Can ultrashort-TE (UTE) MRI sequences on a 3-T clinical scanner detect signal directly from collagen protons: freeze – dry and D 2 O exchange studies of cortical bone and Achilles tendon specimens

Ultrashort-TE (UTE) sequences can obtain signal directly from short- T 2 , collagen-rich tissues. It is generally accepted that bound and free water can be detected with UTE techniques, but the ability to detect protons directly on the collagen molecule remains controversial. In this study, we investigated the potential of UTE sequences on a 3-T clinical scanner to detect collagen protons via freeze – drying and D 2 O – H 2 O exchange studies. Experiments were performed on bovine cortical bone and human Achilles tendon specimens, which were either subject to freeze – drying for over 66h or D 2 O – H 2 O exchange for 6days. Specimens were imaged using two- and three-dimensional UTE with Cones trajectory techniques with a minimum TE of 8 μ s at 3T. UTE images before treatment showed high signal from all specimens with bi-component T 2 * behavior. Bovine cortical bone showed a shorter T 2 * component of 0.36ms and a longer T 2 * component of 2.30ms with fractions of 78.2% and 21.8% by volume, respectively. Achilles tendon showed a shorter T 2 * component of 1.22ms and a longer T 2 * component of 15.1ms with fractions of 81.1% and 18.9% by volume, respectively. Imaging after freeze – drying or D 2 O – H 2 O exchange resulted in either the absence or near-absence of signal. These results indicate that bound and free water are the sole sources of UTE signal in bovine cortical bone and human Achilles tendon samples on a clinical 3-T scanner. Protons on the native collagen molecule are not directly visible when imaged using UTE sequences. Copyright © 2016 John Wiley & Sons, Ltd.


.INTRODUCTION
Collagen is the most abundant protein in the human body and is the main component of connective tissue (1). The collagen molecule is composed of a three-stranded arrangement of α-helices which are stabilized by a ladder of hydrogen bonds, with additional support from stereoelectronic effects and post-translational modifications, such as hydroxylation and cross-linking (2). Alterations of collagen structure, amount and type are of clinical interest in nearly all biological systems, including musculoskeletal, neurological, cardiovascular and respiratory (3)(4)(5)(6).
Of the clinically utilized imaging modalities, MRI has the potential to provide the most information about collagen. However, the highly anisotropic structure of collagen causes short transverse relaxation times, which result in little or no signal when imaged using conventional MR sequences. Ultrashort-TE (UTE) sequences, with TEs that range from 0.008 to 0.50 ms, have been increasingly used on clinical MRI systems to directly image collagen-rich short-T 2 tissues or tissue components, such as cortical bone, calcified cartilage, menisci, ligaments and tendons (3,7,8).
Although UTE sequences can obtain signal directly from collagen-rich tissues, the precise origin of this signal remains controversial. Potential candidates include: (i) protons on the collagen molecule; (ii) collagen-associated water with various amounts of restricted motion (also known as 'bound' water); and (iii) unrestricted water surrounding collagen (also known as 'free' or 'bulk' water). With regard to restricted collagenassociated water, three distinct compartments have been identified, including water bridges (single and double bridges between α-helices too distant to allow direct hydrogen bonds), cleft water (residing in the groove-like depressions of the triple helix) and interfacial monolayer (also known as surface or hydration layer) water, in order of increasing mobility and energy (9,10). Using NMRI at different levels of hydration, Fullerton et al. (10) proposed a mono-exponential decay pattern of all three restricted water compartments due to fast exchange. Using clinical MR scanners, quantitative UTE techniques have been employed with bi-component T 2 * analysis to evaluate collagenassociated bound and free water compartments (11)(12)(13).
More recently, a study by Siu et al. (14) has suggested that UTE sequences can detect signal directly from protons on the collagen molecule, with a reported mean short T 2 * of 0.75 ± 0.05 ms and a mean chemical shift of -3.56 ± 0.01 ppm relative to water at 7 T. The purpose of this study is to further investigate whether UTE sequences can detect signal from collagen protons via freeze-dry and D 2 O exchange studies of cortical bone and Achilles tendon specimens at 3 T.

Sample preparation
Two bovine cortical bone samples (2 × 2 × 6 mm 3 ) and two cadaveric human Achilles tendon samples (1 cm in length) were prepared for this study. The bone samples were sectioned from two fresh femoral mid-shaft bovine specimens obtained from a local slaughterhouse using a low-speed diamond saw (Isomet 1000, Buehler, Lake Bluff, IL, USA) with constant water irrigation. The two tendon samples were sectioned from two human ankle specimens obtained from our institutional anatomical laboratory using a scalpel. All samples were stored in phosphate-buffered saline (PBS) solution for 24 h prior to use.

Pulse sequences and imaging protocol
All samples were imaged with two-dimensional (2D), nonslice-selective, UTE sequences and three-dimensional UTE with Cones trajectory (3D Cones) sequences on a GE 3-T Signa TwinSpeed MR scanner (GE Healthcare Technologies, Milwaukee, MI, USA) which had a maximum gradient strength of 40 mT/m and a maximum slew rate of 150 mT/m/ms. The non-slice-selective 2D UTE sequence employed a short rectangular pulse excitation (duration, 32 μs), followed by 2D radial ramp sampling with a minimal nominal TE of 8 μs. The 3D UTE Cones sequence employed a short rectangular pulse (duration, 32 μs) for nonslice-selective excitation, followed by 3D spiral sampling on the Cones sequence (15). A home-built 1-in-diameter birdcage T/R coil was used for signal excitation and reception for both 2D and 3D UTE imaging.
For 2D UTE morphological imaging, the following parameters were used: TR = 100 ms; flip angle, 10°; bandwidth, 62. for Achilles tendon. For T 2 * measurement, the image flip angle was low (i.e. 10°) and the pulse repetition time (i.e. 100 ms) was sufficiently long to eliminate the T 1 dependence, especially for cortical bone with a short T 1 (16,17). The total scan time for T 2 * measurement was 11 min and 25 s for both cortical bone and Achilles tendon.

Experimental procedures
The bovine cortical bone and human Achilles tendon samples were blot-dried and then imaged with the above morphological and quantitative UTE imaging protocols. After the initial 2D and 3D UTE imaging, one bovine bone sample and one human Achilles tendon sample were lyophilized using a Labconco Lyph-Lock 4.5-L freeze-dry system (model 77510-00, Labconco Corp., Kansas City, MO, USA) for over 66 h. The weight of each sample was measured before and after freeze-drying using a Mettler-Toledo AL-104 digital balance (Mettler-Toledo, Schwerzenbach, Switzerland) with a precision of 0.1 mg. After freeze-drying, each sample was stored in a sealed tube and warmed to room temperature, and then imaged again with the same protocol as described above.
The other bovine bone sample and the other human Achilles tendon sample were subjected to a D 2 O-H 2 O exchange study. Each sample was placed in a 20-mL syringe filled with D 2 O solution for exchange in the refrigerator for 6 days. Each syringe was flushed with fresh D 2 O every 2 days in order to achieve more complete D 2 O-H 2 O exchange. Finally, the same UTE MRI protocol was applied to the D 2 O-exchanged samples.

Data analysis
A bi-component model was used for the quantification of both bound and free water components, which assumes that both bound and free water contribute to the UTE signal. The following bi-component analysis model was employed to analyze the T 2 * values of both bound and free water and their relative fractions: where a constant noise term was fitted to account for background noise. S bw and S fw represent the magnetization of the bound and free water components, respectively, and T 2S * and T 2L * are their T 2 * relaxation times. The analysis algorithm was written in Matlab (The MathWorks Inc., Natick, MA, USA) and was executed offline on the DICOM images obtained by the protocols described above. The program allowed the placement of regions of interest (ROIs) on the first UTE image of the series, which was then copied onto each of the subsequent images. The mean intensity within each of the ROIs was used for subsequent curve fitting. The bi-component exponential fitting model shown in Equation [1] was used to fit the UTE T 2 * images. Noise was estimated automatically using a maximum likelihood estimation algorithm. were the experimental and fitted data points. For the bicomponent fitting of cortical bone and Achilles tendon, the residuals were 0.4% and 2.2%, respectively. The residuals of bi-component fitting were very low, which demonstrates that the bi-exponential model describes the observed signal decay well. Figures 3, 4 show 2D UTE and 3D Cones imaging of the bovine cortical bone sample and human Achilles tendon sample, respectively, before and after freeze-drying. High signal was observed for both samples before freeze-drying. The fascicular pattern in the blot-dried Achilles tendon was nicely depicted.
However, after freeze-drying, near-zero signal was observed for bovine cortical bone, and only a few bright spots were observed near the central region of the Achilles tendon sample. The bright spots are probably from residual water surviving the freezedrying of 66 h. Longer freeze-drying may be necessary to remove all detectable water in the Achilles tendon sample. These results suggest that freeze-drying removes the source of the MR signal, namely both bound water and free water. Collagen protons are expected to survive the freeze-drying process and, when imaged using 2D and 3D UTE sequences, show zero signal. Therefore, collagen protons in cortical bone and the Achilles tendon are 'invisible' with UTE sequences on clinical MR scanners. Figure 5 shows 2D UTE and 3D Cones images of the other Achilles tendon sample before and after D 2 O exchange. High signal was observed before D 2 O exchange, but the signal dropped to near zero after D 2 O exchange, confirming that the signal foci in the tendon images after freeze-drying were generated from residual water which survived the lyophilization process. Similar results were observed for cortical bone (not  shown) with high signal when imaged with 2D UTE and 3D Cones before D 2 O exchange, but near-zero signal after D 2 O exchange.
From the results, the water protons can be effectively removed from the cortical bone and Achilles tendon by both freeze-drying and D 2 O-H 2 O exchange. The cortical bone showed an absence of signal after both freeze-drying and D 2 O-H 2 O exchange. Although there was a small focus of signal intensity in the central region of the Achilles tendon after freeze-drying, the D 2 O exchange experiment confirmed that this focus of signal in the tendon images was generated from residual water after freeze-drying.

DISCUSSION
Short-T 2 structures that were previously invisible when imaged with conventional MR sequences, including bone and tendon, can now be visualized using UTE techniques. However, an understanding of the source of signal is important for image interpretation, particularly in the setting of in vivo translation, where MRI may be a surrogate for more invasive techniques. In our study, we have shown that bound and free water are the sources of UTE signal in bovine cortical bone and human Achilles tendon samples on a clinical 3-T scanner. Specifically, the absence of signal in both bone and tendon samples after freeze-drying indicates that protons on the native collagen molecule are not directly visible using UTE sequences.
Our results contrast with those recently published by Siu et al. (14), who suggested that collagen protons could be detected and quantified using UTE sequences at 7 T. However, there are a number of differences between our studies. Siu et al. (14) performed their experiments on a hydrolyzed collagen solution, which causes cleavage of the molecule into smaller peptides. This results in an amorphous state of collagen, lacking either partially or completely the tertiary, secondary and primary structure of native collagen (18). The destruction of the highly ordered, 3D structure of collagen would be expected to alter the T 2 * values of the collagen protons. Presumably, the collagen proton T 2 * values would increase, and this could result in their detection using UTE sequences, in a similar manner to macromolecular protons on myelin which may be detectable using UTE techniques (19,20). The effect of this on the stability and accuracy of bi-component fitting, which requires sufficient separation of the two modeled pools, is unclear (21,22). In addition, at higher field strengths, such as 7 T, oscillations in signals resulting from non-water offresonance spins further degrade bi-exponential fitting (23).
Direct detection of collagen protons is possible using spectroscopic techniques at high field strengths. This has been shown by Siu et al. (14) and  (24,25). Modeling of UTE-MT images acquired with a series of MT frequency offsets and power may provide quantitative assessment of protons in free water, bound water and collagen in short-T 2 tissues, such as cortical bone and the Achilles tendon. The precise relationship between these three proton pools in the native in vivo condition and with regard to healthy versus pathological tissue remains to be elucidated. UTE-MT imaging combined with bi-component analysis would provide information from all three pools and will be the focus of future work. The answer to the question of which proton pool is most sensitive and discriminative of pathological change would be of great clinical significance. Along the lines of clinical translation, our study utilized the 3D UTE Cones sequence with the correction of linear eddy current components rather than the 3D radial UTE sequence. Although the 3D radial UTE sequence may be more immune to hardware imperfections with a shorter readout window with a linear acquisition trajectory, the 3D UTE Cones sequence demonstrates higher scan time efficiency with a spiral acquisition trajectory. The 3D Cones sequence also allows anisotropic spatial encoding with lower slice resolution than in-plane resolution to achieve higher SNR and less scan time (26).
Our study has several limitations. First, freeze-drying is an elegant way to remove water whilst largely preserving structure, but does not result in a completely anhydrous state. Previous studies using collagen-rich tissues have shown residual water measuring less than 3% (27). This is confirmed by the central hyperintensities present in the Achilles tendon sample after lyophilization. However, this may be considered to be the most economic method achievable, whilst largely maintaining structure. Second, freeze-drying results in some minor degree of protein denaturation. The process of drying removes water bridges which play a role in stabilization (28). However, the expected change in structure is probably small (2) and we would not expect large changes in collagen proton T 2 * values. Third, D 2 O exchange for 6 days, even in the refrigerator, may lead to significant tissue degradation. Fourth, as a large amount of collagen protons are not exchangeable with D 2 O (e.g. -CH 2 -, -CH 3 ), the absence of any signal after D 2 O exchange probably remains a valid indicator that collagen protons are not directly visible with UTE sequences on clinical MR scanners. Fifth, both freeze-drying and D 2 O exchange may change the mobility of the collagen molecule, as it interacts closely with the surrounding water; in this case, the two treatments themselves may alter the T 2 * value of the proton signal specifically associated with collagen. Future high-field spectroscopic studies are needed on samples before and after freeze-drying and D 2 O flushing to show whether or not the collagen signals observable under these conditions are affected. Sixth, all samples in this study were maintained at room temperature during scanning. Although it is likely that quantitative relaxation measurements would be affected by differences between room and body temperature, we do not anticipate that the main results of our study, which show that protons on the native collagen molecule are invisible with UTE sequences, would be significantly affected. Finally, it would be of interest to investigate whether other imaging techniques, such as zeroecho-time (ZTE) imaging or sweep imaging with Fourier transformation (SWIFT), could conceivably be able to detect the signals of collagen protons directly (29,30).

CONCLUSIONS
We have demonstrated that bound and free water, but not collagen protons, are the sources of signal in cortical bone and Achilles tendon samples when imaged with UTE sequences on a clinical 3-T MR scanner.