New Tools for Genetically Targeting Myeloid Populations in the Central Nervous System

As the resident macrophages of the brain and spinal cord, microglia are crucial for the phagocytosis of infectious agents, apoptotic cells and synapses. Developmentally, microglia originate from the embryonic yolk sac and serve important roles in the sculpting of neonatal neural circuits. During brain injury or infection, bone-marrow derived macrophages invade neural tissue, making it difficult to distinguish between invading macrophages and resident microglia. In addition to circulation-derived monocytes, other non-microglial central nervous system (CNS) macrophage subtypes include borderzone (meningeal and perivascular) and choroid plexus macrophages. To distinguish between resident microglia and these other CNS macrophage subtypes, we generated a P2ry12-CreER mouse line. P2RY12 is a microglial-specific nucleotide sensing GPCR that is important for microglial response to tissue damage. Using immunofluorescent labeling and flow cytometry experiments, we show that P2ry12-CreER recombination is exceptionally specific to parenchymal microglia. We also perform ribosome immunoprecipitations and transcriptional profiling of P2ry12-CreER recombined cells, using a Cre-dependent Rpl22-HA mouse line. By identifying genes enriched in this dataset that are not correspondingly enriched in a broader Cx3CR1-CreER; Rpl22 dataset, we isolate a number of borderzone macrophage-specific transcripts, including the gene PF4. Using a PF4-Cre mouse line, we show that PF4 expression robustly marks borderzone macrophages. Together, we demonstrate two new methods to genetically target distinct CNS macrophage subtypes.


Cx3cr1-CreER
, have significant advantages compared to their predecessors, but still have a number of drawbacks. For instance, Cx3cr1 is expressed in other immune cell types in the brain, including perivascular, choroid plexus and meningeal macrophages (Goldmann et al., 2016). Cx3cr1-CreER also labels circulating monocytes, which results in the labeling of peripherally-derived cells in the context of CNS injury or disease. Furthermore, Cx3cr1-CreER mouse lines are haplo-insufficient for Cx3cr1, which may affect microglia functions, such as synaptic pruning (Rogers et al., 2011).
Here, we describe a P2ry12-CreER knock-in mouse line that we have developed to specifically label brain microglia. P2RY12 is a nucleotide sensing metabotropic GPCR in the P2Y family of GPCRs that has an important role in the microglial "sensome" (Hickman et al., 2013). P2RY12 regulates the microglial response to tissue damage and the adoption of the amoeboid "activated" microglial morphology (Bernier et al., 2019;Haynes et al., 2006). We found that P2ry12-CreER, unlike the commonly used Cx3cr1-CreER, specifically labeled brain microglia, but not perivascular or meningeal macrophages. Using this line, we performed ribosomal immunoprecipitations and transcriptional profiling of "pure" microglia. By comparing this dataset to existing ribosomal profiles of CNS macrophages and subtracting genes that are shared by both datasets, we identify a number of borderzone macrophage markers, including PF4. Analysis of the PF4-Cre mouse line shows that PF4-Cre recombination robustly labels perivascular and meningeal macrophages. Altogether, we show that P2ry12-CreER and PF4-Cre can be used to genetically target parenchymal microglia and perivascular macrophages, two distinct CNS macrophage subtypes.
To genetically label P2RY12(+) microglia, we generated a P2ry12-2A-CreER knock-in mouse line (hereafter called P2ry12-CreER) ( Figure 1A) using CRISPR-facilitated homologous recombination. The targeting construct was designed to preserve P2ry12 expression, with the P2ry12 stop codon replaced by a ribosome-skipping 2A-fusion peptide coding sequence, followed by the coding sequence for CreER. We did not find any health problems in either P2ry12-CreER heterozygous or homozygous mice. Immunolabeling of homozygous and heterozygous P2ry12-CreER mice showed no significant change in P2RY12 protein expression or microglial morphology in homozygous compared to wild-type mice ( Fig  1A) indicating that gene knock-in did not affect gene expression or function.
Consistent with our flow cytometry results, we saw no recombination in perivascular macrophages (LYVE1 + macrophages associated with αSMA immuno-labeled vessels), unlike Cx3cr1-CreER, which robustly labels these cells ( Figure 1E). Similarly, we saw no recombination in the meninges of P2ry12-CreER; Ai14 mice, unlike Cx3cr1-CreER; Ai14 mice, which showed recombination in a significant portion of LYVE1 + meningeal macrophages. In the choroid plexus, P2ry12-CreER recombination marked a small number of IBA1 + cells, as well as a small number of IBA1-negative or IBA1-low cells, unlike Cx3cr1-CreER, which showed recombination in most choroid plexus IBA1 + macrophages ( Figure 1G). Some of the P2ry12-CreER; Ai14 recombined cells appeared to be on the surface of the choroid plexus, consistent with a recently published study that revealed a microglia-like, P2ry12-expressing "CP-epi" subpopulation of choroid plexus macrophages, which may be Kolmer's epiplexus cells (Ling et al., 1998;Van Hove et al., 2019).

P2ry12-CreER recombination efficiently and specifically labels embryonic microglia
In our previous studies, we noted that P2RY12 expression is developmentally regulated, increasing in the brain throughout embryonic and early post-natal development (Arnold et al., 2019). To determine whether P2ry12-CreER was capable of recombining embryonic macrophages, we induced pregnant dams carrying P2ry12-CreER; Ai14 pups with a series of three tamoxifen doses at E13.5, E15.5, and E17.5 ( Figure 2A). We collected these pups at E18.5 and analyzed their degree and pattern of recombination. Examination of these brains revealed widespread recombination in IBA1 + parenchymal microglia. Recombination was seen in essentially all parenchymal IBA1 + cells, which were most dense in the subventricular zone and lateral migratory stream of the amygdala, where at this age there are significant numbers of migrating GABAergic interneurons ( Figure 2B-B'', C, D). Unlike the widespread parenchymal IBA1 + recombination, we observed sparse recombination in the embryonic meninges that was limited to LYVE1 + macrophages ( Figure 2C, E). Embryonic recombination in IBA1 + cells of the choroid plexus was widespread, unlike in the adult ( Figure 2F). We did not observe any embryonic recombination in blood vessels or other cell types in the brain at this age.

Expression of P2ry12-CreER in other tissues and in circulating monocytes
To determine the specificity of P2ry12-CreER-mediated recombination, we examined recombination in non-microglial cell types in the brain, circulation and in a variety of organs. In the brain, we saw no recombination in macroglia (oligodendrocytes and astrocytes), or neurons ( Figure 3A). In organs of P2ry12-CreER; Ai14; Cx3Cr1-eGFP mice, we found sparse recombination that often overlapped with EGFP expression, suggesting P2ry12 expression mainly marks a small subset of macrophages in other tissues ( Figure 3B). Sparse recombination was seen in all organs analyzed (spleen, intestine, heart, skin, lung, liver, muscle, thymus, spleen), except for the kidney, which did not show any recombination (not shown). Strong splenic recombination was seen in a subset of Cx3cr1-eGFP + cells of the marginal zone between the red and white pulp, an area that contains macrophages that surveil the bloodstream (Borges da Silva et al., 2015). In the lung, intestine and heart, P2ry12-CreER; Ai14 recombination occurred in a subset of Cx3cr1-eGFP cells; in the liver and thymus, P2ry12-CreER; Ai14 recombination partially overlapped with Cx3cr1-eGFP.

P2ry12-CreER-dependent ribosomal profiling enriches for transcripts of true brain microglia
We next studied the mRNA transcriptional profile of true microglia using P2ry12-CreER mice. Classically, mRNA profiling of defined cell types requires preparation of single cell suspensions followed by cell sorting, RNA isolation and sequencing. These manipulations are inefficient and introduce bias, especially in macrophage populations, which are difficult to mechanically or enzymatically separate. Furthermore, microglia become activated and show significant transcriptional changes upon dissociation and sorting (Haimon et al., 2018a). To avoid these pitfalls, we performed TRAP-Seq (translating ribosome affinity purification followed by mRNA sequencing) (Sanz et al., 2009) which allows for cell-specific isolation of mRNA without these manipulations. We generated P2ry12-CreER; Rpl22-HA mice, which, upon recombination, express a hemagglutanin (HA) epitope tagged RPL22 riboprotein that can be isolated by immunoprecipitation ( Figure 4A). Upon tamoxifen-induced recombination in P2ry12-CreER; Rpl22-HA mice, we found robust labeling of microglia by Rpl22-HA, as shown by immunofluorescent costaining for HA and IBA1 ( Figure 4B). qPCR analysis of immunoprecipitations from P2ry12-CreER; Rpl22-HA mice revealed significant enrichment for microglial transcripts such as Tmem119, P2ry12 and HexB (not shown). After this qPCR validation, libraries from P2ry12-CreER; Rpl22 immunoprecipitations were generated and sequenced (workflow in 4C). Analysis of the transcripts enriched in P2ry12-CreER; Rpl22 libraries revealed significant enrichment for many known microglial markers, but did not show enrichment for markers of other cell types, such as neurons, oligodendrocytes and astrocytes ( Figure 4D).
The transcriptional profile of P2ry12-CreER recombined cells was quite similar to recently published profiles derived from Cx3cr1-CreER; Rpl22 immunoprecipitations (Haimon et al., 2018b). By comparing the relative enrichment values for individual genes in our P2ry12-CreER; Rpl22 data (X axis, Figure 4E) to Cx3cr1-CreER; Rpl22 enrichment values for the same gene (Y axis, Figure 4E), we identified a number of enriched genes that were either common or unique to each dataset. Genes common to both datasets included macrophage markers such as Pu.1, Aif1 (which codes for IBA1), Csf1r and CD11b. In line with our histological and cell-sorting data however, there were many genes enriched in Cx3cr1-CreER; Rpl22 IPs that were not enriched in P2ry12-CreER; Rpl22 IPs. Some of these genes have recently been identified in single-cell sequencing studies as markers of border-associated macrophages (Van Hove et al., 2019), suggesting that our comparison differentiates genes expressed in true microglia from those expressed in non-parenchymal macrophages in the CNS. This subtractive approach also identified a number of genes that were preferentially enriched in the P2ry12-CreER; Rpl22 population, which represent potential microglial specific markers. As expected, P2ry12 was in this group, as were a number of other P2Y and P2X family members, such as Gpr34, P2ry10b, 13, 6 and P2rx1, 7.
One gene that had significant enrichment in Cx3cr1-CreER; Rpl22 IPs but marked depletion in P2ry12-CreER; Rpl22 IPs was the gene PF4, also known as Cxcl4 ( Figure 4E). Given its strong enrichment in Cx3cr1-CreER expressing cells and lack of enrichment P2ry12-CreER expressing cells, PF4 appeared to be a good candidate marker of perivascular and/or meningeal macrophages. We explored the lineage of cells marked by PF4 expression by examining cells recombined by a PF4-Cre transgenic mouse line (Tiedt et al., 2007). Indeed, we found that in PF4-Cre; Ai14 mice, perivascular and meningeal macrophages are strongly recombined ( Figure 4F). PF4-Cre; Ai14 mice also had a low degree of microglial recombination, which appeared as clumps of cells scattered throughout the brain, suggestive of sporadic recombination and subsequent clonal expansion.

Discussion:
In recent years, knock-in and transgenic mouse lines using the fractalkine receptor gene Cx3cr1 (e.g. Cx3cr1-Cre, -CreER and -eGFP) have been some of the best and most frequently used tools for targeting microglia (Goldmann et al., 2013b;Parkhurst et al., 2013). Other strategies, targeting genes such as Tie2 (Cre, CreMERM), CD11b (Cre, L10aGFP), and CSF1R (LacZ, eGFP, CreMERM) (M. L. Bennett and F. C. Bennett, 2019;Wieghofer et al., 2015) have been used in landmark studies to determine the yolk-sac origin of microglia, and to characterize the transcriptional architecture of these cells. However, despite their usefulness, many of these microglia-targeting mouse lines mark other myeloid cell types in the CNS. For instance, Cx3cr1-based mouse lines mark borderzone and choroid plexus macrophages in addition to microglia. Cx3cr1 is also expressed in circulating T cells, B cells, NK cells and monocytes, which means that in the context of injury, Cx3CR1-Cre and eGFP are incapable of distinguishing between resident macrophages and other invading immune cells ( Bar-On et al., 2010;Fogg et al., 2006;Imai et al., 1997;Liu et al., 2009). The tamoxifen-inducible Cx3cr1-CreER allele avoids this problem to a degree, but requires a significant waiting period after the time of recombination for recombined circulating cells to be depleted. This waiting period is inconvenient for most models of CNS injury and is incompatible with timesensitive studies, such as models of neonatal brain injury. In addition to issues regarding specificity, Cx3cr1 knock-in alleles (eGFP, Cre, and CreER) (Goldmann et al., 2013a;Parkhurst et al., 2013) disrupt the Cx3cr1 gene, which may result in Cx3cr1-dependent effects on synaptic pruning (Rogers et al., 2011). Aside from Cx3CR1, knock-in GFP and CreER lines have been generated for the microglial-enriched transcription factor Sall1 (Buttgereit et al., 2016). However, significant caveats for these lines exist, including the fact that Sall1-GFP and CreER both disrupt the function of the Sall1 gene, and that Sall1-CreER marks a broad range of non-myeloid CNS cell types (Chappell-Maor et al., 2019).
Here, we show by immunohistochemistry, flow cytometry, and ribosomal profiling that P2ry12-CreER has a number of advantages over existing microglial targeting Cre and CreER mouse lines. One such advantage is that P2ry12-CreER-recombination avoids circulating monocytes, meaning that in the context of injury, resident microglia can be distinguished from invading monocyte-derived macrophages. P2ry12-CreER also avoids targeting borderzone macrophages of the meninges or perivascular space. As perivascular inflammation is a hallmark of many neurological diseases, such as multiple sclerosis and Alzheimer's disease, the ability to separately manipulate microglial and borderzone macrophages in mouse models of these diseases will be critical.
The P2ry12-CreER mouse line is particularly well-suited for the labeling and tracking of microglia during CNS brain injury. By genetically labeling microglia, transcriptional profiling can be performed at different stages of CNS injury, from the initial inflammatory response to later regenerative stages, without contamination by invading peripheral monocytes. One major barrier to studying microglia in the context of injury is that microglia undergo significant transcriptional changes when activated, making it difficult to distinguish activated microglia from other macrophage subtypes. For instance, upon activation, microglia have been shown to downregulate "true" microglia markers (Tmem119, P2ry12) and upregulate other macrophage markers (CD45, CD86, Cd53, TLR4, TLR7) (Haimon et al., 2018b;Jordão et al., 2019). Microglial activation occurs in response to CNS injury, but is also induced by methods used to prepare microglia for analysis by flow cytometry. By genetically labeling and profiling microglia with P2RY12-CreER and TRAP-seq, these transcriptional changes associated with flow cytometry can be avoided (Haimon et al., 2018a).
The recent publication of a number of Tmem119-targeted mouse alleles (Tmem119-GFP, RFP, CreER) serve to highlight some of the advantages of the P2ry12-CreER line described here (Kaiser and Feng, 2019;Ruan et al., 2020). Regarding the specificity of Tmem119-based lines, both Tmem119-CreER and -GFP appear to show low but not insignificant levels of recombination or expression in brain-derived CD11b+; C45int/hi myeloid cells (1.9% and 6.9% for CreER recombination and GFP expression respectively). In the circulating blood, no Tmem119-GFP(+) monocytes were reported, but 7 days after the last tamoxifen injection, 3% of circulating CD45(+) monocytes were labeled by Tmem119-CreER. In a separate publication, using a Tmem119-RFP mouse line and a percoll gradient to isolate CNS macrophages, 5.3% of RFP+ cells of CNS macrophages were shown to be CD11b + ; CD45 hi , similar to the Tmem119-GFP report. In the neonatal P1 brain, Tmem119-GFP showed widespread expression overlapping with the vascular endothelial cell marker, CD31. In adulthood, prolonged administration of tamoxifen to Tmem119-CreER mice induced recombination in non-IBA1(+) cells. This recombination was quite strong in IBA1(-) meningeal cells and in cells adjacent to large penetrating blood vessels, likely representing meningeal and perivascular fibroblasts, respectively. Indeed, based on single cell sequencing, Tmem119 is highly expressed in CNS resident fibroblasts (http://betsholtzlab.org/VascularSingleCells/database.html). In summary, despite robust labeling of microglia in Tmem119-based lines, recombination in CD11b+; C45int/hi CNS cells and circulating monocytes by flow cytometry, the recombination in mural and meningeal cells, and the significant degree of Tmem119-GFP endothelial expression in neonatal brains suggest that P2ry12 may be a more useful gene for genetic targeting of microglia.
P2ry12-CreER can be used to delete floxed alleles in embryonic or adult microglia. Here, we show that in addition to the fluorescent TdTomato-expressing Ai14 reporter mouse line , P2ry12-CreER can be used to recombine the Rpl22-HA ribotrap allele for microglia-specific transcriptional profiling. Interestingly, using a subtractive approach, we found a number of genes that had very high enrichment in Cx3cr1-CreER; Rpl22 samples but no enrichment in P2ry12-CreER; Rpl22 samples. Some of these markers (Mrc1, F13a1, Ms4a7 and MMP9) were recently identified in other studies that used physical dissociation of the dural membranes from the neural parenchyma to separate borderzone macrophages from other CNS macrophage subtypes (Jordão et al., 2019;Van Hove et al., 2019). After isolating macrophages from these preparations, these groups performed single-cell sequencing to identify gene signatures of microglia, border-associated and choroid plexus macrophages. One major drawback to strategies that rely on the physical dissection of these populations is that this process results in significant levels of cross contamination from microglia and other cells that that are in close contact with the meningeal layers. Furthermore, physical dissection of the meningeal layers does not separate perivascular macrophages of the penetrating vessels from the parenchymal microglia. In contrast, P2ry12-CreER recombination provides a more precise and complete genetic segregation of these populations that minimizes the transcriptional changes that are known to occur during sorting. The differential enrichment observed in our subtractive approach revealed several transcripts only expressed in border-associated macrophages. We focused on one such gene, PF4, and found that PF4-Cre robustly and preferentially recombines borderzone macrophages. We did observe a low degree of sporadic microglial recombination in PF4-Cre; Ai14 mice, which may be due to short-term expression of PF4, perhaps upon microglial activation, or developmental labeling. Together, our studies find that with P2ry12-CreER and PF4-Cre, both parenchymal and borderzone CNS macrophages can be genetically dissociated.
CNS macrophages regulate a wide range of processes from early development to adulthood in health and disease. Here, we provide improved methods for genetically targeting microglia and borderzone macrophages. Use of these genetic tools will open new experimental avenues to better define the specific functions of CNS macrophage subclasses. For instance, microglial or borderzone macrophages ablation experiments can be performed with conditional Csf1R ablation, or conditional diphtheria toxin receptor expression. Another particularly useful function of these lines will be to track and transcriptionally profile distinct macrophages classes in the context of CNS injury or neurodegenerative disease.

Material and Methods:
Mice: P2ry12-CreER mice were generated by CRISPR-facilitated homologous recombination. The P2ry12 coding sequence was joined to CreER by deletion of the P2ry12 stop codon and in-frame 3' insertion of P2A-CreER. The P2A fusion peptide contains a 5' GSG motif, which has been shown to significantly increase fusion protein cleavage efficiency (Szymczak-Workman et al., 2012). P2ry12-CreER mice were generated by pronuclear injection of the P2ry12-CreER targeting construct and a Cas9+gRNA expression plasmid (Biocytogen) into fertilized mouse eggs, followed by adoptive embryo transfer. Founder mice and progeny were screened by PCR, and proper genomic integration was verified by Southern blot and genomic sequencing. Recombination was induced by three doses of tamoxifen dissolved in corn oil, administered by oral gavage every other day (150uL of 20mg/mL). For embryonic mouse inductions, pregnant dams were given tamoxifen (150uL of 20mg/mL) on E13.5, E15.5 and E17.5, for a total of three gavage injections. All mouse work was performed in accordance with UCSF Institutional Animal Care and Use Committee protocols. The P2ry12-CreER mouse line will be deposited at Jackson Labs and at the Mutant Mouse Resource and Research Center (MMMRC).
Flow cytometry: Microglial flow cytometry was performed using a percol gradient isolation strategy. Single-cell suspensions were prepared and centrifuged over a 30%/70% discontinuous Percoll gradient (GE Healthcare) and mononuclear cells were isolated from the interface. Flow cytometry was performed on a FACS Aria III using the FACSDiva 8.0 software (BD Biosciences) and was analyzed using FlowJo v10.6.1. TdTomato+ cells were analyzed for CD45 and CD11b expression following initial gatings for forward scatter, singlets and live cells. Blood smears and flow cytometry were performed on adult mice following three tamoxifen inductions as described above. Blood was collected from the facial vein and smears were imaged for EGFP and TdTomato expression. Prior to immunostaining, blood samples used for flow cytometry were depleted of erythrocytes using an ammonium chloride lysis buffer (.15M NH4Cl, 10mM NaHCO3, 1.2mM EDTA) wash. Blood flow cytometry was performed on a BD FACSVerse machine and analyzed using FlowJo v10.6.1.

Ribosomal immunoprecipitations:
Whole brains of P2ry12-CreER; Rpl22-HA and Cx3cr1-Cre; Rpl22-HA mice were dounce homogenized in 5mL of homogenization buffer (1% NP-40, .1M KCL, 50 mM Tris pH 7.4, .02 mg/mL cycloheximide, .012M MgCl2, RNAse inhibitors, Heparin, .5 mM DTT). Tissue lysates were centrifuged at 10k-rcf for 10 min at 4 degrees to obtain a supernatant ribosomal fraction. An 80uL input sample was taken at this point to serve as a comparison to immunoprecipitated samples. Rabbit monoclonal anti-HA antibody (Cell Signaling C29F4, 2.5uL per mL) was added to the supernatant and incubated for 4 hours at 4 degrees. Iron-linked Protein A beads were added to the ribosomal fraction/antibody mixture (40uL per mL) and incubated for an additional 4 hours before being magnetically precipitated. Immunoprecipitate/beads were washed with a high salt buffer (.3M KCL, 1% NP-40, 50mM Tris pH 7.4, .012M MgCL2, cyclohexamide, .5 mM DTT) 3 times and then eluted with 350 uL of buffer RLT (Qiagen RNeasy kit). Input and immunoprecipitated RNA was purified on column (Qiagen RNeasy kit) and frozen as aliquots at -80 degrees. qPCR: cDNA from ribosomal immunoprecipitations was generated using Maxima First Strand cDNA Synthesis Kit (Thermo scientific) and qPCRs were performed using SYBR Green qPCR Mastermix (Bimake) on a BioRad CFX Connect thermocycler. Quantitative comparisons between immunoprecipitations and input samples were made using delta-delta Ct normalization, with GAPDH as a normalizing control.

Library preparation and sequencing:
Total RNA quality was assessed by spectrophotometer (NanoDrop, Thermo Fisher Scientific Inc., Waltham, MA) and with an Agilent Fragment Analyzer (Agilent Technologies, Palo Alto, CA). All RNA samples had integrity scores greater than 9, reflecting high-quality, intact RNA. RNA sequencing libraries were generated using the Nugen Universal Plus mRNA-Seq kit with multiplexing primers, according to the manufacturer's protocol (NuGen Technologies, Inc., San Carlos, CA). Fragment size distributions were assessed using Agilent's Fragment Analyzer DNA high-sensitivity kit. Library concentrations were measured using KAPA Library Quantification Kits (Kapa Biosystems, Inc., Woburn, MA). Equal amounts of indexed libraries were pooled and single end 50 bp sequencing were performed on the Illumina HiSeq 4000 at the UCSF Center for Advanced Technology (CAT) (http://cat.ucsf.edu).

RNA sequencing analysis:
Raw reads were pseudo-mapped to the mouse transcriptome using Salmon 0.13.1 (Patro et al., 2017). Quantified transcripts were passed to Tximport 1.12 for gene-level summarization and differential expression analysis was performed with DESeq2 1.24, using alpha = 0.05 (Love et al., 2015) (Love et al., 2014) . Genes with total read counts below 20 in all samples were eliminated from the analyses. Data has been submited to the Gene Expression Omnibus (GEO) repository for datasets, the accension number for this dataset is: GSE138333.