Phenotypic impacts of CSF1R deficiencies in humans and model organisms

Mϕ proliferation, differentiation, and survival are controlled by signals from the Mϕ CSF receptor (CSF1R). Mono‐allelic gain‐of‐function mutations in CSF1R in humans are associated with an autosomal‐dominant leukodystrophy and bi‐allelic loss‐of‐function mutations with recessive skeletal dysplasia, brain disorders, and developmental anomalies. Most of the phenotypes observed in these human disease states are also observed in mice and rats with loss‐of‐function mutations in Csf1r or in Csf1 encoding one of its two ligands. Studies in rodent models also highlight the importance of genetic background and likely epistatic interactions between Csf1r and other loci. The impacts of Csf1r mutations on the brain are usually attributed solely to direct impacts on microglial number and function. However, analysis of hypomorphic Csf1r mutants in mice and several other lines of evidence suggest that primary hydrocephalus and loss of the physiological functions of Mϕs in the periphery contribute to the development of brain pathology. In this review, we outline the evidence that CSF1R is expressed exclusively in mononuclear phagocytes and explore the mechanisms linking CSF1R mutations to pleiotropic impacts on postnatal growth and development.

(reviewed in ref. 22 and 23) and unified the diagnosis with a number of other disease entities with similar presentation (e.g., pigmented orthochromatic leukodystrophy or POLD). The brain pathology and symptoms of ALSP vary significantly between affected individuals and the disease has been misdiagnosed antemortem as various other dementias and neurodegenerative diseases. For example, Sassi et al. 24 identified 3 likely pathogenic CSF1R mutations in a cohort of 465 lateonset Alzheimer's patients.
A distinct recessive disease has more recently been associated with loss-of-function alleles at the CSF1R locus. 25,26 Patients lacking CSF1R in the recessive disease had almost complete loss of microglia, the M s of the brain, as well as defects in skeletal development and osteosclerosis. As far as we are aware, no patients have been described in whom there is homozygous loss-of-function mutation in either CSF1 or IL34. The contact amino acids involved in interactions in the CSF1-CSF1R and IL-34-CSF1R complexes have been dissected in crystal structures. 27,28 In the large human exome sequence collection ExAC (exac.broadinstitute.org), there are no non-synonymous mutations in CSF1 that potentially affect binding of CSF1 to the receptor, and only 2 rare variants in IL-34 (E111K and W116G) that could alter IL34 binding. However, there is likely to be expression variation between individuals. SNPs within the CSF1 locus have been associated with Paget's disease, an abnormality of bone resorption, 29 most likely associated with overexpression of CSF1 and excessive osteoclast activation. The level of circulating CSF1 in a very large cohort of coronary artery disease patients was correlated with distinct and relatively common cis-acting variants at the CSF1 locus and in turn with susceptibility to disease. 30 In this article, we will critically review the interpretation of studies of Csf1, Il34, and Csf1r mutations in experimental animals and their relevance to the human genetic diseases.

The cell-type and tissue specificity of Csf1r expression
A key piece of knowledge required to interpret the impact of Csf1r mutations is the site of transcript and protein expression. The main site of expression of Csf1r (aside from expression in placental trophoblasts) is undoubtedly in cells of the monocyte-M lineage. Reporter transgenes driven by the Csf1r promoter have been used to locate M s throughout embryonic development and in tissues of adult mice, [31][32][33] rats, 34 sheep, 35 and even chickens. 14,15 The same mouse Csf1r promoter driving Cre recombinase has also been used in lineage trace experiments in the embryo 36 without any evidence of expression outside myeloid lineages.
During mouse embryonic development, Csf1r mRNA is first detected in the ectoplacental cone early after implantation and in isolated M -like cells in the yolk sac. 37 Localization of Csf1r mRNA by whole mount in situ hybridization is consistent with restriction to myeloid cells in the embryo. 37 Although they are abundant and actively involved in clearance of apoptotic cells, 37 M s in the embryo are not required for organogenesis. Ablation of Csf1r-dependent M s by anti-CSF1R treatment of the mother has no effect on embryonic development 38 and Csf1r-deficient mice and rats are indistinguishable from their littermates at birth. 39,40 The restriction of Csf1r expression to cells of the M lineage is also supported by network analysis in both mouse and human systems. In mouse development, the appearance of Csf1r mRNA is strongly correlated with expression of other known M markers in a time course of embryo gene expression. 41 In the large promoter-based transcriptomic atlas produced by the FANTOM Consortium, there is a single cluster of M -specific transcription start sites in both mouse and human. 13 There is no detectable CSF1R expression in non-myeloid primary cells or cell lines of multiple lineages. Interestingly, CSF1R mRNA is also tightly correlated with a M signature in gene expression profiles from a wide range of human solid tumors 42 suggesting that it is exclusively expressed by tumor-associated M s and ectopic expression in tumor cells is not common. The molecular basis for Mspecific transcription of Csf1r and reporter gene expression has been reviewed elsewhere. 1,20 An inducible Fas-based suicide gene driven by the Csf1r promoter (the so-called MAFIA mouse) has been applied to functional studies of M s in vivo without evidence of ectopic expression or adverse impacts on other cell types. 43 In spite of this compelling evidence that Csf1r expression is restricted to cells of the monocyte/M lineage, the recent studies of the CSF1R homozygous mutation in human patients 25,26 and reviews of CSF1R roles in embryonic and postnatal development (e.g., ref. 44) cite a small number of studies that claim to demonstrate expression of Csf1r in non-hematopoietic cells including neuronal progenitors, intestinal and renal epithelial cells, and cells of the female reproductive system. If these reports are correct, then some of the pleiotropic impacts of CSF1R mutations in patients and experimental animals might be attributed to defects in non-myeloid cells. It is therefore timely to re-evaluate the evidence from these reports for non-myeloid expression of CSF1R.
There are several caveats to each of the studies claiming functional expression of CSF1R outside the mononuclear phagocyte lineage, notably in relation to the specificity of anti-CSF1R Abs. In the mouse brain, Sierra et al. 45 provided detailed evidence that the Csf1r-EGFP reporter gene produced by our laboratory 32 is restricted in its expression to microglia and perivascular M s at all stages of postnatal development and in injury and ageing models. Using additional Csf1r reporter genes in mice and distinct anti-CSF1R Abs, we 31-33 and others 46,47 have shown that expression of the transgene and CSF1R protein is restricted to microglia and M s at all stages of brain development. By contrast, Nandi et al. 48 reported that Csf1r mRNA was expressed in neuronal progenitors and used Ab staining to demonstrate high levels of expression of CSF1R protein in these cells in the early postnatal period. The specificity of their Ab binding was based upon lack of binding to Csf1r −/− mouse brains, but curiously, the Ab did not appear to detect microglia in wild-type mice. Their study would suggest quite high levels of Csf1r mRNA in the developing mouse brain. This is clearly not evident from analysis of deep RNAseq data, including the time course of embryonic development 41 in which expression is low and tightly correlated with increased expression of other known M -specific transcripts. In detailed network analysis of the transcriptomes of isolated cells from mouse and human brain, Csf1r is clearly part of a microglia-M signature. 49,50 Based upon conditional deletion of Csf1r with a Nestin-cre transgene and direct impacts of added CSF1 or IL-34 to "microglia-free" forebrain cultures (selected based upon Nestin-EGFP expression), Nandi et al. 48 proposed that CSF1 acts directly on neuronal stem cells. The interpretation of these studies depends on the view that the Nestin promoter is not active in microglia. However, Nestin mRNA is detectable in isolated microglia at levels higher than total brain (see data in ref. 13 and www.biogps.org). A subsequent study 51 also used Nestin-cre to provide evidence that following depletion with a CSF1R kinase inhibitor (see below) microglia repopulate from a Nestinexpressing progenitor. The transient expression of Nestin in microglia self-renewing following ablation was confirmed by another group using a different model. 52 The most widely cited study claiming neuronal expression of Csf1r 53 investigated the potential of CSF1R ligands as treatment for dementia and claimed that systemic administration of CSF1 or IL-34 could ameliorate neurotoxicity associated with excitotoxic injury. The authors were not able to detect specific staining for CSF1R protein in wild-type brain that was absent in Csf1r −/mice using any of 6 commercially available polyclonal anti-mouse CSF1R Abs. However, based upon in situ hybridization, they reported expression of Csf1r mRNA in scattered neurons in normal uninjured mouse brain, notably in the neuron-dense hippocampal pyramidal layers, and an apparent increase in labeled cells in this region following kainic acid-induced injury.
The underlying assumption was that there are no microglia among the neuronal cells. Using an amplified Csf1r-ECFP reporter gene, we detect positive cells in the same regions, but they clearly resemble microglia (Fig. 1A). Both this transgene and our Csf1r-EGFP and Csf1r-mApple transgenes are stringently restricted to CD45 + hematopoietic cells in digests of brain, and within that set, co-expressed with the myeloid marker CD11b. [31][32][33] The study by Luo et al. 53

used 2 other
Csf1r reporter gene systems; the MAFIA mouse (which has an EGFP reporter separated by an internal ribosomal re-entry site 43 ) and a conditional Csf1r-Cre-dependent EGFP reporter, to demonstrate apparent induction of Csf1r in injured neurons. They detected EGFP by Ab staining rather than direct imaging of EGFP fluorescence, so even if the signal reflects some level of inducible promoter activity, the signal is not quantitative. Finally, Luo et al. 53 generated a conditional knockout by crossing a floxed allele of Csf1r to CaMKIIA-cre to delete the Csf1r gene in neurons. These mice were reported to be more susceptible to kainic acid injury. But the controls in this case were mice that lacked the CaMKIIa-cre transgene. There is a need for caution in interpreting both this result and those obtained with Nestin-cre.
High-level expression of cre recombinase can clearly have impacts on cellular function. 54 It is entirely possible that expression of cre in neurons directly impacts their functions including the production of CSF1 and IL-34 and the sensitivity to toxic challenge. In overview, we consider there is compelling evidence for the exclusive expression of Csf1r mRNA and CSF1R protein in the brain in microglia and M s.
In the intestine, in Csf1 op/op and Csf1r −/− mice, Paneth cells were lost and there was disordered differentiation of epithelia including an excess of goblet cells. [55][56][57] In the mouse, Csf1r was apparently expressed functionally by Paneth cells and by other epithelial cells in both small and large intestine based upon staining with a commercial rabbit polyclonal Ab against CSF1R. [55][56][57] Conditional deletion of Csf1r with a tamoxifen-inducible Villin-cre reproduced intestinal epithelial disruptions associated with Csf1r mutation leading to the conclusion that Csf1r function is intrinsic to epithelial cells. 55 Consistent with the cautionary note above about cre recombinase, Bohin et al. 58 subsequently showed that tamoxifen-inducible Villin-cre activation per se led to DNA damage and cleavage of cryptic LoxP sites in intestinal stem cells. A secondary concern with the inducible cre recombinase system is that tamoxifen is not a neutral agonist, especially when applied to M biology (reviewed in ref.  Figure 1B shows the intimate association between crypt-associated M s and intestinal stem cells. This conclusion was supported by mRNA analysis of isolated intestinal cell populations. Furthermore, conditional deletion of Csf1r using a constitutive Villin-cre had no effect on epithelial differentiation. Importantly, Paneth cells were not actually depleted by anti-CSF1R; their expression of markers such as lysozyme and defensins was lost indicating that M s control their differentiation rather than their survival. 59 In the rat, we also saw no effect of the Csf1r null mutation on the presence of Paneth cells or on overall villus architecture. 40 We have not yet investi-  63 appeared to detect phosphorylated CSF1R in damaged renal epithelium using anti-CSF1R Ab. In the former study, protein detection was based upon the same rabbit antimouse CSF1R preparation used by Nandi et al. 48 The authors noted that Csf1r mRNA expression was considerably lower than in M s and contamination by M s was not excluded. Based upon conditional deletion using an Itgam-diptheria toxin receptor transgene, Menke et al. 60 argued that M s make a minor contribution to CSF1-dependent repair. Our own study in a similar renal ischemia model, which reproduced the beneficial effect of CSF1 treatment, strongly favors the M as the mediator of tissue repair and the exclusive responder to CSF1 administration. 61 This conclusion is supported by subsequent  63 Menke et al. 60 described an apparent increase in Csf1r-EGFP expression in epithelial cells in response to renal injury but in our view, the apparent increase in EGFP fluorescence was attributable to M infiltration of the damaged epithelial layers and autofluorescence of tubular casts. 61 In neither study was there any evidence of expression of the Csf1r-EGFP reporter genes in undamaged renal epithelium consistent with the original description of the transgene. 31,32 As in the brain, the increasing abundance of Csf1r mRNA in the kidney during embryonic and postnatal development correlated closely with other M markers. 41 In summary, the claim that CSF1 signals directly to renal tubular epithelial cells in any circumstance is not strongly supported. There are high-affinity mAbs against murine CSF1R that detect expression in isolated monocytes and progenitors (e.g., ref. 60 and 64) but they have not been used successfully to detect the protein in tissues. In fact, it is intrinsically unlikely that the CSF1R protein is detectable at high levels in tissues because it turns over constantly upon ligand binding (see below). As an alternative, we developed an AF647-conjugated version of CSF1. This protein bound specifically to monocytes isolated from the blood of mice and rats and when injected into mice localized specifically to tissue M s. 31,40,65 The restriction of CSF1R expression to M s and microglia is also supported by studies using Csf1r kinase inhibitors. Elmore et al. 51 reported the almost complete elimination of microglia from the mouse brain using selective CSF1R inhibitors. There was no evidence of a phenotypic impact and gene expression profiling revealed only the loss of known microglia-associated transcripts. Since this original report, the inhibitor PLX3397 has been used extensively in studies of the functions of microglia in brain and retinal development and homeostasis (e.g. ref. [66][67][68][69], and references therein) without any evidence of effects on non-myeloid cells. The lack of effect is actually surprising. Contrary to the way it is portrayed explicitly in many publications, PLX3397 is not a specific CSF1R kinase inhibitor; it is also an effective inhibitor of related kinases KIT and FLT3 [70][71][72] and likely mediates its effects on microglia in part by interacting with other kinase targets. Another orally available CSF1R kinase inhibitor, GW2580, also penetrates the brain but unlike PLX3397, it prevents microglial proliferation/selfrenewal without impacting on survival. 46 Taking all of these data together, we believe there is no reason to consider Csf1r expression outside of myeloid lineages in the interpretation of mutant phenotypes in experimental animals or humans.

CSF1R signal transduction
The binding of CSF1 or IL34 to CSF1R and the downstream signaling events have been reviewed in detail by Stanley and Chitu. 73 CSF1 signal transduction has mostly been studied in mature M s, osteoclasts, or cell lines, and mainly in mouse systems where CSF1 is not produced by M s themselves and endogenous/autocrine CSF1 signaling is therefore not an issue. For obvious reasons, the CSF1 response is also commonly studied in cells that have been deprived of growth factor to allow up-regulation of surface receptor and the analysis of a synchronous response to receptor ligation.
In broad outline, studies of CSF1 signaling have shown that ligand binding induces dimerization of the receptor and release of the kinase domain from an auto-inhibited conformation leading to initial tyrosine phosphorylation and ubiquitination of a membrane proximal domain.
Trans-phosphorylation of individual tyrosine residues in the intracellular domain then provides a scaffold for recruitment of several different effector pathways linked separately to survival, increased cell motility, proliferation, and specific gene regulation. One of these pathways is the classical SOS-GRB2-RAS-RAF-MAPK pathway that was first dissected in detail in the analysis of "sevenless/pointed" pathway in Drosophila. 74 The same pathway from CSF1R through SOS/GRB2, RAS, RAF, and the MAP kinases ERK1/ ERK2 leads in M s to phosphorylation of the transcription factor ETS2 on the pointed domain.
ETS2 then interacts with AP1 transcription factors on a conserved Ras response element to activate transcription of urokinase plasminogen activator (Plau) [75][76][77][78] in the same way that AP1 (Jun) interacts with pointed to induce photoreceptors in the Drosophila eye. 79 Following initial signal generation, the SOS-GRB2 complex dissociates from the receptor, there is further tyrosine and serine phosphorylation and a cascade of ubiquitination culminating in degradation of both ligand and receptor in lysosomes. 73 Internalization and degradation of the receptor is not blocked by inhibitors of receptor kinase activity. 80 81 and CSF1 must be present continuously in order for cells to enter S phase and subsequently undergo cell division. Because CSF1 is internalized and degraded, binding at 37 • C is irreversible and ligand is depleted from the medium. The degradation of CSF1 by proliferating mouse bone marrow-derived M s is saturated at concentrations that are required to drive entry into the S phase if the cell cycle. 9 As a consequence of the rapid degradation of the ligand, the dose-response curve for CSF1 action on any measured outcome in cell culture is very steep. 82 It is actually not a concentration dependence, but a titration of the amount of CSF1 available per cell per hour. If the cells exhaust the supply of growth factor at any time, the signaling cascade terminates. For example, the CSF1-dependent phosphorylation of the MAP kinases ERK1/ERK2 and of their target ETS2 in M s is sustained for as long as CSF1 is present. 75 The outcome of signaling also depends upon the cell population. Mature peritoneal M s are more effective at internalization and degradation of CSF1 than BMDM but do not undergo proliferation; they can compete in vitro for the available CSF1. 82,83 The M s of the liver clear CSF1 from the blood 84 thereby maintaining a low circulating concentration (∼20 ng/ml 39 ) that is less than saturating for M -mediated clearance by the receptor (∼70-100 ng/ml).
The circulating CSF1 concentration in vivo is also sub-stimulatory for monocyte production by the bone marrow and for proliferation and regulated gene expression in resident tissue M s. As a consequence, the entire mononuclear phagocyte system can respond to increased CSF1 availability. Administration of CSF1 85 or a CSF1-Fc fusion protein (which has a longer circulating half-life 7,31,86,87 ) leads to both expansion of the blood monocyte pool and proliferation of resident M s in all organs. It also induces expression by M s of Plau and other target genes with similar regulatory elements (e.g., Mmp9). A striking and unexpected consequence of CSF1-Fc treatment is extensive hepatocyte proliferation leading to a rapid expansion of the size of the liver in mice, rats, and pigs. 34,86,87 This finding indicates that M s contribute to the homeostatic regulation of liver size relative to body size. 86     Another group 107 studied a more limited backcross of the Csf1 op/op to this genetic background and found that the homozygotes could be maintained to adulthood with careful husbandry and feeding.
The idiosyncrasies of C57Bl/6 mice as a model for M biology have been reviewed elsewhere. 1 One relevant feature is that they have an intrinsically low bone density. Female C57BL/6 develop spontaneous osteoporosis at a relatively young age, and this can be blocked by anti-CSF1R treatment. 88 If anything, the ligand mutation in C57BL/6 mice is more severe than the receptor mutation. Curiously, heterozygous mutation of Csf1r abolished the pre-weaning mortality of the Csf1 op/op mutation. 39 There is a similar paradox in the case of  111 The impact of the Il34 mutation is consistent with the major sites of expression of Il34 in both mice and humans. 13 Within the mouse brain, Il34 and Csf1 have distinct and largely non-overlapping distributions across regions, and the knockouts of the 2 ligands accordingly show distinct impacts on regional microglial densities. 106 Studies in vitro do not support the idea that CSF1 and IL-34 have any differential signaling effects on microglia. 112 A second receptor for IL-34, PTPRZ, has been identified 113 but thus far no phenotype has been described in the mouse Il34 knockout that is incompatible with effects solely mediated by CSF1R. Recent studies have extended the analysis of differential Il34 dependency to distinct niches in the retina 114 and to M s of the kidney. 115 By contrast to mice, extensive profiling of human brain regions in the FANTOM5 project did not indicate a significant excess of IL34 over CSF1 mRNA, nor any region specificity. 13 However, the transcriptomic data do identify separate promoters/transcription start sites associated with expression of IL34 in skin and brain in both mouse and human.
As discussed above, some impacts of Csf1r mutations on the brain have also been attributed to direct actions on neuronal cells. 48 Erblich et al. 47

The rat as an alternative rodent model of Csf1r deficiency
There have been considerably fewer studies of CSF1R signaling biology in the rat. The toothless rat (Csf1 tl/tl ) has a frame-shift mutation in the Csf1 gene that ablates function. 118,119 Most studies of the Csf1 tl/tl rat have focused on the bone phenotype and the control of tooth eruption.
By contrast to the Csf1 op/op mouse, which retains some osteoclasts and recovers with age, the Csf1 tl/tl rat has an almost complete loss of osteoclasts, chondrodysplasia and unremitting osteopetrosis that was only partly overcome by postnatal CSF1 administration. [119][120][121][122][123][124] Like Csf1 op/op and Csf1r −/− mice, Csf1 tl/tl rats also exhibited severe postnatal growth retardation, which was associated with deficiencies in the growth-hormone/IGF1 axis. 123,125 We recently generated a Csf1r knockout rat by homologous recombination in embryonic stem cells. 40

F I G U R E 2
The effect of Csf1r mutation in rats on skeletal development. The X-ray image compares the skeletal development of an 11-week-old female Csf1r −/− rat with a control wild-type littermate. 40 Note in particular the increased calcification of the skull base (arrows), a phenotype shared with patients with homozygous CSF1R mutation. 25 There is also an overall reduction in body size and increased calcification of the long bones (especially the hind limbs) and the entire vertebral column, also a feature of the human syndrome 25 of several brain regions (hippocampus, olfactory bulb, striatum, pituitary) revealed the selective loss of many known microglia-associated transcripts. However, there was no effect on expression of genes associated with neuronal progenitors (e.g., Dcx, Cux1); thus providing no support for a non-redundant functional role for Csf1r in growth or survival of neuronal progenitor cells.
Since this original study, where the line was an early backcross to an inbred line, we have bred the rat Csf1r knockout fully to a fully inbred dark agouti and a fully outbred (Sprague-Dawley) genetic background.
On the pure inbred background, survival to weaning is somewhat more compromised and we see more severe ventricular enlargement in the brain and almost complete loss of the olfactory bulbs similar to the

Growth deficiency in CSF1R-deficient animals
Mutations of Csf1r in both mice 39 and rats 40 lead to reduced body weight. The growth retardation in Csf1 tl/tl rats was reportedly associated with an almost complete loss of circulating IGF1. 123,125 The Csf1r −/− rats were indistinguishable from litter mates at birth but in common with Csf1 tl/tl rats, their growth rate declined rapidly. 40 The impacts of Csf1r mutations have some obvious similarities to growth hormone (GH)/IGF1 mutations (reviewed in ref. 126). Like Csf1r −/− animals, GH-deficient (Gh lit/lit ) or GH receptor-deficient (Ghr −/− ) mice are born normal size and the growth defect manifests from around 2 weeks of age. Igf1 deficiency has a greater impact on embryonic growth than Gh or Ghr mutation but as is the case with Csf1r mutation, the perinatal lethality depends on genetic background. 127 Although the liver is the main source of IGF1 in the circulation, conditional deletion of Igf1 in hepatocytes did not cause a substantial reduction in postnatal growth. 126 Mouse M s grown in CSF1 also express very high levels of Igf1 mRNA initiated from a separate promoter from that used in the liver. 125 Chitu and Stanley 44  is generated in muscle as a pro-IGF1 form that requires processing. 132 Conditional deletion of muscle-specific Igf1 expression can also reduce circulating IGF1 and impair somatic growth 133 whereas M -expressed Igf1 appears essential for muscle regeneration following injury. 134 As well as contributing directly as a source of IGF1 production, M s are obvious candidates for a role in proteolytic processing of both pro-IGF1 and IGFBPs. Whereas the precise mechanism is unclear and probably complex, CSF1R mutation clearly impacts the GH-IGF1 axis and many of the pleiotropic consequences in rodents are probably linked to that impact.

A hypomorphic Csf1r mutation in mice
The   39,40 This is also the case in Csf1r ΔFIRE/+ mice. 135 In the expression profiles of the brains of heterozygous csf1r +/rats and Csf1r ΔFIRE/+ mice, there is a 50% reduction in Csf1r mRNA. Nevertheless, there is no significant change in any other transcript in response to the 50% loss of Csf1r. 40,135 The lack of dosage compensation is rather surprising since CSF1 can induce down-regulation Csf1r mRNA in M s. 143 One might have anticipated that reduced CSF1R signals would permit up-regulation of expression from the wild-type allele. A further puzzling finding is that despite the 50% reduction in Csf1r mRNA and CSF1R protein in heterozygous mutant mice no increase in circulating CSF1 was detected. 39 It seems that haploinsufficiency for Csf1r has little impact in mice or rats and is unlikely to explain the human dominant disease.
By contrast to the ALSP-associated mutations, the CSF1R mutations described in the recessive syndrome with skeletal symptoms 25,26 all appear to result in complete or partial loss of function or expres- binding either CSF1 or IL34 27,28 the proline is immediately adjacent to a conserved cysteine involved in the immunoglobulin fold, and is conserved in chickens. 142 Both this mutation, and a mutation in the tyrosine kinase domain (K627del) that was also present as a compound heterozygote in affected individuals, retained some biological activity when expressed in a reporter system. 25 Oosterhof et al. 26  One of the two recent reports of the recessive disease 26 promoted the zebrafish as an alternative model of Csf1r deficiency, in part because the mutation in inbred mice is apparently much more severe than the human disease. It is certainly the case that the function of Csf1r in the generation of M s and microglia is conserved in fish 16,17,152 as it is in birds that provide an alternative tractable model in which development can be monitored in ovo. 15 We have produced a Csf1r deletion in chickens and observed the same severe growth retardation (post-hatch) seen in mice and rats (DAH, A. Balic; unpublished).
One complexity of working with zebrafish, other than the quite distinct skeletal and hematopoietic biology, is that much of the genome is duplicated and there are two Csf1r loci with partially redundant functions.
The generation of an allelic series with graded loss of 1 to 4 copies of Csf1r indicated that microglial numbers are sensitive to Csf1r dosage and further, that as in mice and rats (see above) some peripheral M populations were less dependent upon Csf1r. 152 However, as noted above, 50% loss of Csf1r in the brain of rats and mice does not com- Adult-onset patients with homozygous CSF1R mutations have not been reported to exhibit the severe postnatal growth retardation seen in mice and rats. 25 We suggest that the impact of mutation on the GH/IGF1 axis requires the complex loss of CSF1-CSF1R activity and peripheral M populations and it may have contributed to the more severe cases with infant mortality. In any case, the absence of this phenotypic impact supports the argument above that osteopetrosis per se is insufficient explanation for growth retardation. Otherwise the differences from mutant animal phenotypes are not that great. It is clear from the mouse and rat mutations of Csf1r that microglia and osteoclasts are CSF1R-dependent. Other populations that share CSF1Rdependence with microglia, such as Langerhans cells, peritoneal M s, and heart and kidney M s, have not been studied in either ALSP or homozygous/compound heterozygous mutant patients. ALSP patients, who certainly do have residual CSF1R activity, do not exhibit the skeletal phenotypes described in the recessive disease but they do show ectopic calcification in the brain. 146 As discussed above, the osteo- but they are also relatively CSF1R-independent in animals. 40 There is clearly some impact of ALSP mutations in the periphery, since there is a defect in the CSF1-dependent generation of so-called non-classical monocytes in patients. 154 So, further studies of peripheral M populations in ALSP patients are needed in order to fully understand the disease process.

CONCLUSIONS
In overview, we suggest that mouse and rat mutations provide informative and predictive models of the pathology of human CSF1R deficiency provided account is taken of genetic background and the variable sensitivity of different M populations to CSF1R loss of function. The evidence for functional expression of CSF1R in non-hematopoietic cells is not compelling and accordingly all of the phenotypes associated with mutation of CSF1R or its ligands can be attributed to their impacts on mononuclear phagocyte biology. In keeping with that conclusion, all of the pleiotropic impacts of a Csf1r mutation in mice can be overcome by neonatal bone marrow transplantation. 155,156 The rat Csf1r −/− model with improved postnatal viability offers the opportunity to test therapies that might reverse the adverse phenotypes of human CSF1R mutations later in postnatal development or even in adults. Transplantation and other interventions proven efficacious in mouse and rat models are likely to provide insight into the human condition and may offer promise to patients with these rare diseases.