Gut Microbial Dysbiosis is Correlated with Stroke Severity Markers in Aged Rats Following Stroke

6 Background: An imbalanced gut microbial community, or dysbiosis, has been shown to occur 7 following stroke. It is possible that this dysbiosis negatively impacts stroke recovery and 8 rehabilitation. Species level resolution measurements of the gut microbiome following stroke 9 are needed to develop and test precision interventions such as probiotic or fecal microbiota 10 transplant therapies that target the gut microbiome following stroke. Previous studies have used 11 16S rRNA amplicon sequencing in young male mice to obtain broad profiling of the gut 12 microbiome at the genus level following stroke, but further investigations will be needed with 13 whole genome shotgun sequencing in aged rats of both sexes to obtain species level resolution 14 in a model which will better translate to the demographics of human stroke patients. 15 Results: 39 aged male and female rats underwent middle cerebral artery occlusion. Fecal 16 samples were collected before stroke and three days post stroke to measure gut microbiome. 17 Machine learning was used to identify the top ranked bacteria which were changed following 18 stroke. MRI imaging was used to obtain infarct and edema size and cerebral blood flow (CBF). 19 ELISA was used to obtain inflammatory markers. 20 Dysbiosis was demonstrated by an increase in pathogenic bacteria such as Butyricimonas 21 virosa (15.52 fold change, p<0.0001), Bacteroides vulgatus (7.36 fold change, p<0.0001), and 22 Escherichia coli (47.67 fold change, p<0.0001). These bacteria were positively associated with 23 infarct and edema size and with the inflammatory markers Ccl19, Ccl24, IL17a, IL3, and 24 complement C5; they were negatively correlated with CBF. Conversely, beneficial bacteria 25 such as Ruminococcus flavefaciens (0.14 fold change, p<0.0001), Akkermansia muciniphila 26 (0.78 fold change, p<0.0001), and Lactobacillus murinus (0.40 fold change, p<0.0001) were 27 decreased following stroke and associated with all the previous parameters in the opposite 28

(chlorohexidine scrub) prior to 70% EtOH and then a betadine solution. Maintenance isoflurane 119 was maintained at 2.5% in O2 was delivered via a nosecone placed in line with the binner 120 tubeQ (gas delivery tube) of the anesthesia circuit. Under bnear sterileQ conditions and with 121 the use of a Zeiss operating microscope (Carl Zeiss AG, Gottingen, Germany) at 4 to 25 122 magnification, the procedure was performed. First, the skin was opened with a midline vertical 123 incision, and the underlying submandibular gland bluntly dissected in the midline to produce 124 left and right lobes, which were retracted laterally. Division of the omohyoid muscle, then 125 dissection medial to the right sternocleidomastoid (SCM) muscle was used to expose the 126 common carotid artery (CCA), which was separated from the vagus nerve. tie that was applied just proximal to the first, leaving enough space in between the two ties to 134 cut the artery with micro scissors. At this point, blunt dissection was used to isolate the internal 135 carotid artery (ICA) and its collateral, the pterygopalatine artery. Next, microvascular 136 aneurysm clips (Mizuho, Beverly, MA, USA) were applied to the CCA and the ICA. A 5-0 137 PDS II monofilament embolus (Ethicon, Cornelia, GA, USA), was introduced into an 138 arteriotomy hole-produced with a 26-gauge hypodermic needle-in the reflected ECA stump 139 and fed distally into the ICA. At this time, a collar suture at the base of the ECA stump was 140 tightened around the embolus, and the ICA clamp was removed. The embolus was advanced 141 20 mm from the carotid bifurcation, with care taken to avoid entrance into the pterygopalatine 142 artery. 143 For the transient occlusion, the same steps were done as stated with the pMCAO, with the 144 exception that Doccol Corporation silicone rubber-coated monofilaments were used for the 145 occlusion of the middle cerebral artery (MCA). Multiple sized Doccol monofilaments are used 146 in the MCAO surgery depending on the sex and weight of the rat. Two 18-inch length of 5-0 147 silk suture were used for the ligation of the external carotid artery (ECA) to secure the ECA 148 stump, and the entry point of the monofilament into the ECA/ICA bifurcation. The third 5-0 149 silk suture was used to secure the monofilament within the ECA. A micro-serrefines arterial 150 clamp (FST, Fine Science Tools, #18055-01) was used to occlude the internal carotid artery 151 (ICA) and common carotid artery (CCA) prior to advancement of the monofilament into the 152 MCA. After 5 hours, the embolus was gently removed and the collar suture at the base of the 153 ECA stump tightened. The skin was closed with 3-0 nylon suture (Ethicon, Cornelia, GA, 154 USA), anesthesia discontinued, and the animal allowed to recover. Animals used for control 155 underwent a neck dissection and coagulation of the external carotid artery, but no manipulation 156 or occlusion of the common or internal carotid arteries. 157

Post-surgical fluid management and pain control
158 Immediately post-operatively the animals received 2 ml of sterile saline (0.9%) subcutaneous. 159 An additional 1 ml of saline was given if extra blood loss occurred during surgery. The animals 160 were injected with sterile filtered PBS pH 7.4 at 6 (for the p-MCAO), 24, 48, and 72 hours 161 post-MCAO. The animals were weighed every morning post-MCAO to determine dehydration. 162 Hydration status was checked by pinching up or "tenting" the skin over the nape of the neck. 163 The skin should immediately relax into its normal position. If the skin remains tented longer 164 than normal, the rat was deemed dehydrated, and saline was given. Per DLAR guidelines, rats 165 can receive up to 10 ml at a time and no more than 2 ml at any one location per 6 hr. If 166 warranted, additional saline (1-2 ml) will be given in addition to 6, 24, 48, and 72 hr. Also, we 167 added an additional water bottle in each cage to allow more avail-ability to free water for the 168 rats to consume and moistened food was provided on the bottom of the cage to encourage 169 feeding and additional water intake. Post-surgical pain control was managed with carprofen, 170 which is based on weight of the animal. Animal weights are taken prior to surgery (pMCAO) 171 and daily until animals are euthanized at 72 hr. (post MRI). The animals received a dosage of 172 carprofen 5mg/kg prior to surgery and every 24 hr. for three days post-pMCAO until 72 hr. 173 when they were euthanized (post MRI). Termination of survival criteria include that all animals 174 were weighed and monitored, especially for dehydration and pain, each morning post surgery. 175 This includes specific attention to the animal as a whole, as well as incision sights. If symptoms 176 such as pain, fatigue, loss of energy, excess energy, ruffled hair coat, reluctance to move, failure 177 to groom or feed, hypoactivity, hyperactivity, restlessness, self-trauma, aggressiveness, ataxia, 178 pale mucous membranes, cyanosis, rapid, shallow and/or labored breathing, cachexia, 179 porphyria, soiled anogenital area, inactivity, failure to respond to stimuli, lack of 180 inquisitiveness, vocalization, and/or hunched posture were observed, the research team 181 obtained advice from the vivarium veterinary staff on how best to intervene to alleviate 182 discomfort; if that was not possible the animal was euthanatized. Additional checks were made 183 in the afternoon if there was any rat of concern. The animals were removed from the study if 184 adverse signs persisted despite carprofen and treatment past 24 hr. If the signs fail to resolve, 185 the vivarium veterinarian was consulted and decided the time course when such animals were 186 euthanized. Additionally, weight loss greater than 20% (emaciated appearance, rapid weight 187 loss over two days) was considered an endpoint. Rapid weight loss was considered greater than 188 10% a day for two days. 189 with an average of 2.25% isoflurane in oxygen, while female rats were anesthetized with an 214 average of 1.75% isoflurane in oxygen using an MRI compatible CWE Inc. equipment 215 (Ardmore, PA). They were held in place on a Bruker scanning bed with a tooth bar, ear bars, 216 and tape. Body temperature, heart rate, and respiratory rate were continuously monitored 217 throughout the MRI scans (SA Instruments, Inc., Stony Brook, NY). The animal's body 218 temperatures were maintained at 37°C with a water heating system built into the scanning bed. 219

Microbiome Sequencing
The scanning procedure took approximately 40-60 mins. per animal. 220 The MR images were analyzed by a blinded neuroradiologist who visually identified infarct 221 volume and edema volume. These volumes were counted, and this number was normalized to 222 the number of images counted to provide a per section count. The volume of brain parenchyma 223 demonstrating infarct volume visibly affected was calculated by manual segmentation using 224 ITK-SNAP software (www.itksnap.org, version 3.6) 13 . The volume of brain parenchyma 225 visibly affected by T2 hyperintensity (edema volume) was calculated in a similar fashion. The 226 data are given as absolute volume in cubic millimeters. The calculation was based on all slices 227 from each MR sequence. Cerebral perfusion values of the area of lesion within the ipsilateral 228 hemisphere, and the equivalent region within the contralateral hemisphere were generated 229 using the quantification as previously described. 14 Receptor Array from Qiagen. Delta Delta CT was calculated using the fold change of the gene 244 expression measurement from pre to 3-day. 245

Statistical analysis
246 Descriptive microbiome analyses were performed with CosmosID bioinformatics software to 247 generate alpha diversity, beta diversity, and relative abundance data. Alpha diversities amongst 248 groups were compared using Wilcoxon Rank Sum test. Beta diversities amongst groups were 249 compared using PermANOVA. Relative abundance data was compared to measures of stroke 250 severity as determined by imaging (infarct size, edema size, CBF) using general linear models 251 within the MaAsLin 2 R package 18 . Random forest was used to determine top bacterial species 252 that were changed following stroke using the randomForest R package 19 . All imaging variables 253 in the study were transformed to meet assumptions of normality. The transformation 254 procedures began with Shapiro-Wilks and for measures with p < 0.05, the variables were square 255 root transformed. A p-value of 0.05 was set a priori to determine statistical significance. 256

257
We analyzed all rats before and after middle cerebral artery occlusion and considered sex, 258 surgery type, and treatment with LIF or PBS in the analysis. We administered a leukemia 259 inhibitory factor (LIF) treatment on half of the rats based on previous work suggesting that LIF 260 is an anti-inflammatory that regulates the immune/inflammatory response to stroke 20 . The rats 261 had an average of 96.50 mm 3 infarct size, 131.0 mm 3 edema size, and 1.31 ml/g/min CBF from 262 a permanent occlusion and 31.46 mm 3 infarct size, 102.1 mm 3 edema size, and 2.16 ml/g/min 263 CBF from a transient occlusion. Infarct and edema volumes were not significantly different 264 between sex, treatment group, or occlusion type. No significant difference in CBF was detected 265 between sex or treatment, but, as expected, a significant difference occurred between 266 permanent and transient occlusion in CBF (Fig. 1). 267 The aged rat gut microbiome is disrupted following stroke 268 We performed an analysis on the gut microbial communities of the aged rats before and after 269 stroke. Comparing the alpha diversity before and after stroke, we found that richness and 270 evenness increased from 3.818 on the Shannon diversity index 21 to 4.178 ( Fig. 2A). There were 271 no differences in the change of alpha diversity between sex, treatment, or occlusion type. 272 Comparing the beta diversity before and after stroke, we found that the microbial communities 273 were significantly different between baseline and stroke (p=0.0001), but no significant 274 microbial community differences were detected based on sex, treatment, or occlusion type. 275 ( Fig. 2B and Supplementary Table 1). 276 We investigated specific differences in the relative abundance of the major bacterial phyla in 277 the gut (Fig. 3). We found increases in proteobacteria and Bacteroidetes and decreases in 278 firmicutes, verrucomicrobia, and actinobacteria following stroke (Supplementary Table 2A). 279 This translates to a sharp decrease in the firmicutes to bacteroidetes ratio. Using linear 280 regression, the major bacterial phyla predict infarct size with an R 2 =0.3866 and edema size 281 with an R 2 =0.6022 (Supplementary Table 2B). 282 The top 13 disrupted bacterial species following stroke 283 We investigated specific differences in the relative abundance of the major bacterial species in 284 the gut. There was a total of 29 species increased and 23 species decreased following stroke 285 ( We investigated potential interactions between bacterial species in predicting infarct size, 299 edema size, and CBF (Supplementary Table 4). Using a feasible solution algorithm (FSA) 300 for finding interactions, we found that decreases in Lachnospiraceae bacterium A2 and 301 Lactobacillus murinus predict infarct size, but a combination of the two predicts a dramatic 302 increase in the prediction value with an R 2 =0.6206. Decreases in Lachnospiraceae bacterium 303 A4 and Lactobacillus murinus predict edema size, but a combination of the two have stronger 304 predictive ability with an R 2 =0.6454. Decreases in Adlercreutzia equolifaciens and 305 Desulfovibrio desulfuricans predict CBF, but again, a combination of the two has a stronger 306 prediction with an R 2 =0.8093. 307 Bacterial community disruptions following stroke are correlated 308 with stroke severity markers 309 We investigated the correlation of all the bacterial species with infarct size and edema size 310 (Table 3). Using the MaAsLin 2 R package 18 , which automatically normalizes and transforms 311 all variables in preparation for linear regression, we correlated metagenomic sequencing with 312 imaging variables of stroke severity. Twenty-seven bacterial species were positively correlated 313 and 19 negatively correlated with infarct volume. Thirty species were positively correlated, 314 and 31 species were negatively correlated with edema volume. No species were correlated with 315

CBF. 316
Bacterial community disruptions following stroke are correlated 317 with rises in inflammatory markers 318 We investigated the association of inflammatory markers with gut microbiome changes (Table  319 4). Using an Rt2 PCR array 22 to test the difference between inflammatory genes expressed 320 before and after stroke in a subsample of the rats, we found all the markers that were associated 321 with the changes in gut microbiome. There were 22 bacterial species changed with stroke that 322 were also correlated with changes in inflammatory markers. There were 49 total inflammatory 323 markers that were increased in association with bacterial changes (Supplementary Table 5). 324

325
To our knowledge, we are the first to report on the gut microbial changes with species level 326 resolution in aged male and female rats and to correlate these changes with clinical MRI 327 imaging markers of stroke and inflammatory markers. Following stroke, we found that alpha 328 diversity significantly increased, beta diversity significantly changed, and 5 of the 6 major 329 bacterial phyla were altered. Using machine learning, the top 13 bacterial species that predict 330 whether a sample came from the baseline or post-stroke time point. These bacterial species had 331 independent significant correlations with infarct size, edema size, and CBF. We also identified 332 several species whose interactions with one another were significant in correlating with stroke 333 imaging outcomes. Finally, we found 49 inflammatory markers that correlated with the changes 334 in microbiome from stroke. These changes are representative of a shift from beneficial to 335 pathogenic bacterial species following stroke which results in an increased inflammatory 336 response. 337 Figure 4 summarizes the changes in gut microbial communities in response to stroke. 338 Following stroke there is a significant shift in the gut microbiome, with alterations to 52 major 339 bacterial species. These bacterial fluctuations shift the environment to a more inflammatory 340 state that adversely affect injury. The microbial community dysbiosis is likely due to the 341 increased gut permeability and decreased gut motility in addition to the immunodepression 342 caused by the amplified stress response (increased sympathetic nervous system response and 343 hypothalamic-pituitary-adrenal (HPA) axis response) following stroke 23 . Previous groups have 344 reported a decrease in alpha diversity following stroke in a mouse model 6 and an increase in a 345 human model 24 . Our findings are consistent with others who have seen that microbial 346 communities differ before and after stroke based on measures of beta diversity 25 . We did not 347 find any significant differences in the microbiome between males and females. Some groups 348 have found sex differences in the microbiome that are largely attributed to hormone 349 differences 26 . It is possible that we did not see these differences because the female rats we 350 used are aged and reproductively senescent. 351 We saw increases in proteobacteria following stroke. In previous studies, proteobacteria have 352 been associated with increased cognitive impairment following stroke 27 . Dysbiosis related to 353 metabolic disorders, inflammation, and cancer is often related to an increase in 354 proteobacteria 28,29 . This is possibly due to increased oxygen content in the gut following 355 increases in inflammation, providing an optimal environment for these facultative anaerobes 30 . 356 We also saw decreases in firmicutes and increases in bacteroidetes species. Decreased 357 firmicutes have also been associated with Alzheimer's disease 31 . Obesity is often characterized 358 by a significantly increased firmicutes to bacteroidetes (F/B) ratio 32 ; interestingly, our study 359 found that stroke has the opposite effect on F/B ratio. Actinobacteria was significantly 360 decreased following stroke. Actinobacteria downregulates inflammation by production of IL-361 4 and IL-13 33 and is known to have anti-biofilm properties against pathogenic bacteria 34 . It is 362 possible that a decrease in actinobacteria allows other pathogenic bacteria to flourish. 363 Of the bacteria we found that are increased following stroke, many were of the bacteroides 364 species. Bacteroides species have the ability to reduce oxygen levels and breakdown food 365 products to liberate fucose and sialic acid residues from glycoproteins that can be consumed 366 by other microorganisms, including pathogens. Higher bacteroides species are associated with 367 type I diabetes 35 . Bacteroides vulgatus and Bacteroides dorei reduce gut microbial 368 lipopolysaccharide production and inhibit atherosclerosis 36 , but they are also associated with 369 insulin resistance, altered bile acid metabolism, and reduced interleukin-22 secretion 37 . 370 Butyricimonas virosa, Escherichia coli, and Parabacteroides distasonis were also elevated 371 following stroke. An increase of Butyricimonas virosa has also been seen in divers with high 372 occupational exposure to a hyperoxic environment 38 , which is very different from the hypoxic 373 environment of stroke. Escherichia coli is a very common commensal bacteria that has the 374 potential to cause extraintestinal infections based on its genome content and phenotypic traits 39 375 and is famous for causing post-stroke infections, especially pneumonia. Parabacteroides 376 distasonis has been shown to alleviate obesity and metabolic dysfunctions via production of 377 succinate and secondary bile acids 40 , which is interesting since stroke is often associated with 378 obesity and metabolic dysfunctions. 379 380 Many bacteria which are generally considered beneficial were decreased following stroke 381 including akkermansia, lactobacillus, and ruminococcus species. Akkermansia muciniphila is 382 a mucin-degrading bacterium 41 that can be increased with fasting 42 that is known to improve 383 host metabolic functions and immune responses 43 . Lactobacillus murinus can combat 384 inflammaging 44 , and a reduction of L. murinus due to high salt consumption has been 385 associated with an increase in proinflammatory TH17 cells 45 , which have been correlated with 386 post stroke dysbiosis and secondary injury 46 . Lactobacillus reuteri was also significantly 387 reduced following stroke. A randomized control trial in children showed administration of L. 388 reuteri as a probiotic to be useful in treating constipation in children 47 . Constipation is a 389 common morbidity in stroke, and administration of this species could help to alleviate 390 symptoms. Ruminococcus flavefaciens has also been shown to decrease the therapeutic effects 391 of antidepressants, having implications for the treatment of post-stroke depression. 392 Many of the bacterial changes were associated with increases in inflammatory markers. The 393 major markers that were increased were CCL19, CCL24, IL-17A, IL-3, and complement factor 394 C5. CCL19 is a chemokine that is commonly upregulated as a result of viral infections 48 , and 395 attracts dendritic cells and T lymphocytes 49 ; it promotes thymocyte development, secondary 396 lymphoid organogenesis, high affinity antibody responses, regulatory and memory T cell 397 function, and lymphocyte egress from tissues organs 50,51 . CCL19 suppresses angiogenesis and 398 can inhibit proliferation, migration, and sprouting responses of tumors 52 . CCL19 has previously 399 been found to be upregulated following stroke after damage to the intestinal epithelium 53 and 400 has been shown to facilitate T-cell migration to the insult site and microglial activation 401 following stroke 54 . CCL24 plays an important role in pathological processes of skin and lung 402 inflammation and fibrosis 55 and regulates inflammatory and fibrotic activities through its 403 receptor, CCR3 56 . CCR3 is a mediator of neural cell death 57 . In host defense, IL-17A has been 404 shown to be mostly beneficial against infection caused by extracellular bacteria and fungi 58 and 405 IL-17A has been shown to be increased following stroke, especially in males 59 . IL3 is strongly 406 associated with brain volume variation and plays pivotal roles in the expansion and 407 maintenance of the neural progenitor pool and the number of surviving neurons 60 ; our work 408 has previously identified IL3 increased in the spleen with our aged rat model of stroke 20

429
We found that alpha diversity significantly increased following stroke irrespective of sex, 430 treatment, or occlusion type. Beta diversity was also significantly different, with increases in 431 proteobacteria and decreases in the firmicutes to bacteroidetes ratio. Random forest analysis 432 revealed the top 13 species changes as a result of stroke including increases in Butyricimonas 433 virosa and Escherichia coli and decreases in Akkermansia muciniphila and Bacteroides dorei. 434 Correlation analysis revealed that these species changes were associated with increased infarct 435 and edema sizes following stroke. Furthermore, the bacterial changes were associated with 436 increases in inflammatory markers, notably Ccl19, Ccl24, IL17a, IL3, and complement C5.    Diversity changes following stroke. A) Alpha diversity as measured by the Shannon diversity index detecting species richness and evenness is increased following stroke. There is no difference in change across sex, treatment, or stroke type. B) Beta Diversity as measured by Bray-Curtis method comparing how different samples are Phyla changes as a result of stroke. Relative Abundance shows phyla composition before and after stroke.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. RatSupplementalTablesandFigures.pdf