Voluntary running does not increase capillary blood flow but promotes neurogenesis and short-term memory in the APP/PS1 mouse model of Alzheimer’s disease

Exercise exerts a beneficial effect on the major pathological and clinical symptoms associated with Alzheimer’ s disease in humans and mouse models of the disease. While numerous mechanisms for such benefits from exercise have been proposed, a clear understanding of the causal links remains elusive. Recent studies also suggest that cerebral blood flow in the brain of both Alzheimer’ s patients and mouse models of the disease is decreased and that the cognitive symptoms can be improved when blood flow is restored. We therefore hypothesized that the mitigating effect of exercise on the development and progression of Alzheimer’ s disease may be mediated through an increase in the otherwise reduced brain blood flow. To test this idea, we examined the impact of three months of voluntary wheel running in ∼1-year-old APP/PS1 mice on short-term memory function, brain inflammation, amyloid deposition, and cerebral blood flow. Our findings that exercise led to improved memory function, a trend toward reduced brain inflammation, markedly increased neurogenesis in the dentate gyrus, and no changes in amyloid-beta deposits are consistent with other reports on the impact of exercise on the progression of Alzheimer’ s related symptoms in mouse models. Notably, we did not observe any impact of wheel running on overall cortical blood flow nor on the incidence of non-flowing capillaries, a mechanism we recently identified as one contributing factor to cerebral blood flow deficits in mouse models of Alzheimer’ s disease. Overall, our results replicate previous findings that exercise is able to ameliorate certain aspects of Alzheimer’ s disease pathology, but show that this benefit does not appear to act through increases in cerebral blood flow.

The novel object recognition test (NOR) was used to evaluate object-identity memory and 182 explorative behavior. The testing protocol was identical to the object replacement test, except 183 mice were returned to their home cage for 90 min in between trials and one of the initial objects 184 was replaced, at the same location, with a novel object for the testing period. The preference 185 score was analogously determined from the testing period data and calculated was as 186 (exploration time of the novel object/exploration time of both objects). 187

SURGICAL PREPARATION 189
For cranial window implantation, mice were anesthetized under 3% isoflurane and then 190 maintained at 1.5 -2% isoflurane in 100% oxygen. Once fully sedated, mice were 191 subcutaneously administered dexamethasone (0.025 mg per 100 g; Phoenix 192 Pharm, Inc.) to reduce post-surgical inflammation, atropine (0.005 mg per 100 g; 193 54925-063-10, Med-Pharmex, Inc.) to prevent lung secretions, and ketoprofen (0.5 mg per 100 194 g; Zoetis, Inc.) to reduce post-surgical inflammation and provide post-surgical analgesia. Mice 195 were then provided with atropine (0.005 mg per 100 g) and 5% glucose in saline (1 ml per 100 196 g) every hour while anesthetized. The hair was shaved from the back of the neck up to the 197 eyes. Mice were placed on a stereotaxic frame over a feedback-controlled heating blanket to 198 ensure body temperature remained at 37° C. The head was firmly secured and eye ointment 199 was applied to prevent the animal's eyes from drying out. The operating area was then 200 sterilized by wiping the skin with iodine and 70% ethanol three times. The mice were given 0.1 201 ml bupivacaine at the site of incision to serve as a local anesthetic. The skin over the top of 202 the skull was removed and the skull exposed. Using a high-speed drill with different sized bits, 203 a ~6-mm diameter craniotomy was performed over the cerebral cortex, rostral to the lambda 204 point and caudal to bregma. The exposed brain was then covered with a sterile 8-mm diameter 205 glass coverslip which was glued to the skull surface using cyanoacrylate adhesive. Using 206 dental cement, a small well around the window was created. After completion of the 207 craniotomy, mice were returned to their cages and were subcutaneously administered 208 ketoprofen (0.5 mg per 100 g) and dexamethasone (0.025 mg per 100 g) once daily for three 209 days, and their cages were placed on a heating pad during this time. Animals were given three 210 weeks to recover from the surgery before imaging experiments. 211

IN-VIVO TWO-PHOTON MICROSCOPY 213
For imaging sessions, mice were anesthetized with 3% isoflurane and then maintained at 214 1.5 -2% isoflurane in 100% oxygen. Atropine and glucose were provided, as described above. 215 Eye ointment was applied to prevent the eyes from drying out. Mice were placed on a 216 stereotactic frame over a feedback-controlled heating pad to keep body temperature at 217 37° C. To fluorescently label the microvasculature, Texas Red dextran (50 μl, 2.5% w/v, 218 molecular weight (MW) = 70,000 kDA, Thermo Fisher Scientific) in saline was injected retro-219 orbitally. Although we did not use these labels for any analysis, these mice also had 220 Rhodamine 6G (0.1 ml, 1 mg/ml in 0.9% saline, Acros Organics, Pure) injected into the 221 bloodstream to label leukocytes and blood platelets and Hoechst 33342 (50 μl, 4.8 mg/ml in 222 0.9% saline, Thermo Fisher Scientific) to label leukocytes. A custom-built two-photon excitation 223 fluorescence (2PEF) microscope was used to acquire three-dimensional images of the cortical 224 vasculature and to measure red blood cell flow in specific capillaries. Imaging was performed 225 with 830-nm, 75-fs pulses from a Ti-Sapphire laser oscillator (Vision S, Coherent). Lasers were 226 scanned by galvanometric scanners (1 frame/second) and ScanImage software was used to 227 control data acquisition [64]. For obtaining broad maps of the cortical surface vasculature, a 228 4x magnification air objective (numerical aperture of 0.28, Olympus) was used. For high-229 resolution imaging, a 25x water-immersion objective lens (numerical aperture of 0.95, 230 Olympus) was used. The emitted fluorescence from Texas Red was detected on a 231 photomultiplier tube through an emission filter with a 641-nm center wavelength and a 75-nm 232 bandwidth. Stacks of images were created by repeatedly taking images axially spaced at 1 μm 233 up to a depth of ~300 μm. Additionally, centerline line scans and image stacks across the 234 diameter of ~15 capillaries within the imaging area were obtained in each mouse. 235

ANALYSIS OF CAPILLARY BLOOD FLOW AND VESSEL DIAMETER 237
Blood flow velocities were determined based on the movement of red blood cells (RBCs). Since 238 the injected dye (Texas Red) labels the blood plasma only, RBCs are seen as dark patches 239 within the capillaries. We acquired repetitive line scans along the centerline of individual 240 capillaries, forming space-time images with diagonal streaks due to moving RBCs. As 241 previously described [65], we used a Radon transform-based algorithm to determine the slope 242 of these streaks and thus quantify RBC flow speed. Post-hoc sensitivity analysis (G*Power; 243 a=0.05, b=0.80) with our realized SD and means between running and sedentary groups 244 suggested we could detect a 30% change in flow speed. Capillary diameters were extracted 245 from image stacks taken along with each line scan. 246

CROWD-SOURCED SCORING OF CAPILLARIES AS FLOWING OR STALLED 248
In the 2PEF microscopy used to take three-dimensional image stacks of the vasculature, the 249 intravenously injected fluorescent dye labels the blood plasma, but not red blood cells. 250 Because each capillary segment was visible for multiple frames in the image stack, flowing 251 segments showed different patterns of fluorescent and dark patches in successive frames due 252 to moving blood cells. In capillary segments with stalled blood flow, the dark shadows from 253 non-moving cells remained fixed across all frames where the capillary segment was visible. 254 We used a purpose-built citizen science data analysis platform, StallCatchers.com, that 255 enabled volunteers to score ~26,000 individual capillary segments as either flowing or not 256 using 2PEF image stacks. We recently reported on the methodological details and validation 257 of this StallCatchers based scoring [58]. Briefly, we used a convolutional neural network, called 258 DeepVess, to segment the 2PEF image stack into voxels that were within vs. outside the 259 vasculature [66]. Individual capillary segments were identified using standard dilation and 260 thinning operations to define vessel centerlines [67], with capillary segments defined as the 261 path between two junctions. To restrict this analysis to capillaries, we excluded all segments 262 with diameter greater than 10 µm. Image stacks were then created that each had a single 263 identified capillary segment outlined and these stacks were analyzed by citizen scientists using 264 StallCatchers to score the identified segment as flowing or stalled. Each segment was scored 265 by multiple volunteers, each of whom had a sensitivity defined by their performance on capillary 266 segments we knew to be flowing or stalled, and we computed a weighted "crowd confidence" 267 score representing the likelihood of each segment being stalled. Laboratory researchers then 268 looked at these capillary segments as a final validation, starting with segments with the highest 269 crowd confidence score for being stalled. The crowd confidence score ranged from 0 to 1, and 270 laboratory researchers examined vessels with a score between 0.5 and 1, a total of 259 271 capillary segments. For crowd confidence scores between 0.9 and 1, laboratory researchers 272 concluded 95% were stalled, while for crowd confidence scores between 0.5 and 0.6, only 1 273 out of 80 vessels were not flowing. The initial scoring by citizen scientists decreased the 274 number of capillary segments laboratory researchers needed to evaluate by a factor of 100. 275 We report the density of non-flowing capillaries as stalls per cubic millimeter. 276

CHARACTERIZATION OF GEOMETRIC PROPERTIES OF CORTICAL CAPILLARIES 278
For each identified capillary segment from the 2PEF image stacks that were segmented using 279 DeepVess [66], we calculated the diameter (averaged along the length of the segment), the 280 segment length (distance along the centerline of the vessel between two junctions), and the 281 tortuosity (segment length divided by the Euclidean distance between the two junctions). 282

IMMUNOSTAINING OF BRAIN TISSUE 284
For evaluating cell proliferation, mice received intraperitoneal injection of 5-ethynyl-2'-285 deoxyuridine (EdU; E10415, ThermoFisher Scientific) at a dose of 25 mg/kg body weight every 286 day for four days before harvesting the brains. Mice were sacrificed by lethal injection of 287 pentobarbital (5 mg/100 mg). Brains were extracted and cut in half along the center line. One 288 hemisphere was snap frozen in liquid nitrogen for future protein extraction and the other half 289 was kept in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) for 48 hours at 290 4° C and subsequently placed in 30% sucrose for histological analysis. Immunohistochemistry 291 was performed on a set of coronal sections cut on a cryotome with OCT media at a thickness 292 of 30 μm. From each mouse, every sixth section was mounted, washed with PBS, and blocked 293 1 hour at room temperature (3% goat serum, 0.1% Triton-X100 in PBS). Sections were then 294 incubated overnight at 4°C with primary antibodies against Iba1 (1:500, rabbit anti-mouse Iba1; 295 019-19741, WAKO) and GFAP (1:500, chicken anti-mouse GFAP; ab53554, Abcam) in the 296 same buffer. Secondary antibodies (goat anti-chicken Alexa Fluor 488, goat anti-rabbit Alexa 297 Fluor 594; ThermoFisher Scientific) were used at 1:300 dilution and added to the slides for 3 298 hours at room temperature. Cell proliferation was detected through use of the Click-iTTM EdU 299 Cell Proliferation Kit for Imaging (Alexa Fluor 647 dye, ThermoFisher Scientific). Methoxy-X04 300 was used to counterstain amyloid deposits (1 mg/ml MeO-X04 (5 mg/ml in 10% DMSO, 45% 301 propylene glycol, and 45% saline) for 15 minutes at room temperature, 4920, Tocris). Hoechst 302 33342 was used to label cell nuclei (3 μg/ml, ThermoFisher Scientific). Sections were washed 303 with PBS before mounting with Richard-Allan Scientific Mounting Medium (4112APG, 304 ThermoFisher Scientific). Images were obtained using confocal microscopy (Zeiss 305 Examiner.D1 AXIO) operated with Zen 1.1.2 software. Z-stack images of the hippocampal and 306 cortical regions of each slide were acquired with 1 μm optical section thickness (three adjacent 307 images between the suprapyramidal blade of the granule cell layer of the dentate gyrus and 308 CA1, one image in CA3, and two adjacent images in the cerebral cortex taken directly toward 309 the cortical surface from the hippocampus). Images were then binarized using a manually 310 determined threshold. Appropriate thresholds varied between mice and were adjusted to 311 ensure that all morphologically relevant objects were recognized. The fraction of pixels above 312 threshold (%Area) and the integrated density (product of mean gray value and area) was 313 determined across sections for cortical and hippocampal regions. All sections were stained 314 and imaged in parallel. Researchers were blinded in this analysis as to whether mice were part 315 of the running or sedentary group. 316

ELISA ASSAY 318
The frozen half-brains were weighed and homogenized in 1 ml PBS containing complete 319 protease inhibitor (Roche Applied Science) and 1 mM AEBSF (Sigma) using a Dounce 320 homogenizer. The homogenates were sonicated for 5 min and centrifuged at 14,000 g for 30 321 min at 4° C. The supernatant (PBS-soluble fraction) was removed and stored at -80° C. The 322 pellet was re-dissolved in 0.5 ml 70% formic acid, sonicated for 5 min, and centrifuged at 323 14,000 g for 30 min at 4° C, and the supernatant was removed and neutralized using 1 M Tris 324 buffer at pH 9 (insoluble fraction). Protein concentration was measured using the Pierce BCA 325 Protein Assay (ThermoFischer Scientific). The extracts were then diluted to equalize different 326 protein concentrations. These samples were analyzed by sandwich ELISA for Aβ1-40 and 327 Aβ1-42 using commercial ELISA kits and following the manufacturer's protocol 328 (ThermoFischer Scientific). The Aβ concentration was calculated by comparing the sample 329 absorbance with the absorbance of known concentrations of synthetic Aβ1-40 or Aβ1-42 330 standards assayed on the same plate. Data was acquired with a Synergy HT plate reader 331 (Biotek) and analyzed using Gen5 software (BioTek) and Prism8 (GraphPad). 332

STATISTICAL ANALYSIS 334
To determine statistical significance of differences between groups, the data was first tested 335 for normality using the Shapiro-Wilk normality test. In case of normality, the statistical 336 comparison was performed using an unpaired t-test for comparison between two groups. In 337 case of non-normal distribution, the Mann-Whitney test (two groups) was used. For the 338 analysis of blood flow speed and capillary diameter, a nested two-way analysis of variance 339 (ANOVA) was carried out to account for possible interactions between individual mice. 340 P-values less than 0.05 were considered statistically significant and we used a standardized 341 set of significance indicators in figures: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. 342 Boxplots show the median with a black line; the mean is indicated with a red line. The box 343 spans between the 25th and the 75th percentile of the data, defined as the interquartile range 344 (IQR). Whiskers of the boxplots extend from the lowest datum within 1.5 times the IQR of the 345 lower quartile of the data to the highest datum within 1.5 times the IQR of the highest quartile 346 of the data. Correlation analysis was performed via the Pearson Product-Moment Correlation, 347 with best fit lines indicated in red. Statistical analysis was performed and graphs were created 348 using Prism8 (GraphPad).

RUNNING DISTANCE 353
To determine the effects of three months of voluntary exercise in 13 month old APP/PS1 mice, 354 we provided single-housed mice with unrestricted access to a monitored running wheel inside 355 their home cages (Fig 1A) [68]. All mice reliably ran on the wheels on a daily basis, with just 356 under a factor of three variance in the average daily running distance between individual 357 animals ( Fig 1B). Sedentary, control APP/PS1 mice were singly housed for the same duration, 358 but without access to a running wheel.

361
Timeline of the study. After three months of voluntary exercise on a wheel running (or standard housing for the 362 sedentary control group), mice underwent behavioral testing. Next, a craniotomy was performed, after which mice 363 were allowed three weeks to fully recover. The cortical vasculature was then imaged with in-vivo two photon 364 excitation fluorescence microscopy. Finally, after four consecutive days of EdU-injections, mice were euthanized. To investigate cognitive function in running versus sedentary APP/PS1 mice, we performed a 370 variety of behavioral tests aimed at testing different aspects of memory performance. Running 371 mice achieved a significantly higher preference score in the object replacement test (OR) than 372 sedentary mice, i.e. they spent significantly more time exploring the replaced object than the 373 object in the familiar location as compared to the sedentary mice (Fig 2A; P = 0.014). There 374 was also a significant correlation between individual running distances and OR performance 375 ( Fig 2B; R 2 = 0.58 P = 0.028). Conversely, we did not observe an increase in spontaneous 376 alternation between the three arms of the Y-maze (Fig 2C), nor increased exploration time for 377 a novel object in the novel object recognition (NOR) test (Fig 2E), nor increased time spent in 378 the center of the arena in an open field (OF) test (Fig. 2G) in running as compared to sedentary 379 APP/PS1 mice. Similarly, none of these measures was significantly correlated with individual 380 running distances (Fig 2D, 2F, and 2H). Additionally, there was no significant difference in time 381 spent exploring objects in the OR test (Fig S1A) or NOR test (Fig S1B), nor in the number of 382 arm entries in the Y-maze (Fig S1C), nor in the average track length during the OF test (Fig  383   S1D) between running and sedentary groups.

BRAIN INFLAMMATION NOR AMYLOID DEPOSITION 396
Exercise has been shown to exert its effects through multiple pathways, with increased 397 hippocampal neurogenesis being one of the most consistently described effects in both 398 patients [69], and in AD mouse models [70,71]. Consistent with these previous findings, we 399 found an increased number of proliferating neuronal stem cells in the dentate gyrus of running, 400 as compared to sedentary, mice, as detected by the number of EdU-positive cells (Fig 3A and  401 3B; P = 0.03). In the same hippocampal tissue sections, as well as in cortical slices (Fig 3C), 402 we further examined the density of Iba1 and GFAP staining to quantify any differences in the 403 density of microglia and astrocytes, respectively, providing a measure of brain inflammation. 404 We also quantified the density of amyloid plaques that were labeled with Methoxy-X04. We 405 found no notable differences in microglia (Fig 3D) or astrocyte (Fig 3E) density, nor in the 406 number of amyloid deposits (Fig 3F) between running and sedentary APP/PS1 mice in both 407 hippocampal and cortical regions. Using ELISA assays, we quantified the concentration of 408 Aβ1-40 and Aβ1-42 in brain lysates and saw no differences between running and sedentary 409 mice in both the soluble and insoluble fractions (Fig. 4).

VOLUNTARY RUNNING DID NOT AFFECT CAPILLARY BLOOD FLOW 432
The trends toward improved cognitive function (Fig 2), the clear increase in neurogenesis, as 433 well as the lack of a notable impact on brain inflammation and amyloid deposition ( Fig. 3 and  434 4) that we observed are consistent with previous studies of the impact of exercise on mouse 435 models of AD [72]. We next sought to determine if these exercise-mediated changes were 436 correlated with an increase in brain blood flow, perhaps linked to a decrease in the incidence 437 of non-flowing capillaries. We used a crowd-sourced approach to score individual capillary 438 segments as flowing or stalled based on the motion of red blood cells (which appear as dark 439 patches within the fluorescently-labeled blood plasma in 2PEF image stacks) (Fig 5A). 440 Contrary to our initial hypothesis, we did not observe a decrease in the incidence of stalled 441 capillaries in running APP/PS1 mice as compared to sedentary controls (Fig 5B). We further 442 quantified red blood cell flow speed and vessel diameter in cortical capillaries (Fig. 5C), and 443 saw no differences, on average, between running and sedentary APP/PS1 mice (Fig 5D). 444 Capillary speed and diameter were also not found to be correlated with the total distance ran 445 by animals (Fig S2). To further explore potential exercise-induced changes in the brain 446 vasculature, we used a convolutional neural network-based segmentation algorithm, 447 20 DeepVess [66], to segment 3D image stacks of the cortical vasculature of running ( Fig 6A) and 448 sedentary (Fig 6B) APP/PS1 mice and we characterized the capillary density (Fig 6C), capillary 449 diameter (Fig 6D), capillary segment length (Fig 6E), and capillary tortuosity (Fig 6F). None of 450 these parameters differed between running and sedentary APP/PS1 mice.

469
Physical activity is correlated with attenuation of cognitive impairment, amelioration of age-470 related changes in the brain, and reduced risk of dementia [73,74]. While exercise has been 471 found to have a positive effect on cognition and brain health in aging individuals [75], the 472 underlying mechanisms have not been fully elucidated. In this study, we investigated the 473 effects of several months of voluntary wheel running on memory function, AD-related brain 474 pathology, and cortical blood flow in aged APP/PS1 mice. 475 Impairments in spatial learning and memory in APP/PS1 mice have been reported in 476 mice as young as seven months of age [76,77]. We found performance in the object 477 replacement task in 12-month-old APP/PS1 mice was markedly improved by three months of 478 voluntary wheel running, as compared to sedentary controls, while no improvements were 479 detected in the novel object recognition task and Y-maze tasks. This exercise-related 480 improvement in performance on some memory-related tasks but not on others is consistent 481 with similarly mixed impacts of exercise in previous studies of memory function in mouse 482 models of AD [72] and in AD patients [78]. While a variety of brain regions are likely involved 483 in each of these memory tasks, it has been shown that memory of an object's spatial location 484 (evaluated with object replacement task) is highly dependent on hippocampal regions, while 485 memory of an object's intrinsic characteristics (evaluated with novel object task) also involves 486 significant contributions from other brain regions, such as the temporal lobe [79][80][81]. It is 487 possible that exercise differentially improves function in different brain regions and this 488 contributes to the mixed impact of exercise on different memory tasks. 489 Consistent with a broad consensus in the literature [37,40], we observed increased 490 neural stem cell proliferation in the dentate gyrus of the hippocampus in exercising APP/PS1 491 mice, as compared to sedentary controls. Exercise did not, however, lead to changes in brain 492 inflammation, as assayed by microglia and astrocyte density, or in amyloid pathology, as 493 assayed by amyloid plaque density and amyloid-beta monomer concentrations. A recent study 494 in the 5xFAD mouse model of AD similarly found no decreases in measures of brain 495 inflammation or in the density of amyloid deposits due to exercise [82]. 496 We have shown that neutrophils plug a small fraction of brain capillary segments in the 497 APP/PS1 and 5xFAD mouse models of AD, and that this contributes to the overall brain blood 498 flow reductions seen in these mice [57]. Contrary to our initial hypothesis that exercise could 499 decrease the microvascular dysfunction that underlies these capillary stalls, we did not find a 500 decrease in the number of non-flowing capillaries in running APP/PS1 mice, as compared to 501 sedentary controls. Further, we did not detect any differences in average capillary blood flow 502 speeds between running and sedentary APP/PS1 mice. While there is clear consensus that 503 reduced brain blood flow in both AD patients and mouse models is a key feature of the disease, 504 the effects of exercise on cortical blood flow in AD remain up for debate. Consistent with our 505 findings in APP/PS1 mice, however, recent studies in patients with dementia (diagnosed as 506 mild to moderate AD in Refs. [83] and [84]) showed that 3 times per week 60 min of aerobic 507 exercise led to improved cognition [83,85,86], but this effect was not correlated with increased 508 cerebral blood flow [84,87]. Finally, we did not observe changes in the density or geometry 509 of cortical capillaries in APP/PS1 mice that exercised, compared to sedentary controls. 510 While we did not observe an impact of exercise on the capillary stalling phenomena we 511 recently tied to brain blood flow deficits in AD mouse models, there are other aspects of cortical 512 microvascular dysfunction that have been shown to occur in mouse models of AD that we did 513 not examine. For example, mouse models of AD have shown increased blood brain barrier 514