Ischemic Stroke Reduces Bone Perfusion and Alters Osteovascular Structure

Rationale Stroke patients lose bone mass and experience fracture at an elevated rate. Although functional intraosseous vasculature is necessary for skeletal maintenance, the effect of stroke on osteovasculature is unknown. Objective To characterize changes to osteovascular function, structure, and composition following mild-to-moderate-severity ischemic stroke in mice, both with and without exercise therapy. Methods and Results Twelve-week-old male mice (n=27) received either a stroke (middle cerebral artery occlusion) or sham procedure, followed by four weeks of daily treadmill or sedentary activity. Intraosseous perfusion, measured weekly in the proximal tibial metaphysis, was reduced by stroke for two weeks. In the second week of recovery, exercise nearly restored perfusion to sham levels, and perfusion tended to be lower in the stroke-affected limb. At the conclusion of the study, osteovascular structure was assessed with contrast-enhanced computed tomography in the distal femoral metaphysis. Stroke significantly increased osteovascular volume and branching but reduced the relative number of blood vessels close to bone surfaces (6-22 μm away) and increased the relative number more than 52 μm away. These differences in vessel proximity to bone were driven by changes in the stroke-exercise group, indicating compounded effects of stroke and exercise. Exercise, but not stroke, nearly reduced the amount of osteogenic Type H blood vessels in the proximal tibial metaphysis, quantified with immunofluorescence microscopy. Conclusions This study is the first to examine the effects of stroke on osteovasculature. Stroke increased the amount of osteovasculature, but since blood vessels close to bone are associated with bone remodeling, the shift in osteovascular structure could play a role in bone loss following stroke. The exercise-induced reduction in the amount of Type H vessels and the stroke-exercise effect on osteovascular structure suggest moderate aerobic activity may have detrimental effects on bone remodeling during early stroke recovery.


Introduction
sham procedure. After acclimation, exercise groups performed a standard moderate treadmill   [23,34]. Blood vessels that are positive for both EMCN and CD31 (defined as Type H) 173 have been shown to regulate bone-vascular crosstalk and couple osteo-and angiogenesis in long bones [24]. Samples were prepared and labeled using a previously described procedure [35], 175 producing 50-µm-thick longitudinal sections of the proximal tibial metaphysis labeled for EMCN 176 and CD31. Nuclear staining was performed with DAPI. 177 Immediately following staining, sections were imaged at 20X on a Zeiss Laser Scanning   For analysis #1, LDF measures of perfusion were compared between surgery and activity 195 groups at each timepoint and limb using a mixed hierarchical linear model (procedure GLIMMIX) 196 with interaction between all terms [36]. LDF was measured across four timepoints (Weeks 1-4) on each limb of each mouse. Effect differences between surgery and activity groups were compared 198 within each timepoint (i.e., stroke-exercise vs. stroke-sedentary at Week 2) using least squares 199 means and Tukey-Kramer adjustments for multiple comparisons. 200 For analysis #2, the stability of the tibial perfusion measurements throughout the surgery 201 was examined using the slope of the 30-minute LDF measurement taken during the surgical 202 procedure. To determine if the slope for each group was different from zero, which would indicate 203 that the perfusion measurement varied and was not stable, an unpaired t-test was performed for the 204 sham group, and a Wilcoxon rank sum test was performed for the stroke group. Slopes were also 205 compared between the stroke and sham groups with a Wilcoxon two-sample test.

206
For analysis #3, distribution data from CE-CT (vessel thickness, blood vessel surface to 207 bone surface distance) were compared between surgery and activity groups using a mixed 208 hierarchical linear model (procedure GLIMMIX) with interaction between terms. Effect 209 differences were compared between groups within histogram bins (i.e., stroke-exercise vs. stroke- average Ves.Th, Ves.D, branches, junctions, triple points, and quadruple points) were compared 213 between surgery and activity groups using a repeated measures factorial model (procedure 214 MIXED) but without the 'bin' repeated factor. Immunofluorescence parameters were compared 215 between surgery and activity groups using a standard two-factor general linear model (procedure 216 GLM) with interaction and Tukey adjustments for multiple comparisons.

217
For analysis #4, the same repeated measures models from analyses 1 and 3 were used, but 218 least squares means were compared between limbs within surgery group (i.e., affected vs. 219 unaffected side within the stroke group).
Immunofluorescence and LDF data during MCAo are presented as mean ± standard 221 deviation. Data analyzed using least squares means (repeated LDF measures, CE-CT data) are 222 presented as least squares mean ± 95% confidence interval. Repeated LDF and CE-CT data are 223 presented as the least squares mean for both limbs per mouse unless otherwise noted.

226
Bone Perfusion 227 Tibial perfusion was reduced in both limbs for two weeks following stroke, with 23% lower 228 perfusion in the stroke groups relative to the sham groups at Week 1 (p = 0.0064) and 24% lower 229 perfusion at Week 2 (p = 0.0061), but perfusion levels were similar between surgery groups at 230 Week 3 (p = 0.71) and Week 4 (p = 0.60) (Fig. 2B). Perfusion was nearly greater in exercise 231 groups, relative to sedentary groups, at Week 2 (21%, main effect p = 0.050), primarily driven by 232 the nearly increased perfusion in stroke-exercise (30% relative to stroke-sedentary, p = 0.054), not  During the surgeries (occluded for stroke, not occluded for sham), tibial perfusion 240 measurements remained stable throughout, with slopes that were not significantly different from 241 zero for either stroke (p = 0.39) or sham (p = 0.14). The slopes were similar between stroke (-242 0.0063 ± 0.0415 PU/min) and sham (0.0049 ± 0.0106 PU/min) groups (p = 0.98). These results demonstrate that cerebral ischemia does not directly impact blood supply to the tibia during the 244 occlusion, suggesting that any changes to osteovasculature during the recovery period result from 245 more systemic effects.   Adipocyte microstructure was not significantly affected by stroke or exercise, with similar 277 values for adipocyte volume fraction, density, and thickness across groups (Table I)

290
This study is the first to report the effects of ischemic stroke on osteovasculature. Middle 291 cerebral artery occlusion in mice was associated with increased blood vessel volume within the 292 distal femoral metaphysis, primarily due to increased vascular branching and a greater number of 293 larger vessels. These changes followed reduced perfusion in the proximal tibia for two weeks 294 following stroke, with relatively lower blood supply to the affected than unaffected bone in stroke 295 groups but not sham groups. Furthermore, the effects of stroke and exercise on the proximity of 296 vessels to bone surfaces were driven primarily by changes in the stroke-exercise group, yielding   In this study, stroke decreased functional blood supply within the proximal tibia for two may be a compensatory response to offset perfusion deficits. However, a study examining the 327 progression of these changes over time, and at the same site, is needed to assess the relative timing.

328
Ischemic stroke is associated with many conditions that could contribute to the reduced 329 intraosseous perfusion observed, including increased vascular tone, increased vascular resistance, 330 or decreased vasodilation. Vascular tone, which constricts blood vessels and decreases perfusion, 331 was higher in cutaneous blood vessels in the paretic hands of patients with ischemic lesions [40] 332 and in denervated arteries in rabbit ears [41], suggesting that stroke-related damage to the central 333 nervous system may contribute to the reduced intraosseous perfusion. Stroke has been associated 334 with increased vascular elasticity (a measure of vascular resistance) in the forearm [15], and resistance arteries in cutaneous blood vessels in the paretic arms of stroke patients were also less 336 responsive to exogenous acetylcholine-induced vasodilation [42], which would increase vascular 337 resistance. However, the effect of stroke on the arteries that supply bone are unknown, and future 338 studies are required to determine whether increased vessel volume and branching are 339 compensatory for changes to vascular tone, resistance, or sensitivity to vasodilators following 340 stroke within long bones. Exercise restored the stroke-related perfusion deficits by the second 341 week of recovery but also reduced osteovascular volume relative to stroke-sedentary. Moderate  Despite the increased amount of osteovasculature, the reduction in vessel-bone proximity, 373 resulting primarily from loss of small vessels, may explain the lack of exercise-induced gains in 374 bone microstructure following stroke, which we found in the same region using the same bones as 375 in this study [27]. 376 In this study, we extended previous clinical findings that limb perfusion is reduced 377 following ischemic stroke and demonstrated for the first time, using a mouse model, that 378 intraosseous perfusion is also reduced during early stroke recovery, particularly in the affected 379 limb, and notably even in the absence of limb disuse. These functional deficits occurred in 380 conjunction with changes in osteovascular structure, including potentially compensatory increases mitigated the negative effects of stroke on bone perfusion, the combination of stroke and exercise 383 altered the vessel proximity to bone to a less osteogenic arrangement with fewer small vessels 384 located near bone. These findings suggest that exercise like this moderate treadmill regime may 385 not be beneficial for osteovascular structure, although more studies are needed to examine if these 386 effects extend to other exercise therapies or rehabilitation strategies. Examining potential 387 mechanisms for these changes in osteovascular structure and function following stroke may 388 provide insight for mitigating skeletal fragility in human stroke patients.
Acknowledgments 390 We thank Dr. Eva Johannes and Dr. Mariusz Zareba for confocal microscopy support; and Dr.

391
Consuelo Arellano for statistical consulting. This work was performed in part at the Cellular and             For the stroke or sham surgery, an incision was made over the neck midline, and the right common carotid artery (CCA), external carotid artery (ECA), and internal carotid artery (ICA) were exposed. In the sham group, saline was added to the neck incision to prevent tissue from drying out, CBF was monitored, and tibial perfusion was recorded for 30 minutes. In the stroke group, temporary ligations were made to the CCA and ICA, while two permanent ligations were made to the ECA, and the vessel was cut between them. Baseline CBF values were collected by loosening the CCA ligation for 2 minutes. The CCA ligation was then retightened, and a thin 6-0 nylon monofilament occluder with a silicon-coated tip (Doccol Corporation, Redlands, CA) was passed through the ECA to the ICA and MCA origin until an 80% reduction in CBF relative to baseline was noted and maintained. The size of the silicon coating was selected based on body mass, per the manufacturer's instructions, and either a 1-2 or 2-3 mm long, 0.20-0.24 mm diameter coating was selected. Saline was added to the incision to prevent tissue from drying out. The occluding filament was left in place for 30 minutes, and tibial perfusion was recorded. After 30 minutes, the occluder was gently retracted, the ECA was permanently ligated, and temporary ligations were removed. For both sham and stroke procedures, the LDF probes were removed, and an intraincisional injection of bupivacaine (2 mg/kg, Marcaine, Hospira, Lake Forest, IL) was administered to the neck. The neck and skull incisions were sutured closed, and the hindlimb incision was closed with tissue adhesive (VetBond™, 3M Company, St. Paul, MN). Triple antibiotic ointment and 4% lidocaine cream were applied to all incision sites, and a subcutaneous injection of carprofen (7 mg/kg, Rimadyl, Zoetis, Parsippany, NJ) was administered. Bupivacaine was administered via subcutaneous injection for the first two days of recovery. The severity of stroke impairments was examined weekly following surgery by assessing sensorimotor function with neuroscores [4].
LDF provides a functional measure of blood flow in long bones that is influenced by the amount of blood vessels, blood flow velocity, blood vessel permeability, and blood vessel size [6]. We previously showed that our modified, less invasive LDF procedure can be performed serially without inducing inflammation or gait abnormalities [7]. After 6-8 hours of fasting, anesthesia was induced with 4% isoflurane in pure oxygen and maintained with about 2% isoflurane throughout the 15-to 20-minute-long procedure. The fur over both proximal tibiae was shaved. Mice were placed supine on a heated pad, and the hindlimbs were secured with tape. Using the same methods as described above, an incision was made over one tibia, the periosteum was gently removed, and the LDF needle probe was placed firmly against bone with the micromanipulator. A 30-sec measurement was recorded, the probe was removed and repositioned, and a second 30-sec measurement was recorded. The incision was closed with tissue adhesive, and triple antibiotic ointment and lidocaine cream were applied. The procedure was repeated for the contralateral tibia.
Each weekly perfusion measurement was composed of the weighted mean of the two 30-sec long measurements for that bone. Branching parameters of the osteovascular network, including total number of branches, number of junctions, number of triple points (e.g., junctions with three branches), and number of quadruple points (e.g., junctions with four branches), were quantified using the 'Skeleton 3D' and 'Analyze Skeleton' plugins in BoneJ (FIJI v. 1.51n) [9,10], applied to the binarized blood vessels.
The distance between blood vessels and bone surfaces was calculated with custom code in MATLAB ® (R2018, The MathWorks, Natick, MA). The proximity of blood vessels to bone surfaces was also examined. Similar to the procedure described above, a 1.6-mm-long VOI was manually selected in Slicer (v. 4.11.0) [11], starting at the distal growth plate and extending proximally into the diaphysis, including the cortical bone. The VOI was exported to MATLAB ® , and bone tissue, adipocytes, and blood vessels were binarized and processed using the same global threshold values and processing steps described above. The distribution of distances between blood vessel surfaces and bone surfaces (Ves.S-BS distance) was calculated using the 'bwgeodesic' function in MATLAB ® .

Osteovascular Composition (Tibia)
Osteovascular composition was examined using immunofluorescence confocal microscopy in bone tissue sections from a subset of affected tibiae (n = 3 sham-sedentary, n = 2 sham-exercise, n = 5 stroke-sedentary, n = 4 stroke-exercise) labeled for markers of non-arterial endothelial cells (endomucin, EMCN) and endothelial cells in sinuses, arterioles, venules, and capillaries (CD31) [12,13]. Blood vessels that are positive for both EMCN and CD31 (defined as the following: 1) effect of ischemic stroke and exercise on intraosseous perfusion during stroke recovery; 2) acute effect of ischemic stroke on intraosseous perfusion during surgery; 3) effect of stroke and exercise on osteovascular structure and composition; and 4) whether stroke differentially affects intraosseous perfusion or osteovascular structure in the affected vs.
For analysis #1, LDF measures of perfusion were compared between surgery and activity groups at each timepoint and limb using a mixed hierarchical linear model (procedure GLIMMIX) with interaction between all terms [16]. Each mouse was considered a subject (replicate), assigned randomly to a group comprised of the combination of surgery and activity groups. Surgery groups (sham or stroke) and activity groups (sedentary or exercise) were modeled as fixed factors, limb as a nested within-subject factor, and timepoint as a longitudinal repeated measure. LDF was measured across four timepoints (Weeks 1-4) at each observational unit (each limb of each mouse).
Variation among subjects within each group combination (i.e., sham-sedentary) was considered a random effect. Residuals were modeled using a compound symmetry covariance structure. A modified Kenward-Roger approximation was used to calculate denominator degrees of freedom and standard error of fixed effects [16]. Effect differences between surgery and activity groups were compared within each timepoint (i.e., stroke-exercise vs. stroke-sedentary at Week 2) using least squares means and Tukey-Kramer adjustments for multiple comparisons.
For analysis #2, the stability of the tibial perfusion measurements throughout the surgery was examined using the slope of the 30-minute LDF measurement taken during the surgical procedure. To determine if the slope for each group was different from zero, which would indicate that the perfusion measurement varied and was not stable, an unpaired t-test was performed for the sham group, and a Wilcoxon rank sum test was performed for the stroke group. Slopes were also compared between the stroke and sham groups with a Wilcoxon two-sample test.
For analysis #3, distribution data from CE-CT (vessel thickness, blood vessel surface to bone surface distance) were compared between surgery and activity groups using a mixed hierarchical linear model (procedure GLIMMIX) with interaction between terms. The model was similar to the model used for analysis #1, except the repeated factor 'week' was replaced with histogram 'bins'. Variation among subjects within each group combination (i.e., sham-sedentary) was considered a random effect that was modeled with a heterogenous variance components method across each surgery group (sham or stroke). Variation among each group combination was modeled as an intercept for each group linear predictor. A modified Kenward-Roger approximation was used to calculate denominator degrees of freedom and standard error of fixed effects. Effect differences were compared between groups within histogram bins (i.e., strokeexercise vs. stroke-sedentary within bin 'a') using least squares means with Tukey-Kramer adjustments for multiple comparisons. All other CE-CT parameters (Ad.V/Ma.V, average Ad.Th, Ad.D, Ves.V/Ma.V, average Ves.Th, Ves.D, branches, junctions, triple points, and quadruple points) were compared between surgery and activity groups using a repeated measures factorial model (procedure MIXED) with interaction between terms. Surgery and activity groups were modeled as fixed factors, while limb was modeled as a repeated measure. Residual variance was modeled with compound symmetry covariance, and effect differences were compared between groups using least squares means with Tukey-Kramer adjustments for multiple comparisons.
Immunofluorescence parameters were compared between surgery and activity groups using a standard two-factor general linear model (procedure GLM) with interaction and Tukey adjustments for multiple comparisons.