http://informahealthcare.com/cts ISSN: 0300-8207 (print), 1607-8438 (electronic) Connect Tissue Res, 2015; 56(2): 68–75 ! 2015 Informa Healthcare USA, Inc. DOI: 10.3109/03008207.2015.1005209

ULTRASTRUCTURE IMAGING ORIGINAL RESEARCH

Effects of fixation and demineralization on bone collagen D-spacing as analyzed by atomic force microscopy Joseph M. Wallace1,2 1

Department of Biomedical Engineering, Indiana University-Purdue University at Indianapolis, Indianapolis, IN, USA and 2Department of Orthopaedic Surgery, Indiana University School of Medicine, Indianapolis, IN, USA Abstract

Keywords

Purpose/Aim: Collagen’s role in bone is often considered secondary. As increased attention is paid to collagen, understanding the impact of tissue preservation is important in interpreting experimental results. The goal of this study was to test the hypothesis that bone fixation prior to demineralization would maintain its collagen ultrastructure in an undisturbed state when analyzed using Atomic Force Microscopy (AFM). Materials/Methods: The anterior diaphysis of a pig femur was cut into 6 mm pieces along its length. Samples were mounted, polished and randomly assigned to control or fixation groups (n ¼ 5/group). Fixation samples were fixed for 24 h prior to demineralization. All samples were briefly demineralized to expose collagen, and imaged using AFM. Mouse tail tendons were also analyzed to explore effects of dehydration and fixation. Measurements from each bone sample were averaged and compared using a Mann–Whitney U-test. Tendon sample means were compared using RMANOVA. To investigate differences in D-spacing distributions, Kolmogorov– Smirnov tests were used. Results: Fixation decreased D-spacing variability within and between bone samples and induced or maintained a higher average D-spacing versus control by shifting the D-spacing population upward. Tendon data indicate that fixing and drying samples leaves collagen near its undisturbed and hydrated native state. Discussion: Fixation in bone prior to demineralization decreased D-spacing variability. D-spacing was shifted upward in fixed samples, indicating that collagen is stretched with mineral present and relaxes upon its removal. The ability to decrease variability in bone suggests that fixation might increase the power to detect changes in collagen due to disease or other pressures.

2-Dimensional fast Fourier transform, glutaraldehyde, nanoscale, paraformaldehyde, ultrastructure

Introduction Bone has an elegant structural hierarchy which spans 9–10 orders of magnitude in length scale. While non-collagenous proteins and water play important roles (1,2), bone is primarily a two-phase composite material made of a flexible organic matrix (90% Type I collagen, hereafter referred to as collagen) impregnated with and surrounded by a stiffer reinforcing mineral phase. The combination of strength and toughness that are characteristic of bone are derived from the intimate interaction between these nanoscale constituents of vastly differing mechanical properties (3–5). However, given its complex composite nature and structural hierarchy, relationships between structure, quality and function in bone are largely not understood. Although much basic and clinical

Correspondence: Dr Joseph M. Wallace, Department of Biomedical Engineering, Indiana University-Purdue University at Indianapolis, 723 W Michigan St. SL220D, Indianapolis, IN 46202. USA. Tel: (317) 2742448. Fax: (317) 278-2455. E-mail: [email protected]

History Received 26 June 2014 Revised 22 September 2014 Accepted 27 October 2014 Published online 27 January 2015

research focuses on bone’s mineral phase, collagen’s role in bone health is often considered secondary. Recent work has begun to uncover the key role collagen plays in determining the material and mechanical properties of bone (6–9). As the most abundant protein in mammals (10), the synthesis of collagen begins as a polypeptide chain of amino acids which takes on a left-handed helical configuration [the a-chain (11,12)]. Three a-chains (two a1, one a2 in the case of Type I collagen) then counter-wind into a heterotrimeric righthanded triple helix. Non-helical propeptide ends are enzymatically cleaved and the resulting tropocollagen molecules assemble into a parallel/staggered twist to form a threedimensional collagen fibril (13), the primary structural building block of many tissues. The enzyme lysyl oxidase then catalyzes the conversion of amino groups of specific lysine or hydroxylysine residues to reactive aldehydes which later form stabilizing covalent crosslinks (14). Due to space between the ends of the tropocollagen molecules and an offset from row to row, fibrils have a characteristic oscillating topographical feature known as the D-periodicity or D-spacing (Figure 1).

DOI: 10.3109/03008207.2015.1005209

Effects of fixation and demineralization on bone collagen D-spacing

69

Figure 1. Height and error image of collagen. This figure shows a typical height (A) and error (B) image obtained while imaging a bone’s surface in the current study. The image is 3.5 mm  3.5 mm and was obtained in peak force tapping mode. The false color image in panel A has a total height range of approximately 50 nm. Note the rich collagen surface with D-spacing of fibrils easily visible as a striped pattern on individual fibrils.

Several recent studies utilizing atomic force microscopy (AFM) have demonstrated that collagen’s D-spacing exists with a distribution of values rather than being a singular value of 67 nm as originally postulated (15). Distributions have been reported in bone (7), tendon (16), dentin (17) and skin (18). Unpublished work has verified distributions in collagen produced in vitro by fibroblasts and osteoblasts, and in collagen purified from skin and tendon. Investigations of disease [genetic (16,19) and surgically induced (7,20,21)], pharmacological treatment (22) and mechanical stimulation (20) have all been undertaken in an attempt to understand how collagen’s morphology may be a window into the nanoscale state of bone. These studies reproducibly demonstrate that D-spacing distributions exist in all Type I collagen-based tissues, and these distributions shift under a variety of experimental conditions. D-spacing provides information on the internal structure and state of tropocollagen molecules, which can be driven by structural defects or alterations in post-translation modifications during synthesis. In addition, information about the state of enzymatic and non-enzymatic crosslinks can be ascertained. Data suggest that diseases which have a direct impact on collagen synthesis (e.g. osteoporosis and osteogenesis imperfecta, OI) tend to shift D-spacing distributions downward (7,16,18,19,23). In most cases, crosslinking (both enzymatic and non-enzymatic) shifts D-spacing upward (6,24). Attempts have been made to directly correlate changes in D-spacing to larger-scale mechanical properties. Atomic force microscopy was used to extract mechanical information from collagen in tendon. Phenotypic changes in fibril morphology in tendon from OI mice were accompanied by an increased elastic modulus (16). In diabetic rat tendon, increased fibril modulus was correlated with increased mechanical stiffness and strength at larger length scales

(24). A recent characterization of OI bone also revealed connections between nanoscale changes in collagen and whole bone functional deficiencies, although not directly (23). Together, these studies demonstrate that changes in D-spacing and fibril mechanics are present in diseased tissues, and may be related to structural mechanical alterations. One previous study of note directly compared bone and tendon from the same animal imaged using the same AFM under the same imaging conditions (6). The mean D-spacing for bone was nearly 2.5 nm lower than that of tendon and the distribution in bone had more than double the overall range. The majority of tendon’s distribution (82%) was contained within a 2.5 nm range of its mean, while within the same range of bone’s mean, only 46% of the distribution was found. This observation of a wider and shifted distribution of D-spacing in mineralized bone versus non-mineralized tendon raises the possibility that something inherent about the tissues drives these distribution differences. Bone is mineralized and as part of the preparation process for AFM imaging, some mineral must be removed to expose the underlying collagen for analysis. Since a portion of this mineral exists within the confines of the fibrils being analyzed (i.e. intrafibrillar mineral), it is possible that its removal causes a portion of that structure to collapse without the mineral supporting it. This collapse could lead to a more heterogeneous fibril population and a shift in the population center. Although analyzing the tissue with mineral removed does give insight into collagen’s structure and exposes phenotypes in mineralized tissues, it could also be important to preserve the structure prior to mineral removal. As a means to prevent these changes, fixation can be used to chemically crosslink the structure prior to mineral removal (25,26). These samples can then be used for quantitative analyses of native structure in bone (27–29). The goal of this study was to test

70

J. M. Wallace

Connect Tissue Res, 2015; 56(2): 68–75

Figure 2. Bone sample preparation. To isolate the diaphysis, proximal and distal ends of the femur were removed (A, grayed area indicates portion of bone that was removed, CT image obtained from 3dcadbrowser.com). The anterior 1/2 of the bone was cut away (B) and ground flat using silicon carbide sandpaper (C). This bone sample was segmented into 10  6 mm pieces (D). Samples were mounted on a steel disc and polished using a 3 mm diamond suspension (E). Following sonication in water, samples were stored wrapped in saline-soaked gauze at 20  C.

the hypothesis that fixation of bone prior to demineralization would maintain its collagen structure in an undisturbed state. To test this hypothesis, bone samples were fixed, demineralized and imaged or demineralized and imaged without fixation using AFM. D-spacing of individual collagen fibrils was analyzed using a 2D fast Fourier Transform (2D FFT) approach. In addition, to study the effects of fixation and dehydration on collagen D-spacing and attempt to define a native state, non-mineralized mouse tail tendon samples were investigated.

Materials and methods Bone tissue collection A right porcine hind quarter was obtained fresh from a local butcher. Soft tissue was immediately removed and the femur was cleaned of any remaining adherent periosteum. To isolate cortical bone in the diaphysis, the proximal and distal ends were removed under constant water irrigation using a bone band saw equipped with a diamond-coated blade (Mar-med Inc., Cleveland, OH) leaving behind an approximately 6 cm long piece of bone (Figure 2). Marrow was carefully removed using a combination of a scalpel and a high-pressure water irrigation system (Water Pik Inc., Fort Collins, CO). The anterior 1/2 of the bone’s thickness was carefully cut away and the cut surface was ground flat using 320 grit wet silicon carbide sandpaper to reduce sample tilt and overall thickness. The bone was then cut into ten 6 mm long pieces along its length. Each sample was mounted anterior-side up to a steel AFM disc using cyanoacrylate glue, and polished using a 3 mm water-based diamond suspension as previously described [Buehler Ltd., Lake Bluff, IL; (6,30)]. Samples were sonicated in ultrapure water to remove polishing debris, wrapped in gauze soaked with phosphate-buffered saline (PBS) and frozen at 20  C until needed.

Bone sample processing – fixation and demineralization The 10 samples were randomly assigned to one of the two treatment groups (control and fixed). For control group samples, surface mineral was removed and the underlying collagen fibrils were exposed through demineralization in 0.5 M ethylenediaminetetraacetic acid (EDTA) at a pH of 8.0 as previously described (6,30). Samples were submerged in ultrapure water and brought to room temperature, then placed in EDTA with shaking for 20 min followed by sonication in ultrapure water for 5 min. This process was repeated three times. A final cycle of demineralization occurred in fresh EDTA for 10 min followed by 5 min of sonication in water. The sample was rinsed in water and allowed to air dry prior to imaging. Samples in the fixed group were placed in 3 ml of the fixation buffer at room temperature for 30 min, then at 4  C for 24 h prior to demineralization. The fixation buffer contained 0.8% paraformaldehyde and 0.2% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA) in PBS (Life Technologies, Grand Island, NY) at a of pH 7.4 (31). After the initial fixation, samples in the fixed group were demineralized under the same conditions as the control group with the exception that the 0.5 M EDTA contained 0.2% paraformaldehyde and 0.05% glutaraldehyde as a postfixative (31). Bone samples atomic force microscopy imaging and analysis Methods associated with AFM imaging and analysis have been described elsewhere (6,30). Briefly, 3.5 mm  3.5 mm images were obtained in peak force tapping mode using a Bruker Catalyst AFM equipped with ScanAsyst Fluid + probes (Bruker, Santa Barbara, CA). Peak force error images were analyzed to investigate the D-periodic

DOI: 10.3109/03008207.2015.1005209

Effects of fixation and demineralization on bone collagen D-spacing

spacing of individual fibrils (Figure 1). A total of 60–65 fibrils were analyzed from at least five locations in each sample. To measure the D-spacing of each fibril, a twodimensional Fast Fourier Transform (2D FFT) was employed. Briefly, the 2D FFT was performed on a rectangular region of interest (ROI), and the D-period was determined from the first harmonic peak in the power spectrum (SPIP v5.1.5, Image Metrology, Hørsholm, Denmark). Tendon collection, processing and imaging Five female mice from the C57BL/6 background strain were obtained from Harlan at 10 weeks of age with prior IACUC approval (protocol number SC210R). Mice were sacrificed by CO2 inhalation and tails were removed whole and processed for imaging. The skin of each tail was pulled back at the base and removed in a base-to-tip direction to expose the underlying tendon bundles. From each tail, 50 mm lengths of individual fascicles were removed and placed in PBS. Fascicles was placed on a glass slide and flattened with curved forceps. For samples that were imaged unfixed, the sample was immediately imaged in fluid then dried in ambient conditions and imaged dry. For those samples undergoing fixation, the fixation protocol was identical to the bone samples above. Samples were placed in 20 ml of the fixation buffer at room temperature for 30 min, then at 4  C for 24 h. Samples were then washed for 15 min in water and imaged wet. Following wet imaging, the sample was dried in ambient conditions and imaged dry. 5 mm  5 mm images were obtained in peak force tapping mode (ScanAsyst Fluid + probes). D-spacing was analyzed from 2 to 3 locations in each of 2–3 fascicles for a total of 55–60 fibrils per tail and condition.

71

Results Bone results All measurements in a sample were averaged, yielding a single D-spacing measure for that sample. The five control values were then compared with the five fixed values. The mean values were 64.9 ± 0.4 nm and 65.8 ± 0.2 nm for control and fixed, respectively. The data indicate that fixation resulted in a significant increase in mean fibril D-spacing (p ¼ 0.008). In Figure 3, the data from each bone and the overall pooled data from all measurements in each group are displayed as boxplots (Control: panel A, Fixed: panel B). These graphs are intentionally shown on the same scale to demonstrate differences in data spread. The box represents the interquartile region (IQR, the middle 50% of data), while the line in the box is the median and the diamond is the mean. The whiskers extend from the top of the third quartile up by 1.5 times the IQR and from the bottom of the second quartile down by the same amount. In addition, all values outside of the whisker range are indicated with closed circles to demonstrate the full data range. Qualitatively, it is apparent that measurements in

Statistical analysis All statistical analyses utilized SPSS (Version 21.0, IBM, Armonk, NY). For all investigations, a value of p50.05 was considered significant unless noted. To investigate the effect of fixation in bone, values measured from each sample were averaged, yielding a single D-spacing for that sample. Despite verifying normality of residuals and homoscedasticity, mean values were compared using a non-parametric Mann–Whitney U-test since the 10 bones did not come from independent biological samples. To investigate differences in D-spacing distributions between groups, histograms (bin size ¼ 1 nm) and the cumulative distribution function (CDF) of all fibrils measured in each group were also computed. A Kolmogorov– Smirnov (KS) test was applied to the distribution data (Control: n ¼ 334, Fixed: n ¼ 316). In tendons, values measured from each mouse and imaging condition were averaged, yielding a single D-spacing value. Comparisons between groups were made using a repeated measures ANOVA (RMANOVA; with before and after drying being the repeated measure). To investigate differences in D-spacing distributions between groups, histograms and CDFs of all fibrils measured in each group were also computed. Kolmogorov–Smirnov tests were then applied to the distribution data. Because of the multiple comparisons being made, a Bonferroni correction was applied reducing the p value needed to attain significance to p50.01.

Figure 3. Boxplot representation of D periodic spacing values in bone. This figure shows the boxplot for each control (A) and fixed (B) sample and the overall boxplot for all samples pooled in each group. The box is the interquartile region (IQR) from all measured fibrils (middle 50% of the data), the line inside of the box is the median, and the diamond is the mean. The whiskers on the box represent the Q1–1.5*IQR and Q3 + 1.5*IQR for that group. Values between 1.5*IQR and 3*IQR away from Q1 or Q3 are mild outliers (shown as circles). Outliers were included in all analyses and are only shown to indicate the full data range. When the mean values from the five control samples (64.9 ± 0.4 nm) were compared to the five fixed samples (65.8 ± 0.2 nm), the difference in D-spacing was significant (p ¼ 0.008).

72

J. M. Wallace

the fixed samples were less variable. The overall data range in the fixed group was 6.6 nm (with an average of 5.2 ± 0.7 nm in the five samples), while the range in the control group was nearly double (12.0 nm with an average of 7.8 ± 2.2 nm). Differences in the data spread and distribution are more readily apparent in Figure 4. From the histograms in Figure 4(A), one can observe that the distribution in the fixed group was narrower and shifted towards higher spacing values. The overall standard deviation from the fixed group (1.1 nm from n ¼ 316 fibrils) was nearly half that of the control group (1.9 nm from n ¼ 334 fibrils). The majority of the fixed distribution (75.5%) was contained within a 2.5 nm range of its mean (mean ± 1.25 nm), while only 51% of the control sample distribution was found within a similar 2.5 nm of its mean. The control sample range needed to be increased to 4 nm to include the same percentage of its distribution (74%). The strong negative skewness of the control distribution (0.70) versus the mild skewness in the fixed samples (0.32) further demonstrates that the control population had a long tail towards lower spacing values. The CDF was computed and shown to be significantly different between groups (p50.001, Figure 4B). The CDFs clearly demonstrate that the fixed population distribution was shifted higher versus control. In addition, the steep almost

Figure 4. Histogram and CDF for D-periodic spacing in bone. Panel A shows the histogram and panel B shows the CDF computed from measurements in each sample group (n ¼ 334 for control and n ¼ 316 for fixed). In panel B, the mean of each group is marked with a star. The fixed population was shifted to higher spacing values and the distribution was visibly narrower. The Kolmogorov–Smirnov test indicated a significant difference in the two populations (p50.001).

Connect Tissue Res, 2015; 56(2): 68–75

linear slope in the center of the fixed distribution is reflective of the sharp peak noted in its histogram. The control slope is shallower, reflective of its widened distribution. Mouse tail tendon results All measurements in a sample and condition were averaged, yielding a single D-spacing measure for that sample. For unfixed samples, the mean values were 67.7 ± 0.2 nm and 68.1 ± 0.3 nm for wet and dried, respectively. For fixed samples, the mean values were 67.5 ± 0.3 nm and 67.7 ± 0.4 nm for wet and dried, respectively. RMANOVA indicated a significant difference for the main effect of dehydration (p ¼ 0.03) but no effect of fixation nor an interaction between dehydration and fixation. The full data set from each group is displayed as boxplots in Figure 5. The overall data range in the groups ranged from 4.5 nm (fixed wet and fixed dry) to 5.4 nm (unfixed wet). Cumulative distribution functions were also computed for each group (Figure 6A). As indicated by the tight data ranges noted above, all four distributions in tendon were narrower than those observed in bone in the current study (Figure 6B). The largest population standard deviation was 0.8 nm in tendon versus the 1.1 nm seen in fixed bone samples. In unfixed tendon samples, the dried population was shifted to significantly higher D-spacing values versus the wet population (blue versus green: rightmost versus second rightmost curve, p50.001). In fixed samples, there was no significant shift when comparing dried versus wet samples (yellow versus gray: leftmost versus second leftmost curve, p ¼ 0.079). There was a significant effect of fixation alone when comparing unfixed wet samples to fixed wet (blue versus yellow, p ¼ 0.003). The fixed sample population was pushed downward for most of its range. The effect of fixation alone was exacerbated when these samples were dried and compared (green versus gray, p50.001). Because the fixed population did not change when dried, this change was primarily driven by the large shift in unfixed samples upon drying. Finally, the comparison between unfixed wet and fixed dried samples showed no significant difference, with the CDFs being nearly on top of one another (blue versus gray, p ¼ 0.181).

Figure 5. Boxplot representation of D-periodic spacing values in tail tendon. This figure shows the boxplot for each tail tendon treatment group. A RMANOVA indicated a significant difference for the main effect of dehydration (p ¼ 0.03) but no effect of fixation nor an interactive effect between fixation and dehydration.

DOI: 10.3109/03008207.2015.1005209

Effects of fixation and demineralization on bone collagen D-spacing

Figure 6. Cumulative distribution function for D-periodic spacing in tail tendon. Panel A shows the CDF computed from measurements in each tendon sample group. In panel B, a comparison between tail tendon and bone CDFs is shown. In unfixed tendon, the dried population (rightmost curve) was shifted to significantly higher D-spacing values versus the wet population (second rightmost curve, p50.001). In fixed samples, there was no significant shift when comparing wet (leftmost curve) versus dried samples (second leftmost curve, p ¼ 0.079). There was a significant effect of fixation alone when comparing unfixed wet samples to fixed wet (p ¼ 0.003), and this effect was exacerbated when these samples were dried and compared (p50.001). Finally, the comparison between unfixed wet and fixed dried samples showed no significant difference (p ¼ 0.181).

Discussion The main goal of this study was to investigate the effects of chemical fixation in bone prior to demineralization on an important metric of collagen’s nanoscale morphology, the characteristic axial D-spacing. It was hypothesized that fixation would maintain the structure of collagen in an undisturbed state, mitigating effects caused by the removal of mineral. Data indicate that fixation profoundly impacted collagen’s structure, decreasing the variability in D-spacing both within and between samples and either inducing (or maintaining) a higher average D-spacing versus samples demineralized without fixation. This tendency to decrease variability, especially in bone samples which are inherently more variable than other tissues, suggests that fixation might increase our power to detect changes in collagen due to disease or other external stimuli or pressures. An important point to note is that although fixation may preserve morphological features of interest, the effects of fixation on mechanical properties in collagen-based tissues are almost certainly significant and would likely lead to stiffening due to

73

crosslinking. If the goal of a study is to ultimately perform mechanical characterizations on the sample, the use of chemical fixation should be avoided. Studies are currently underway to determine how fixation impacts our ability to detect phenotypic difference in various murine disease models. Since mineral exists both within and around collagen fibrils in bone, imaging techniques such as AFM and electron microscopy (31) require the removal of some or all of the minerals if collagen morphology is the target for analysis. In the current study, chemical fixation was used to preserve collagen’s structure prior to demineralization. Fixation of a tissue by immersion in a chemical fixative occurs through a sequence of actions which differs depending on the fixative being used (25). The goal of fixation is to preserve ultrastructural integrity and relationships which might otherwise be lost due to dehydration or sample processing (26,31). Aldehydes (e.g. paraformaldehyde and glutaraldehyde) work by reacting with proteins to form covalent crosslinks. In the case of collagen, these aldehydes react with lysine or hydroxylysine residues (similar to what occurs during enzymatic crosslinking in collagen). Secondary reactions which depend on the fixative follow. Presumably, these crosslinks lock the structure in place and prevent changes that occur as mineral and water are removed. A mild combination of these fixatives was used here (31,32). As a note, the upward shift in D-spacing seen with fixation is similar to mild upward shifts previously noted in diabetic bone and tendon (6,24). The proposed mechanism at work in the diabetic tissues is the presence of advanced glycation end products (AGEs) or permanent non-enzymatic crosslinks which form when reducing sugars react with free amino groups in proteins. The data together suggest that AGEs may partially stabilize the collagen ultrastructure in a process similar to chemical fixation (33). In the current analysis in bone, the assumption was made that the mild fixative used allows for a study of baseline, undisturbed bone (i.e. fully hydrated and fully mineralized). To date, it has not been possible to study D-spacing in fully hydrated bone as fibrils are swollen and fail to exhibit adequate D-spacing for analysis. In addition, fully mineralized bone contains little exposed collagen that is available for Dspacing analysis. Tail tendon, also based on Type I collagen, can serve as a surrogate for such an analysis to define an undisturbed or native state. A secondary set of experiments was undertaken to investigate the effects of fixation and dehydration in type I collagen using mouse tail tendons. These samples were either imaged in fluid, then dried and imaged dry or were first fixed for 24 h and imaged hydrated, then dried and imaged dry. Several important conclusions were drawn from these experiments. Similar to a previous study in tendon (16), the D-spacing population was shifted to higher values when unfixed dry samples were compared with unfixed wet samples (Figure 6). In fixed samples, although the population was also shifted slightly higher when dried, this shift was not significant (p ¼ 0.079). This p value was marginal, but the KS test is robust and very sensitive to population changes. A p value greater than 0.05 (0.01 in this case because of the Bonferroni correction) clearly indicates that this difference is not relevant. This observation suggests

74

J. M. Wallace

that fixation, on its own, mitigates the effects of dehydration. There was a significant effect of fixation alone when comparing unfixed wet samples to fixed wet samples (p ¼ 0.003). The fixed sample population was pushed downward for most of its range. Interestingly, this was not the same type of effect seen in bone where the effect of fixation was to push the population upward and tighten it (Figure 4). However, those bone samples were imaged dried. In tendon, the effect of fixation alone was exacerbated when samples were dried (p50.001). Because the fixed population did not change when dried, this change was primarily driven by the large shift in unfixed samples upon drying. While these previous observations were interesting and important, the finding that was most germane to the interpretation of the current bone data was that the D-spacing distribution from fixed and dried tendon fibrils was indistinguishable from unfixed fibrils imaged wet (p ¼ 0.181). This finding suggests that fixing and drying samples leaves collagen near its undisturbed and hydrated native state. Although tendon is not the same tissue as bone, and the presence of mineral may alter the native state, these data support the idea that dried and fixed samples are an appropriate approximation of the native state in bone. Several AFM-based experiments have investigated D-spacing in bone and other collagen-based tissues. D-spacing distributions have been observed in all tissues studied, but those in bone are consistently wider and shifted towards lower spacing values versus non-mineralized tissues (6,30,34). A lower average D-spacing in mineralized dentin versus non-mineralized tendon was also recently been shown using transmission electron microscopy (31). The sharper peak of the D-spacing distribution noted here in the fixed bone samples resembles data obtained in mouse tail tendon samples, although the mean value was still shifted lower by 2.3 nm in fixed bone (Figure 6B). In fact, all tendon distributions measured in the current study were narrower than those seen in bone and were shifted upward to higher spacing values. This finding suggests that while fixation may mitigate the widening of the distribution due to the removal of mineral, the collagen ultrastructure of the two tissues may still exhibit fundamental differences. The structure of collagen in bone relies on mineral (4) and ultrastructural features of collagen would be expected to change when that mineral is removed (33). Previous observations in bone have been interpreted to mean that upon removal of some mineral during processing, collagen relaxes which widens the distribution and shifts a portion of the D-spacing towards lower values. If we take the fixed samples to represent collagen’s structure in bone when fully mineralized, the effect of fixation was to minimize the downward shift in D-spacing by preventing this relaxation. This interpretation would indicate that when mineral is present, collagen in bone is stretched out and under a tensile prestress (33), contradictory to other studies suggesting compressive pre-stresses and decreased D-spacing in mineralized bone (35,36). This conflicting observation has yet to be reconciled. One possible explanation is that D-spacing here is measured in individual fibrils and may capture aspects of collagen’s ultrastructure that are lost using other techniques which average over a larger scale [e.g. neutron scattering,

Connect Tissue Res, 2015; 56(2): 68–75

X-ray diffraction (35,36)]. Another possibility is that water and mineral are not fully removed here as opposed to other studies. The light demineralization with EDTA likely kept much of the mineral in place, and air drying was not enough to drive out all bound water. The fixed samples would have had their structure locked in place in a fully hydrated and mineralized state and the D-spacing would be reflective of this. The comparison here between fixed and control samples was therefore not one that also considered full dehydration and demineralization. A full demineralization was attempted with and without fixation. However, the samples were grossly deformed after drying and were unsuitable for AFM analysis. Some limitations of the current study deserve mentioning. The goal of this study was to focus on the effects of chemical fixation in bone prior to demineralization. However, demineralization in bone prior to fixing is an important additional study that is needed to understand what effects, if any, fixation of collagen in bone has after the removal of mineral. The use of a single bone from a single animal may be considered a limitation, but this choice was made to decrease biological variability and increase the chance of detecting the effect of interest. Within a single bone, regional effects along the bone’s length are expected. For this reason, the 10 samples from within the anterior quadrant were randomized into the two treatment groups. Finally, comparing these effects in a treatment group where differences in D-spacing are expected is an important next step. The argument can be made that in previous work where changes in bone were detected (e.g. in a model osteogenesis imperfect [OI], (19)), the observed effect was driven by the removal of mineral and the collapse of the defective collagen. In vivo, collagen would be supported by the mineral so the observed difference may not be important. This argument may have merit, but observing the altered collagen is still important to understanding functional outcomes at the nanoscale and larger length scales. In addition, a phenotype may still be present with mineral in place and the decreased variability from fixation may increase our power to detect it. As noted above, studies are currently underway to answer these questions. In conclusion, this study demonstrated that fixation in bone prior to demineralization decreases the variability in collagen D-spacing within and between samples. The mean D-spacing was shifted upward in fixed versus non-fixed samples, indicating that collagen is stretched with mineral present and relaxes when that mineral is removed. The ability to decrease variability in bone might enhance the chance of detecting changes in collagen due to disease or external stimuli.

Declaration of interest The author wish to confirm that there are no known conflicts of interest associated with this publication and there has been no financial support for this work that could have influenced its outcome. This work was supported by IUPUI departmental startup funds, Research Support Funds Grant from the IUPUI Office of the Vice Chancellor for Research and funding from the IUPUI Biomechanics and Biomaterials Research Center.

DOI: 10.3109/03008207.2015.1005209

Effects of fixation and demineralization on bone collagen D-spacing

References 1. Thurner PJ, Chen CG, Ionova-Martin S, Sun L, Harman A, Porter A, Ager Iii JW, Ritchie RO, Alliston T. Osteopontin deficiency increases bone fragility but preserves bone mass. Bone 2010;46: 1564–73. 2. Poundarik AA, Diab T, Sroga GE, Ural A, Boskey AL, Gundberg CM, Vashishth D. Dilatational band formation in bone. Proc Natl Acad Sci 2012;109:19178–83. 3. Ritchie RO, Buehler MJ, Hansma P. Plasticity and toughness in bone. Phys Today 2009;62:41–7. 4. Landis WJ. The strength of a calcified tissue depends in part on the molecular structure and organization of its constituent mineral crystals in their organic matrix. Bone 1995;16:533–44. 5. Gupta HS, Wagermaier W, Zickler GA, Raz-Ben Aroush D, Funari SS, Roschger P, Wagner HD, Fratzl P. Nanoscale deformation mechanisms in bone. Nano Lett 2005;5:2108–11. 6. Hammond MA, Gallant MA, Burr DB, Wallace JM. Nanoscale changes in collagen are reflected in physical and mechanical properties of bone at the microscale in diabetic rats. Bone 2014;60: 26–32. 7. Wallace JM, Erickson B, Les CM, Orr BG, Holl MMB. Distribution of type I collagen morphologies in bone: relation to estrogen depletion. Bone 2010;46:1349–54. 8. Nyman JS, Reyes M, Wang X. Effect of ultrastructural changes on the toughness of bone. Micron (Oxford, England: 1993); Micron (Oxford, England: 1993) 2005;36:566–82. 9. Makowski AJ, Uppuganti S, Wadeer SA, Whitehead JM, Rowland BJ, Granke M, Mahadevan-Jansen A, Yang X, Nyman JS. The loss of activating transcription factor 4 (ATF4) reduces bone toughness and fracture toughness. Bone 2014;62:1–9. 10. Kadler KE, Holmes DF, Trotter JA, Chapman JA. Collagen fibril formation. Biochem J 1996;316:1–11. 11. Canty EG, Kadler KE. Procollagen trafficking, processing and fibrillogenesis. J Cell Sci 2005;118:1341–53. 12. Petruska JA, Hodge AJ. A subunit model for the tropocollagen macromolecule. Proc Natl Acad Sci USA 1964;51:871. 13. Orgel JPRO, Irving TC, Miller A, Wess TJ. Microfibrillar structure of type I collagen in situ. Proc Natl Acad Sci 2006;103:9001–5. 14. Knott L, Bailey AJ. Collagen cross-links in mineralizing tissues: a review of their chemistry, function, and clinical relevance. Bone 1998;22:181–7. 15. Hodge AJ, Petruska JA. Recent studies with the electron microscope on ordered aggregates of the tropocollagen molecule. In: Ramachandran GN, ed. Aspects of protein structure. New York: Academic Press; 1963:289–306. 16. Kemp A, Harding C, Cabral WA, Marini JC, Wallace JM. Effects of tissue hydration on nanoscale structural morphology and mechanics of individual type I collagen fibrils in the Brtl mouse model of Osteogenesis Imperfecta. J Struct Biol 2012;180:428–38. 17. Wallace JM, Chen QS, Fang M, Erickson B, Orr BG, Holl MMB. Type I collagen exists as a distribution of nanoscale morphologies in teeth, bones, and tendons. Langmuir 2010;26:7349–54. 18. Fang M, Liroff KG, Turner AS, Les CM, Orr BG, Holl MMB. Estrogen depletion results in nanoscale morphology changes in dermal collagen. J Invest Dermatol 2012;132:1731–97. 19. Wallace JM, Orr BG, Marini JC, Banaszak Holl MM. Nanoscale morphology of type I collagen is altered in the Brtl mouse model of Osteogenesis Imperfecta. J Struct Biol 2011;173:146–52.

75

20. Warden SJ, Galley MR, Hurd AL, Wallace JM, Gallant MA, Richard JS, George LA. skeletal exercise in young female rats provides lifelong benefits to cortical bone properties independent of an artifical menopause. Endocrinology 2013;154:3178–87. 21. Newman CL, Moe SM, Chen NX, Hammond MA, Wallace JM, Nyman JS, Allen MR. Cortical bone mechanical properties are altered in an animal model of progressive chronic kidney disease. PLoS One 2014;9:e99262. 22. Gallant MA, Brown DM, Hammond MA, Wallace J, Du J, Deymier-Black A, Stock S, Allen M, Burr DB. Bone cellindependent benefits of raloxifene on the skeleton: a novel mechanism for improving material properties. Bone 2014;61: 191–200. 23. Bart ZA, Wallace JM, Hammond MA. Multi-scale analysis of bone chemistry, morphology and mechanics in the OIM model of Osteogenesis Imperfecta. Connect Tissue Res 2014;55:4–8. 24. Diaz-Gonzalez A, Gallant MA, Burr DB, Wallace JM. Multiscale analysis of morphology and mechanics in tail tendon from the ZDSD rat model of type 2 diabetes. J Biomech 2014;47:681–6. 25. Chapman JA, Tzaphlidou M, Meek KM, Kadler KE. The collagen fibril – a model system for studying the staining and fixation of a protein. Electron Microsc Rev 1990;3:143–82. 26. An YH, Martin KL. Handbook of histology methods for bone and cartilage. Totowa (NJ): Humana Press Inc.; 2003. 27. Reznikov N, Almany-Magal R, Shahar R, Weiner S. Threedimensional imaging of collagen fibril organization in rat circumferential lamellar bone using a dual beam electron microscope reveals ordered and disordered sub-lamellar structures. Bone 2013; 52:676–83. 28. Almany Magal R, Reznikov N, Shahar R, Weiner S. Threedimensional structure of minipig fibrolamellar bone: adaptation to axial loading. J Struct Biol 2014;186:253–64. 29. Alexander B, Daulton TL, Genin GM, Lipner J, Pasteris JD, Wopenka B, Thomopoulos S. The nanometre-scale physiology of bone: steric modelling and scanning transmission electron microscopy of collagen–mineral structure. J Royal Soc Interface 2012;9: 1774–86. 30. Erickson B, Fang M, Wallace JM, Orr BG, Les CM, Banaszak Holl MM. Nanoscale structure of type I collagen fibrils: auantitative measurement of D-spacing. Biotechnol J 2013;8:117–26. 31. Quan BD, Sone ED. Cryo-TEM analysis of collagen fibrillar structure. Meth Enzymol 2012;532:189–205. 32. Karnovsky M. A formaldehyde-glutaraldehyde fixative of high osmolarity for use in electron microscopy. J Cell Biol 1965;27: 137Az8A. 33. Yeni YN, Schaffler MB, Gibson G, Fyhrie DP. Prestress due to dimensional changes caused by demineralization: a potential mechanism for microcracking in bone. Ann Biomed Eng 2002; 30:217–25. 34. Fang M, Goldstein EL, Turner AS, Les CM, Orr BG, Fisher GJ, Welch KB, Rothman ED, Banaszak Holl MM. Type I collagen Dspacing in fibril bundles of dermis, tendon, and bone: bridging between nano-and micro-level tissue hierarchy. ACS Nano 2012;6: 9503–14. 35. Almer J, Stock S. Internal strains and stresses measured in cortical bone via high-energy X-ray diffraction. J Struct Biol 2005;152: 14–27. 36. Bonar LC, Lees S, Mook H. Neutron diffraction studies of collagen in fully mineralized bone. J Molec Biol 1985;181:265–70.

Copyright of Connective Tissue Research is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.