Angiotensin II-mediated neuroinflammation in the hippocampus contributes to heart failure-induced neuronal deficits and cognitive impairment in rats

Ferdinand Althammer Georgia State University Ranjan Roy Georgia State University https://orcid.org/0000-0001-6234-8043 Matthew Kirchner Georgia State University Kathryn Whitley Georgia State University Steven Davis Georgia State University Juliana Montanez Georgia State University Hildebrando Ferreira-Neto Georgia State University Jessica Danh Georgia State University https://orcid.org/0000-0001-7739-0287 Rafaela Feresin Georgia State University https://orcid.org/0000-0003-0649-2774 Vinicia Biancardi Auburn University https://orcid.org/0000-0001-7301-1702 Marise Parent Georgia State University Javier Stern (  jstern@gsu.edu ) Georgia State University


Introduction
supportive of a pro-inflammatory state that strongly correlated with higher expression of proinflammatory 77 cytokines. These pro-inflammatory changes correlated with the progression and severity of the disease. We 78 found evidence for altered hippocampal neuronal excitability and substantial hippocampal apoptosis in HF 79 rats. Intriguingly, we observed an increased number of vessel-associated microglia, elevated expression of 80 hypoxia markers, angiogenesis and compromised hippocampal blood-brain barrier (BBB) integrity in HF 81 rats. Most of these changes correlated with an increased expression of AT1R mRNA levels in hippocampal 82 microglia. Importantly, treatment of HF rats with an AT1R antagonist improved all neuroinflammation-83 associated parameters, without affecting however the hippocampal hypoxic state. Finally, AT1R blockade 84 ameliorated cognitive deficits in HF rats. Collectively, our results highlight a novel mechanism of AngII-85 AT1aR-mediated hippocampal microglial activation in HF, which progresses over time and has detrimental 86 effects on hippocampal function and hippocampal-dependent memory.

89
Heart failure induces a pro-inflammatory microglial phenotype in the hippocampus

90
To assess potential microglia morphological differences between sham and HF rats, we stained 91 brain sections containing DH and VH for the microglial marker IBA1 (n=8 for both groups, Fig. 1a, Extended 92 Fig. 1a-f). We found larger microglia somatic volumes in HF rats (Fig. 1b,c 37 ), which could be indicative of 93 a pro-inflammatory microglial phenotype. This was further confirmed using an Imaris-assisted microglia 94 morphometry approach 17 that showed significant changes in microglia surface area, cell volume, and 95 filament length that are consistent with a proinflammatory phenotype (Fig. 1d-f). Interestingly, these 96 changes were dependent on the progression of the disease, with animals at later stages of HF displaying 97 more profound changes in microglial morphology. We recently reported lack of microglial morphological 98 changes in the somatosensory cortex (S1BF) of HF rats 17 . To further confirm that microglial morphology 99 changes in HF are not a brain-wide phenomenon, we analyzed also here the prelimbic cortex (PLC) and 100 found no difference between sham and HF rats (Extended Fig. 1g-i). To obtain more detailed information 101 about changes in microglial morphometry, we created heatmaps of 256 randomly selected microglia (32 per 102 animal) that display the maximum length and Sholl values for each microglia for both Sham and HF rats 103 (Fig. 1g, h). On average, microglia from HF rats displayed a reduced reach (i.e. maximum extension of 104 filaments) and diminished maximum complexity (peak Sholl value). Next, we compared between both 105 groups the proportions of microglia that were considered pro-inflammatory using our previously established 106 conservative threshold ( 17 , peak Sholl value <10). In sham rats, we found 14.9% and 15.3% (in VH and DH, 107 respectively) of all microglia to display a pro-inflammatory profile, while the value was increased in HF rats 108 (Fig. 1i, j; 41.2% and 36.2% for VH and DH, respectively). No such changes were observed in the PLC, 109 consistent with our initial findings (Extended Fig. 1j, k). Importantly, we found a significant negative 110 correlation between the degree of microglia cell activation and the corresponding rat echocardiography EF 111 value (Fig. 1k, l), suggesting that the microglial status within the hippocampus is dependent on the severity 112 of cardiac compromise in HF. The HF-induced decrease in microglia morphology complexity was further 113 confirmed by Sholl analysis (Fig. 1m). Counting of microglia revealed no differences in the total number or 114 density of microglia cells between the two groups, indicative of lack of microglia cell proliferation during HF 115 (Fig. 1n). Assessment of mRNA transcripts via qPCR revealed a significant increase in several 116 neuroinflammation-associated genes within the VH and DH (Fig. 1o), but not in the PLC (Extended Fig.   117 1l). Taken together, these findings suggest that hippocampal microglia undergo a morphological transition 118 towards a pro-inflammatory phenotype in HF which is progressive in time, and dependent on the severity of 119 the disease.

131
A core feature of neuroinflammation is the intricate interaction between microglia and astrocytes.

132
Microglia release C1q, TNFa and IL-1a to activate astrocytes, which in turn become neurotoxic (A1 133 phenotype) and induce apoptosis in neighboring neurons in affected brain areas 20 . To investigate whether 134 HF induced a morphological transition of astrocytes towards a pro-inflammatory A1 phenotype, we stained 135 astrocytes using GFAP and glutamine synthetase (GS) as markers which predominantly stain astrocytic 136 processes and astrocyte soma, respectively (Fig. 2h, i). We found that HF astrocytes in the VH and DH, 137 but not the PLC, had substantially increased soma volume and visible swelling of GFAP-labeled processes, 138 indicative of a hypertrophic phenotype 38 and reactive astrocytes 39 (Extended Fig. 2f-h). In line with this, we 139 found lower levels of neuroprotective A2-and higher levels of neurotoxic A1-related mRNA transcripts in 140 the hippocampus of HF rats (Fig. 2j). Similar to microglia changes, we did not observe changes in astrocyte 141 morphology (Extended Fig. 2h), nor an increase in the number of astrocytes in the PLC (Extended Fig.   142 2i). Taken together, these findings are consistent with the previously described interaction of microglia and 143 astrocytes during neuroinflammation 20,40 , and support that in addition to a microglia proinflammatory state,

144
the hippocampus of HF rats shows evidence for an astrocytic shift from a neuroprotective to a neurotoxic 145 state.

147
Evidence for hippocampal apoptosis in HF rats

148
To probe for potential neuroinflammation-induced apoptosis in the hippocampus of HF rats, we 149 performed a TUNEL assay to assess apoptotic levels in the DH and VH of sham and HF rats. This study 150 was done at two different time points (6w and 16w post HF) to be able to discriminate between early and 151 late apoptosis in our HF model (Fig. 3a, b, Extended Fig. 3 a, b). We observed frequent apoptotic clusters 152 in both DH and VH of HF rats, while those clusters were almost entirely absent in sham rats (Fig. 3 c, d).

153
To corroborate these findings, we stained brain sections of sham and HF rats with a cleaved caspase-3 154 (cCasp3) antibody and found significant more apoptotic clusters at 6w and 16w post HF surgery in the 155 hippocampus ( Fig. 3e-j). Interestingly, we observed a reduction in cell number at 16w, but not 6w post HF 156 surgery (Fig. 3g, j), suggesting continued apoptosis that manifests in significant decreases in cell numbers 157 only in later stages of HF. In addition to the observed apoptotic clusters and reduced cell number, we also 158 found significant thinning of pyramidal cell layers both in the DH and VH of HF rats (Extended Fig. 3 c-

159
which is in line with previous reports of hippocampal shrinkage in HF rats 13 .

160
To further assess hippocampal neuronal damage during HF we patched ex vivo pyramidal neurons 161 in dorsal CA1 in both sham and HF rats (n= 11 neurons/ 3 rats and 9 neurons/3 rats respectively) and 162 assessed changes in membrane excitability and firing discharge properties (Fig. 3 k-o, Extended Fig. 3f).

163
While resting membrane potential was similar between CA1 neurons in sham and HF, we observed a 164 significantly decreased input resistance which was evident as a decreased slope in the current/voltage plots 165 ( Fig. 3l). Moreover, we found a significantly diminished input/output function in CA1 neurons in HF rats (i.e. 166 decreased number of evoked action potentials per stimulation, Fig. 3o). AP amplitude, AP threshold, AP 167 half-width, peak AHP amplitude, sAHP amplitude, or AHP decay tau did not differ between sham and HF 168 rats (Extended Fig. 3f). Taken together, these findings indicate an early onset of HF-induced hippocampal 169 apoptosis, along with a blunted overall neuronal excitability and decreased ability to fire action potentials in 170 response to an incoming stimulus.

173
Myocardial infarction with reduced ejection fraction results in a hypoxic environment that can have 174 far-reaching consequences on brain vascularization, metabolism and ultimately brain function 41-44 . However, 175 whether tissue hypoxia and/or pathological angiogenesis occurs in the hippocampus of HF rats, and whether 176 it is mechanistically linked to the neuroinflammatory process during this condition remains unknown. Thus, 177 we performed qPCR time series for Hif1a and Hif2a, two widely used hypoxia markers, at 6-, 8-and 12-178 weeks post HF. We found a significant increase in both Hif1a and Hif2a in the VH and DH of HF rats.

179
Intriguingly, while the magnitude of the increase in Hif1a levels diminished over time, Hif2a kept increasing 180 over time (Fig. 4a, b). In line with previous studies showing that Hif2a promotes pathological angiogenesis 181 under various conditions 45,46 , we assessed for changes in hippocampal vascular density in sham and HF 182 rats at two different time points (6 and 12-weeks post HF) (Fig. 4c-h). We observed a prominent and 183 progressing increase in vascularization in both DH and VH of HF rats (Fig. 4e, f, h), as well as in the 184 somatosensory cortex, paraventricular nucleus of the hypothalamus (PVN) and central nucleus of the 185 amygdala (CeA) (Extended Fig. 4a-e). Finally, we analyzed vessel branching points for each animal and 186 found significantly higher numbers of vessel branching points in HF rats, with a progressive increase over 187 time (Extended Figure 4f). These findings support hypoxia and hyper-vascularization in multiple brain 188 regions in rats with HF.

191
Overactivation of the renin-angiotensin system (RAS) is a hallmark of HF, and disturbed angiotensin 192 II (AngII) signaling is thought to underlie many of the pathophysiological findings and symptoms associated 193 with early and late HF stages 8,19,28 . AngII is a pro-inflammatory neuropeptide 25 and we recently reported that 194 AngII type 1a receptors (AT1aR) are present in hypothalamic microglia 30,31,47 , while others have shown the 195 same in hippocampal microglia 33 . Thus, we first performed non-cell type-specific qPCR to determine 196 whether AT1aRs were upregulated in HF rats. Indeed, we found a time-dependent, progressive increase of 197 AT1aRs both in the DH and VH of HF rats when compared to age-matched sham controls (Fig. 5a).

198
Interestingly, we could confirm these findings in PVN and CeA, where we previously reported microglial 199 activation 17 , but not the somatosensory cortex or PLC (Extended Fig. 5a). To determine whether the degree 200 of AT1aR expression was linked to microglial deramification, we performed RNAScope for AT1aR mRNA 201 in combination with IHC staining against IBA1 (Fig. 5b-f) and we analyzed the peak Sholl values in AT1aR

202
-negative and AT1aR-positive microglia. We found that AT1aR positive microglia were significantly less 203 complex than their AT1aR-negative counterparts, suggesting a potential role of AT1aR in regulation of 204 microglial morphology and functional state (Fig.5e). In addition, we observed a drastic increase (6-fold) in 205 the number of AT1aR mRNA positive microglia in both DH and VH of HF rats (Fig. 5f, g), as well as a 206 significant 1.4-fold increase in AT1aR mRNA expression in hippocampal microglia of HF rats (Extended 207 Fig. 5b). In hippocampal neurons, a 2.8-fold increase in AT1aR mRNA expression was also observed in 208 HF rats (Extended Fig. 5c-e), however, unlike hippocampal microglia, the total number of AT1aR mRNA-209 positive neurons in HF rats did not increase (Fig. 5f). We did not find evidence for AT1aR receptors in 210 hippocampal astrocytes in sham or HF rats (Extended Fig. 5f).

211
Following our findings showing increased cytokine production and microglial ATa1R upregulation, 212 we sought to better understand the time course of these concomitantly occurring phenomena. Thus, we 213 performed qPCR at a very early stage (10 days post HF surgery) for IBA1, cytokines and AT1aR mRNA 214 ( Fig. 6 a, b). We not only found elevated levels of AT1aR mRNA at this stage, but also observed that the 215 increase in AT1aR mRNA preceded the increases in IBA1, TNFa, IL1b and IL6 (Fig. 1). Thus, it is tempting 216 to propose that AT1aR-mediated activation of hippocampal microglia in HF rats contributes to the induction 217 of the observed morphological changes and subsequent induction of microglial cytokine production.

219
Increased number of vessel-associated, AT1aR-positive microglia during HF

220
A recent study showed that during systemic inflammation microglia migrate towards blood vessels 221 and promote BBB stability through Claudin-5 48 . However, upon sustained inflammation, microglia begin to 222 phagocytose astrocytic AQP4-positive endfeet in a CD68-dependent manner, thereby compromising BBB 223 integrity 48 . Inspired by this discovery, we sought to investigate potential changes in hippocampal vessel-224 associated microglia following myocardial infarction. Thus, we combined IHC against IBA1 and AQP4 and 225 analyzed the number of vessel-associated microglia in Sham and HF rats in VH and DH. Intriguingly, we 226 found a stark increase in the number of vessel-associated microglia, accompanied by a decrease in 227 parenchymal microglia (DH and VH combined, Fig. 7a-c). Interestingly, we observed a substantial variation 228 in shape and AQP4-positive contact points of vessel-associated microglia, which prompted us to further 229 characterize them (Fig. 7d, Extended Fig. 6a). We categorized vessel-associated microglia as Type I 230 (microglial filaments contacting the vessel), Type II (partial microglial soma contacting the vessel) and Type 231 III (extensive microglial soma contacting the vessel). We found that the relative incidence of Type I-III 232 microglia varied between Sham and HF rats (Fig. 7e), with Type I predominating in Sham rats but Type III 233 predominating in HF rats. Interestingly, Type III microglia in HF rats frequently protruded into blood vessels 234 (Fig. 7f, Extended Fig. 6b), which is in line with observations recently been made by another group 48 .

235
Vessel protrusion by microglia was much less frequently observed in Sham rats (Extended Fig. 6c). Finally,

236
we performed RNAScope in situ hybridization for AT1aR mRNA in combination with AQP4 and IBA1 IHC, 237 and found that the vast majority of Type III vessel-associated microglia in Sham and HF rats were AT1aR-238 positive ( Fig. 7g-i). Based on these findings, we propose that AT1aRs not only promote microglia activation 239 in the hippocampus, but potentially stimulate their migration towards blood vessels and disruption of BBB 240 integrity.

242
Increased BBB permeability and co-localization of circulating AngII with microglia in HF rats

243
Given the profound changes in microglial migration, hypoxia, angiogenesis and AT1aR upregulation 244 (Figures 4-6), we hypothesized that a compromised BBB could be a plausible mechanism contributing to 245 the concomitantly-occurring phenomena 49,50 . As soon as 2-4 weeks post HF surgery, we found significantly 246 larger leakage of intravascularly delivered FITC10 in the DH, VH and PVN, (but not somatosensory cortex 247 or PLC) of HF rats compared to Sham rats (Fig. 8 a-c, Extended Fig. 7a-d). Three-dimensional 248 reconstruction of blood vessels and extravasated dyes revealed that FITC predominantly accumulated at 249 what seemed to be neuronal/cellular surfaces (Fig. 8b). We previously reported that AngII gains access to 250 the CNS in a hypertensive rat model, and specifically co-localizes with microglia 31 . Thus, we hypothesized 251 that a similar mechanism could underlie the HF-induced, AT1aR-driven neuroinflammation. Indeed, 252 following intravascularly delivered fluorescently-labeled AngIIfluo,, we found a significantly higher degree of 253 IBA1-positive microglia cells displaying bound leaked AngIIfluo in HF rats compared to Sham rats (Fig. 8d, 254 e, Extended Fig. 7e). In addition, we found significant larger amounts of parenchymal AngIIfluo in HF rats 255 than Sham rats (Extended Fig. 7f), again corroborating our findings of increased BBB permeability and 256 subsequent leakage. Interestingly, the average amount of AngIIfluo taken up by HF microglia was 1.8-times 257 higher than in Sham animals (Extended Fig. 7g), which could potentially be explained by the enrichment 258 of AT1aRs (Extended Fig. 5b) in HF microglia. Finally, we aimed to determine whether AngIIfluo-labeled 259 microglia were vessel-associated. To this end, we co-infused animals with Rho70 to label blood vessels and 260 found that the vast majority of AngIIfluo-positive microglia was vessel-associated in HF rats ( Fig. 8f,

264
Based on the results above suggesting a causal link between HF-induced exacerbated microglial 265 AngII signaling, astrocyte activation, elevated cytokine levels and apoptosis (Figs. 1-3), we hypothesized 266 AT1aRs mediated these effects, and thus predicted that blocking AT1R-mediated signaling would 267 ameliorate these pathological alterations. In addition, we were interested to determine whether this 268 treatment could reverse our recently reported cognitive impairments in HF rats 35 . To this end, we randomly 269 allocated HF rats to be given either normal drinking water or water containing the AT1R blocker losartan 31 270 (20 mg/kg/day) for 12 weeks after the myocardial infarction. While losartan did not affect cardiac function 271 (Extended Fig. 8a), it did significantly decrease both systolic and diastolic blood pressure (Extended Fig.   272 8b, c), supporting the efficacy of the treatment 30 . In addition, losartan did not affect weight gain (Extended 273 Fig. 8d) nor total water consumption (Extended Fig. 8e). We found that losartan significantly reduced 274 mRNA levels for IBA1, GFAP, IL-1, IL-6 and TNF-a, but not AT1aR, Hif-1a or Hif-2a in the VH and DH ( Fig.   275 9a, b). We also found that losartan significantly reversed previously observed changes in microglia 276 morphology including cell volume, surface area and filament length in both hippocampal subdivisions 277 ( Fig.9c-f). In addition, losartan almost entirely reversed the previously observed changes in astrocyte 278 morphology (Extended Fig. 8f). Importantly, losartan also significantly diminished cCasp3 staining both in 279 the VH and DH (Fig. 9g, h). In addition, we found than losartan significantly decreased the number of vessel-280 associated microglia (Extended Fig. 8g, h). Finally, losartan almost entirely reversed the earlier described 281 HF-induced shift (Fig. 7) in Type I-III microglia subtypes (Extended Fig. 8i).

282
We recently reported that, compared to sham rats, HF rats displayed signs of spatial and emotional 283 memory impairments 35 both of which are associated with altered hippocampal function 51,52 .To determine 284 whether AT1aR-induced neuroinflammation contributed to these effects, we repeated the spontaneous 285 alternation and inhibitory avoidance tests (IA) in HF and losartan-treated HF rats (Fig. 9i). We found that 286 losartan-treated rats displayed significantly more percent alternations (Fig. 9j), suggestive of improved 287 spatial working memory. Although losartan affected the sequence of arm entries (i.e., percent alternation),

288
it did not affect the number of arms the rats entered in the maze (Fig. 9k), thereby ruling out that the 289 observed behavioral changes were merely a result of changes in activity levels. Losartan-treated HF rats 290 also had significantly higher retention latencies during the IA memory test, suggesting that it improved 291 emotional memory (Fig. 9l). Of note, there were no differences in the training latencies ( Fig. 9m), further 292 supporting the interpretation that losartan did not affect activity levels. Taken together, these findings 293 suggest that blocking AT1R-mediated AngII signaling partially prevents neuroinflammation and subsequent 294 apoptosis, thereby reversing cognitive impairments observed in HF rats.

297
A growing body of clinical studies supports a high degree of comorbidity between cardiovascular 298 diseases and cognitive decline 53-55 . In fact, 20-40% of all HF patients develop major depression and elevated 299 anxiety 2,53,54,56 , along with memory-associated symptoms, which usually appear later than the 300 cardiovascular and autonomic-related symptoms 3 . These cognitive and mood disorders have also been 301 observed in experimental animal models of HF, including the rat and mouse left coronary ligation model 57-302 59 . Still, the mechanisms underlying HF-induced cognitive impairments remain largely unexplored. In our 303 current study, we focused on the hippocampus, given that normal hippocampal function is paramount for 304 mood and memory, and that alterations in hippocampal signaling have been linked to cognitive decline, 305 depression and anxiety 11,60 . In fact, hippocampal shrinkage and cell loss after HF have been observed in 306 rats and humans 13,14 . In addition, cognitive impairment in HF rats has been recently demonstrated by our 307 group 35 and others 36 .

308
Microglia, the resident immune cells of the CNS, are highly dynamic cells that surveil the brain state 309 and respond to injury by migration and substantial changes in cellular morphology 61 , leading to a pro-310 inflammatory state and secretion of cytokines. As part of this process, microglia directly interact with 311 astrocytes via cytokine signaling to further promote the neuroinflammatory cascade, eventually resulting in 312 neuronal damage and death 20,40 . Using our recently developed morphometric profiler to assess microglial 313 morphology at various stages in sham and HF rats 17 , we found significant microglial process retraction, 314 somatic swelling and reduction of surface area, all of which are in line with a pro-inflammatory microglial 315 phenotype 17,62,63 . Intriguingly, we not only observed time-dependent microglia morphological changes during 316 the progression of the disease, but also brain region-specific changes, highlighting potential brain 317 differences in susceptibility and resilience to HF-induced neuroinflammation. Moreover, these changes were 318 dependent on the severity of cardiac compromise in HF. The robust and significant negative correlation we 319 report between cytokine level expression (e.g. IL1b and TNFa) with microglial complexity markers, further 320 support the microglia pro-inflammatory status in the hippocampus of HF rats.

321
A core feature of the neuroinflammatory process is the intricate interaction between microglia and 322 astrocytes. Consistent with our previous findings in the PVN and amygdala 17 , we report here swelling of 323 astrocyte processes in DH and VH, as well as an astrocytic shift from a neuroprotective to a neurotoxic  were AT1aR-positive. Finally, a causal link between AngII-AT1aR and neuroinflammation is more 343 compellingly supported by the fact that treating HF rats with the AT1R antagonist losartan substantially 344 reversed numerous microglial morphometric changes and the elevated cytokine levels observed in non-345 treated HF rats. Importantly, AT1R blockade also significantly diminished hippocampal apoptosis, further 346 supporting a mechanistic role for AngII-AT1R in mediating neuroinflammation and neuronal death in the 347 hippocampus of HF rats. These results are in line with a previous study showing that candesartan (another 348 AT1R blocker) ameliorated brain inflammation following LPS injection 67 . Finally, it is worth highlighting that 349 neuroinflammatory markers were evident only in brain regions where an expression/upregulation of AT1aR 350 was observed (e.g., hippocampus/PVN, but not PLC or SSC). Thus, the brain region specificity of the 351 neuroinflammatory response during HF could be dependent, at least in part, on a concomitant region-352 specific upregulation of AT1aRs in this condition. Clearly, additional work is needed to more conclusively 353 asses this. An important caveat to take into consideration is that losartan blocks AT1R signaling in all cells, 354 not just microglia. Thus, a contribution of AT1R-expressing neurons to neuroinflammation and apoptosis 355 cannot be ruled out entirely. Finally, we observed a losartan-induced reduction in astrocytic soma volume 356 as well as a reduction in GFAP mRNA, even though we found no evidence for astrocytic AT1aR receptors.

357
These results in also in line with previous findings showing that astrocyte signaling in neuroinflammation is 358 downstream to activated microglia 20 .

359
It was recently showed that during systemic inflammation microglia migrate towards blood vessels 360 and promote BBB stability through Claudin-5 48 . However, upon sustained inflammation, microglia begin to 361 phagocytose astrocytic AQP4-positive endfeet in a CD68-dependent manner, thereby compromising BBB 362 integrity 48 . In this context, we found a robust increase in the number of vessel-associated microglia, 363 accompanied by a decrease in parenchymal microglia in HF rats. We categorized vessel-associated 364 microglia as Type I (microglial filaments contacting the vessel), Type II (partial microglial soma contacting 365 the vessel) and Type III (extensive microglial soma contacting the vessel). Importantly, we found a shift from

373
HF, but that this signaling unit could also stimulate their migration towards blood vessels to promote 374 disruption of BBB integrity. To test whether this was the case, we used a well-established approach to 375 quantitatively assess BBB permeability using intravascular infusions of fluorescently-labeled dyes 31 . Our 376 results showed significant BBB leakage in the DH and VH (as well as PVN, but not in the somatosensory 377 cortex or PLC) in HF rats, which was sufficient enough to allow circulating AngII to leak into the hippocampal 378 parenchyma. Interestingly, and in line with a previous study in hypertensive rats 31 we found leaked AngII to 379 be bound predominantly to microglial cells. Taken together, these results, along with the fact that AT1R 380 blockade prevented BBB leakage in HF rats, provide compelling evidence that AngII-AT1aR activation 381 during HF contributes to microglial cell activation and recruitment into a vascular pool, leading ultimately to 382 disruption of BBB integrity. These findings are in line with a recent study from our lab showing the 383 involvement of AT1Rs in BBB disruption in hypertensive rats 69 .

384
Altogether, the above data strongly support an AngII-AT1aR-mediated pro-inflammatory microglia 385 state, an astrocytic shift from neuroprotective to neurotoxic phenotype, neuronal dysfunction and apoptosis, 386 and BBB disruption in HF rats. Importantly, these factors have been shown to contribute to cognitive decline 387 and mood disorders in neurodegenerative conditions 70-72 . We recently reported that, compared to sham rats,

388
HF rats displayed signs of spatial and emotional memory impairments 35 , both of which are associated with 389 altered hippocampal function 51,52 . Importantly, we report here that losartan-treated HF rats showed 390 significant cognitive improvements compared to their non-treated counterparts, as shown by improved 391 performance both during the spontaneous alternation and the inhibitory avoidance testing, while the number 392 of arm entries and training latencies were not affected. These findings are highly relevant from a clinical 393 standpoint and suggest that exacerbated AngII-AT1aR signaling in the hippocampus of HF rats is a pivotal 394 mechanism driving the cognitive impairment observed after HF. We selected to deliver losartan orally, to 395 mimic the conventional clinical route of administration in patients. We acknowledge however that this limits 396 our ability to determine the site of action of this drug to mediate the reported effects. Still, we believe this 397 limitation is mitigated by the fact that AT1aRs were upregulated in very selective brain regions, particularly 398 the hippocampus, a well-established brain region involved in memory and cognitive functions.

399
In addition to neuroinflammation, hypoxia, due to the overall compromised cardiac output and

426
Heart failure surgery and Echocardiography

460
RNAScope reagents were purchased from acdbio (PN320881). Nuclease-free water and PBS were 461 purchased from Fisher Scientific. Brains were processed as described under Immunohistochemistry using 462 nuclease-free PBS, water, PBS and sucrose. We followed the manufacturer's protocol with a few

531
For the assessment of BBB integrity, we performed intra-carotid infusion of two dextran dyes with different 532 molecular weight (FITCH 10kDa and Rho 70kDa) as well as fluorescently-labeled AngII (3µmol/L, Anaspec, 533 CA). Rats were anesthetized with Ketamine/Xylazine (60/8 mg/mL, respectively) and a non-occluding 534 catheter filled with the fluorescent dyes was inserted into the left internal carotid artery as previously 535 described 31 . We used both dyes at 10mg/mL, 2.86µl/g each, with an injection volume of 1mL per rat. We 536 infused the dyes using 0.9% saline and dyes were allowed to circulate for 30 mins. Rats were then 537 decapitated and left in 4% PFA for 48hrs, followed by 30% sucrose for 2-3 days at 4°C. Brains were 538 sectioned using a Cryostat and 40µm-thick sections were mounted for confocal imaging. To detect and 539 quantify the amount of leaked, extravasated FITC 10kDa or AngIIfluo we used a digital subtraction approach, 540 in which we subtracted the vessel-bound FITC10 from the total FITC10 (extravascular + intravascular) in 541 Fiji as described before 31 .

543
Losartan treatment 544 HF rats were randomly allocated two either HF or HF + losartan groups and EF were compared 545 post hoc to verify that there were no significant differences in the severity of the myocardial infarction.

546
Losartan-treated rats received losartan (20mg/kg/day) in the drinking water starting 1 week after the HF 547 surgery until they were sacrificed for analysis 13-weeks post-surgery. Weight gain and water consumption 548 was monitored bi-weekly to guarantee adequate consumption of losartan-containing water and comparable 549 food intake between the groups.

563
Intra-carotid artery infusion for the assessment of blood-brain barrier permeability 564 Carotid artery infusion was performed as in our previous study 31 . Briefly, with adequate 565 ketamine/xylazine anesthesia (80/20 mg/kg/bw, respectively), rats were placed supine on the surgical table 566 and the neck was shaved. A longitudinal incision was made followed by blunt dissection of omohyoid muscle 567 to expose the underlying right common carotid artery (CCA). The vagus nerve and connective tissue 568 surrounding the CCA were carefully dissected. Using 4-0 silk suture, the caudal end of the CCA was ligated and a vascular clamp was applied to the rostral end just above the bifurcation of CCA. An incision was made 570 and a PE50 tubing was inserted in a retrograde fashion, and then 4-0 silk suture was used to stabilize the 571 tubing followed by the removal of vascular clamp. All dyes were infused slowly (1 ml over 2 mins) and 572 allowed to circulate for 30 mins before extraction of the brains. Following post-fixation, we acquired confocal 573 images of brain sections containing somatosensory cortex, DH, VH, PVN and CeA and performed binary 574 reconstruction and subtraction of vessel-confined fluorescent signal of confocal images to assess the levels 575 of extravasated dyes in the brain parenchyma, as we previously reported 31 .

577
TUNEL apoptosis staining 578 TUNEL staining for the identification of apoptotic cells was performed using the Neurotacs TM II In 579 Situ Apoptosis Kit (Trevigen, 4823-30-K) according to the manufacturer's protocol. We used two different 580 approaches (number of apoptotic clusters and area fraction) to measure and quantify the degree of 581 hippocampal cell death (Extended Fig. 3a, b). The number of apoptotic clusters were counted manually 582 using the cell count function in Fiji. Area fraction was assessed using the thresholding function as described with 95% O2/5% CO2. Horizontal slices were cut at 250 µm thickness and placed in a holding chamber 595 containing aCSF bubbled with 95% O2/5% CO2. The aCSF is identical in composition to the sucrose solution, 596 but the 200mM sucrose was replaced with 119 mM NaCl. The slice chamber was warmed using a water 597 bath at 32°C for 20 mins before placement at room temperature for at least 40 mins.

599
Horizontal slices containing the dorsal hippocampus (CA1) were placed into a specimen chamber on the 600 stage of a Nikon Eclipse FN1 microscope and perfused constantly (~3 ml/min) with aCSF bubbled 601 continuously with 95% O2/5% CO2 and warmed to 32°C. CA1 neurons were visualized using the Dragonfly

639
Three days after SA testing, rats were trained in the one-trial IA task, which assesses emotional, 640 long-term memory 78 . All testing was conducted during the light phase between 9:00 am and 12:00 pm and 641 the apparatus was cleaned with 70% ethanol after each rat was tested. For both training and testing, rats 642 were placed in a polycarbonate trough-shaped apparatus (84 x 20 cm x 34 cm) that was divided into a 643 lighted (24 cm) and dark compartment (60 cm) by a retractable polycarbonate guillotine door. The dark 644 compartment had a metal floor through which shock could be delivered. The lighted compartment was 645 illuminated by a 60 W lamp and was the only light source in the room. For training, a rat was placed in the 646 lighted side of the apparatus facing away from the door and the door was lowered when it turned around or 647 after 12 sec passed. As soon as the rat entered the dark compartment with all four paws, the door was 648 closed and a 0.5 mA/1 sec footshock (Coulbourn) was administered, and then the rats was removed from 649 the apparatus 5 sed later. The current level was verified using a digital multimeter (AstroAI AM33D) before, 650 during, and after the experiment. The latency to enter the dark compartment was recorded. The retention 651 test was conducted 24hrs later using the same procedure, with the exception that footshock was not 652 administered. Latency to enter the dark compartment (maximum 600 sec) was recorded and used as an

1238
Ejection fraction of all rats used for this study (151 total, n=56 sham, n=95 HF). c-f Quantification of microglia 1239 volume and filament length via three-dimensional reconstruction reveals no significant differences across   Type I-III vessel-associated microglia in HF and HF + Losartan rats (n=5 per group). Scale bar 100µm (g).