|Home | About | Journals | Submit | Contact Us | Français|
Mouse Hoxb8 mutants show unexpected behavior manifested by compulsive grooming and hair removal similar to humans with OCD-spectrum disorder trichotillomania. Since Hox gene disruption often has pleiotropic effects, the root cause of this behavioral deficit was unclear. Here we report that, in the brain, Hoxb8 cell lineage exclusively labels bone marrow-derived microglia. Further, this pathological behavior is rescued by transplantation with wild-type bone marrow. It has been suggested that the grooming dysfunction results from a nociceptive defect, also exhibited by Hoxb8 mutant mice. However, bone marrow transplantation does not rescue the sensory defect. Also, disruption of Hoxb8 in the hematopoietic lineage recapitulates pathological grooming, without conferring nociceptive insensitivity. Conversely, disruption of Hoxb8 in the spinal cord, results in generating the sensory defects, without induction of pathological grooming. Immunological dysfunctions have been associated with neuropsychiatric disorders but the causative relationships are unclear. In this mouse, a distinct compulsive behavioral disorder is associated with mutant microglia.
Grooming in mammals is an innate, stereotypic behavior with a well defined syntax (Berridge et al., 1987). The head is invariably groomed first, followed by body regions and finally the anogenital region and tail. This cephalocaudal progression of grooming is defined as the “syntactic groom chain”. Previous studies have shown that multiple regions of the rodent brain, notably the brainstem, striatum and cortex are used to implement the syntactic groom chain (Aldridge, 1993; Berridge, 1989; Berridge and Whishaw, 1992).
Mice homozygous for a loss of function mutation in Hoxb8 show excessive grooming. The syntax of grooming appears normal, but the number of incidences per unit time and the duration of grooming bouts are increased (Greer and Capecchi, 2002). However, the behavior is pathological, for it leads to hair removal and self-inflicted open skin lesions at the over groomed sites. This behavior is very similar to that described for humans with the OCD-spectrum disorder, trichotillomania, where compulsive removal of hair is also a hallmark. This disorder is quite common in humans with an occurrence ranging from 1.9-2.5 per 100 in seven separate international communities (Horwath and Weissman, 2000). Curiously, these mutant mice also excessively groom their wild type cage mates. This aspect of the phenotype suggested that the peripheral nervous system is not likely responsible for the excessive grooming behavior (Greer and Capecchi, 2002).
Hoxb8 mutant mice also show altered response to nociceptive and thermal stimuli, which have been attributed to deficiencies in the formation and organization of interneurons in the dorsal spinal cord laminae I and II that receive the majority of nociceptive inputs (Holstege et al., 2008). Holstege, et al. further suggested that the excessive and pathological grooming defects previously described in Hoxb8 mutants result from to these sensory spinal cord defects.
It was quite unexpected that disruption of a Hox gene should result in a distinct behavioral deficit such as excessive and pathological grooming (Greer and Capecchi, 2002). Hox genes are normally involved in establishing body plans by providing positional values along the major axes of the embryo (Capecchi, 1997). However, Hox genes also have direct roles in the formation of multiple tissues and organs, including the formation of the hematopoietic system, and, with respect to Hoxb8, maintenance and differentiation of myeloid progenitor cells, one of the two known sources of microglia (Kawasaki and Taira, 2004; Krishnaraju et al., 1997; Perkins and Cory, 1993).
Since implementation of grooming in rodents is rooted within the brain, we anticipated that Hoxb8 would be expressed in a neural circuit that modulates grooming behavior. Instead we have surprisingly found that, in the brain, the site that generates and implements grooming behavior, the only detectable cells derived from Hoxb8 cell lineage are microglia. Secondly, we demonstrate that normal bone marrow transplantation into lethally irradiated Hoxb8 mutant mice rescues the excessive pathological grooming behavior, without correcting the spinal cord defects. Thirdly, conditional restriction of Hoxb8 deletion to the hematopoietic system results in mice with the excessive grooming and hair removal behavioral defects, without induction of the nociceptive/spinal cord defects. Finally and conversely, conditional deletion of Hoxb8 in the spinal cord, generates mice with the spinal cord sensory defects, but with normal grooming behavior.
These experiments strongly support the hypothesis that the excessive pathological grooming behavior observed in Hoxb8 mutant mice originates from defective microglia, thus directly connecting hematopoietic function to mouse behavior. The extensive role of microglia, as the brain's monitor and responder of immune activity, in the normal function of our brain is becoming increasingly apparent. As examples, immunological dysfunctions have been widely linked to many psychiatric disorders including obsessive compulsive disorder (OCD), major depression, bipolar disorder, autism, schizophrenia, and Alzheimer disease (Ashwood et al., 2006; da Rocha et al., 2008; Kronfol and Remick, 2000; Lang et al., 2007; Leonard and Myint, 2009; Strous and Shoenfeld, 2006). In addition, results from genome wide association studies suggest that genes whose dysfunction have been implicated in immune dysfunction and/or signaling, contribute to increased susceptibility to the above mentioned mental disorders (Hounie et al., 2008; Purcell et al., 2009; Shi et al., 2009; Stefansson et al., 2009).
Unfortunately, animal models that directly associate distinct behavioral deficits with defective microglia have been lacking. Here we provide such a model which should allow interrogation, at the molecular genetic and cellular levels, the roles of microglia in promoting normal behavior and how perturbation of microglia leads to pathological behavior.
Hoxb8 mutant behavior is characterized by excessive pathological grooming. Previously, we determined the number and duration of grooming bouts from continuous video recording of mouse activity (Greer and Capecchi, 2002). This procedure was robust, but very labor intensive. More recently, we have been using technology developed by B. V. Metris based on the use of very sensitive vibration detectors (Laboras platforms). Each activity such as drinking, eating, rearing, climbing, locomotion, immobility, grooming and scratching is associated with characteristic patterns of vibration, which are continuously recorded. A computer algorithm then interprets specific vibration patterns as individual behaviors. The advantage of this approach is that behavior is classified automatically, and collected non-obtrusively, over any chosen period of time. We typically monitor activity over 24-hour periods. We have evaluated the Laboras platforms for their assessment of time spent grooming by co-monitoring mouse activity using our video camera system. The Laboras platforms are remarkably accurate, comparable to human classification and far less labor intensive. Supplemental Figure S1A shows the average time spent grooming for 25 Hoxb8 mutant mice and 22 controls over 24-hour periods as measured by the Laboras platforms. These results compare very well with those previously obtained by analyzing continuous video recordings, illustrating that on average Hoxb8 mutant mice spend approximately twice as much time grooming as their wild-type littermates (Greer and Capecchi, 2002). The penetrance for excessive grooming in Hoxb8 mutant mice is 100%.
The expression pattern of Hoxb8 in the adult brain is broad (Greer and Capecchi, 2002). However, the expression level is very low and dispersed in the brain making it difficult to identify the cell type(s) expressing Hoxb8. To identify the Hoxb8 cell lineage in the mouse brain, we generated a Hoxb8-IRESCre driver that could be used to activate Cre dependent LacZ or YFP reporter genes targeted to the ubiquitously expressed ROSA26 locus (supplemental Figure S1B; Soriano, 1999). In mice carrying both the Hoxb8-ICre driver and ROSA26-YFP reporter alleles, activation of Hoxb8 expression also triggers YFP production. Brains of such mice were collected at pre and post-natal stages and examined by immunohistochemistry. In adult brains, YFP positive cells can be found throughout the brain, but consistent with previous results, predominantly in the cerebral cortex, striatum, olfactory bulb and brainstem (data not shown). These cells appear morphologically to be microglia and indeed co-express the general microglia marker CD11b and Iba1, a marker of activated microglia (Figures 1 A-C and data not shown) indicating that Hoxb8 is expressed in microglia or their progenitor cells. Notably, not all microglia in the brain are YFP positive, suggesting that Hoxb8 expression was present only in a subpopulation of microglia or their progenitors (~40% of total, see Figure 7C).
In newborn mice, very few Hoxb8-labeled cells are observed in the brain, and these cells are found predominantly in the choroid plexus (Figure 1D), meninges, and the ventricular lining, with their numbers declining with increased distance from the ventricular zone. This gradient suggests migration of YFP positive cells from the ventricular zone into the forebrain areas. Between P2 and P14, YFP positive cell count in the mouse brain increases dramatically and is then maintained at this high level (Figures 1E-F and data not shown).
Although the origin of microglia is still debated, there is general agreement that at least one subpopulation is of bone marrow origin, [i.e. derived from circulating monocytes; (Kaur et al., 2001; Ransohoff and Perry, 2009)]. The time of first appearance and the site of entry of Hoxb8-labeled microglia is consistent with this subpopulation being of hematopoietic, bone marrow-derived origin.
To assess if the Hoxb8 mutation affects the number of microglia present in the adult brain, comparable sections of the brain were evaluated for the presence of Iba1 positive cells in six Hoxb8 mutants and six control mice. We consistently observe an approximate 15% reduction of total number of microglia present in Hoxb8 mutant vs. control mice (Supplemental Figure 2). Currently, we cannot specifically label Hoxb8 mutant (i.e. Hoxb8−1−) microglial lineage because our Hoxb8-Cre driver is a “knockin” IRES-Cre driver which does not itself affect Hoxb8 function. Therefore, the above count for reduction of microglia in Hoxb8 mutant mice may represent an underestimate of the actual reduction of the hemotopoietic bone marrow derived microglia subpopulation, should such loss be partially compensated by an increase in the resident non-Hoxb8 expressing microglial subpopulation.
The presence of Hoxb8 lineage in the hematopoietic compartment was directly tested. Peripheral blood was collected and sorted by size and then by fluorescently labeled antibodies to cell surface antigens that distinguish the different types of white blood cells. All tested hematopoietic lineages, including platelets, granulocytes/monocytes, B cells and T cells, exhibited Hoxb8-YFP signal (Figures 2A-F). Observation of the YFP signal in both myeloid and lymphoid lineages is suggestive of Hoxb8 expression in stem cells or multipotent progenitor cells. To further support this hypothesis, bone marrow cells from Hoxb8-ICre/Rosa26 YFP mice were collected and examined by fluorescence microscopy. Most of the bone marrow cells were YFP positive (Figure 2G). In addition, the Hoxb8 -YFP+ signal was present in Sca1+C-kit+ cells, consistent with Hoxb8 expression in hematopoietic stem cells (Figure 2H). It has been reported that Hoxb8 is not expressed in mature hematopoietic cells (Kongsuwan et al., 1989; Petrini et al., 1992). The above results suggest that Hoxb8 is expressed at early stages of hematopoiesis but subsequently down regulated in fully differentiated blood cells.
We showed above that the only cells derived from Hoxb8 cell lineage detectable in the adult mouse brain are microglia, likely of bone marrow origin. To investigate whether a dysfunction in the hematopoietic system is responsible for the Hoxb8-excessive grooming and hair removal phenotypes, bone marrow transplantation experiments were conducted. Four different experimental groups were included. Bone marrow cells were collected from both wild-type or Hoxb8 mutant adult mice and transplanted into either irradiated Hoxb8 mutant or wild-type mice at two months of age, respectively. The development or regrowth of hairless patches, as well as grooming times, were monitored in all mice receiving transplants over a five month period. In the control group, in which irradiated wild-type mice received wild-type bone marrow cells, the grooming behavior remained normal, and no hairless patches were detected for the duration of the experiment (N = 10, data not shown). In the second control group, irradiated Hoxb8 mutant mice received Hoxb8 mutant bone marrow. These animals showed deteriorating health, and six out of eleven animals in this group died before the observation period was completed, presumably due to difficulty to re-establish bone marrow with a mutant erythropoietic lineage. The surviving recipients continued to exhibit very severe grooming and self mutilation phenotypes, without detectable hair regrowth during the observation period. Figure 3A shows one of the Hoxb8 mutants four weeks after transplantation with normal bone marrow. In the group of irradiated Hoxb8 mutant mice receiving normal, wild-type bone marrow cells, the hairless patches continued to develop for the first few weeks after transplantation. However, starting from three months after bone marrow transplantation, six of ten animals showed extensive regrowth of hair in the hairless areas and healing of open lesions. Four of them fully recovered and were indistinguishable from wild-type mice (Figures 3B-C). Grooming time was assessed on the Laboras platforms four months after bone marrow transplantation. The grooming times of these animals (N=6) decreased significantly to levels comparable to those of control mice (Figure 3D). Finally, among the irradiated wild-type mice that were transplanted with bone marrow collected from Hoxb8 mutant mice, two of ten animals developed hairless patches very similar to the phenotype observed in Hoxb8 mutant mice (Figure 3E). These mice (N=2) also showed increased grooming times relative to wild-type animals, but not increases as high as normally shown in Hoxb8 mutant mice. (Figure 3F)
The above experiments demonstrate that transplantation of normal bone marrow can efficiently rescue the Hoxb8 mutant grooming phenotypes, including restoration of the hairless patches, healing of open lesions, and reduction of excessive grooming times back to baseline. In separate experiments we have shown that CAG-GFP labeled bone marrow (Ikawa et al., 1995) transferred into wild-type irradiated mice can be detected in brain microglia four weeks after transplantation (supplemental Figure S3A) and increases by 12 weeks after transplantation (supplemental Figure S3B). The lower efficiency of conferring the Hoxb8 mutant phenotype to irradiated wild-type mice following transplantation of Hoxb8 mutant bone marrow, compared to efficient rescue of the mutant phenotype by transplantation of normal bone marrow into irradiated Hoxb8 mutant mice, appears to be a consequence of the lower robustness of Hoxb8 mutant bone marrow. Reconstitution of lethally irradiated wild-type mice with bone marrow containing 50% GFP labeled wild-type cells and 50% unlabeled Hoxb8 mutant cells showed that myeloid and T cells derived from Hoxb8 mutant bone marrows were at a measureable disadvantage relative to the normal GFP labeled wild-type cells blood lineages (supplemental Figure S3C). In contrast, B cells derived from Hoxb8 mutant bone marrow were at a competitive advantage relative to the wild-type cells.
The role of Hoxb8 in hematopoiesis has not been fully elucidated. In the literature, based on experiments involving Hoxb8 over expression, a case has been made for Hoxb8 involvement in the maintenance and differentiation of the myeloid progenitor pool (Kawasaki and Taira, 2004; Krishnaraju et al., 1997; Perkins and Cory, 1993). The very early expression of Hoxb8 during hematopoiesis and our bone marrow transplantation competition studies suggest a broader role for Hoxb8 in hematopoiesis, affecting both the myeloid and lymphoid lineages. Do T and B cells contribute to the compulsive grooming phenotype observed in Hoxb8 mutant mice? To test this possibility, bone marrow transplantations using bone marrow derived from RAG2 mice, which do not produce T and B cells, into irradiated Hoxb8 mutant mice were carried out. These experiments show that RAG2 bone marrow can rescue the excessive pathological grooming phenotype. However, the robustness of rescue is not as high as with wild-type bone marrow either in terms of the extent of hair replenishment of the hairless patches (supplemental Figures S3D and E) or in time spent grooming by these animals (supplemental Figure S3F). These experiments suggest that T and B cells do not have a primary role in the induction of the Hoxb8 mutant pathological grooming phenotype, but that T and B cell deficiencies may contribute to the severity of the phenotype.
Holstege et al. (2008) have reported that mice mutant for Hoxb8 show attenuated responses to noxious and thermal stimuli, and display a reduction and disorganization of interneurons in laminae I and II of the dorsal horn of the spinal cord, which receive the majority of the nociceptive and thermal sensations. We have observed very similar insensitivity to noxious and thermal stimuli, as well as dorsal horn spinal cord defects in our Hoxb8 mutant mice (Figure 4). Figures 4A and B show intact cervical spinal cord sections from wild-type and Hoxb8 mutant mice stained with a general neuronal marker anti-Neu N. It is apparent from these histological sections that the number of neuronal cell bodies present in the dorsal horn of Hoxb8 mutant mice is significantly decreased relative to those from wild-type mice. Immunohistochemical analysis of the spinal cord (shown at lumbar levels 4 and 5) further illustrate that in Hoxb8 mutant mice, relative to control mice, disorganization and reduction in numbers are apparent in both the input sensory fibers, labeled with CGRP, as well as interneurons in laminea I & II, labeled for calbindin and calretinin (Figures 4D-I). Nociceptive insensitivity was demonstrated by significantly greater latency time required by Hoxb8 mutant mice to respond to heat, relative to wild-type control mice (Figure 4C). Hostege et al. (2008) further suggested that the nociceptive/spinal cord defects account for the excessive grooming and hair removal phenotypes that we previously described.
A puzzling aspect of the hair removal phenotype described by Holstege et al. (2008) is that it appears quite different from the excessive grooming and hair removal phenotype that we observe in our Hoxb8 mutant mice. In their study, the hairless patches and skin lesions are very localized to the dorsal rump, (Holstege et al., 2008), and appear more consistent with the consequences of scratching a chronic itch. We observe a gradual progression of hair removal along most of the ventral surface of the mouse and extending to the lateral surfaces, which correlates with the consequences of an excessive normal grooming pattern (see for example Figure 3A). The hair is removed from the over-groomed areas in our mutant mice by the use of their teeth and accumulates in between their incisors, reflecting an extension of normal grooming behavior rather than scratching with their hind paws (Greer and Capecchi, 2002). What we observe and have reported in our Hoxb8 mutant mice is that the grooming syntax does not appear to be altered, but rather the number and duration of grooming bouts are increased. This aspect of the grooming phenotype has not been reported by Holstege et al. (2008). The marked difference in the pattern of hair removal and the excessive normal grooming observed in our mutant animals, rather than excessive scratching, suggests that the two groups might be studying different behavioral paradigms in their respective Hoxb8 mutant mice.
Scratching in rodents is a rather simple movement made by the hind limbs (Brash et al., 2005) that can be distinguished from grooming by the Laboras platforms. To determine if our Hoxb8 mutants exhibited excessive scratching, eight Hoxb8 mutants and eight wild-type mice were placed on the Laboras platforms set to score scratching. No significant differences in time spent scratching were detected between the Hoxb8 mutant and wild-type controls (supplemental Figure S4A). Only in mutant animals that had developed severe lesions as a consequence of compulsive, excessive grooming could increased scratching begin to be observed. For comparison we provide measurements of time spend grooming over the same duration of time.
Consistent with the localized scratching caused by a response to chronic itching, Holstege et al. (2008) reported that their behavior could be alleviated by localized injection of lidocaine. We tested whether our Hoxb8 mutant mice would respond to lidocaine treatment. Eight Hoxb8 mutant mice and eight control sibling mice were injected with lidocaine into regions where hairless patches had developed and placed on Laboras platforms to measure their grooming periods. Lidocaine treatment did not alter the grooming behavior of either the Hoxb8 mutant or wild-type mice (supplemental Figure S4B).
The first insight that the excessive grooming and hair removal behavior could be separated from the sensory spinal cord defects in our Hoxb8 mutant mice was gained from the transplantation experiments. Irradiated Hoxb8 mutant mice transplanted with normal bone marrow show normal grooming behavior, but retained their thermal stimuli insensitivity spinal cord defects (Figure 5). The latency of response to thermal stimuli in the bone marrow-rescued animals was comparable to that observed in Hoxb8 mutant mice and significantly longer than in wild-type mice. Thus, while normal bone marrow transplants into irradiated Hoxb8 mutant mice efficiently rescued the pathological grooming defect, it did not restore the defective pain response.
To further explore the causality of the Hoxb8 mutant phenotype, we utilized Tie2 Cre/loxP based conditional mutagenesis. Tie2 was originally considered an endothelial cell marker. However, Constien et al. (2001) showed that a Tie2Cre transgenic line displayed loxP mediated recombination in all hematopoietic cells, as well as in endothelial cells. In the brain, Tie2 lineage is present in blood vessels and microglia, but not in interneurons of the dorsal spinal cord (Figure 6A and data not shown).
In order to perform tissue-specific deletion of Hoxb8, we constructed a conditional allele in which the entire coding sequence of this gene was flanked by Lox511 sites (supplemental Figure S1C). Mice carrying the Hoxb8 conditional allele were crossed to Tie2Cre males to produce conditional mutants. Five conditional mutants carrying Tie2Cre and homozygous for the Hoxb8 conditional mutation were collected, and their grooming behavior analyzed. Four out of the five mice developed hairless patches very similar to the patterns observed in Hoxb8 mutant mice (Figure 6B). The grooming times, measured on the Laboras platforms, were significantly longer than control siblings and comparable to those observed in Hoxb8 mutant mice (N = 5, Figure 6C). However, these mice did not exhibit heat insensitivity (Figure 6D), and immunohistochemical analysis of their dorsal spinal cord laminae I and II labeled for calbindin and calretinin appeared normal when compared to Hoxb8 mutants (Figure 6E-J). Thus, restricted deletion of Hoxb8 in the hematopoietic system is sufficient to induce pathological grooming in mice, and importantly, the presence of the nociceptive defects are not observed or required for induction of the aberrant grooming behavior.
Next we wanted to determine if conditional deletion of Hoxb8 in the spinal cord recapitulates the nociceptive defects and whether or not they would be associated with the pathological grooming defects. Hoxc8 cell lineage in the spinal cord has a broad pattern of expression very similar to that of Hoxb8 (Figure 7A). However, relative to Hoxb8, Hoxc8 is poorly expressed in the hematopoietic system. Most importantly, the number of Hoxc8Cre/Rosa26 YFP labeled microglia in the brain is much lower (i.e. more than 10 fold lower) when compared to Hoxb8-labeled cell lineage (compare, for example, Figure 7B to Figure 1F, both derived from adult cortical brain sections). Quantitation of YFP+ Iba1+ cell counts relative to Iba1+ cells, which labels microglia, using the respective Hoxb8ICre and Hoxc8 ICre drivers is shown in Figure 7C. Since Hoxc8 is strongly expressed in the spinal cord, but poorly represented within microglia, the Hoxc8ICre mouse can be used to determine whether or not induction of the nociceptive defects are sufficient to initiate the pathological grooming defects. Ten such animals (i.e. Hoxc8ICre; Hoxb8c/c) were collected. None of these mutant animals developed the hair removal and skin lesions patterns typical of Hoxb8 mutant mice (Figure 7D). Also the time engaged in grooming by these mice was not significantly greater than in their control siblings (Figure 7E). However the latency of their response to heat was significantly longer than control siblings and was comparable to that of Hoxb8 mutant mice (Figure 7F). Further, examination of these mutants by immunohistochemistry with calbindin and calretinin antibodies showed that ablating Hoxb8 in the Hoxc8 expression domain fully recapitulated the dorsal horn spinal cord defects typical of Hoxb8 mutant mice (Figure 7G-L). Thus, conditional deletion of Hoxb8 in the Hoxc8 expression domain (i.e. the spinal cord) cleanly separated the nociception/dorsal spinal cord defects from the excessive pathological grooming defects. Selective induction of the Hoxb8 associated nociceptive/spinal cord defects (Hoxc8Cre) is not sufficient to induce the pathological grooming defects.
Herein we provide strong support for the hypothesis that the pathological grooming behavior observed in Hoxb8 mutant mice results from a deficiency in microglia. In support of this hypothesis we have shown that the only detectable Hoxb8 labeled cell lineage in the brain, (the source of the complex, innate behavioral grooming syntax), is microglia. Second, disruption of Hoxb8 function results in the reduction of the total number of microglia in adult mouse brains (i.e. a microglia phenotype). Further, the excessive pathological grooming behavior in Hoxb8 mutant mice can be rescued by transplantation with normal bone marrow. Finally, restricted deletion of Hoxb8 in the hematopoietic system (Tie2Cre) recapitulates the excessive pathological grooming behavior in these mice, while restricted disruption of Hoxb8 in the spinal cord (Hoxc8Cre) does not.
There appear to be two principle sources of microglia in the mouse, a resident population that is present in the brain early during embryogenesis prior to vascularization (Alliot et al., 1999), and a second population of bone marrow origin, derived from circulating monocytes, that migrate into the brain through the vascular system shortly after birth (Kaur et al., 2001; Ransohoff and Perry, 2009). The kinetics of infiltration of Hoxb8 labeled microglia into the brain is consistent with this population being the bone marrow-derived subpopulation. As such, the Hoxb8 lineage provides a useful molecular marker for distinguishing between these two microglial subpopulations. Do they have similar or different roles in the brain? Molecular markers allow genetic interrogation of the system. What would be the consequences of selectively ablating only one population? Interestingly, although Hoxb8 labeled microglia represent only 40% of the total microglial population present within the adult brain, selective inactivation of Hoxb8 in this subpopulation, is sufficient to induce the pathological grooming behavior. This fact would favor the hypothesis that the two microglial subpopulations present in the brain are performing distinguishable roles.
Microglia could affect neuronal activity and behavior by a number of mechanisms, including the secretion of cytokines that stimulate or inhibit neuronal activity, and work in parallel with neurotransmitters. Microglia have also been reported to function in regulating neuronal cell death during embryogenesis (Frade and Barde, 1998; Marin-Teva et al., 2004). Absence of appropriate cell death during neurogenesis could manifest itself later as aberrant behavior. Finally, the experiments of Wake et al. (2009) illustrating that microglia processes are very dynamic and engage in intimate contacts with synapses are particularly intriguing. They observed that the duration of contact at synapses is dependent on neuronal activity. From the above, it is becoming apparent that due to their mobility and dynamic contacts with synapses, microglia could represent an additional system for stabilizing and managing neural networks. By virtue of their high abundance in the cortex, including the frontal orbital regions and basal ganglia, the microglia of Hoxb8 lineage are positioned in close proximity to the pathways controlling repetitive behavior.
Obsessive compulsive disorder in human patients is associated with three principal brain regions: the prefrontal cortex, particularly the orbitofrontal cortex and anterior cingulated cortex; basal ganglia, including dorsal striatum and globus pallidus; and thalamus, namely the dorsal medial nucleus (Graybiel and Rauch, 2000; Huey et al., 2008) . Excessive grooming in rodents is widely believed to mimic the key traits of obsessive compulsive disorders. Although the syntactic grooming chain can be fully executed in rats decerebrated at various mesencephalic levels, indicating that neural circuits specifying the basic sequential structure are all present within the brainstem (Berridge, 1989), cortex and striatum play important roles in modulating the initiation and completion of grooming bouts (Berridge and Whishaw, 1992). Dopamine is a prominent neurotransmitter for implementation of the grooming pattern (Taylor et al. 2010). However, Welch, et al. (2007) have reported OCD-like behaviors in Sapap3 mutant mice that have reduced synaptic transmission in glutamanergic cortico-striatal circuits, and they further showed that the compulsive grooming in these mutants is alleviated with serotonin reuptake blockers. Thus, it is apparent that multiple brain regions and signaling pathways control the frequency of repetitive behaviors.
An alternative hypothesis has been put forward that the pathological grooming observed in Hoxb8 mutant mice is due to the sensory defects resulting from impaired formation of the spinal cord (Holstege et al., 2008). However, all of the experiments that we have presented contradict this hypothesis and have instead indentified a defect in the hematopoietic system, and more specifically a deficiency in microglia as likely causative for the aberrant grooming behavior. Our experiments have clearly separated the pathological grooming behavior from the sensory spinal cord defects to distinct cellular compartments. However, our experiments do not rule out the possibility that at later stages of the pathology the spinal sensory defects could exacerbate the consequences of excessive grooming.
Some of the apparent differences in the interpretation of Hoxb8 mutant phenotype by the Deschamps and our laboratories may result from monitoring different behavioral features. We monitor the time spent grooming (on Laboras platforms) whose excesses in Hoxb8 mutant mice leads to pathological behavior and very broad hair removal and skin lesion patterns. Grooming can be distinguished from scratching by the Laboras platforms. We did not observe increased scratching in our Hoxb8 mutant mice except modest increases at the very late stages of the pathology when lesions become apparent and the animals would normally be euthanized. The Deschamps laboratory reports very localized hair removal and skin lesions which appear more consistent with a response to a localized chronic itch. What accounts for the differences in phenotypic outcomes between the two Hoxb8 mutant mice? Differences in genetic backgrounds is not likely to be a strong contributor since both lines have been crossed to a predominantly C57Bl/6J background.
Notably, the Hoxb8 mutant alleles are different. This is a concern because the density of genes is very high within the Hox complex, and there are many non-coding RNA transcripts as well as protein encoding transcripts transcribed within this complex (Mainguy et al., 2007). The Deschamps allele is a LacZ knockin into the first exon of Hoxb8. Our allele was generated by introduction of a nonsense codon in the first exon and a loxP site into the second. Both inserts are small relative to the LacZ insert. We have shown that a knockin allele of neor into the Hoxb8 locus shows additional phenotypes due to perturbation of neighboring Hox gene expression, which disappear upon removal of the neor gene (Greer and Capecchi, 2002). Similarly, the LacZ gene could perturb neighboring RNA and/or protein expression, and thereby altering phenotypes. Consistent with this interpretation, in mice homozygous for a knockin allele of CreER™ inserted into the first exon of Hoxb8 (supplemental Figure S1E), we have observed the localized hair removal and skin lesion phenotype described by Holstege et al, 2008 in their Hoxb8 mutant (supplemental Figure S4D). This new phenotype shows up at a low penetrance (~10%) in addition to our characteristic Hoxb8 mutant grooming phenotype.
We have demonstrated that a deficiency in hematopoiesis of Hoxb8 mutant mice is causal to the pathological grooming deficit observed in these mutant animals. This deficit is correctable by normal bone marrow transplants. We have argued that the grooming malbehavior is primarily manifested by a deficit in microglia derived from bone marrow. However, a deficiency in T and/or B cells may also contribute to the severity of the behavioral pathology. Also, we have not ruled out other cellular members of the hematopoietic system as potential contributors to Hoxb8 pathological grooming.
Why couple behavior such as grooming to the host's immune system? From an evolutionary perspective it may make perfect sense to couple a behavior such as grooming, whose purpose is to reduce pathogen count with the cellular machinery, the innate and adaptive immune systems, used to eliminate pathogens.
In summary, we have provided strong support for the hypothesis that the excessive pathological grooming behavior exhibited by Hoxb8 mutant mice is caused by a defect in microglia. That a behavioral deficit could be corrected by bone marrow transplantation is indeed surprising. The therapeutic implications of our study on amelioration of neurological behavioral deficits in humans have not escaped us. This mouse model provides an opportunity to determine how impaired microglia results in generating such a distinct compulsive behavioral anomaly. Further, since the Hoxb8 lineage specifically marks the microglia subpopulation derived from bone marrow, this mouse can also be used to genetically interrogate this cell subpopulation relative to the resident microglial subpopulation.
The lines harboring the IRES-Cre knock-in cassettes in the Hoxb8 and Hoxc8 loci and the floxed, conditional Hoxb8 allele were generated here and are detailed in the Extended Experimental Procedures.
Grooming times and scratching behaviors were determined on Laboras platforms (Metris B.V., Netherlands). The animals were placed on the apparatus 6-8 hours before animal behavior recording started, their behavior was recorded for 24-hour periods and the data classified into several behavioral categories including eating, drinking, climbing, locomotion, grooming and immobility by the Laboras software. To determine the time spent scratching, a separate testing module was used, which specifically recognizes scratching behavior. Data was collected between 8PM and 12PM, when rodents are most active.
To measure responses to thermal stimulation, experimental animals were placed on a 53°C hot plate (Stoelting Corp.). The latency period required for the animals to respond, by licking their hind paws or by jumping was recorded. All animal experiments carried out in this study were reviewed and approved by the Institutional Animal Care and Use Committee of the University of Utah.
Bone marrow cells were harvested as previously described (Wang et al., 2006) and transplanted into lethally irradiated mice. Further details are provided in the Extended Experimental Procedures.
Peripheral blood samples were collected by retro-orbital bleeding with heparinized capillary tubes and processed as previously described (Spangrude et al., 2006). Peripheral blood cells were incubated for 20 min on ice with PE-Mac-1 and PE-Gr-1 for myeloid lineage analysis, Biotin-CD19/Avidin-APC/AF750 for B cell analysis, and APC-CD4 and APC-CD8 for T cell analysis. For bone marrow analysis, isolated immature bone marrow cells were incubated with AF647-c-Kit and PE-Sca-1 to identify early hematopoietic progenitors and stem cells. Prepared cells were analyzed with a BD FACScan flow cytometer (BD Biosciences, San Jose, CA). Additional reagents are listed in the Extended Experimental Procedures.
All statistical analysis was performed on raw data for each group by one-way analysis of variance, followed by a Tukey post hoc test, or Student's t-test. Differences among groups were considered significant if the probability of error was less than 0.05
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.