2.1.

Imaging Hardware and Acquisition

Optical projection/emission images are acquired of a sample suspended in (for example) an agarose gel, mounted vertically on a rotating platform inside a small aquarium made from antireflection-coated glass. The gel rotates inside a high-refractive-index matched fluid, which minimizes refraction, enabling straight-line projection/emission images to be acquired. In optical-CT (transmission mode) light from a uniform back-light traverses through the sample to image formation at the CCD. Projections are acquired in a “step-and-image” manner at small angular increments (e.g., $0.3\mathrm{deg}$ ) through a complete 360-deg revolution. In optical-ECT (emission mode) excitation light causes longer wavelength fluorescent light to be emitted within the sample, which is then also imaged on the CCD, as illustrated. The use of telecentric lenses in the optical imaging system facilitates accurate optical-CT and optical-ECT images, as image formation occurs from light traveling parallel to the optic axis. Acquisition proceeds in the same step-and-image procedure, with the same settings and constraints as already described. A key difference is that narrow-bandwidth optical filters are used to select both the excitation and emitting wavelengths. High-quality filters greatly enhance effective image quality by reducing any contaminant autofluorescence. A DSRed2 filter set was used to select for the excitation and emission wavelengths corresponding to RFP (558 and $583\mathrm{nm}$ , respectively). Exposure of the sample to excitation light was minimized by utilization of a shutter, which permitted exposure only during image acquisition. Consistent interprojection normalization was achieved by keeping the exposure time constant for each projection.

Once a complete set of projection or emission images has been acquired, 3-D reconstructions of the sample were created using the commercial COBRA Feldkamp filtered back-projection algorithm (Exxim Computing Corp, Pleasantown, California). Input parameters were adjusted to reflect the geometries of the image acquisition. In principle, the spatial resolution can approach the resolution of the CCD camera as the resolution of the lens is usually far higher. In practice, however, resolution is lost through the tomographic reconstruction and artifacts arising from motion point spread function and projection noise. Present prototype systems have been reported to have a modulation transfer function (MTF) $\sim 10\%$ at $20\mathrm{lp}∕\mathrm{mm}$ , but substantial improvements are expected. The microoptical-CT scanner presented in Fig. 1 represents a second-generation and very fast scanning system when compared with the first-generation optical-CT devices described elsewhere. ^{20, 21, 22, 29, 30} The time for a typical acquisition of 360 projections was about $5\min$ .

2.2.

Optical Clearing for Optical-CT and Optical-ECT

A significant difference between the optical and x-ray imaging analogs is that the poor optical transmission of biological tissue necessitates ex vivo sample preparation to improve optical transmission (the optical clearing process). Despite this limitation, accurate “in vivo” functional information is entirely feasible because optical stains and fluorescing labels can be applied in vivo, such that representative staining/labeling is achieved for subsequent imaging. Meaningful imaging of any functional parameter therefore requires preservation of the condition of that stain/label through the excision and sample preparation procedures. Optical clearing can be achieved by dehydration and reperfusion of the tissue with a transparent solution of high refractive index close to that of the cell nuclear and organelle membranes. In general, better-reconstructed image quality (both optical-CT and optical-ECT) is directly associated with better optical clearing, provided the signal of interest (here isotonic-ink in the vasculature, and fluorescent proteins) are not affected by the optical clearing process. Illustrative projection images through various unsectioned tissue samples that were cleared and labeled using the processes described later are shown in Fig. 2a . The images were acquired using the system shown in Fig. 1a, with a backlight source of wavelength $633\mathrm{nm}$ . Corresponding transmission characteristics of the samples relative to methyl-salicylate are shown in Fig. 2b. All samples were imaged immersed in methyl salicylate. The lowest transmission was observed for the rat heart because of the high concentration of ink taken up in the vasculature and ventricle chambers. The highest transmission is observed in the brain, where ink uptake was greatly reduced due to the blood brain barrier.

Fig. 2

(a) Light transmission projection images of a variety of optically cleared unsectioned tissue samples, acquired using the imaging system illustrated in Fig. 1a, with a backlight of wavelength $633\mathrm{nm}$ and (b) light absorbtion characteristics of these samples acquired with a spectrophotometer, relative to pure methyl salicylate.

2.2.2.

“Fluorescence-friendly” optical clearing

Oldham ²⁷ investigated the feasibility of 3-D imaging the viable tumor burden in HCT116 tumor xenografts. The HCT116 cell line had been transfected with a gene coding for constitutively expressed³³ RFP. Viable tumor cells therefore express red fluorescence when exposed to the excitation wavelength. The initial challenge was to develop a tissue optical-clearing procedure that would preserve RFP fluorescence in the cleared tissue. A series of plating experiments were performed, where the transfected HCT116 tumor cells were exposed to a variety of tissue-fixing and clearing agents representing different potential clearing processes. Full details of the clearing procedures are given in Ref. 28. The effect on RFP and GFP fluorescence is shown in Fig. 3 . The significant result from these experiments is that substantial fluorescence preservation was achieved when the initial cell fixation was in ethanol. Fixation in either PFA or methanol resulted in almost complete loss of fluorescence [Figs. 3A, 3B]. Furthermore, after ethanol fixation, the cells proved robust to subsequent exposure to either clearing agent BABBs or MetSal. To successfully image RFP in whole xeneograft tumors, the results of the plating experiments must be transferred to whole tumor specimens. This was achieved by first performing perfusion-fixation of the tumors in situ, by aortic cannulation and gravitational drip feed of ethanol. Tumors were then placed in ethanol at 4°C overnight, and set in 0.75% agarose the next day. The clearing procedure involved perfusion graded solutions of water:ethanol, and then ethanol:methyl-salicylate. The total clearing process for tissue samples of order $1{\mathrm{cm}}^3$ typically takes from 1 to 3 weeks, depending on size of sample, under gentle motion-assisted diffusion processes at room temperature. A similar procedure was used irrespective of the particular agent used for dehydration and clearing steps. The sample is initially placed in a 75:25 ethanol:water solution, which is changed to 100% ethanol after $12\mathrm{h}$ , and then refreshed on a daily basis for 4 days. Complete dehydration is necessary to avoid precipitation of opaque regions of agarose, which may occur if water and clearing agent react with agarose gel. The sample is then placed in a 50:50 solution of ethanol:clearing agent. After $24\mathrm{h}$ , the solution is changed to 100% clearing agent, which is also replenished on a daily basis.

Fig. 3

Fluorescent light intensity from cells labeled with RFP and GFP, after exposure to various fixing and clearing solutions. Two cell lines were tested, human colon cancer (HCT116) and human breast cancer (4T1). Comparisons are shown against wild-type (WT) controls, which had no fluorescent labeling. (A) Effect of fixing agents on RFP intensity, (B) effect of fixing agents on GFP intensity, (C) effect of clearing agents on RFP, and (D) effect of clearing agents on GFP. (Based on Sakhalkar ²⁸)

3.1.

Imaging Xenograft Tumors

Oldham ²⁷ presented preliminary data imaging HCT116RFP xenograft colon cancer tumors, containing constitutive RFP labeling, grown on the hind legs of nude mice, following the procedures of an Institutional Animal Care and Use Committee (IACUC) approved protocol. When the tumors had grown to a $\sim 1\mathrm{cm}$ length, labeling/staining of the tumor microvasculature was achieved by tail vein injection and subsequent natural circulation around the body of a double bolus of isotonic india ink and fluorescent probe (lectin conjugated with FITC). The carbon-based ink particles circulate in the blood stream and are phagocytosed inside endothelial cells of the vessels, thereby marking patent microvasculature.³⁴ At 5 to $10\min$ postinfusion, the mouse was sacrificed and the tumor removed for sample preparation. Lectin actively binds to endothelial cells of the microvasculature, providing independent labeling of the microvasculature amenable for optical-ECT imaging. The implementation of both passive and active labeling of microvasculature enables cross-validation and comparison of both techniques. A single optical transmission or projection image of the whole HCT116 tumor is shown in Fig. 4A , and contrasted with micro-x-ray-CT and micro-MRI images acquired with state of the art small animal imaging systems^{35, 36} [Figs. 4B, 4C] to highlight the excellent contrast and resolution of optical-CT. Exquisite visualization of the microvasculature was observed in the optical projection, although the 3-D nature of the vascular network is lost. In this instance, the vasculature is primarily seen on the periphery of the tumor, with a few larger vessels penetrating to the tumor core. As these tumors were implanted subcutaneously, especially dense and intensive vasculature is seen along one side of the tumor where it was attached to the underlying fascia of the animal.

Fig. 4

(A) Projection image of the HCT116 tumor after optical clearing, (B) In vivo micro-CT of the same tumor on hind leg of mouse. The micro-CT image is reconstructed at $100\text{-}\mathrm{\mu }\mathrm{m}$ pixel resolution and involved application of a new vascular contrast agent (fenestra). (C) in vivo T1-weighted contrast-enhanced micro-MRI image of a similar tumor of same size acquired on the 2T system at the CIVM (Centre for Invivo Microscopy at Duke). MRI contrast was achieved with a continuous infusion of magnevist. The location of the tumor is indicated by the dashed line. (From Oldham ²⁷)

Quantitative information is available from CT reconstructions of the projections. Figures 5A, 5B, 5C show such a reconstruction from 240 projections acquired at 1.5-deg increments. The original projection images $(1040\times 1388\mathrm{pixels})$ were downsized to $748\times 998$ (to match present software restrictions) and reconstructed on a $512\times 512\times 512$ grid. The pixel dimensions in the image are thus $\sim 30\mathrm{\mu }\mathrm{m}$ , although a greater number of projections must be acquired to meet the Nyquist criteria for this resolution limit. Significant vascular penetration is observed to be limited to the lower part of the tumor [Fig. 5B], indicating this region was relatively well perfused. The corresponding reconstructions of the emitting FITC distribution within the tumor, acquired with the FITC filters, are shown in Figs. 5D, 5E, 5F. In general, a clear correlation and agreement is observed between the optical-CT and optical-ECT images. Well-perfused regions appear bright in the optical-CT images (corresponding to regions high ink absorption) and also as bright regions in the optical-ECT images (where the scale is inverted such that light pixel values correspond to high emission of light and hence high concentration of FITC). The HCT116 tumor was also imaged in optical-ECT mode with DSRed2 filter set [Figs. 5G to 5H]. This image is significant as it represents the 3-D distribution of RFP emitted by viable tumor cells. The correlations between the corresponding views in Fig. 5 are striking, and clearly show that regions of high RFP expression correlate closely with the well-perfused regions. This makes intuitive sense, as one would expect the more viable regions of the tumor to correlate with perfusion. A precise interpretation is complex due to the novelty of these techniques, and requires reference to more established imaging modalities. Further interpretation of the optical-ECT images would require an attenuation correction, similar to that routinely encountered in SPECT imaging. Comparison with conventional histological sections [Fig. 5J] provided strong supporting evidence for the conclusions derived above from the optical imaging modalities. The peripheral band of well-perfused viable cells, inferred from all three optical reconstructions [Figs. 5B, 5E, 5H] exhibits strong H&E staining in histological section (Fig. 5J) and therefore viable cells in this region. The large central areas devoid of vasculature and viable cells, as determined from the optical modalities, are indeed found to be devoid of H&E stain, indicating regions of necrosis.

Fig. 5

(A) to (C) Orthogonal views of microvasculature of a HCT116 tumor stained with light-absorbing ink and imaged by optical-CT incorporating a white light source. Light regions of high perfusion are clearly visible as containing relatively high amounts of light attenuating ink. (D) to (F) microvasculature of a HCT116 tumor stained with FITC-conjugated lectin and imaged by optical-ECT incorporating the FITC excitation and emission filter set. (G) to (I) the distribution of viable tumor cells in the HCT116 tumor obtained by optical-ECT using the DSRed2 filter set to select for RFP excitation and emission. (J) conventional H&E stained histological section corresponding to views (B), (E), and (H).

The images shown in Fig. 5 were acquired with a prototype microscope-based optical system that incorporated nontelecentric optical components with limited suitability for tomographic imaging due primarily to limited depth of field. More recent images of similar tumors acquired with the bench system of Fig. 1 are shown in Fig. 6 . A significant improvement in image quality is observed, through the removal of artifacts associated with the limited depth of field of the microscope system. Figure 6 also presents the first reconstructions of endogenous gene expression in three dimensions in xenograft tumors using optical-ECT. This was achieved using FITC filter sets and an HCT116 cell line that was double labeled with reporter proteins. Constitutive RFP-labeled viable tumor (as in the preceding), but secondary GFP labeling was incorporated to report the expression of the HIF1 gene.³³ While further work (for example, development of an attenuation correction for optical-ECT) and independent validation are required to fully understand the interpretation of these images, the potential utility is evident.

Fig. 6

Orthogonal views through 3-D reconstructions of HCT116 xenograft tumor imaged by optical-CT and optical-ECT. The tumors were triply labeled, with RFP and GFP flourescent reporter proteins, and light absorbing ink in the vasculature. Constitutive RFP labels the viable tumor burden, and GFP labeled HIF1 expression. The upper row is optical-CT images of vasculature. The middle row is optical-ECT images of the RFP (viable tumor) distribution. The lower row is optical-ECT images of the GFP (HIF1) distribution.

3.2.

Whole Rodent Organ Imaging

Vascular staining in rodent organs was achieved by tail vein injection of isotonic ink as described in Sec. 3.1. Organs were prepared for optical-CT imaging using the procedure outlined in Sec. 2.1. Reconstructed images of a rat heart and lung are shown in Figs. 7 and 8 , respectively. Postimaging histological study of these samples confirmed that the lighter regions in these images correspond to tissues with higher optical attenuation (greater ink staining), and higher perfusion.²⁶ In the heart images, the most prominent features are the coronary arteries, ventricles, and other areas of high blood content. Excellent detail of general heart structure is also observed. Similar observations are made in the lung images of Fig. 8. The major and minor bronchial airways are clearly visible, as are interesting variations in perfusion around the lung.

Fig. 7

Orthogonal views (A) to (C) through a 3-D optical-CT reconstruction of a normal rat heart and (D) a surface rendered image. Labels are LV/RV, right and left ventricles; AV, aortic valve; RA, right atrium; PA, pulmonary arch; AA, ascending aorta; PV, pulmonary veins; and CA, coronary arteries.

Fig. 8

(A) to (C) Orthogonal views through a 3-D optical-CT reconstruction of a normal rat lung and (D) surface rendered image.