Xenotransplantation of Human PSC-derived Microglia Creates a Chimeric Mouse Brain Model that Recapitulates Features of Adult Human Microglia

Microglia, the brain-resident macrophages, exhibit highly dynamic functions in neurodevelopment and neurodegeneration. Human microglia possess unique features as compared to mouse microglia, but our understanding of human microglial functions is largely limited by an inability to obtain human microglia under resting, homeostatic states. We developed a human pluripotent stem cell (hPSC)-based microglial chimeric mouse brain model by transplanting hPSC-derived primitive macrophage precursors into neonatal mouse brains. The engrafted human microglia widely disperse in the brain and replace mouse microglia in corpus callosum at 6 months post-transplantation. Single-cell RNA-sequencing of the hPSC microglial chimeric mouse brains reveals that xenografted hPSC-derived microglia largely retain human microglial identity, as they exhibit signature gene expression patterns consistent with physiological human microglia and recapitulate heterogeneity of adult human microglia. Importantly, the chimeric mouse brain also models species-specific transcriptomic differences in the expression of neurological disease-risk genes in microglia. This model will serve as a novel tool to study the role of human microglia in brain development and degeneration.


2
As the resident macrophages of the central nervous system (CNS), microglia play critical roles in 3 maintenance of CNS homeostasis and regulation of a broad range of neuronal responses 1, 2 . Recent 4 studies indicate that dysfunction of microglia contributes to neurodevelopmental and neurodegenerative 5 diseases, including Alzheimer's disease (AD) 3-7 . Moreover, genome-wide association studies have 6 shown that many neurological disease risk genes, particularly many AD risk genes, are highly and 7 sometimes exclusively expressed by microglia [8][9][10] . These observations provide a compelling incentive 8 to investigate the role of microglia in models of abnormal brain development and neurodegeneration.

9
Most studies of microglia largely rely on rodent microglia. However, there is increasing evidence that 10 rodent microglia are not able to faithfully mirror the biology of human microglia 11 . In particular, recent 11 transcriptomic studies have clearly demonstrated that a number of immune genes, not identified as part 12 of the mouse microglial signature, were abundantly expressed in human microglia 8,12 . Moreover, a 13 limited overlap was observed in microglial genes regulated during aging and neurodegeneration 14 between mice and humans, indicating that human and mouse microglia age differently under normal 15 and diseased conditions 12, 13 . These findings argue for the development of species-specific research 16 tools to investigate microglial functions in human brain development, aging, and neurodegeneration.

17
Functional human brain tissue is scarcely available. In addition, given the considerable 18 sensitivity of microglia to environmental changes 8 , the properties of available human microglia isolated 19 from surgically resected brain tissue may vary significantly, due to different disease states of the 20 patients and the multi-step procedures used for microglia purification. In order to study human microglia 21 in a relatively homeostatic state, many scientists have turned to human pluripotent stem cells (hPSCs).

22
Recent advances in stem cell technology have led to the efficient generation of microglia from hPSCs 23 14-19 , providing an unlimited source of human microglia to study their function. However, when cultured 24 alone or co-cultured with neurons and astrocytes in 2-dimensional (2D) or 3D organoid/spheroid culture 25 systems, these hPSC-derived microglia best resemble fetal or early postnatal human microglia, as 26 indicated by much lower expression of key microglial molecules such as TREM2, TMEM119, and 27 P2RY12 in the hPSC-derived microglia, as compared to microglia derived from adult human brain 28 tissue 16, 18, 20 . Thus, even with these novel in vitro models, it has been challenging to advance 29 understanding of human microglial function in adult ages or in neurodegeneration during aging.

30
Recent studies from us 21, 22 and others 23-25 have demonstrated that neonatally engrafted 31 human neural or macroglial (oligodendroglial and astroglial) progenitor cells can largely repopulate and 32 functionally integrate into the adult host rodent brain or spinal cord, generating widespread chimerism.

33
This human-mouse chimeric approach provides unique opportunities for studying the pathophysiology 34 of the human cells within an intact brain. In this study, we developed a hPSC microglial chimeric mouse 35 brain model, by transplanting hPSC-derived microglia into neonatal mouse brains. The engrafted 11 weeks after plating, hPSC-derived hPMPs emerged into the supernatant and were continuously produced 12 for more than 3 months. The cumulative yield of PMPs was around 40-fold higher than the number of 13 input hPSCs ( Figure 1A), similar to the results from previous studies 16,18 . PMPs are produced in a Myb-14 independent manner that closely recapitulated primitive hematopoiesis 1 . We confirmed the identity of 15 these hPSC-derived PMPs by staining with CD235, a marker for YS primitive hematopoietic progenitors 16 26, 27 , and CD43, a marker for hematopoietic progenitor-like cells 26,27 . As shown in Figure 1B, over 95% 17 of the hPSC-derived PMPs expressed both markers. Moreover, the human PMPs are highly proliferative 18 as indicated by Ki67 staining (95.4 ± 2.2%, n = 4) ( Figure 1B). Using this method, we routinely obtain 19 ample numbers of hPSC-derived PMPs with high purity as required for cell transplantation experiments.

20
We engrafted hPSC-derived PMPs into the brains of postnatal day 0 (P0) immunodeficient mice 21 that are Rag2/IL2rg-deficient and also express the human forms of CSF1, which facilitates the survival 22 of xenografted human myeloid cells and other leukocytes. We deposited cells into the white matter 23 overlying the hippocampus and sites within the hippocampal formation ( Figure 1C). In order to visualize 24 the distribution of donor-derived microglia, at 6 months post-transplantation, we stained the mouse 25 brain sections with human-specific antibody recognizing TMEM119 (hTMEM119). TMEM119 is a 26 marker that is only expressed by microglia, but not other macrophages 14, 17, 28 . We found that the 27 donor-derived hTMEM119 + microglia migrated long distances along the corpus callosum to reach the 28 olfactory bulb ( Figure 1D). The is consistent with the observation in the developing brain that microglia 29 use white matter tracts as guiding structures for migration and that they enter different brain regions 29 .

30
As early as 3 weeks post-transplantation, donor-derived microglia had already migrated along corpus 31 callosum and passed through the rostral migratory stream to the olfactory bulb ( Figure 1E). At 6 months 32 post-transplantation, human microglia widely dispersed in multiple brain regions, including olfactory 33 bulb, hippocampus, and cerebral cortex, and exhibited a highly ramified morphology ( Figure 1F and G) 34 typical of resting microglia. Similar to our previous studies 24, 25 , we assessed the engraftment 35 efficiency and degree of chimerization by quantifying the percentage of hTMEM119 + cells among total 36 DAPI + cells in the forebrain in sagittal brain sections covering regions from 0.3 to 2.4 mm lateral to 37 midline and found that about 8% of the total cells were human microglia in the 6-month-old mouse 38 brains ( Figure 1D and L). Frequently, we also observed clusters of human microglia in the cerebellum 39 ( Figure 1H), which might be a result from the strong ability of immune cells trafficking along blood 40 vessels and/or the choroid plexus 30 . These results demonstrate that hPSC-derived PMPs survive in 41 mouse brain and that they migrate to a variety of structures.

42
To examine whether transplanted hPSC-derived PMPs efficiently differentiated to microglia in 43 the mouse brain, we double-stained brain sections for both human nuclei (hN) and hTMEM119. As  Ki67. As shown in Figure 1K and N, at 3 weeks post-transplantation, about 17% (16.9 ± 5.7%, n = 8) of 51 hN + transplanted cells expressed Ki67, indicating that these cells were capable of proliferating in the 1 mouse brain. At 6 months post-transplantation, the percentage of proliferating cells dramatically 2 decreased and less than 2% (1.7 ± 0.8%, n = 7) of total engrafted cells were Ki67 positive. These Ki67 + 3 proliferating human cells mainly localized in the subventricular zone, the walls along lateral ventricles, 4 corpus callosum, and olfactory bulb ( Figure 1K and supplementary Figure 1B). Taken together, these 5 findings demonstrate that engrafted hPSC-derive PMPs differentiate to microglia, generating a mouse 6 brain with a high degree of human microglial chimerism in the forebrain.

8
Human PSC-derived microglia undergo morphological maturation and are functional in the mouse 9 brain 10 Compared with three weeks post-transplantation, hPSCs-derived microglia appeared to exhibit more 11 complex processes at 6 months post-transplantation ( Figure 1E and F). Moreover, even at the same 12 stage, hPSC-derived microglia in the cerebral cortex seemed to exhibit much more complex 13 morphology, compared with the hPSCs-derived microglia in the corpus callosum and cerebellum 14 ( Figure 1G and H, supplementary Figure 1B

20
and gene expression profiles between wild type and Rag2 -/mice 38 . Building upon that, we also 21 compared the differences between xenografted hPSC-derived microglia vs. host mouse microglia. We 22 double-stained the brain sections with human and mouse specific TMEM119 (hTMEM119 and 23 mTMEM119, respectively) antibodies to distinguish hPSC-derived microglia and mouse host microglia.

24
As shown in Figure 2A, in 6 months old mice, both hPSCs-derived microglia and mouse microglia were 25 seen in the cerebral cortex and hippocampus. Notably, hPSCs-derived microglia seemed to expel 26 mouse microglia, as indicated by the observation that mouse microglia mainly resided in distal regions 27 in the cerebral cortex and hippocampus. Particularly, in the corpus callosum, mouse microglia were 28 rarely seen, and the vast majority of microglia were hPSC-derived microglia, indicating that hPSC-29 derived microglia replaced the host mouse microglia. In the cerebral cortex, hTMEM119 + hPSC-derived 30 microglia exhibited much more complex processes at 8 weeks and 6 months post-transplantation than 31 those cells at 3 weeks post-transplantation, as indicated by the increased number of endpoints ( Figure   32 2C). The total length of processes of hPSC-derived microglia also significantly increased from week 3 33 to week 8 and month 6 ( Figure 2D), suggesting the gradual maturation of hPSC-derived microglia in 34 mouse brain. We further examined the morphological differences between hPSC-derived microglia vs.

50
We then investigated whether hPSCs-derived microglia were also able to prune synapse in the mouse brain. By double staining human microglia marker hCD45 with a synapse marker PSD95, we found that 1 some PSD95 + puncta localized inside of the hCD45 + processes of hPSC-derived microglia. This 2 engulfment of synaptic materials was observed from 3 weeks to 6 months post-transplantation, most 3 prominently seen at 8 weeks post-transplantation, diminishing to only a few at 6 months ( Figure 2G and 4 F). By double staining hCD45 with oligodendroglial marker Olig2, we found that hPSCs-derived 5 microglia in white matter clearly engulfed Olig2 + oligodendroglia at 3 weeks post-transplantation 6 (supplementary Figure 2B). Microglia, together with endothelial cells, pericytes and astrocytes, form the 7 functional blood-brain barrier. We double-stained the brain sections with hCD45 and laminin, a marker 8 that has been commonly used to visualize vascular structures in the mammalian brain 45 . We found 9 that hPSC-derived microglia clustered around and were closely affiliated with blood vessels in both grey 10 matter and white matter across different brain regions including the olfactory bulb ( Figure 2H and 11 supplementary Figure 2C). Taken together, hPSC-derived microglia show variable morphologies in a 12 spatiotemporal manner, morphologically differ from the host mouse microglia, and are functional in 13 mouse brain.

15
Single-cell RNA-sequencing of hiPSC microglial chimeric mouse brain identifies a gene 16 expression signature consistent with adult human microglia

17
Homeostatic human microglia at adult stages are difficult to obtain, because microglia are highly 18 sensitive to environmental changes and microglia derived from adult human brain tissue-derived are 19 usually purified through multi-step procedures that can change their biological properties significantly 8 .

20
In addition, microglia derived from hPSCs using all current differentiation protocols largely resemble 21 fetal or early postnatal human microglia 16,18,20 . We hypothesize that hPSC microglial chimeric mice 22 may provide a unique opportunity to study biological properties of adult human microglia, because the 23 engrafted hPSC-derived microglia are likely to exhibit an expedited maturation process promoted by the 24 maturing environment in the mouse brain. To test this hypothesis, we examined transcriptomic profiles 25 of hiPSC-derived microglia developed in the in vivo homeostatic mouse environment using single-cell 26 RNA-sequencing (scRNA-seq). We collected brain regions where engrafted hiPSCs-derived microglia 27 preferentially dispersed, including the cerebral cortex, hippocampus, corpus callosum, and olfactory 28 bulb, from 6-month-old chimeric mouse brain for scRNA-seq. Owing to the wide distribution and high 29 abundance of hiPSC-derived microglia in those brain regions, we were able to capture ample number 30 of hiPSC-derived microglia for scRNA-seq without using FACS sorting, which has the potential to 31 impact transcriptional profiles through extended ex vivo manipulation. After brain tissue dissociation 32 with papain and centrifugation to remove debris and myelin, single cell suspensions were directly 33 subjected to droplet-based 10X Genomic RNA-seq isolation (  accounted for about 7% of total cells ( Figure S3D). Of note, a cross-correlation analysis of clustered 50 cell types showed that Xeno MG had a highest correlation coefficient value (0.765) with mouse microglia, consistent with a microglial identity of the engrafted human cells ( Figure S3E). Furthermore, 1 the expression of a set of canonical microglial genes (C1QA, CX3CR, TREM2, CSFRR, and P2RY12) 2 was only detected in Xeno MG and mouse microglia clusters ( Figure 3D). We examined the expression 3 of the top 30 human microglial signature genes reported by a previous study 8 . We found that the vast 4 majority of human microglia signature genes were exclusively or abundantly expressed in Xeno MG, as 5 compared to the other types of cells in chimeric mouse brains ( Figure 3E). Furthermore, we compared 6 the transcriptomic profile of Xeno MG with a published dataset generated from human brain tissue-7 derived human microglia 12 . A significant correlation was observed between Xeno MG and the published 8 dataset 12 ( Figure S4A), further confirming the human microglial identity of the engrafted human cells.

9
As shown in Figures 1 and 2, the highly ramified morphology of hiPSC-derived microglia strongly 10 indicate that they exhibit a quiescent and non-activated state. To confirm this, we examined expression 11 of several pro-inflammatory cytokines to assess the impact of the tissue preparation procedures on the 12 microglial state. We found very minor expression of acute pro-inflammatory cytokines such as IL-1β, IL-13 1α and TNF-α 63 64, 65 ( Figure S4B). In contrast, the pro-inflammatory cytokine, IL-6 and an anti-14 inflammatory cytokine, IL-10, were nearly undetectable, and expression of these pro-inflammatory    Figure S4C). Taken together, these results demonstrate that Xeno MG developed in the mouse 42 brain highly resemble adult human microglia and faithfully recapitulate heterogeneity of adult human 43 microglia.

45
Transcriptomic profiling analysis reveals differences between co-resident Xeno MG and mouse 46 microglia Previous studies reported differences in transcriptomic profiles between human and mouse microglia 8, 48 12 . In the chimeric mouse brain, as xenografted hiPSC-derived microglia and host mouse microglia 49 developed in the same brain environment, this model may provide a unique opportunity to directly 50 examine the differences between human and mouse microglia. Xeno MG and host mouse microglia 51 clusters obtained from 4 independent samples of 6-month-old chimeric mouse brains were used for the 1 following comparison ( Figure 4A). We first compared the average levels of microglial gene transcripts in 2 Xeno MG with orthologous gene transcripts in host mouse microglia. Consistent with previous findings 3 8, 12 , the comparison between Xeno MG and mouse microglial transcriptomes demonstrated similar 4 gene expression patterns overall (r 2 = 0.553; p < 2.2 x 10 -16 ), and the majority of orthologous genes 5 pairs (14,488 of 15,058; 96.2%) were expressed within a twofold range ( Figure 4B). Using a cut-off of 6 2-fold difference and an FDR of 0.05, we identified that 248 gene transcripts were preferentially 7 expressed in human microglia, whereas 247 gene transcripts were preferentially expressed in mouse 8 microglia ( Figure 4D, Table S2). Importantly, previously-reported signature genes expressed in human 9 microglia 8 , including SPPI, A2M, and C3, and signature genes expressed in mouse microglia, including 10 HEXB, SPARC, and SERINC3, were all differentially expressed in our sequencing data ( Figure 4B and 11 C), indicating the high fidelity of our samples in resembling previously-identified human vs. mouse 12 microglial gene expression profiles. To explore the function of genes that were highly expressed human 13 microglia, we further performed Gene Ontology (GO) term analysis. Many significantly enriched terms 14 were associated with the innate immune activity of microglia, such as "immune system response," 15 "regulation of immune system process response," and "leukocyte activation," which is consistent with a 16 recent study 12, 69 . In addition, we observed enrichment of genes associated with "regulation of cell 17 adhesion," "cytoplasmic translation," and "peptide biosynthetic process" ( Figure 4E). These results

18
suggest that compared to the host mouse microglia, Xeno MG may be more immunocompetent, as

24
and mouse microglia from our chimeric mouse brain preparation. Expression of disease risk genes, as 25 reported in a recent study 8 , had a highly similar differential expression pattern in co-resident mouse 26 and human microglia ( Figure 4F, G and Figure S4D, S4E). Specifically, with respect to AD, we found 27 that out of 14 AD genes, 10 genes, including Apoc1, Sorl2, and Mpzl1, were more abundantly 28 expressed in Xeno MG than in mouse microglia ( Figure 4F and H). Similarly, out of the 20 PD genes 29 listed in a previous report 8 , 18 genes, such as Vps13c, Snca, Fgf20, Mnnrn1, and Lrrk2, had the same 30 trend of differential expression with greater expression in Xeno MG than in mouse microglia ( Figure 4G 31 and H). We also found that some of the disease genes were preferentially expressed in mouse 32 microglia, such as Syt11and Gba in PD. Altogether, these observations demonstrate that our hPSC 33 microglial chimeric mouse brain can faithfully model the transcriptomic differences between human and 34 mouse microglia, and this new model will serve as a new tool for modeling human neurological 35 disorders that involve dysfunction of microglia.

2
Humanized mouse models, in which the immune system is reconstituted by cells of human origin, have 3 been well-established and provide powerful tools for studying cancer, inflammatory and infectious 4 disease, and human hematopoiesis 71 . However, there are no previous studies reporting a mouse 5 model in which the brain is largely repopulated by human brain-resident immune cells, microglia. In this 6 study, by engrafting neonatal mice with hPSC-derived PMPs, we demonstrate the generation of 7 chimeric mouse brains in which hPSC-derived microglia widely disperse. We propose that the following 8 three reasons may account for the generation of human microglial chimeric mouse brain. First, as 9 compared to other types of neural cells, microglial cells are unique in that they turn over remarkably 10 quickly, allowing the vast majority of the population to be renewed several times during a lifetime 72-74 .

11
Previous studies have shown that neonatally transplanted human macroglial or neural progenitor cells 12 can outcompete and largely replace the host mouse brain cells 21,22,75 . In this study, we also observe 13 that the hPSC-derived PMPs are highly proliferative prior to transplantation and transplanted cells 14 divide for at least 6 months in the mouse host brain. Therefore, the nature of high turnover rate of         (Table S3). Sample/barcode identifiers for the human-specific data were isolated and 24 matching gene symbols were converted from human to mouse. Sample/barcode identifiers not 25 matching this cluster were assumed to be mouse, and these were trimmed to retain only mouse gene 26 symbols matching the homology list. The resulting tables were merged for subsequent analysis in 27 Seurat.

28
For comparisons among sources of human microglia, raw RNAseq reads from the human-29 specific cluster were pooled by sample and aligned with reference human genome (hg38) using

43
Immunostaining and cell counting 44 Mouse brains fixed with 4% paraformaldehyde were processed and cryo-sectioned for 45 immunofluorescence staining 95,96 . The primary antibodies were listed in supplementary Table 4. Slides

49
The relative fluorescence intensity was presented as normalized value to the control group. The cells 50 were counted with ImageJ software. For brain sections, at least five consecutive sections of each brain 1 and at least 7 mice in each group were counted. Engraftment efficiency and degree of chimerization 2 were assessed by quantifying the percentage of hN + cells among total DAPI + cells in sagittal brain 3 sections, as reported in the previous studies 24, 25 . The cell counting was performed on every fifteenth 4 sagittal brain section with a distance of 300 µm, covering brain regions from 0.3 to 2.4 mm lateral to the 5 midline (seven to eight sections from each mouse brain were used).

7
Data analysis 8 All data represent mean ± s.e.m. When only two independent groups were compared, significance was 9 determined by two-tailed unpaired t-test with Welch's correction. When three or more groups were 10 compared, one-way ANOVA with Bonferroni post hoc test or two-way ANOVA was used. A P value less 11 than 0.05 was considered significant. The analyses were done in GraphPad Prism v.5.

13
1 This work was in part supported by grants from the NIH (R21HD091512 and R01NS102382 to P.J.) 2 R.P.H. was supported by R01ES026057, R01AA023797, and U10AA008401. We would like to thank 3 Dr. Brian Daniels from Rutgers University for critical reading of the manuscript. project and wrote the manuscript together with R.X. and input from all co-authors.

11
Competing Financial Interests

12
The authors declare no competing financial interests.