Introduction

Stroke is a leading cause of disability in adults, with a total direct and indirect economic cost of over $40 billion in the US in 2013–20141. Spasticity and muscle stiffness are common sequelae following stroke, with reported prevalence of 19–92%2,3,4. Post-stroke spasticity and muscle stiffness severely affect the quality of life of individuals with stroke and their caregivers, and drive up direct healthcare costs four-fold during the first year alone5. In addition, muscle spasticity and stiffness are also problematic for patients with multiple sclerosis, traumatic brain injury, spinal cord injury and cerebral palsy6,7,8,9.

The generally accepted neural mechanism underlying spasticity is a decrease in suppression of the spinal stretch reflexes resulting from reduced cortical influence on the spinal cord, causing hyperreflexia and muscle over activity10. Following neural injury, patients with spasticity and muscle stiffness also have weak muscles11 and conventional treatment of muscle over activity can further increase muscle weakness12. Furthermore, while hyperreflexia develops relatively early after neurologic injury, muscle stiffness develops over the ensuing weeks and months13. This increases the resistance to movement making it more challenging to move and participate in therapy. Non-neural mechanisms have therefore been proposed to explain the increased resistance to movement which may result from secondary changes in the structure or function of skeletal muscles secondary to the neural injury14. However, it is not clear how muscle properties change as a result of the neural injury15. The hyaluronan hypothesis proposes that the accumulation of the hyaluronan, within the extra-cellular matrix (ECM) of muscles contributes to the development of muscle stiffness16. Hyaluronan is a high molecular weight GAG in the ECM, which acts as a lubricant to aid the sliding of muscle fibers and bundles during movement17. Histological studies using animal models have shown that immobilization of skeletal muscle leads to the accumulation of hyaluronan18. Following a stroke, atrophy of muscle fibers leads to an increase in the ECM relative to the muscle fibers19, perhaps due to the accumulation of fat or hyaluronan, but this has not been shown conclusively. However, a recent case series showed that off-label, intramuscular injections of the enzyme hyaluronidase which hydrolyzes hyaluronan, resulted in reduced muscle stiffness and increased the range of passive and active motion in the post-stroke arm with beneficial effects lasting over three months16,20.

We used T1ρ mapping to image GAG content in upper limb muscles. T1ρ contrast is an endogenous MRI contrast mechanism that refers to the spin lattice relaxation time constant in the rotating magnetic field, and measures the transverse magnetization decay in the presence of a spin-lock radiofrequency (RF) field21. T1ρ contrast is sensitive to lower energy interactions22 related to the chemical exchange between extra-cellular water and macromolecules, and is well-suited to characterize proteoglycans23. T1ρ mapping allows quantification of GAG content, and has been used to quantify proteoglycan content in cartilage23,24, muscle25,26,27 and intervertebral discs28,29.

The purpose of this study was to i) provide imaging evidence of increased GAG content in patients with post-stroke muscle stiffness; and ii) to determine the effect of hyaluronidase treatment on intramuscular GAG content.

Results

Figure 1(A) shows a schematic representation of the data processing pipeline for T1ρ mapping. The data from multiple spin lock times (2–55 ms) are used to generate the T1ρ maps. Data for each pixel across all spin lock times are modeled using a 3-parameter non-linear, mono-exponential decay, and the resulting T1ρ relaxation time for each pixel is displayed as a T1ρ map. Regions of interest (ROIs) for the biceps and triceps muscles are manually drawn (Fig. 1(B)).

Figure 1
figure 1

Schematic representation describing the T1ρ mapping process. (A) The T1ρ weighted images from all TSL durations are fit on a pixel by pixel basis using a 3 parameter fit mono-exponential model to fit the data in each voxel to produce a T1ρ map. (B) Shows typical manually drawn ROIs for biceps and triceps muscles.

Full size image

Controls vs patients

Figure 2(A) shows representative T1ρ maps for three slices in a control subject. Figure 2(B) shows three representative T1ρ map slices of the affected arm of a patient with post-stroke muscle stiffness before treatment. In the controls, the mean T1ρ relaxation time in the biceps was 26.7 ± 0.54 ms, and in the triceps it was 30.29 ± 2.23 ms. In the patients, prior to treatment, the mean T1ρ relaxation time in the biceps was 34.04 ± 4.39 ms and in the triceps it was 34.06 ± 4.62 ms. In the biceps, a significant difference was observed between the controls and the patients pre-injection (P = 0.006), while the triceps did not show a significant difference (P = 0.14).

Figure 2
figure 2

T1ρ maps for a representative control subject and patient (pre-treatment). (A) Shows T1ρ maps of 3 representative slices overlaid over anatomy in a control subject. (B) Shows T1ρ maps of three representative slices overlaid over anatomy in a patient with post-stroke muscle stiffness prior to hyaluronidase injection treatment.

Full size image

Patients-pre- vs post-treatment

Pre- vs post-treatment analysis used only the patients who came for post-treatment imaging (n = 3). Figure 3 shows representative T1ρ maps in the three patients before injection (A) and after the hyaluronidase injections (B). The slices shown in (A) and (B) correspond to approximately similar slice positions in the affected arm.

Figure 3
figure 3

Representative T1ρ maps before and after hyaluronidase injection treatment. (A) Show T1ρ maps of representative slices overlaid over anatomy from 3 patients with post-stroke muscle stiffness before injections were administered (B) Shows T1ρ maps of the same patients in (A) at approximately similar slice locations following hyaluronidase injection treatment. Note the difference in shape of the muscle before and after the injections.

Full size image

The mean T1ρ relaxation time before injection treatment in this cohort for the biceps was 35.48 ± 3.38 ms and for the triceps was 35.36 ± 4.93 ms. Following the hyaluronidase injections, the mean T1ρ relaxation time in the biceps decreased to 29.45 ± 1.23 ms and the mean value in the triceps decreased to 32.91 ± 4.9 ms. A significant difference was observed between pre- and post-injection conditions for the biceps (P = 0.077), but not for the triceps (P = 0.65). To account for the small number of patients in the pre- vs post injection analysis, we reduced the significance level for this comparison to 0.1.

Figure 4A shows the comparison between controls and pre-injection patients, and Fig. 4B shows the comparison between pre-and post-injection conditions. Table 1 summarizes the results in the biceps and triceps.

Figure 4
figure 4

Comparison of mean T1ρ values between groups. (A) Shows the comparison of mean T1ρ values in the biceps and triceps ROIs between controls and pre-injection patients (N = 5). The T1ρ values expressed as mean ± SD are shown. In the biceps ROI, there is a significant different in the mean T1ρ value (P = 0.005), while a significant difference in the triceps ROI is not observed (P = 0.14) (B) Shows the comparison of mean T1ρ values in the biceps and triceps ROIs between pre- vs post-injection patients expressed as mean ± SD values. For the biceps ROI, a significant difference was observed between pre- and post-injection patients (P = 0.077), but a significant difference was not observed for the triceps ROI between pre- and post-injection (P = 0.65). Significance level set to 0.1, owing to the lower sample size (N = 3).

Full size image
Table 1 Demographic Information.
Full size table

Dixon water-fat imaging

Figure 5 shows the results from the Dixon water-fat imaging. Figure 5(C) shows the Dixon water-fat imaging in a healthy volunteer showing the water image (a), fat image (b), water fraction (c) and fat fraction (d) expressed as a percent. Figure 5(D) shows the same results from the affected arm of a patient prior to treatment. Although both the biceps and triceps showed higher fat fractions in patients, a significant difference between the fat-fractions of the control subjects and patients was not observed in the biceps (P = 0.063), and triceps (P = 0.19) (Table 2).

Figure 5
figure 5

Assessment of fatty infiltration in upper arm muscles using Dixon water-fat imaging. (A) Shows the T1ρ map of a healthy control subject, (B) shows the T1ρ map of patient before treatment (C) shows the calculated results from the Dixon water-fat imaging showing the water image, fat image, water fraction and fat fraction in the control subject in A and in the patient in B before treatment (D). Although the triceps and biceps ROIs showed increases in fat fraction in the skeletal muscle of patients, the differences were not significant compared to controls (P = 0.063 for biceps, P = 0.19 for triceps) suggesting that fatty infiltration did not play a big role in the elevation of T1ρ values.

Full size image
Table 2 Table summarizing data analysis for T1ρ imaging and Dixon water-fat imaging in the control subjects and patients.
Full size table

Model of evolution of muscle biochemical and structural changes post-stroke

Figure 6 shows a schematic representation of the evolution of the arm muscle micro-environment post-stroke. In normal skeletal muscle, the ECM consists of a thin layer of collagen fibers types I and III, which form the endomysium surrounding each muscle fiber, the perimysium around muscle fiber bundles, and the epimysium surrounding the entire muscle. A thin layer of hyaluronan is present adjacent to the collagenous endomysium, perimysium and epimysium17, and its thickness is continuously remodeled in response to mechanical stimuli (Fig. 6A). Following neural injury, inflammation, skeletal muscle paralysis and immobility can lead to an imbalance in the turnover of the ECM resulting in increased hyaluronan production and reduced degradation, and the accumulation of hyaluronan in the ECM; this can increase the viscosity of the ECM causing the muscle fibers to stick to one another which can increase the resistance to movement (Fig. 6B). This is the stage of muscle stiffness. At this stage, the collagen content of the endomysium and perimysium may not be dramatically different from normal and disuse may lead to minor muscle atrophy. Continued inflammation, immobility and/or disuse of muscles due to reduced cortical input and muscle stiffness can initiate a complex pathologic pathway involving transforming growth factor (TGF-β), Rho associated protein kinase (ROCK) and mitogen associated protein kinase (MAPK) resulting in fibrogenesis30,31,32. The excess hyaluronan in the ECM is then replaced by collagen, leading to permanent and irreversible thickening of the endomysium and perimysium, and alteration in the structure and function of the muscle14 manifested as contracture (Fig. 6C). This stage may be characterized by marked muscle atrophy.

Figure 6
figure 6

Model of the evolution of muscle biochemical/structural changes post stroke. (A) Shows a cross-section of a normal skeletal muscle, (B) shows the evolution of muscle stiffness following post-stroke paralysis and immobility that results in accumulation of hyaluronan in the extracellular space, increasing the viscosity of the ECM, and causing the muscle fibers to stick together (C) shows that continued inflammation, immobility and muscle stiffness initiates fibrosis leading to the replacement of hyaluronan by collagen, thickened endomysium and perimysium, and muscle fiber atrophy. Note that while (C) depicting fibrosis is an irreversible stage, data from literature suggests that the processes between stages depicted in (A) and (B) are reversible and may represent a potential therapeutic target.

Full size image

Discussion

In this study, we demonstrate that T1ρ MR imaging can be employed to assess the GAG content in muscles post stroke in comparison to control subjects. The study characterized the effect of hyaluronidase injections post stroke, and provides the first in vivo imaging evidence for the hyaluronan hypothesis of muscle stiffness.

This study used T1ρ imaging to quantify the amount of GAG in the upper arm muscles. The main advantages of using this T1ρ imaging technique are that it is a completely non-invasive method which does not employ exogenous contrast agents, thus making it safe and well-suited to image patients with stroke, with the added advantage that repeated imaging can be conducted at multiple time points for longitudinal monitoring. Furthermore, T1ρ imaging is sensitive to the chemical exchange of large macromolecules such as hyaluronan with protons in bulk water making it a useful tool to quantify GAG content in muscles before and after treatment. Importantly, it can shed light on the underlying mechanisms of the vexing problem of muscle stiffness after neural injury.

Hyaluronan is a non-sulfated negatively charged polysaccharide composed of repeating chains of two monosaccharides: ß(1,4)-N-acetyl-D-glucosamine and ß(1,3)-D-glucuronic acid, with a molecular formula of C33H54N2O23. T1ρ mapping reflects the chemical exchange of protons in water with the hydroxyl (-OH) groups of hyaluronan. In vivo, depending on the local environment, T1ρ contrast may depend on both exchangeable (chemical exchange rate, pH, concentration, f) and non-exchangeable (dipole-dipole and scalar interactions) mechanisms. In addition, inflammation may also affect T1ρ contrast. Supplementary Fig. 1 shows a schematic diagram of a two-pool model of exchangeable macromolecule proton pools with bulk water affecting T1ρ contrast. Each pool is characterized by its own properties (relaxation rate R1, R2), solute concentration (fB) and the chemical exchange between the pools (kBA, fB* kBA).

In vivo, increased T1ρ values reflect greater deposits of hyaluronan in the extra-cellular space of the muscles, noted in both the biceps and triceps post-stroke. The increased standard deviation of T1ρ values observed reflects the variability in patients compared to the control subjects. A number of factors may affect these values such as the concentration of hyaluronan, degree of muscle atrophy, presence of fibrosis, and duration of muscle stiffness in individual patients. In Fig. 3, along with reduced T1ρ values, we note the difference in the shape of the muscles among all patients in the pre-injection vs post-injection images. We attribute this to reduced viscosity in the ECM of the muscle after hyaluronidase injections and relaxation of the muscle as a result of reduced hyaluronan content. In Fig. 4, although there were reductions post-treatment in mean T1ρ numbers for both the biceps and triceps, a larger effect was observed in the biceps. Limited elbow extension post-stroke is attributed to stiffness in the biceps, which is typically treated more aggressively33. Our results demonstrate that hyaluronan accumulates in both the flexor and extensor compartments, and that both muscle groups may require treatment as visualized using imaging.

Intervention at the stage of muscle stiffness to reduce the imbalance in turnover of the ECM represents a potential therapeutic target. Local intramuscular injection of hyaluronidase into the affected muscles hydrolyzes the excess hyaluronan and reduces the viscoelasticity of the ECM in the affected muscle. Thus muscle stiffness may be reversible with the administration of hyaluronidase, but further studies are needed to investigate its therapeutic potential. The results shown here suggest that using T1ρ imaging as a tool to quantify GAG content and monitor treatment response shows significant potential to enhance the management of post-stroke muscle stiffness, which is a daunting and difficult problem to treat. The results also support the use of hyaluronidase in improving muscle characteristics on T1ρ imaging in patients with post-stroke muscle stiffness.

This study has the following limitations. The patient cohort used in this study is small. Further studies are underway to quantify GAG content in patients with post-stroke muscle stiffness which will include a larger diverse patient population to assess the robustness, repeatability and reproducibility of this technique. The imaging sequence used is not commonly available in all medical centers and may be seen as a barrier to entry for clinical adoption, and may require requesting the vendor for imaging sequences that can perform T1ρ imaging. Nevertheless, this imaging technique for the quantification of GAG can easily be translated to other disorders that result in muscle spasticity and stiffness to understand the non-neural contributions of changes in the ECM, such as in multiple sclerosis, traumatic brain injury, cerebral palsy, and spinal cord injury.

In summary, this study provides T1ρ MR imaging evidence of accumulation of GAGs, specifically hyaluronan, in the upper limb muscles of patients with post-stroke muscle stiffness.

Methods

Study design

This study was approved by New York University Langone Health’s institutional review board (IRB) and was health insurance portability and accountability act (HIPAA) compliant. This was a prospective, non-randomized imaging study to determine intramuscular GAG content using proton T1ρ relaxation mapping. All subjects gave written informed consent after explanation of the study and the protocol, as per the IRB guidelines.

Subjects

Five post-stroke patients with upper limb muscle stiffness were recruited for this study (2 males/3 females, age = 52 ± 5 years). The patients were English speaking, adults, with chronic post-stroke moderate-to-severe muscle stiffness (Table 1). The patients did not have any contraindication to MRI (MR non-compliant implants, pacemakers, or claustrophobia). Five healthy volunteers (1 male/4 females, age = 27 ± 2 years) with no known history of mental illness, or physical injury to the arms, without contraindication to MRI were also recruited.

Study protocol

For the patients a pre-injection MRI scan (N = 5) of the affected upper arm was performed. Then they received off-label intramuscular hyaluronidase injections (Hylenex, Halozyme Therapeutics, Inc) for clinical treatment of their muscle stiffness unrelated to the imaging research study. The intramuscular dose of hyaluronidase was similar to those reported in a previous article and included several muscles besides the imaged arm muscles16. Approximately 1–4 weeks following the injections, the patients were scanned again using the same MRI protocol. For post-injection imaging, only 3 of the 5 injected subjects were scanned. Only the subjects having pre- and post-injection MRI scans were included in the pre- Vs post-treatment analysis. Healthy volunteers were scanned once using the same MRI protocol as the patients.

Imaging protocol

All imaging scans were done on a clinical 3 T MRI scanner (Prisma, Siemens Healthineers, Erlangen, Germany). An 8-coil flexible receive array coil was wrapped around the arm. A 3D-turbo-FLASH (fast, low angle shot) MRI sequence with a customized T1ρ preparation module was used to enable varying spin lock durations (TSL). A paired self-compensated spin-lock pulse was used to minimize B0 and B1 variations34. The sequence parameters included FOV = 130 mm, matrix size = 256 × 64 × 64, TR = 1500 ms, resolution = 0.5 × 2 × 2 mm2, spin-lock frequency = 500 Hz, 10 TSL durations = 2, 4, 6, 8, 10, 15, 25, 35, 45, 55 ms. acquisition duration = ~10 minutes (Fig. 2). The MRI body coil was used for transmission, and vendor supplied flexible receive array coils (8 coil elements each) were wrapped around the affected arm (right arm for controls) for imaging. To investigate if fatty infiltration in skeletal muscle played a role in patient results, Dixon based methods were used to separate fat and water distribution using the iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL)35 technique. The Dixon water/fat imaging parameters included: TR = 9.3 ms, TE = (2.26, 3.08, 3.90) ms, FOV = 180 mm, matrix size = 128 × 128, resolution = 1.4 × 1.4 mm2, total acquisition time ~ 4.5 min.

Data analysis and statistics

Manual ROIs were drawn for the biceps and triceps muscles. The mono-exponential T1ρ mapping was performed by fitting the signal intensity at different spin-lock durations for each pixel using a three-parameter non-linear mono-exponential model:

$$S=A\cdot {e}^{-TSL/{T}_{1\rho }}+{A}_{0}$$
(1)

where S is the signal intensity, A is the amplitude, TSL is the spin lock duration, T1ρ is the mono-exponential relaxation time in the rotating frame, and A0 is the average noise level.

For the Dixon water-fat imaging, IDEAL post-processing was used to separate the water and fat images from the 3-echo images. Water-fractions and fat-fractions were calculated from the water and fat images. The fat-fraction maps were used to quantify the fat infiltration in the muscles.

The mean values and standard deviations in the manually drawn ROIs were calculated across the patients and controls. Wilcoxon Mann-Whitney test was used for comparing the pre- vs post-treatment conditions in patients’ vs controls. A P-value of less than 0.05 was used for analysis as the threshold to reject the null hypothesis. For the comparison between pre and post-injection patients, a P-value of 0.1 was used due to the small number of subjects.