Development of the mammary gland largely occurs postnatally. Under the influence of ovarian hormones during puberty, the rudimentary epithelial ducts elongate and branch to form a ductal tree that fills the fat pad in the adult virgin female. The invasive front of the duct is known as the terminal end bud (TEB), a highly proliferative multi-layered epithelial structure surrounded by supporting stroma. During puberty, the terminal end buds grow rapidly through the mammary gland fat pad, with ductal extension occurring at a rate of approximately 1 mm per day (Hinck and Silberstein,2005). Once puberty is completed, the terminal end buds disappear and are replaced by terminal end ducts, and the hormonal influence during the estrous cycle causes secondary branching and limited alveolar development.
Colony-stimulating factor-1 (CSF-1) is a primary regulator of mononuclear phagocyte proliferation, survival, and differentiation, as well as being chemotactic and chemokinetic to these cells (Stanley,1986). In non-pregnant mice, with the exception of oocytes, CSF-1 receptor expression is limited to cells of the mononuclear phagocyte lineage and their precursors (Pollard and Stanley,1996). Therefore, cells of these lineages mediate all effects of CSF-1. Mice carrying an inactivating mutation in the Csf1 gene, osteopetrotic (Csf1op), have a greatly reduced number of macrophages in most tissues and a severe depletion of osteoclasts (Wiktor-Jedrzejczak et al.,1990). This leads to osteopetrosis, diminished fertility, and developmental defects in a number of organs (Wiktor-Jedrzejczak et al.,1990; Cohen et al.,1999). In particular, mammary gland development during puberty and pregnancy is retarded (Pollard and Hennighausen,1994; Gouon-Evans et al.,2000). During puberty, these mice display reduced terminal end bud numbers, branching, and ductal elongation, suggesting CSF-1-responsive macrophages are necessary for adequate terminal end bud development (Gouon-Evans et al.,2000). The requirement for CSF-1 specifically in the mammary gland was shown by studies employing a CSF-1 transgenic mouse strain, where CSF-1 expression was restricted to the mammary gland epithelium, crossed onto the Csf1op/Csf1op strain, whereby the mice retained all their Csf1op/Csf1op phenotypes with the exception of reduced pubertal mammary gland development (Nguyen and Pollard,2002). Furthermore, hormone replacement studies showed that these effects were independent of the systemic hormonal environment and detailed analyses showed that only macrophages express the CSF-1R in the mammary gland (Gouon-Evans et al.,2000). Together these experiments showed a cell autonomous role for macrophages within the mammary gland.
Multi-photon microscopy is a relatively new technology that surpasses conventional confocal imaging in the ability to image fluorescent signals from biological tissues. Femtosecond pulsed low-energy photons are used, leading to greater depth of imaging through tissue, reduced toxicity, and photobleaching, and this technology can, therefore, be used to image living mammary tissue (Wang et al.,2002; Condeelis and Segall,2003). Another advantage of multi-photon microscopy over confocal microscopy is its ability to utilize a naturally occurring phenomenon known as second harmonic generation (SHG), which causes the resonant emission of polarized light from alpha triple helical matrices, including collagen type I, II, and III, acto-myosin, and tubulin when these structures are subjected to low-energy photons (Campagnola et al.,2001; Stoller et al.,2002a,b; Stoller et al.,2003; Mohler et al.,2003; Zipfel et al.,2003; Sahai et al.,2005).
In order to look more closely at the effect of macrophage deficiency on terminal end bud development during puberty, we employed multi-photon microscopy to image mammary gland whole mounts and construct three-dimensional images. Using the phenomenon of second harmonic generation, we were able to visualize long type I collagen fibers in the immediate vicinity of the terminal end bud. Together with macrophage-specific GFP expression, we have used multi-photon microscopy to explore relationships between extracellular matrix organization and structure of the terminal end bud. Our studies reveal an essential and previously undescribed role for macrophages in type I collagen fibrillogenesis and in the structural organization of this rapidly developing organ.
Multiphoton Microscopy Reveals Macrophages Are Associated With Fibrillar Structures in the Mammary Gland
Using the multi-photon microscope, we were able to look at 3-dimensional extracellular components in the mammary gland in greater detail than ever before (Fig. 1). Using fixed mammary glands from mice transgenic for the CSF-1R-eGFP macrophage reporter, we observed large primary and small secondary terminal end buds (Fig. 1A–H and I–L, respectively) and the surrounding SHG signal together with macrophages marked by their green fluorescence. Long fibers emitting light by SHG were located on the baso-lateral side of these terminal end buds (Fig. 1A,E,I). In large primary terminal end buds, these fibers tended to be close to the sides of the terminal end bud, and project upwards in the direction of growth (Fig. 1A,E), or in the case of newly formed smaller terminal end buds project outwards perpendicular to the terminal end bud (Fig. 1I). A large amount of SHG signal was also found around blood vessels (Fig. 1H, white arrow). Macrophage accumulation was visible around the sides of the terminal end buds (Fig. 1B,F,J), as was previously described in thin sections using immunohistological analysis (Gouon-Evans et al.,2000). By merging the images generated from the blue (SHG), green (macrophages), and red (propidium iodide) filters, we found that the macrophages tended to be associated with the long fibers that were detected by SHG (Fig 1D,H,L).
Type I Collagen Fibers Are the Source of Second Harmonic Generation
Given the distinct absorption and emission spectra of the SHG signal, and what is known about the composition of the extracellular matrix in the mammary gland, it is expected that the signal is generated principally by collagen I fibers (Zipfel et al.,2003). To test this hypothesis, we stained frozen sections of mammary gland with a collagen I antibody and looked for co-localization with SHG signal (Fig. 2). By merging images obtained by SHG (Fig. 2A), antibody binding (Fig. 2B), and propidium iodide staining (Fig. 2C), we observed a large extent of co-localization between collagen I antibody (red pseudo-color) and SHG (green pseudo-color) signal in the vicinity of terminal end buds (Fig 2D). When observed at higher magnification, the SHG signal and the collagen I antibody staining almost completely overlapped (Fig. 2E–H). The antibody detected more signal overall than SHG. This is expected to be due to the antibody recognizing both fibrillar and less fibrillar forms of collagen. A secondary antibody control revealed that SHG by the blue filter (Fig. 2I) was detected in the absence of collagen I antibody staining by the green filter (Fig. 2J).
It was also noted that some areas detected by SHG did not stain with the collagen I antibody. This may be due to more fibrillous collagen bundles excluding antibody binding or to different alpha helix–containing proteins in these bundles. To determine that these bundles and the remainder of the SHG were given by collagen I fibers, we treated sections with collagenase. Pre-treatment with mild collagenase digestion destroyed most of the SHG detected in the blue filter, including those in the large bundles that had previously resisted antibody-staining, indicating that they also contained collagen. In addition, anti-collagen antibody signal detected in the green filter was lost, confirming the enzyme had the appropriate specificity (Fig. 2M and N, respectively), although some antibody staining was still observed within the terminal end buds between the body cells (Fig. 2N and P). This residual antibody staining may be from undigested by collagen subunits that still retain antibody specificity. Taken together, these studies confirm that the SHG signal detected in the mammary gland is caused by fibrillar type I collagen.
Macrophages Direct the Shape of Terminal End Buds
The depth of imaging that the multi-photon microscope permits enabled us to construct 3-dimensional images of propidium iodide–stained terminal end buds by stacking 2-dimensional images together. Using this approach, we analyzed the effect of the CSF-1 null mutation on terminal end bud structure (Fig. 3). Primary terminal end buds from +/Csf1op macrophage replete females were oblong in shape (Fig. 3A). In contrast, primary terminal end buds from Csf1op/Csf1op mice appeared shorter and rounder (Fig. 3B). This change in shape was due to macrophages acting directly on the mammary gland, as Csf1op/Csf1op females expressing a CSF-1 transgene that restores the macrophage population around the terminal end bud specifically in the mammary gland also had the characteristic oblong terminal end bud shape (Fig. 3C). Quantification of these parameters using Image J software analysis revealed that, indeed, terminal end buds from Csf1op/Csf1op were shorter in length than wildtype, while no differences were observed in width or height (Fig. 3D). Terminal end buds from Csf1op/Csf1op mice were also more circular (circularity index of 1 is a perfect circle) (Fig. 3E). No significant difference in epithelial area was detected between +/Csf1op and Csf1op/Csf1op mice although in the transgenic CSF-1 add back there was a modest increase in the terminal end bud epithelial area (Fig. 3F).
Macrophages Promote Collagen Fibrillogenesis
The effect of CSF-1 null mutation on collagen I fibers surrounding the terminal end bud was investigated in fixed mammary gland whole mounts from +/Csf1op and Csf1op/Csf1op (Fig. 4A,B, respectively). By stacking 2-dimensional images together to create a 3-dimensional view of fibrillar collagen surrounding the terminal end bud, the relative abundance of collagen was quantified in the 255-bit image using Image J software. A 40% reduction in collagen I detected by second harmonic generation was observed in Csf1op/Csf1op mice compared to their heterozygous controls (Fig. 4C).
To establish whether there was any association between the amount of collagen fibers and the shape of terminal end buds, results from +/Csf1op and Csf1op/Csf1op were pooled for linear regression analysis. There was, indeed, a significant (P = 0.036) albeit weak (r2 = 0.37) inverse relationship between circularity of the terminal end bud and collagen fiber abundance (measured by SHG) around that terminal end bud.
To determine whether the reduction in fibrillar collagen detected in Csf1op/Csf1op mice was due to a reduction in collagen synthesis or in fibrillar organization, we took advantage of the different forms of collagen detected by SHG versus collagen I antibody staining. We previously observed that only fibrillar collagen was detected by SHG, while the antibody recognized both fibrillar and less fibrillous forms of collagen I. When collagen I surrounding terminal end buds was analyzed in frozen sections using SHG together with immunofluorescent staining, there was no difference in the amount of collagen I detected by antibody staining in Csf1op/Csf1op females compared to +/Csf1op (Fig. 5A). However, a reduction in collagen I detected by SHG signal was observed (Fig. 5B), in agreement with our previous results on fixed mammary gland whole mounts (Fig. 4C). These results suggest that the role of macrophages in collagen synthesis is in the organization of collagen into fibrillar bundles, rather than in the production or secretion of monomeric collagen.
Motility of Macrophages in the Pubertal Mammary Gland
The interaction between macrophages and collagen fibers in the developing mammary gland were observed by intravital imaging using the multi-photon microscope. An area in the expected vicinity of terminal end buds of 5-week-old CSF-1R-eGFP mice, which contained second harmonic signal similar to those observed around terminal end buds, was observed for a period of 30 min, with z-series images taken every 2 min (see Supplemental Movie 1, which can be viewed at www.interscience.wiley.com/jpages/1058-8388/suppmat). Green macrophages were found to be motile in the developing mammary gland, moving at a speed of approximately 5 μm/min, and were observed crawling on the collagen matrix, and crossing between the fibers.
We have previously reported that depletion of macrophages by genetic removal of CSF-1 resulted in pubertal ductal development that was delayed in its initiation, and that progressed more slowly with fewer primary and secondary branches (Gouon-Evans et al.,2000). Furthermore, there was a significant disruption of patterning as the ductal front advanced through the fat pad (Gouon-Evans et al.,2000). These defects could be rescued by specifically expressing CSF-1 in the mammary epithelium thereby recruiting macrophages to the developing terminal end bud (Nguyen and Pollard,2002). These data suggest that in addition to intrinsic growth of the terminal end bud and influence of the surrounding inductive fat, macrophages enhance the rate of ductal outgrowth and are involved in its patterning. But how these hematopoietic cells might influence ductal growth is unclear. They do not appear to affect mammary gland epithelial or stromal cell proliferation, or serum estradiol-17β levels during puberty (Gouon-Evans et al.,2000). A new approach was, therefore, necessary to address this question.
The use of multi-photon microscopy is a powerful tool to analyze 3-dimensional structures in biology. This method provides quantitative and qualitative results not easily produced by conventional histological methods and confocal microscopy, and can be utilized to investigate relationships between structure and function in a variety of mutants and transgenic mouse models, both intravitally and ex vivo. We used this method to explore the action of macrophages on terminal end bud formation during mammary development. By analyzing the 3-dimensional shape of terminal end buds and the surrounding collagen I matrix, we conclude that reduced macrophage accumulation causes reduced fibrillar collagen organization and a rounder, less elongated terminal end bud shape. We hypothesize that macrophages organize type I collagen supramolecular structure around the terminal end bud, and together with other extracellular matrix components, direct the shape of the terminal end bud.
A benefit of multi-photon microscopy that we have utilized in our current study is the phenomenon of SHG. We observed a large amount of SHG signal in the matrix surrounding the terminal end buds in the developing mammary gland. We have shown the SHG signal to be caused principally by collagen I fibers in the mammary gland. Long fibers of this collagen were observed baso-lateral to terminal end buds. We also observed long fibers of collagen running the length of the mammary gland and, interestingly, these fibers often linked up to other terminal end buds. We found that macrophages tended to be associated with the collagen fibers.
Early studies by Daniel et al. (1984) using ex-vivo culture and fragment implants in collagen 1 gels showed that collagen was required for the regenerating mammary gland and perhaps for normal development. In these experiments, we show that in the absence of macrophages, much less fibrillar collagen surrounds the terminal end bud in the stroma. Using the genetic add back of CSF-1 specifically in the mammary epithelium that recruits macrophages back to the mammary gland (Nguyen and Pollard,2002), we were able to restore terminal end bud shape. Since macrophages are the only CSF-1R-expressing cells (Gouon-Evans et al.,2000) and there was no restoration of other mutant phenotypes (Nguyen and Pollard,2002), these data demonstrate that the rescued phenotypes were a direct effect of macrophages on mammary development.
Macrophages do not promote production of collagen units, which are approximately 300 nm long (Kadler,2004) and were detected in abundance by immunohistochemical staining of Csf1op/Csf1op mammary glands. Rather, macrophages are involved in its assembly into long >100 μm organized fibers. How macrophages might achieve this is not known. We and others have noted that macrophages have a strong affinity for collagen fibers and increased motility when they are present (Boswell and Swan,1984; Condeelis and Segall,2003). This has been shown previously by studies in tumors and now in the context of normal development. We hypothesize that macrophages crawling along the fibers, circling and crossing from one fiber to a neighboring fiber, may serve to enhance end-to-end fusion of collagen subunits. Alternatively, macrophage-secreted factor(s) may stimulate the production of lysyl oxidase by fibroblasts (Koslowski et al.,2003), which catalyzes cross-link formation in collagen fibrillogenesis (Eyre et al.,1984).
In conclusion, we have shown that macrophages in the mammary gland affect the shape of the terminal end bud and its surrounding collagen I matrix. We propose that macrophages play a critical role in an interplay between other stromal cells such as fibroblasts, whereby collagen is synthesized and secreted, and organized into bundles in a pattern that provides a framework for the terminal end bud to elongate. Presumably as the terminal end bud continues to develop, collagen fibers are broken down (perhaps by secretion of matrix metalloproteinases) and then remodeled, so that terminal end bud structure and surrounding matrix may be maintained as the ducts grow forward through the mammary gland fat pad.
Whole Mount Preparation
Inguinal mammary glands were dissected from pubertal (5 weeks) Csf1op/Csf1op and +/Csf1op littermates (morphologically and physiologically similar to homozygous wild type mice; Pollard and Stanley,1996) and stretched across a glass slide. In some experiments, mammary glands from Csf1op/Csf1op mice carrying a tetracycline regulatable CSF-1 transgene targeted to the mammary gland with a tet-operated MMTV transgene (Nguyen and Pollard,2002) were used. The slides were immersed in 10% formalin overnight at 4°C, defatted for 5 h in toluene, and stained with propidium iodide (Molecular Probes, Eugene, OR). The whole mounts were dehydrated in ethanol (70, 90, and 100% for 15 min each) and further defatted overnight in toluene. The whole mounts were stored in methyl salicylate before imaging. Whole mount mammary glands from mice carrying a GFP transgene driven by the CSF-1 receptor promoter (CSF-1R-eGFP) (Sasmono et al.,2003) were fixed in 1.5% paraformaldehyde with 20% sucrose overnight and defatted and stained as above.
Dissected mammary glands were frozen in OCT on dry ice, and stored at −80°C. Sections (40 μm) were cut on a cryostat and dried onto positively charged glass slides. Some sections were incubated with collagenase type III (Sigma, Eugene, MO) (50 U/ml) for 30 min at 37°C, and then rinsed 3 times for 5 min each in PBS containing calcium and magnesium. Sections were blocked with 10% donkey serum for 10 min at room temperature, excess fluid was removed from the slide, then incubated with 1:40 dilution of rabbit anti-mouse collagen I antibody (Chemicon, Temecula, CA) for 1 hr. Following 3×5-min PBS washes, the slides were incubated in 1:200 dilution of alexa fluor-488 donkey anti-rabbit secondary antibody (Molecular Probes) with propidium iodide for 1 hr. After washing as before, the sections were mounted with fluorescent mounting medium (Vector Laboratories, Burlingame, CA) under glass coverslips. Sections were stored in the dark at 4°C until imaging.
Multi-Photon Imaging of Ex Vivo and In Vivo Mammary Glands
Whole mount mammary glands were imaged using a Biorad Radiance 2000 multi-photon microscope with an inverted Olympus IX70 connected to a Spectra Physics Tsunami Ti-Sapphire laser as detailed elsewhere (Wyckoff et al.,2005). Briefly, all images were collected using non-de-scanned detectors, and using a 20× Plan Apo 0.7NA (air) objective, covering a field of 542 × 542 μm in the XY axis, or a 60× LUMPlan/IR 0.9NA (water) objective, covering a field of 181 × 181 μm. The filters used on ex vivo mammary glands were 450/80 nm (blue), 515/30 nm (green), and 590/LP nm (red) (Chroma, VT). For live imaging, simultaneous dual filters of 450/30 nm (blue) and 515/30 nm (green) were used. All images were collected at a laser wavelength of 850 nm, taking z-series images at 2-μm steps. The blue filter detected fluorescence generated by second harmonic generation (SHG), the green filter was used for GFP and alexa-488 conjugated secondary antibody, and the red filter detected fluorescence emitted from propidium iodide–stained cells.
CSF-1R-eGFP female mice at 5 weeks of age were anaesthetized under 2.5% isoflurane (Abbott Laboratories, Abbott Park, IL). A midline incision was made, and the 4th mammary gland was dissected away from the underlying peritoneal wall. The mouse was then placed on the heated (30°C) microscope stage with the exposed mammary gland fat pad resting on a coverslip over the objective. The location of terminal end buds was approximated using the mammary gland lymph node as a reference point and a region of high macrophage density was found under epifluorescence. This area was imaged using multi-photon covering a z-series of 8 μm in 2μm steps for 30 min, with a new z-series being taken at 2-min intervals. The mouse remained anaesthetized for the duration of the experiment under 2% isoflurane, and was sacrificed immediately after the experiment was finished.
As a large amount of variation exists between TEBs within the same mammary gland, it was important to select primary TEBs for analysis in an unbiased manner. To be included in the analysis, the entire TEB of the central invading front had to be scannable under multi-photon imaging, and not have a branch point within 500 μm of the base of the TEB. The 5 TEBs meeting this criteria most distal to the nipple were selected for analysis. However, in some Csf1op/Csf1op mice, 5 TEBs filling this criteria were not available, as Csf1op/Csf1op mice tend to have fewer TEBs (Gouon-Evans et al.,2000); in this case at least 2 TEBs were analyzed. The images were compiled and analyzed using Image J software (http://rsb.info.nih.gov/ij/) as described previously (Wyckoff et al.,2005). Terminal end bud length, width, height, and area were calculated. Examples of the length and width are illustrated in Figure 3. Height was calculated by projecting the TEB in three dimensions and measuring the largest distance between the top and bottom of the TEB. Circularity of the structure was quantified, whereby a value of 1 is a perfect circle. Abundance of SHG and collagen I antibody staining was quantified by highlighting the area 150 μm immediately around (not within) the terminal end bud, and measuring the amount of white in each pixel of the 255-bit image. The data from each TEB was then averaged to give one data point per mouse (n = 5 mice in each group) and analyzed using a Student's t-test with SPSS software, statistical significance being inferred with P < 0.05.
The authors thank the staff of Core A of PO1 CA100324 for expert technical assistance in these studies and to Dr. Elaine Lin and Jim Lee for their help in obtaining the CSF-1 transgenic mouse samples. J.W. Pollard is the Sheldon and Betty E. Feinberg Senior Faculty Scholar in Cancer Research. W.V.I. is an Australian NHMRC CJ Martin Fellow, no. 250473.