Astroglial Hmgb1 Regulates Postnatal Astrocyte Morphogenesis and Cerebrovascular Maturation PDF

Article https://doi.org/10.1038/s41467-023-40682-3 Astroglial Hmgb1 regulates postnatal astrocyte morphogenesis and...

Article https://doi.org/10.1038/s41467-023-40682-3 Astroglial Hmgb1 regulates postnatal astrocyte morphogenesis and cerebrovascular maturation Received: 19 May 2023 Moises Freitas-Andrade 1, Cesar H. Comin2, Peter Van Dyken3, Julie Ouellette1,3, Joanna Raman-Nair1,3, Nicole Blakeley1,3, Qing Yan Liu4,5, Sonia Leclerc4, Accepted: 31 July 2023 Youlian Pan 6, Ziying Liu 6, Micaël Carrier 7, Karan Thakur1, Alexandre Savard1, Gareth M. Rurak8, Marie-Ève Tremblay 7, Natalina Salmaso8, Luciano da F. Costa9, Gianfilippo Coppola10 & Baptiste Lacoste 1,3,11 Check for updates 1234567890():,; 1234567890():,; Astrocytes are intimately linked with brain blood vessels, an essential rela- tionship for neuronal function. However, astroglial factors driving these phy- sical and functional associations during postnatal brain development have yet to be identified. By characterizing structural and transcriptional changes in mouse cortical astrocytes during the first two postnatal weeks, we find that high-mobility group box 1 (Hmgb1), normally upregulated with injury and involved in adult cerebrovascular repair, is highly expressed in astrocytes at birth and then decreases rapidly. Astrocyte-selective ablation of Hmgb1 at birth affects astrocyte morphology and endfoot placement, alters distribution of endfoot proteins connexin43 and aquaporin-4, induces transcriptional changes in astrocytes related to cytoskeleton remodeling, and profoundly disrupts endothelial ultrastructure. While lack of astroglial Hmgb1 does not affect the blood-brain barrier or angiogenesis postnatally, it impairs neuro- vascular coupling and behavior in adult mice. These findings identify astroglial Hmgb1 as an important player in postnatal gliovascular maturation. Astrocytes in the mature brain are physically and functionally cou- called endfeet contact blood vessel walls and the communication pled to blood vessels, a multicellular ensemble referred to as the between these structures can mutually influence growth and gliovascular unit, which regulates cerebral blood flow and the maturation10–12. However, exactly when, and how, the brain glio- blood-brain barrier to support neurotransmission and maintain vascular unit is established is not known. neuronal health1,2. In mice, the genesis of astrocytes is initiated in During the first two postnatal weeks of brain development, the late prenatal period, and postnatally these glial cells proliferate astroglial morphogenesis is highly active, with primary branches rapidly and undergo extensive morphological changes3–6, coincid- sprouting from the cell body and dividing into finer processes5,7. ing with the expansion of cerebrovascular networks5,7–9. During Astroglial processes continue to ramify until a highly complex postnatal brain development, perivascular astroglial processes spongiform-like morphology is achieved, which requires extensive 1 Neuroscience Program, The Ottawa Hospital Research Institute, Ottawa, ON, Canada. 2Federal University of São Carlos, Department of Computer Science, São Carlos, Brazil. 3Cellular & Molecular Medicine, University of Ottawa, Ottawa, ON, Canada. 4National Research Council of Canada, Human Health and Therapeutics, Ottawa, ON, Canada. 5Department of Biochemistry Microbiology and Immunology, Faculty of Medicine, University of Ottawa, Ottawa, ON, Canada. 6Digital Technologies, National Research Council of Canada, Ottawa, ON, Canada. 7Division of Medical Sciences, University of Victoria, Victoria, BC, Canada. 8Department of Neuroscience, Carleton University, Ottawa, ON, Canada. 9University of São Paulo, São Carlos Institute of Physics, FCM-USP, São Paulo, Brazil. 10Yale School of Medicine, Dept. of Pathology, New Haven, CT, USA. 11University of Ottawa Brain and Mind Research Institute, Ottawa, ON, Canada. e-mail: [email protected] Nature Communications | (2023)14:4965 1 Article https://doi.org/10.1038/s41467-023-40682-3 cytoskeletal remodeling. However, very little is known about the time differentiation of radial glia into astrocytes in the developing cortex34. course and molecular machinery of astroglial morphogenesis13. To confirm these observations obtained by immunofluorescence, we While astroglial endfeet are separated from endothelial cells and utilized transmission electron microscopy (TEM) to investigate post- pericytes by the basal lamina, extensive signaling occurs between natal development of the gliovascular interface at the ultrastructural astrocytes, pericytes and endothelial cells14–18. Moreover, a recent level (Fig. 1f, Supplementary Fig. 2). TEM showed that microvessels study reported that a large portion of the astrocyte proteome is were not yet contacted by perivascular endfeet at P0. By P7, however, dedicated to astroglial endfeet, highlighting the complex nature of this large immature endfeet partially covered microvessels (Fig. 1f, h). cellular compartment19. Yet, the fundamental principles guiding end- Vascular coverage by endfeet observed at P14 via EGFP (Fig. 1a) foot placement around brain blood vessels remain elusive20. translated into 100% coverage on TEM micrographs at the same age Neuronal activity and metabolism are inextricably “coupled” via (Fig. 1f, h). Between P7 and P21, the number of astroglial endfeet the gliovascular unit1,21,22. Abnormal gliovascular growth has been increased, while the total endfoot area decreased (Fig. 1h), illustrating a observed in neurodevelopmental disorders and is associated with refinement of these developing structures. Altogether this shows that long-term neurological consequences23–28. Moreover, astrocyte dys- perivascular endfoot placement establishes around P7 and is complete function is linked to most, if not all, neuropsychiatric conditions29,30. at P14, which is in line with observations made by Gilbert et al.12. For example, coverage of blood vessels by astroglial endfeet was sig- nificantly reduced in subjects with major depressive disorders28. High mobility group box 1 (Hmgb1) is highly expressed in Therefore, understanding the fundamental mechanisms of astrocyte astrocytes at birth morphogenesis and gliovascular maturation is crucial to develop Given the morphological changes observed concomitantly in astro- strategies that promote healthy brain development. Here, we char- cytes and microvessels between P0 and P14, we next sought to identify acterize the early postnatal temporal profile of cellular and molecular temporal changes in gene expression associated with gliovascular changes at the gliovascular interface in the mouse cerebral cortex, and growth during that period. We first interrogated our published TRAP- we identify a mechanism that regulates gliovascular development. We Seq database33, and then confirmed our observations using in situ show that a key molecular factor produced by astrocytes, namely multiplex RNA sequencing in astrocyte-enriched areas of interest HMGB1, regulates astrocyte morphogenesis via regulation of the (AOIs; Fig. 2a–c and Supplementary Fig. 3a–g). Consistent with major cytoskeleton, controlling endfoot placement and endfoot protein changes in astroglial morphology between P0 and P14, more genes distribution during postnatal brain development, with long-term appeared differentially expressed in astrocyte-enriched AOIs at P5 implications in neurovascular coupling and behavior. (Supplementary Fig. 4). When focusing on genes known as involved in cerebrovascular remodeling, we consistently found that Hmgb1 was Results highly expressed by astrocytes at birth (Fig. 2a, b), even though past Mouse brain astrocytes initiate vascular endfoot coverage dur- work has focused on the role of this gene in the adult brain, where it is ing the first postnatal week upregulated post-injury35. Expression of Hmgb1 in astrocytes was To thoroughly assess the temporal profile of gliovascular development markedly higher than canonical pro-angiogenic genes, such as Vegfa or and set the stage for subsequent molecular investigations, we Wnt7a (Fig. 2a, b). Both Hmgb1 gene expression and HMGB1 protein employed transgenic mice producing the enhanced green fluorescent levels decreased between P0 and P14 alongside maturation of the protein (EGFP) under the control of pan-astroglial Aldehyde Dehy- gliovascular unit (Fig. 2b, Supplementary Fig. 5a, b). HMGB1 protein drogenase 1 Family Member L131 (Aldh1l1) promotor. Analysis of EGFP- was also produced by neurons and pericytes, but microglia and labeled astrocytes and CD31-labeled endothelial cells (ECs) showed endothelial cells failed to display detectable HMGB1 nuclear immu- that cortical astrocytes and microvessels increase in complexity over noreactivity (Fig. 2e–g, Supplementary Fig. 5c, d). Yet, endothelial cells the first two postnatal weeks with little change during the third post- at P14 expressed Hmgb1 gene (mRNA), even at higher levels than natal week (Fig. 1a, Supplementary Fig. 1a). Astrocyte morphogenesis astrocytes (Supplementary Fig. 9g), consistent with a recent adult was highly active during the first 5 postnatal days (Supplementary mouse database36,37 and suggesting important post-transcriptional Fig. 1b), followed by an active phase of angiogenesis up to P14 (Fig. 1a, regulation for this gene. Given the high expression of Hmgb1 in Supplementary Fig. 1a)6,8,32. At P0, perivascular astrocytes exhibited astrocytes at birth and the role of HMGB1 in adult neurovascular repair, simple main branches aligned along the scaffold of blood vessels we hypothesized that HMGB1 regulates astrocyte-dependent gliovas- (Fig. 1a), and astrocyte cell bodies were in close apposition with cular maturation during this critical period. To test this, we generated microvessels compared to other timepoints (Fig. 1d). By P5, astroglial Hmgb1ΔAstro mice (Aldh1l1-CreERT2+/−;Hmgb1flox/flox) for conditional, primary processes ramified into a highly branched morphology astrocyte-specific ablation of Hmgb1 starting at P0 (Fig. 2h, and Sup- (Fig. 1b, Supplementary Fig. 1b). With cortical expansion, astrocytes plementary Fig. 5e, f) and we confirmed selective HMGB1 removal from positioned their cell body farther away from blood vessels and then astrocytes as early as P7 (Fig. 2i). At P14, loss of HMGB1 in Hmgb1ΔAstro progressively sent extensions back towards these vessels. This was also astrocytes did not affect HMGB1 production by either neurons or illustrated by the significant reduction in the proportion of astrocytes pericytes (Fig. 3a, b and Supplementary Fig. 5d). directly contacting vessels at P5 (Fig. 1c). After P5, astrocyte cell bodies remained distant from the vessel wall (Fig. 1d), but EGFP immunor- Postnatal ablation of astroglial Hmgb1 affects endfoot matura- eactivity on vessels increased sharply between P5 and P14 (Fig. 1e), tion around blood vessels consistent with a period of endfoot placement. Increases in astroglial Vessel density and branching appeared unaffected in Hmgb1ΔAstro mice extensions’ branching and length continued until P21 (Supplementary compared to control littermates at P14 (Supplementary Fig. 6a), sug- Fig. 1b), and the dominant vertical orientation alongside vessels seen at gesting that astroglial HMGB1 is not required for postnatal angiogen- birth was progressively replaced by a seemingly random orientation by esis. While astrocyte cell density and surface ratio were unaltered in P21 (Supplementary Fig. 1c). We also observed a rise in astrocyte cell the cerebral cortex of Hmgb1ΔAstro mice (Supplementary Fig. 6b), we density during that period (Supplementary Fig. 1a), consistent with found that distribution of gap junction protein connexin43 (Cx43), recent findings6,33, with a peak in proliferation of astrocytes around P4 expressed by astroglial processes and endfeet38,39, was significantly correlating strongly with EC proliferation (Supplementary Fig. 1d). In altered throughout the cerebral cortex of Hmgb1ΔAstro mice (Fig. 3c). addition, we observed glial acidic fibrillary protein (GFAP)-labeled Disrupted distribution of Cx43 was particularly evident on larger radial glial processes at P5 and P7 (Supplementary Fig. 1a), coinciding microvessels (Fig. 3c, d). We also measured a ~3.5-fold increase in with emergence of GFAP-positive astrocytes from P7, and with the the number of astrocytes exhibiting high density “patches” of Cx43 Nature Communications | (2023)14:4965 2 Article https://doi.org/10.1038/s41467-023-40682-3 a CD31 (1) ALDH1L1-EGFP (2) GFAP (3) Merge (1-3) Segmentations (1,2) P0 50 Pm P14 200 Pm b P0 P5 P7 P14 P21 CD31 / EGFP 100 Pm c 1.0 P < 0.0001 P < 0.0001 d Astrocyte cell bodies e EGFP immunoreactivity contacting blood vessels Fraction of astrocytes 0.8 0.6 P = 0.0314 n=4 n=4 0.4 0.2 n=4 0.0 P0 P5 P7 P14 P21 Postnatal day Distance from vessel wall (Pm) Distance from vessel wall (Pm) f P0 P7 P14 P21 D RBC RBC L L 1 Pm A Astrocytic processes/endfeet Total endfoot area / capillary (nm ) 2 Pericyte processes h 150 n=4 0.6 n=4 P = 0.0013 2.0 107 n=4 Number of endfeet /Pm BM Endothelium g P0 P7 P < 0.0001 Astrocyte coverage 1.5 107 of capillaries (%) 250 nm 100 0.4 D D S 1.0 107 P = 0.0214 BM P BM S 50 0.2 P S 5.0 106 P = 0.0263 S S vs. P0 N.D. vs. P7 N.D. vs. P7 E E 0 0.0 0.0 P0 P7 P14 P21 P0 P7 P14 P21 P0 P7 P14 P21 L Postnatal day Postnatal day Postnatal day puncta in and around their cell body compared to control littermates We also found that AQP4 protein levels were unaffected in Hmgb1ΔAstro (Fig. 3c, e). The number of clearly identifiable endfeet delineated by mice (Fig. 3h), confirming that these alterations were in AQP4 dis- Cx43 on blood vessels was lower in Hmgb1ΔAstro mice at P14 (Fig. 3c, e), tribution but not production. We then tested whether these changes in while both Cx43 protein levels and total Cx43 immunoreactivity per endfoot-enriched proteins were associated with altered endfoot pla- vessel were unchanged (Fig. 3f, Supplementary Fig. 6c, d). Aquaporin-4 cement around microvessels. Using immunofluorescence, we mea- (AQP4), another astroglial endfoot marker, also displayed altered sured a significantly altered distribution of ALDH1L1 immunoreactivity distribution with significant decrease in vascular coverage in onto CD31+ microvessels (Fig. 4a). In Hmgb1ΔAstro mice at P14, a higher Hmgb1ΔAstro mice at P14 compared to age-matched controls (Fig. 3g). number of vessels displayed reduced astroglial coverage, whereas a Nature Communications | (2023)14:4965 3 Article https://doi.org/10.1038/s41467-023-40682-3 Fig. 1 | Gliovascular maturation during postnatal development of the cerebral from blood vessel wall between P0 and P21. Data are mean ± SEM. f Electron cortex. a Fluorescence micrographs of somatosensory cortex from Aldh1l1-eGFP micrographs show temporal developmental profile of astrocyte coverage (pseu- male mice at P0 and P14. Endothelial marker CD31 (red); astrocyte markers EGFP docoloured in green) around microvessels (pseudocoloured in red). Pericytes can (green) and GFAP (blue). Arrows point to astroglial endfeet around microvessels. also be observed (pseudocoloured in yellow). Red asterisks indicate areas lacking Blue arrowheads indicate radial glia. Segmentations represent reconstructed astroglial coverage. Red arrows point at inter-endfoot walls. A astrocyte, D neuronal astrocytes and vessels quantified to evaluate vessel density and branching, as well dendrite, L lumen, RBC red blood cells. g Higher magnifications corresponding to as astrocyte density and fraction contacting vessels. b Fluorescence micrographs of dashed boxes in (f) illustrating fine features of astrocytic endfeet at zones lacking somatosensory cortex from Aldh1l1-eGFP mice at P0, P5, P7, P14 and P21. CD31 endfoot coverage. BM basement membrane, D neuronal dendrite, E endothelium, L (red); EGFP (green). White asterisks indicate perivascular astrocyte cell bodies lumen, P pericyte, S dendritic spine. Black arrows, BM. Green arrows point at closely associated with microvessels at P0. White arrows indicate astroglial endfeet astrocyte endfoot cytoplasm. h Analysis at the nanoscale level of astrocyte extending and contacting microvessels from P7. c Segmentations (see in a) were maturation and endfoot coverage during postnatal brain development. Data are quantified to evaluate the fraction of astrocytes contacting vessels. Data are whisker boxes (min to max, center line indicating median) in (g). *p < 0.05, whisker boxes (min to max, center line indicating median). *p < 0.05, ***p < 0.001 **p < 0.01, ***p < 0.001 (One-way ANOVA and Tukey’s post-hoc test). All displayed (One-way ANOVA and Tukey’s post-hoc test). d, e Quantifications of astrocyte cell microscopy images are representative of experiments repeated in 4 mice per body density (d) or EGFP immunoreactivity (e) as a function of the distance (μm) group, with similar results. Source data are provided as a Source Data file. lower number of vessels displayed elevated (i.e. normal) coverage postnatal maturation of astrocytes and that this is independent of compared to controls. At the ultrastructural level (TEM), endfoot cortical lamination. morphology appeared disorganized in Hmgb1ΔAstro mice and exhibited a ~1.5-fold increase in the number of smaller, fragmented endfeet Astroglial Hmgb1 regulates the expression of genes related to around microvessels, with some perivascular areas lacking astroglial cytoskeletal remodeling coverage (Fig. 4b, Supplementary Fig. 7a, b). Furthermore, endothelial Next, to unmask the molecular underpinnings of altered gliovascular morphology appeared dramatically altered in Hmgb1ΔAstro mice at P14, structure at P14 in Hmgb1ΔAstro mice, we isolated primary cortical with increased number (2–3-fold) of vascular profiles displaying mac- astrocytes and ECs from Hmgb1ΔAstro and control littermates and per- ropinocytic extensions and vacuoles (Fig. 4b, Supplementary Fig. 7b, formed deep RNA sequencing on each cell type from individual mice c). We also observed several instances where endothelial protrusions (Fig. 6a, b and Supplementary Fig. 9b, d). In contrast to the markedly extended into astroglial endfeet, but only in Hmgb1ΔAstro mice (Sup- disturbed ultrastructure of ECs in Hmgb1ΔAstro mice at P14, tran- plementary Fig. 7d). Given the morphological alterations measured in scriptomic analysis showed little-to-no change in endothelial gene endothelial cells from Hmgb1ΔAstro mice at P14, we sought to determine expression (Supplementary Fig. 9c). Compared to ECs, astrocytes from whether blood-brain barrier (BBB) integrity was compromised at this Hmgb1ΔAstro mice displayed a larger number of differentially-expressed age in mutant animals. Injection of small tracer AlexaFluor555- genes (DEGs) (Fig. 6c, d). In particular, four astroglial DEGs down- conjugated Cadaverine (~1 kDa) did not reveal any leakage in the cer- regulated in Hmgb1ΔAstro astrocytes have been associated with cytos- ebral cortex of Hmgb1ΔAstro mice compared to controls (Fig. 4c). keletal remodeling, namely Slit1, involved in neuronal Moreover, levels of tight and adherens junction proteins appeared morphogenesis41,42, Camk1g encoding Calcium/calmodulin-dependent similar between mutant and control mice (Fig. 4d, Supplementary protein kinase Iγ (CaMKIγ) which regulates activity-dependent den- Fig. 7e). In addition, despite disturbed endothelial morphology, tight dritic growth via cytoskeletal remodeling43,44, Arhgef28 encoding Rho junction ultrastructure and the total number of caveolae-type vesicles guanine nucleotide exchange factor (RGNEF) involved in microtubule appeared normal in Hmgb1ΔAstro mice (Supplementary Fig. 10a), alto- network remodeling45,46, as well as Ptger4 encoding prostaglandin E gether suggesting an intact BBB despite endfoot misplacement. receptor 4 (EP4) which controls actin cytoskeleton remodeling47 (Fig. 6e, Supplementary Fig. 9d). Of note, activation of EP4 was shown Postnatal ablation of astroglial Hmgb1 affects astrocyte to inhibit actin ring formation in osteoclasts47. Consistent with these morphogenesis observations, and possibly linked to Ptger4 downregulation, we con- Given the astroglial endfoot phenotype observed in Hmgb1ΔAstro sistently observed significantly more Hmgb1ΔAstro astrocytes presenting mice, we assessed whether loss of HMGB1 affects overall astroglial a somatic actin ring compared to control astrocytes when stained with morphology in vivo and in vitro. First, a systematic in vivo mor- F-actin-binding phalloidin (Fig. 6f, Supplementary Fig. 9e, f). Fewer phological analysis was performed on ALDH1L1-immunostained transcripts were significantly upregulated (FRD < 0.05; FC > 2) in brain sections from mutant and control littermates at P14 (Fig. 5a, Hmgb1ΔAstro astrocytes, including Teddm2, encoding the epididymal Supplementary Fig. 8). Astrocytes from Hmgb1ΔAstro mice displayed a protein Me9, and Lurap1l, predicted to function as an adaptor involved significant decrease in several morphological metrics, including in the regulation of cell shape48. Collectively, these data suggest that reduced number and total length of main branches (Fig. 5a). Second, HMGB1 regulates cytoskeletal arrangement in postnatal astrocytes. to determine whether aberrant astrocyte morphology is a cell- autonomous phenotype, primary astrocyte cultures were prepared Loss of astroglial Hmgb1 postnatally affects cerebrovascular from Hmgb1ΔAstro and control littermates and immunostained for function and behavior into adulthood GFAP and HMGB1 (Fig. 5b, c). Similar to what was observed in vivo, Considering the morphological and transcriptional alterations mea- primary astrocytes isolated from Hmgb1ΔAstro mice exhibited sig- sured in postnatal astrocytes from Hmgb1ΔAstro mice, we sought to nificant changes in morphology, with less extensions and a more determine consequences of loss of astroglial HMGB1 on gliovascular flattened cell body compared to control astrocytes (Fig. 5d, e). unit function in adult Hmgb1ΔAstro animals. We assessed neurovascular Considering the roles played by ECs and astrocytes (including Cx43) coupling in the primary somatosensory (S1) cortex of young adult in the maturation of the cerebral cortex and neuronal (P50) Hmgb1ΔAstro mice and control littermates. Upon whisker stimula- networks11,24,40, immunofluorescent analysis of three neuronal mar- tion in Hmgb1ΔAstro mice, we measured reduced cerebral blood flow kers (NeuN, CTIP2 and POU3F2) was performed to test whether lack (CBF) responses (% increase), as well as reduced CBF response ampli- of astroglial Hmgb1 affects cortical growth. No difference in neu- tude during stimulation (Fig. 7a). Other hemodynamic parameters ronal layering across cortical depth was observed between the were unaffected by the conditional mutation (Supplementary Fig. 10c). Hmgb1ΔAstro and control mice at P14 (Supplementary Fig. 9a), alto- The reduction in CBF response exhibited by Hmgb1ΔAstro mice appeared gether demonstrating that astroglial Hmgb1 is required for proper independent of neuronal activation, as no change in cFos induction Nature Communications | (2023)14:4965 4 Article https://doi.org/10.1038/s41467-023-40682-3 a 8 TRAP-Seq database b 10 Spatial transcriptomics c 10 Spatial transcriptomics (astroglial AOIs) Log2 normalized read counts 9 Translation level (RPKM) 7 8 P = 0.0352 8 P = 0.0073 6 7 -log10(FDR) P < 0.0001 5 6 P < 0.0001 6 P = 0.0290 P = 0.0242 5 Hmgb1 4 P = 0.0003 4 3 4 3 2 Hmgb1 2 n = 4-5 2 n=4 1 Vegfa 1 FDR = 0.05 vs. P1 vs. P0 n=4 Wnt7a 0 0 0 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 P1 P4 P7 P14 P0 P5 P14 Higher at P0 log2(FC) Higher at P14 Age Age d HMGB1 / ALDH1L1 / DAPI Astrocytes HMGB1 immunofluorescence intensity 50 P0 P5 P7 P14 40 P = 0.0147 (arb.units) 30 P < 0.0001 20 10 n=4 vs. P0 0 50 Pm 100 Pm P0 P5 P14 Age e Neurons P14 f Microglia P14 g P0 Endothelium HMGB1 / NeuN HMGB1 / CD31 HMGB1 / Iba1 P14 25 Pm 50 Pm 100 Pm 50 Pm 100 Pm 100 Pm HMGB1 / ALDH1L1 / DAPI h i P7 Hmgb1flx/flx +Tam. Hmgb1'Astro Tamoxifen Tissue/cell collection Function (50 Pl, 1 mg/mL) (IF, EM, RNAseq) (CBF, behaviour) P0 P1 P2 P14 P50 50 Pm 100 Pm was observed in S1 cortex of whisker-stimulated Hmgb1ΔAstro animals immunoreactivity (Supplementary Fig. 10b). Finally, we hypothesized (Fig. 7a, Supplementary Fig. 10c). We also tested whether altered CBF that all the alterations reported above in Hmgb1ΔAstro mice may lead to in adult mutants correlated with aberrant distribution of cortical Cx43. behavioral changes in adult mice. Hmgb1ΔAstro mice displayed sig- Consistent with our findings at P14, the Cx43 phenotype persisted in nificantly reduced marble-burying behavior compared to control lit- adult Hmgb1ΔAstro mice compared to age-matched controls, yet the size termates (Fig. 7c), an indicator of stereotypic and/or repetitive of abnormal Cx43 puncta patches appeared higher at P50 compared to behavior. While Hmgb1ΔAstro and control mice showed comparable P14 in Hmgb1ΔAstro mice (Fig. 7b). However, protein levels of Cx43 were behaviors in the open field and novel object recognition tasks (Fig. 7d, significantly lower in mutant mice compared to littermate controls Supplementary Fig. 10e, f), mutant mice travelled more distance dur- (Supplementary Fig. 10d). In addition, protein levels of AQP4 appeared ing the 5-minute novel object test trial (Fig. 7d). In the elevated plus slightly higher in P50 mutant mice compared to age-matched controls maze test, more entries in closed arms were measured for Hmgb1ΔAstro (Supplementary Fig. 10d). Of note, no astrogliosis, was found in either mice compared to their control littermates (Fig. 7e, Supplementary control or mutant mice at P50 as assessed by low GFAP Fig. 10g), suggesting an anxiety-like phenotype in mutant animals. Nature Communications | (2023)14:4965 5 Article https://doi.org/10.1038/s41467-023-40682-3 Fig. 2 | HMGB1 is highly expressed in astrocytes around birth in the mouse HMGB1 immunoreactivity at P0, P5 and P14. Data are mean ± SEM. *p < 0.05, cerebral cortex. a Expression (reads per kilo base per million mapped reads, ***p < 0.001 (One-way ANOVA and Tukey’s post-hoc test). e Fluorescence micro- RPKM) of Hmgb1 relative to known pro-angiogenic astroglial genes at P1, P4, P7 and graph showing HMGB1 (red) in NeuN+ neurons (green) at P14. Inset displayed higher P14 from Star Database (https://stardb.cu-bic.ca/). Data are mean ± SEM. *p < 0.05, magnification. f Fluorescence micrograph showing lack of HMGB1 (red) in Iba1+ **p < 0.01, ***p < 0.001 (One-way ANOVA and Tukey’s post-hoc test). Comparisons microglia (green) from P14 mice. Inset displayed higher magnification, and arrows were made with respect to P1. b In situ expression (normalized read counts) of point at microglial cells. g Fluorescence micrographs of somatosensory cortical Hmgb1 relative to known pro-angiogenic genes at P0, P5 and P14 in astrocyte- sections immunostained with CD31 (green) and HMGB1 (red) at P0 and P14. Insets enriched (ALDH1L1+CD31−) areas of interest (AOIs). Data are mean ± SEM. *p < 0.05, displayed higher magnifications. h Diagram of mating strategy and tamoxifen **p < 0.01, ***p < 0.001 (One-way ANOVA and Tukey’s post-hoc test). Comparisons injections for selective Hmgb1 ablation from astrocytes, with experimental end were made with respect to P0. c Volcano plot (False discovery rate, FDR, versus points. i Fluorescence micrographs showing immunostaining for ALDH1L1+ astro- Log2 fold change, FC) to visualize astroglial genes with differential expression cytes (green) and HMGB1 (red) in control and Aldh1l1-CreERT2;Hmgb1flx/flx (or between P0 and P14 in situ. d Left, Fluorescence micrographs of immunostained Hmgb1ΔAstro) mice. Insets displayed higher magnifications. Arrows and arrowheads somatosensory cortex sections at specific time-points during postnatal develop- point to astrocytes with (left) or without (right) HMGB1. All displayed microscopy ment. Insets displayed higher magnifications. Arrows indicate HMGB1 immunor- images are representative of experiments repeated in at least 4 mice per group, eactivity (red) in ALDH1L1+ astrocytes (green). Right, Quantification of astroglial with similar results. Source data are provided as a Source Data file. Discussion zebrafish, reported aberrant brain development, with altered neural In this study, we analyzed gliovascular development, focusing on the progenitor proliferation and survival63,64. Here, we provide evidence time course of astrocyte maturation and endfoot placement around revealing an important role for astroglial Hmgb1 in gliovascular cortical microvessels, as well as on the cellular and molecular development. maturation of astrocytes with respect to cerebrovascular growth. We Postnatally, Hmgb1ΔAstro mice displayed profound morphological found that the first seven days after birth are critical for the maturation changes at the gliovascular unit (both in astrocytes and ECs) as of astrocytes and for recruitment of their endfeet at the microvascular revealed by in vivo and in vitro experiments. However, the BBB was not wall in the mouse cerebral cortex. We reveal that perivascular endfoot affected in these mutant mice, since neither small-molecule leakage coverage begins to establish around P7 in the cerebral cortex, and that nor tight junction protein difference were observed between the complexity of astrocyte morphology and transcriptome evolves Hmgb1ΔAstro mice and their control littermates. Previous studies have alongside postnatal growth of cortical vessels. At P5, marked tran- shown that removal of astroglial endfeet does not perturb the BBB, scriptional changes are apparent in astrocytes compared to other consistent with our findings in Hmgb1ΔAstro mice at P14. On the one time-points during postnatal brain growth. Genes associated with hand, the gliovascular unit exhibits plasticity for instance upon brain cytoskeletal remodeling, such as Mapt49, Rasgrf250 and Gap43 linked to injury, and astroglial endfeet replacement around blood vessels has astrocyte arborization and elongation51, highlight the dynamic nature been observed after laser ablation of endfeet65 or following two- of astrocyte morphogenesis early after birth. Prior to our work, several photon chemical apoptotic ablation of perivascular astrocytes66. studies had reported elements of postnatal astrocyte maturation4,6,52,53, However, whether single astrocytes are forming more vascular con- but the precise time course underlying gliovascular maturation in the tacts, as a compensatory mechanism, or more astrocytes are sending brain was unknown. Gilbert et al.12 reported a mechanism by which the processes to the same vascular portion Hmgb1ΔAstro mice remains to be astrocyte-specific membrane protein MLC1 is critical for astrocyte elucidated. We found that while Hmgb1ΔAstro astrocytes displayed orientation and polarity towards developing vessels12, which could be reduced main branches in vivo and in vitro, the number of smaller at play during postnatal gliovascular growth. Another recent study astrocytic endfeet appeared increased in vivo using TEM; This dis- focusing on the retinal vasculature reported that constitutive lack of crepancy may reflect the limitation in resolution of immuno- Adenomatosis polyposis coli downregulated-1 (Apcdd1) in mice led to fluorescence (IF) detection by algorithms to fully delineate precocious maturation of astrocytes with increased Aqp4 expression morphological characteristics of single astrocytes. In addition, it will and more extensive perivascular endfoot coverage at P1454. Testing therefore be important to verify if endfoot replacement can be whether Hmgb1 expression is modulated in Apcdd1−/− or Mlc1-/- mice observed in adult Hmgb1ΔAstro mice. On the other hand, our observa- could provide a valuable genetic framework to study gliovascular tions imply that a contact-dependent regulation of the BBB by astro- maturation. glial endfeet during postnatal development is unlikely. However, this Our investigation into molecular players associated with gliovas- does not rule out a regulation of the BBB by factors released from cular unit formation led us to identify HMGB1 as a molecular factor in astrocytes, with implications for neurological disorders67. astrocytes to regulate both astroglial morphogenesis and gliovascular Several genes linked to cytoskeletal regulation were identified as unit maturation. HMGB1 is an evolutionarily conserved non-histone significantly downregulated in Hmgb1ΔAstro astrocytes, including Slit1, protein that is widely present in the nucleus of eukaryotic cells. It plays Camk1g and Arhgef28. In particular, Ptger4, encoding EP4 which reg- important roles in stabilizing DNA and regulating transcription55. ulates actin ring formation47, was downregulated in Hmgb1ΔAstro astro- HMGB1 has multiple functions depending on its location in the nucleus cytes. This observation correlated with a high proportion of Hmgb1ΔAstro or cytosol, and it can be released actively from stressed cells, or pas- astrocytes exhibiting actin rings in cell culture compared to control sively after necrotic cell death56. Under pathological conditions in the cells. While very little is known about the function of HMGB1 in adult brain, HMGB1 is upregulated and secreted to promote reparative astrocytes, a recent in vitro study showed that overexpression of angiogenesis and induction of inflammatory pathways35,57–59. In our nuclear HMGB1 in astrocytes significantly increased the methylation of study, Hmgb1 gene was highly expressed at birth (i.e., in a developing, the SAM and SH3 domain-containing 1 (SASH1) gene, encoding a healthy brain) and rapidly decreased by the end of the second post- scaffolding protein associated with cytoskeletal regulation, and this natal week. We propose that increased production of HMGB1 under was linked to reduced cell adhesion and increased migration68. pathological conditions in the adult brain relates to the induction of a One hallmark of astrocyte differentiation is the robust expression developmental genetic program promoting growth and repair. Simi- of Cx43, the most abundant connexin in the brain, involved in exten- larly, others have shown that developmental processes are reactivated sive coupling between astrocytes via gap junction69. This coupling in pathological conditions in adults60–62, however a role for astroglial allows for rapid communication between astroglial networks in HMGB1 in neurodevelopment has been overlooked. Few studies using response to neuronal demands70–72. In the cortex of Hmgb1ΔAstro mice, mice constitutively lacking HMGB1 (Hmgb1-/-), or Hmgb1 knockdown in despite the difference in Cx43 protein levels at P50, Cx43 protein Nature Communications | (2023)14:4965 6 Article https://doi.org/10.1038/s41467-023-40682-3 a Hmgb1flx/flx +Tam. Hmgb1'Astro b Hmgb1flx/flx +Tam. Hmgb1'Astro HMGB1 / ALDH1L1 / DAPI P14 P14 HMGB1 / NeuN / DAPI 25 Pm 50 Pm 25 Pm 50 Pm c Hmgb1flx/flx +Tam. Hmgb1'Astro P14 P14 ALDH1L1 / Cx43 100 Pm P14 P14 50 Pm d e 5 Pm < diameter < 15 Pm P = 0.0002 f Control Mutant Number of “Cx43-high” astrocyte patches /mm2 P14 CD31 / Cx43 0.06 0.010 Hmgb1flx/flx +Tam. ~43 kDa Cx43-delineated endfeet /Pm Number of clearly identifiable Hmgb1'Astro 0.008 2.0 Cx43 (cortex) (relative to loading control) flx/flx 0.04 P = 0.0022 Hmgb1 +Tam. 0.006 1.5 Hmgb1 'Astro Cx43 levels 0.004 1.0 0.02 0.002 0.5 50 Pm n=6 n=8 0.00 0.000 n=4 0.0 Control Hmgb1'Astro Control Hmgb1'Astro g Hmgb1flx/flx +Tam. Hmgb1'Astro P14 h Control Mutant CD31 / Aquaporin-4 1.0 ~30 kDa 0.8 (relative to loading control) AQP4 (cortex) 2.5 Coverage ratio flx/flx P = 0.0193 Hmgb1 +Tam. 0.6 2.0 'Astro Hmgb1 AQP4 levels 0.4 1.5 1.0 0.2 0.5 n=6 n=4 50 Pm 100 Pm 0 0.0 distribution appeared severely impaired at both P14 and P50. This ependymal cells into astrocytes75. Taken together, HMGB1 may func- might be explained by the aberrant cytoskeletal rearrangement in tion as a transcriptional regulator in developing astrocytes to regulate Hmgb1ΔAstro astrocytes. It is possible that F-actin rings prevented the cytoskeletal reorganization and cellular morphogenesis. distribution of Cx43 beyond the cell body into astroglial extensions. The reduced morphological complexity of Hmgb1ΔAstro astrocytes, Indeed, the C-terminal tail of Cx43 interacts with the actin cytoskele- observed both in vitro and in vivo, affected endfoot placement at the ton, a substrate for proper transport to the cell membrane40,73. Cx43 gliovascular interface, as measured at the TEM level and by endfoot was also reported to control astrocyte morphogenesis via a channel- markers Cx43 and AQP4. Interestingly, the spaces let by the Hmgb1ΔAstro independent mechanism likely mediated by its adhesive properties74. endfeet on blood vessels (TEM images) are small in comparison to the Of note, HMGB1 was shown to promote differentiation of spinal length of vessels with no AQP4 signal (immunofluorescent images) in Nature Communications | (2023)14:4965 7 Article https://doi.org/10.1038/s41467-023-40682-3 Fig. 3 | Ablation of HMGB1 in newborn astrocytes affects astroglial endfoot patches (right). Data are whisker boxes (min to max, center line indicating median). protein distribution in the postnatal cerebral cortex. a Fluorescence micro- **p < 0.01, ***p < 0.001 (two-tailed Mann–Whitney’s test). f Cx43 immunoblot of graphs showing immunostaining for ALDH1L1+ astrocytes (green) and HMGB1 (red) mouse cerebral cortex from control (n = 4) and Hmgb1ΔAstro (n = 4) mice. Graph at P14. Asterisks delineate blood vessels. Insets are higher magnifications. Arrows shows quantification of immunoblot signal normalized to the loading control and point to astrocytes expressing (left) or lacking (right) HMGB1. b Fluorescence relative to control values. Data are whisker boxes (min to max, center line indicating micrographs of neurons (NeuN, green) and HMGB1 (red). c Top left panel: Fluor- median). g Left panels, Fluorescence micrographs showing aquaporin-4 immu- escence micrograph showing normal Cx43 immunolabelling (cyan) in astrocytes noreactivity (red) on CD31+ blood vessels (green) from control (left) or Hmgb1ΔAstro (magenta) in control (Hmgb1flx/flx) mice. Top right panel: Fluorescence micrograph (right) brain sections. Right panel, coverage ratio of aquaporin-4 on CD31+ blood showing disrupted Cx43 distribution in Hmgb1ΔAstro mice. Arrowheads indicate vessels. Data are whisker boxes (min to max, center line indicating median). increased Cx43 within astrocyte cell bodies. Arrows point to patches of dense Cx43 *p < 0.05 (two-tailed Mann–Whitney’s test). h Aquaporin-4 immunoblot in mouse immunoreactivity. Bottom left panels: higher magnifications of Cx43 weblike pat- cerebral cortex from control (n = 4) and Hmgb1ΔAstro (n = 4) mice. Graph shows terns (small arrows) delineating endfeet along a vessel. Bottom right panels: higher quantification of immunoblot signal normalized to the loading control and relative magnifications of Cx43 disruption (yellow asterisks), and of Cx43-high patches to control values. Data are whisker boxes (min to max, center line indicating (yellow arrows), on astrocytes. d Fluorescence micrographs showing Cx43 (cyan) median). All microscopy images displayed in this figure are representative of on CD31+ blood vessels (magenta) from control (left) or Hmgb1ΔAstro (right) mice. experiments repeated in at least 6 mice per group, with similar results. Source data e Average number of clearly-identifiable, Cx43-delineated endfeet (left) and Cx43 are provided as a Source Data file. Hmgb1ΔAstro mice; However, we did not detect differences in AQP4 accordance with the guidelines of the Canadian Council on Animal protein levels. Taken together, like the aberrant distribution observed Care (Breeding Protocol: CMM-3317; Experimental Protocol: for Cx43 in Hmgb1ΔAstro astrocytes, AQP4 distribution at the endfeet CMM-3956). (but not production) may have been affected by HMGB1 ablation. Of note, F-actin cytoskeleton plays a primary role for AQP4 plasma Mouse husbandry. All mice were bred in house and housed maximum membrane localization and during cell adhesion76. While vessel five per cage with free access to food and water. Mice are on a 12/12 branching and density were not affected by loss of astroglial HMGB1, light cycle (7AM On / 7PM Off). Mice undergoing behavioral testing are suggesting normal angiogenesis, brain EC morphology was also on an inverted light cycle. Animal temperature for rodent rooms is severely impaired in P14 Hmgb1ΔAstro mice. Brain ECs in P14 Hmgb1ΔAstro 21 °C–23 °C, with humidity of 40–60%. To assess astrocyte coverage/ mice were reminiscent of ECs from newly formed capillaries77, and density, Aldh1l1-eGFP (BAC) males (Jackson laboratory, Stock No. appeared morphologically closer to ECs we observed in wild-type 026033; B6 background) were crossed with WT “Noncarrier” females brains at P0. Yet, little-to-no transcriptional change was measured in (Jackson laboratory). Conditional knockout of Hmgb1 was achieved by P14 Hmgb1ΔAstro ECs, suggesting a post-transcriptional (contact- breeding Aldh1L1-Cre/ERT2 BAC transgenic male animals88 (Jackson dependent or via released signals) regulation of EC maturation by laboratory, Stock No. 031008; C57BL/6N-congenic background) into a astrocytes. background of female mice carrying a loxP-flanked Hmgb1 gene Changes observed at the gliovascular unit in Hmgb1ΔAstro mice at (Hmgb1flx/flx)89 (Jackson laboratory, Stock No: 031274; B6 background). P14 were followed by neurovascular coupling abnormalities in adults. Animals were backcrossed to generate Hmgb1flx/flx offspring with or This may be due to either aberrant Cx43 signaling in astrocytic without Aldh1L1-Cre/ERT2. Offspring and breeding pairs were all con- endfeet78,79 or abnormal astrocyte coverage affecting vascular smooth firmed by PCR. Littermate mice of both sexes were included in muscle cell contractility12,80. Moreover, adult Hmgb1ΔAstro mice dis- experiments at age postnatal day (P)0, P5, P7, P14, P21 or P50, as played behavioural changes, with reduced repetitive movements, specified in the text and figure legends. For tissue extraction, pups 7- increased locomotion in a cognitive task and more entries in closed day-old and younger were decapitated using sharp scissors. For ani- arms of the elevated plus maze, overall indicative of increased anxiety. mals 14-day-old and older, anesthesia (Ketamine/xylazine, 100 mg/kg While we did not draw causal links, we propose that neurovascular and 10 mg/kg respectively, i.p.) was used prior to euthanasia by cer- deficits and altered distribution of Cx43, which persisted in adult vical disclocation. Hmgb1ΔAstro astrocytes, may contribute to the behavioural phenotype. Of note, Cx43 plays an important role in healthy synaptic transmission Genotyping. Heterozygous Aldh1l1-eGFP offspring were identified and cognition in mice39. The long-term implications associated with through PCR. Primers used were: 5′-GAACAGGCGAAAGCGTTAAG-3′ loss of astroglial HMGB1 on aging and resilience to brain injury remain (23136, forward), 5′-GTAAACCTCCTGGCCAAACA-3′ (18707, reverse), to be established. 5′- CTAGGCCACAGAATTGAAAGATCT-3′ (oIMR7338, forward), and 5′- Cerebrovascular dysfunction, resulting from altered neurovas- GTAGGTGGAAATTCTAGCATCATCC-3′ (oIMR7339, reverse) with PCR cular maturation or neurovascular uncoupling, is a precipitating factor products of 550 bp (transgene) and 324 bp (internal positive control). in neurological disorders81,82. Cognitive decline with reduced cerebral For Cre genotyping, the following primers were used: 5′-GCAAGTT- blood flow has been described as an early marker of vascular GAATAACCGGAAATGGTT-3′ (forward) and 5′-AGGGTGTTA- dementia83 and Alzheimer’s disease84, both of which were linked to TAAGCAATCCCC AGAA-3′ (reverse), with a 250-bp PCR product. For astrocyte dysfunction29. Cerebrovascular impairments in neurodeve- Hmgb1flx genotyping, 5′-AAAGTTTGATGCGAACACG-3′ (forward) and lopmental conditions such as autism spectrum disorders24 and 5′-TGATCTCAAGAGTAGGCACAGG-3′ (reverse), with a 400-bp mutant schizophrenia85 have also been associated with astrocyte and 328-bp WT PCR product. malfunction86,87 and reduced cerebral blood flow in adults26. As aber- rant astrocyte morphogenesis and gliovascular development may Immunohistochemistry and immunofluorescence predispose the adult brain to cognitive decline and/or impair vascular Immunofluorescence for two-dimensional (2D) imaging. For post- responses to neuronal injury, elucidating the role of HMGB1 in glio- natal development of astrocyte and vascular beds, brains from P0, P5, vascular plasticity may lead to therapies for neuroprotection. P7, P14 and P21 Aldh1l1-eGFP (BAC) males were fixed in 4% paraf- ormaldehyde (PFA) overnight at 4 °C. Cortices were embedded in Methods agarose and cut into serial, 50 μm-thick, coronal sections using a Animals vibratome (VT1000S, Leica), then processed free-floating. Sections All animal procedures including euthanasia were approved by the were blocked using a solution containing 10% donkey serum, 0.5% University of Ottawa Animal Care Committee and were conducted in Triton X-100 in 50 mM PBS (‘0.5% PBT’) and 0.5% cold water fish skin Nature Communications | (2023)14:4965 8 Article https://doi.org/10.1038/s41467-023-40682-3 a CD31 / ALDH1L1 Binarized ALDH1L1 / vascular profile outlines Hmgb1flx/flx +Tam. P14 flx/flx Hmgb1 +Tam. n = 5 'Astro Hmgb1 n=5 Frequency (number vascular profiles) 70 60 50 50 Pm 40 30 P = 0.0066 Hmgb1'Astro 20 10 0 0 5 10 15 20 25 30 35 40