Volumetric tomography of fluorescent proteins through small animals in vivo

Zacharakis et al. 10.1073/pnas.0504628102.

Supporting Information

Files in this Data Supplement:

Supporting Figure 6
Supporting Table 1
Supporting Figure 7
Supporting Figure 8

Supporting Figure 6

Fig. 6. Schematic representation of the experimental setup showing the laser source, the light delivery optical fibers (thick red lines), the charge-coupled device (CCD) detector, the optical scanning switch, and the imaging chamber. The two branches used for reflectance (r) and transillumination (t) measurements are also shown. The system used a highly sensitive CCD camera (VersArray, Princeton Instruments, Trenton, NJ), similar to the one used for bioluminescence imaging. The light source was an Argon-Ion laser (Melles Griot Laser Corp., Carlsbad, CA) delivering »40mW of light power onto the animals investigated, at 488 and 514 nm. Light was delivered through a 100-mm multimode fiber (Thorlabs, Newton, NJ) to the input collimator of a custom-made optical scanning device (Nutfield Technology, Windham, NH) that offers multiprojection noncontact photon delivery using a set of two galvanometer-controlled mirrors that scan a focused laser spot (300 mm diameter) through a telecentric lens onto the animal. During the experiments, the animals were placed horizontally on the imaging plate and slightly compressed with the covering glass down to 1.2 cm. The chamber was then filled with an Intralipid (Fresenius Kabi Clayton, Clayton, NC) and India ink solution for minimizing diffuse-wave mismatches. The solution’s optical properties were m_a = 1.25 cm^–1 and m_s' = 16 cm^–1 for absorption and reduced scattering coefficient, respectively, and match optimally the animal boundaries: the middle torso is expected to be of higher absorption. Two types of measurements were acquired during the experiments: fluorescence measurements using a band-pass interference filter centered on 510 ± 5 nm (Andover Corp., Salem, NH) with CCD exposure time of 30 s and intrinsic light measurements obtained by using a band-pass interference filter centered on 488 ± 1.5 nm and exposure time of 0.5 s. The volume of interest was discretized into volume elements according to the size of the field of view of each experiment (usually 23 axial, 23 sagittal, and 17 coronal slices), with a typical voxel size of 1 × 1 ´ 0.7 mm³. The forward model was constructed using Eq.1 in the main text and inverted using the algebraic reconstruction technique with positive restriction. The reconstruction procedure involved solving the inverse problem, U^nB = Wn, where W represents the weight matrix mapping the vector of fluorochrome concentrations

into the measurement vector U^nB. Typical inversion times were 5 min on a 2-GHz Intel Pentium 4 processor for 100 iterations.

Supporting Figure 7

Fig. 7. Histology images of the GFP lung tumor. (a) White light image. (b) Fluorescence image in gray scale. (c) Fluorescence image with pseudocolor. The images correspond to a 30-mm-thick histology slice of tissue excised from the lung. Two different types of tissue are clearly seen in the white light image. In the two fluorescence images, however, the separation of the two types of tissue, with normal lung tissue on the left-hand and GFP-expressing tumor tissue on the right-hand side, is clearly visualized because only the tumor tissue exhibits fluorescence activity. The slices were created with a microtome (CM1900, Leica Microsystems, Bannockburn, IL), and the images were obtained with a fluorescence microscope (Zeiss Axiovert 100TV).

Supporting Figure 8

Fig. 8. Three different views of a 3D rendering of the skeleton and skin of the mouse based on the computed tomography (CT) data with an overlay of the optical fluorescent protein tomography (FMT) reconstruction, here shown with red color by the red arrows for better illustration of the location of the tumor inside the lung. All three images correspond to day 20 after the injection of the cancer cells. CT and optical examinations were successively obtained. Special care was taken, using appropriate mouse holders, that the mice were positioned in a similar way in both modalities, and surface information was used to guide the coregistration using a semiautomatic procedure that allocates the centers of each acquisition using user input and then automatically aligns the images using geometrical information from both modalities based on known markers on both images.

Table 1. Spectral strengths of the fluorescence of GFP- and YFP-expressing cells at the two spectral regions of interest

	Spectral strengths
Protein	510 ± 5 nm	570 ± 5 nm
GFP	0.8301	0.2963
YFP	0.1699	0.7037

These spectral strengths were measured deep in the body of a mouse from transillumination data. The numbers are normalized to the sum of the raw counts of both GFP and YFP signals at the two spectral bands.