Aging blood factors promote CD8+ T cell infiltration in the adult mouse brain


 Aging coincides with major changes in brain immunity that aid in a decline in neuronal function. Here, we postulate that systemic, pro-aging factors contribute to immunological changes that occur within the brain during aging. To investigate this hypothesis, we first characterized the immune landscape upon aging in 20-month-old mice using cytometry by time-of-flight (CyTOF) and observed that memory T cells expanded in the circulation and that specifically effector CD8+ T cells expressing programmed cell death protein 1 (PD-1) and tissue-resident memory CD8+ T cells accumulated in the aged brain. Injections of plasma derived from 20-month-old mice into 5-month-old receiving mice decreased the frequency of splenic and circulating naïve T cells, increased memory CD8+ T cells and non-classical, patrolling monocytes in the spleen, and elevated levels of regulatory T cells and non-classical monocytes in the blood. Finally, we found an accumulation of CD8+ T cells within the brain parenchyma of plasma-treated mice, which coincided with expression of vascular cell adhesion molecule 1 (VCAM-1), a mediator of immune cell trafficking, on the brain vasculature and increased numbers of proliferating microglia. Taken together, our data highlight a role for CD8+ T cells in the aged brain and suggests involvement of age-associated systemic factors that mediate CD8+ T cell migration and expansion in the brain parenchyma.


Introduction
During aging, the immune system becomes less e cient in mounting an effective response against pathogens and harmful processes such as the accumulation of senescent cells, misfolded proteins and cellular degeneration. This decline in immune function, called immunosenescence, has also been associated with age-related neurodegenerative diseases like Alzheimer's disease (AD) and AD-related dementias 1 and goes hand-in-hand with an increase in basal levels of in ammatory factors, a phenomenon known as in ammaging 2 . The brain its immune system mainly consists out of its own unique immune cells, microglia, and harbors relatively low numbers of peripheral immune cells as T, B and natural killer (NK) cells, neutrophils, macrophages, dendritic cells (DCs) and other myeloid-derived cells (MdCs) 3,4 . During aging however, peripheral immune cells increasingly in ltrate and expand in the brain 4 . Memory T cells 5 and speci cally CD8 + T cells 6,7,8 accumulate in the brain during aging where they modify microglia homeostasis and can potentiate in ammation and leukocyte recruitment. Interestingly, CD3 + T cells found in the aged brain are clonally expanded and exhibited higher interferon-y signaling than their peripheral counterparts, indicating antigen experience 9 . In AD patients, CD8 + T cells have been increasingly found in the cerebral spinal uid and brain parenchyma 10,11 which coincided with higher numbers of circulating effector memory CD8 + T cells that negatively associated with cognition 11 . Next to memory T cells, a subset of circulating B cells, termed age-associated B cells, accumulates in the elderly as well 12 and B cell depletion strategies ameliorated disease progression in an AD model 13 . Together, these studies provide strong evidence that aging and age-related diseases as AD are associated with changes in adaptive immune cells in both the circulation and the brain.

Local and distant immune communication is tightly controlled through cell-cell contacts and via
numerous soluble systemic factors like cytokines, hormones and growth factors, among others. During the human lifespan, the plasma proteome shows three waves of changes, re ecting aspects of aging of different cell types and tissues 14 . In aged mice, blood plasma contained higher levels of pro-in ammatory cytokines and chemokines which affect brain function 15 . Parabiosis experiments have shown that an aged systemic environment negatively impacts neurogenesis, cognitive function and activates microglia of young mice 15,16 . Contrariwise, plasma derived from young and physical active mice is able to rejuvenate multiple tissues including the brain [17][18][19] and rescue neuroin ammation and neurodegeneration in a mouse model for AD and acute brain in ammation 17,20 .
Still, despite efforts made to dissect underlying age-related immune changes that contribute to neurodegeneration, parts of the puzzle remain incomplete. Here, we pro led immune cells using cytometry by time-of-ight (CyTOF) of the blood and brain of old and young-adult mice and investigated if exposure to plasma derived from old mice induces age-related immunological changes in younger adult mice. Altogether, our data show that normal aging changes both the immune landscape of the brain and the periphery and these changes can be (partly) mimicked by exposing younger mice to plasma of aged mice.

Results
Aging changes the peripheral immune landscape We rst determined the impact of aging on circulating immune cell populations within the blood of old mice (20 months) by comparing it to younger adult mice (6 months) (Fig. 1a). Immune cells isolated from the blood were stained with a 37-heavy metal isotope-tagged antibody panel for mass cytometry containing surface and nuclear markers to de ne immune cell populations (Table 1). Data were visualized in a UMAP (Fig. 1b) and cells were appointed to clusters by Phenotyping by Accelerated Re ned Community 21 (PARC). Finally, PARC-derived clusters were manually merged into biologically relevant immune cell subsets (Fig. S1a).
B cells increased in frequency at the expense of all myeloid cell populations (e.g. monocytes, neutrophils, eosinophils and myeloid-derived cells (MdCs)) in the circulation of old mice (Fig. 1c) and expressed higher levels of CX3CR1 (Fig. S1b). CD8 + T cells as a percentage of CD45 + immune cells did not change, while the proportion of CD4 + T cells decreased (Fig. 1c). Within myeloid cells we detected a lower frequency of classical, in ammatory Ly6C high monocytes while eosinophils took up a larger proportion (Fig. 1d). Among T cells, CD8 + T cells increased in percentage whereas CD4 + T cells decreased in old mice (Fig. 1e). More in depth phenotyping of T cells (Fig. 1f-g) showed an increase in CD4 + and CD8 + memory T cells expressing CD44 + CD127 + at the expense of CD44 -CD127 + naïve populations and an increase in CD8 + effector T cells de ned as CD44 + CD127 - (Fig. 1h). Memory CD8 + T cells (CD8 + mem), the most abundant circulating T cell population in old mice, PD1 + KLRG1effector CD8 + T cells (CD8 + eff1), Ly6Cmemory CD4 + T cells (CD4 + mem1) and NKT cells were increased old mice, while the proportion of CD44 low KLRG1 -CD103quiescent regulatory T cells (Tregs; CD4 + treg1) and of all gamma delta T cells (γδ T cells) decreased (Fig. 1i). Altogether, the aged peripheral landscape is marked by a high proportion of B cells and a decline in myeloid subsets. Within the T cells, effector and especially memory populations were increased at the expense of naïve, regulatory and γδ T-cell subsets.
Aging changes the brain immune landscape Next, we investigated how the brain immune landscape changed with aging. For this purpose, immune cells were isolated from the brain of old (20 months) and younger adult (6 months) mice, stained with our CyTOF antibody panel and subsets were characterized ( Fig. 2a-b; Fig. S1c). Upon aging, the absolute number of non-microglial immune cells quadrupled, while numbers of microglia remained the same (Fig.  2c). Consequently, every non-microglial immune cell type we were able to detect was enriched in the total pool of immune cells (Fig. S1d). Among the non-microglial immune cells in the aged brain, speci cally B cells, plasma cells and neutrophils accumulated (Fig. 2d). In addition, B cells derived from the brain of old mice expressed higher levels of LFA1, CD5 and CD38 (Fig. S1e). Concomitantly, a subset of borderassociated macrophages (BAMs), which include macrophages resident in the perivascular spaces of brain vessels, lining the meninges, and the choroid plexus 4 and DCs decreased in old mice together with NK cells (Fig. 2d). BAMs can be further subdivided based on CD38 and MHC class II (MHCII) expression 4 . Imputing the subset of BAMs and DCs for further sub-clustering revealed different subsets of BAMs within the brain (Fig. 2e-f) and a shift towards more CD38 + MHCII + BAMs in the aged brain at the expense of CD38 -MHCII + BAMs (Fig. 2g). Conventional DCs (cDCs) can be divided into two major subsets called cDC1s and cDC2s. cDC1s engage in cross-presentation or the priming of naive CD8 + T cells through MHCI, while cDC2s are specialized in co-stimulation and induce naïve CD4 + T cells through presentation on MHCII. Within the pool of DCs, cDC1s were increased in the aged brain, while cDC2s decreased and the percentage of monocyte-derived DCs (moDCs) remained unchanged (Fig. 2h). Together, these data show that non-microglial immune cells accumulate in the brain during aging, and that the frequency of B cells, plasma cells, neutrophils, BAMs, DCs and NK cells changes among the brain-immune in ltrates.
Subsets of PD1 + effector and memory T cells accumulate in the aged brain Next, we investigated if aging changed dominancy of certain CD8 + and CD4 + subsets in the brain. Brie y, T cells were re-clustered, and naïve, effector and memory phenotypes were assigned when possible ( Fig. 2i-j). With aging, the proportion of CD8 + T cells increased in the brain as a percentage of all T cells (Fig.   2k). Among CD4 + T cells, naïve cells decreased while the fraction of memory CD4 + T cells expanded (Fig.   2k). A closer look into T-cell subsets within CD8 + , CD4 + and double-negative (DN) T cells showed a striking expansion of CD8 + PD1 + effector T cells (CD8 + eff3) in the brain from 2.2% ± 1.5% in adult mice to 18.6% ± 3.6% in old mice (Fig. 2l). This 'CD8 + eff3' effector population was very similar to the PD1 + effector CD8 + T cells found increased in the blood of old mice before (CD8 + eff1; Fig. 1i). The aged brain harbored higher levels of CD103 + tissue-resident memory CD8 + T cells (CD8 + trm), Ly6C -CX3CR1memory CD4 + T cells (CD4 + mem3) and ICOS + PD1 + Tregs (CD4 + treg1; Fig. 2l). In addition, the frequency of Ly6C -KLRG1 -PD1effector CD8 + T cells (CD8 + eff4), Tbet -PD1effector CD4 + T cells (CD4 + eff2), CX3CR1 high Ly6Chighly polarized cytotoxic effector memory CD4 + T cells (CD4 + mem1) and Ly6C + DN T cells (DN2) was decreased in the aged brain (Fig. 2l). In sum, the aged brain re ects partly age-associated changes in circulating immune subsets, including the expansion of PD1 + effector CD8 + T cells, memory T cells and the decline in naïve T-cell subsets, but in contrast to the blood, also harbors an increased frequency of ICOS + PD1 + regulatory T cells.
Aged plasma increases Ly6C low PDL1 + non-classical monocytes and memory CD8 + T cells in the periphery We show that aging rearranges the composition of immune cells in the circulation and that the brain re ects parts of these changes. Next, we hypothesized that exposure to age-associated systemic factors might drive these shifts in peripheral and central immunity. To investigate this, 5-month-old mice were exposed to blood plasma derived from 20-month-old mice (termed: 'aged plasma') biweekly for 4 weeks (Fig. 3a). Then, immune cells were isolated from the spleen, the largest blood ltrating secondary lymphoid organ, and from the blood, where after immune cell populations were de ned using our CyTOF panel (Fig. 3b-c; Fig. S2a-c). Treatment with aged plasma increased the frequency of splenic and circulating Ly6C low PDL1 + non-classical monocytes (Fig. 3d), also known as non-classical patrolling monocytes, but not that of other immune populations ( Fig. S2d-f). Among peripheral T cells (Fig. 1e- S3a-d), exposure to aged plasma decreased the frequency of naïve CD8 + T cells in both spleen and blood, and of naïve CD4 + T cells in blood only ( Fig. 3g-h), mimicking our earlier ndings in old mice (Fig. 1h). In the spleen, the decrease in naïve CD8 + T cells was concomitant with an increase in memory CD8 + T cells ( Fig. 3g). Speci cally, aged plasma increased a population of splenic Ly6C + Tbet dim memory CD8 + T cells that expressed co-inhibitory molecule TIGIT (CD8 + mem2; Fig. 3i), while CD44 low KLRG1 -CD103quiescent Tregs increased in the circulation (CD4 + treg2; Fig. 3j). The latter nding in contrast to the circulation of aged animals, where this regulatory T cell population was decreased (CD4 + treg1; Fig. 1i). Together, these ndings show that aged systemic factors increase levels of non-classical monocytes in the spleen and circulation, and raise levels of splenic memory CD8 + TIGIT + T cells and of circulating quiescent Tregs.
Aged plasma induces CD8 + T cell and plasmacytoid DC accumulation in the brain Next, we set out to nd if aged plasma also induced, either directly or indirectly, immune changes within the brain parenchyma. First, immune cells were isolated from the brains of aged plasma-treated mice and phenotyped using CyTOF ( Fig. 4a-b; Fig. S4a). While the ratio between microglia and non-microglial immune cells was not affected (Fig. 4c), we found that the frequency of CD3 + CD4 -T cells, presumably CD8 + T cells, doubled in the brain of aged plasma-treated mice (Fig. 4d). CD3 + CD4 + T cells and other subsets of non-microglial immune cells derived from the initial clustering did not change ( Fig. 4d; Fig.  S4b). Deeper phenotyping of BAMs, DCs and T cells revealed however an increase in plasmacytoid DCs (pDCs; Fig. 4e-g), which are prominent secretors of type 1 interferons (IFNs). Among CD3 + T cells, further subsetting revealed an increase Ly6C + CD4 -T cells that expressed LFA1 ( Fig. 4h-i), a common marker of effector/memory T cells. Immunohistochemical analysis con rmed the in ltration of CD45 + immune cells and CD45 + CD8 + T cells in the brain parenchyma after aged plasma exposure, while numbers of CD45 + CD4 + T cells were less affected ( Fig. 4j-k). Tra cking of immune cells into the brain is regulated by cell adhesion molecules on the brain vasculature such as VCAM-1. In line with previous ndings 16 , we found that aged plasma increased the number of VCAM-1 + vessels in young-adult animals ( Fig. 4l-m), whereas levels of ICAM-1 and PECAM-1 were not changed ( Fig. S4c-h). Higher levels of CD45 + immune cell in ltrates in the brain parenchyma correlated modestly (r = 0.65), but on a trend level (P = 0.08), to the increase in VCAM-1 + vessels in the plasma-treated animals (Fig. 4n) and moderate or not to levels of PECAM-1 and ICAM-1, respectively (Fig. S4e, h). Aged plasma did not affect vascular permeability, as assessed by brinogen, a blood coagulation protein deposited in the brain upon vascular dysfunction ( Fig. S4i-j), nor vessel density (Fig. S4l). In addition, vascular permeability was not related to increased numbers of CD45 + immune cell in ltrates (Fig. S4k). Altogether, these ndings indicate that pro-aging factors in plasma increase speci cally pDCs and CD8 + T cells in the brain and might facilitate immune cell tra cking of T cells into the brain via VCAM-1 expression on the brain vasculature.

Aged plasma increases the number of proliferating microglia but does not cause synaptic loss
To investigate if aged plasma would also affect the number and phenotype of microglial and neuronal cells, animals were injected for 3 consecutive days with 5′-bromo-2′-deoxyuridine (BrdU), which is incorporated into DNA during S-phase, starting the day after the last aged plasma injection. Interestingly, we found that aged plasma increased absolute numbers of cortical BrdU + IBA1 + proliferating microglia cells ( Fig. 5a-b) but did not raise levels of BrdU + IBA1 + microglia in the hippocampus (Fig. 5c). Neither did aged plasma affect the proliferation of IBA1cortical or hippocampal cells (as astrocytes, neurons, oligodendrocytes, endothelial cells, etc.; Fig. 5b-c), change numbers of BrdU + cells in the meninges (Fig.  5d) or affected the overall microglial density ( Fig. 5e-g). Aged plasma exposure did however decrease microglial expression of homeostatic marker P2Y12 in the hippocampus ( Fig. 5h-i), while microglial morphological features were unaffected ( Fig. 5j-l). Aged plasma has shown to induce memory de cits in adult mice before 15 , however, we did not nd a loss of Homer1 + post-synapses in the stratum radiatum of the hippocampus (Fig. 5m-n), nor a change in synaptic pruning by microglia (Fig. 5o) or in microglial lysosomal size (Fig. 5p. In addition, we did not observe cortical neuronal loss ( Fig. 5q-r). Altogether, these data show that aged plasma increases the number of proliferating cortical microglia and reduced hippocampal microglial P2Y12 expression, but does not impact the loss of cortical neurons nor changed the amount of hippocampal post-synapses, or their pruning by microglia.

Discussion
We here systematically assessed age-associated changes in the immune landscape of the mouse circulation and brain and investigated whether exposure to aged plasma could induce similar changes in younger adult mice. We observed that the immune pro le of aged mice is dominated by B cells and memory T cells and that circulating and brain-resident T cells of old animals show an exhausted phenotype, indicated by increased expression of PD1. Exposure of young-adult mice to plasma from aged animals mimicked the decrease in naïve T cells and increase in memory T cells in the periphery, and raised numbers of CD8 + T cells within the brain parenchyma. The latter coinciding with increased VCAM-1 on the brain vasculature. In addition, exposure to aged plasma increased the fraction of splenic and circulating non-classical monocytes, of pDCs amongst brain-derived immune cells and augmented numbers of proliferating cortical microglia. Altogether, our data provides an in-depth characterization of immune cells, particularly T-cell subsets, present in the circulation and brain of old mice and indicates that systemic, pro-aging factors drive, at least partly, CD8 + T cell in ltration in the aged brain. In addition, we establish new and validate previously found changes associated with aging in other immune subsets (e.g. B, NK, DC and myeloid cells).
Immune cells from the circulation and brain of aged animals showed a decline in the CD4 + :CD8 + ratio, particularly due to increased numbers of memory CD8 + T cells. Recent studies show that CD8 + T cells are also elevated in the aged human brain 7,8 and even more in the circulation and brain of AD patients 10,11 .
Speci cally, tissue-resident memory T cells, which have been found in the elderly human brain before 5 , accumulated in the aged mouse brain. We also found PD1 + effector CD8 + T cells that highly expressed CD38 to be increased in the circulation and brains of old mice. CD38 + T cells are associated with recent activation 22 and lower cytotoxic capacity 23 , while the PD1 high subset is associated with CD8 + T cell dysfunction and exhaustion 24 . In addition, a population of ICOS + PD1 + Tregs expanded in the aged mouse brain. ICOS expression marks an activated regulatory subset with strong inhibitory capacity 25 , but in combination with PD1 expression also associates with Treg exhaustion and increased IFN-γ secretion 26 . Future research should reveal whether high levels of PD1 + effector CD8 + T cells and exhausted Tregs in the aged brain present dysfunctional T cells that are unable to yield effective T cell responses and leave the brain vulnerable to external stressors. In addition, we found that the frequency of B cells and plasma cells highly increased with age in both the circulation and brain. Circulating B cells of aged mice showed increased expression of CX3CR1, which has been previously linked to B-cell mediated immunosuppression 27 , while B cells in the aged brain expressed higher levels of LFA1, CD5 and CD38, which associate with B cell receptor (BCR) signaling and B cell activation 28-30 . Taken together, our data shows that B cells and memory T cells are increasingly present in the aged mouse blood and brain, and that different T-cell subsets show signs of exhaustion with high expression of inhibitory PD1.
Aged blood harbors increased levels of immune-derived cytokines and chemokines that have detrimental effects on the brain 15 . In an attempt to uncover if blood-borne factors present in the systemic milieu can induce the age-dependent immune changes found before, we repeatedly exposed young-adult mice to aged plasma for a duration of 4 weeks. These repeated injections led to an overall decline in circulating CD4 + and CD8 + naïve T cells and of naïve splenic CD8 + T cells while splenic CD4 + T cell populations did not change. Exposure to aged plasma also increased memory CD8 + T cells in the spleen, of which speci cally Tbet dim memory T cells that also expressed Ly6C and TIGIT. Ly6C is expressed by most central memory CD8 + T cells 31 which provide central immune immunosurveillance, sustain immune responses by proliferation in secondary lymphoid organs like the spleen, have a long-term survival and are suppliers of new effector memory cells 32 . TIGIT is a co-inhibitory molecule that can protect T cell recall responses by restricting terminal differentiation 33 and was upregulated on T cells from aged mice who exhibited features of exhaustion 34 . In contrast to our old mice however, aged plasma increased levels of circulating naïve CD103 − KLRG1 − Tregs as well. Finally, aged plasma exposure corroborated the age-associated reduction in Ly6C high /Ly6C low monocyte ratio by elevating levels of circulating and splenic Ly6C low PDL1 + non-classical monocytes. Non-classical monocytes engage in long-term migration, also called patrolling, along the endothelium and often exert anti-in ammatory and homeostatic effects [35][36][37] . Together, our data shows that exposure to aged plasma mirrors some age-associated peripheral immune changes, including an overall decline in naïve T cells, an increase in memory CD8 + T cells that upregulated expression of co-inhibitory molecules and a reduced Ly6C high in ammatory to Ly6C low patrolling monocyte ratio.
Interestingly, exposure to aged plasma also elevated numbers of LFA1 + Ly6C + effector/memory CD8 + T cells in the brains of plasma-treated mice. Due to the low number of cells in this cluster analysis, we were not able to make a distinction between effector or memory phenotypes. Interestingly, the increase of CD8 + T cells in the brain parenchyma associated with increased VCAM-1 levels on the brain vasculature. The latter nding con rming earlier data in which plasma from both old mice and humans increased VCAM-1 expression on brain endothelial cells 16 . VCAM-1 binds to VLA-4, which is expressed by leukocytes and is used for tethering, rm adhesion and transmigration of leukocytes into the brain. We did not nd upregulation of other adhesion molecules such as ICAM-1 or PECAM-1. Equivalent, others also did not detect increased levels of ICAM-1, E-selection and P-selectin on the brain vasculature after exposure to aged plasma nor during normal aging 16 . Fibrinogen leakage into the brain, which is indicative of brain vasculature leakage, was also not increased after aged plasma exposure. As such, aged plasma either directly or indirectly via ongoing in ammatory responses, induces VCAM-1 expression on the brain vasculature which associates with more effector/memory T cells in the brain parenchyma. In addition, we found that pDCs were increased in the brains of plasma-treated mice. pDCs are the main source of type I IFNs and directly impact T cell responses through antigen presentation. Interestingly, a type 1 IFN signature in the brain was associated with aging before and negatively affected brain function 38,39 .
Recent research shows that CD8 + T cells induce phenotypic changes in microglia and neurons 40 and are involved in the killing of neurons [41][42][43] , inhibition of neural stem cell proliferation 9 , synaptic elimination 44,45 and associate with a decline in synaptic plasticity and cognition in AD pathology 10,11 . Indeed, we found that aged plasma increased the number of proliferating microglia in the cortex, and decreased expression of the purinoceptor P2Y12 in hippocampal microglia, a homeostatic microglial marker 46 that is involved in microglia-mediated neuronal regulation 47 . However, we did not observe synaptic or neuronal loss, nor altered synaptic pruning by microglia. Others have reported that aged plasma decreased hippocampal long-term potentiation, neurogenesis and performance on learning and memory tasks in younger mice of 2-3 months approximately 15,16,48 . Possibly, effects of aged plasma on neuronal and synaptic processes are more profound at earlier ages.
Future studies can now focus on identifying the factors in aged plasma that are responsible for the observed immunological changes. Since plasma was dialyzed prior to injection, most growth factors, lipids, metabolites and other small molecules were removed, leaving the main candidates responsible for the immunologic effects plasma-derived proteins such as cytokines, chemokines and hormones. We found that age-associated systemic factors in plasma mimicked the loss of naïve T cells and the expansion of memory T cells in the periphery, and re ected CD8 + T cell accumulation in the brain. However, we did not nd age-associated changes in B cells, NK cells or neutrophils after exposure to aged plasma. Maybe, these age-associated changes occur after prolonged plasma exposure or are induced by other aging factors that are not present in our aged plasma.
Earlier reports have characterized immune cells in the steady-state brain 3 and during aging and neurodegeneration 4 . These well performed studies changed our view of the brain as an immune privileged organ but focused mostly on de ning different myeloid immune cell subsets, while T cells were not investigated in depth. Recent studies found that T cells play an important role in neurodegenerative diseases 10,11 and accumulate in the aged brain [5][6][7][8][9] . As such, we were determined to further investigate which speci c T-cell subsets were altered in brain aging and if age-associated systemic factors were involved. Our data emphasizes above all the accumulation of effector and memory T cells with an exhausted phenotype in the aged circulation and brain, and shows that systemic exposure to aged plasma can induce migration and/or expansion of CD8 + T cells in the brain parenchyma of young-adult mice. This connection between age-associated blood-born factors and CD8 + T cell recruitment opens potential ways through which a possible CD8 + T cell mediated decline in brain function can be counteracted via systemic modulation. Future research will have to determine routes of aged-plasma induced CD8 + T cell migration and expansion within the brain, investigate their interaction with other brain resident cells and study long-term effects of exposure to age-associated systemic factors.

Animal model
Experiments were performed using male C57BL/6 wild-type (WT) mice (The Jackson Laboratory [strain C57BL/6J; stock number 000664; http://jaxmice.jax.org/). All animals were bred on location, handled regularly and group housed (max. 5) under standard laboratory conditions (12-hour light/dark cycle, 21°C, 50% humidity) in the presence of cage enrichment (nesting material). Water and food were available ad libitum. Experiments were reviewed and approved by the Animal Ethics Committee of the Central Authority for Scienti c Experiments on Animals of the Netherlands (CCD, approval protocol AVD115002016659), and acting in accordance with the Directive of the European Parliament and of the Council of the European Union of September 22 nd 2010 (2010/63/EU).

Plasma collection and injection
Mouse plasma was collected from WT mice (average age 20 months + /-2 months). Mice were euthanized using Euthanimal, then plasma was collected by intracardial bleed at time of sacri ce. Plasma was prepared from blood collected with edetic acid followed by centrifugation at 1000g for 10 min at 4°C (brake 1). Plasma aliquots were stored at −80°C until further use. Prior to administration, plasma was pooled and dialyzed using 3.5-kDa Slide-A-Lyzer Dialysis Cassettes (Thermo Scienti c) in PBS to remove edetic acid (2x 2 hour at RT, and 1x o/n at 4°C in 4 L PBS). WT mice were systemically treated with 8 plasma injections (150 µL), or with PBS injections as control, intravenously via the tail vein over a period of 28 days, starting at 5 months of age (± 1 week). Mice were sacri ced the day after the last plasma injection for immune cell isolations. Mice used in immunohistochemical analyses received after the last day of plasma treatment, intraperitoneal injections of BrdU (150mg/kg in PBS; Thermo Fisher) for 3 consecutive days and were directly sacri ced the day after the last BrdU injection.
Immune cell isolations from brain, spleen and blood Tissue collection Mice were rst euthanized using Euthanimal. Blood was collected by intracardial bleed at time of sacri ce with a needle containing edetic acid via the right ventricle. Trans-cardinal perfusion was performed through the left ventricle with a peristaltic pump using PBS (no Ca2 + or Mg2 + ) for 5 min (3 mL/min). After perfusion, mice were decapitated and the brain and spleen were removed. For the 6month-old "young-adult" and 20 month-old "aged" mice, total brain was used for immune cell isolation and subsequent CyTOF analysis. For the plasma or PBS-treated animals, cortical and hippocampal brain structures were removed prior isolation. Finally, tissues were stored in in RPMI (RPMI 1640, Thermo Fisher) on ice until further processing.

Spleen and brain dissection and digestion
Spleens and brains were transferred to a 24-well plate and thoroughly dissected with sharp scissors in 1 mL of digestive mix (2 U/mL Liberase TL, #5401020001). The 1 mL cell suspension was transferred to 12-mL round bottom polypropylene tubes containing 1 mL of 2 U/mL Liberase TL. A magnetic stirrer was added to each tube and the suspensions were incubated in a 37ºC stove on a magnetic stir plate at 200 RPM for 20 min (spleen) and 50 min (brain). Then, 10 mL of ice-cold RPHE (RPMI 1640, 10% FCS, 10 mM EDTA, 20 mM HEPES, 50 µm 2-mercaptoethanol) was added to each tube and the tubes were put in a glass bucket containing ice on top of a stir plate and stirred for 10 min at 500 RPM. Next, the cell suspension was ltered through a 70 µm cell strainer into a 50-mL canonical tube and 30 mL of RP10 (RPMI 1640, 10% FCS, 1% penicillin/streptomycin, 1% glutamine) was added.

Spleen processing
Splenocytes were then centrifuged for 7 min at 1500 RPM, brake 5. The pellet was resuspended in 5 mL ACK red blood cell lysis buffer (Thermo Fisher), transferred to a 15 mL tube and incubated for 5 min on room temperature (RT) with occasional shaking. Next, 8 mL of RP10 was added and samples were centrifuged for 7 min at 1500 RPM, brake 5. Supernatant was carefully removed. Cell pellets were suspended in 1 mL RP10 for counting followed by cell viability staining.

Brain processing
Myelin was separated and removed from the brain using a density gradient. The pellet was resuspended in 70% Percoll, and 30% Percoll was carefully layered on top. Following centrifugation at 900g for 30 min at 22ºC, immune cells in the 70-30% interphase were collected into 10 mL of RP10 buffer. Cell pellets were suspended in 1 mL RP10 for counting followed by cell viability staining.
Blood processing Blood samples were centrifuged at 300g for 10 min (9 acc/6 dec) at 4ºC. Plasma was removed from the upper layer and remaining blood samples were transferred to a 15 mL tube and 5 mL ACK red blood cell lysis buffer (Thermo Fisher) was added. Samples were incubated for 5 min on RT with occasional shaking. Next, 8 mL of RP10 was added and samples were centrifuged for 7 min at 1500 RPM, brake 5. Supernatant was carefully removed and samples underwent another round of red blood cell lysis. Then, 10 mL of RP10 was added to the samples and centrifuged 7 min at 1500 RPM, brake 5. Cell pellets were suspended in 1 mL RP10 for counting followed by cell viability staining.

CyTOF antibody labeling and titration
Antibody labelling with the indicated metal tag was performed using the MaxPAR antibody conjugation kit (Fluidigm) according to the manufacturer's instructions. Puri cation of the bound antibody was performed with high-performance liquid chromatography (Thermo Fisher) and subsequently concentrated by ltering with a 10-kDa lter (Merck Millipore) in a swing-out bucket at 4000 RPM for 15 min. The end volume was determined and an equal volume of antibody stabilizer buffer (Fluidigm; supplemented with 0.05% sodium azide) was added before the antibodies were stored at 4°C. All antibodies used in this study were titrated using both xed and un xed thawed immune cells and the most optimal concentrations with the least spillover were chosen.

CyTOF staining
Reagents were cooled on ice, centrifugation steps were performed at 800g for 7 min at 4°C (9 acc/7 dec) and incubations were performed at RT. Samples were thawed rapidly and washed twice with 8 mL of Maxpar PBS (Fluidigm) and transferred to a 96 well V-bottom plate. After centrifugation, cells were washed twice with 150 μL of 1X Barcode Perm Buffer (Fluidigm, #201057) and incubated with the appropriate palladium barcodes (Fluidigm, #201060) in 1X Barcode Perm Buffer for 30 min. After centrifugation, samples were washed twice with 150 μL of cell staining buffer (CSB) (Fluidigm, #201068) and the cells from all samples were pooled. Cells in this combined sample were counted and after centrifugation incubated with TruStain FcX anti-mouse CD16/32 (BioLegend, #101320) diluted in CSB (1:100) for 10 min. Surface antibody cocktails (Table 1) were prepared fresh in CSB and added to the combined sample in a cell to antibody-cocktail ratio of 3 x 10 6 cells per 100 uL. This was followed by a 30 min incubation. After the incubation cells were washed twice with CSB, followed by a wash with Maxpar PBS. Then, cells were xated with 1 mL of freshly made 1.6% PFA (Thermo Fisher, #28906) in Maxpar PBS for 10 min. After centrifugation, cells were permeabilized with 1 mL of FoxP3 Fix/Perm working solution (eBioscience, #00-5523) for 30 min, followed by two washes with 1X Permeabilisation Buffer (eBioscience, #00-5523). Then, a freshly prepared nuclear antibody cocktail (Table 1)   Runs took approximately 30 -45 min. During the day, tuning of the machine was performed during startup and after 4 hours of sample acquisition. Within each barcoded set of samples, one reference sample was included to correct for differences in staining intensity between barcodes due to technical variation in the staining protocol or daily changes in instrument functioning, as will be discussed later.

Generation of the CyTOF reference sample
The reference sample contained a mixture of immune cells derived from the mouse spleen, blood, brain, liver, bone-marrow and tumor biopsies. Part of the spleen and bone-marrow-derived immune cells were stimulated with 30 ng/mL PMA and 1 µg/mL ionomycin (both Sigma-Aldrich) for 120 min at 37°C to induce expression of transcription factors that were included in the CyTOF panel. All cells were combined, stained for viability as described earlier, xated and stored in aliquots at -80°C until further use.
CyTOF data analysis Acquired samples were randomized using Gaussian negative half zero randomization in CyTOF Software version 6.7. The FCS les were normalized using bead normalization, concatenated and debarcoded using the CyTOF Software version 6.7. Batch alignment for blood and spleen samples was then performed using CytoNorm 49 for each tissue and experiment (e.g. young/old and PBS/plasma-injected) separately. This batch alignment however induced noise in the staining of the brain samples and so the decision was made to work with the 'raw' data there instead. The processed FCS les were then uploaded into OMIQ data analysis software (http://www.omiq.ai) and data analysis was performed. First, normalization beads, cell debris and cell doublets were removed from the data using the DNA staining, bead intensity and Gaussian parameters. Next, live cells showing negative reactivity for viability marker Cell-ID™ Cisplatin-198Pt and dim-to-positive reactivity for CD45-89Y were selected and used for further processing steps. Data were visualized in UMAPs and tSNEs and cells were appointed to clusters by Phenotyping by Accelerated Re ned Community 21 (PARC) and PhenoGraph 50 algorithms. Visual inspection of 1) PARC-and PhenoGraph-derived clusters overlaid on corresponding UMAPs and tSNEs derived from the same sample, and 2) clustered heatmaps that compare median marker expression between clusters, guided the manual merging of clusters and their annotation into biologically relevant immune cell subsets. After the 'general' immune analysis, T cell clusters (and for some instances myeloid clusters as well) were selected, merged and subsampled after which the cluster analysis was repeated to ensure optimal resolution of the underlying data. Counts per immune cell subset were then exported and further analyzed in Graphpad Prism 8.2.1.

Immuno uorescence
After perfusion, mice were decapitated and brains were isolated for post-xation overnight in 4% PFA (4°C) and subsequent incubation for 48 hours in 30% sucrose in PBS solution (4°C) and stored at -80°C. In addition, per mouse, 20 μm thick coronal sections were collected and mounted on Superfrost glass slides (Menzel-glaser, Germany) stored at -20C. For the staining, antigen retrieval was performed at 95°C for 30 min in sodium citrate (pH 6). Sections were incubated with either neutralized donkey serum, neutralized goat serum or neutralized rabbit serum for 30 min at RT. Sections were incubated overnight with primary antibodies in PBS (Table 2) at 4 o C. After incubation, the sections were stained with the appropriate secondary antibodies in PBS. Last, sections were counterstained with DAPI 1:10000 (DAKO) for 10 min. Microscopy and data analysis T cell in ltration in the brain Per mouse, minimal one complete coronal brain section was imaged using a wide-eld microscope (Nikon TI2) at a 1.39 µm/pixel resolution. The CD45 + CD4 + and CD45 + CD8 + T cells were manually counted by a blinded observer and normalized to the total slide area. T cells inside the vessel and the meninges, visualized with Coll-IV, were excluded from the analysis.
Expression of cell adhesion molecules on the brain vasculature Per mouse, 4 confocal images were obtained on a Leica TCS SP8 HyD confocal microscope using a 0.63 µm/pixel resolution, a z-stepsize of 0.33 µm and a total of 10 z-steps. The 3 μm depth z-stack was compressed into one 2D-image using a maximum projection. The percentage of positive vessels was determined using NIS Elements (Nikon) by dividing the total number of Coll-IV + and ICAM-1 + / VCAM-1 + / PECAM-1 + vessels by the total number of Coll-IV + vessels. Thresholds were manually set per ROI by a blinded observer. Vessel density was determined by measuring and dividing the total Coll-IV + area by the total measured tissue area.

Blood vessel permeability
Per mouse, 4 confocal images were obtained on a Leica TCS SP8 HyD confocal microscope using a 0.18 µm/pixel resolution, a z-stepsize of 0.33 µm and a total of 10 z-steps. The 3 μm depth z-stack was compressed into one 2D-image using a maximum projection. Brain vessel permeability was determined by studying the area of extravascular brinogen. First, using NIS Elements (Nikon), a threshold was set manually per ROI by a blinded observer for both Coll-IV and brinogen. Per image, only the brinogen + Coll-IVarea outside the vessels was measured and normalized to the total amount of Coll-IV + vessel area.
Microglial phenotype, number and proliferation Per mouse, two wide eld images were acquired on a Vectra 3.0 spectral imaging system (PerkinElmer) using a resolution of 0.497 µm/pixel. Sections labeled for IBA1 together with P2Y12 and BrdU were analyzed using NIS Elements (Nikon). In short, mean uorescence intensity of P2Y12 in microglia was quanti ed by measuring the total uorescence intensity of P2Y12 within the IBA1 + mask. Microglia density and proliferation was determined by manually counting the number of IBA1 + and IBA1 + BrdU + cells per area by a blinded observer.

Microglial morphology
Per mouse, 10 confocal images were obtained on a Leica TCS SP8 HyD confocal microscope using a 0.09 µm/pixel resolution, a z-stepsize of 0.3 µm and a total of 11 z-steps. The 3 µm depth z-stack was compressed into one 2D-image using a maximum projection. Each z-stack contained one IBA1 + cell located in the stratum radiatum of the hippocampus. IBA1 + microglia were manually traced using FIJI (NIH). Traced microglia were analyzed using the Sholl Analysis Plugin 51 with a 0.2 µm step size from the cell soma. Soma surface area of the traced microglia was measured using the freehand selection tool in FIJI.

Neuronal cell count
Per mouse, two wide eld images of HuC/D stained sections were acquired on a Vectra 3.0 spectral imaging system (PerkinElmer) using a resolution of 0.497 µm/pixel. Cortical neuronal density was determined by manually counting the number of HuC/D + cells per cortical area by a blinded observer.

Synapse density and pruning
Per mouse, 10 confocal images each containing one IBA1 + cell using a 0.09 µm/pixel resolution, a zstepsize of 0.3 µm and a total of 11 z-steps, were acquired from the stratum radiatum area of the hippocampus. To measure synaptic density, Homer1 + spots were created using NIS Elements (Nikon) with a spot diameter of 0.5 μm in a manually selected area lacking IBA1 + cellular processes or cell nuclei.
The density of Homer1 + spots was then calculated by dividing the number of spots by the total volume of the selected region. To investigate synaptic pruning by microglia, manual thresholds were set for Lamp1 and IBA1 to obtain only Lamp1 + IBA1 + volumes. Next Homer1 + spots located within the Lamp1 + IBA1 + volumes were counted. Lysosomal size was calculated by dividing the total Lamp1 + IBA1 + volume by the total IBA1 + volume.

Statistics
Graphpad Prism 8.2.1 was used for all statistical tests. Shapiro-Wilk and F tests were performed to test for normality and equality of variances, respectively, after which appropriate tests were selected. The unpaired two-tailed Student's t test with or without Welch's correction for unequal variances for normally distributed data, or the Mann-Whitney test for non-normally distributed data. To examine the strength of association between two variables, either Pearson or Spearman correlation was used depending on the normality of the data. Data were judged to be statistically signi cant when P < 0.05 and, if signi cant, reported in the gures using the signi cance levels indicated in the gure legends.