Using CARS and SHG on the same microscope platform (see Figure S2
), we imaged the composition and structural organization of mammary gland and tumor stroma. We found that multimodal NLO imaging enables simultaneous visualization of adipocytes, collagen fibrils, and tubular structures representing blood capillaries (, supplementary movie 1
). Because imaging blood capillaries without labeling has not been reported previously, we labeled the endothelial cells lining the blood vessels with fluorescein isothiocyanate conjugated isolectin B4
The TPEF image in exhibited strong FITC-IB4
staining of tubular structures, as well as activated macrophages, in the mammary gland (supplementary movie 2
We observed that most adipocytes were surrounded by collagen fibrils in the mammary gland (, supplementary movie 3
). In addition, we showed that CARS enabled imaging of tumor cells in mammary tumors without any labeling (). Tumor cell visualization was possible owing to weak CARS signals arising from the cell membrane’s lipid bilayer and strong CARS signals from numerous small lipid droplets surrounding the nucleus, which gives a dark contrast in the CARS image (supplementary movie 4
). Tumor cell visualization was further confirmed by DiOC18 fluorescence imaging (see Figure S4
In contrast to the orientation of collagen fibrils around adipocytes in the mammary gland, we found that most collagen fibrils were located at the outer perimeter of the tumor mass in mammary tumor (see , and Figure S5
). This spatial organization of collagen relative to tumor mass is consistent with a previous report by Ahmed and colleagues in which SHG and TPEF microscopy were used to image the organization of collagen and mammary tumor cells expressing green fluorescent protein in transgenic mice, respectively.13
However, the use of CARS microscopy to image tumor cells does not require any labeling and should be applicable to image any type of tumor.
Figure 1 Nonlinear optical imaging of mammary gland and tumor stromal composition. A, Adipocytes and blood capillaries (arrows) imaged with coherent anti-Stokes Raman scattering (red) and collagen fibrils imaged with second harmonic generation (SHG) (green) in (more ...)
With the demonstrated capability of CARS and SHG for label-free imaging of adipocytes and collagen, respectively, we proceeded to evaluate the impact of obesity on their composition in mammary gland and tumor stroma. We imaged mammary tissues collected from three rats for each of the four experimental groups: lean rat chow, obese Western, lean rat chow tumor, and obese Western tumor. Because SHG intensity is directly correlated with the concentration of collagen fibril type I, we evaluated the relative collagen content of a tissue based on total SHG intensity.14
We found that there were strong correlations between obesity, ECM collagen content, and adipocyte size. In the mammary gland, we observed higher collagen content in lean rats on the rat chow diet compared with obese rats on the Western diet (). Conversely, in mammary tumor stroma, we observed higher collagen content in obese rats on the Western diet compared with lean rats on the rat chow diet (). To systematically compare the ECM collagen content from one animal with that of the other, we defined a fixed analysis volume and collected total SHG signals from nine volumes in the mammary gland or tumor stroma of each rat (see Methods
; Figure S6
). The total SHG collagen intensity as a function of rats in different diet groups is plotted in (see Table S1
). Consistent with our observation, quantitative analysis of SHG intensity indicated that obesity decreases collagen content by an average of 5-fold in the mammary gland and increases collagen content by an average of 14-fold in mammary tumor stroma.
Figure 2 Coherent anti-Stokes Raman scattering imaging of adipocytes (red) and second harmonic generation imaging of collagen fibrils (green) to evaluate the impact of obesity on mammary gland and tumor stromal composition. Representative images (single frames) (more ...)
Figure 3 Analysis of the impact of obesity on mammary gland and tumor stromal composition. Mammary gland and tumor tissues of three rats from each animal group were analyzed for collagen content and diameter of lipid droplets (LD) of adipocytes. A, Total second (more ...)
To analyze the impact of obesity on adipogenesis in mammary stroma, we used CARS to evaluate the size of lipid droplets in adipocytes. By focusing CARS excitation beams at the equatorial planes, we measured the diameters of 100 lipid droplets of adipocytes in each rat in both diet groups (see Table S1
). The average lipid droplet diameter as a function of lean and obese rats is plotted in . We found that, on average, the diameters of lipid droplets from obese Western rats were two fold larger than those from lean rat chow rats. This observation indicates that obesity increases the size of lipid droplets in mammary adipocytes.
Finally, we analyzed standard histologic tissue sections by CARS and SHG and compared the images with those obtained from fresh tissues. We found that most of the SHG collagen signal came from areas surrounding mammary ducts and terminal end buds in mammary gland tissue sections (). Consistent with fresh tissue analysis, we observed that most collagen signal was found at the perimeter of the tumor mass (). However, when comparing two-dimensional images of histology samples, we could not observe any correlation between obesity, the size of lipid droplets of adipocytes, and relative collagen content in mammary gland or tumor stroma. Given the fact that tissue sections were not sliced at the equatorial plane of the adipocytes, it became obvious that we could not compare the size of lipid droplets of adipocytes from one histologic tissue section with another. In addition, each tissue section was approximately 5 µm thick, less than the average size of a tumor cell (≈10 µm); therefore, collagen fibrils present primarily outside the tumor mass could not be assayed. It is conceivable that serial sectioning of tissues and careful three-dimensional image reconstruction analysis of standard histology samples should allow evaluation of the impact of obesity on mammary gland and tumor stroma. However, given the tedious works involved with serial histology section analysis, a simpler alternative can be found with NLO imaging, which has intrinsic three-dimensional sectioning capability.11
Taken together, our results show that three-dimensional imaging of mammary tissues enabled analysis of the impact of obesity on mammary stroma not readily accessible by standard two-dimensional histologic evaluation.
Figure 4 Coherent anti-Stokes Raman scattering imaging of lipid (red) and second harmonic generation imaging of collagen fibrils (green) of standard histologic tissue sections. Histology of the mammary gland of (A) a lean rat chow rat (LRC), (B) an obese Western (more ...)