Astrocyte-secreted chordin-like 1 regulates spine density after ischemic injury

Ischemic injury occurs when the brain is deprived of blood flow, preventing cells from receiving essential nutrients. The injury core is the brain region directly deprived and is surrounded by the peri-infarct area, the region with recovery potential. In the peri-infarct area neurons undergo acute loss of dendritic spines, which modifies synaptic plasticity and determines neuronal survival. Astrocytes can be protective or detrimental to the ischemic injury response depending on the specific stage, yet we lack clear understanding of the underlying mechanisms. Chordin-like 1 (Chrdl1) is an astrocyte-secreted protein that promotes synaptic maturation and limits experience-dependent plasticity in the mouse visual cortex. Given this plasticity-limiting function we asked if Chrdl1 regulates the response to ischemic injury, modelled using photothrombosis (PT). We find that Chrdl1 mRNA is upregulated in astrocytes in the peri-infarct area in both acute and sub-acute phases post-PT. To determine the impact of increased Chrdl1 on the response to PT we analyzed Chrdl1 knock-out mice. We find that absence of Chrdl1 prevents ischemia-induced spine loss in the peri-infarct area and reduces cell death in the core, without impacting gliosis. These findings highlight the important role of astrocyte-secreted proteins in regulating structural plasticity in response to brain ischemic injuries.

www.nature.com/scientificreports/ the sub-acute phase there is increased expression of genes related to synaptogenesis and spinogenesis 9 , which enhances endogenous plasticity and spontaneous recovery. Thus, the study of structural plasticity in the context of ischemia holds great interest in order to understand the mechanisms that make post-stroke plasticity and spontaneous recovery the most efficient to achieve functional recovery. Astrocytes play important roles in homeostasis of the central nervous system (CNS), and some of these functions are of particular importance in the context of ischemic injuries, pointing to astrocytes as potential targets for neuroprotection 10 . Astrocytes are important components of the neurovascular unit contributing to the integrity of the blood-brain barrier (BBB); they play an important role in glial scar formation; and they are responsible for clearing excessive glutamate from the extracellular space which otherwise triggers excitotoxic events 11 . In addition, some evidence has shown how astrocytes and microglia may be involved in diverse CNS disorders and pathologies, including ischemic stroke 12,13 . Astrocytes, in response to injury and disease, go through changes in their morphology, molecular profile and functionality, a response termed astrogliosis, regulated by signaling pathways that are yet to be fully defined 14 . Astrocytes are not isolated, and in fact interact with a variety of cells that surround the peri-infarct area to regulate different responses such as inflammation or tissue replacement 8 . In response to stroke, a range of inflammatory cells, such as neutrophils, macrophages, T cells and microglia, get activated and infiltrate the injury site, promoting both detrimental and beneficial effects 15 . Microglia are crucial in the regulation of the immune response in the CNS during acute injuries like stroke. They undergo morphological changes and they are recruited to the injury site where they release a series of proinflammatory molecules with potential injurious effects at early stages after injury 16 . At chronic post-injury stages, microglia release anti-inflammatory and neurotrophic factors with neuroprotective functions 16 . Both astrocytes and microglia partake in inflammatory mechanisms during ischemia, which promotes both detrimental and beneficial effects 15 .
In the healthy brain astrocytes modulate the development, maturation and function of synapses through the secretion of diverse factors 17 , and some of these factors have been linked to neuroprotective mechanisms in ischemic injuries. For example, thrombospondins-1 and -2 (TSP-1 and -2) are astrocyte-secreted proteins involved in the formation of silent synapses 18 . Elimination of TSP-1 and -2 leads to a greater loss of synapses in response to ischemic lesions which impairs functional recovery, therefore establishing an important role for astrocyte-secreted proteins in synaptic regulation after stroke 19 . We previously identified chordin-like 1 (Chrdl1), which is enriched in expression in astrocytes in upper layers of the cortex and the striatum, as an astrocytesecreted protein that induces synapse maturation by recruiting GluA2 AMPA glutamate receptors to synaptic sites 20 . During normal development the visual critical period is a time when endogenous plasticity is enhanced in the visual cortex, permitting remodeling of visual circuits in response to a modification in visual input 21 . This endogenous plasticity is greatly diminished at older ages. We demonstrated that the absence of Chrdl1 increases experience-dependent plasticity in a visual sensory deprivation paradigm, an effect that was observed not only during periods of high endogenous plasticity i.e. the visual critical period, but beyond this time into adulthood, identifying Chrdl1 as an important negative regulator of synaptic plasticity 20 . A study blocking a different plasticity-limiting molecule that is expressed by neurons, PirB, found this manipulation also promoted enhanced experience-dependent plasticity, and further proved beneficial in the context of ischemic stroke by reducing the size of the injury, improving motor recovery, and decreasing astrocyte reactivity 22,23 . This suggests that targeting other endogenous plasticity-limiting molecules, such as Chrdl1, may be beneficial in recovery from ischemic injuries.
To enable functional recovery after ischemia, plasticity potential is necessary to promote functional and structural changes to compensate for lost synapses 24 . As Chrdl1 regulates synaptic maturation and limits plasticity, we asked what role Chrdl1 plays in regulating the neuronal response to PT-injury. We found that Chrdl1 is significantly upregulated during the acute phase (24 h post-injury) in both the peri-infarct and contralateral hemispheres, whereas at later timepoints during the sub-acute phase (7 days post-injury) the expression of Chrdl1 remains elevated only in the peri-infarct area, and returns to physiological levels during the chronic phase (30 days post-injury). Elimination of Chrdl1 in vivo reduces cell death and prevents the characteristic spine loss associated with ischemic injuries that has been previously reported 25 . Therefore, we hypothesize that the increase in plasticity-potential in the absence of the astrocyte-secreted protein Chrdl1 reduces cell death, permits faster remodeling in the peri-infarct area and potentially promotes functional recovery.

Results
Chrdl1 expression increases in astrocytes in response to ischemic conditions. In response to ischemia caused by MCAO (middle cerebral artery occlusion), the expression of various astrocytic genes is altered 26 . For example, SPARC and TSP-1, both astrocyte-secreted factors with synaptogenic roles, are upregulated during the acute phase and they have been linked to functional recovery 19,27 . Here we used photothrombosis (PT) to cause ischemic injury in the visual cortex of 4-month-old male mice to study the role of Chrdl1. The effects were analyzed during the acute (24 h post-injury), sub-acute (7 days post-injury) and chronic (30 days post-injury) phases, the post-injury stages characterized by different plasticity levels. The sub-acute phase is considered the time window during which, given the enhanced endogenous plasticity, recovery and neural repair are facilitated 7 .
Due to its role in synaptic plasticity, and previous microarray data showing that the expression of different astrocytic genes is altered following MCAO 26 , we hypothesized that Chrdl1 expression could be regulated in response to ischemia in confined areas of the injured brain. By fluorescence in situ hybridization (FISH), we analyzed mRNA expression of Chrdl1 in the peri-infarct area (ipsilateral, ipsi) and in the homologous contralateral (contra) hemisphere (Fig. 1a). We found that expression of Chrdl1, which we previously demonstrated to be astrocyte-enriched 20 , was upregulated in the peri-infarct area of WT mice during the acute phase (24 h after insult) (Fig. 1b, www.nature.com/scientificreports/ observed in non-injured tissue on the contralateral hemisphere to the lesion (Fig. 1b,c, contra: PT, 2.25 ± 0.24-fold compared to sham). During the sub-acute phase (7 days post-injury) the expression of Chrdl1 in the contralateral hemisphere was restored to physiological levels (Fig. 1d,e contra: PT, 1.47 ± 0.12-fold compared to sham), whereas the increased expression in the peri-infarct area was persistent and significantly increased (Fig. 1d,e ipsi: PT, 4.24 ± 0.47-fold compared to sham). During the chronic phase (30 days post-injury) expression of Chrdl1 returned to physiological levels in both ipsilateral and contralateral hemispheres (Fig. 1f,g, ipsi: PT, 1.11 ± 0.42fold compared to sham, contra: PT, 0.76 ± 0.23-fold compared to sham). We found that upregulation of Chrdl1 was confined to the peri-infarct area, and mainly in the upper layers of the visual cortex, the region that shows highest expression of Chrdl1 in the healthy brain (7 days post-injury, sub-acute phase Fig. 1h).
To determine if the upregulation of Chrdl1 was astrocytic, and not due to upregulation from a non-astrocyte cell type, we used FISH in layers 2/3 of the visual cortex to analyze Chrdl1 mRNA overlap with markers for astrocytes (Slc1a3, also known as Glutamate Aspartate Transporter, GLAST) and neurons (Tubb3) in the peri-infarct area. This agreed with our previous study 20 where we found that Chrdl1 was mainly expressed by astrocytes, and we verified that ischemic injury does not trigger upregulation of Chrdl1 from a non-astrocyte cell population. We found that Chrdl1 mRNA was overlapping with GLAST in the acute phase (Fig. 2a,d ipsi: sham, 74.39 ± 3.75%, PT 80.73 ± 0.76%), the sub-acute phase (Fig. 2b,e ipsi: sham 78.91 ± 1.33%, PT 88.63 ± 1.98) and the chronic phase (Fig. 2c,f ipsi: sham, 63.82 ± 0.97%, PT, 67.88 ± 1.06%) with no significant differences between animals subjected to PT or sham surgeries (Table S1). We found that at every time point, nearly 100% of GLAST + cells in WT mice express Chrdl1 in both sham and PT conditions ( Fig. S2b-d). We also assessed Chrdl1 signal from individual astrocytes, as described before 28 , and found that individual astrocytes (GLAST + cells) significantly increased their Chrdl1 expression at 24 h (Fig. 2g, (Fig. 1). These results suggest that the upregulation of astrocytic Chrdl1 after ischemic injury may trigger mechanisms that hinder synaptic plasticity, potentially obstructing synaptic regeneration and hampering recovery.
We also analyzed the expression of the astrocyte-enriched gene Slc1a3 (GLAST) to see whether the changes in expression that we observed in Chrdl1 were specific to Chrdl1 or shared by other astrocyte genes. GLAST was chosen due to its important roles in glutamate uptake in the context of excitotoxic injuries such as ischemia 29 . We found that GLAST was stable in the peri-infarct area (Fig. S1a,b ipsi: PT, 1.23 ± 0.06-fold compared to sham) and in the homologous contralateral region during the acute phase ( Fig. S1a,b, contra: PT 1.18 ± 0.26fold compared to sham). During the sub-acute phase, and similar to what we observed with the expression of Chrdl1, GLAST expression increased in the peri-infarct area (Fig. S1c,d, ipsi: PT, 1.63 ± 0.21-fold compared to sham), while in the homologous contralateral region there was no significant upregulation (Fig. S1c,d contra: PT, 1.55 ± 0.10-fold compared to sham). During the chronic phase, expression of GLAST returned to physiological levels ( Fig. S1e,f, ipsi: PT 0.62 ± 0.18-fold compared to sham; contra: PT 0.73 ± 0.14-fold compared to sham). These results demonstrate that both Chrdl1 and GLAST expression are upregulated in response to ischemic injury but with distinct time courses, with Chrdl1 upregulation showing more rapid changes than GLAST in response to the same sort of injury.
Reactive astrogliosis is not impaired in Chrdl1 KO mice after PT injury. As a consequence of ischemic injury, astrocytes become reactive in the peri-infarct area and undergo molecular, morphological and functional changes that drive the formation of a border surrounding the core of the injury 30 . Glial fibrillary acidic protein (GFAP) is a cytoskeletal protein that is upregulated in astrocytes in response to different types of CNS injury, and it is an indicator of reactive astrogliosis 30 . In a previous study, elimination of the plasticitylimiting neuronal molecular PirB led to reduction in GFAP levels, indicating a decline in astrocyte reactivity 22 . We assessed whether Chrdl1 played a role in attenuation or intensification of reactive astrogliosis in the periinfarct area by performing PT in male Chrdl1 KO (-/y) and WT (+ /y) mice and analyzing GFAP stained area in the peri-infarct region at different post-injury stages. Overall we found a significant increase of GFAP restricted to the peri-infarct area at 7 days post-injury (Fig. 3a), when GFAP upregulation has been previously reported 31 . During the acute phase we did not detect significant increases in GFAP in either WT or Chrdl1 KO peri-infarct areas (Fig. 3b,c 24 h post-injury, WT: PT 1.31 ± 0.35-fold compared to WT sham; Chrdl1 KO: sham 0.66 ± 0.14fold compared to WT sham; Chrdl1 KO: PT 0.96 ± 0.56-fold compared to WT sham). During the sub-acute phase, we detected an increased intensity in GFAP staining which was significant, but independent of Chrdl1 expression (Fig. 3d,e, 7 days post-injury, WT: PT 14.82 ± 1.65-fold compared to WT sham; Chrdl1 KO: sham 0.56 ± 0.19-fold compared to WT sham; Chrdl1 KO PT 11.22 ± 3.50-fold compared to WT sham). There were no variations in GFAP levels in the contralateral hemisphere during the acute (Fig. S3a,b) or sub-acute phases (Fig. S3c,d). Our results indicate that Chrdl1 does not regulate GFAP, a protein involved in reactive astrogliosis and in the formation of the astrocytic border in certain pathological contexts 32 .
Absence of Chrdl1 does not affect the microglia response after PT injury. Microglia have important roles in neuroinflammation; however, it is unclear what determines their beneficial or detrimental functions in response to ischemic injury. The use of Iba1 has been previously used in the context of stroke to study microglia activation, and it has been shown that in response to ischemia, Iba1 + microglial cells increase in the core of the injury which indicates activation of a microglial response 33 . Thus, we used Iba1 as an appropriate marker to identify the presence of microglia infiltrating to the core of the ischemic injury (7 days post-injury, sub-acute phase Fig. 4a). We assessed if there was a role for Chrdl1 in microglia activation by performing Iba1 staining of WT and Chrdl1 KO tissue after PT. In the core of the injury Iba1 signal was not evident during the first 24 h www.nature.com/scientificreports/ www.nature.com/scientificreports/  www.nature.com/scientificreports/ post-injury corresponding to the acute phase (Fig. 4b,c, 24 h post-injury, WT: PT 0.97 ± 0.55-fold compared to WT sham; Chrdl1 KO: sham 2.09 ± 1.06-fold compared to WT sham; Chrdl1 KO PT 1.76 ± 0.99-fold compared to WT sham), but it was apparent at 7 days post-injury, during the sub-acute phase, in both WT and Chrdl1 KO mice (Fig. 4d,e, 7 days post-injury, WT: PT 21.18 ± 2.02-fold compared to WT sham; Chrdl1 KO: sham 0.67 ± 0.29-fold compared to WT sham; Chrdl1 KO PT 17.42 ± 2.51-fold compared to WT sham). In both WT and Chrdl1 KO mice, differences in microglia recruitment were not observed in the contralateral hemisphere during the acute phase (Fig. S4a,b) or the sub-acute phase (Fig. S4c,d). Our results indicate that Chrdl1 does not have a large effect on the microglia response to ischemic injury.
Absence of Chrdl1 does not affect PT injury volume. One factor that determines the severity of the injury is the size of the brain region affected by the deprivation of blood supply. Increasing plasticity can help repair and remap neuronal circuits that were damaged during the ischemic episode 34 . It has been shown that manipulating astrocytic or neuronal plasticity-regulating molecules contributes to reducing the size of the injury and improved functional recovery 19,22 . In our previous study we found that constitutive Chrdl1 KO mice display enhanced experience-dependent plasticity in the visual system 20 , thus we hypothesized that Chrdl1 KO mice would show decreased injury volume after an ischemic lesion. However, absence of Chrdl1 did not affect the volume of the injury during the acute phase ( Fig. 5a, (Fig. S5c,d). This demonstrates that Chrdl1 does not impact the overall size of the core of the injury. However, use of TTC staining does not inform about the potential survival of some cells in the core of the injury.
Cell death is reduced in Chrdl1 KO mice after PT injury. Although the core of the injury is generally considered tissue that cannot be rescued, some cells may survive and be the basis for future recovery. Without intervention damage from the core can extend into the peri-infarct area, aggravating the functional outcomes. Thus, we investigated the extent of cell death in the core of the injury of Chrdl1 KO compared to WT mice during the acute phase, the post-injury stage when acute cell death occurs 7 . To do this, we assessed in situ cell death using TUNEL technology, and found that Chrdl1 KO mice displayed reduced apoptotic cell death levels compared to WT 24 h post-injury as quantified by the % of TUNEL positive cells within the core of the injury (Fig. 6, WT sham 0.46 ± 0.08%; WT PT 12.15 ± 0.53%; Chrdl1 KO sham 0.83 ± 0.33%; Chrdl1 KO PT 7.72 ± 0.89%). No apoptotic cell death was observed in the contralateral hemisphere (Fig. S6). Although our results indicate a small difference, they suggest that absence of Chrdl1 may attenuate injurious signaling from the core into the periinfarct area, where there is tissue that holds potential for recovery and repair.
Absence of Chrdl1 prevents spine loss in the peri-infarct area. Immediately after ischemia, there is an acute loss of dendritic spines in the peri-infarct area in WT mice. The decreased spine density is severe during the acute phase, with signs of potential recovery during the sub-acute phase 25 . The absence of Chrdl1 does not affect spine density or morphology in uninjured adult mice 20 , but we hypothesized that it may have an effect in response to ischemic injury, as was observed in other studies where plasticity-regulating proteins were eliminated 36 . To this end, WT and Chrdl1 KO mice where layer 5 neurons express the fluorescent protein YFP were subjected to sham surgeries or PT in the visual cortex, and spine density of secondary dendrites extending into layers 2/3 of the visual cortex were analyzed (Fig. 7a). WT mice subjected to PT injury displayed severe loss of spines in the peri-infarct area during the acute phase (Fig. 7b,c, 24 h post-injury: WT sham 1.17 ± 0.08 protrusion/µm; WT PT, 0.50 ± 0.11 protrusion/µm), a well-documented consequence of ischemic lesions 25,37,38 . Interestingly, Chrdl1 KO mice did not show the characteristic ischemic-induced decrease in spine density during the acute phase (Fig. 7b,c, 24 h post-injury: Chrdl1 KO sham 1.10 ± 0.15 protrusion/µm; Chrdl1 KO PT 1.22 ± 0.17 protrusion/µm), suggesting that absence of Chrdl1 prevents the initial degeneration of spines in the peri-infarct area. We observed that spine loss in the peri-infarct area was persistent in WT mice during the sub-acute phase, though not as severe as initially observed (Fig. 7f,g, 7 days post-injury: WT sham 1.46 ± 0.12 protrusion/µm; WT PT 0.91 ± 0.08 protrusion/µm). During the sub-acute phase, Chrdl1 KO mice displayed spine numbers similar to those mice that went under sham surgeries (Fig. 7f,g, 7 days post-injury: Chrdl1 KO sham 1.47 ± 0.15 protrusion/µm; Chrdl1 KO PT 1.25 ± 0.08 protrusion/µm), suggesting that absence of Chrdl1 protects from spine loss rather than delaying the neurodegenerative processes that drive spine elimination after injury. Neither WT nor Chrdl1 KO mice showed variation in spine density in the contralateral hemisphere (Fig. S7a,b,e,f). Previous studies have extensively investigated the link between morphology and maturity of dendritic spines 39,40 . It is established that dendritic spines with mushroom morphologies are characteristic of mature www.nature.com/scientificreports/ spines and considered to establish strong synaptic connections, as opposed to those with thin and stubby morphologies that are highly dynamic and predominant during developmental stages and plasticity processes 39,40 . One possibility is that Chrdl1 KO mice experience spine loss with a fast turnover during the early acute phase (first 24 h post-injury), in which case an increase in immature spine morphologies would be expected compared to mice that went under sham surgeries. We found that WT mice showed a significant decrease in mature spines after injury (Fig. 7d, WT sham 0.33 ± 0.02 protrusion/µm; WT PT 0.10 ± 0.03 protrusion/µm). Chrdl1 KO did not show any alterations in mature spine density in the peri-infarct area during the acute phase (Fig. 7d, Chrdl1 KO sham 0.31 ± 0.06 protrusion/µm; Chrdl1 KO PT 0.33 ± 0.04 protrusion/µm). During the acute phase, WT mice also showed a decrease in immature spine density (Fig. 7e, WT sham 0.84 ± 0.08 protrusion/µm; WT PT 0.40 ± 0.08 protrusion/µm), whereas Chrdl1 KO mice had a constant density of immature spines after injury compared to sham animals (Fig. 7e, Chrdl1 KO sham 0.79 ± 0.12 protrusion/µm; Chrdl1 KO PT 0.89 ± 0.13 protrusion/µm). During the sub-acute phase, the WT mice showed restoration of mature spine density (Fig. 7h, WT sham 0.28 ± 0.06 protrusion/µm; WT PT 0.20 ± 0.01 protrusion/µm), while the Chrdl1 KO mice showed no changes in the mature spine density (Fig. 7h, Chrdl1 KO sham 0.31 ± 0.08 protrusion/µm; Chrdl1 KO PT 0.29 ± 0.06 protrusion/µm). On the other hand, density of immature spines in the WT mice during the sub-acute www.nature.com/scientificreports/ phase remained low (Fig. 7i, WT sham 1.18 ± 0.13 protrusion/µm; WT PT 0.71 ± 0.08 protrusion/µm). During the sub-acute phase Chrdl1 KO mice did not experience reduction of immature spines as the density of the spines with immature morphology remained at comparable levels to the acute phase (Fig. 7i, Chrdl1 KO sham 1.16 ± 0.21 protrusion/µm; Chrdl1 KO PT 0.96 ± 0.02 protrusion/µm). There were no significant morphological changes in the contralateral hemisphere of WT or Chrdl1 KO in response to ischemic insult (Fig. S7c,d,g,h). This indicates that the absence of Chrdl1 protected the peri-infarct neurons from the typical spine loss observed in www.nature.com/scientificreports/ WT mice in response to ischemic injury, maintaining consistent spine density and steady proportion of mature and immature spine morphologies.

Discussion
In this study we investigated the contribution of the astrocyte-enriched synapse-regulating protein Chrdl1 to the response to PT injury. We found that Chrdl1 is upregulated after injury, and that removal of Chrdl1 prevents neuronal dendritic spine loss that is characteristic of the acute phase after injury and reduces cell death during the early stages post-injury. We further found that this protective role of Chrdl1 is independent of reactive gliosis and does not impact on the initial size of the lesion. We found that Chrdl1 expression is upregulated in the acute phase after PT in the peri-infarct area and in the contralateral hemisphere, and that the upregulation is persistent during the sub-acute phase only in the periinfarct area, with return to physiological levels at later time points during the chronic phase (Figs. 1b-g, 2g-i). Contralateral alterations are possible in response to a focal injury as observed before 41 but the mechanisms are yet to be fully described. The expression of Chrdl1 is astrocyte-enriched and ischemic conditions do not trigger expression in non-astrocyte cell populations (Fig. 2). We also found that GLAST (Slc1a3) is increased in expression in the sub-acute phase, but not during the acute or chronic phases (Fig. S1). In ischemia there is a massive release of glutamate into the synaptic cleft that ultimately triggers excitotoxicity and cellular death, and astrocytes play important roles in re-uptake of excessive glutamate including through GLAST 42 . The differential temporal changes of Chrdl1 and GLAST expression in response to PT suggest that other astrocyte factors may also display differential temporal regulation in response to the same kind of injury. In the future it will be important to assess astrocyte gene expression during the acute, sub-acute and chronic phases, as they may dictate the endogenous plasticity levels that determine functional recovery and circuit remodeling in response to injury.
The increased expression of Chrdl1 may represent an inhibitory role of astrocytes in response to ischemic conditions, as Chrdl1 appears to limit plasticity during the sub-acute phase (Fig. 7). This suggests that eliminating Chrdl1 further enhances post-injury endogenous plasticity during the sub-acute phase. Studies in other injury and disease settings found that Chrdl1 is upregulated in an animal model of spinal cord injury 43 , while in chronic disease animal models, such as Huntington's disease and amyotrophic lateral sclerosis, there is no alteration in Chrdl1 expression 44,45 . This suggests that the upregulation of Chrdl1 is context-specific to CNS acute injury, and that the functional role of Chrdl1 that we show here is limited to these settings. In future it will be important to assess Chrdl1 expression in human tissue after acute injury such as ischemic stroke. The results presented here relate to RNA levels of Chrdl1 and suggest that protein secretion may be increased in response to ischemic injury. However, due to technical limitations, protein levels of secreted Chrdl1 in the peri-infarct area have not been possible to assess in the present work, and it will be a necessary analysis for future studies.
Glial cells respond to diverse types of injury and disease through a series of molecular and morphological changes that group under the term reactive gliosis. In the case of astrocytes, reactive astrogliosis in its most severe form leads to the formation of the glial scar 30 . A previous study showed that elimination of plasticity-limiting molecules, such as PirB, leads to reduction in GFAP levels, thus decreasing astrocyte activation 22 . According to our results based on GFAP immunohistochemistry, Chrdl1 absence does not influence reactive astrogliosis (Fig. 3, Fig. S3), suggesting that formation of an astrocytic border surrounding the core of the injury is Chrdl1independent in the context of this model. This can explain why the absence of Chrdl1 does not affect the volume of the injury in Chrdl1 KO mice when compared to WT mice (Fig. 5, Fig. S5), as reactive astrogliosis and the associated border formation is not affected by knocking out Chrdl1. The effects of Chrdl1 appear to be related to neuronal plasticity mechanisms in response to external stimuli, such as sensory deprivation 20 or injury, rather than intrinsic mechanisms of astrocytes themselves that lead to reactive astrogliosis and formation of the astrocytic border. Further experiments to analyze other reactive markers like Lcn2 or Serpina3n among others 26 will clarify whether Chrdl1 regulates astrogliosis in ischemic conditions. The immune response in ischemic injury is regulated by various glial cell types that also interact among themselves to regulate inflammation, and these glia-mediated mechanisms can exert both beneficial and detrimental effects 16 . Astrocytes and microglia are two important components of the neurovascular unit, and both are activated in response to ischemia, as well as other types of CNS injuries 46 . It has been previously shown that activated microglia can induce reactive astrogliosis with neurotoxic effects 47 , but the mechanisms that regulate microglia activation are still not completely clear. In the present study we found that absence of Chrdl1 does not affect microglia activation, as demonstrated by Iba1 immunostaining (Fig. 4, Fig. S4). This suggests that Chrdl1 does not play a role in the inflammatory response in this injury model, an important feature that leads to further damage beyond the core and the peri-infarct areas, contributing to the extension of the injury and worse outcomes. Therefore, the beneficial effects of Chrdl1 removal on dendritic spine number does not appear to be related to inflammation or astrogliosis. A previous study demonstrated that overexpression of Noggin, a BMP (Bone Morphogenetic Protein) antagonist similar to Chrdl1, promoted increased Iba1 + microglia during the sub-acute phase in a mouse model of pMCAO (permanent MCAO) 48 , which promoted reduction in the infarct volume. In future it will be interesting to study the effects of Chrdl1 overexpression on the microglial response and whether this may influence the injury volume.
During the first hours and days following injury, cells in the core and in the peri-infarct area follow apoptotic cascades triggered by excitotoxic events and energy failure due to the interruption of blood flow 49 . Cell survival and the volume of the injury have important implications in the functional outcome from ischemic conditions. Survival of neuronal and non-neuronal cells in the core of the injury can be crucial for repair and recovery, as that prevents release of pro-apoptotic proteins that contribute to increased injury volume 50 . Normally, larger injury sizes are more likely to have fatal outcomes 51 . Lack of Chrdl1 promotes enhanced experience-dependent plasticity in the brains of mice 20  www.nature.com/scientificreports/ injury through reduction of the core volume, as has been shown for other plasticity limiting molecules 22 , and thus, decline in cell death. We found, however, that absence of Chrdl1 did not have any effects on the volume of the injury (Fig. 5, Fig. S5), but Chrdl1 KO mice showed a significant reduction, albeit small, in cell death during the acute phase (Fig. 6). This indicates that absence of Chrdl1 may have beneficial roles during the early postinjury stages. This is in contrast to the study where Noggin was overexpressed, promoting reduction in injury size during the sub-acute phase and improving behavioral outcomes 48 , supporting the hypothesis that different BMP inhibitors may display differences in their mode of action. Interestingly, a previous study on thrombospondin-1 and -2, two astrocyte-secreted synaptogenic proteins, showed that their absence did not play a role in the size of the injury, but impaired functional recovery 19 . However some studies suggest that injury size does not correlate with functional deficits 52 , in contrast with reports from other groups 51 . Our results indicate that although Chrdl1 does not influence injury volume, it may limit apoptotic signaling pathways at early stages post-injury. In the future it will be interesting to assess the levels of other proteins that have been previously identified to regulate the size of the injury and cell death during ischemic conditions, and assess whether their expression is altered in Chrdl1 KO mice to trigger mechanisms that maintain injury volume while promoting cell survival.
In response to ischemic conditions neurons in the peri-infarct region undergo deep remodeling, featuring spine loss and morphological modifications which are especially severe during the acute phase 25 . During later stages post-injury, spontaneous mechanisms promote spinogenesis 6 , but those mechanisms remain unclear. Since we found an increased expression of Chrdl1 in response to ischemia that may hinder synaptic remodeling, we hypothesized that eliminating Chrdl1 could protect spine loss or promote faster spine turnover to compensate for the ischemic damage. Chrdl1 KO mice did not show the characteristic spine loss in the peri-infarct area in response to ischemic lesion as we found no difference in dendritic spine density in the Chrdl1 KO mice between 24 h and 7 days post-PT (Fig. 7, Fig. S7). We did not observe any alteration in spine morphology, suggesting that absence of Chrdl1 protects spines from injury-driven pruning. Time-and region-dependent Chrdl1 upregulation may hinder remodeling of damaged circuits by preventing formation and/or restructuring of neuronal synapses during the sub-acute phase. In the future live imaging experiments will help to determine whether spines in Chrdl1 KO mice are stable and protected in response to ischemic lesion, or experience a rapid turnover after injury. Alternatively, elimination of Chrdl1 may be facilitating a delay in dendritic spine loss, which may offer a wider time window for recovery. Performing electrophysiological recordings from peri-infarct neurons in Chrdl1 KO will inform whether the remaining dendritic spines retain synaptic activity. Additionally, it will be important to perform behavioral analysis to assess if functional recovery is improved by blocking astrocytesecreted Chrdl1 at different time points, or if the protection of spines observed in Chrdl1 KO mice offers a longer time window for recovery.
Our work shows that increased expression of Chrdl1 can be a detrimental contribution from astrocytes that impedes recovery by limiting plasticity and circuit remodeling. Elimination of Chrdl1 may promote protective effects in the visual cortex due to Chrdl1 role as a regulator of AMPAR-mediated synaptic maturation and synaptic plasticity as we previously reported 20 . It is known that different neuronal populations display differential surface levels of GluA2-containing AMPARs in ischemic-like conditions 53 , and this is linked to the differential vulnerability that neuronal subpopulations show to ischemic injury 53,54 . Thus, in the future it will be important to analyze the expression of Chrdl1 in other brain regions in response to ischemic conditions, as Chrdl1 may have variable effects on different neuronal populations. One possibility is that upregulation of Chrdl1 in response to injury is exclusive to upper layers of cortex and striatum where Chrdl1 is expressed in the healthy brain, and additional astrocyte-secreted factors may regulate post-injury plasticity in other brain regions.
An alternative explanation for the potential protective effect in absence of Chrdl1 after injury is through its role as an inhibitor of BMP signaling. It has been previously shown that administration of BMP7 24 h after ischemic injury improves behavioral recovery 55 . Post-injury upregulation of Chrdl1, a known BMP antagonist with affinity for BMP4, 6 and 7 56 , could be blocking the beneficial effects of BMPs after ischemic injury. Blockade of Chrdl1 after ischemia may therefore represent a new approach to promote the neurorestorative roles of BMPs.  www.nature.com/scientificreports/ However, different BMPs may play diverse roles in injury, promoting recovery or delayed cell death. In a different study it was found that overexpressing the BMP antagonist Noggin had neuroprotective effects by reducing the size of the injury 48 . Noggin has high affinity for BMP4, like Chrdl1 56 . Therefore, it is important to determine the specific post-injury phase in which to manipulate Chrld1 in order to promote beneficial effects.
In conclusion, our results indicate that astrocyte-secreted factors play a role in structural plasticity and cell death after ischemic insult. The restricted regional expression of Chrdl1 suggests that alternative astrocytesecreted factors that regulate plasticity and spine density in other brain regions may also mediate plasticity mechanisms and cell death in response to region-specific injury in a time-dependent manner, in an antagonistic or synergistic manner with Chrdl1. This study highlights the importance of dissecting astrocyte-mediated plasticity mechanisms to understand the limitations in circuit remodeling in response to injury and other pathologies of the CNS, and design new strategies to promote neuroprotection and recovery at different timepoints post-injury.

Methods
All animal work was approved by the Institutional Animal Care and Use Committee (IACUC) of the Salk Institute for Biological Studies and performed in accordance with their guidelines. ARRIVE guidelines were followed in all animal experiments.
Animals. Mice were housed in the Salk Institute animal facility at a light cycle of 12 h light:12 h dark, and access to water and food ad libitum.

Photothrombotic ischemic injury.
Male mice at 4 months of age were placed on a stereotaxic frame and anesthetized by 2% isoflurane in oxygen by constant flow via nose cone. Mice were retro-orbitally injected with 10 mg/ml Rose Bengal (Fisher R323-25) in saline (0.9% NaCl) at a dose of 25 mg/kg. Rose Bengal was prepared fresh before surgeries and protected from light. Control mice subjected to sham surgeries were injected with the equivalent volume of saline (0.9% NaCl) with no Rose Bengal, and followed the same procedure. After injection, mice were prepared for surgery during the 5 min allowed for Rose Bengal dye to diffuse. Fur was removed and the scalp was disinfected with alternative swipes of 70% ethanol and betadine. During surgery body temperature was monitored and maintained at 37 ± 0.5 °C with a rectal probe and feedback-controlled heating pad. An incision through the midline with surgical scissors was made to expose the skull and locate Bregma. In this study, the brain region to apply photothrombosis is located at 3.28 mm posterior and 2.80 mm lateral to Bregma. After 5 min from Rose Bengal injection and location of stereotaxic coordinates, the brain region was illuminated through the skull for 10 min with a 520 nm diode laser (Thor labs) set at 2 mm diameter and 10 mW power. Skull was then washed with saline 0.9% NaCl and the incision was closed with Vetbond tissue adhesive (3 M 1469Sb). Triple antibiotic and 2% lidocaine were applied to the closed incision, and the mouse was returned to a clean cage. All analysis were made in the peri-infarct area which has been previously defined as the area contained within 200 µm from the border of the core of the injury 57,58 .
Tissue collection and preparation. Brains were collected at 24 h, 7 days, or 30 days post-photothrombosis (PT) to study the acute, sub-acute or chronic phases, respectively. Mice were injected intraperitoneally with 100 mg/kg Ketamine (Victor Medical Company) and 20 mg/kg Xylazine (Anased) mix and subjected to transcardial perfusion. Perfusion was performed with PBS to obtain fresh frozen tissue for fluorescent in situ hybridization (FISH). Those brains were collected, embedded in OCT (Scigen 4583) and stored at − 80 °C until analysis. For the rest of experiments, perfusion was performed with PBS followed by 4% PFA (paraformaldehyde) to obtain fixed tissue. Brains were collected and stored in 4% PFA at 4 °C overnight, after which they were washed three times with PBS and transferred to 30% sucrose and stored for 3 days at 4 °C for cryoprotection. Those brains were embedded in TFM (General data healthcare TFM-5), frozen in dry ice/100% ethanol and stored at − 80 °C until analysis. For TTC staining, mice were euthanized by intraperitoneal injection of ketamine and xylazine mix and brains were collected without prior perfusion. For every experiment, a minimum of three brain sections were imaged per mouse, and a minimum of 3 mice per condition (sham or PT). N refers to the specific number of mice used in each experiment.

Fluorescent in situ hybridization (FISH) and analysis.
Coronal sections were obtained from WT C57BL/6 J mice that went under sham surgeries or PT and euthanized 24 h, 7 days or 30 days after surgery to analyze expression of Chrdl1. Sections were made at a thickness of 16 µm, cut with a cryostat (Hacker Industries OTF5000) at coordinates corresponding with the core of the injury (3.28 mm posterior and 2.80 mm lateral to Bregma). Fluorescent in situ hybridizations (FISH, ACDbio 320850) were performed according to manufacturer's instructions with some modifications. Brain slices were pre-treated with Protease IV for 20 min. Probes used were Chrdl1 (ACDbio 442811), Slc1a3/GLAST (ACDbio 430781-C2), and Tubb3 (ACDbio 423391-C3). For negative control we used a 3-plex negative probe (ACDbio 320871) to determine the background fluorescence and the cut-off threshold to determine a positive stained cell (Fig. S2) Fig. 1h was taken on a Zeiss LSM 700 confocal microscope using a 20x/0.8 NA objective as a 16-bit image mosaic of stitched tiles with 10% overlap, image size 10,244 × 6556 pixels (pixel size 0.31 × 0.31 µm). Maximal intensity orthogonal projection images were produced with ImageJ for all subsequent analyses. The experimenter performing the analysis was blind to condition (sham or PT). For Fig. 1, fluorescence intensity was measured using ImageJ (NIH, RRID:SCR_003070), and calculated as mean intensity of the ROI multiplied by % of area stained. For Fig. 2d-f, cells co-expressing Chrdl1 signal and Tubb3 or Slc1a3 signal were counted using ImageJ multi-point function. In all cases, a consistent threshold appropriate for visualization of the stained area was applied, using a negative control to determine the cut-off for a positive stained cell. All results were normalized to the sham condition. In Fig. S2b, cells expressing Chrdl1 were counted and plotted as a % of Slc1a3 + cells. For Fig. 2g-i, we used a semi-automatic custom-made macro for ImageJ 28 . Astrocytes were identified using the Slc1a3 signal to create an ROI around each astrocyte. Then a consistent threshold was applied to the Chrdl1 probe signal across samples, and the threshold area was recorded for each astrocyte.
GFAP and Iba1 staining. YFP-Chrdl1 KO and YFP-WT male mice were subjected to sham surgeries or PT, and perfused 24 h or 7 days later with PBS and 4% PFA to collect fixed tissue. Fixed brains were sliced in the cryostat at a thickness of 16 µm and stained for GFAP or Iba1. Upon collection, slices were incubated in blocking solution (5% goat serum, 0.3% Triton X-100 in PBS) for 1 h at room temperature. Slices were then incubated with rabbit polyclonal primary antibody anti-GFAP (Abcam ab7260, RRID:AB_305808) at a dilution 1:500 or rabbit polyclonal anti-Iba1 at a dilution 1:250 (Wako 016-20001, RRID:AB_839506) in antibody buffer (5% goat serum, 100 mM lysine, 0.3% Triton X-100 in PBS) overnight at 4 °C in a humidified chamber. For the negative controls, primary antibody was omitted (Fig. S3E). Slices were washed three times with PBS and incubated with secondary anti-rabbit Alexa 594 (Thermo Fisher Scientific A11012, RRID:AB_141359) in antibody buffer at a dilution of 1:500 for 2 h at room temperature. Slices were washed three times with PBS and mounted using SlowFade gold antifade mountant with DAPI (Thermo Fisher Scientific S36939), covered with coverslips 22 × 50 mm 1.5 thickness and sealed with nail polish. The core and the peri-infarct areas in the ipsilateral hemisphere (Figs. 3a and 4a) and equivalent regions in the contralateral hemisphere were imaged using a fluorescent microscope Zeiss Axio Imager.Z2. Images were taken using a 10x/0.45 NA objective as 16-bit mosaics of 9 tiles with 10% overlap. To measure GFAP signal, a region of interest (ROI) of 1000 × 1000 pixels (pixel size 0.645 × 0.645 µm) was selected in the peri-infarct area, and in the homologous contralateral area. For Iba1 signal measurements, an ROI in the core of the injury was selected. GFAP or Iba1 signal was measured using ImageJ and calculated as stain positive (GFAP + or Iba1 +) area of the ROI. All results were normalized to WT sham.
2,3,5-Triphenyltetrazolium chloride (TTC) staining. Chrdl1 KO and WT male mice were subjected to sham surgeries or PT and brains were extracted without prior perfusion 24 h or 7 days later after euthanasia by intraperitoneal injection of overdose of ketamine/xylazine mixture. Brains were cut in 1 mm thick slices and incubated in a pre-warm solution of 2% 2,3,5-Triphenyltetrazolium chloride (TTC, Sigma T8877) in PBS for 15 min at 37 °C, protected from light. Sections were transferred to cold 4% PFA, and later imaged using × 0.8 magnification on a stereo zoom microscope (Nikon SMZ-445). To ensure the right orientation of the brain sections, a small nick was made on the contralateral ventral region. Injury and hemisphere volumes were measured with ImageJ using the freehand selection tool and measuring the area. Volume was calculated as area by thickness of the slice (1 mm). Injury volume was expressed as % of ipsilateral hemisphere. At least three biological replicates were used for each time point (24 h, 7 days) and condition (sham, PT), and slices at 0.7, 1.7, 2.7 and 3.7 mm posterior to Bregma were analyzed, the slice at 3.7 mm posterior to Bregma being the posterior limit of the infarct. To evaluate the edematous expansion in the Chrdl1 KO compared to the WT we analyzed the difference between the volume of the ipsilateral hemispheres and the contralateral hemispheres, and used calculations previously described 35 : ipsilateral hemispheric volume (%) = [(Σ(I i − C i ))/(ΣC i )]*100, where I i is the area of the ipsilateral hemisphere, including core of the injury, of slice i and C i is the area of the contralateral hemisphere of slice i.
Cell death detection. WT and Chrdl1 KO male mice littermates were subjected to sham surgeries or PT and their brains were collected 24 h later after 4% PFA perfusion. Brains were sliced in 16 µm-thick section and collected on Superfrost Plus micro slides (VWR 48311-703). In situ cell death detection kit TMR red (Roche 12156792910) was used to detect cell death using TUNEL technology in layers 2/3 of the visual cortex in the core of the injury and the contralateral hemisphere. The manufacturer's instructions were followed. Brain sections were fixed with 4% PFA for 20 min at room temperature. Then brain sections were washed with PBS for 30 min at room temperature and permeabilized with 0.1% Triton X-100, 0.1% sodium citrate for 2 min at 4 °C. Brain sections were washed twice with PBS and incubated with the TUNEL mix (label solution + enzyme terminal transferase) for 60 min at 37 °C in a humidified chamber. For the positive controls, brain sections were prior treated with recombinant DNase I for 10 min at room temperature. For the negative controls, brain sections were