Previous analysis of the Csf1rko on the outbred background focussed on surviving adult female animals [13]. The postnatal mortality was variable and apparently more penetrant in males. The impact of the mutation on the inbred DA background remained distinct from the pre-weaning lethality reported in inbred Csf1rko mice [3,14,15]. There was some evidence of early postnatal mortality, especially in first litters, but the large majority of Csf1rko pups survived. The genotype frequencies at weaning from a heterozygous mating were WT 0.27, heterozygote 0.53, Csf1rko 0.19, based upon 65 litters. The majority of homozygous DA Csf1rko animals that survived to weaning were sacrificed because of breathing difficulties by 10 weeks.

To dissect the mechanisms underlying the Csf1rko phenotype, we first examined the effect of the Csf1rko on macrophage populations in the embryo. IBA1 staining labelled the abundant macrophage populations in the liver and throughout the WT embryos at around 10.5 days of gestation whereas staining was almost absent in Csf1rko embryos in the same litters (Fig 1A and 1B). A similar lack of IBA1⁺ cells was observed at later gestational age (12.5–13.5dpc; not shown). One of the major functions of macrophages in the embryo is the clearance of apoptotic cells [16]. To examine this further we performed TUNEL (terminal transferase-mediated dUTP-biotin nick end-labelling) staining to detect apoptotic cells. There was no evidence of accumulation of TUNEL⁺ staining except for occasional cells in the liver (Fig 1C and 1D). Locations where CSF1R⁺ macrophages are actively involved in phagocytosis of apoptotic cells include the interdigital region and the pharyngeal arches [17]. However, we saw no evidence of delay in clearance of dying cells (e.g. accumulation of pyknotic nuclei) in these regions despite the complete absence of IBA1⁺ cells in the Csf1rko embryos (Fig 1B).

Like the outbred animals, the inbred Csf1rko rats were indistinguishable from littermates at birth in either body weight or morphology. By 3 wks they were less than 50% of the weight of WT controls. Most organs, including the liver, were proportionally reduced but the absolute weight of the brain was unaffected (S1A and S1B Fig). The relative lack of effect of the Csf1rko on the adult brain in outbred rats, aside from lateral ventricular enlargement, was described previously [13]. The very limited impacts of the Csf1rko on postnatal development of multiple brain regions on the inbred background are reported elsewhere [18].

As reported previously, the relative growth advantage of males over females was also abolished (S1C Fig). By 7 wks of age, the shape of the brain was clearly different between WT and Csf1rko rats consistent with the radical difference in skull shape (S1D–S1F Fig). The reduced somatic growth was associated with significant reductions in cellular proliferation in major organs such as liver, lung and kidney as detected by localisation of Ki67 (Fig 2A–2F).

The skeletal phenotype of the outbred adult Csf1rko rat [13] closely resembles that of human CSF1R-deficient patients, referred to as dysosteosclerosis [8]. Dai et al. [19] reported disorganized matrix, reduced mineralization and osteoblast deficiency in the bones of 2–3 wk old Csf1rko mice. As shown in Fig 3A, analysis of the juvenile inbred Csf1rko rats revealed that a delay in skeletal calcification was already evident in newborns. By 3 wks of age secondary ossification centers of long bones and the small bones in the fore and hind paws were clearly deficient (Fig 3B and 3C). Unlike the hyper-calcified skull base [13], a shared feature with human bi-allelic CSF1R mutation [8,10,20], the cranial case of the Csf1rko rats remained hypomineralized even in adults and closure of the sutures was impaired (Fig 3D). The delay in postnatal musculoskeletal development was reflected in muscle. Sections through muscle at 7 days, 3 wks and 7 wks stained with laminin (Fig 3E and 3F) demonstrate that the relative cellularity was similar and the reduced muscle mass in the Csf1rko was primarily associated with a reduction in muscle fibre diameter (i.e. failure of hypertrophy).

There is a well-known homeostatic relationship between liver and body weight (reviewed in [21,22]). In the outbred adult Csf1rko rats analysed previously there was approximately 70% reduction in Kupffer cells (KC) detected in the liver, associated with selective loss of KC-enriched transcripts detected using microarrays [23]. To our knowledge there has been no previous analysis of the role of macrophages in postnatal liver development. In rodents, relative liver mass increases several fold in the postnatal period reaching the adult liver/body weight (LBW) ratio by around 4 wks of age [24,25]. The postnatal proliferative expansion of the liver is associated with profound changes in gene expression. The FANTOM5 consortium generated a dense developmental time course of RNA expression in mouse liver using cap analysis of gene expression (CAGE). Network analysis of these data revealed clusters of co-expressed genes with distinct temporal profiles including postnatal expansion of a macrophage-related cluster [26]. To assess the role of CSF1R in this hepatic differentiation/maturation process in the rat, we generated expression profiles of the inbred male and female Csf1rko and WT rats at 3 wks of age by RNA-seq. S1A Table contains the primary data and S1B Table shows 2760 differentially-expressed genes (DEG) distinguishing WT and Csf1rko livers (FDR <0.05). The DEG were grouped into a number of categories based upon known functions: (i) growth factors; (ii) Kupffer cell/macrophage-associated; (iii) cell cycle-related; (iv) lipid metabolism; (v) liver function/maturation-associated. Fig 4A summarises representative DEG in each category. The genes significantly reduced in the Csf1rko included Ghr, Igf1 and Igfals, with the latter most affected. Igf1 and Ghr closely paralleled each other in individual WT and Csf1rko samples (S1A Table). By contrast, the genes encoding the major IGF1 inhibitory binding proteins Igfbp1 and Igfbp2 (which in mice were massively induced at birth and declined rapidly thereafter; [26]) were highly-expressed and amongst the most over-expressed transcripts in the 3 wk-old Csf1rko liver relative to WT. Aside from Igf1, known direct GH targets (e.g. Socs2, Socs3, Cish; [27]) were unaffected suggesting that GH signalling was not impaired.

To identify gene co-expression networks, we analysed the data using the analysis and visualisation tool BioLayout (http://biolayout.org). The sample-to-sample network graph for all transcripts revealed a clear separation based upon genotype but a lack of separation based upon sex (Fig 4B). Clusters of co-expressed transcripts identified from a gene-to-gene correlation network are summarised in S1C Table and gene ontology (GO) terms for the largest clusters in S1D Table. Fig 4C shows the average profiles of the 4 co-regulated gene clusters that distinguished the WT and Csf1rko. The three clusters with lower average expression in the Csf1rko are consistent with the relative loss of Kupffer cells (Cluster 6), reduced proliferation and cell cycle (Cluster 5) and delayed liver-specific differentiation (Cluster 2). Cluster 3, which contains transcripts expressed more highly, but variably, in the Csf1rko compared to WT is enriched for transcripts associated with mitochondria and oxidative phosphorylation.

As observed previously in outbred juvenile WT rats [28] the livers of both WT and Csf1rko DA rats were mildly steatotic at 3 wks of age. Whereas this resolved with age in WT there was progressive and extensive steatosis in the liver of older Csf1rko rats (Fig 5A) which was not noted previously [13]. Because of the failure of tooth eruption, the Csf1rko rats were maintained on a modified diet (see Materials and Methods) but steatosis was not observed in the livers of WT rats maintained for 12 wks with access to the same diet (not shown). Mice with the ligand mutation (Csf1^op/op) have been reported to have a substantial deficit in insulin-producing cells in the pancreatic islets [29]. The selective loss of visceral adipose tissue [28], which was also evident in the inbred line, might suggest the reverse in the Csf1rko rat. We therefore measured circulating fed glucose and insulin levels and found small but significant reduction in fed insulin and glucose levels in the Csf1rko at 7 wks of age (Fig 5B and 5C).

Further analysis of juvenile rats indicated that the developmental delay in the inbred Csf1rko was not confined to the liver. One phenotype we did not note previously was almost complete involution of the thymus which was almost undetectable by 7 weeks of age (not shown). A subset of major organs is shown in S2 Fig. In the rat, the process of nephrogenesis continues in the immediate postnatal period [30]; the impact of the Csf1rko on the kidney was not previously analysed [13]. By 7 wks of age, we observed profound renal medullary hypoplasia, which was so severe that in most Csf1rko rats examined there was just a bud of papilla within the central pelvicalyceal space. The cortex was correspondingly hyperplastic, but the tubules and glomeruli appeared relatively normal (S2A and S2B Fig).

The intestines of outbred adult Csf1rko rats were not grossly abnormal and by contrast to Csf1^op/op mice, they were not Paneth cell deficient [13]. To dissect more subtle impacts of the Csf1rko in the inbred line we performed quantitative analysis of villus architecture in the ileum. There were significant reductions in the length and width of the crypts and villi, and the thickness of the submucosa in the Csf1rko (S2C–S2F Fig). Interestingly, given the reported role of CSF1R dependent macrophages in M cell differentiation in mice [31], we noted for the first time that macroscopic Peyer’s patches were almost undetectable in the Csf1rko rats at 3 wks and remained so at later ages.

The main welfare concern with the inbred Csf1rko is the development of progressive breathing difficulty. We considered the possibility that macrophage deficiency might also impact postnatal development of the lung. S2G Fig shows images of the inflated lungs of WT and Csf1rko rats at 3 wks stained with aldehyde-fuchsin to highlight elastin fibres and S2H Fig quantitates the airway space. There was no unequivocal evidence of impaired alveolisation. Flow cytometric analysis of disaggregated lung tissue revealed an increase in granulocytes in Csf1rko rats by 9 wks (S2I and S1J Figs), as seen in peripheral blood and BM, but there remained no histological evidence of inflammation that could explain the impaired respiratory function.

As in the outbred line [13], male and female inbred Csf1rko rats lacked development of primary and secondary reproductive organs. In males, the prostate and seminal vesicles and in females the uterus were so under-developed as to be almost undetectable (not shown). CSF1R-dependent macrophages have also been implicated in the process of branching morphogenesis in the mammary gland in mice [32]. This was not previously analysed in the rat. S3 Fig shows a comparison of mammary gland development in Csf1rko and WT female DA rats at around 9 wks. The negative impact of the mutation on ductal development was evident from staining with mature epithelial cell markers (KRT5, E-cadherin).

Aside from limited analysis using CD68 as a marker in adult liver and spleen, the effect of the Csf1rko on resident macrophages other than microglia was not examined previously [13]. To visualise resident macrophages we crossed the Csf1rko back to the Csf1r-mApple reporter transgene on the outbred SD background [33]. Fig 6 shows detection of the mApple transgene in a diverse set of tissues from WT and Csf1rko rats on this genetic background at weaning. In most tissues there was complete loss of Csf1r-mApple expressing cells aside from occasional monocyte-like cells and granulocytes in the vessels. This includes the abundant resident macrophage populations in smooth and skeletal muscle, kidney, pancreas, adipose, salivary and adrenal glands that were not recognised or analysed previously in mice. The pancreas contains numerous small lymph nodes that contain abundant Csf1r-mApple-positive cells. Whereas peri-acinar and islet macrophage populations were entirely lost in the Csf1rko, the lymph node-associated populations were partly retained and highlighted in whole mounts (Fig 6D). Partial reductions in resident populations were observed in intestinal mucosa, liver and lung (Fig 6C, 6H and 6J) Note that the resident Csf1r-mApple expressing cells in non-lymphoid tissues that are missing in the Csf11rko would include populations classified as dendritic cells, which are also CSF1R-dependent in mice [34].

To determine whether the major developmental abnormalities in the Csf1rko rat were autonomous to hematopoietic lineage cells and reversible we treated a cohort of 3 wk-old DA Csf1rko rats with WT congenic BM cells by IP injection without ablation of recipient BM. By 2–3 wks post BM transfer (BMT) the body weight gain of recipients diverged from Csf1rko controls and continued to increase thereafter (Fig 7A). The recipients did not fully recover the deficit in growth. At necropsy up to 6 months post BMT, the body weights were around 10% lower than WT littermates (Fig 7A) but visceral adipose was fully restored in both sexes and major organ histology was indistinguishable from controls.

The appearance of the animals changed rapidly, notably the skull and limb/foot morphology, so that by 6 wks post-transfer they resembled WT animals (Fig 7F–7H). The steatosis observed in untreated Csf1rko rats remained 4 wks post BMT but resolved by 9 wks (Fig 7I). Many BMT recipients developed teeth which required regular trimming. Both males and females became sexually mature and fertile and produced multiple litters. The females were able to suckle their offspring indicating that the failure of mammary development was reversible as confirmed in S3 Fig.

The Csf1^tl/tl rat has been reported to have a deficiency in circulating IGF1 [35] and the transcriptomic analysis of the liver indicated dysregulation of the GHR/IGF1/IGFBP system. Accordingly, we measured IGF1 and GH in the circulation of the Csf1rko rats. In the WT DA rats, there was a postweaning surge in IGF1 followed by a gradual decline (Fig 7B) whereas IGF1 was barely detectable in the serum of Csf1rko DA rats at any age. By contrast, the circulating concentration of GH was unaffected (Fig 7C). The DA rat is relatively small and slow-growing compared to outbred lines such as Sprague-Dawley (SD). To test the impact of genetic background, we crossed the mutation back out to SD for two generations. Despite the more rapid postnatal growth of SD and 2–3 fold higher adult body weight compared to DA (S4A Fig) both the time course and magnitude of circulating IGF1 was very similar between the WT inbred and outbred lines (Fig 4B). Unlike the almost complete loss of IGF1 seen in the DA Csf1rko line, the surge of circulating IGF1 in the postnatal period was readily detected, albeit reduced and delayed, in the SD Csf1rko rats (S4B Fig). Nevertheless, the SD Csf1rko rats showed the same growth arrest around 7 weeks of age as the DA albeit at peak body weight of 80-100g rather than 50-70g (S4A Fig). In the DA Csf1rko rats that received WT BM, IGF1 was only partly restored. It was first detectable by 4 wks post transfer (week 7) and peaked at 6 wks post BMT (week 9) (Fig 7B). However, the divergence of body weight gain in the BMT recipients compared to untreated Csf1rko rats was evident within 2 weeks before IGF1 was detectable and peak levels did not recapitulate the postnatal surge.

The Csf1r-mApple transgene is not currently available on an inbred background to enable analysis of tissue macrophage recovery following BMT. For this purpose, as in the embryo (Fig 1), we used IBA1 as a marker. Although most commonly used as a microglial marker, IBA1 is widely expressed by tissue macrophages in mice [36]. Immunohistochemical localisation of IBA1 in selected organs from the inbred Csf1rko and litter mate control rats at 3 and 7 wks is shown in Figs 8 and 9. In most WT tissues, IBA1 immunoreactivity was restricted to abundant interstitial stellate cells resembling macrophages. The morphology, abundance and location of IBA1⁺ populations was comparable to the Csf1r-mApple transgene (Fig 6) and much more extensive than seen with anti-CD68 (ED1) [13]. The Csf1rko led to substantial or complete loss of macrophage-like IBA1⁺ cells in all organs examined and there was no evidence of recovery with age.

The majority of IBA1⁺ tissue macrophage populations were restored by BMT. Resident peritoneal macrophages which were absent in the Csf1rko were restored to the level of WT controls. Fig 10A shows IBA1 staining of multiple organs of adult (22 wks) rats that had received WT bone marrow at weaning indicating the restoration of tissue IBA1⁺ cell populations to levels similar to littermate controls. Note in particular that the visceral adipose, which was restored following BMT, was populated with IBA1⁺ macrophages. BMT prevented premature thymic involution and the thymus of BMT recipients contained abundant IBA1⁺ cells. IBA1 is also expressed during sperm maturation, and IBA staining highlights the loss of mature sperm in testis (Fig 9) and their restoration following BMT (Fig 10). The repopulation of the liver with macrophages of donor origin was confirmed by analysis of the restoration of the expression of Csf1r (Fig 7D) and Adgre1 mRNA (Fig 7E). The brains of Csf1rko rats lack microglia and brain-associated macrophages detected with anti-IBA1, alongside the loss of Aif1 mRNA encoding this marker and many other microglia-associated transcripts [13,18]. Fig 10B compares IBA1 staining in the cortex of adult WT, Csf1rko and BMT rats at 16 weeks post-BMT. The images are representative of all brain regions in all recipients examined. The IBA1⁺ cells in the BMT recipients adopted the same regular spacing and similar density to IBA1⁺ microglia detected in age-matched litter-mates. However, the cell morphology of the BMT-derived cells throughout the brain was quite distinct, more stellate and less ramified than typical microglia. The ventricular enlargement observed in juvenile Csf1rko rats was not reversed by BMT, but also had not progressed. The mechanisms underlying hydrocephalus in both mouse and rat Csf1rko and bi-allelic human CSF1R mutations are unknown [6] but likely unrelated to microglial deficiency [37]. Otherwise, we could detect no difference between BMT recipient and WT littermate brains.

The concentration of CSF1 in the circulation of WT animals is relatively low due to receptor-mediated clearance mainly by the macrophages of the liver and spleen [38]. Accordingly, increased circulating CSF1 was detected in the serum of adult Csf1rko rats by Western blot [13]. We reasoned that injected BM cells likely responded to the elevated CSF1 and successful restoration of CSF1R-expressing tissue macrophage populations could be monitored by measuring circulating CSF1 by ELISA. Indeed, CSF1 was massively elevated in juvenile Csf1rko rats at weaning and declined rapidly following BMT (S5A Fig). Given the granulocyte accumulation seen in the Csf1rko we also assayed CSF3 (granulocyte colony-stimulating factor) in the same samples, and there was no significant impact of the Csf1rko (S5B Fig).

In overview, Csf1rko rats are deficient in resident macrophages in most organs throughout postnatal development and rescue by BMT is associated with restoration of these populations to WT density.

Csf1rko mice were reported to have enlarged spleens and evidence of extramedullary hematopoiesis [39]. This is not the case in Csf1rko rats. Given the apparent lack of macrophages within hematopoietic islands in the fetal liver (Fig 1) we were especially interested in resident BM macrophages, which are believed to be an essential component of the hematopoietic niche [40]. As previously reported in outbred animals [13], the inbred Csf1rko rats were entirely deficient in osteoclasts expressing tartrate-resistant acid phosphatase (TRAP) (not shown). However, IBA1⁺ positive island macrophages were detectable in the residual BM cavity of Csf1rko rats at 7 weeks with similar stellate morphology to WT (Fig 11A).

Like adult outbred Csf1rko rats [13], juvenile (3 wks) and adult (9 wks) inbred Csf1rko rats exhibited a 3–5 fold increase in circulating granulocytes (Fig 11B and 11C). The Csf1rko on the outbred background was reported to be around 70% monocyte-deficient [13]. Monocyte sub-populations were not analysed. In the rat, monocyte sub-populations are distinguished by reciprocal expression of the markers CD43 and HIS48, with HIS48^Hi and CD43^Hi monocytes corresponding to so-called classical and non-classical monocytes, respectively [33]. Consistent with the role of CSF1R signals in their differentiation, the CD43^Hi monocytes were selectively lost in juvenile and adult Csf1rko rats. However, there was no corresponding increase in the classical HIS48^Hi monocytes (Fig 11D). In keeping with previous analysis, there was also a small but significant reduction in circulating B cells whereas T cells were unaffected (Fig 11E).

Because of the osteopetrosis, BM cells could only be obtained from Csf1rko by crushing the femurs. There was a ~2-fold reduction in CD45⁺ leukocytes (as a proportion of live cells) recovered from Csf1rko BM and a relative increase in granulocytes (Fig 11F and 11G). After excluding granulocytes (HIS48^Hi/SSC^Hi), three monocyte/macrophage populations could be distinguished: two putative monocyte populations paralleling peripheral blood HIS48^Hi and CD43^Hi monocyte profiles, and a CD172A^Low/CD4^int BM resident macrophage population. The classical (HIS48^Hi) monocytes were more abundant in WT BM than non-classical (CD43^Hi) monocytes (the reverse of peripheral blood) and their relative abundance was unaffected by the Csf1rko. The HIS48^-ve/CD43^lo/CD172A^lo/CD4^int resident BM macrophages (Fig 11H) as well as B cells (Fig 11I), were selectively depleted in Csf1rko BM.

Surprisingly, the BMT did not restore the CD43^Hi blood monocyte population in the rescued Csf1rko rats whereas the granulocytosis and B cell deficiency in peripheral blood was completely resolved (Fig 12A–12F). The BM cavity was patent and the populations of IBA1⁺ hematopoietic island macrophages and TRAP⁺ osteoclasts were indistinguishable from WT litter mates (Fig 12G). Accordingly, the yield of CD45⁺ leukocytes from BM obtained by flushing was similar to WT. The CD172^Lo/CD4⁺/HIS48^-/CD43^Lo resident macrophage population was partially restored and granulocyte and B cell numbers were normalised as in the blood (Fig 12H–12J). In standard liquid cultures used routinely to generate BM-derived macrophages in our laboratory [41] WT control BM cells produced a confluent macrophage culture (Fig 12K) By contrast, cultures of cells harvested from Csf1rko BMT recipients contained only small numbers of large adherent macrophages and there was no increase in cellularity with time.

Our results demonstrate that CSF1R-dependent macrophages are essential for early postnatal growth and organ development in the rat. The exclusive impact on postnatal growth distinguishes the Csf1rko from Igf1 and Igf2 mutations which impact the growth of the embryo [42]. Interestingly, although Csf1r is highly-expressed in placenta, the lack of impact of the mutation on embryonic growth also indicates that placental function is CSF1R-independent. It is likely that any loss of macrophage-derived trophic factors in the embryo is mitigated to some extent by placental and maternal-derived growth factors. Severe postnatal growth retardation is not evident in human patients with bi-allelic CSF1R mutations [8,10,11,20]. This may be indirect evidence that the mutant alleles in these individuals are hypomorphic, rather than complete loss-of-function. The only definitive human homozygous CSF1R null mutation described thus far was associated with severe osteopetrosis, brain developmental defects and infant mortality [43].

The somatomedin hypothesis proposed that somatic growth is controlled by pituitary GH acting on the liver to control the production of IGF1. Numerous analyses of conditional mutations of Ghr, Igf1 and Igf1r in various tissues and cell types in mice paint a more complex picture [27,44,45]. Notably, conditional deletions of Igf1 and Igf1r in chondrocyte and osteoblast lineage cells produce substantial reductions in overall somatic growth rates (reviewed in [46]). Hence, as suggested by Chitu & Stanley [3], the defects in skeletal growth and maturation that we showed were already evident in the Csf1rko at birth (Fig 1) are likely one underlying cause of reduced postnatal somatic growth rate.

By contrast, conditional deletion of Igf1 in the mouse liver had a marginal effect on somatic growth despite 70–90% loss of circulating IGF1 [44,45]. Like the GH-deficient dwarf rat [47] and GHR-deficient mice [48] the Csf1rko rats have greatly-reduced circulating IGF1. The loss of IGF1 was also present, albeit less severe, on the outbred background. However, the Csf1rko rats are not GH-deficient, consistent with unchanged levels of GH mRNA in the pituitary [13,18], and the phenotypes of GH and Csf1rko mutations are quite distinct. Dwarf rats have a slower overall growth rate but do not exhibit the early growth arrest and major organ phenotypes and morbidity seen in the Csf1rko irrespective of genetic background. Indeed, GH/GHR deficiency in rats and mice is associated with increased longevity [49]. Macrophages themselves produce an array of growth factors, including IGF1, that are implicated in development and tissue repair [7,50]. Meta-analysis of mouse resident tissue macrophage gene expression [34] revealed constitutive high expression of Igf1 and Igfbp4 mRNA but macrophages are clearly not the only extrahepatic source. Rat macrophages grown in CSF1 also express abundant Igf1 mRNA [41] and given their relative abundance, could be a significant source of IGF1 in tissues. However, conditional deletion of Igf1 in myeloid cells in mice had no reported effect on somatic growth [51]. Furthermore, depletion of microglia in the brain of the Csf1rko rat had no effect on Igf1 mRNA [13,18]. It is notable that the phenotypes we have described in the Csf1rko rat are considerably more severe than those reported in Csf1^tl/tl rats, which are also deficient in circulating IGF1 [35]. As described originally [52] and confirmed on an inbred background [53], Csf1^tl/tl rats achieve body weights of 250-300g, are male fertile and have normal longevity. This suggests that in rats the alternative ligand, IL34, has significant non-redundant roles in macrophage homeostasis and function in multiple organs during postnatal development. Accordingly, we suggest that CSF1R-dependent macrophages act indirectly to regulate circulating IGF1 mainly through their effects on hepatocyte proliferation and maturation but this is not linked directly to the impacts on somatic growth.

The set of hepatic genes correlated with Csf1r in network analysis (Fig 4A, S1C Table) includes Cd68 and Aif1, encoding CD68 and IBA1 respectively, and confirms a rat Kupffer cell (KC) signature [23] that includes marker genes enriched in this resident population in mice (e.g. Cd5l, Clec4f, Timd4, Vsig4) [54,55] as well as highly-expressed genes involved in iron metabolism/erythrophagocytosis (Cd163, Slc40a1, Timd2). The liver has a unique ability to regenerate following partial hepatectomy to return to a constant liver-body weight ratio, a so-called hepatostat [21]. Administration of CSF1 to adult mice, rats or pigs [33,56–58] overcomes this constraint, drives hepatocyte proliferation and accelerates regeneration following partial hepatectomy.

During the postnatal period in rat liver a phase of hyperplasia is followed by hypertrophy and structural maturation to form mature hepatic sinusoids lined by a single layer of hepatocytes [24]. In addition to impacts of the Csf1rko on the IGF1 axis in the liver, transcriptional profiling indicated a broader delay in this postnatal hepatocyte maturation exemplified by the retention of fetal liver-expressed genes such as lipoprotein lipase (Lpl)[59], fetal liver hepcidin (Hamp) [60] and the fetal amino acid transporter Slc38a2 [61]. Two other genes highly enriched in the liver of Csf1rko rats were the cold shock-inducible gene Rbm3 and stress-inducible Gadd45b. Expression of Rbm3 and Gadd45b and their potential impacts on liver metabolism [62,63] could be a consequence and/or a cause of the reduced adiposity and hepatic steatosis of the Csf1rko. Conversely, the relative loss of several highly expressed liver-specific genes is likely to exert additional pleiotropic effects on other organs. Amongst proteins encoded by the most CSF1R-dependent transcripts, SERPINA6 (transcortin) is the major binding protein for circulating glucocorticoids implicated in regulation of the response to GH [64]. In mice, Serpina6 is highly-expressed in fetal liver, declines to almost undetectable level at birth, and is then re-induced in parallel with Ghr and Igf1 [26].

The liver of 7-day old rats contains abundant lipid droplets [28] but this usually resolves rapidly. The coordinated reduction of numerous genes involved in lipid metabolism (Fig 4A) in the liver in 3 wk old Csf1rko rats is the reciprocal of the increase observed following CSF1 treatment of neonates [28] and likely contributes to both the progressive steatosis and the lack of visceral adipose in the animals. Both of these phenotypes were reversed by the transfer of WT BM cells (Figs 7 and 10). The coordinated regulation of genes involved in lipid metabolism may be related to the reduced expression of the transcriptional repressor Hes6 [65] in the Csf1rko.

The precise mechanisms underlying reduced proliferation in the postnatal Csf1rko liver are not evident from the transcriptome analysis, which did not reveal the loss of known growth regulators or indirect evidence of deficient receptor signalling. Expression of potential candidates is highlighted separately in S1E Table. It is also notable that the Csf1rko has no impact on expression of markers of other non-parenchymal cells, endothelial cells (Pecam1, Cdh5) or hepatic stellate cells (Pdgfrb). Postnatal growth of the liver involves β-catenin signalling linked to E-cadherin and the receptor for hepatocyte growth factor (MET). Conditional deletion of Ctnnb1 in hepatocytes of mice led to a 15–25% decrease in LBW ratio at postnatal day 30 [66]. Conditional deletion of Yap1, a downstream component of the hippo kinase pathway, also led to reduced hepatocyte proliferation and reduced LBW [25]. By contrast, the grossly-reduced hepatocyte proliferation seen in the juvenile Csf1rko rats is not liver-specific and does not lead to reduced LBW ratio. The regulation of hepatocyte proliferation has been studied mainly in the context of regeneration and involves a complex array of growth factors; both locally-derived and present in the circulation (reviewed in [22]). As noted in the introduction, CSF1 treatment of newborn rats can promote the selective growth of the liver [28]. However, the intrinsic hepatostat remains functional in the Csf1rko rat and the reduced hepatocyte proliferation may partly be a consequence of deficient somatic growth that is also CSF1R-dependent.

Although there are similarities in some phenotypes, the osteosclerotic bone phenotype in the Csf1rko is quite distinct from osteopenia associated with GH or IGF1 deficiency [67]. The novel phenotype we identified in the cranial case relates to the process of intramembranous ossification. Intramembranous bones ossify directly from preosteogenic condensations of multipotent mesenchymal stem cells without a chondrocyte intermediate. Like endochondral ossification, this process is dependent upon angiogenesis (reviewed in [68]). The process of bone formation in the cranial case is initiated by postnatal expansion of bone from the sutures formed during embryonic development [69]. The ossification defect reported in Csf1rko mice was attributed to the lack of CSF1R-dependent osteoclasts [19]. However, osteoblast maintenance and calcification requires input from osteomacs, a bone-associated macrophage population distinct from osteoclasts [70]. These CSF1-responsive macrophages promote the processes of both endochondral and intramembranous ossification during bone repair in vivo [71,72]. Analysis of the Csf1^tl/tl rat revealed a defect in osteoblast attachment to bone surfaces and the absence of prominent stress fibres [73]. Hence, the selective failure of intramembranous ossification in the cranial case, and the increased calcification of the base of the skull, may reflect distinct functions of osteomacs and osteoclasts, both dependent on CSF1R.

The Csf1rko rat resembles the effect of bi-allelic human CSF1R mutation in the pronounced under-modelling of the digits [8]. Detailed analysis of digital development in the mouse [74] revealed morphologic and calcification patterns in the subarticular regions that were also distinct from archetypal physeal endochondral ossification. The defects we observe in the Csf1rko rats resemble skeletal dysplasia caused by mutations in the gene encoding the macrophage-enriched transcription factor MAFB [74] and more generally various forms of malignant osteopetrosis (reviewed in [75]).

The combined analysis using the Csf1r-mApple reporter (Fig 6) and localization of IBA1 (Figs 8 and 9) highlights both the abundance of resident tissue mononuclear phagocytes in WT rats and the almost complete depletion of these cells in the Csf1rko regardless of genetic background. The loss of monocytes in Csf1rko rats [13] was shown here to be selective for the CD43^Hi ‘non-classical’ monocyte population which is the major population in the rat [33]. Analysis of a hypomorphic Csf1r mutation [37] and the selective impact of anti-CSF1R antibody [76] indicates that CSF1R is not required for monocyte production, whereas tissue macrophages require CSF1R signals for survival. The lack of accumulation of classical (His48^Hi) monocytes in the Csf1rko despite the block to maturation could reflect more rapid transit of these cells in the circulation. For example, the striking periportal concentration of the residual IBA1⁺ Kupffer cells in the liver and reduced numbers in the gut lamina propria could be associated with continuous extravasation of monocytes and subsequent short half-life in the tissue.

In many different organ systems, depletion of resident tissue macrophages permits monocyte extravasation to occupy the vacant territory or niche (reviewed in [1,2]). It is rather striking that the rescue of tissue macrophage populations following BMT (Fig 10) restores them to almost precisely WT levels. Guilliams et al. [1] favor the idea that each macrophage occupies a defined niche responding mainly to local CSF1. An alternative view is that the macrophages are territorial and the regular distribution in every tissue (confirmed in Figs 8–10) is determined by mutual repulsion [2]. That view favors a role for systemic CSF1 stimulation which may vary between tissues. In mice, tissue-specific postnatal macrophage defects associated with the Csf1^op/op mutation are mitigated by trans-placental transfer of maternal CSF1 [77] and were partly compensated by administration of CSF1 in the postnatal period [78]. Based upon the rapid reduction in circulating CSF1 in the BMT recipients (S5A Fig) and the ability of an exogenous CSF1-Fc fusion protein to expand macrophage populations in all tissues in the neonatal and adult rat [28,33] we suggest that the entire adult mononuclear phagocyte system is regulated in a coordinated manner by CSF1 availability. Following BMT, macrophage repopulation occurs until balance is restored between CSF1 production and CSF1R-mediated clearance/utilisation.

Analysis of the role of macrophages in postnatal development based upon Csf1r and Csf1 mutations in mice has focussed on one organ at a time, emphasising local functions [3]. The loss of circulating IGF1 is just one example of the systemic consequences of the Csf1rko. We clearly cannot separate direct local roles of macrophages in the regulation of proliferation and differentiation in liver, brain, skeleton, muscle, lung, kidney, gut, thymus, adipose and gonads from systemic consequences of developmental abnormalities in every other organ. For example, although growth hormone (Gh) mRNA was not reduced in the pituitary of Csf1rko rats, RNA-seq profiling revealed a significant reduction in cell-cycle associated transcripts and relative reductions in Fshb and Lh in males and prolactin (Prl) in both sexes [18].

Given the large numbers of macrophages in the developing embryo and their roles in clearance of apoptotic cells [16] it is surprising that there is so little apparent impact of their almost complete absence in the Csf1rko on prenatal development. “Amateur” phagocytes can evidently replace the phagocytic functions of macrophages in embryonic development. In keeping with this view, although genes encoding lysosomal enzymes and endosome-associated proteins are highly-expressed by microglia, the relative expression of these genes was unaffected in brains of Csf1rko rats or microglia-deficient mice [13,37].

Bennett et al [14] reported that pre-weaning lethality could be prevented in around 50% of inbred Csf1rko mice by neonatal transfer of WT BM cells. Their study focussed on reconstitution of microglia and did not examine the contribution of the donor cells to the hematopoietic compartment. Importantly, they showed that rescue was independent of the monocyte chemotactic receptor, CCR2. Like these authors, we were able to fully repopulate a myeloid population in the brain following BMT in a Csf1rko recipient (Fig 10B). The less-ramified stellate morphology we observed was not reported directly in the mouse study, but is evident in the published images [14]. Ongoing studies address the question of whether these BM-derived “microglia” reconstitute expression of the microglial gene expression signature that is lost in the Csf1rko brain [18].

In principle, neonatal transfer in the published study [14] could allow WT hematopoietic stem cells (HSC) to populate the developing bone marrow. In the rat, we were actually able to reverse and rescue the Csf1rko phenotype and resident tissue macrophage populations, including those of the brain, with 100% success by transfer at weaning when the developmental defects including severe osteopetrosis were already evident. The recipients were not treated to deplete HSC and were not deficient in other blood cell lineages. The reversal of osteosclerosis and expansion of the hematopoietic niche in the long bones likely requires the generation of CSF1R-dependent osteoclasts and resident osteomacs [40,79] that were replenished in the BMT recipients (Fig 12). Similarly, CSF1 treatment restored TRAP⁺ osteoclast populations and active resorption in the bone of Csf1^tl/tl rats [80]. The granulocytosis and B cell deficits observed in marrow and blood of the Csf1rko rats were reversed by BMT despite the lack of contribution of donor cells to the BM compartment. We eliminated any increase in the key growth factor, CSF3, as a mechanism underlying granulocytosis (S5 Fig). Neutrophils have a short half-life in the circulation [81]. In part, the granulocytosis in the Csf1rko rat probably reflects the loss of Kupffer cells and selected splenic macrophages which clear senescent neutrophils via phosphatidylserine-mediated phagocytosis [82]. The Csf1r-mApple reporter transgene is expressed by neutrophils, but in most tissues of the Csf1rko rat there was no evidence of any residual positive myeloid cells (Fig 6) suggesting that the loss of macrophages also compromised neutrophil extravasation. Neutrophils migrate constitutively into healthy tissues and their homeostasis depends upon the CXCR4/CXCL12 axis [81]. Interestingly, Cxcl12 mRNA was highly-expressed by rat liver and down-regulated in the Csf1rko (S1 Table).

The lack of CSF1 responsiveness in BM in the rescued Csf1rko rats (Fig 12K) is consistent with the failure of the BMT to rescue circulating monocyte numbers and indicates that the transplant does not contribute to the HSC and myeloid progenitor populations. This is not surprising since the recipients received no conditioning and there is no evidence of hematopoietic insufficiency. The implication is that there is a committed progenitor in rat BM that can directly provide long term engraftment of tissue mononuclear phagocyte populations without a monocyte intermediate. The ability of hematopoietic stem and progenitor cells to traffic through lymph and blood and to populate resident myeloid populations directly has been reported in mice [83]. The key donor population may be related to the CSF1-responsive clonogenic monocyte/macrophage progenitor described in mouse BM [84]. On the other hand, a recent study indicated that descendants of the yolk sac erythro-myeloid progenitor that can give rise to tissue macrophages and osteoclasts, may be present in the circulation [85]. Studies in the chick also revealed the existence of a BM progenitor that could produce long term macrophage-restricted chimerism when injected into the embryo prior to the onset of definitive hematopoiesis [4]. Further characterization of BM macrophage progenitor populations in the rat will depend on development of markers that are not currently available. The current study demonstrates the potential utility of the Csf1r-mApple transgene, which is currently being backcrossed to the DA background.

The second key implication of the rescue of the Csf1rko relates to the origins of resident macrophages. There is an emerging view based upon lineage trace models in mice that most tissue macrophages are seeded during embryonic development and thereafter maintained by self-renewal [1,2]. Whether or not this model can be extended to other species, the long-lived effectiveness of the rescue of Csf1rko rats indicates that macrophage territories in every organ can be occupied by cells of BM origin and maintained in the absence of CSF1R-expressing monocytes.

In conclusion, we have shown that severe developmental abnormalities in inbred Csf1rko rats can be reversed by WT bone marrow. The phenotypic rescue achieved by BMT is consistent with evidence that CSF1R expression is entirely restricted to MPS lineage cells (Reviewed in [6,86]) and all impacts of Csf1rko mutation are attributable to their absence.