1. Introduction

The tumor stroma consists of the non-malignant cells in a tumor, the basement membrane, and the extracellular matrix (ECM), and is known to play a significant role in tumor growth [1]. Phenotypic alterations in stromal cells surrounding the malignant cancer cells, especially fibroblasts, contribute to tumor progression [2]. These in turn modify the production and degradation of components of the ECM, resulting in a “reactive stroma” that is a hallmark of the neoplastic transformation [1, 2] (See Fig. 1). This occurs primarily through the increased synthesis and deposition of ECM molecules such as type-I collagen, as well as the degradation of these molecules via elevated synthesis of ECM-degrading proteases (e.g. the matrix metalloproteinases, or MMPs) [1]. This highly altered ECM in turn is believed to play an active role in tumor progression and is known to affect tumor cell migration, to modify the availability of growth factors, and to directly signal tumor cells through integrins [1–3]. As a result the character of the tumor ECM is a highly useful prognostic factor: In pathology practice, breast carcinomas are graded based on the structure and density of ECM staining, while the presence of fibrotic foci in invasive ductal carcinoma of the breast indicates poor prognosis [4]. Furthermore, elevated serum markers of collagen synthesis [5, 6] and degradation [7] are indicators of breast cancer aggressiveness and poor patient survival.

 

Fig. 1. Enhanced stromal deposition characteristic of many tumors. Masson’s Trichrome staining of TG1-1 tumor cells grown in the mammary fat pad of FVB mice (left) as well as the healthy mammary fat pad of FVB mice (right). Abundant bands of ECM, primarily collagen, are evident throughout the tumor tissue as a blue staining (left), and are largely confined to isolated ducts in the healthy mammary fat pad (blue ring in center of right image). Images are 600 µm across.

Download Full Size | PDF

The central role that the reactive stroma plays in tumor growth, progression, and metastasis means that there is a significant interest in understanding the reactive stroma itself, as well as the relationship between normal stroma and reactive tumor stroma [1]. Second harmonic generation (SHG) has proven to be a useful window into the amount and organization of fibrillar collagen in biological tissues due to its relative specificity and the fact that it is an intrinsic signal. Recently SHG has begun to offer important insight into the reactive stroma of the living tumor: Experimentally-induced and naturally occurring carcinomas have been detected as an alteration in overall SHG intensity [8–11], and that intensity has been shown to scale with overall collagen content [12]. SHG imaging of collagen fibrils has provided insight into the metastatic pattern of malignant cells [11, 13], and revealed the dynamic evolution of collagen content and fibril morphology during tumor growth and treatment [12].

These reports utilized SHG intensity information to reveal the overall morphology of tumor collagen fibrils (location, orientation, and length), and to provide a diagnostic ability to detect malignancy (via scoring of morphology or of average intensity). However, the coherent nature of SHG means that it provides a great deal more information about the structure of the collagen fibrils than is contained simply in its intensity, and hence may reveal important information about the underlying biology of the reactive stroma.

The relationship between incoming laser polarization, outgoing SHG polarization, and macromolecular orientation, reveals the orientation angle of the individual SHG scatterers with respect to the global macromolecule axis [14]. Consequently SHG polarization can be used to provide details of the angular orientation of individual collagen triple helices as they are assembled into collagen fibrils, with single fibril resolution. This is interesting because the reactive stroma of the tumor consists of abnormal proportions of matrix molecules such as collagen I, III, fibrin, and fibronectin [1, 15, 16], and alterations in these ratios, such as that of collagen I to collagen III, are known to alter the angular distribution of triple helices in the overall collagen fibril [17]. These differences in the molecular constituents of healthy versus tumor stromas predict a possible difference in the angular orientation in the individual triple helices.

The relationship between the forward-scattered (F) and backward-scattered (B) SHG signal provides information about the length scale over which the scatterers are appropriately ordered (i.e. SHG-producing) along the laser axis [18, 19]. Therefore this F/B ratio can be used to estimate the overall spatial distribution of appropriately ordered triple helices within collagen fibrils. This is interesting because the reactive tumor stroma is typified by an altered collagen degradation machinery, with elevations of various collagen degrading enzymes such as the MMPs [1, 20]. These upregulated degradation mechanisms predict a possible difference in the spatial distribution of ordered scatterers in individual tumor collagen fibrils.

In this paper we will specifically utilize the polarization properties of SHG as well as the scattering directionality to explore the relationship between reactive and healthy stroma. We will investigate these optical properties in collagen in two murine breast tumor models, both grown in the mammary fat pad of the mouse, and compare them to collagen from the healthy mammary fat pad in the same mouse strain. The models we will study are the TG1-1 murine mammary tumor cell line which arose spontaneously in a FVB/N neu-transgenic mouse [21], as well as the 4T1 murine mammary adenocarcinoma cell line, which arose spontaneously in the BALB strain [22]. Both tumor models exhibit enhanced collagen deposition characteristic of a tumor reactive stroma (see Fig. 2).

 

Fig. 2. Anti-Collagen I antibody staining of TG1-1 (left) and 4T1 (right) tumor sections, with DAB contrast. Both tumor types show the enhanced ECM deposition characteristic of tumor reactive stroma, evidenced by enhanced dark brown contrast around “islands” of lightly stained tumor cells. Left image is 600 µm across, right is 1.2 mm.

Download Full Size | PDF

2. Experimental methods

2.1 Apparatus

The general experimental apparatus is shown in Fig. 3. SHG signal was generated by a Spectra Physics MaiTai Ti:Sapphire laser providing 100 fs pulses at 80 MHz and 810 nm. Beam scanning and image acquisition were performed with an Olympus Fluoview FV300 scanning system interfaced with an Olympus BX61WI upright microscope. The focusing objectives are two identical Olympus UMPLFL20XW water immersion lenses (20×, 0.5 N.A.). Objective 1 was used to focus the excitation beam on the sample and at the same time collect the backscattered SHG signal. The backscattered SHG signal was then separated from the excitation beam by a dichroic mirror (Chroma 670 DCSX) and a band pass filter centered at 405 nm (Chroma HQ405/30m-2P) placed after the objective back aperture, and detected by a photomultiplier tube (HC125-02, Hamamatsu). The forward propagating SHG signal was collected by the second objective, reflected by a silver mirror and passed through two filters before being detected by the second PMT. Filter 2 is a short pass filter (Chroma E700SP-2P) that is used to block the 810 nm excitation beam. Filter 3 is a band pass filter centered at 405 nm (Chroma HQ405/30m-2P).

 

Fig. 3. Experimental Apparatus.

Download Full Size | PDF

2.2a Sample preparation: cell culture

TG1-1 murine mammary tumor cells (courtesy of E. L. Steele Laboratory, Cambridge, MA) and 4T1 murine mammary adenocarcinoma (American Type Culture Collection, Manassas, VA) were grown in T-75 tissue culture flasks (Corning, Corning, NY) containing DMEM supplemented with 4.0 g/L glucose, 10% fetal calf serum (FCS), and 1% penicillin/streptavidin (Invitrogen, Carlsbad, CA). Cells were grown to 90% confluence and harvested with 0.25% trypsin/EDTA (Invitrogen). The trypsin reaction was stopped by adding DMEM containing 10% FCS. Cells were washed three times in sterile Dulbecco’s phosphate buffered saline (DPBS: Invitrogen) by centrifuging at 200xg for 10 minutes at 4°C. After the final wash, cells were resuspended in sterile DPBS and kept on ice until injection.

2.2b Sample preparation: animal husbandry

All animal techniques were developed and practiced in accordance with the policies of the University Committee on Animal Resources. Female FVB/NJ or BALB/c mice (Jackson Labs, Bar Harbor, ME) at 8 weeks old were anesthetized intraperitoneally with 90 mg ketamine hydrochloride and 9 mg xylazine (Hospira, Lake Forest, IL) per kg body weight. Hair was removed from the ventral surface by use of a depilatory agent. Concentrated tumor cell suspension was injected into the inguinal mammary glands of each animal (a total of 4 injections per animal) at 0.05 mL per injection site. Tumors were allowed to grow to 0.5 cm before being resected from the animal. To ensure maximal variance across animals, only one tumor was taken from each animal for imaging purposes. Mammary fat pads (MFPs) were resected from animals which did not undergo tumor implantation. Female Wistar rats at 1 year old were anesthetized intraperitoneally with 90 mg ketamine hydrochloride and 9 mg xylazine (Hospira, Lake Forest, IL) per kg body weight and sacrificed. Tails were cut into ~1 cm discs with a razor blade and collagenous tendon was teased out onto a coverslip with tweezers.

2.2c. Sample preparation: tissue sectioning for SHG imaging

Excised tissues for SHG imaging were sliced into ~100 um sections on a freezing microtome, then mounted in saline between two 150 um coverslips (VWR International, West Chester, PA).

2.2 d. Sample preparation: staining.

For Masson’s Trichrome staining, samples were excised from the animal and fixed in 10% formalin, then embedded in paraffin and cut into 5 µm sections on a vibrating microtome. The resultant sections were mounted on glass slides and stained with Masson’s Trichrome. For anti-collagen I antibody staining, tissue was removed, fixed in buffered (pH 7.4) 4% paraformaldehyde, cryoprotected in 10 and 30% sucrose, and cut at 30 µm on a cryostat. Free floating sections were incubated in a 1:200 dilution of a rabbit polyclonal antibody against Collagen 1 protein isolated from rabbit (AbCam 21286, Cambridge, MA). An anti-rabbit Vector peroxidase kit (Vecotr Laboratories, Burlingame, CA) was used to complete the reaction, and the reaction product was intensified with nickel sulfate.

2.2 e. Sample preparation: Electron Microscopy.

Mouse normal mammary gland and tumor specimens were fixed overnight at 4oC in 0.1M phosphate buffered 2.5% glutaraldehyde, postfixed in phosphate buffered 1.0% osmium peroxide for 60 minutes, and dehydrated in a graded series of ethanol to 100 %. The samples were then transitioned to propylene oxide, infiltrated with EPON/Araldite resin, embedded in fresh resin and polymerized for two days at 70oC. One micron sections were cut from epoxy tissue blocks onto glass slides and stained with a 1.0% Toluidine Blue stain to find the appropriate areas to examine ultrastructurally. Thin sections were cut with a diamond knife at 70nm and placed onto 200 mesh copper grids which were sequentially stained with 2.0% aqueous uranyl acetate and lead citrate. The grids were examined using a Hitachi 7100 transmission electron microscope and digital images were captured using a Megaview III digital camera with “AnalySIS” (Soft Imaging Systems, Lakewood, California) software.

2.3a Polarization analysis: equipment

For polarization analysis, excitation polarization was fixed parallel to the dichroic mirror surface to avoid elliptical polarizing effects of the primary beamsplitter. This was necessitated by the observation that 810 nm light of>500:1 polarization was degraded to as low as 45:1 upon reflection from the primary dichroic at a polarization angle of 45 degrees, while it maintained a polarization of 250:1 if the polarization angle was fixed parallel to the dichroic surface. The forward-scattered SHG signal was collected with Objective 2 (see Fig. 3), and a polarization analyzer consisting of a cube beam splitter (GL10-A, Newport, Irvine, CA) on a rotary stage was placed directly in front of the PMT.

2.3b Polarization analysis: theory

An incoming excitation beam with electric field vector E⃗ will induce polarization P⃗ in the fibril given by:

In the case of second harmonic generation from a system with cylindrical symmetry (such as the collagen fibril), the second term is relevant, and simplifies to [23]:

where ŝ is the direction of collagen orientation and the coefficients a, b, and c are related to elements of the second-order nonlinear susceptibility tensor χ⁽²⁾. If the resultant SHG signal is detected through a polarizer oriented at a direction ê), the detected SHG intensity is I_e∝)(P⃑)·ê)². This is given by

Where α is the angle between the collagen fibril and the polarizer axis and φ is the angle between the fibril and excitation beam polarization. Note that this assumes that the fibril is perpendicular to the optical axis. If we consider a z-directed excitation beam interacting with a y-aligned fibril then when α=0:

and when $\alpha =\frac{\pi }{2}$ :

Therefore:

If the collagen molecule is considered to be a cylindrically symmetric collection of single-axis scatterers (i.e. elements of the individual triple helices that assemble into a collagen fibril) with a constant polar angle θ and a random azimuthal angle ϕ, there are only two independent elements of χ⁽²⁾, the nonlinear susceptibility tensor, with the result that [14]:

Where β is the hyperpolarizability of the individual single-axis scatterers, of density N. This simplification assumes that the susceptibility tensor has Kleinman symmetry regardless of whether the excitation energy is significantly off-resonance [14]. In this case, Eq. (6) becomes:

And a measurement of I_y and I_x at a known φ allow the deduction of θ, the polar angle of the SHG scatterers within a given fibril. Note that in Williams et al. [18],

Where I_p is the total intensity measured with a perpendicularly polarized illumination beam and the fitting parameter ρ is used as a measure of the axial polarizing effects of the fibril. Hence:

And we see that the ρ term of Williams et al and others [18, 23, 24] is also equivalent to a measurement of θ, the polar angle of the SHG scatterers in a given fibril.

We measured θ in 100 um thick acute tissue slices mounted in saline between two coverslips. Each sample was imaged 36 times on an MPLSM with an excitation beam of fixed polarization angle, with each of the 36 images made through an analyzer rotated in 10 degree increments. With that data set, individual fibrils of arbitrary orientation can then be selected, and their intensity versus α can be plotted as in Fig. 4 by drawing a narrow region of interest around the fibril. The chosen fibril’s angle φ with respect to the excitation polarization direction can be manually measured from one image using ImageJ (Freeware website), and I_x and I_y can then be extracted from the radar plot. Note that absence of photodamage can (and should) be verified by ensuring that I(0)=I(360), as shown in Fig. 4.

 

Fig. 4. Radar plot of detected intensity versus analyzer angle. This represents the intensity versus analyzer angle for each of five selected fibrils. In combination with the measured angle of the fibril relative to the laser polarization (vertical in the above graph), I_x, I_y, and hence θ can be extracted from this data for each fibril.

Download Full Size | PDF

2.4 Forwards/Backwards Scattering Analysis

Using the apparatus described in Fig. 3, we measured the detected forward-directed and backward-directed SHG signal in 100 um thick acute tissue slices mounted in saline between two coverslips. In order to derive the true F/B ratio from these measurements, the effects of scattering as SHG light transits the tissue as well as the relative efficiencies of the two detection pathways must be accounted for. In the manner of Williams et al [18], we applied a dilute solution of 10 µm diameter blue fluorescent polystyrene beads (10m365415, Invitrogen) to the surface of our sample, and F/B ratio images were normalized such that the F/B ratio of in-focus beads was set equal to the ratio measured on the same day in beads in saline between two coverslips [18]. During each imaging session, these images were further corrected for relative detector inefficiencies by replacing the matched 405/30 SHG filters with matched 535/30 filters, and the F/B ratio of free FITC was measured and normalized to one.

To quantify F/B images, the background signal was first subtracted from the individual forward and backward images, then the F/B ratio of all pixels was calculated. In regions of the original F or B images where there was no significant collagen SHG the F/B ratio fluctuates due to small variations in background noise around a pixel count of zero, and lacks physical meaning. Consequently, an intensity threshold was chosen based upon the B image, producing a binary image mask which set the background pixels to zero and the foreground (e.g. collagen) pixels to one. This mask was multiplied by the F/B image, setting the varying background pixels to zero. Image J was then used to calculate the average pixel count of the resultant masked F/B image, and this was divided by the average pixel count of the binary mask image, producing the average F/B ratio of all pixels within fibrils (i.e. all pixels above threshold in the original B image).

To quantify F/B ratio versus apparent fibril diameter, a blinded observer analyzed 24 fibrils in four tumor samples and 24 fibrils in four mammary fat pad samples by drawing a line transverse to the fibril in ImageJ and fitting the intensity profile along that line to a Gaussian. The e-2 diameter of the fit is then reported as the apparent fibril diameter. A small ROI was also drawn within the imaged fibril at the same location to quantify the average F/B value. The spatial resolution of our system (e_-2 radius of the PSF, 0.5 NA lens overfilled with 810 nm excitation light) is 670 nm, so this analysis was limited to those fibrils with an apparent e^-2 diameter of 1.0 um or greater. The PSF will certainly affect the measurement of the diameter of these fibrils over these length scales, but without a priori knowledge of the “true” radial spatial distribution of SHG scatterers in the fibril (i.e. solid rod versus hollow cylinder, etc.), we chose not to perform any deconvolution calculation and present the raw data with the independent variable labeled as “apparent” fibril diameter. The resultant scientific conclusion, that there is no statistically significant relationship between F/B ratio and apparent fibril diameter, will be reinforced by any deconvolution calculation as this will serve to stretch the independent variable in Fig. 7 (see below) and render the F/B distribution even flatter.

2.5 Fibril orientation.

In a complex three-dimensional network such as the tumor extracellular matrix, the orientation of the individual fibrils relative to the laser axis can vary considerably. If the laser is propagating along the z axis, the fibril orientation relative to the x and y axes can be easily determined by image inspection, but orientation relative to the z axis (i.e. how closely the fibril lies in the plane of the image) will also affect the results of these measurements and must be evaluated. In our F/B measurements, the thresholding step described in section 2.4 above serves to eliminate pixels from fibrils that are oriented at a significant angle out of the plane of the image (i.e. less perpendicular to the z axis), as they are less intense in the SHG images that produce the mask. Manual analysis of the visible fibrils remaining after the masking step reveals that the average length of fibrils analyzed for F/B ratio was 102±21 µm and 153±25 µm in the TG1-1 and FVB MFP samples respectively (91 and 96 fibrils in N=4 animals), and 116±26 µm and 170±14 µm in the 4T1 and BALB MFP samples respectively (110 and 102 fibrils in N=5 animals). If we model individual fibrils as straight segments diving through the image plane, with an image plane thickness of ~12 microns (e^-2 z diameter of the PSF), then a visible fibril length of 100 microns corresponds to an angular deviation relative to the plane of tan(α)=12/100, and α~7 degrees. This suggests that the fibrils analyzed were not oriented at significant angles out of the plane of the image, and can be considered perpendicular to the optical axis. The SHG polarization analysis was performed with manually selected fibrils, and for this reason we chose fibrils for analysis with a minimum visible length of 100 µm.

2.5 Statistical analysis

Unless stated otherwise, all judgments of statistical significance are based upon the Student’s T Test with a threshold for significance of p<0.05.

3. Results

3.1 Comparison of polarization properties of collagen fibrils in tumor and normal mammary fat pad tissue

The relationship between incoming laser polarization, outgoing SHG polarization, and macromolecular orientation, reveals the orientation angle of the individual SHG scatterers with respect to the global macromolecule axis [14]. Using the polarization analysis method described above we found that θ, the angle of the elemental SHG scatterers relative to the fibril axis, is 51.43±0.91 degrees (N=4 tumors with six measurements of θ per tumor) in TG1-1 tumor collagen and is 50.61±1.3 degrees (N=4 fat pads with six measurements of θ per pad) in healthy FVB MFP. These two populations are not statistically significantly different (p>0.05). Likewise θ is 51.78±4.0 degrees (N=5 tumors with 11 measurements of θ per tumor) in 4T1 tumor collagen and is 51.71±4.5 degrees (N=5 fat pads with 11 measurements of θ per pad) in healthy BALB MFP, also not statistically significantly different (p>0.05), and neither is statistically significantly different from TG1-1 nor FVB MFP. This is surprising because the tumor reactive stroma is characterized by alterations in synthesis of collagen I, collagen III, as well as other ECM constituents [1, 15, 16], and the ratio of different collagen components (e.g. collagen I/III ratios [17]) are known to modify the angular assembly of triple helices into the overall collagen fibril. However, it has already been shown that SHG is generated by an appropriately ordered fibrillar subpopulation of the overall collagen in tumor stroma [12]: the polarization data therefore suggests that the altered collagen synthesis properties typical of the reactive tumor stroma do not influence the angular assembly of the ordered (i.e. SHG-producing) subpopulation of fibrillar collagen.

Our measurements of θ for tumor and MFP collagen in mice are in close agreement with lower resolution measurements of ρ in rat tail collagen [23, 24] (ρ~1.2-2.0, i.e. θ~49°, converted via Eq. (10)) and approximately match the 45.3° pitch angle of the collagen glycine proline helix [14]. Surprisingly, however, they appear to disagree with rat tail collagen measurements of ρ=2.6±0.2 (i.e. θ=41.3±1.1°, converted via Eq. (10)) that had the same single-fibril resolution as our measurements [18]. Therefore we also measured θ in rat tail collagen and FVB mouse tail collagen, determining that θ=42.4±1.8° (12 measurements in N=2 animals) in rat tail collagen and θ=51.8±3.5° (18 measurements in N=5 animals) in FVB mouse tail collagen. To sample an entirely different organ system we also determined that θ=53.7±5.0° (25 measurements in N=4 animals) in acute sections of FVB mouse colonic submucosa, an organ chosen for its highly elastic collagen structure. See Fig. 5 for a summary of the results. All measurements in mouse organs are statistically significantly different from our measurements in rat (p<0.05), and are not statistically significantly different from each other (p>0.05).

 

Fig. 5. Measured values of θ in rat tail collagen as well as in different organ systems in the mouse. Error bars are standard deviations. All measurements in mouse organs are statistically significantly different from our measurements in rat (P<0.05), and are not statistically significantly different from each other (p>0.05).

Download Full Size | PDF

Taken together, our results suggest that θ=42° when measured with single fibril resolution in rat tail collagen, in agreement with similarly high resolution work done by Williams et al. [18]. However, in mice, θ=51° when measured with single fibril resolution in tail, colon, or tumor, in these two different mouse strains. The difference between low- and high-resolution rat tail collagen measurements are likely assigned to differences in resolution, as previously suggested [18], and the similarity between our high resolution mouse collagen measurements and low resolution rat collagen measurements is merely coincidence. Overall, our polarization data suggests that not only does the reactive tumor stroma retain an appropriately ordered (i.e. SHG-producing) subpopulation of fibrillar collagen that is assembled in a similar fashion (i.e. with similar θ) to the ordered subpopulation in the mammary fat pad, but that this ordered subpopulation is assembled in a uniform manner throughout the animal, while in different species (at least in mice versus rats) the assembly can be altered, producing differing values of θ.

3.2 Comparison of the forward- and backward-scattering properties of collagen fibrils in tumor and healthy mammary fat pad tissue

The relationship between the forward-scattered and backward-scattered SHG signal provides information about the length scale, along the laser axis, over which the collagen fibril is appropriately ordered to produce SHG. When a highly focused laser beam is used to produce SHG from inhomogenously distributed ordered scatterers, the forward-scattered to backward-scattered (F/B) ratio approaches 1 as the scatterers’ spatial distribution becomes significantly smaller than the excitation wavelength, and rapidly increases as the distribution size approaches the excitation wavelength [19]. Therefore this F/B ratio can be used to estimate the overall spatial distribution of ordered triple helices in collagen fibrils.

We measured the average F/B ratio of the collagen fibrils in 100 um acute slices of TG1-1 tumor and healthy MFP tissue using the methods described above. Qualitatively, differences in morphology between TG1-1 SHG images and healthy MFP images are readily apparent in the resultant forward-scattered and backward-scattered images (see Fig. 4). In tumor tissue the backward-scattered SHG reveals evident thick bands of collagen, and the SHG within those bands is generally more uniform and homogenous than the forward-scattered image, which exhibits marked bright fibrils embedded in the bands (i.e. higher individual F/B ratios, see Fig. 6 (a), (b)). In healthy MFP tissue the collagen tends to be arranged in more delicate fibrillar structures (see Fig. 6 (c), (d)), does not share the evident “thick band” organization of the tumor collagen, and is instead more evenly spread across the image. If these images are quantified as discussed above, the average F/B ratio for TG1-1 tumor tissue is 33.8±7.7 (20 quantified images in N=4 animals) and 34.5±11.8 (20 quantified images in N=4 animals) in healthy FVB MFP. There was no statistically significant difference between the two populations (p>0.05). Likewise, the average F/B ratio for 4T1 tumor tissue is 44.5±15 (25 quantified images in N=5 animals) and 36.3±17 (25 quantified images in N=5 animals) in healthy BALB MFP, also not statistically significantly different (p>0.05), and neither is statistically significantly different from TG1-1 nor FVB MFP.

 

Fig. 6. Top Row: (a) Backward-, and (b) Forward-scattered image of collagen in an acute slice of a TG1-1 mammary adenocarcinoma. Bottom Row: (c) Backward-, and d) Forward-scattered image of collagen in an acute slice of a healthy FVB mammary fat pad. The small dots in the images are calibration beads. Images are 680 microns across.

Download Full Size | PDF

Interestingly, the F/B ratio did not vary significantly with the apparent diameter of collagen fibrils in either TG1-1 or FVB MFP (N=3 fibrils in each of 4 tumors and 4 MFPs: see Fig. 7). This F/B ratio analysis of collagen ordering with single fibril resolution has also been performed on rat tail collagen by Williams et al. [18]. They found an F/B ratio of ~1, also independent of fibril diameter (we confirmed those F/B values on our system, producing an F/B ratio for rat tail collagen of 1.21±0.14 (N=5 animals)). Williams et al calculated that an F/B ratio of ~1, independent of fibril diameter, corresponds to collagen fibrils that are appropriately ordered only in a thin shell of <20 nm thickness, with a center that is not capable of producing SHG due to a defect in collagen ordering [18]. According to the calculations of Mertz et al. [19], an F/B ratio of ~34 under 810 nm excitation corresponds to a ~70 nm Gaussian distributed cluster of scatterers. Hence we are presented with two possible interpretations of an F/B ratio of ~34 that does not vary with fibril diameter: collagen fibrils in murine tumors and in healthy MFP consist of either 1) an SHG-producing shell with a thickness of ~70 nm surrounding a center that is not capable of producing SHG due to a defect in collagen ordering, or 2) a collection of ~70 nm diameter rods, appropriately ordered throughout their interior, whose number varies with visible fibril thickness, but whose diameter remains constant. To determine the correct interpretation, we performed Electron Microscopy on TG1-1 and 4T1 tumors, as well as healthy mammary fat pads. We observed that the collagen in each of these samples consisted of multiple rod-like structures with a characteristic diameter on the order of the ~70 nm predicted by SHG (see Fig. 8). This suggests that in mouse tumor and mammary fat pad, collagen consists of bundles of ~70 nm diameter rods and, unlike the rat tail, this collagen is appropriately ordered throughout the rods and contributes to SHG signal.

 

Fig. 7. The F/B ratio does not vary with apparent diameter of collagen fibrils. A fit straight line slope of zero is within the 95% confidence interval. Note that the PSF has an e^-2 radius of 0.67 um. Consequently the stated apparent diameter is a convolution of the Gaussian PSF with the true fibril diameter. However, a deconvolution of the Gaussian PSF with the unknown true distribution (i.e. solid rod, hollow tube, etc.) will rescale the independent variable but will not impart a trend in the data which is not already apparent in the above plots.

Download Full Size | PDF

 

Fig. 8. TEM images of collagen in TG1-1 tumor (a), 4T1 tumor (b), and the corresponding healthy FVB and BALB mammary fat pads (c and d, respectively) consists of bundles of rods with a relatively constant characteristic diameter on the order of the ~70 nm predicted by SHG F/B ratio. Collagen fibrils intersecting the image plane transversely appear as streaks (a, b, d) and fibrils intersecting the plane perpendicularly appear as discs (c and d).

Download Full Size | PDF

The similarity between the average F/B ratio between these mammary tumors and healthy mammary fat pads suggests that the average spatial extent of ordering (along the laser axis, transverse to the fibril axis) is the same. The fact that the F/B ratio is ~34 and does not vary significantly with apparent fibril size, combined with the TEM images, suggest the collagen in all of these samples consists of collections of ~70 nm rods, appropriately ordered throughout to produce SHG. These results are somewhat surprising, as the reactive tumor stroma is typified by an altered collagen degradation machinery, with elevations of various MMPs [1, 20], and the SHG-producing collagen in the tumor samples should be accessible to enzymatic degradation (and hence detectable disruption of its ordering). It therefore seems likely that the SHG-generating subpopulation of collagen in tumors is somehow protected from the elevated levels of MMPs that are hallmarks of the reactive stroma [1, 20], perhaps by a coating of molecules such as decorin [25]. Consequently, it seems likely that MMPs in the reactive stroma primarily interact with the subpopulation of collagen which does not produce detectable SHG, and that degradation and synthesis of collagen in that inappropriately ordered subpopulation is the primary cause of the plasma-borne markers of collagen synthesis and degradation which are hallmarks of poor prognosis in breast cancer patients [5–7].

4. Conclusion.

We have shown that the angle of the SHG scatterers relative to the fibril axis in collagen fibrils in TG1-1 and 4T1 murine mammary tumors is equal to that in fibrils in the healthy mammary fat pad. We have also shown that the spatial distribution of ordered (i.e. SHG-producing) triple helices in collagen fibrils in TG1-1 and 4T1 tumors is equal to that in fibrils in the healthy mammary fat pad. Specifically, the triple helices form ~70 nm diameter rods, with multiple rods clustering together to produce larger diameter fibrils. Unlike collagen in the rat tail, these rods are appropriately ordered for SHG generation throughout their diameter. These observations are somewhat surprising in light of the myriad differences in both collagen synthesis and degradation between the tumor reactive stroma and the healthy mammary fat pad. They suggest that the subpopulation of collagen that is appropriately ordered to produce SHG is a “stable pool” whose location may be altered in tumorigenesis (i.e. collected in bands versus diffusely distributed, as shown in Fig. 6(a) vs 6(c) but whose production mechanism remains unchanged (i.e. similar θ) and which is somehow protected from the altered degradative environment (i.e. similar F/B ratio).

We have also shown that there are fundamental species-specific differences in these two key structural properties of SHG-producing collagen fibrils. In rat tail collagen the angle of SHG scatterers in collagen fibrils is significantly different from that in mouse tail collagen, while this angle remains unchanged between mouse tail, colonic submucosa, mammary fat pad, and mammary tumor tissue (across two different strains of mice). Similarly, the spatial distribution of ordered triple helices within collagen fibrils is significantly different in rat tail collagen versus mouse mammary fat pad while this distribution remains unchanged in mouse mammary tumor tissue (across two different strains). This suggests that these two properties are most significantly altered when the collagen structure itself is altered at the genetic level (i.e. rat genome vs mouse genome) versus alterations in the synthesis rates, ratios of collagen subtypes, or levels of degradation enzymes, such as found in the tumor reactive stroma. Therefore these optical properties may be useful in studying genetic disorders of collagen, such as in Osteogenesis Imperfecta [26].