Combining irradiation, bone-marrow transplant, and microglial depletion, achieves long-term and complete peripheral myeloid cell engraftment in the brain
Here, we set out to explore as proof-of-principle the clinical feasibility of a pharmacological-based method for near complete myeloid cell replacement. It is well-known that BM-derived cells enter the brain following whole-body irradiation (49), however, cell engraftment is limited. To this end, we and others (22), recently developed a unique paradigm, which utilizes whole body irradiation, BM transplant, and a CSF1R inhibitor (CSF1Ri) treatment to achieve near complete parenchymal engraftment by donor peripheral myeloid cells (Fig. 1A-C). Although this method presents therapeutic potential for the replacement of dysfunctional myeloid cells, no studies to date have investigated the long-term consequences of such replacement. Given that microglia, non-parenchymal macrophages, and monocytes represent distinct heterogeneous populations of myeloid cells and it remains unclear whether these cells share overlapping functional capabilities, we set out to investigate the long-term consequences of microglial replacement with bone marrow (BM)-derived peripheral myeloid cells. To accomplish this, three experimental groups were generated: 1) WT CON: C57BL/6 wildtype (WT) control mice, 2) GFP-BM CON: WT mice subjected to whole body irradiation and retro-orbital infusion of donor CAG-green fluorescent protein (GFP) BM cells, and 3) GFP-BM REPOP: WT mice subjected to whole body irradiation, GFP-BM transplant, and CSF1Ri-induced repopulation. It should be noted that WT CON received no irradiation, transplantation of GFP BM cells, or CSF1Ri treatment. Peripheral percent chimerism was evaluated after ~ 12 wks via blood granulocytes and averaged ~ 98.0% in both irradiated groups (Fig. 1G). Previous studies have shown that high-dose cranial irradiation increases blood-brain barrier (BBB) permeability, albeit transiently (50), and so animals were provided with a 3.5mo BBB recovery period prior to further treatment. After 3.5mo, mice were treated for 14d with the CSF1R inhibitor PLX3397 at a dose of 600 ppm to deplete the microglial compartment. PLX3397 was then withdrawn, effectively stimulating myeloid cell repopulation. In previous studies, we and others have shown that repopulation of the microglial niche derives from the proliferation of surviving cells (7, 21, 51). However, in this study we found that repopulation under exceptional conditions (i.e. following whole body irradiation and microglial depletion) derives from BM-derived myeloid cells and results in the near complete replacement of the microglial niche with BM-derived peripheral myeloid cells (i.e. monocytes) within 14d. To explore the long-term consequences of this engraftment, animals remained on control chow for an additional 6mo prior to sacrifice (~ 10mo after irradiation). At this timepoint, peripheral percent chimerism was evaluated via BM-collected hematopoietic stem cell (HSC) GFP expression and averaged ~ 97% in both irradiated groups (Fig. 1H).
Immunohistochemical examination of brains ~ 10mo after irradiation and ~ 7mo after PLX3397 treatment revealed the extent of BM-derived cell parenchymal engraftment. As expected, no GFP + cells were found in control brains (Fig. 1D). Irradiation alone, without microglial depletion/repopulation, caused partial engraftment of GFP + BM-derived cells, particularly in meningeal and perivascular spaces, in line with prior reports (Fig. 1D; (52, 53)). Approximately 20% of parenchymal IBA1 + cells were GFP+, indicating that irradiation alone allows for limited access and engraftment of monocytes into the CNS (Fig. 1E). On the other hand, irradiated mice treated with CSF1Ri, allowing for myeloid cell repopulation (GFP-BM REPOP mice), displayed extensive (~ 80%) GFP + myeloid cell chimerism, essentially replacing the entire microglial tissue with BM-derived cells (Fig. 1E-F).
Lasting effects of irradiation on BBB integrity, neurogenesis, and cell proliferation
Previous studies have shown that irradiation, with accompanying monocyte infiltration, has lasting effects on BBB integrity (50, 54–56). Despite an observed influx of peripheral-derived myeloid cells, BBB integrity was not compromised, as assessed by Prussian blue, IgG, and fibrinogen staining ~ 10mo post-irradiation (Fig. 2A-D).
Another potent and long-term effect of irradiation involves impairing neurogenesis and cell proliferation (57, 58). In line with these studies, we observe a significant decrease in doublecortin, a marker for neuronal precursor cells/neurogenesis, in the subgranular zone in the dentate gyrus of the hippocampus, a major site of neurogenesis in the mouse brain, following irradiation, which remains reduced following monocyte infiltration (Fig. 2E-F). Monocytes are short-lived cells, with a half-life of approximately 1-7d in humans, whereas studies have shown that microglia exhibit much slower turnover rates with an average lifespan of 4.2 years in humans (59, 60). To assess how the engrafted monocyte population is maintained over time, we explored cell proliferation using Ki67 and found no significant differences in IBA1 + cell proliferation. In fact, we detected no to little Ki67 + IBA1 + cells in the cortex across all groups (Fig. 2G-H). In addition to examining these known long-term effects of irradiation, these data also suggest that once peripheral donor cells engraft the CNS, they are long-lived in the brain with very little cell turnover.
Myeloid Cell Characterization After Long-term Peripheral Myeloid Cell Engraftment
Studies have shown that monocytes downregulate their canonical markers upon infiltration and differentiation in the CNS. Moreover, the loss of the homeostatic microglial signature (e.g. Sall1, Pu.1, Tmem119, Cx3cr1, and P2ry12) is associated with CNS disorders (61), thus we sought to characterize myeloid cell surface marker expression and morphology following irradiation and microglial replacement. Immunohistochemistry utilizing ionized binding adaptor molecule 1 (IBA1), a marker common to all myeloid cells, and P2RY12, a microglial-specific marker, was performed on brains collected ~ 7mo following CSF1Ri-induced peripheral myeloid cell infiltration (Fig. 3A). GFP-BM REPOP mice exhibited an elevated number of IBA1 + cells compared to controls (Fig. 3A-D and quantified in 3D). These IBA1 + myeloid cells also displayed distinct morphological differences from controls; in which they displayed smaller cell bodies (Fig. 3E), larger dendrites/cell processes (Fig. 3F) and reduced dendritic complexity (Fig. 3G); consistent with BM-derived cells exhibiting more amoeboid-like morphology compared to microglia.
As expected, all IBA1 + cells in control brains exhibited extensive P2RY12 immunoreactivity, and all GFP + cells (i.e. BM/peripheral-derived cells) displayed little to no P2RY12 + staining (Fig. 3A, H). In addition, it appears that irradiation induces a significant loss of P2RY12 expression. Similar results were seen with TMEM119 staining (Fig. 3I, J). These data indicate that the engrafting BM-derived myeloid cells, even following 7mo of residence in the brain, continue to lack microglial signature expression. Furthermore, monocyte-engrafted brains displayed significantly increased CD68, a lysosomal marker associated with a heightened activation or phagocytic state, staining compared to controls and irradiated brains (Fig. 3K-L). Collectively, these results show that although monocytes can take up residence in the brain, they maintain a unique cell surface marker and morphological identity within the brain parenchyma, even after extended periods of residence in the brain.
After evaluating all IBA1 + cells, we next sought to delineate infiltrating monocytes from microglia unaltered by treatment, referred to as remaining microglia. Here, remaining microglia form pockets of P2RY12 + spaces within the brain (Fig. 3A; outlined by white dotted lines). Analysis of images in regions of extensive infiltrating monocyte engraftment or extensive remaining microglia presence shows that the loss in the microglial homeostatic signature (i.e. TMEM119) is specific to monocytes. Expression of TMEM119 in remaining microglia is not significantly altered by irradiation or repopulation (Fig. 3I, J). In addition, we also analyzed AXL, a protein enriched in phagocytic cells, and observe significant increased expression of AXL in infiltrating monocytes in both irradiated groups, as well as, remaining microglia in GFP-BM REPOP. However, AXL expression appears most elevated in infiltrating monocytes in GFP-BM REPOP mice (Fig. 3M-O).
Transcriptional Changes Following Long-term Peripheral Myeloid Cell Engraftment
After detecting alterations in myeloid cell expression, we continued our investigations by addressing the impact of long-term peripheral myeloid cell engraftment at the transcriptional level. Since previous studies have shown that microglial heterogeneity is detected in a distinct brain region-specific manner by bulk tissue RNA sequencing (RNAseq) (62–65), we performed RNAseq analysis on three brain regions (cortex, hippocampus, and thalamus/striatum). Gene expression can be explored at http://rnaseq.mind.uci.edu/green/long-term_monocytes/. To initially identify gene expression changes due to both irradiation and the brain-wide presence of monocytes, we compared the transcriptomes of GFP-BM REPOP mice to controls (WT CON), which resulted in 2695 differentially expressed genes (DEGs; FDR < 0.05) across all brain regions. Notably, the effects were most prominent in the thalamus (2401 DEGs), followed by the hippocampus (370 DEGs), with few DEGs detected in the cortex (36 DEGs; Fig. 4A). We next examined transcriptional changes due to irradiation alone (i.e. WT-CON vs. GFP-BM CON), and due to the presence of monocytes (i.e. GFP-BM CON vs. GFP-BM REPOP; Fig. 4A) and display the results as volcano plots (Fig. 4B-D). Separating the effects of irradiation and monocyte infiltration, we observe that the hippocampus and thalamus are more vulnerable to irradiation-induced gene changes, compared to the cortex, which show little to no changes. Of note, the most significant transcriptional changes due to monocyte infiltration occur exclusively in the hippocampus. Overall, these results demonstrate that different brain regions exhibit selective vulnerabilities to irradiation and the infiltration of monocytes. Using the DEGs present in the transcriptional comparison between WT CON vs. GFP-BM REPOP mice across all three brain regions, we were able to build a unique long-term monocyte-related signature (Fig. 4E). Among these genes most prominently upregulated in monocyte-engrafted mice, we identified Apobec1, the chemokine receptor Ccr1, the C-type lectin Mrc1, and several members of the Ms4a cluster (Ms4a6b, Ms4a6c, and Ms4a7). In this comparison known microglial specific genes were also downregulated, including Crybb1, Slc2a5, Siglech (selected genes shown in Fig. S1A).
To understand and place transcriptional changes associated with microglial-monocyte replacement and irradiation within three distinct brain regions on a broader system-level scale, we next conducted weighted gene co-expression network analysis (WGCNA) (48). Twenty-two distinct modules were identified and given color-based names (Fig. 4F). The correlation between modules is shown in Fig. S1B, highlighting modules that behave similarly to one another. We initially focused on modules that exhibited the highest correlation with monocyte presence in the brain (“treatment” variable) and identified a single module: darkgreen (Fig. 4G-I). Eigengene values for darkgreen showed consistent increases in GFP-BM REPOP mice across all three brain regions, reflected in the provided heatmap (Fig. 4H). Furthermore, cell-type enrichment analysis (Fig. 4P) demonstrated that darkgreen genes are myeloid-enriched and included previously identified peripheral myeloid-related genes: Apobec1, Apoe, Clec7a, Ms4a6c, and Ms4a7. Accordingly, pathway analyses showed gene-ontology (GO) terms associated with immune responses, including endocytosis (Fig. 4I). In addition to identifying a monocyte-specific signature module, we also identified several modules associated with irradiation across all three brain regions. Here, we provide one module: salmon (Fig. 4J-L). Cell-type enrichment analysis shows that the salmon module is significantly enriched with interneuron genes (Fig. 4P). Functional annotation of the salmon module shows that it is enriched in gene-ontology (GO) terms related to DNA repair, including RNA splicing and processing, consistent with the known effects of radiation on DNA and cranial irradiation on the brain (Fig. 4L). It should be noted that several modules (red, lightgreen, and brown) were identified reflecting that monocyte engraftment lead to the reversal of irradiation-induced effects, specifically in the hippocampus (Fig. 4M-O). Cell enrichment analysis revealed the brown module to be highly enriched for astrocyte-expressed (astro) genes, and pathway analysis shows that these genes are involved in cilium organization and assembly (Fig. 4P). Cilia are small microtubule-based signaling projections that are known to regulate cell division (66). Analysis shows the red module (Fig. S1C-E) to be enriched for pyramidal neurons in the somatosensory (SS) cortex and lightgreen module (Fig. S1F-H) enriched for pyramidal neurons in the CA1 region of the hippocampus (Fig. 4P). Pathway analysis of the red module shows it is enriched in GO terms related to macroautophagy, synaptic transmission and axon development. Together, these data indicate that irradiation and the presence of repopulating monocytes play a role in altering astrocyte and neuronal-related genes in the brain, as well as, exhibits distinct effects in different brain regions, including the cortex and hippocampus.
The differential and brain region-dependent effects of long-term peripheral myeloid cell engraftment on astrocytic and neuronal properties
To further explore these differential effects, we next assessed astrocytes and neurons in various regions of the brain using immunohistochemical analysis. Astrocytes, the other major glial cell of the CNS, were stained using astrocyte markers glial fibrillary acidic protein (GFAP) and S100β (Fig. 5A-C). GFAP+ (Fig. 5D) and S100β+ (Fig. 5G) cells were significantly elevated in the cortex of GFP-BM REPOP mice. A significant elevation in S100β + cells was also found in hippocampus in these mice (Fig. 5H), but no differences were detected in the number of GFAP + cells (Fig. 5E), indicating that monocyte-engrafted brains show heightened astrocyte numbers for specific astrocyte subsets in a brain region-dependent manner. Furthermore, we observed a significant increase in GFAP + staining intensity in the hippocampus of BM-GFP CON mice, indicating that astrocytes in the hippocampus of irradiated mice are not elevated in cell number, but appear more activated, and that this activation is reversed by monocyte engraftment (Fig. 5F). These data are in line with transcriptional findings that monocyte engraftment reverses irradiation-induced astrocyte-associated gene changes (Fig. 4M-P).
In addition to astrocyte alterations, we also explored neurons and associated structural properties (e.g. axons and synapses). No gross differences were detected in neuronal numbers following irradiation or long-term monocyte engraftment, as assessed by neuronal marker NeuN (Fig. 5I-J). However, as has been previously reported by others, microtubule-associated protein 2 (MAP2) was significantly decreased in both brain regions (i.e. cortex, hippocampus) in GFP-BM CON mice (Fig. 5K, N-O) (67–69). Consistent with previous findings in the hippocampus, a significant reduction in MAP2 expression was not seen in GFP-BM REPOP mice, indicating a reversal in irradiation-induced changes following monocyte engraftment. Next, we explored synaptic alterations using postsynaptic density 95 (PSD-95; Fig. 5K), a major scaffolding protein expressed on excitatory synapses, and synaptic vesicle glycoprotein 2A (SV2A; Fig. 5L), a synaptic vesicle protein expressed ubiquitously in the CNS present on GABAergic and glutamatergic presynaptic terminals (70). A significant loss of PSD95+ synapses was found in GFP-BM REPOP mice in the cortex, and a significant loss was found in GFP-BM CON mice in the hippocampus but not in GFP-BM REPOP mice (Fig. 5K, O-P). In the cortex, monocytes appear to facilitate PSD95 losses. In line with this, we observe PSD95+ staining within IBA1+ bodies in GFP-BM REPOP mice in the cortex (Fig. 5K1). No significant alterations in SV2A staining were detected (Fig. 5L, Q-R). Together, these data indicate that irradiation and monocyte engraftment can have differential effects on astrocytic and neuronal properties in a brain-region dependent manner. These data also confirm that in the hippocampus monocyte engraftment appears to reverse some irradiation-induced effects.
Effects of long-term peripheral myeloid cell engraftment on behavior and cognition
After observing changes at the cellular level, we next sought to determine whether irradiation and/or myeloid cell replacement leads to short or long-term alterations in behavior or cognition. To accomplish this, mice underwent behavioral and cognitive testing 1mo and 6mo following CSF1Ri-induced repopulation (Fig. 1A). Mice were first tested on the elevated plus maze (Fig. 5A-D). Neither irradiated (GFP-BM CON) nor repopulated (GFP-BM REPOP) mice exhibited changes in anxiety during this task at 1mo (Fig. 6B) or 6mo (Fig. 6D) post-repopulation compared to controls (WT CON), and no motor differences were recorded during this task (Fig. 6A,C). Despite the lack of changes in the elevated plus maze, we did observe significant differences during open field analyses in distance traveled (Fig. 6E) in both transplanted groups compared to WT CON at 1mo, showing that irradiation impacts locomotion. However, these significant alterations were transient and not detected at 6mo (Fig. 6H). To assess recognition memory, mice were tested using the novel place recognition test. In this task, we observed no significant differences between groups (Fig. 6F-G, 5I-J), however, we do observe a slight increase in memory as measured by preference index in GFP-BM REPOP mice at 1mo. To further investigate these memory alterations, hippocampal-dependent learning and memory was assessed using the contextual fear conditioning apparatus (Fig. 6K-N). No significant differences were detected between groups as measured by % freezing at 1mo (Fig. 6K-L) and 6 mo (Fig. 6M-N), indicating that hippocampal learning and memory is not altered by irradiation or the presence of monocytes. Additional behavior and cognitive testing were added at 6 mo, including the Crawley’s sociability test (Fig. 6O-P) and spontaneous alternation Y-maze (Fig. 6Q-S). No significant differences in general sociability were found between groups, as assessed by time spent in empty and stranger mouse-filled chambers (Fig. 6P). The spontaneous alternation Y-maze serves as an index for active retrograde working memory, which relies on the proper functioning of many brain regions, including the hippocampus and cortex (71). Here, we observed the most overt behavioral alterations as a result of irradiation. Both GFP-BM CON and GFP-BM REPOP exhibited significantly reduced zone alternation frequency compared to WT CON (Fig. 6S, indicating that working memory may be impaired by irradiation. It should be noted that the number of total arm entries was slightly reduced in GFP-BM REPOP mice (Fig. 6R). Together, these findings indicate that by most behavioral tests irradiation and monocyte engraftment result in little changes to behavior and cognition; however, there appears to be a significant long-term deficit in retrograde working memory, which is associated with the temporal lobe and prefrontal cortex.