Inactivating Mutation in IRF8 Promotes Osteoclast Transcriptional Programs and Increases Susceptibility to Tooth Root Resorption

Abstract

This is the first study to report a novel mutation in the interferon regulatory factor 8 gene (IRF8^G388S) associated with Multiple Idiopathic Tooth Root Resorption, a form of periodontal disease. The IRF8^G388S variant in the highly conserved C-terminal motif is predicted to alter the protein structure, likely impairing IRF8 function. Functional assays demonstrated that the IRF8^G388S mutant promoted osteoclastogenesis and failed to inhibit NFATc1-dependent transcriptional activation when compared to IRF8^WT control. Further, similar to subjects with heterozygous IRF8^G388S mutation, Irf8^+/− mice exhibited increased osteoclast activity in the mandibular alveolar bone surrounding molar teeth. Immunohistochemistry illustrated increased NFATc1 expression in the dentoalveolar region of Irf8^−/− and Irf8^+/− mice when compared to Irf8^+/+ controls. Genome-wide analyses revealed that IRF8 constitutively bound to regulatory regions of several thousand genes in osteoclast precursors, and genetic aberration of IRF8 significantly enhanced many osteoclast-specific transcripts. Collectively, this study delineates the critical role of IRF8 in defining osteoclast lineage and osteoclast transcriptional program, which may help in better understanding of various osteoclast-mediated disorders, including periodontal disease.

Introduction

Periodontal disease affects nearly half of adults aged 30 years and above ⁽¹⁾. Although the disease process is initiated by the oral microbial biofilm, it is the hyperactive immunoinflammatory response to the microflora and subsequent activation of osteoclasts that cause destruction of tooth-supporting structures ⁽²⁾. Currently, the molecular processes governing osteoclast-mediated tissue destruction in periodontal disease remain unclear.

Tooth root resorption mediated by osteoclasts/odontoclasts is a normal physiologic process occurring in primary dentition that leads to exfoliation of deciduous teeth ⁽³⁾. However, root resorption in permanent teeth is largely pathological ⁽⁴⁾. Multiple Idiopathic Cervical Root Resorption (MICRR) is a form of osteoclast-mediated pathologic root resorption that affects multiple teeth within the dentition and its etiology is unknown ^(5–7). Clinically, MICRR lesions are asymptomatic, non-carious, and unassociated with signs of overt gingival inflammation, increased pocket depth, or tooth mobility. Histologically, numerous resorptive areas are noted along the root surface with evidence of osteoclasts/odontoclasts contained in Howship’s lacunae ^(5–7). MICRR lesions are often aggressive in nature and resistant to interventions, resulting in tooth loss ^(5–7). Previous reports suggest a genetic predisposition to MICRR ^(5,8).

By examining a rare cohort of familial susceptibility to MICRR, this study identified a novel mutation in the interferon regulatory factor 8 gene (IRF8 ^G388S) that leads to increased osteoclastogenesis. As an initial step in understanding the MICRR disease process, this study focused on investigating the effects of IRF8^G388S mutation on osteoclast regulation and further characterizing the genome-wide effects of IRF8 on osteoclastogenesis. Our results delineate important IRF8-mediated regulators of osteoclastogenesis and provide novel insights into the changes in osteoclast transcriptome profile induced by IRF8 deficiency.

MATERIALS AND METHODS

Detailed materials and methods are provided in the supplementary section.

Human subjects

Research was performed under the protocol approved by the National Institutes of Health (NIH) Office of Human Subjects Research Protections (OHSRP) (protocol# 11497). All subjects agreed to participate in the study and provided saliva samples for genetic analysis.

Whole-exome sequencing, Sanger sequencing, and molecular modeling

Genomic DNA was isolated from whole saliva using an all-in-one collection kit (DNA Genotek Inc.). Whole-exome sequencing was performed by a service provider (Otogenetics) using Illumina® HiSeq sequencers (Illumina) with 2×100+ bp paired reads. Using bioinformatics and statistical genetics tools, likely disease-causing mutations were prioritized based on quality score, allele frequency, functional impact, probable inheritance models, and expert evaluation. Verification of the variant was performed by Sanger sequencing. Sequencing was performed with Big Dye Terminators v3.1 (Applied Biosystems) and aligned to the consensus sequence NM_002163.2. The effects of IRF8 mutation were evaluated in silico using software including PolyPhen-2, PROVEAN, and SIFT. Multiple sequence alignments of IRF8 from different species were performed using Clustal Omega.

Mice

The Irf8^−/− mice used in this study have been previously described ⁽⁹⁾. Animal procedures were performed according to protocols approved by the NIAMS and NICHD Institutional Animal Care and Use Committees (ASP# A017–09-05 and 17–044, respectively). Micro-CT and histological analyses were performed on Irf8^+/+, Irf8^+/−, and Irf8^−/− mice.

Plasmids and vectors

EGFP-tagged human IRF8 wild type (hIRF8^WT), G388S mutant (hIRF8^G388S), and a larger deletion around the G388 residue (hIRF8^379−389) were generated using site-directed mutagenesis. The functional consequences of these mutants were determined by performing cellular and functional studies including luciferase reporter assay and protein stability assay.

In vitro Osteoclast Assay

Primary bone marrow cells were harvested from femurs and tibias of 8–10-week-old Irf8^+/+ and Irf8^−/− mice and cultured overnight in osteoclast media supplemented with 20 ng/ml M-CSF (R&D Systems). The non-adherent cell population was recovered and then selected for CD11b+ cells (CD11b microbeads, Miltenyi Biotec) and re-plated in 12-well plates (Corning) at 5×10⁵ cells/cm² (plated in triplicates) in osteoclast media supplemented with retroviral soup and 20 ng/ml M-CSF. The following day, cells were re-infected with retrovirus and 40 ng/mL RANKL. Later, media were replaced every 2 two days with 20 ng/ml M-CSF and 40 ng/mL RANKL to generate osteoclasts. RNA was extracted for RT-qPCR analysis, protein was extracted for western blot analysis, or cells stained for TRAP. The resorption activity of osteoclasts was examined using Osteo Assay Surface Plates.

RNA-seq analysis and ChIP-seq calling

Total RNA was first extracted using RNeasy mini kit (Qiagen) and RNA integrity was analyzed with TapeStation (Agilent Technologies). mRNA purification and fragmentation, cDNA synthesis, and target amplification were performed using the Illumina NeoPrep system (Illumina). Pooled cDNA libraries were sequenced using the Illumina HiSeq 2000 platform (Illumina). Data from independent runs of 50 bp single-end reads were mapped to the mouse genome (UCSC genome browser, mm9 assembly) using TopHat 2.1.0. Mapped reads were analyzed using Partek GS v6.6 software and normalized to RPKM (reads per kilobase of exon per million mapped). Previously published ChIP-seq profile immunoprecipitated from murine BMMs ⁽¹⁰⁾ was used to map genome-wide IRF8 binding sites.

Results

MICRR Patients and Clinical History

We studied a two-generation family with four MICRR affected and four unaffected family members (Fig. 1A). The family’s detailed medical history, dental history, and management of MICRR has been previously reported ^(5,6). Briefly, the index case proband (I:1) initially at the age of 63 years (current age 97 years) presented with a history of MICRR affecting multiple teeth within the dentition. Over several decades, the resorptive lesions progressed with a total of 19 teeth becoming affected and leading to extraction/exfoliation of 12 teeth. In addition, the proband’s two sons (II:2 and II:4) and one daughter (II:3) also developed MICRR during fourth to sixth decades of life. In all affected subjects, the resorptive lesions were localized to the cervical portion of tooth root structure, and they were asymptomatic and unassociated with any known predisposing factors. All family members reported a non-contributory medical history. The affected family members reported no history of immunological abnormalities or severe bacterial and viral infections indicative of immune cell deficit. Blood workup as part of their routine medical examinations showed no deficiencies in immune cells.

Fig. 1. Genetic analysis of IRF8 mutation in familial MICRR.

(A) Pedigree of affected family. Also shown are IRF8 variants detected by whole-exome sequencing and current age of the family members. Asterisk indicates non-detection of IRF8 G388S variant in II:4. (B) Schematic representation of human IRF8 gene showing the position of G388S variant in the C-terminal (bold red). DBD denotes DNA-binding domain, and IAD denotes IRF association domain. E=exons. (C) Multiple sequence alignment of IRF8 orthologs for region downstream of the IAD domain (G388 region is shown in box; residues conserved across species are marked by *). (D) Confirmation of heterozygous IRF8 variant c.1219 G>A (G388S) in affected family members I:1, II:2, and II:3 by Sanger sequencing. (E) Structural model of wild-type (G388) and mutant (G388S) IRF8. The rainbow spectrum indicates order of protein sequences from N-terminal (blue) to C- terminal (red). The substituted G388S form shows the C-terminal portion turned to the protein interior, resulting in a dramatically altered protein structure.

Genetic Analysis and Identification of Candidate Gene

No genetic analysis was done on the kindred previously, therefore, we performed whole-exome sequencing using genomic DNA obtained from saliva. Excellent average coverage per-base was observed for all samples. A total of 22,342 variants were identified across all samples, including SNPs, INDELs, and mCNVs. After filtering out variants that did not pass quality control, 10 missense variants segregated in the affected individuals I:1, II:2, II:3, and they were absent in all unaffected family members (Table S1 lists variants). To identify risk alleles, we focused on variants involved in abnormal osteoclast regulation, craniofacial development, and inflammatory diseases. Among the identified SNPs, the most prominent candidate was a novel missense variant (c.1219 G>A; p.Gly388Ser) in the C-terminal region of IRF8 (Fig. 1B). Alignment view of this variant is presented in Fig. S1. The IRF8 G388 residue is highly conserved among mammalian species (Fig. 1C and Fig. S2). Whole-exome sequencing revealed proband (I:1) and two of his affected children (II:2, II:3) were heterozygous for G388S, which was confirmed by Sanger sequencing (Fig. 1D). The identified G388S variant is absent in the ClinVar, dbSNP or ESP databases, but was found as a heterozygous variant in 1 out of 119,482 chromosomes (allelic frequency of 8.369×10⁻⁶) in the ExAC database (Table S1). There are currently no homozygous or compound heterozygous nonsynonymous IRF8 G388S mutations found in public databases.

The C-terminal mutation within exon 9 of IRF8 gene is immediately adjacent to the IRF association domain (IAD), a critical regulatory domain that mediates homo- and heterodimeric interactions among IRFs, and between IRFs and other transcriptional co-modulators ^(11–14). Conformational and bioinformatics analyses based on closely related IRF family members (since IRF8 crystal structure is unsolved) suggested that the G388S substitution would disrupt the alpha helix motif and cause the C-terminal portion to be turned towards the protein interior, potentially masking the terminal 37 amino acids (389–426) and resulting in an altered protein structure (Fig. 1E). The altered protein structure is predicted to impact DNA-binding specificity/affinity and transactivation potential of IRF8. In addition, analysis of known phosphorylation and serine binding motifs suggested that the G388S substitution could also create novel glycosylation, phosphorylation, and/or protein interaction sites that could potentially affect dimerization or DNA binding (Figs. S3 and S4).

Inexplicably, the G388S variant was not detected in the fourth affected individual (II:4), which could be related to mosaicism. This subject shared only two missense variants identified in other affected family members - RET (c.166C>A; p.L56M) and SLC45A1 (c.2080G>C; p.V694L) (Table S1). In contrast to the IRF8 G388S mutation, RET and SLC45A1 variants have been reported more commonly in dbSNP, ESP, and ExAC databases. The RET L56M variant is reported 301 times in ExAC database (allelic frequency of 2.497×10⁻³) with one homozygous case. The SLC45A1 V694L variant is reported 62 times in ExAC database (allelic frequency of 5.117×10⁻⁴) with one homozygous case. Of the identified variants, none apart from IRF8 have an established role in osteoclast regulation and bone metabolism ^(15,16).

To select genes for functional validation, we further investigated whether a skeletal phenotype had been reported for prioritized genes in the Mouse Genome Informatics (MGI; “URL”) or International Mouse Phenotyping Consortium (IMPC; “URL”) databases. We found that among the prioritized genes, only Irf8 knockout mouse data included skeletal abnormalities ^(15,16). Another IRF family member, Irf1, has been reported to have a role in bone metabolism, and Irf1 ablation results in skeletal abnormalities ⁽¹⁷⁾. Based on these findings, IRF8 was selected for functional validation of the G388S substitution.

IRF8^G388S Variant Leads to Increased Osteoclastogenesis

To determine how the G388S mutation affects IRF8 function in osteoclast regulation, we elected to study osteoclast differentiation using Irf8^−/− mice bone marrow progenitor cells transduced with human IRF8^G388S. IRF8 is well conserved among species, and IRF8’s role in myeloid cell development and inflammation has been well documented using Irf8^−/− mice ^(9,18–21). We created expression vectors of EGFP-tagged human IRF8 wild type (hIRF8^WT), G388S mutant (hIRF8^G388S), and a larger deletion around the G388 residue (hIRF8^379−389) and expressed them in Irf8^−/− bone marrow macrophages (BMMs) using a retroviral vector. BMMs were incubated with M-CSF and RANKL to induce osteoclastogenesis. Immunofluorescent staining of GFP signals and RT-qPCR analysis of hIRF8 mRNA showed that their expression levels were comparable between groups, verifying that the G388S mutant and 379–389 deletion did not affect IRF8 expression (Fig. 2A). Notably, hIRF8^G388S and hIRF8^379−389 transduced BMMs exhibited increased osteoclast formation, resorption activity, and osteoclast-specific gene expression when compared to hIRF8^WT (Fig. 2A–C). The average size of TRAP-positive multinucleated cells (≥3 nuclei) was increased by ~3-fold in both hIRF8^G388S and hIRF8^379−389 groups, when compared to hIRF8^WT (Fig. 2B).

Fig. 2. IRF8 G388S mutation leads to increased osteoclastogenesis.

(A) Top panel shows fluorescent images of GFP-positive multinucleated cell (osteoclast) formation due to retrovirus-mediated exogenous expression of hIRF8^WT, and hIRF8^G388S and hIRF8^379−389 mutants in BMMs stimulated with M-CSF and RANKL. Scale bar, 50 μm. Lower panel shows relative expression of hIRF8 retroviral vectors in unstimulated (D0) and RANKL-stimulated (D4 and D8) BMMs, measured by RT-qPCR analysis. (B) TRAP staining (top) and resorption activity (bottom) of osteoclasts formed by exogenously expressed hIRF8^WT, and hIRF8^G388S and hIRF8^379−389 mutants. Scale bar, 100 μm. In lower most panel, bar graphs show the quantified number of TRAP⁺ cells, average cell size, and osteoclast resorption activity in each group. Experiments were performed at least in triplicate, and results are representative of more than three experiments. (C) RT-qPCR analysis of osteoclast-specific genes in unstimulated (D0) and RANKL-stimulated (D4 and D8) BMMs. Data compared against hIRF8 WT. (D) Luciferase assay was performed to determine the effects of hIRF8^G388S and hIRF8^379−389 mutants on the transcriptional activity of NFATc1 through activation of human Ctsk promoter. The numbers indicate the amount of DNA transfected (nanograms) for each variant. (E) The effect of hIRF8^G388S and hIRF8^379−389 variants on the transactivation of Ctsk promoter by hIRF8^WT was evaluated by co-transfecting 293T cells with WT and mutant vectors. (F) Interaction between hIRF8^WT, hIRF8^G388S and NFATc1 was assessed by coimmunoprecipitation. IB, immunoblotting. ****P <0.0001, ***P <0.001, **P <0.01, *P<0.05 and NS=non-significant (one-way ANOVA).

It has been reported that the IAD domain of IRF8 (IAD1: N/200–377aa) physically interacts with the TAD-A domain of NFATc1 (N/1–205aa) to inhibit NFATc1 nuclear translocation and activation of downstream target genes ⁽¹³⁾. The identified G388S mutation is 11 amino acids downstream of the IAD domain, which could impact IRF8 interaction with NFATc1. Therefore, we examined the effects of hIRF8^G388S and hIRF8^379−389 mutations on the transcriptional activity of NFATc1 through activation of human Ctsk promoter ⁽²²⁾. Overexpression of Nfatc1 gene activated the Ctsk promoter, whereas simultaneous expression of hIRF8^WT reduced the activity of the promoter to control levels (Fig. 2D), indicating that hIRF8^WT inhibits the transcriptional activity of NFATc1. Conversely, both hIRF8^G388S and hIRF8^379−389 mutants failed to inhibit NFATc1-dependent transcriptional activation of Ctsk promoter (Fig. 2D), suggesting that the mutations disrupt functional binding of IRF8 with NFATc1. We then investigated whether the G388S variant functions in a dominant negative manner. We co-expressed hIRF8^WT and mutant alleles in 293 cells and observed loss of inhibition of NFATc1-dependent transcriptional activation of Ctsk promoter with increasing levels of mutants hIRF8^G388S and hIRF8^379−389 added to a fixed level of nonvariant hIRF8^WT (Fig. 2E). These data suggest that the heterozygous G388S mutation results in a dominant-negative IRF8 allele, which competes with function of the hIRF8^WT protein and impairs the suppression of NFATc1 transactivation in vitro.

Furthermore, we examined whether hIRF8^G388S interacts with NFATc1 using coimmunoprecipitation assay. Immunoprecipitation of Myc-NFATc1 resulted in lack of co-precipitation of IRF8, indicating that hIRF8^G388S failed to interact with NFATc1 when compared to hIRF8WT (Fig. 2F). In addition, we investigated the stability of the hIRF8^G388S and hIRF8^379−389 variants in 293 cells treated with cycloheximide (CHX), an inhibitor of protein synthesis. Over a 12-hour period, we did not observe enhanced degradation of hIRF8^G388S or hIRF8^379−389 variants when compared to hIRF8^WT (Fig. S5), suggesting that the mutant proteins are stable and that the loss of IRF8 function is more likely because of its inability to interact with other transcription factors such as NFATc1.

IRF8 Deficiency Shapes Osteoclast Transcriptomes

While the role of IRF8 in osteoclast regulation has been established ^(15,16), how globally IRF8 controls transcription during osteoclast formation and how critical IRF8 is for defining osteoclast lineage is not fully understood. Therefore, it was important to delineate the global role of IRF8 in osteoclast differentiation before analyzing the global effects of hIRF8^G388S mutation. We performed RNA-seq analysis of Irf8^+/+ and Irf8^−/−BMMs prior to (day 0 (D0)) and after RANKL stimulation (day 4 (D4) and day 8 (D8)). Principal components analysis (PCA) clustered the samples along two axes: component 1 associated with RANKL stimulation, and component 2 associated with mouse genotype (Fig. 3A). About 50% of the genes (n=4,910) regulated by RANKL were common to both Irf8^+/+ and Irf8^−/− groups, whereas 28% of the genes were unique to Irf8^−/−, and 22% of the genes were unique to Irf8^+/+ (Fig. 3B). One-way ANOVA analysis identified significantly increased expression of genes in Irf8^−/− vs. Irf8^+/+ cells at D0, D4 and D8 (fold change >2 or < −2, FDR<0.01, Fig. 3C). Hierarchical clustering of RANKL-responsive genes identified four major categories of transcripts (Fig. 3D). Group 1 genes (n=1,945) expression levels were increased by RANKL in both groups, but were more pronounced in Irf8^−/− vs. Irf8^+/+; Group 2 genes (n=527) were significantly increased by RANKL in Irf8^+/+ cultures but not in Irf8^−/− cells; Group 3 genes (n=656) were more profoundly increased in Irf8^−/− BMMs at D0; and Group 4 genes (n=1,971) were significantly downregulated by RANKL in both groups, but more distinct in Irf8^−/− vs. Irf8^+/+ cells. GO annotations associated these gene groups with different enriched functions (Fig. 3E). Upregulated genes in Group 1 included several well-established osteoclast markers such as Nfatc1, Acp5, Ctsk, Calcr, Myc, Mmp9, Itgb3, Dcstamp, Ocstamp, and many other novel genes that may positively regulate osteoclastogenesis (Fig. 3F). Downregulated genes in Group 4 included Cybb, Lamp1, Apoe, Ly86, Cd40, Mafb, Tnfrsf11b, Irf1, Il-10, Def6, Runx1, Stat1, and many others that are important for immune response and regulation of bone metabolism.

Fig. 3. Osteoclastogenesis transcriptional program is regulated by Irf8 deficiency.

(A) PCA of RNA-seq data from Irf8^+/+ and Irf8^−/− BMMs before and after RANKL treatment. (B) Venn diagram depicting overlaps between RANKL responsive genes in Irf8^+/+ and Irf8^−/− groups at D4 and D8. (C) Volcano plot illustrates significant differences in gene expression pattern between Irf8^−/− vs. Irf8^+/+ cells (one-way ANOVA, fold change >2 or < −2, FDR<0.01). Red dots represent non-RANKL responsive genes, green dots represent RANKL responsive genes, and black dots represent no significant change. (D) Hierarchical clustered heat map of transcripts regulated in Irf8^+/+ and Irf8^−/− cells during osteoclast differentiation process. Differential genes clustered into four groups. (E) Enriched functions for each group based on gene ontology. (F) Expression level changes for selected osteoclast-specific markers (n=78) in Irf8^+/+ and Irf8^−/− groups. Red shaded heat map indicates the presence of one or more IRF8 binding sites. (G) RNA-seq gene expression data was analyzed using GSEA, which showed strong enrichment of osteoclast-specific signatures in Irf8^−/− vs. Irf8^+/+ group. Data in heat maps are presented as standardized RPKM scale.

Prominent increase or decrease in gene expression was first observed at D4, which may stem from an enrichment or depletion of certain mRNAs in the adherent cell population at D4 when compared to undifferentiated mononuclear BMMs at D0. However, in Irf8^−/− cells, the basal expression (D0) of many established positive regulators of osteoclastogenesis (n=73 genes) were enriched and negative regulators (n=104 genes) were diminished (Irf8^−/− vs. Irf8^+/+, fold change >2 or < −2, Table S2 and S3). These results suggest that the major differentiation program and lineage commitment is defined before D4, and that genetic ablation of Irf8 initiates osteoclast lineage transcriptional changes and defines its differentiation path at a very initial stage.

To study the effect of Irf8 deficiency on ontogeny and activity of osteoclast and osteoclast precursor populations, we performed GSEA analyses comparing the Irf8-associated gene expression profile with cell-specific transcript signatures obtained from Ingenuity Pathway Analysis (IPA) and previously published datasets ⁽²³⁾. GSEA analysis showed strong enrichment of osteoclast-specific signatures in Irf8^−/− vs. Irf8^+/+ cells (Fig. 3G).

The fact that several genes differentially regulated by RANKL were significantly higher or lower in Irf8^−/− vs. Irf8^+/+ cells, suggests that many of them may be direct transcriptional targets of IRF8. To establish a correlation between gene expression and IRF8 binding, we mapped genome-wide IRF8 binding sites onto our RNA-seq data using a previously published ChIP-seq profile immunoprecipitated from murine BMMs ⁽¹⁰⁾. We noted IRF8 was constitutively bound to thousands of genes (7,887, Table S4) in BMMs at D0. In agreement with our hypothesis, we noted clusters with significant enrichment of transcripts or depletion of transcripts from genes close to IRF8 binding sites in the Irf8^−/− group when compared with Irf8^+/+ (Fig. 3F and Fig. S6). The enriched genes (e.g., Acp5, Dcstamp, Oscar, Fig. S6) were associated predominantly with positive osteoclast regulation and repressed genes included negative regulators of osteoclastogenesis (e.g., Bcl6, Rbp-j, Mafb).

IRF8^G388S Mutation Promotes Osteoclastogenesis Transcriptional Program

To gain insight into functional consequences of hIRF8^G388S mutation on osteoclastogenesis, we compared genome-wide expression profiles of Irf8^−/− BMMs transduced with mock (empty), hIRF8^WT, hIRF8^G388S, or hIRF8^379−389 expression vectors, followed by RANKL stimulation. hIRF8 expression levels were comparable between groups and absent in mock as measured by EGFP RPKM count (Fig. 4A). Transfer of the hIRF8 gene into Irf8^−/− BMMs restored substantial gene expression profile, which demonstrates sufficient levels of hIRF8 transduction (Fig. 4A). PCA illustrated a major effect of RANKL (PC1) and genotype (PC2) on osteoclast gene expression programs (Fig. 4B). Between hIRF8^WT and hIRF8^G388S groups, approximately 54% of the genes were commonly regulated, whereas 30% of the genes were unique to hIRF8^WT, and 15% of the genes were unique to hIRF8^G388S (Fig. 4C). One-way ANOVA analysis identified genes with increased expression (D0: 343; D4:865; D8: 238) and genes with decreased expression (D0: 131; D4: 883; D8: 90) in hIRF8^G388S cells when compared to hIRF8^WT (Fig. 4F. hIRF8^G388S vs. hIRF8^WT, fold change >1.5 or < −1.5, P<0.05). These lists are expected to represent genes that are differentially expressed among BMMs (D0) vs. preosteoclasts (D4) vs. osteoclasts (D8), which are regulated by IRF8 and expected to be more or less numerous in MICRR affected individuals vs. unaffected controls. Hierarchical clustering of RANKL-responsive genes identified four major categories of transcripts (Fig. 4D). GO annotations associated these groups with different functions (Fig. 4E). Similar to Irf8^−/− cells, upregulated genes in hIRF8^G388S (Group 1) cells included several known osteoclast markers such as Nfatc1, Acp5, Ctsk, Junb, Mmp14, Mmp9, Fos, Dcstamp, Clcn7, Src, Traf1, Actn1 (Fig. S7). Both RNA-Seq and RT-qPCR showed similar expression changes for several osteoclast-related genes (Fig. 4G).

Fig. 4. hIRF8^G388S mutation promotes osteoclastogenesis transcriptional program.

(A) Top panel shows relative expression of hIRF8 in unstimulated (D0) and RANKL-stimulated (D4 and D8) BMMs, measured by EGFP RPKM count. Lower panel shows exogenous hIRF8 expression rescues the gene expression profile noted in Irf8^−/− BMMs transduced with empty vector (mock) (2-fold change, FDR<0.01). (B) PCA of RNA-seq data from hIRF8^WT, hIRF8^G388S and hIRF8^379−389 transduced BMMs before and after RANKL treatment. (C) RANKL responsive genes that are common in hIRF8^WT and hIRF8^G388S groups at D4 and D8. (D) Expression level of genes regulated by RANKL treatment in different groups. Hierarchical clustering segregated differentially expressed genes clustered into four groups. (E) Enriched functions for each group based on gene ontology. (F) Volcano plot illustrates significant differences in gene expression pattern between hIRF8^G388S vs. hIRF8^WT cells (one-way ANOVA, fold change >1.5 or < −1.5, P<0.05). Red dots represent significantly upregulated genes, green dots represent significantly downregulated genes, and black dots represent no significant change. (G) Relative expression of osteoclast-specific genes in hIRF8^G388S group compared to hIRF8^WT. Both RNA-Seq and RT-qPCR analyses show similar expression changes. (H) GSEA analysis showed strong enrichment of osteoclast-specific signatures in hIRF8^G388S group vs. hIRF8^WT. (I) Venn diagram illustrates correlation between Irf8^−/−, Irf8^+/+, hIRF8^G388S, and hIRF8^WT gene expression identified by RNA-seq. Data in heat maps are presented as standardized RPKM scale.

As expected from the root resorption phenotype noted in MICRR patients, the GSEA analysis showed strong enrichment of osteoclast-specific signatures in hIRF8^G388S group vs. hIRF8^WT (Fig. 4H and Fig. S8). These results are consistent with the osteoclast-specific transcript signature enriched in Irf8^−/− vs. Irf8^+/+ cells, suggesting that the increased osteoclastogenesis noted in Irf8^−/− and hIRF8^G388S cultures is a consequence of impaired IRF8 function. We also compared the list of transcripts altered in hIRF8^G388S cultures with a list of direct transcriptional targets of IRF8. We noted a strong and significant enhancement of transcripts in the hIRF8^G388S group from genes predominantly associated with osteoclast regulation, many of which are direct IRF8 transcriptional targets (Fig. S6 and Fig. S7). These results are consistent with Irf8^−/− profile (Fig 3F and Fig. S6). The correlation between Irf8^−/−, Irf8^+/+, hIRF8^G388S, and hIRF8^WT RNA-seq results is illustrated in Fig. 4I.

IRF8^G388S Mutation has minimal Impact on Transcriptome in IFN-γ+LPS Treated Macrophages

IRF8, along with IRF1 and PU.1, is known to direct expression of a subset of genes important for macrophage antimicrobial defenses and production of inflammatory cytokines ^(10,19,24). IRF8 deletion results in impaired myeloid cell development, macrophage activity, and dendritic cell function ^{(10,19,20,24–26)}. In order to determine the effects of hIRF8^G388S and hIRF8^379−389 mutants on gene expression programs essential for macrophage function, we performed RNA-seq analysis in resting and IFN-γ+LPS– activated macrophages. PCA illustrated no major difference between hIRF8^WT and mutants on LPS-inducible genes (Fig. 5A). hIRF8 mRNA expression levels were comparable between groups and absent in mock as determined by RT-qPCR analysis (Fig. S9) and EGFP RPKM count (Fig. 5B). IFN-γ caused upregulation of several genes and a subset of those genes were further upregulated with LPS stimulation at 4 and 8 hr (Fig 5C and 5D). Hierarchical clustering of genes (Fig. 5D) showed minimal difference between hIRF8^WT and both mutant forms. Furthermore, RT-qPCR analysis of macrophage-related genes showed marginal difference between hIRF8^WT and mutants (Fig. 5E).

Fig. 5. hIRF8^G388S has minimal impact on transcriptome in IFN-γ+LPS treated macrophages.

(A) PCA of RNA-seq data from resting and IFN-γ+LPS–activated macrophages transduced with Mock, hIRF8^WT, hIRF8^G388S and hIRF8^379–389. (B) Relative expression of hIRF8 retroviral vectors in untreated and IFN-γ+LPS treated macrophages, measured by measured by EGFP RPKM count. (C) Venn diagram depicting overlaps between IFN-γ+LPS genes in Mock, hIRF8^WT, hIRF8^G388S and hIRF8^379−389 groups at 4hr and 8hr. (D) Hierarchical clustering of genes shows minimal difference between hIRF8^WT and mutants. (E) RT-qPCR analysis of resting and IFN-γ+LPS–activated macrophages. Data compared against hIRF8 WT. Data in heat map are presented as standardized RPKM scale.

IRF8 Deficiency Leads to Increased Osteoclast Activity in Mice Jaw Bones

While previous findings ⁽¹⁶⁾ reveal that Irf8^−/− mice exhibit an osteoporosis phenotype, the effect of heterozygous IRF8 deficiency on osteoclast activity, particularly in the craniofacial region remains unknown. Therefore, we performed phenotypic analysis of femurs and mandibles in Irf8^−/−, Irf8^+/−, and Irf8^+/+ mice. Micro-CT analysis of femurs revealed that Irf8^+/− mice display decreases in total amount of bone, trabecular BMD, trabecular number, and trabecular thickness when compared with Irf8^+/+ mice (Fig. 6A). Histological analyses of femurs and mandibles by TRAP staining showed a substantial increase in osteoclast numbers and resorption activity in Irf8^−/− and Irf8^+/− mice when compared to Irf8^+/+ controls (Fig. 6B and 6C). The increased number of osteoclasts lining the alveolar bone (surrounding molar teeth) were spontaneously noted in Irf8^−/− and Irf8^+/− mice without exogenous inflammatory challenge. Immunohistochemistry illustrated increased NFATc1 expression in the bone marrow and dentoalveolar region of Irf8^−/− and Irf8^+/− mice when compared to Irf8^+/+ controls (Fig. S10 and Fig. 6D). The loss of Irf8 function may directly lead to the increased expression of NFATc1 in these mice. Furthermore, the increased osteoclast numbers and NFATc1 expression may “prime” or “predispose” the periodontium for developing MICRR, and other factors such as subgingival bacterial inflammation and/or trauma to the cervical root surface may collectively lead to the development of MICRR. Together, these data suggest that the increased osteoclast activity noted in Irf8^−/− and Irf8^+/− mice is a consequence of Irf8 deficiency, which supports an important parallel to our patients with heterozygous IRF8 mutation and increased frequency of osteoclast activity in the jaw bones.

Fig. 6. Irf8 deficient mice exhibit increased osteoclast activity in femurs and jaw bones.

(A) Microcomputed tomography of the femurs of 8.5-week-old Irf8^+/+, Irf8^+/−, and Irf8^−/− mice (left), and bone morphometric analysis of femurs (right) (n=5 for each group). BV/TV, bone volume per tissue volume; Tb.N, trabecular number; Tb.Th, trabecular bone thickness; BMD, bone mineral density. Top, longitudinal view; middle, axial view of the cortical bone in midshaft region; bottom, axial view of the trabecular bone in metaphyseal region. Scale bar, 0.5 mm. (B) Histological analysis of femurs (12-week-old mice) and (C) mandibles (4-week-old mice) by TRAP staining demonstrate increased osteoclast activity (arrows indicate osteoclasts) in Irf8^−/− and Irf8^+/− mice when compared to Irf8^+/+ mice. Lower left panel shows quantified osteoclast numbers. Femur scale bar, 100 μm. MN scale bar 200μm and 100μm. (D) Immunohistochemical staining shows increased NFATc1 expression in the dentoalveolar region of Irf8^−/− and Irf8^+/− mice when compared to Irf8^+/+ controls (4-week-old mice). D=dentin, B=bone, PDL=periodontal ligament. Scale bar, 50 μm. ***P <0.001, **P <0.01, *P<0.05 and NS=nonsignificant (one-way ANOVA). n=5 mice/group for micro-CT analysis, and n=4 mice/group for histological analysis.

Discussion

By adopting a multi-parameter screening and phenotyping approach, this is the first study to identify a novel IRF8 mutation associated with MICRR and delineate the genome-wide effects of IRF8 on osteoclast transcriptome. Bioinformatics analysis predicted that substitution of serine for glycine at the highly conserved 388 residue alters the protein structure, likely impairing IRF8 function. Consistent with this notion, we noted hIRF8^G388S mutant promoted increased osteoclastogenesis and failed to physically interact with NFATc1 and inhibit NFATc1-dependent transcriptional activation. RNA-seq analyses revealed that hIRF8^G388S mutation promoted global expression of osteoclast-specific genes, consistent with the transcripts noted from Irf8^−/− cells, suggesting that the increased osteoclastogenesis noted in both Irf8^−/− and hIRF8^G388S cultures is a consequence of impaired IRF8 function. Further, in parallel to the heterozygous hIRF8^G388S mutation, we noted Irf8^+/− mice exhibit a gene dosage-dependent skeletal phenotype due to increased osteoclast activity, which is reported here for the first time.

IRF8, a member of the IRF family of transcription factors (IRF1–9), plays an important role in myeloid cell differentiation, immune response, bone metabolism, and transcription of type I IFN and IFN-inducible genes ^{(9,16,20,26–29)}. IRF8 deficiency in mice is characterized by absence of monocytes, DCs, NK cells, and a lack of IL-12 and IFNγ production ^(28,29). Similarly, IRF8 mutations identified to date in humans cause a spectrum of immune cell phenotypes ^(25,30–32). The homozygous K108E variant in the DNA-binding domain leads to severe immunodeficiency, characterized by complete lack of monocytes, DCs, IL-12, IFN-γ and TNF-α production ^(25,30). The heterozygous T80A variant in the key DNA-binding helix is associated with milder immunodeficiency and a selective depletion of CD1c⁺ compartment of CD11c⁺ circulating DCs ⁽²⁵⁾. The compound heterozygous mutations R83C/R291Q lead to complex immunodeficiency syndrome caused by DC and monocyte deficiency ⁽³¹⁾. The biallelic A201V/P224L mutation within the IAD domain leads to NK cell deficiency with decreased NK cell number and CD56^dim subset, which is also observed in Irf8^−/− mice but not in Irf8^+/− mice ⁽³²⁾. In contrast to the K108E mutation, the A201V/P224L variant is able to normally activate IRF1- and PU.1-dependent transcription, which is manifested by a subtle DC deficiency and normal production of IFN- γ and TNF-α. Collectively, these results underscore the varying impact of different IRF8 mutations on immune cells. However, no detailed dental or skeletal phenotypes have been reported in patients harboring these mutations except for a history of oral candidiasis and bone marrow histology showing normal osteoclast activity in the 10-week-old infant carrying K108E variant ⁽²⁵⁾.

In contrast to these previously reported mutations, the heterozygous G388S variant promoted increased osteoclastogenesis with minimal impact on macrophage gene signatures. Our results suggest that the G388S variant has a more profound impact on osteoclast differentiation. The mechanisms underlying the differential impact of IRF8 mutations on immune cells vs. osteoclasts remain unclear. By itself, IRF8 possesses weak transcriptional activity. However, it can function both as a transcriptional activator and repressor by forming different DNA-binding heterocomplexes with multiple partners, including IRF, ETS and NFAT family members ^(11–14). It is plausible that variants in DNA-binding domain vs. activation domain may differentially affect IRF8 heterodimerization with other partners, leading to differing phenotypes. Several reports indicate that effects of gene dosage and site-specific mutations are frequently observed in lineage-determining transcription factors regulated by super-enhancer structures, such as the IRF8 ⁽³³⁾. Indeed, we noted gene dosage-dependent effect on Irf8^+/− vs. Irf8^−/− mouse skeletal phenotypes (Fig. 6). Further in agreement, genome wide association studies have identified IRF8-associated polymorphisms as risk factors for several inflammatory diseases including systemic lupus erythematosus ⁽³⁴⁾, and multiple sclerosis ⁽³⁵⁾. However, to date, no studies have implicated IRF8 mutations in periodontal disease susceptibility.

Comparative studies in Irf8^−/− and hIRF8^G388S cells offered us the opportunity to identify IRF8 transcriptional targets in osteoclastogenesis and gene expression changes associated with IRF8 deficiency. While RANKL promoted osteoclast differentiation in all groups, the differences were pronounced in Irf8^−/− and hIRF8^G388S cultures when compared to their respective WT groups (Figs. 2, 3, and 4). Some of the differentially regulated genes in both Irf8^−/− cells and hIRF8^G388S cells included positive and negative regulators of osteoclastogenesis. In agreement with previous studies ^(15,36), we noted metabolites involved in glycolysis, TCA cycle, oxidative phosphorylation, and electron transport were significantly enriched in RANKL-stimulated Irf8^−/− and hIRF8^G388S BMMs when compared with their respective WT cells. Furthermore, consistent with previously established osteoclast signaling pathways ⁽³⁶⁾, molecules related to RANK-RANKL, NF-kB, PI3K-Akt-mTOR, MAPK and TGFβ signaling were induced during the differentiation process (Fig. S11), and more noticeably in RANKL-stimulated Irf8^−/− and hIRF8^G388S cells. We noticed that IRF8 constitutively bound to regulatory regions of several thousand genes in osteoclast precursors (7,887, Table S4). The loss of IRF8 binding with RANKL treatment or genetic aberration leads to enhanced expression of genes (e.g., Nftac1, Acp5, Oscar) important for osteoclast regulation.

Similar to IRF8, IRF1 functions as a negative regulator of osteoclastogenesis ⁽¹⁷⁾. In this study, we noted IRF1 and IRF7 expression levels were significantly downregulated during osteoclast differentiation process and more profoundly in Irf8^−/− vs. Irf8^+/+ cells. Considering the critical and complimentary roles played by IRF family members in myeloid cell differentiation and chronic inflammatory diseases, further studies are needed to investigate the role of other IRF family members in bone metabolism. Among the list of negative regulators of osteoclastogenesis, RBP-J has been shown to function upstream of IRF8 and augment IRF8 expression ⁽³⁷⁾. In this study, we noted that RBP-J expression was initially higher in Irf8^−/− vs. Irf8^+/+ BMMs at D0 (mean RPKM, 67 vs. 37), however, its expression was downregulated with RANKL treatment and more significantly in Irf8^−/− vs. Irf8^+/+ cells at D4 (9 vs. 15) and D8 (10 vs. 27). We further noted that RBP-J has multiple IRF8 binding sites, indicating that RBP-J is a direct transcriptional target of IRF8 in BMMs. Collectively, these results suggest that IRF8 and RBP-J may regulate each other, and there may be other complex mechanisms involved during the osteoclast differentiation process that regulate both of these transcription factors. Furthermore, Nishikawa et al., reported that DNA methylation by Dnmt3a regulates osteoclastogenesis via epigenetic repression of Irf8, and Irf8 acts downstream of Dnmt3a ⁽¹⁵⁾. Consistent with this finding, we noted Dnmt3a expression was increased during osteoclast differentiation process and was more distinct in Irf8^−/− (D4) and hIRF8^G388S (D8) cells when compared to their respective WT groups. Furthermore, our results indicate that Dnmt3a has multiple IRF8 binding sites. The expression of Dnmt3a in BMMs may be controlled by IRF8 binding and IRF8 silencing during osteoclastogenesis may allow upregulation of Dnmt3a. A proposed signaling cascade involved in RANKL-RANK induced osteoclastogenesis is illustrated in Fig. 7.

Fig. 7. Signaling cascade in RANKL-RANK induced osteoclastogenesis: A model of IRF8-mediated regulation of osteoclast differentiation.

RANKL binds to its receptor RANK, which results in the recruitment of TRAF6 and activation of NF-κB and MAPK pathways. NF-κB pathway induces c-Fos expression via IKKs. MAPK pathway results in the activation of the Jun proteins. Together, the c-Fos and Jun proteins associate to form the AP-1 complex. AP-1 along with NF-κB, NFATc2, and NFATc1 itself drives the expression of NFATc1, the master regulator of osteoclastogenesis. Activated NFATc1 binds to NFAT-binding sites on its own promoter, stimulating its own expression and leading to autoamplification loop. NFATc1 cooperates with other transcription factors, such as AP-1, PU.1, CREB, and MITF to promote the expression of various osteoclastogenic genes. On the other hand, NFATc1 expression is negatively regulated by transcription factors, such as IRF8, MafB, Bcl6, and LRF, some of which are inhibited by Dnmt3a, RBP-J, and Blimp1. IRF8 in turn may control these inhibitors.

Importantly, the cohort described in this study provided us a unique opportunity to investigate a novel mutation in IRF8 and longitudinally study the pathogenesis of MICRR. Our in vitro functional data support the hypothesis that loss of IRF8 function in patients may lead to increased osteoclast activity and subsequent development of MICRR. While affected subjects presented extensive osteoclast-mediated cervical root resorption, they reported no other systemic bone disorders. The late onset of disease, including lack of recognized effects on skeleton, could be in part due to the heterozygosity of the mutation and/or other compensatory actions such as increased bone formation to offset resorption. In agreement with this concept, enhanced bone turnover and remodeling is observed in Irf8^−/− mice due to accelerated rate of bone resorption accompanied by increased rate of bone formation ⁽¹⁶⁾. In addition to its effect on osteoclasts, IRF8 is known to modulate TLR signaling and contribute to the cross-talk between IFN-γ and TLR signal pathways ⁽²¹⁾. With age, the junctional epithelium migrates apically creating a deeper periodontal pocket or sulcus, exposing tooth root cementum to bacterial biofilm that accumulates in the sulcus. Thus, loss of IRF8 function would not only activate osteoclastogenesis but may also fail to regulate TLR-induced inflammatory destruction in response to bacterial LPS ⁽¹⁶⁾. Therefore, age, subgingival plaque, IRF8 dysfunction, and other factors such as occlusal trauma or parafunctional habits may collectively contribute to the development of MICRR. This notion further helps explain why root resorption was not observed in Irf8^−/− and Irf8^+/− mice, despite documenting increased osteoclast activity in the alveolar bone. In addition, the fact that MICRR is not always noted in patients with overactive osteoclast-related disorders, further supports that MICRR is a multifactorial disease and involves etiological factors specific to the oral cavity.

Empirical observations suggest that MICRR may not be a monogenic disease, and other regulators upstream or downstream of IRF8 pathway (e.g., NFATc1 gain-of-function mutation) could similarly contribute to root resorption. Inexplicably, the G388S variant was not detected in the fourth affected individual (II:4), which could be related to mosaicism. Due to lack of access to peripheral blood, genetic sequencing was performed using genomic DNA extracted from saliva. Saliva primarily contains leukocytes and buccal epithelial cells. If the relative proportion of mutant carrying somatic cells differ or are lower, the ability to detect the mutation would be substantially reduced. Inter-tissue variation of mutant cell frequencies has been observed previously ⁽³⁸⁾. One example is Proteus syndrome, characterized by overgrowth of bones and skin, in which mosaic mutations of the AKT1 gene are found in multiple tissues but rarely in hematopoietic cells ⁽³⁹⁾. Another example of an autosomal dominant disorder demonstrating somatic mosaicism is segmental Neurofibromatosis Type 1 (NF1) ⁽⁴⁰⁾. In patients with mosaicism, the proportion of cells with NF1 micro deletions vary between peripheral leukocytes and buccal smears or peripheral skin fibroblasts (40).

The lack of immunophenotyping data and detailed characterization of skeletal phenotype in MICRR patients is a major limitation of this study. To overcome this limitation, IRF8^G388S knock-in mice are being generated. These knock-in mice will help in studying the effects of IRF8^G388S mutation on osteoclast regulation, tooth root resorption, as well as its effect on other immune cells (e.g., DCs, NK cells). In conclusion, this is the first study to report a novel mutation associated with MICRR and illustrate global gene expression changes induced by IRF8 deficiency during osteoclast differentiation process. Our results support the growing recognition that transcription factors important for immune response may also play a role in osteoclast regulation. Further studies are needed to identify metabolic programs and transcription factors that regulate epigenetic changes and osteoclast-specific enhancers, which will help in developing prevention and treatment strategies for various bone disorders including periodontal disease.