Disruption of the Hoxb8 gene results in chronic anxiety and pathological overgrooming in mice. Using bilateral intracerebral cell transplantation, we demonstrate that mutant Hoxb8 microglia are causative for both behaviors. Mice contain two microglia lineages, Hoxb8 and non-Hoxb8 microglia. We proposed that the two lineages work as a binary system, in opposition to each other with Hoxb8 microglia functioning to reduce anxiety and grooming (function as brakes), whereas non-Hoxb8 microglia increase the levels of both behaviors (function as accelerators). This model makes a strong, unexpected prediction: mice containing only wild-type canonical non-Hoxb8 microglia should exhibit pathological levels of grooming and anxiety. We demonstrate that this is the case, providing strong support for both microglia functioning as a binary system and for the 'Accelerator/Brake' model. Since mice containing only non-Hoxb8 microglia represent mice with a loss of Hoxb8 function due to the absence of Hoxb8 microglia, the more intensive pathology associated with Hoxb8 mutant mice must reflect that mutant mice have both gain and loss of function components. We identify and quantify the relative contribution of each component.

Microglia are the immune cells of the brain [1,2,3,4]. In mice, microglia arise from two hematopoietic progenitor pools with distinct ontogenies and functions [5, 6]. The first, which we refer to as canonical non-Hoxb8 microglia, are derived from hematopoietic progenitors born in the yolk sac at embryonic day (E) 7.5 followed by direct migration into the developing brain at E9.5, following vasculature formation [7]. The second, Hoxb8 microglia progenitors, are descendants of Hoxb8 hematopoietic progenitors also born in the yolk sac, but a day later, which migrate to the AGM (aorta-gonad-mesonephros) and fetal liver, where they are amplified 16 and 280-fold, respectively, prior to their migration into the developing brain starting at E12.5 [5, 6]. Hoxb8 and non-Hoxb8 microglia are further distinguishable because of their differential dependence on Myb function for survival [6]. Importantly, both microglial populations are capable of self-renewal. Once the blood-brain barrier is formed, there is minimal infiltration of microglial progenitors from the periphery under non-pathological conditions. Steady-state levels of both populations of microglia are maintained through a combination of proliferation and apoptosis [5, 7, 8].

In the adult brain the total microglia population is composed of ~75% canonical non-Hoxb8 and ~25% Hoxb8 microglia. Molecularly, the two populations are very similar, with only ~20 genes differentially expressed ([5]; and unpublished results). Both populations of microglia express equivalent levels of Sall1, Tmem119, and P2ry12 that distinguish them from other brain macrophages and reinforce the identity of Hoxb8 microglia as bonafide parenchymal microglia ([5] and herein). The two microglia populations are also distinguishable with respect to function [5, 9, 10].

We previously reported that transplantation of mouse wild-type (WT) bone marrow cells rescued pathological overgrooming in Hoxb8 mutant mice [11], suggesting that the cells participating in the rescue could have been microglia. However, the ontogeny of microglia had yet to be defined [5, 7]. Subsequent papers have clarified that the transplanted cells responsible for the above rescue were not microglia but rather bone marrow-derived macrophages (i.e. monocytes, [5, 6]. Subsequently, it has been shown that monocyte-derived macrophages can infiltrate the mouse brain, irrespective of blood-brain barrier damage, and become microglia-like with respect to their transcription profiles and perform microglia-like functions [12,13,14,15]. Several studies have revealed that the replacement of unhealthy recipient microglia with healthy donor-derived microglia or monocyte-derived macrophages result in the rescue of several neurological disorders [11, 16,17,18,19]. In essence, our prior 2010 experiments were correct. They demonstrated the therapeutic potential of bone marrow-derived myeloid cells, but based on experiments reported subsequently, the interpretation of the Chen et al. 2010 experiments [11] have required adjustment to comply with the later ontogeny data (i.e. rescue of the behavioral phenotype of Hoxb8 mutant mice was accomplished in the earlier experiments by monocyte-derived macrophages, rather than microglia).

Disruption of the Hoxb8 gene in mice gives rise to both chronic anxiety and pathological overgrooming (resembling Trichotillomania, a human obsessive-compulsive spectrum disorder [OCSD]) [5, 9, 11]. Further, the only cells in the brain that are labeled with a Hoxb8 cell lineage marker are a subpopulation of microglia which we now designate as Hoxb8 microglia [5]. Therefore, are defective Hoxb8 microglia causative for the two behavioral pathologies observed in Hoxb8 mutant mice? Herein, we addressed this question by transplantation of purified, cell-sorted Hoxb8 microglia E12.5 fetal liver progenitors derived from WT and mutant Hoxb8 mice, into recipient mouse brains largely devoid of microglia. These two mice are identical except for the source of Hoxb8 microglia progenitors (i.e. Hoxb8 mutant or WT mice). Only mice containing transplanted Hoxb8 microglia progenitor cells derived from Hoxb8 mutant mice show, upon maturation, pathological levels of anxiety and grooming characteristic of Hoxb8 mutant mice. Unlike mice transplanted with Hoxb8 mutant-derived microglia progenitors, mice transplanted with Hoxb8 progenitor cells derived from WT mice on maturation, show WT, low levels of anxiety and grooming. From these experiments we conclude that defective, mutant Hoxb8 microglia are causative for both pathological behaviors in Hoxb8 mutant mice. This conclusion is seminal to our microglia story. The results allow focus on defective, mutant Hoxb8 microglia to determine mechanism of how disruption of the Hoxb8 gene leads to chronic anxiety and pathological overgrooming in Hoxb8 mutant mice.

Two genetic methods were used to generate transplantation recipient mice devoid of microglia, a prerequisite for efficient engraftment of donor microglia: conditional mutagenesis of the Csfr1 gene in Cx3cr1; Csf1r mice and Csf1r mice. The Csf1r gene is required for viability of all microglia [7, 20, 21]. Deletion of both copies of the fms-intronic regulatory element (FIRE) in the Csf1r gene produces mice lacking all microglia, while preserving resident border-associated macrophages (BAMs) [20]. Csf1r homozygous mice have recently been used for cell transplantation experiments in a mouse model of Alzheimer's Disease [16]. Either Cx3cr1; Csf1r mice or Csf1r mice operate as effective recipient hosts for microglia transplantation experiments. However, for two reasons, homozygous Csf1r mice are our preferred recipients for microglia transplantation. First, Csf1r homozygous mice are viable and fertile, which allows 100% rather than 25% of the recipient mice to be used for transplantation experiments. Second, Csf1r homozygous recipient mice contain no detectable endogenous microglia. On the other hand, Cx3cr1; Csf1r recipient mice retain a residue of endogenous microglia (of up to 17% in adult transplanted mice). These results are in keeping with results reported by [16, 20, 22, 23]. Thus, the use of mice containing the ΔFIRE enhancer deletion allows us to generate transplantation recipient mice faster, easier and more efficiently.

In the above experiments, we assume that the transplanted microglia progenitor cells reintroduced into mouse brain devoid of microglia mature into functional microglia. Consistent with this interpretation, transplanting Hoxb8 microglia progenitor cells derived from WT or mutant Hoxb8 mice exhibit very different but predictable behaviors. To further test this query, we also determined that mice transplanted with WT Hoxb8 microglia progenitor cells are capable of, upon maturation, inducing grooming behavior in response to optogenetic stimulation of these cells in the medial prefrontal cortex (mPFC) as previously shown in untransplanted mice [10].

We have recently reported that optogenetic stimulation of WT Hoxb8 microglia in specific regions of the brain induces anxiety, grooming, or both, whereas concurrent optogenetic stimulation of both populations of microglia, Hoxb8 and non-Hoxb8 microglia, do not [10]. To resolve those two observations, we proposed a model for microglia function in which the two populations of microglia function as a binary system in opposition to each other: WT non-Hoxb8 microglia function to increase levels of anxiety and grooming (function as accelerators), whereas WT Hoxb8 microglia function to lower the levels of these two behaviors (function as brakes). This model (the Accelerator/Brake model) provides a biological reason for the existence of the two populations of microglia in mice and provides a mechanism by which these two populations, when working in concert, provide fine-tuning of the levels of anxiety and grooming to adapt to changing environmental conditions [10]. The model makes a strong prediction. Mice containing only canonical WT non-Hoxb8 microglia, should, in the absence of "brakes", exhibit high and potentially pathological levels of grooming and anxiety compared to mice containing only WT Hoxb8 microglia, or the appropriate mixture of both WT Hoxb8 and WT non-Hoxb8 microglia. Herein, using microglia cell transplantation, we illustrate that mice having only purified WT non-Hoxb8 microglia do exhibit high pathological levels of both anxiety and grooming. In contrast, mice containing only WT Hoxb8 microglia, no microglia, or the appropriate ratio of both microglia exhibit low, WT levels of both behaviors. These experiments provide strong support for the Accelerator/Brake model of microglia function.

Significantly, the levels of anxiety and overgrooming observed in mice transplanted with only WT non-Hoxb8 mice are not as high as those observed in Hoxb8 mutant mice. Since mice containing only WT non-Hoxb8 microglia do not contain any Hoxb8 microglia, by definition their behavior should reflect a "pure" loss of the Hoxb8 function phenotype. However, Hoxb8 mutant mice show even higher levels of pathological overgrooming and anxiety than mice containing only WT non-Hoxb8 microglia. This observation argues that the disruption of the Hoxb8 gene in mice generates not only a loss of Hoxb8 gene function, but also a gain of function component that contributes to the mutant phenotype. In genetics, a gene disruption that results in both loss and gain of function components is not uncommon and often occurs when the gene product participates in protein complexes. Hox proteins commonly participate in transcriptional repressive complexes. The direct comparison of grooming behavior of mice containing only non-Hoxb8 microglia with mice mutant for Hoxb8 allowed identification and quantification of the gain and loss of function components of the Hoxb8 gene disruption that characterize Hoxb8 mutant mice.