Integrating spatially resolved 3D MALDI imaging mass spectrometry with in vivo MRI

Transcription

1 Integrating spatially resolved 3D MALDI imaging mass spectrometry with in vivo MRI Tuhin K Sinha, Sheerin Khatib-Shahidi, Thomas E Yankeelov, Khubaib Mapara, Moneeb Ehtesham, D Shannon Cornett, Benoit M Dawant, Richard M Caprioli & John C Gore Supplementary figures and text: Supplementary Figure 1. Example 2D MALDI IMS data on a coronal mouse brain section. Supplementary Figure 2. Reconstructed blockface volumes for a mouse head and whole mouse. Supplementary Figure 3. Spatially resolved volumetric MALDI IMS in a mouse head. Supplementary Figure 4. Spatially resolved three-dimensional MALDI volumes coregistered to in vivo magnetic resonance imaging data from a tumor laden mouse brain. Supplementary Methods Supplementary Videos 1 3 are available on the Nature Methods website.

2 Supplementary Figure 1 Example 2D MALDI IMS data on a coronal mouse brain section. (a) Image of the tissue subjected to the MALDI IMS process. (b-d) Integrated ion images for peaks designated in the average mass spectra shown in (e). (e) The units for the x-axis represents (m/z), and the y-axis represents intensity in arbitrary units.

3 Supplementary Figure 2 Reconstructed blockface volumes for a mouse head and whole mouse. The mouse head was sectioned coronally in an anterior to posterior fashion. (a) The volume before reconstruction. (b) The volume after reconstruction. Substructures in the brain are more clearly resolved in the reconstructed volume. Features of interest include the left eye of the mouse, the jaw, and corpus callosum and ventricles in the brain. (c) and (d) The unreconstructed and reconstructed volumes for the whole mouse, respectively. Gross anatomic features are readily apparent in the reconstructed volume, including: the spine, the liver, the stomach, the heart, and the brain. The X, Y, and Z axes are labeled to provide physical space dimensions and spatial cues for the orientation of the blockface volume (units in centimeters).

4 Supplementary Figure 3 Spatially resolved volumetric MALDI IMS in a mouse head. (a) and (e) Ion images for m/z 5826 and 14182, respectively, overlaid onto the corresponding blockface image. (b-d) and (f-h) Three MALDI IMS ion images concatenated into a volume and rendered along an oblique image plane through the mouse head. The oblique plane is provide to demonstrate the spatial extent of the mouse head and also the volumetric MALDI IMS data. The X, Y, and Z axes are labeled in centimeters.

5 Supplementary Figure 4

6 Spatially resolved 3D MALDI volumes co-registered to in vivo MR imaging data from a tumor laden mouse brain. The top row shows three different views of an oblique blockface section and coronal T2 image coregistered to each other. The oblique image plane is presented to give reference to the extent of the mouse brain. The coronal image shows contrast variations related to the tumor and injection insult. Each subsequent shows a MALDI IMS volume of 20 slices volume rendered in the same views as the first row. The following m/z peaks are represented: 6741, 7346, 13820, 14128, and 14966, respectively. The injection insult is on the left hand side of the T2 image. Rows 4 and 6 show two peaks which are spatially localized to the injection insult. We believe the peak in row 4, m/z 13820, to be associated with the tumor as it manifests in the contralateral ventricle. Peak m/z is more isolated it its distribution and we believe it to be associated with the response to injection injury.

7 Supplementary Methods Establishment of Intra-cranial Glioma in a Mouse: Six to eight week old athymic nude (nu/nu) mice were obtained (Charles River Laboratories, Wilmington, MA). Following intraperitoneal anesthesia using Ketamine (30 mg/kg) xylazine (30 mg/kg), the right striatum was stereotactically injected with 10 5 GL26- glioma cells in 2.5µL of PBS. All animal use was performed in strict accordance with Animal Care and Use Committee guidelines in force at Vanderbilt University Medical Center. Blockface reconstruction The microtome and image acquisition system were not intrinsically registered, so the precise location of each blockface image acquired was not known. Briefly, in our microtome, the ice-block was moved across a stationary blade to section the tissue. After sectioning, the blockface cannot be returned to its previous position accurately and precisely, causing misalignment between serial blockface images. As a result, a retrospective inter-slice registration was used to align consecutive blockface images. For each set of blockface images, the second slice was aligned to the first. The registered slice was then used as the target of alignment for the next consecutive slice in the blockface volume. This process was iterated upon until all of the images had been registered, and concatenating these together yielded an accurately reconstructed blockface volume. An automatic intensity-based, pattern-matching algorithm was used for inter-slice registration based upon Normalized Mutual Information 9. The NMI metric is optimal when two mis-registered images are correctly aligned, and therefore optimizing NMI as a function of transformation parameters allows one to correctly register two mis-registered images. In this paper, NMI was optimized for x-y translation and in-plane rotation to

8 register sequential images in the reconstruction process. The specific implementation of the algorithm used in this paper was provided by the Insight Toolkit ( It is important to note that registration via NMI requires no fiduciary systems; the NMI metric is calculated using the overlapping intensity values in the two images being registered. Thus, the blockface volume reconstruction is completely automatic. MALDI IMS Data Acquisition Tissue samples were prepared using techniques described in a previous study 3. Collected tissue sections were transferred using rice paper to gold-coated MALDI target plates (Applied Biosystems Inc.) and spraycoated with a 25 mg/ml sinapinic acid matrix solution prepared in 60% acetonitrile, 0.5% trifluoroacetic acid. Approximately 10 ml of matrix solution were needed to produce a homogeneous matrix crystal layer. Matrix coated samples were then analyzed on a linear MALDI-time-of-flight mass spectrometer (Autoflex II, Bruker Daltonics Inc.) equipped with a Smartbeam laser operating at 100 Hz. 3D Spatially Resolved MALDI IMS Post-Processing After acquisition a series of post-processing steps were performed to recreate spatially resolved 3D MALDI IMS data. The first step was to align the MALDI mass spectra to the targeting image used by the mass spectrometer. The spectrometer reports the locations of the acquired spectra within the field-of-view (FOV) in terms of internal motor coordinates. These internal motor coordinates are related to a targeting image using training fiducials, which were selected at the time of acquisition. Pixel locations of the training fiducials in the targeting image are registered to their corresponding motor coordinates using an affine alignment. Once this relationship was established, locations described in internal motor coordinates within the mass spectrometer could be transformed into the targeting image s pixel space.

9 The final step required in transforming the MALDI IMS data to their spatial locations was the alignment of the targeting image and corresponding reconstructed blockface image. This alignment was provided via contour based registration of shape features in each image, such as the outline of the brain, head, or whole-body section. The contours were manually highlighted, although automatic methods could also be employed, and an iterative closest point (ICP) algorithm was used to align the corresponding contours 1. This transform places targeting image coordinates into blockface image coordinates. The slice location, or z-offset, in the axial direction was calculated using the slice number and the thickness of each slice, and was appended to the 2D blockface coordinate to provide a 3D physical space location for each MALDI IMS spectra. The result of these steps provided a continuous transformation from the 2D motor coordinates to 3D animal specific coordinates (see Supplementary Fig. 3). MR imaging in vivo The tumor implant was allowed to develop for three weeks and then imaged at 7T (Varian, Inc.), equipped with a 38mm birdcage coil, to produce quantitative parametric images of the T 1, T 2, and ADC. T 1 maps were obtained using multiple flip-angle gradientecho multi-slice images (96 x 96 x 9, FOV = 19.2 x 19.2 x 1.8 mm, TR = 200 msec, TE = 4.5 msec, = {15, 30, 45, 60, 75, 90} degrees). The weighted images were then fitted, non-linearly, to the following function to generate voxel-by-voxel measurements of T 1 : T 1w ( ) = C 0 (1-e -TR/T 1)*sin( )/(1-cos( )*e -TR/T 1) + C 1, (1) where, T 1w ( ) is the per-voxel measured signal at each flip angle, TR is the repetition time for the MR experiment, T 1 is the fitted parameter, and C 0 and C 1 are constants used to account for baseline signal and measurement offset, respectively. T 2 maps were obtained using multi-echo multi-slice spin-echo images (96 x 96 x 15, FOV = 19.2 x 19.2 x 3.0

10 mm, TR = 4 sec, TE = {12, 24, 36, 48} msec). The weighted images were then fitted, non-linearly, to the following function to generate voxel-by-voxel measurements of T 2 : T 2w (TE) = C 0 *e -TE/T 2 + C 1 (2) where, T 2w (TE) is the per-voxel measured signal at each echo-time, TE is the echo-time for the MR experiment, T 2 is the fitted parameter and C 0 and C 1 are used to account for baseline signal and measurement offset, respectively. ADC maps were made using multiple diffusion gradient spin-echo multi-slice images (96 x 96 x 15, FOV = 19.2 x 19.2 x 3.0 mm, TR = 2 sec, TE = 30 msec, b-values = {0, 100, 200, 400, 800} s 2 /mm). The weighted images were then fitted, linearly, to the following function to generate voxel-by-voxel measurements of ADC: ln(s b )= -(b-value)*adc + C 0 (3) where ln(s b ) is the per-voxel logarithm of each diffusion weighted image, b-value is the magnitude of diffusion weighting for the MR experiment, ADC is the fitted parameter, and C 0 accounts for the measurement offset. All MR measurements were made with an isotropic voxel size of 200 m. References. 1. Besl, P.J. and H.D. McKay, A method for registration of 3-D shapes. Pattern Analysis and Machine Intelligence, IEEE Transactions on, (2): p