Identi cation and characterization of a non- conventional CD45 negative perivascular macrophage population within the mouse brain

Perivascular macrophages (pvM) are closely associated with cerebral vasculature and play an essential 25 role in drainage of the brain and regulation of the immune response. Here, using reporter mouse models 26 and immunofluorescence on sections and whole brain, flow cytometry and single cell sequencing, we 27 identify a Lyve1 + brain perivascular population lacking classical macrophage markers such as CD45 28 and Cx3cr1. We named the new non-conventional CD45 negative perivascular macrophages pvM2. 29 These cells have a similar location, morphology and phagocytic function as conventional pvM. The 30 pvM2 are not derived from hematopoietic stem cells, as they are negative in the Vav tdT lineage tracing 31 model. They increase in number after photothrombotic induced stroke established by flow cytometry 32 and 3D immunofluorescence analysis. Since CD45 negative cells were typically excluded from 33 macrophage studies, the presence of pvM2 has been previously missed and their role is of importance 34 to assess in the brain disease models. The staining dead after FACS were acquired on the (BD Biosciences). Data analyses were done FlowJo


INTRODUCTION 37
The central nervous system (CNS) has long been considered immune privileged, devoid of immune 38 cells other than microglia and without classical lymphatic vessels. The re-discovery of the dural 39 lymphatic network highlights a route for the drainage of the brain to the periphery 1-6 . How drainage 40 from the CNS parenchyma occurs towards lymphatics and draining lymph nodes is unknown, but there 41 are indications this happens via the foramen at the base of the skull 3 . 42 Besides microglia, several myeloid populations have now been characterized and shown to be essential 43 for brain homeostasis and in brain diseases 7 . These myeloid populations include the so-called non-44 parenchymal or border macrophages which can be classified in the perivascular, subdural meningeal 45 and choroid plexus macrophages. These macrophages are established during development by 46 embryonic precursors derived from yolk sac precursors and are not replaced by blood monocytes 47 Prox1 CreErt2 ;Rosa26 tdT ( Supplementary Fig. 2D) reporter mouse models 24,34 . We analyzed the 111 Prox1 CreErt2 ;Rosa26 tdT brain 2 weeks after Tamoxifen injection and did not observe Prox1 + lymphatic 112 endothelial cells, ruling out any LEC identity within the brain parenchyma ( Supplementary Fig. 2B). 113 While the dura mater contains lymphatic vessels, no conventional lymphatics have been described 114 within the parenchyma nor the leptomeninges. Similarly, while we did not observe luminized 115 lymphatic vessels within the pia mater, we did find single Lyve1 + Prox1 + lymphatic endothelial cells in 116 the pia mater ( Supplementary Fig. 2B-C) as was also recently described in mammals 35 . 117 To further rule out astrocyte, glial or fibroblast identity, we analyzed immunofluorescence staining for 118 Aquaporin-4 (AQP4) and GFAP ( Supplementary Fig. 2D), ER-TR7 ( Supplementary Fig. 2E), 119 PDGFRβ ( Supplementary Fig. 2F) and observed these cells to be negative for all. Furthermore, we 120 excluded neural crest cell origin using the Wnt1 Cre ; Rosa26 tdT reporter mouse model (Supplementary 121 cytometry. Using Cx3cr1 GFP brain parenchyma cell suspension devoid of meninges, conventional pvM 125 were identified as Lyve1 + CD45 int/high while microglia were excluded as they are Lyve1 -(gating strategy 126 in Supplementary Fig. 3A). These Lyve1 + CD45cells also lacked CX3CR1, confirming IF section 127 stainings (Fig. 1). After 2 days after birth (P2) we observed a Lyve1 + CD45 -CX3CR1 -F4/80 + 128 population which peaked at P14 and P21 (Fig. 3A-D). Subsequently, we observed lowered numbers in 129 adult and 1-year old brain parenchyma ( Fig. 3E-F, absolute Lyve1 + CD45 -CX3CR1 -F4/80 + cell 130 numbers in Fig. 3G). Also in the Spi1 GFP (encoding PU.1) brain parenchyma we observed a CD45 -131 PU.1 -F4/80 + population, next to the conventional CD45 int/high populations ( Supplementary Fig. 3C), 132 albeit less Lyve1 + CD45 -PU.1 -F4/80 + cells were present compared to the Lyve1 + CD45 -CX3CR1 -133 population observed in the Cx3cr1 GFP brain parenchyma. 134

135
We confirmed absence of Ptprc (encoding CD45) expression within the Lyve1 + CD45cells by sorting 136 this population and determination of Ptprc mRNA levels by qPCR and observed no Ptprc expression 137 within the Lyve1 + CD45population (Fig. 4A, sorting strategy in Supplementary Fig. 4A). To further 138 characterize the transcriptome of the pvM population, we analyzed the single cell RNA sequencing 139 dataset of the non-neuronal cell population in mouse brain cortex (GSE133283), thus including the 140 CD45cells (Fig. 4B, C). We identified macrophages expressing the characteristic macrophage 141 markers Mrc1, Cd68 and Fcgr3 (Fig.S4B). This macrophage cluster was segregated into at least two 6 lower or undetectable Ptprc or Cx3cr1 expression and fewer cells with lower Spi1 expression (Fig.4C). 144 However, these cells were not segregated from other cells by unsupervised clustering analysis, 145 confirming their identity as macrophages. 146 Since we ruled out a lymphatic identity, established that the cells expressed macrophage markers and 147 clustered together with the conventional pvM in the single cell sequencing, we name these cells non-148 conventional CD45perivascular macrophages (pvM2). 149 PU.1 is considered to be a master regulator for macrophage differentiation. Therefore, we analyzed the 150 Spi1 GFP/GFP mouse model, in which GFP is knocked into the Spi1 locus (encoding PU.1) rendering this 151 gene inactive. Homozygous mice deficient for Spi1 lack all macrophages and die around birth 8,27 . In 152 E18.5 Spi1 GFP/+ embryos, we identified Lyve1 + PU.1 + CD45 + and Lyve1 + PU.1 -CD45 -pvM (pvM2) 153 We tested a possible hematopoietic stem cell (HSC) -bone-marrow derived hematopoietic origin of 158 this population using the Vav Cre ; Rosa26 tdT (hereafter named Vav tdT ) reporter mouse model. Vav 159 expression starts around embryonic day 8.5 (E8.5) in the hemogenic endothelium of the dorsal aorta in 160 aorta-gonad-mesonephros (AGM) region, which gives rise to the HSC and monocytes 36 . By 161 immunofluorescence staining of Vav tdT brain sections, we identified a Lyve1 + CD45 + Vav-lineage + and 162 a Lyve1 + CD45 -Vav-lineagepopulation (Fig. 5C). Therefore, we concluded that the Lyve1 + CD45 -163 cells were not derived from HSC or bone-marrow derived monocytes. 164

Phagocytosis by pvM2 165
Besides their role as source for chemokines and growth factors to regulate an immune response, pvM 166 have a phagocytic function, which could be part of a broader role of these cells for tissue homeostasis 167 19,37 . In order to address a role of the pvM2 in fluid drainage or macromolecule clearance, we injected 168 10kD Dextran-AlexaFluor647 and Acetylated LDL-AlexaFluor594 in the lateral ventricle of 169 Cx3cr1 GFP mouse brains. To exclude recirculation through the blood stream, we analyzed the brain 10 170 minutes after injection 3 . We observed that both Lyve1 + CX3CR1and Lyve1 + CX3CR1 + cells 171 phagocytized the injected dyes near the ventricle, in the superior-( Fig. 6A-C) and in the inferior cortex 172 ( Fig. 6D-F). We concluded that Lyve1 + CX3CR1cells were able to efficiently take up lipoproteins and 173 glycoproteins in the range of at least 3 to 10kDa similar as conventional pvM. The close proximity of 7 injection-site within the lateral and connected 3V towards the pia mater (Fig. 6G). Very few Lyve1 and 176 Prox1mOrange2 positive nuclei were observed within the pia mater (Fig. 6G). 177

Involvement in central nervous system diseases 178
It was shown that in stroke pvM can influence the blood-brain-barrier function 38,39 . In patients with 179 intracerebral hemorrhage and focal cerebral ischemia, a CD163 + cell accumulation around brain blood 180 vessels was observed, which also contained myelin 40 . Moreover, it was not clear whether these cells 181 are blood-derived monocytes or pvM since they were not characterized in detail (Holfelder K. et al., 182 2011). 183 To assess the role of the pvM2 in cerebrovascular pathologies, we investigated a photothrombotic 184 model of ischemic stroke. PT was induced on 8 weeks old male mice. In this model, we observed an 185 increase in Lyve1 staining density in whole-mount stained brains (Fig. 7A, method for quantification 186 of staining density in whole mount brain shown in Supplementary Fig. 6A). At day 14 post stroke 187 induction (P14), we noticed a significant increase of Lyve1 staining density (4.2x, p<0.05) within the 188 hippocampus, although no significant difference was observed in the superior cortex. At P30, Lyve1 189 staining density had normalized to control values again (Fig. 7B). We noted an increase in the total 190 Lyve1 + population, including CD45 int and CD45population by flow cytometry (Fig. 7C) absence of labeling in the Vav1 tdT model, we conclude that pvM2 are not bone-marrow derived 207 monocytes. As was shown for pvM in the periphery 19 , pvM2 are also most likely derived from the 208 yolk-sac. Using the Spi1 reporter (encoding the PU.1 protein) and knock-out model, we demonstrate 209 that, similar to myeloid cells, they require PU.1 for differentiation but not for maintenance. They could 210 still have the Spi1 mRNA transcripts, but not he protein or GFP reporter, as we observed more cells 211 positive for Spi1 than for Cx3cr1 or Ptprc in the single cell sequencing analysis. It was previously 212 noted that some macrophages originated from CX3CR1 + precursors but ceased to express Cx3cr1, such 213 as alveolar macrophages, Langerhans and Kupffer cells. These CX3CR1cell coincided with 214 CX3CR1 + cells 42 , as we also observed in the brain parenchyma. Why pvM2 lacked Csf1R and thus 215 likely do not require Csf1 for maintenance is yet unclear. Further experiments using other lineage 216 tracing models are required to reveal their exact origin in relation to the differentiation of conventional 217 pvM. It was observed before that non-hematopoietic ectodermal cells gave rise to phagocytosing cells, 218 resembling Langerhans cells, in the skin of zebrafish 43 . However, it was not shown that these cells 219 depend on PU.1 and moreover were located within the skin of the zebrafish. Whether these cells 220 resemble the murine pvM2 is yet unclear. 221 In adult tissues at steady-state, pvM have important functions related to their perivascular location, 222 such as the regulation of vascular permeability, phagocytosis of blood-transmissible pathogens, antigen 223 presentation or immune regulation 37 . Here, we observed similar phagocytosis of the pvM2 when 224 compared to conventional pvM. Previously, two perivascular monocyte populations were described, 225 being Lyve1 hi or Lyve lo . MHCII was expressed high on the Lyve1 lo population while the Lyve1 hi 9 only from CD45 + cells. CD45 negative non-conventional pvM, if present, were disregarded. Since the 228 pvM2 did not express CD45, MHCII nor CX3CR1, but were Lyve1 + , it is not clear how these cells 229 relate to the previously described perivascular monocytes. 230 Using whole-mount imaging enabled us to establish the pvM contribution specifically within and 231 adjacent to the affected region in ischemic stroke. We established an increase in the total pvM number 232 after stroke, with a concomitant significant increase of the pvM2 population. As the number of the 233 pvM2 increases during stroke, these cells may be involved in the generation of an activated immune 234 status. All previous studies on pvM within the brain analyzed CD45 + cells. This study now for the first 235 time presents evidence of a CD45 -CX3CR1population, named pvM2, which is present within the 236 brain and should be considered in the future to better understand the active role of diverse macrophage 237 populations in brain homeostasis and disease. 238

239
In conclusion, here we demonstrate the existence and function of non-conventional CD45 -pvM2, 240 lacking classical macrophage makers such as CD45, CX3CR1, PU.1, CD206, CD163 and CD11b. 241 These cells are consistently observed within the brain and are likely involved in drainage also due to 242 their blood vessel proximity and involved in an activated immune state such as during ischemic stroke. 243 Although these novel pvM2 increase in number during stroke, their specific role during disease remains 244 unclear. Presently, there are no unique markers available to target these cells specifically, but a newly 245 developed marker should allow a targeted approach to study their specific role in brain physiology in 246 the future.

Cell sorting 278
Cells were prepared and stained with CD45, and Lyve1 antibodies and Fixable NIR as describe above. 279 Live Lyve1 + CD45 -, live Lyve1 + CD45 int and live Lyve1 + CD45 high cells were sort-purified separately 280 using the BD FACS ARIA III SORP sorter. 281

RNA isolation and gene expression analysis 282
After spinning down (7min, 300g, 4°C), cells were resuspended in 1mL TRIzol reagent (Sigma T9424). 283 The suspension was transferred to a phase lock gel heavy tube to facilitate the extraction and 200µL of 284 chloroform was added. The samples were shaken vigorously and incubated for 3 min at room 285 temperature (RT) prior to centrifugation (10min, 12 000g, 4°C). The clear aqueous phase was 286 transferred to a new tube and 2µL of GlycoBlue Coprecipitant (Invitrogen AM9515) was added to help 287 with the isolation prior to adding 500µL of isopropanol. The samples were incubated for 10min at RT 288 and subsequently centrifuged (30min, 12 000g, 4°C). The supernatant was discarded and RNA was 289 washed with 800µL EtOH 70%. The samples were centrifuged (5min, 12 000g, 4°C) and supernatants 290 were discarded. The RNA was air-dried for 15min prior to resuspension in 11.5µL RNase free water. 291 The tubes were subsequently heated at 65°C for 10min and then put on ice.  Fig. 6). 336

Vibratome section immunofluorescence staining and confocal imaging 337
Animals were perfused with PBS/heparin, brains dissected and fixed ON in 4% PFA in PBS at 4°C 338 and subsequently embedded in 1% low melting agarose for generation of 100µm vibratome slices 339 (Leica, VT1000S). Sections were blocked in EBT buffer (EBSS, 0.05% Tx, 1% BSA) containing 10% 340 serum for 2h at RT under agitation. Immunostainings were performed by incubation in primary 341 antibodies for 48h at 4°C in EBT, 3% sera and subsequently with Alexa-fluorochrome coupled 13 Histodenz (Sigma D2158) medium for 48h at RT and subsequently mounted in Histodenz medium. 345 Confocal images were acquired at RT on a confocal microscope (LSM880, Zeiss, Germany), with a 346 20x/0.4 Plan-Apochromat objective and using laser lines at 405, 488, 561, and 633nm for the excitation 347 of AlexaFluor405/GFP/AlexaFluor555/AlexaFluor647 respectively. Fluorescence was recorded in 348 individual channels acquired in a sequential mode using a highly sensitive 32-channel GaAsP detector. 349 Channels were respectively detected using these detection bands: and (D-F) Acetylated-LDL-Alexa594 (42µm maximum intensity projection). Ten minutes after 561 injection, mice were sacrificed and brains dissected. Confocal analysis on brain sections shows Lyve1 562 (white), Dyes (red) and CX3CR1 (green). The pvMs phagocytosed the dyes (red arrows), as well as 563 did conventional pvM (white arrows) (B, C, E and F). (G) Tiled confocal acquisition of a Prox1 mOrange2 564 brain section. Lyve1 + cells in green lined blood vessels stained for CD31 in blue. Lyve1 + cells in green 565 penetrated the brain to the hippocampal fissure (hif) and all the way towards the third ventricle (3V).