Three dimensional multiphoton imaging of fresh and whole mount developing mouse mammary glands

Wild-type FVB mice, or mice carrying an enhanced Green Fluorescent Protein (GFP) transgene under the control of the chicken beta-actin promoter coupled with the cytomegalovirus (CMV) immediate early enhancer (FVB.Cg-Tg (ACTBEGFP) B5Nagy/J strain were obtained from The Jackson Laboratories, Bar Harbor, Maine). HAI-1 transgenic FVB mice (unpublished), that overexpress the Kunitz-type protease inhibitor HAI-1 in their mammary glands were obtained from an in-house breeding colony. Mice were maintained and bred within the Division of Comparative Medicine animal facilities with unrestricted access to food and drinking water. Animals were examined at least once daily. All protocols involving the animals have been approved by the Georgetown University Animal Care and Use Committee (GUACUC) and carried out under the protocol No. 09043. After euthanizing animals according to the approved protocol, mammary tissue (the #3 or #4 glands) was harvested from 6–8 week or six month old wild-type, GFP, or HAI-1 mice (from 1–2 animals each) and spread out on microscope slides or cover glasses (Gold Seal #1) to mimic the arrangement of the tissue in the mouse as closely as possible. The glands were imaged immediately and then fixed and processed for whole mount preparation essentially as previously described [17]. Briefly, the whole mammary gland specimens were fixed in a 3:1 mixture of 100% ethanol and glacial acetic acid for 60 minutes at room temperature, washed with 70% ethanol for 15 minutes, and rinsed with deionized water. The glands were then stained overnight in Carmine Alum (1 g Carmine, #C1022, 2.5 g Aluminium potassium sulphate, #A7167, Sigma St Louis, MO in 500 ml of water) overnight at 4°C. The tissues were rinsed in deionized water and then dehydrated through graded alcohols and cleared in toluene prior to mounting with cover slips or microscope slides with Permount (Fisher Scientific, Pittsburgh PA).

Confocal, bright field, and multiphoton images were collected using a Zeiss510/META/NLO microscope (Carl Zeiss, Thornwood, NY) with Zeiss EC Plan-Neofluar 10x/0.30 M27, a Zeiss LD LCI Plan Apo 25x/0.8 lens (working distance 0.57 at 0.17 mm cover glass), or a Zeiss c-Apochromat 40X/ 1.2 NA W corr. UV–VIS-IR M27. For transmitted SHG signals, a Zeiss achromatic condenser 0.8 H D Ph DIC was used with the polarizer replaced by an IR filter to block excitation IR in the emission. In Figure 1, images were made using a Nikon SMZ-1500 epi-stereofluorescence microscope (Melville, NY) equipped with a Pan Apo 1X Nikon lens and GFP filter or (Ex 480/40 nm, dichroic 505 nm, Em 535/50 nm) [17]. All confocal images were collected using a pinhole diameter of 1 Airy unit and multiphoton images were collected with the pinhole aperture set at 1000 μm (maximum) for internal detectors. Backward scattered second harmonic generation (SGH-B) was detected using an internal detector with maximal pinhole aperture 1000 μm) and forward scattered SHG (SHG-F) using a non-descanned detector (NDD) positioned in the transmission pathway, just above the condenser of the inverted microscope (ChD). Table 1 details imaging conditions for live and whole mount tissue experiments.

An NDD located in the transmitted light path just beyond a Zeiss ACHR COND 0.8 HD PH DIC condenser and IR filter is a useful addition to image transmitted SHG and/or transmitted fluorescence. The major signal from the ChD transmittance NDD detector is the SHG-F originating from collagen fibers and Carmine Alum or GFP signals which are much weaker by comparison. For these studies, these signals were not separated, although addition of emission filters could be used to isolate the transmitted SHG signal if desired. Additional file 1: Figure S1 illustrates the Zeiss510/META/NLO AIM software specific setup to obtain SHG-B, Carmine Alum, and SHG-F signals using a single excitation of 860 nm and internal filter set combinations in Ch2, Ch3, and ChD.

All images are presented in raw form (no post processing) with the exception of the Z-image stacks which were processed for 3D reconstruction as described using Image J 1.44p (NIH, Bethesda, MD), Metamorph Offline ver. 7.7.1.0 (Molecular Devices, Sunnyvale, CA), or Zeiss 510 META (with Physiology Software ver. 3.5 and Multiple Time Series) Software ver. 3.5 (Carl Zeiss MicroImaging, LLC, Thornwood, NJ). Contrast and brightness settings were changed identically for comparisons.

Excised mammary glands from GFP-mice were first examined using conventional bright field with a stereofluorescence microscope (Figure 1A a-c). Next, by way of comparison, live, excised glands from GFP-mice were examined with fluorescence imaging using the same stereofluorescence microscope (Figure 1B a-c). Notably, at higher magnification, the terminal end buds (TEBs) were obscured by the presence of surrounding fat cells in the stroma (Figure 1B c).

Multiphoton imaging of mammary glands was performed to improve resolution and to achieve the benefits of three dimensional imaging (3D). First, emission from GFP and SHG signals were studied in live mouse mammary glands (Additional file 2: Figure S2A) and skin tissue (Additional file 2: Figure S2B-C) by collecting emission scans from 361–704 nm. At excitation 860 nm, GFP emission appeared as a peak at 506 nm with a shoulder at approximately 549 nm (Additional file 2: Figure S2A-B, green curve, eGFP peak). SHG-B appeared as a sharp peak centered at 431 nm (Additional file 2: Figure S2B, SHG-B peak). Images were extracted from the lambda data at Em 404–446 nm, 446–478 nm, 500–532 nm, and 596–730 nm (Additional file 2: Figure S2C). The SHG-B signal was reasonably well separated (Em 404–446, Additional file 2: Figure S2C). However, background contributed to the GFP image at Em 500–532 nm. Images lacking the GFP signal and containing only background signal were observed at Em 446–478 nm and Em 596–703 nm (Additional file 2: Figure S2C). Notably, the peak of the autofluorescent signal was Em 495 nm, whereas the peak of GFP was Em 506 nm.

Using bandpass filters and single track imaging with MP excitation at 860 nm, SHG-B and GFP signals were collected for a single living, ex vivo TEB to generate Z-stacks (Figure 2; Table 1 for imaging details). In red, the SHG-B images depict the changing arrangement of collagen fibers with depth, i.e. a more linear and parallel arrangement of fibers at z =15 μm compared with the more disordered and wavy appearance at a shallower depth of imaging at z = 6 μm (Figure 2A). SHG-B signal disappeared deeper into the tissue beyond the TEB (Figure 2A-B, D). In addition to TEB epithelial cells, stromal cells were observed scattered within the ECM layer surrounding the TEB (Figure 2A, z = 25 μm). The TEB epithelial cells seen at higher magnification include TEB body cells and cap cells (Figure 2C: z = 56 μm, Cap, arrows; z = 35 μm, Body cells). However, TEBs imaged beneath the SHG-B-positive fiber layer lying within the fat tissue of the mammary gland appeared to have shadow artifacts arising from the outline of the fat cells (Figure 2D, arrow).

We next compared SHG-B and SHG-F signals obtained from live tissue. Interestingly, the two signals, SHG-B and SHG-F, imaged different fibrils, even when collected near the surface of the outer ECM layer, whether or not they had the same orientations (Figure 3A and B, compare green and red fibrils in 8 μm and 29 μm XY planes). SHG-F was unfiltered and included the entire spectrum available in ChD, thus contributions from GFP were included so that some cell bodies were observed in Figure 3A (8 μm, arrow). Whereas SHG-B signal was relatively unaffected by the light scattering due to the outlines of fat cells, SHG-F was dramatically affected as visualized in a 3D reconstruction (Figure 3C). This is likely due to the fact that the transmitted signal SHG-F passes through the entire thickness of the tissue containing fat cells prior to detection of the SHG-F by the non-descanned detector (NDD) which was located in the light path after the condenser. In single XY slices, fiber orientation was easily observed by SHG-B (Figure 3D). Average intensity scans at three different ROIs reveal the variation in depth and/or intensity of the SHG-B signal (Figure 3E, graph, peaks of intensity).

A comparison of the cellular and fibrillar structure of TEBs and the surrounding stroma was performed in live and whole mount tissue to determine whether major artifacts are introduced by whole mount preparation. Thus, the same TEBs were imaged in living tissue immediately after excising the gland and then again following fixation, defatting, and Carmine Alum staining. The living and whole mount tissue was first compared using SHG-B and SHG-F imaging of collagen fibers together with either GFP or Carmine Alum fluorescence (Figure 4, Live (GFP) Tissue and Whole Mount, solid arrows indicate the same fibers in live versus whole mount images). The XY scan of the surface fibrillar layer revealed that the fibers were compressed around the TEB in the whole mount preparation, whereas in the live tissue they were less compressed (Figure 4A).

The SHG-F signal was barely detectable in the live tissue, but was prominent in the whole mount (Figure 4A, Live Tissue and Whole Mount, blue, solid arrows). The GFP in live cells is included in the SHG-F (unfiltered) channel (Figure 4A, Live Tissue, dashed arrow, blue). Similarly, in the whole mount, both the SHG and the Carmine Alum signal are seen in the unfiltered SHG-F channel (Figure 4A, Whole Mount, blue, arrow indicates SHG-F, remaining blue is Carmine Alum). In live tissue, individual epithelial cells in the TEB as well as surrounding stromal cells are visualized (Figure 4B, Live Tissue, z = 18). Surrounding the TEB, a layer of fibers is best observed in the Whole Mount preparation (Figure 4, Whole Mount, XZ views). In addition, in the whole mount, SHG-B and SHG-F signals were acquired significantly deeper into the tissue; in fact a layer of fibrils associated with a blood vessel is apparent in the whole mount that is not imaged in the live tissue (Figure 4B, Whole Mount, asterisks XY view, arrows in XZ and YZ views). Figure 4C is a low magnification view of the TEB imaged in Figure 4A-B (asterisk). A kymograph was generated for live and whole mount tissues along a line at approximately the same site and an overlay created to further illustrate the difference in imaging depth of SHG-B between the live and whole mount tissue (Figure 4D, red is whole mount SHG-B and green is live SHG-B, asterisk indicates TEB position). Thus, the combination of SHG-B and SHG-F, together with Carmine Alum signals in whole mounts increases the depth at which tissue architecture can be observed compared with live mammary gland tissue.

Most experiments with mouse glands involve large complex sample collections in which multiple glands (minimum of 2 usually) in multiple animals are harvested for each treatment or time point. Carmine Alum stained whole mount preparations are typically prepared from these experiments so that the data can be analysed according to researcher and instrument availability without concern for sample deterioration. Previously, we had determined that Carmine Alum is fluorescent and can be imaged in 3D [27]. For those experiments, excitation of the Ti-Sapphire MP laser was set to 750 nm and emission was collected in the red channel, 565–615 nm, although mammary gland whole mounts can also be imaged using confocal microscopy with a visible HeNe green laser with excitation at 543 nm (compare Figure 5A. Ex 543 nm with Figure 5C, Ex 750 nm). Bright field imaging of the Carmine Alum staining at the same magnification was not illuminating compared with the fluorescence confocal imaging (compare Figure 5B to 5A).

As an example of the power of 3D imaging of Carmine Alum staining, mammary glands from HAI-1 mice were imaged in the same manner as Figure 5C (Ex 750/EM 565–615 nm). A 3D reconstruction of a TEB is shown in Figure 6A and orthogonal views in Figure 6B. These mice exhibit delayed mammary gland development, and analysis of H&E stained paraffin sections of the mammary glands (but not bright field images) had suggested that the structure of the TEBs was abnormal. 3D imaging of the TEBs made it much easier to appreciate and quantify the nature of their abnormal structure. Furthermore, the evidence for these abnormalities could not be obtained by examination of bright field images (Figure 6C). Orthogonal views (XZ and YZ) to a single plane XY image reveal that the central lumen is not connected with a pocket forming a defect reaching to the surface of the TEB (Figure 6A-B, arrows for abnormal pocket, asterisk marks the central ductal lumen). Examination of a movie through the Z-planes confirms a lack of connection between the two lumens (Additional file 3: Movie S1). Observations of dozens of TEBs to determine the number of abnormal pockets per TEB would have been extremely time consuming and would have to be done using serial sections of traditional paraffin embedded tissue.

In a final experiment, details of the ductal side branches were explored in a mouse model of HAI-1 in which potential abnormalities had been identified previously at the bright field level (Johnson, unpublished data). Imaging of SHG-B, Carmine Alum and SHG-F was performed and the resulting XY, 3D and orthogonal XY, XZ, and YZ images were compared (Figure 7A-D). It is apparent that the single XY slice and 3D views reveal multiple details not available from paraffin sections or bright field imaging of Carmine Alum-stained whole mounts (Figure 7F). First, the many lateral buds present along the duct appear mostly in a single plane (XY) (Figures 7A and B, compare XY slice and 3D views). Second, the lower power views (Figure 7A and B) and the inset of each shown in Figure 7C and D) reveal the association of SHG-B and SHG-F positive fibrils with the duct. The increased fibrillar texture is observed adjacent to the duct on the side containing the lateral buds (Figure 7C, XY views, asterisks). The 3D views illustrate that the fibrillar layer indeed surrounds both the duct and the lateral buds (Figures 7D, 3D, solid arrow), and extends between the buds (Figure 7C, dashed arrows). Carmine Alum staining serves to intensify the epithelial signal, but visualization of ductal cells is still possible when the Carmine Alum image plane is deleted in the multicolour view since unfiltered fluorescence from Carmine Alum appears in the transmitted channel in this particular experiment (Figure 7E).

Figure 8 illustrates more completely the architecture of the collagen fibers surrounding a normal TEB in a Carmine Alum stained whole mount and that the SHG-B signals often predominate at the surface of the gland closest to the cover glass compared with SHG-F signals that are also apparent beneath the TEB (further into the tissue away from the cover glass) (Figure 8A-B). The SGH-B signal (red), although intense on the surface of the gland, is diminished beneath the TEB (Figure 8B, red, note position of TEBs marked by asterisks), whereas the SHG-F signal (green) appears most intensely on the deeper edge of the TEB (Figure 8B, green, note position of TEBs marked by asterisks). In XY regions imaged on either side of the TEB, both SHG-B and SHG-F signals nearer and farther from the cover glass are collected in the final image (Figure 8B, arrows) illustrating the likelihood of a “shadowing” effect on the SHG signals by the Carmine Alum-stained TEB. In the case of SHG-B, the shadowing effect occurs on the far side of the TEB deeper into the tissue, whereas the shadowing effect for SHG-F occurs on the near side of the TEB. This is due to the differing light paths to the detector for each. However, we infer that SHG-B signal is indeed present on the lower aspect of the TEB since in adjacent areas it is present where TEB tissue is not interposed between the source of the SHG-B and the epi-detector. Similarly, we infer that the SHG-F signal is present above or on the near side of the TEB since in areas adjacent to the TEB it appears (Figure 8B, arrows). We conclude that the combination of SHG-B and SHG-F gives a good idea where layers of fibers occur, but in a non-quantitative manner.

The difference in SHG-B and SHG-F signals cannot be explained solely on light scattering properties of Carmine Alum-stained epithelial cells within the TEBs. Both SHG-B and SHG-F signals are seen in fibers within the same layer and in fibers associated in parallel orientations adjacent to TEBs (Figure 9A-D) as well as associated with blood vessels (Figure 9E-H). Some fibers appear to emit both SHG-B and SHG-F signals (Figure 9D). Fibers associated with blood vessels create more intense SHG signals compared with fibers in the rest of the mammary gland stroma, but similarly contain both SHG-B and SHG-F signals originating from different fibers (Figure 9E-H, arrows point to non-vessel associated fibers).

Additional file 4: Figure S3 illustrates the loss of SHG-F signal when a TEB is positioned between the collagen fibrils and the NDD. In this example, the SHG-B signal from collagen fibers surrounding the blood vessels is nearer to the coverslip than the TEB and so the reflected signal is efficiently recovered by the internal PMT detector (Additional file 4: Figure S3A). In contrast, the SHG-F signal must first travel though the thickness of the stromal tissue and then is scattered by the TEB and fails to be detected by the NDD (Additional file 4: Figure S3A). Although the laser penetrates to the TEB, the farther half of the Carmine Alum-stained TEB (relative to the coverslip) is often poorly imaged, but some detail is imaged in the in the SHG-F channel due to the autofluorescent or Carmine Alum bleed through (Additional file 4: Figure S3A-B, note blue to green transition of the TEB with Z-depth). There is a clear loss of SHG-B signal when the Carmine Alum-stained TEB is situated between the coverslip and the source of the SHG-B signal (Additional file 4: Figure S3C, note SHG-B signal in red channel). Thus, using both the signals from SHG-B and SHG-F, the approximate architecture of collagen fiber layers can be discerned. To illustrate this, three dimensional views of a TEB associated with layers of fibers imaged by SHG are shown in Additional file 5: Figure S4 and Additional file 6: Movies S2-5. They illustrate the possible influence of orientation of the layer of fibers upon directionality of TEB elongation and growth since the TEB itself is angled away from the fiber layer with respect to its attached ductal structure. Interpretation of the type of collagen fibers based on SHG-B versus SHG-F detection, however, is limited in these images due to the “shadowing” artifact.

Whole mounts of fixed, unstained GFP and non-GFP mice were compared in order to assess whether the minor emission peaks observed in Carmine Alum lambda emission scans were due to the presence of GFP or autofluorescence (Additional file 10: Figure S8A). ROI’s selected in the ductal epithelium (Additional file 10: Figure S8A, green curve) were compared with ROI’s selected in the ductal space which would be predominantly background (Additional file 10: Figure S8A, red curve). Both GFP and non-GFP glands revealed a major peak at 495 nm in the normalized plots with little difference between the normalized emission curves and no peak coinciding with GFP. The background peak at 495 nm coincides with the background peak previously observed in Carmine Alum lambda scans (Figure 10A, arrows). Background intensity (red curve) was substantially less than the signal from the ductal epithelium when plotted as absolute value rather than normalized intensity (not shown). Background peaks were observed at 735, 860 and 960 nm with 860 nm producing the best signal-to-noise ratio for tissue contrast (Additional file 10: Figure S8B). Since the results were essentially the same for both GFP and non-GFP mice, we conclude that the GFP signal at 506 nm is lost after fixation and processing for unstained whole mounts, and the peak at 495 nm represents background autofluorescence derived from the unstained, but otherwise identically processed gland.

Images suitable for morphological studies were obtained at Ex 860 and 890 nm and included SHG-B and SHG-F signals (Additional file 11: Figure S9A). SHG-B was detected at Ex 800 and 890 nm (Additional file 11: Figure S9B), but not at Ex 735 nm (not shown), and could be separated from the background autofluorescence by using the lambda scan to generate narrow bandwidth images (Additional file 11: Figure S9B, green arrows and SHG-B images). However, SHG-F was observed only at Ex 860 and 890 nm and not at 800 nm (Additional file 11: Figure S9C). The ChD at Ex 800 nm and below captured only the DIC image (Additional file 11: Figure S9C, Ex 800, unfiltered, ChD). Increased signal-to-noise detection of fibers using SHG-B signals could be optimized in unstained whole mounts by narrowing the emission bandwidth to Em 393–425 as illustrated in a lambda scan, extracted emission images, and linescan quantification of pixel intensities (Additional file 12: Figure S10).

Apart from retrospective studies of existing Carmine Alum prepared whole mounts, it is apparent from Z-imaging experiments that preparation of whole mounts, minus Carmine Alum staining, increases the depth of imaging by using autofluorescence of the gland avoiding Carmine Alum “shadowing” by TEBs. In unstained whole mounts, the mammary gland stromal details could be imaged to depths of ~340 μm beneath the unstained TEB, in this case without great loss of signal (Figure 12A-D). Significantly, SHG-B and SHG-F signals appear not to be affected to the same extent by the presence of a Carmine Alum stained TEB; that is, the unstained TEB appears not to shadow the SHG signal in the same way as observed for the Carmine Alum stained TEBs (Figure 12C-E). Figure 12E is a 3D cut showing SHG-B and SHG-F signal. The white arrows in Figure 12C and E illustrate the SHG-B and SHG-F signals below the TEB, deeper in the tissue. In areas lacking TEBs, muscle tissue could be imaged at z depths of up to ~ 200 μm and possibly deeper, although this was not investigated in detail (Additional file 13: Figure S11A-E). We conclude that dense Carmine Alum staining in TEBs itself is largely responsible for the “shadowing” effect below TEBs. The penetration depended in large part on the presence and specific relative location of Carmine Alum-dense ducts and TEBs interposed between light source and detector (Additional file 4: Figure S3).

Live imaging of the GFP mammary gland demonstrated an outer mammary gland layer of strong SHG-B signal characterized by increasingly parallel fiber orientation with Z-depth, and high resolution profiles of GFP-tagged epithelial cells. However, transmitted SHG-F signals were highly scattered resulting in an apparently “shadowed” pattern matching the outline of fat cells within the stroma. At face value, the data suggest a layer of fibers that is approximately 6 μm thick on the surface of the mammary gland, but nothing deeper within the tissue. However, unstained whole mount preparations or those stained with Carmine Alum afforded much more detailed structural information from deeper within the mammary gland. Whereas separate thin layers of oriented fibers could be detected in areas immediately adjacent to TEBs in living tissue near the surface, extensive layers of fibers were detected in whole mount mammary glands. Furthermore, SHG-F signals deeper within the tissue became apparent after whole mount tissue preparation with or without Carmine Alum staining. We conclude that layers of fibers delineate zones that isolate TEBs and ductal growth from surrounding stromal layers.

The fibrillar layers contain both reflected SHG-B and transmitted SHG-F signals. A similar mixture of these signals is found where the fibers run in parallel orientations to the vessels. The SHG signals associated with blood vessels are significantly stronger. Subdivisions of collagen layers within the stroma might divert the directionality of ductal extension by interacting with TEBs, as suggested by images such as Additional file 5: Figure S4, and Additional file 6: Movies S2-3. Dense fibrillar structure is associated with intense ductal bud formation in abnormal glands illustrated in Figure 7. These details were only obtained using a combination of SHG-B and SHG-F. Thus, a combination of reflected and transmitted SHG images together with visualization of ductal structures and TEBs stained with Carmine Alum reveal additional architecture in 3D within mammary glands not previously appreciated.

Previously, SHG signals visualizing fibrillar structures have been identified as primarily due to fibrillar collagen I by means of antibody, collagenase, and in vitro experiments (for example, [22, 30]). Thus, we conclude that the fibrillar structure identified using SHG in these studies are likely collagen I fibrils. Furthermore, using SHG-B, collagen I fibers have been shown associated with TEBs, and their abundance and rigidity demonstrated to be controlled by macrophage activity [6, 22, 31]. Using an imaging platform with both coherent anti-Stokes Raman scattering and second harmonic generation, Le et al. imaged mammary adipocytes, blood capillaries, collagen fibrils, and tumor cells simultaneously and without any labelling [32]. They observed that lipid droplets of adipocytes and collagen content in mammary tumor stroma were both increased in obese animals, measurements that could only have been made by imaging tissue in 3D, and not possible with 2D histology. Although autofluorescence of collagen has been used to measure collagen density in multiphoton experiments [33], the ability to image and quantify both the ductal architecture and density of collagen fibers in archived mammary gland whole mounts provides a simple but powerful tool for mammary gland biology. The evaluation of collagen fibril arrangement and density for breast cancer prognosis was recently reported [8]. The imaging was carried out on archived pathology tissue sections. To understand the basis for the association of collagen structure to prognosis observed in human tissues, studies of animal models of breast cancer are critical. In the present study, we have determined that archived Carmine Alum-stained whole mount tissues are candidates for multiphoton imaging to explore collagen fiber deposition in transgenic animal models of cancer. Molecular, cellular, and tissue imaging has and will continue to provide valuable information on the role of collagen deposition in normal and tumor development.

Zoumi et al. optimized SHG of 3D collagen raft cultures and found peak SHG-B at approximately 800 nm [21]. We found that although 860 nm may not be optimal for SHG, the signal was adequate to provide good images of collagen fibers, and at the same time useful for either GFP imaging (in live tissue) or Carmine Alum (in whole mounts). Imaging depth would likely be improved by imaging at 950 nm excitation with a SHG signal appearing at 475 nm. However, Zoumi et al. observed that the reflected SHG signal declines dramatically and is quite low even at 880 nm [21]. Therefore, use of the SHG-F signal might be preferable when imaging at longer wavelengths such as 960 nm is required. On the other hand, Theodossiou et al. (2006) reported for rat-tendon cryosections that SHG-F signal peaked at 880 nm with secondary maxima about 845, 895, and 915 nm, whereas SHG-B signal had four maxima at 845, 880, 895 and 915 nm [34]. Thus, it is possible that MP excitation at 880 nm would provide sensitivity for both SHG and EGFP [34, 35]. These authors also reported that as the fiber orientation angle with respect to the laser directionality increased from 0 to 90 degrees, SHG intensity also increased. The orientation of the mammary gland on the stage ensures that many of the fibers would be mostly perpendicular to the laser source as the terminal end buds and their associated parallel fibers are mostly perpendicular as well, and thus their intensity would be maximized. However, fibers that are oriented parallel to the laser source would be masked. Indeed, in orthogonal views, fibers are never seen in vertical arrays, but rather as punctate cross sections. To answer this question completely, the tissue must be rotated 90^o so that the fibers are reoriented parallel to the laser beam.

As previously described [23], ECM fibers in the mammary gland visualized using SHG-B appeared to be somewhat more diffuse and irregular in profile compared with the fibers seen using SHG-F (see Figure 9). Previously, the ratio of SHG-F/SHG-B signal was found to change when comparing immature to mature collagen fibers [23]. Examination of the punctate or irregular aspect of reflected SHG signal was therefore proposed as an indicator of the presence of immature fibers. According to this interpretation, in our study, the SHG-B signals identify collagen fibers of more variable fiber maturity compared with the SHG-F signals which would suggest that SHG-F-detected fibers within the gland might have a more crystalline or mature structure. The verity of this is hard to assess. Carmine Alum staining might interfere in some way with SHG detection in addition to reducing signal intensity via a “shadowing” effect. However, this is unlikely since fluorescence and SHG detection occurs by two different mechanisms, and Carmine Alum poorly stained the fibers. Furthermore, SHG-B and SHG-F signals were detected in similar arrangements in unstained, whole mount mammary glands. Another interesting possibility is that whole mount preparation involving fixation and dehydration might actually improve SHG intensity not only by defatting the gland, but also by stabilizing or unmasking the repeating structure of collagen fibrils, especially in the case of SHG-F that was poorly detected in vivo. Thus, further studies are required to examine the utility of the SHG-F compared with SHG-B signals to reveal differences in collagen fiber properties and for quantitative comparison of SHG-detected fibers.

Imaging morphology in Carmine Alum-stained mammary gland whole mounts is limited in that the dense array of Carmine Alum-stained epithelial cells in terminal end buds scatters both the fluorescence and SHG signals depending upon imaging depth and the relative positions of the epithelial or collagen fiber structures with respect to the detectors. Study of unstained whole mounts suggests that the “shadowing” effect of TEBs upon SHG and Carmine Alum signals deeper within the tissue likely arises from the density of Carmine Alum staining itself. In any case, the investigator must pay careful attention to the surrounding structures before interpreting the structural data. Another potential limiting factor to acquiring deep tissue images in whole mounts is the working distance of the imaging lens. TEBs deeper into the whole mounts are accessible with long working distance lenses in the case that there are no intervening epithelial structures between the TEB chosen for imaging and the coverslip surface.

For deepest imaging into the whole mount tissue, it is probably advisable to collect the Carmine Alum signal using the higher emission wavelengths above 623 nm to detect the epithelial cells and eliminate ECM/ stromal contribution from autofluorescence (see Figure 11). At higher excitation and emission wavelengths, tissue penetration would be enhanced in addition. Increasing laser intensity might be important to gain depth of SHG signal and certainly using a non-descanned detector for SHG-B would be optimal. However, if higher intensity laser illumination is used, Carmine Alum should be imaged first at low laser intensity to minimize photobleaching and photodamage. In cases where the researcher is interested to compare SHG-B and SHG-F, appropriately narrow bandpass filters (or ranges) for both reflected and transmitted signals should be employed to block out autofluorescent signal. In the system used for this study, the META detector and collection into ChS can be used to limit the SHG-B to a narrow range around the peak emission to improve the signal to noise of the SHG-B detected collagen fibers and the emission range for Carmine Alum selected to exclude background signal that occurs overlapping with wavelengths at least up to 623 nm (see Additional file 11: Figures S9B and Additional file 12: S10A-C).

Maximal collection of photons is critical to maximize the depth of imaging that is possible, and to do this a non-descanned detector should be used rather than the descanned, confocal detector. In this study, use of a non-descanned detector to collect the SHG-B and Carmine Alum signals was not possible. However, a non-descanned detector with greater sensitivity than the transmitted detector did allow imaging of SHG-F. The presence of high quality non-descanned detectors to maximize photon collection for both transmitted and reflected fluorescence and SHG emissions, would likely increase z-depth detail, even though the operator would also requires a more elaborate set up to minimize stray light that is present even in a darkened room.