Ultrashort echo time T2 ⁄ values decrease in tendons with application of static tensile loads

In early stages of tendon disease, mechanical properties may become altered prior to changes in morpho- logical anatomy. Ultrashort echo time (UTE) magnetic resonance imaging (MRI) can be used to directly detect signal from tissues with very short T2 values, including unique viscoelastic tissues such as ten- dons. The purpose of this study was to use UTE sequences to measure T2 ⁄ , T1 and magnetization transfer ratio (MTR) variations of tendon samples under static tensile loads. Six human peroneal tendons were imaged before and under static loading using UTE sequences on a clinical 3 T MRI scanner. Tendons were divided into two static tensile loading groups: group A that underwent one-step loading (15 N) and group B that underwent two-step loading (15 and 30 N). The T2 ⁄ , T1 and MTR variations were investigated in two selected section regions of interest (ROIs), including whole and core sections. Mean T2 ⁄ values for the ﬁrst step of loading (groups A and B) in both whole section and core section ROIs were signiﬁcantly decreased by 13 ± 7% (P = 0.028) and 16 ± 5% (P = 0.017), respectively. For the second loading step (group B), there was a consistent, but non-signiﬁcant reduction in T2 ⁄ value by 9 ± 2% (P = 0.059) and 7 ± 5% (P = 0.121) for whole and core sections, respectively. Mean T1 did not show any consistent changes for either loading steps (P > 0.05). Mean MTR increased slightly, but not signiﬁcantly for both loading steps (P > 0.05). Signiﬁcant differences were found only in T2 ⁄ values of tendons by static tensile load applica- tion. Therefore, T2 ⁄ monitoring during loading is suggested for quantitative investigation of the tendons biomechanics.


Introduction
Tendons are unique viscoelastic connective tissues that are responsible for transferring the mechanical loads generated by muscles to the bones (Aparecida de Aro et al., 2012;Franchi et al., 2007b). Water, as the main viscous component (Wellen et al., 2005) of tissues, may comprise over 60% of the total tendon weight (Aparecida de Aro et al., 2012)). Of the dry weight of the tendon, 60-85% is composed of highly organized collagenous fibers, particularly collagen type I (Aparecida de Aro et al., 2012). Collagenous fibers count for the elastic component of tendons (Wellen et al., 2005). The mechanical properties and great strength of tendons can be partially explained by their highly organized and hierarchical structure, which consist of collagen fibrils, fibers, bun-dles of fibers and fascicles (Aparecida de Aro et al., 2012;Khan et al., 1999). Tendons demonstrate crimp or zigzag patterns in their microstructure that enable them to withstand, up to 3% strain without damage (Aparecida de Aro et al., 2012;Hansen et al., 2002;Khan et al., 1999).
Early stage diagnosis of tendon diseases is of great interest to the musculoskeletal research community. Different diseases are hypothesized to alter the mechanical performance of tissues earlier than changes in their anatomical morphology. Hence, researchers have attempted to monitor the changes in MR imaging biomarkers after tensile loading in tendon to assess the sensitivity to mechanical alterations (Han et al., 2000;Helmer et al., 2006Helmer et al., , 2004Koff et al., 2014;Mountain et al., 2011;Shao et al., 2016;Syha et al., 2014;Wellen et al., 2005Wellen et al., , 2004, however, only a few of them have used UTE-MRI (Koff et al., 2014;Shao et al., 2016;Syha et al., 2014).
During tendon stretching, two main phenomena occur; first, the number of crimps decrease and the mean crimp angle flattens (Franchi et al., 2007a(Franchi et al., , 2007bLegerlotz et al., 2014;Morgan et al., 2006;Mountain et al., 2011), and second, free water redistributes within Wellen et al., 2005Wellen et al., , 2004 and outside the tissue (Han et al., 2000;Hannafin and Arnoczky, 1994). In addition, the diameter of larger collagen fibrils decreases and less interspace areas can be found in loaded tendons (Morgan et al., 2006). Therefore, MRI properties of tendons may be expected to change under mechanical loading, even though reported changes in the literature have been inconsistent and are not well understood.
Among the few reported UTE-MRI based studies of tendons under the mechanical loading, Syha et al. observed significant increase of UTE-MTR in vivo (i.e., off-resonance radiofrequency saturation ratio, ORS) for tendons, after short-term exercises (Syha et al., 2014). Chang et al. (2015a) also observed a significant increase in UTE-MTR in human tendon under static tensile loads (5-10 N). Koff et al. used UTE-MRI for imaging of cyclically loaded (100 cycles of 45 N) rabbit patellar tendon (Koff et al., 2014). The tendons presented significantly shorter T2 ⁄ values after cyclic loading. In our previous study, (Chang et al., 2015c), UTE-MRI was used for assessing statically loaded human tendons by 10 N loads. No significant changes in UTE biomarkers were observed using single-and bi-component T2 ⁄ analyses. The applied load in our previous study (Chang et al., 2015c), may have been too low to exert significant changes in the tendons, similar to the fact that  did not observe changes in rabbit Achilles tendons when only 5 N tensile load was applied.
The main objective of this study was to determine whether the application of higher static tensile loads (>10 N) on human tendons alter T2 ⁄ , T1, and MTR as measured using UTE sequences.

Sample preparation
Six peroneal tendon samples (longus and brevis) were harvested from the ankles of three donors (all female; 38-, 40-and 95-year-old). Tendon samples were cut into 6 cm length and then distributed into two groups. One sample from each donor (longus or brevis) was selected randomly for group A to undergo a onestep loading protocol (15 N). The three remaining samples were assigned to group B and underwent a two-step loading protocol (15 and 30 N). Tendon samples were subjected to two cycles of freezing and thawing. All samples were immersed in saline for 2 h at 4°C and gently mounted into a plastic loading device fabricated inhouse. During the mounting and installing process, the samples were kept hydrated using saline. The loading device chamber was filled with perfluoropolyether (Fomblin, Ausimont, Thorofare, NJ) to minimize dehydration and susceptibility artifact.

Mechanical loading
An MRI compatible loading device was fabricated using a plastic spring (LL075100U40G, Lee Spring, Brooklyn, NY). The spring was compressed by a manually driven screw, mounted on a notched standard 1 Â 6-inch polyvinyl chloride (PVC) pipe (Fig. 1a). Two sides of the tendon specimens were clamped in ''fixed" and ''sliding" grippers placed on the notched pipe. The two ends of the tendon were fixed firmly in the grippers using two sets of 1/4-inch Nylon bolts and nuts (Fig. 1b). The applied tensile mechanical load was calculated based on the amount of spring deformation and its stiffness. The stiffness of springs was calculated by the manufacturer using a load tester machine (UT 11, Larson Systems Inc., Minneapolis, MN, USA) after four loading sessions from 0 to maximum compression (i.e., 2.5, 5.0, 7.5, 10.0, and 15 mm). This was aimed to reduce the initial plastic creep, to achieve an accurate stiffness value and, more importantly, to reduce stiffness changes along the loading experiment in MRI. Along the four loading sessions, the average calculated stiffness converged from 2.10 to 1.98 N/mm.
For the scans during the unloaded stage, the screw was driven gently until the beginning of spring deformation (1 N approx.) to make sure that the tendon samples were straight. For the scans at loading stage, the load was adjusted after 2 min approximately prior to scanning to compensate the limited relaxation of the tendon.

Data analysis
For each dataset, two regions of interest (ROIs) were selected covering the whole and core regions of the tendons (Fig. 2). Single-component fitting models were utilized for T2 ⁄ decay (SðTEÞ / expðÀTE=T2 Ã Þ þ constant)) and T1 recovery (SðTRÞ / 1 À expðÀTR=T1Þ þ constant) analyses. The MTR for each off-resonance frequency was calculated by the following equation: MTR ¼ ðS0 À SSAT Þ=S0, where S 0 denotes the mean signal intensity acquired without a saturation pulse and S SAT denotes the mean signal intensity acquired with the saturation pulse. All analyses were performed using an in-house developed code in MATLAB (Mathworks, Natick, MA, USA). The T2 ⁄ , T1 and MTR values as well as their pixel maps in loaded tendon samples were compared with the same samples before loading.

Statistical analysis
All statistical analyses were performed using a statistical programming language (R, version 3.2.5, R Development Core Team, Vienna, Austria). The average values of T2 ⁄ , T1 and MTR in unloaded and loaded tendons were compared using the two-tailed paired t-student test. P-values below 0.05 were considered significant.   however the T2 ⁄ reduction was less after the second step of loading (30 N).

Results
The estimated T2 ⁄ of all examined tendons at different loading steps, based on the single component fitting analyses within the whole and the core sections, are presented in Table 1. For all samples in group A and group B, the T2 ⁄ values were reduced by applying mechanical loads. Table 2 presents the estimated T1 values, based on the single component fitting analyses within the whole and the core sections of the tendons, at different loading steps. No clear trend for T1 value change was observed, by applying mechanical tensile load. However, for five of the scanned samples, T1 values were increased at the first loading step.
MTR of all the scanned tendon samples (groups A and B) at 2, 5 and 10 kHz off-resonance frequencies for 1500°MT pulse, are presented in Table 3. The MTR change did not show a consistent trend for all scanned tendon samples. However, for most of the cases there was an increase in MTR.  The summary of average changes in T2 ⁄ , T1 and MTR values within both the ROIs and for both the loading steps are presented in Table 4. The statistical analyses revealed that the mean T2 ⁄ values for the first step of loading, in the whole and the core sections, were decreased significantly by 13 ± 7% (P = 0.028) and 16 ± 5% (P = 0.017), respectively. For the second loading step, changes in T2 ⁄ values were not statistically significant, but showed a decreasing trend (9 ± 2% (P = 0.059) and 7 ± 5% (P = 0.121) for whole and core sections, respectively).

Discussion
Human cadaveric peroneal tendons were studied using UTE-MRI under tensile static loading. This study follows prior works using UTE-MRI for assessing loaded tendons (Chang et al., 2015a(Chang et al., , 2015cKoff et al., 2014;Syha et al., 2014). Monitoring the UTE-MRI based biomarkers of tendons under tensile load application may provide an index for the tendon's mechanical performance, which in turn can be used for early stage diagnosis of tendon diseases.
UTE-MTR and T2 ⁄ were the two studied properties in the literature (Chang et al., 2015a(Chang et al., , 2015cKoff et al., 2014;Syha et al., 2014). UTE-MTR was found to be increasing in vivo after shortterm exercises (Syha et al., 2014) and ex vivo under static tensile loads (5-10 N) (Chang et al., 2015a). Koff et al. observed T2 ⁄ decreasing in tendons after cyclic loading (100 cycles of 45 N) (Koff et al., 2014). Nevertheless, Chang et al. (2015c) did not observe any significant change in T2 ⁄ of tendons under tensile static loads (10 N). This current study helps to realize the impacts of the load magnitude on the tendon MRI-based properties. It was illustrated that the level of changes in the UTE-MRI properties is a function of the applied load magnitude. Specifically, the level of changes in T2 ⁄ was higher for 30 N load compared with 15 N load. It is likely that the 10 N load applied in a previous study (Chang et al., 2015c) was not enough to exert any changes in T2 ⁄ .
The fabricated MRI-compatible loading device with plastic grippers provided a method to apply higher loads compared with our previous study (Chang et al., 2015c). Specifically, by eliminating the need for sutures, we were able to apply > 10 N to the mounted tendons. To the authors' knowledge, the applied loads on the peroneal tendons during daily activities have not been reported in the literature. However, for Achilles tendons, the estimated load peak, during recreational running, may reach up to several times of body weight (Andrew and Jonathan, 2014). This is much higher than applied loads in this study. The loading device employed plastic springs that ended in a very compact and sealed unit which was compatible with a wrist coil, without requiring strings and suspended loads (Chang et al., 2015c;Helmer et al., 2004;Wellen et al., 2005Wellen et al., , 2004. However, the plastic spring may have a limited creep and stiffness reduction along the loading process in MRI. To reduce the stiffness changes along the loading experiment, springs underwent four loading sessions from 0 to maximum compression before scanning. Along the four loading sessions, the average calculated stiffness reduced up to 6% and converged to 1.98 N/mm. Applying the first step static tensile load (15 N) resulted in significant reduction of mean T2 ⁄ in all scanned tendon samples for both whole and core sections (13 ± 7% (P = 0.028) and 16 ± 5% (P = 0.017), respectively, Tables 1 and 4). For the second loading step (30 N), the T2 ⁄ values trended down, but this did not reach statistical significance (Tables 1 and 4). One possible explanation is that the second loading step was performed after an hour of loading by 15 N. Such preconditioning may have prevented the tendons from demonstrating significant changes at the subsequent 30 N load application. T2 ⁄ reductions were consistent with the observations from Koff et al. (2014) in rabbit patellar tendon after cyclically loaded by 45 N. Therefore, both static and cyclic loadings resulted in similar changes in T2 ⁄ . The applied load in our previous study (10 N) and the loading set up (suturing strips passing through the syringe and connecting to suspended weights), were likely limited in their ability to exert significant changes in tendons. Likewise, Helmer et al. (2004) did not observe changes in rabbit Achilles when 5 N tensile load was applied.
Of the biomarkers evaluated in our study, T2 ⁄ was the only one that consistently demonstrated significant changes by applying mechanical loads for all tendon samples. Mean T1 value variations were not statistically significant for both of the loading steps and within both selected ROIs (Table 1 and Table 4). Mean MTR values were increased slightly by loading within both selected ROIs, however, the differences were not statistically significant (Tables 3 and  4).
T2 ⁄ maps demonstrated a shrinkage of high-T2 ⁄ regions by the application of tensile load for all scanned tendons (Fig. 3). High-T2 ⁄ region shrinkage was observed in both core and rim sections of the tendon samples. However, more intense changes were found in the core sections (Fig. 3). Such T2 ⁄ reductions and shifts were expected due to, first, free water movement from denser regions of tendons towards less dense regions and possibly to both ends of tendon mounted in clamps, and second, reduction of crimp number as well as flattening of crimp angles. Specifically, such crimp reductions and flattening result in an angle decrease between collagen fiber axes and static magnetic field (B0), which in turn leads to lower T2 ⁄ values. The lowest and highest values of T2 ⁄ could be found at 0°and 55°angles with respect to B0 (Chang et al., 2015 b;Mengiardi et al., 2006;Mountain et al., 2011). The T2 ⁄ map shifts were not perfectly homogenous, probably due to uneven mechanical stress distribution in tendon cross sections (Slane and Thelen, 2014). Interestingly, the loading led to a slight increase of MTR in the sites of lower T2 ⁄ values (Figs. 3 and 4) which agrees with MTR increases, observed by Syha et al. (2014) in vivo and by Chang et al. (2015a) in vitro. Nevertheless, the MTR increase was obvious only in few samples of this study such as sample B-1 (Fig. 4).
Despite the UTE-MRI techniques, the cilinical MRI has been used very often for loaded tendon studies. Helmer et al. (2004) investigated tendon proton density to monitor intra-tendinous water movement after mechanical loading. Proton density demonstrated an increase in the rim sections of tendons while a decrease was seen in the core sections with 10 N loads, indicating outward water shifts . Remarkably, for a lower tensile load (5 N), no change was observed in the core section. Wellen et al. (2005) studied T2 and T1 of tendons under loading. T2 increased in the rim but decreased in the core sections of tendon after loading. T1 decreased in both rim and core regions (Wellen et al., 2005). Repeated tensile loads after a period of unloading resulted in similar changes in T2 (Helmer et al., 2006). Mountain et al. (2011) also reported a significant T2 increase in tendons after static tensile loading, however the tendons were fixed in formalin after loading. They also examined the tendons in different orientation angles with respect to B0 direction. The T2 values were higher in loaded tendons for all orientation angles (Mountain et al., 2011). Apparent diffusion coefficient (ADC) was also invesitaed in other studies (Han et al., 2000;Wellen et al., 2004). A significant increase in ADC was observed and interpreted as a sign of water reduction in the tendons (Han et al., 2000), particularly in the tendon core .
This study has several limitations. First, a small number of samples were included. As is the nature of pilot studies, our results require a validation with larger sampling sizes. Furthermore, our tendon specimens were obtained from a wide span in age of subjects and future studies should evaluate the possible effects of age. Second, the T2 ⁄ calculations were performed using singlecomponent exponential models. Future studies will include bicomponent models to investigate whether the T2 ⁄ reduction occurrs in the short or long T2 ⁄ components of the tissue. It should be noted that the bi-component models require more data points with higher signal-to-noise ratio in order to ensure appropriate fitting. Third, the tendon samples were scanned only parallel to B0 direction. Although the T2 ⁄ reductions can be explained by changes related to water movement, the reduction and flattening of crimps may also play a role. Future studies evaluating T2 ⁄ may incorporate other tendon orientations, such as scanning at the magic angle (55°), for a comprehensive understanding of UTE-MRI changes of tendon under loading. Moreover, future studies may consider incorporating biomarkers that are known to be less sensitive to orientation angle, such as those generated from UTE-MT with twopool modeling (Ma et al., 2016(Ma et al., , 2017. Fourth, the employed plastic springs may demonstrate limited creep during the loading process and MRI scanning. Future studies should employ active MRIcompatible actuators (e.g., pneumatic actuators) to achieve higher accuracy and reliability. Finally, the degree of change in loaded healthy versus pathologic tendons would be of interest in future studies.

Conclusion
This study investigates the mechanical status of tendons through observations in UTE-MRI biomarkers. T2 ⁄ , T1, and MTR values were measured for human peroneal tendons before and under static tensile mechanical loads. Among examined MRI properties, only T2 ⁄ demonstrated significant changes under mechanical loading (over 12% reduction). T2 ⁄ maps revealed a significant shrinkage of high-T2 ⁄ regions in tendon cross sections which might be caused by, first, intra-tendinous shifts in water and second, alterations in tendon crimp. T2 ⁄ monitoring over the course of loading can be suggested as a new technique for quantitative investigation of the tendons biomechanics.