1. Introduction

Tendon injury also referred to as tendonitis/tendinitis is a major orthopedic condition in human and equine athletes [1–3]. It is caused when the overused tendon loses its natural lubricating function resulting in inflammation, and leading to chronic pain and a long rehabilitation period. The tendon contains predominantly type I collagen and thus several studies have aimed at evaluating tendonitis through polarized light microscopy [4–7], which utilizes the birefringent nature of collagen fibers to produce contrast. It has thus become a conventional method for imaging collagen-based tissues and assessing tendon injury. However, it often involves staining the collagen fibers with picrosirius red to enhance the birefringence and moreover, is limited to two-dimensional imaging. Fortunately, collagen fibers also possess non-centrosymmetry and high second-order susceptibility —properties that permit emission of second-harmonic generation (SHG). In this second-order nonlinear optical scattering process, non-centrosymmetric structures emit light at half the wavelength of the incident light, thereby providing intrinsic contrast when used in an imaging modality—obviating the need for sample staining. In addition, since the intensity of the detected second-harmonic signal depends on the square of the incident light intensity, the generated second-harmonic signal is confined to a sub-femtoliter focal volume, thereby permitting 3D images of collagen-based tissues to be obtained. Thus, it is of no surprise that for the past decade, SHG microscopy has also become a popular technique for obtaining high contrast and highly specific images of collagen fibers in biological tissues [8–10]. The technique has also been used to study several diseases that affect collagen fiber organization [11–13]. However, according to our knowledge, it has not been heretofore applied to assessing collagenase-induced tendon injury—an approach used to simulate the effects of tendonitis.

In this paper, we compare conventional polarized light microscopy and SHG microscopy in their effectiveness for differentiating between normal tendon and collagenase-induced tendon injury. Recently, we showed that spatial Fourier analysis can be combined with SHG microscopy in a single platform, called Fourier transform-second harmonic generation (FT-SHG) microscopy, to quantify collagen fiber organization [14, 15]. We apply FT-SHG, along with the forward-to-backward (F/B) ratio of the emitted SHG signal [16, 17], to quantify the comparison between normal and injured tendon. In section 2, we describe the sample preparation procedure and the experimental setup for both polarized light and SHG microscopy. Results and discussion are provided in section 3, while section 4 both summarizes the work and provides a potential future direction.

2. Sample preparations and experimental setup

2.1 Tissue samples preparation

A clinically normal Quarter horse (age - 4 yrs) was used for the study. All procedures were approved by the University of Illinois Institutional Animal Care and Use Committee. Lameness exams and ultrasonographic evaluation of the flexor tendon were performed prior to inclusion in this study. The horse was pre-medicated with phenylbutazone (2.2 mg/kg), procaine penicillin G (22,000 IU/kg), and received a tetanus toxoid vaccine before induction of collagenase. The superficial digital flexor (SDF) tendon was injected with sterile bacterial collagenase at the mid-metacarpal region, followed by an additional dose of penicillin and phenylbutazone. The horse was maintained on strict stall confinement for 8 weeks and hand walked twice per day for the subsequent 8 weeks (total 16 weeks post-injection). Next, the subject was humanely euthanized and the collagenase-induced injured tendon sample was collected. Control (normal) tendon was also collected from the SDF tendon of the other contra lateral fore limb.

The sample preparation procedures for polarization and SHG microscopy are different; while the SHG microscopy uses cryostat sectioned samples, polarization microscopy utilizes paraffin embedded samples. For the latter procedure, the tendon samples were fixed in 4% paraformaldehyde, dehydrated, and mounted with fresh paraffin wax using a tissue processor (Leica ASP300). 6-μm thick sections were cut using a motorized microtome (Leica RM 2255). The sections were de-waxed and subsequently stained with picrosirius red to enhance the birefringent contrast of collagen fibers. Finally, the sections were dehydrated and secured between two glass cover slips using permount mounting medium.

For cryostat sectioning, the tendon samples were embedded in OCT compound and preserved at −80° C. The samples were brought to −20° C before cutting and 25-μm thick sections were cut using a cryostat (Leica CM3050S). The orientation of the samples was similar to that of paraffin embedded samples. The samples were then thawed and secured between two glass cover slips. In this case, since the samples could not be preserved, they were imaged fresh, i.e., within 24 hrs.

2.2 Experimental setup

2.2.1 Polarized light microscopy

The polarized light microscope used in this study was a standard Axiovert 200M microscope (Zeiss). A halogen lamp was used as the source to illuminate the samples, and a 1388 x 1030-pixel color CCD camera (Zeiss Axiocam MRc) was used for collecting the images. The samples were illuminated through a condenser and the transmitted light scattered from the sample was collected by a 0.8 numerical aperture (NA) objective. The two crossed polarizers placed above and below the sample ensured that the birefringence introduced by stained collagen fibers was converted to intensity for visualization. Polarized images of paraffin-embedded sections of both normal and injured tendons were collected by the camera using the image acquisition software Axiovision (version 4.8.1).

2.2.2 Second-harmonic generation microscopy

The experimental setup used for SHG microscopy is shown in Fig. 1 . The set up was a modified Zeiss LSM 710 microscope system equipped with a tunable Ti: Sapphire laser source that produces 70-fs pulses at a repetition rate of 80 MHz. The excitation wavelength was 780 nm. A quarter waveplate was used to generate circularly polarized light in order to ensure SHG emission from collagen fibers at all orientations. A galvoscanner was used to direct the beam in a raster-scan pattern. The beam was reflected by a short-pass 760-nm dichroic beam splitter and focused onto the sample using a 1.2 NA water-immersion objective. The emitted backward SHG signal was collected by the same objective, whereas the forward signal was collected by a 0.55 NA objective. In both geometries the same two filters were used: one filter (Semrock FF01-680/SP-25) was used to block the laser wavelengths, and the other (Semrock FF01-390/18-25) was a band-pass filter to transmit the SHG signal (390 nm). A photomultiplier tube (Hamamatsu R6357 multialkali) in both geometries was used to record the forward and backward SHG images. For all samples, the average power at the input to the objective was 3 mW. Other parameters such as pixel dwell time, detector gain, offset, and frame averaging were maintained constant between normal and injured tendon samples in both forward and backward directions to enable F/B ratio analysis.

 

Fig. 1 Experimental setup of the SHG microscope with both forward and backward collection geometries. Details are provided in text.

Download Full Size | PDF

3. Results and discussion

3.1 Fourier analysis on polarized light images of normal and injured tendon

Fourier analysis is applied to images of normal and injured tendon acquired using the polarized light microscope. Figure 2 summarizes the Fourier analysis on normal tendon. Images from three adjacent sites of the sample were considered for validation purposes, but only a representative image is shown (Fig. 2a). An orientation plot is superimposed on the polarization image to display the preferred orientation of the observed structures (such as collagen fibers and crimps) in each grid as shown in Fig. 2a. The alternating bright and dark bands (along the horizontal direction) are due to the crimps in the tendon; we also observe sample preparation artifacts such as splits. These structures, along with collagen fibers, influence the calculation of orientation. It is apparent that most structures in normal tendon are oriented roughly in the range ~80°- 125°. Regions that are considered dark (regions with low or negligible signal) and isotropic (regions with fibers aligned along several different orientations) are counted for comparison. It is clear that there are no dark regions; however, a few isotropic regions are observed. In addition, the 2D Fourier transform is applied to the entire tendon image in Fig. 2b in order to provide an estimate of the preferred orientation of the bulk. Here, we observe a broad distribution of spatial frequencies indicating a relatively wide range of orientations as evident from Fig. 2a. A bar plot of the average number of dark, anisotropic, and isotropic regions between the three images of normal tendon is shown in Fig. 2c. The mean values of the number of dark, anisotropic, and isotropic regions are 0, 62.33, and 1.67, respectively, suggesting that most regions are anisotropic (Fig. 2c). In addition, a histogram depicting the distribution of orientation values between the three images is plotted in Fig. 2d. The associated Gaussian fit has a mean of 98.4° and a standard deviation (or spread in orientation angles) of 10.0°.

 

Fig. 2 (a) Polarization image of a normal tendon overlaid with an orientation plot showing the preferred orientation of structures. (b) 2D Fourier transform of the polarization image shown in (a). (c) Bar plot of the number of dark, anisotropic, and isotropic regions averaged between three separate images of normal tendon. (d) Histogram of orientation values obtained from three different regions of normal tendon.

Download Full Size | PDF

In Fig. 3 we apply a similar analysis as in Fig. 2 to a collagenase-induced injured tendon, where again images from three adjacent sites of the sample were considered, but only a single representative image is shown (Fig. 3a). Comparing the polarization images in Figs. 2a and 3a, there is a clear difference in the crimp pattern between the normal and injured tendon. While the normal tendon has a well defined crimp pattern and spacing, the injured tendon has an irregular crimp pattern. However, the values of orientation for the injured tendon lie in a range similar to that of the normal tendon. Moreover, the Fourier transform (Fig. 3b) and the bar plot, representing the average number of dark, anisotropic, and isotropic regions (Fig. 3c) between the three images of injured tendon, are similar to that of the normal tendon (Fig. 2c), thereby making it difficult to distinguish. As evident from Figs. 2d and 3d, the distribution profiles of the histogram orientation for the normal and injured tendons are similar, and the difference in their mean orientations is small (<3.5°) considering their large values in spread (~10°).

 

Fig. 3 (a) Polarization image of an injured tendon overlaid with an orientation plot showing the preferred orientation of structures. (b) 2D Fourier transform of the polarization image shown in (a). (c) Bar plot of the number of dark, anisotropic, and isotropic regions averaged between three separate images of injured tendon. (d) Histogram of orientation values obtained from three different regions of injured tendon.

Download Full Size | PDF

3.2 Fourier analysis on SHG images of normal and injured tendon

Figure 4 summarizes how FT-SHG is used to assess collagen fiber organization in normal tendon. We again obtain images from three adjacent sites of the sample but choose to present only a representative image (Fig. 4a). The overlaid orientation plot showing the orientation of collagen fibers in the SHG image of normal tendon is shown in Fig. 4a. Unlike the polarization images, the SHG images show predominantly the collagen fibers present in the tissue, as is obvious from Fig. 4a. The fibers are extremely well-oriented at ~90° in normal tendon. Although the crimp patterns are not as prominent in the SHG image as they are in the polarization images they can still be observed. The crimps are fairly regular in normal tendon. The 2D Fourier transform is displayed in Fig. 4b. We observe two narrow lobes, horizontal and vertical, corresponding to the highly regular fibers and crimps, respectively (see Fig. 4b). Figure 4c shows the bar plot of the average number of dark, anisotropic, and isotropic regions between the three images of normal tendon. Since the fibers are regularly oriented, all regions are anisotropic. A distribution of orientation values between the three images is shown in the histogram plot in Fig. 4d. We observe that the distribution is narrow and centered at ~91° with a spread of ~3°. In Fig. 4e, an image of the forward-to-backward ratio (F/B ratio) is obtained by dividing the intensity value of each pixel in the forward SHG image by its corresponding pixel in the backward SHG image. In order to mitigate the influence of noise due to the dark regions of backward SHG image, we have considered only the pixels (using a threshold) with considerable signal above the noise floor for F/B ratio analysis. Note that the forward and backward images of normal and injured tendons are obtained using the identical optical setup and conditions, and hence a comparison of their ratios is a feasible metric. Regions where the backward intensity is greater than that of the forward are shown in green, while converse regions are shown in red. Regions with similar intensities are shown in black, while those that fall below the threshold are represented in cyan. The histograms plot of F/B ratios, from three separate images of normal tendon, is also shown in Fig. 4f. As evident from Figs. 4e and 4f, most regions in normal tendon produce comparable forward and backward SHG signal, and very few regions are below the threshold. The values of F/B ratio have a mean of 1.05 and a spread (standard deviation) of 0.50.

 

Fig. 4 (a) Preferred orientation of collagen fibers from the SHG image of normal tendon. (b) 2D Fourier transform of (a). The orientation of the Fourier transform (FT) is ~0° which corresponds to the orientation in real space (RS) of 90°. (c) Bar plot of number of dark, anisotropic, and isotropic regions obtained from three different regions of normal tendon. (d) Histogram of orientation from the three images of normal tendon. (e) Image of F/B ratio per pixel for normal tendon. (f) Histogram of F/B ratios from three different regions of normal tendon.

Download Full Size | PDF

A similar procedure to that carried out in Fig. 4, is shown in Fig. 5 for a representative SHG image of injured tendon. The collagen fibers in the injured tendon shown in Fig. 5a are clearly less ordered than those in normal tendon (Fig. 4a). It is also observed that the crimps are highly irregular in the injured tendon as opposed to fairly regular crimps in the normal tendon; this observation is consistent with the polarization images. Moreover, the Fourier transform of the injured tendon shows a single broad lobe as shown in Fig. 5b, which is an indication of the randomness (isotropy) of fiber orientations. A number of dark (mean=5.67) and isotropic (mean = 6.00) regions also exist in the injured tendon, indicating the disruption of collagen fiber organization, a characteristic feature of collagenase-induced tendon injury. These observations are consistent with those previously reported in the literature [4,7]. The injured tendon has a mean orientation of ~140°and a relatively large spread of ~17°. The F/B ratio image of injured tendon (Fig. 5e) is also different from the normal tendon (Fig. 4e). It contains more regions with intensities below the threshold, which are regions with missing fibers (dislodged fibers) due to disorganization. Figure 5f shows the histogram of F/B ratios from three separate images of injured tendon. We observe an increase in the values of the mean (1.60) and spread (1.11) for injured samples. These differences could be due to several parameters: changes in fiber organization (orientation and spacing between fibers), size, environment, and scattering coefficient of the tissue. Note that although a controlled study is required to determine the cause for the difference, we observe that the F/B ratio is still useful in differentiating normal and injured tendon.

 

Fig. 5 (a) Preferred orientation of collagen fibers from the SHG image of injured tendon. (b) 2D Fourier transform of (a). The orientation of the Fourier transform (FT) is ~40° which corresponds to the orientation in real space (RS) of 130°. (c) Bar plot of number of dark, anisotropic, and isotropic regions obtained from three different regions of injured tendon. (d) Histogram of orientation from the three images of injured tendon. (e) Image of F/B ratio per pixel for injured tendon. (f) Histogram of F/B ratios from three different regions of injured tendon.

Download Full Size | PDF

From the above results, it is clear that SHG microscopy is advantageous over the conventional polarization microscopy for the evaluation of tendon injury. As mentioned before, the principal observable difference between normal and injured tendon images obtained from polarization microscopy is the crimp pattern. However, this technique is not well-suited for quantifying the underlying structural differences in collagen fiber organization between the two tissue types, thereby limiting assessment of injury. Yet, direct assessment of collagen fiber organization is critical in tendon injury. We observe from the above results that, FT-SHG microscopy provides us with the following aids for assessment of tendon injury: 1) noninvasive (i.e., unstained) high-contrast 3D images of collagen fiber organization that permits direct differentiation between healthy and injured tendon; 2) manifestation of the aforementioned crimp pattern (although observed with less contrast than images obtained using polarization microscopy) in healthy tendon; and 3) objective quantification of the differences in collagen fiber organization in healthy and injured tendon. The first two points, in particular, are to be expected since it is well-established that SHG microscopy is sensitive to the intrinsic noncentrosymmetric structure of collagen fibers. However, until recently, the last point on quantification of structural organization has only been made clear when the technique is combined with harmonic analysis. Thus, our results suggest that FT-SHG could be a powerful tool for assessment of damage to collagen fiber organization.

It is also important to highlight an understated advantage that SHG microscopy has over polarization microscopy for tendon injury evaluation. That is, polarization microscopy does not possess the optical sectioning capabilities of SHG microscopy, and, therefore, it cannot avoid artifacts in the surface topology of the sample potentially introduced during sample preparation. Figure 6 gives an example of this for mechanical/cutting artifacts resulting from the use of a microtome. In particular, we observe in Fig. 6a cuts or sectioning artifacts (indicated by arrows) in the polarization image of the surface of healthy tendon tissue. Similarly, breaks in the fiber structure can be seen in Fig. 6b at the surface for the corresponding SHG image. Such artifacts (in both image types) would obviously affect accurate quantification of collagen fiber organization. However, unlike polarization microscopy, an image can be obtained well below the surface, away from the deleterious effects introduced by sample cutting. This is shown in Fig. 6c which is imaged 6 µm below the surface.

 

Fig. 6 (a) Polarized light image of normal tendon sample. (b) Backward SHG from the surface (c) and 6 µm into the sample. Arrows show sectioning artifacts.

Download Full Size | PDF

4. Conclusion

Fourier transform-second harmonic generation imaging has been shown to be capable of evaluating collagenase-induced tendon injury. Clear differences in collagen fiber organization between normal and injured tendon were observed. Injured tendon displayed less ordered collagen fibers in comparison to normal tendon. This observation was further supported by the use of two additional metrics. First, the average number of dark (no/minimal signal) and isotropic (no preferred fiber orientation) regions in the images were increased for injured tendon compared to normal tendon. Second, a histogram of F/B ratios for injured tendons showed a larger spread in its distribution compared to healthy tendon. Observations of collagen fiber disorganization in injured tendons were consistent with what has been reported in the literature. Biomechanically, abnormalities in crimp structure and collagen fiber organization results in a loss of performance of normal functions (e.g., crimps allow longitudinal elongation in response to loads, and collagen fibers provide stiffness) and thus degrades a subject’s ability to move. In addition, FT-SHG microscopy was shown to be more sensitive to assessing tendons injury than polarized light microscopy. It was also shown that the former technique avoids sample preparation artifacts. Future work includes incorporation of in-depth crimp analysis and quantification of other structural features (e.g., texture, fiber density). For extension to potential in vivo studies, an SHG endoscope may be designed comprising of a pulsed laser source, a fiber-optic catheter, and a MEMS-based scanning head. Such a device could aid in assessment of other injuries or diseases that affect collagen fiber organization such as fibrosis and rheumatoid arthritis, and the development of standards for evaluation and diagnostics.