To identify the genetic basis of HDLS, we established an international consortium with ethical approval from the Mayo Clinic Institutional Review Board and collected clinical data, MRI studies, blood and brain tissue samples from families with at least one patient with autopsy- or biopsy-proven HDLS. In total, we collected 14 kindreds from the United States, Norway, Germany and Scotland (Fig. 2). Family VA was selected for genome-wide linkage studies, and non-parametric linkage analyses identified one locus with a lod-score>2.5 (chromosome 5; lod=2.67) and four loci with lod-scores>1.0 (Supplementary Fig. 1). Subsequent parametric linkage analysis identified significant linkage on chromosome 5q34 (lod=3.71, θ=0 at rs13178296), while none of the other loci reached significance (Supplementary Fig. 1). Obligate recombinants narrowed the candidate region to 30.3cM between rs801399 and rs1445716 (Supplementary Fig. 2), corresponding to a ~25Mb genomic interval containing 233 candidate genes.

To generate a list of potential disease-causing mutations, we performed whole-exome sequencing of two pathologically confirmed patients from family VA (VA-21 and VA-24, Fig. 2). We generated variant profiles for each patient and searched for shared heterozygous variants located within the chromosome 5q candidate region. We further predicted that mutations underlying HDLS are likely to be previously unidentified; therefore, we filtered all of the identified base alterations against dbSNP132. This led to the identification of two non-synonymous mutations: c.80C>T (p.S27L) in the gene encoding the 5-hydroxytryptamine receptor 4 (HTR4) and c.2624T>C (p.M875T) in the macrophage colony-stimulating factor 1 receptor gene (CSF1R). Both mutations segregated with disease in the extended family VA and were absent in 660 controls. We therefore searched for additional mutations in a cohort of 13 probands from autopsy- or biopsy-proven HDLS families (Fig. 2). Sanger sequencing of the 6 coding exons of HTR4 and 22 coding exons of CSF1R identified heterozygous CSF1R mutations in all 13 probands, whereas no other mutations in HTR4 were identified (Fig. 3; Supplementary Table 1). Segregation analyses confirmed transmission of the CSF1R mutations and co-segregation with the disease phenotype in all families where DNA from multiple affecteds was available (Fig. 2). We further confirmed the de novo occurrence of one CSF1R mutation in monozygotic twins from family NO, without a family history of HDLS (Supplementary Fig. 3). To confirm the rarity of these mutations, and to provide supporting evidence for pathogenicity, we also sequenced the CSF1R gene in 24 unrelated controls and genotyped the 13 novel mutations in at least 1436 Caucasian controls using Taqman genotyping assays. None of the mutations identified in HDLS patients and no other novel CSF1R mutations were found in controls.

The 14 CSF1R mutations identified in HDLS families are all located in the intracellular tyrosine-kinase domain of CSF1R encoded by exons 12-22. The mutations include 10 missense mutations and one single-codon deletion, all affecting residues highly conserved across species and within members of the CSF1/PDGF receptor family of tyrosine-protein kinases (Kit, FLT3 and PDGFRα/β)¹³ (Fig. 3). We further identified three splice-site mutations, leading to the in-frame deletion of exon 13 (NO) or exon 18 (CA2/FL2), deleting up to 40 consecutive amino acids within the tyrosine kinase domain (Supplementary Fig. 4).

Detailed clinical information was available for 24 patients with proven CSF1R mutations from 14 HDLS families (Table 1). Mean age at onset was 47.2±14.5 years (range 18-78 years), with mean disease duration of 6.0±3.1 years (range 2-11 years) and a mean age at death of 57.2±13.1 years (range 40-84 years). In some families (FL1/CA1/VA), age at onset or death differed by more than 25 years among family members, whereas a monozygotic twin pair (family NO) showed highly similar disease course with ages at onset and death within one year from each other, suggesting that currently unidentified genetic or environmental factors may be important determinants of the age-related disease penetrance. Presenting features and evolving clinical symptoms also varied significantly within and across families, and ante mortem clinical diagnoses in mutation carriers included frontotemporal dementia (FTD), CBS, Alzheimer disease (AD), multiple sclerosis (MS), atypical cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), and Parkinson disease (PD).

Since most patients included in our study were not diagnosed with HDLS, we hypothesized that CSF1R mutation carriers may be present in clinical series of early-onset AD, FTD, CBS, MS and PD, or ischemic stroke patients with additional white matter changes. Sequencing analyses of CSF1R exons 12-22 encoding the protein tyrosine kinase domain in up to 93 Mayo Clinic patients affected with each of these neurological syndromes led to the identification of an additional CSF1R missense mutation c.2509G>T (p.D837Y) in a woman with clinical symptoms resembling CBS (Supplementary Tables 2-3). The identification of a CSF1R mutation in this limited patient series underscores that HDLS may be an under-diagnosed disease.

CSF1R is a cell-surface receptor primarily for the cytokine CSF-1, which regulates the survival, proliferation, differentiation and function of mononuclear phagocytic cells, including microglia of the central nervous system¹⁴. CSF1R is composed of a highly glycosylated extracellular ligand-binding domain, a trans-membrane domain and an intracellular tyrosine-kinase domain¹⁵. Binding of CSF-1 to CSF1R results in the formation of receptor homodimers and subsequent auto-phosphorylation of several tyrosine residues in the cytoplasmic domain¹⁶. CSF1R autophosphorylation precedes CSF1R-dependent phosphorylation of several proteins, including the phosphatase SHP-1 and the kinases Src, PLC-g, PI(3)K, Akt and Erk^16-18. In the brain, CSF1R is predominantly expressed in microglial cells, although low levels of CSF1R have been reported in cultured neurons^19-21. An increase in CSF1R copy number and point mutations leading to constitutive activation of the CSF1R receptor have been associated with tumor development, including hematological malignancies and renal cell carcinomas^22,23.

To assess the functional importance of the CSF1R mutations identified in this study, we first studied the effect of the mutations on CSF1R in vitro. We transiently expressed DDK-tagged wild-type (CSF1R_WT) and mutant (CSF1R_E633K, CSF1R_M766T and CSF1R_M875T) CSF1R in cultured cells. Upon stimulation with CSF-1, autophosphorylation on multiple CSF1R tyrosine-residues was observed for CSF1R_WT, while none of the mutants showed detectable levels of autophosphorylation (Fig. 4 and Supplementary Fig. 5). Since all mutations are in the CSF1R kinase domain, dimerization and/or cell surface expression are unlikely to be affected; however, we cannot exclude this at this time. These preliminary findings suggest that mutant CSF1R kinase activity is abrogated, likely affecting the phosphorylation of downstream targets. We speculate that mutant CSF1R might assemble into non-functional homodimers and wild-type/mutant heterodimers inducing a dominant-negative disease mechanism.

To address whether CSF1R autophosphorylation is also disrupted in HDLS patient samples, we first subjected blood samples from a healthy control and HDLS patient CA1-1 to CSF1R immunoblotting, which revealed no apparent difference in CSF1R total or phosphorylation levels (Supplementary Fig. 6a). Further, CSF1R immunoblotting was performed in frontal cortex brain tissue of healthy controls as well as patients with HDLS. Brain samples from AD and ALS patients were included as neurodegenerative disease controls. Our data showed varied levels of total and phosphorylated CSF1R in these brain samples (Supplementary Fig. 6b); however, statistical analysis did not reveal a significant difference between any of the groups. Although these preliminary in vivo studies do not reveal a defect in autophosphorylation, these findings do not necessarily conflict with the data obtained in cultured cells. First, HDLS patients are heterozygous for the CSF1R mutations and therefore, in contrast to our in-vitro experiments, wild-type receptor is still present in these patients. In our cell culture experiments, CSF1R signaling was down regulated by serum deprivation to minimize basal signaling through this receptor before stimulation with the CSF-1 ligand. CSF-1 is a serum protein, so without this deprivation in vivo, immediate changes in CSF-1-induced CSF1R autophosphorylation may not be apparent as we cannot disregard wild-type receptors at the cell surface that have already been activated. Unfortunately, without access to an immortalized cell line derived from an HDLS patient, we are currently unable to accurately assess acute receptor activation in vivo. Finally, the post-mortem brain samples from HDLS patients included in these studies exhibit extensive degeneration, leaving the possibility that cells with greater disruption of CSF1R signaling are underrepresented in the tissue sample.

Unraveling the genetic etiology of HDLS may significantly contribute to the understanding of other adult-onset leukoencephalopathies. De novo mutations in CSF1R could explain the disease in sporadic patients that have been described with clinical and pathological similarities to HDLS^24-29. Future CSF1R mutation screening may also determine whether HDLS and pigmentary orthochromatic leukodystrophy (POLD) are part of a single clinicopathologic entity, as was recently suggested². Moreover, the discovery of a mutation in a microglial trophic factor receptor may further elucidate the role of microglia in more common white matter disorders, particularly those associated with axonal dystrophy, such as Binswanger’s disease^24,30, multiple sclerosis³¹ and HIV encephalitis³².

Interestingly, our findings also shed new light on Nasu-Hakola disease (NHD), a rare condition characterized by systemic bone cysts and dementia with striking similarities to HDLS^33-35. NHD is caused by recessive loss-of-function mutations in the DAP12/TREM2 protein complex^36,37, which was recently implicated in CSF1R signaling, establishing NHD as a primary microglial disorder³⁸. We speculate that a partial loss of the CSF1R/DAP12 signaling cascade in microglia is responsible for the neurological phenotypes observed in HDLS and NHD, whereas a complete loss of this signaling cascade in bone marrow-derived macrophages is needed for the bone-cysts formation observed in NHD. In support of this hypothesis, a partial loss-of-function mutation in TREM2 in a family with early-onset dementia without bone-cysts was recently reported³⁹. Also, no bone-cysts were reported in any of our HDLS patients and a bone scan in a patient CA1-1 did not show bone fractures, hypomineralization or any other bone structure abnormalities.

In summary, we have shown that mutations affecting the tyrosine-kinase domain of CSF1R underlie the white matter disease of HDLS, establishing HDLS as an important novel member of the recently defined class of primary microglial disorders, called ‘microgliopathies’⁴⁰. Future molecular studies of CSF1R signaling might offer novel insights into microglial physiology and the involvement of this cell type in HDLS and neurodegeneration. Moreover, CSF1R mutation screening in neurodegenerative disease patient series will now allow an accurate diagnosis of HDLS and could facilitate detection of presymptomatic individuals, which is indispensable for therapy development and early treatment.

1

2