressed to J.P, while the corresponding author for MALDI-TOF imaging is U.S.S.

Supporting Information Available: This material is available free of charge via the Internet at http://pubs.acs.org

References (1) Krafft, C.; Dietzek, B.; Popp, J. Analyst 2009, 134, 1046–1057. (2) Schwamborn, K.; Caprioli, R. Nat. Rev. Cancer 2010, 10, 639–646. (3) Rahman, S. J.; Gonzalez, A. L.; Li, M.; Seeley, E. H.; Zimmerman, L. J.; Zhang, X. J.; Manier, M. L.; Olson, S. J.; Shah, R. N.; Miller, A. N.; Putnam, J. B.; Miller, Y. E.; Franklin, W. A.; Blot, W. J.; Carbone, D. P. et al. Cancer Res. 2011, 71, 3009–3017. (4) Lui, H.; Zhao, J.; McLean, D.; Zeng, H. Cancer Res. 2012, 72, 2491–2500. (5) Alexandrov, T.; Becker, M.; Guntinas-Lichius, O.; Ernst, G.; von Eggeling, F. J. Cancer Res. Clin. Oncol. 2013, 139, 85–95. (6) Bielecki, C.; Bocklitz, T.; Schmitt, M.; Krafft, C.; Marquardt, C.; Gharbi, A.; Knösel, T.; Stallmach, A.; Popp, J. J.B.O. 2012, 17, 076030–076030.


Page 17 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(7) Delaney, J.; Urbanek, A.; Wehder, L.; Perelaer, J.; Crecelius, A.; von Eggeling, F.; Schubert, U. ACS Comb. Sci. 2011, 13, 218–222. (8) Steven, R. T.; Bunch, J. Anal. Bioanal. Chem. 2013, 1–10. (9) Popp, J., Tuchin, V. V., Chiou, A., Heinemann, S., Eds. Handbook of Biophotonics: Vol. 1: Basics and Techniques; Wiley-VCH, 2011. (10) Masyuko, R.; Lanni, E. J.; Sweedler, J. V.; Bohn, P. W. Analyst 2013, 138, 1924. (11) Petit, V. W.; Réfrégiers, M.; Guettier, C.; Jamme, F.; Sebanayakam, K.; Brunelle, A.; Laprévote, O.; Dumas, P.; Le Naour, F. Anal. Chem. 2010, 82, 3963–3968. (12) Li, Z.; Chu, L.-Q.; Sweedler, J. V.; Bohn, P. W. Anal. Chem. 2010, 82, 2608–2611. (13) R Development Core Team, R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria, 2007. (14) Gibb, S. readBrukerFlexData: Reads mass spectrometry data in Bruker *flex format. 2012; R package version 1.2.4. (15) Morhac, M. Peaks: Peaks. 2008; R package version 0.2. (16) Fortran code by H. Akima R port by Albrecht Gebhardt aspline function by Thomas Petzoldt ([email protected]) enhancements and corrections by Martin Maechler, akima: Interpolation of irregularly spaced data. 2009; R package version 0.5-4. (17) Wolthuis, R.; Tjiang, G. C. H.; Puppels, G. J.; Schut, T. C. B. J. Raman Spectrosc. 2006, 37, 447–466. (18) Schaaff, T. G.; McMahon, J.; Todd, P. J. Anal. Chem. 2002, 74, 4361–4369. (19) Vidová, V.; Novák, P.; Strohalm, M.; Pól, J.; Havlícek, V.; Volný, M. Anal. Chem. 2010, 82, 4994–4997. ACS Paragon 17 Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(20) Notingher, I.; Hench, L. Expert Rev. Med. Devices 2006, 3, 215–234. (21) Krafft, C.; Knetschke, T.; Siegner, A.; Funk, R. H. W.; Salzer, R. Vib. Spectrosc. 2003, 32, 75–83. (22) O’Brien, J. S.; Sampson, E. L. J. Lipid Res. 1965, 6, 537–544. (23) Hasegawa, T. Handbook of Vibrational Spectroscopy; John Wiley and Sons, Ltd., 2002.


Page 18 of 19

Page 19 of 19

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Graphical TOC Entry MALDI-TOF imaging

Correlation

Raman imaging

1 mm

for TOC only


Deeper Understanding of Biological Tissue ... - ACS Publications

Recommend Documents