Chronic Non-bacterial Osteomyelitis and Autoinflammatory Bone Diseases

Introduction

Chronic non-bacterial osteomyelitis (CNO), also known as chronic recurrent multifocal osteomyelitis (CRMO), is an autoinflammatory bone disease that mainly affects children with an average age of onset around 9 years of age [1–5]. The female to male ratio is about 2:1 [1–5]. The nomenclature can be confusing as there have been multiple names for sterile osteomyelitis syndromes including SAPHO (synovitis, acne, pustulosis, hyperostosis and osteitis) syndrome (which is the term used most often by adult rheumatologists), diffuse sclerosing osteomyelitis of mandible (which is used more by the oral surgeons and dentists), and nonbacterial osteomyelitis (NBO) among many others used in the literature. In this review, we will use the term CNO.

Genetic and immunologic basis of CNO

Evidence supports that sterile osteomyelitis can have a genetic component. There are reports of familial clustering of CRMO, affected siblings, and affected parent-child duos. Many individuals with CNO have a personal or family history of psoriasis, inflammatory bowel disease or inflammatory arthritis; occurring in up to 50% of their close relatives [2,6]. In addition, mice, dogs (particularly Weimaraners) and non-human primates can also develop sterile multifocal osteomyelitis [7–10] that may share common mechanisms as humans. Genetic analysis of syndromic forms of human CNO and murine models of the disease have allowed the identification of multiple susceptibility genes for sterile osteomyelitis [8,11–15].

Majeed syndrome was descried in 1989 as an autosomal recessive syndrome that presents with early-onset CNO, dyserythropoietic anemia with or without a neutrophilic dermatosis [16]. In 2005, mutations in LPIN2 were identified using homozygosity mapping in two affected families by Majeed syndrome, which was subsequently confirmed in additional families [11,17]. Treatment of Majeed syndrome with cytokine blockers demonstrates the importance of IL-1β in disease pathogenesis as the associated osteomyelitis and systemic inflammation improves with IL-1β, but not TNF, blocking agents [18–20]. The anemia also improves by IL-1β blocking agent but it is not clear if the improvement is due to control of chronic inflammation or the reversal of the dyserythropoiesis. The latter has yet to be confirmed because repeat bone marrow biopsies have not been clinically indicated after treatment due to overall clinical improvement.

Research at the bench and bedside has confirmed the critical role of IL-1β and places Majeed syndrome in the category of a Nlrp3 inflammasomopathy [18,21]. Balboa’s group demonstrated that LIPIN2 deficiency or under expression results in reduced intracellular cholesterol, aberrant P2X7 receptor function in LPS + ATP stimulated macrophages accompanied by downstream activation of the Nlrp3 inflammasome and increased production of IL-1β [21]. Interestingly, this effect on the receptor and inflammasome activation can be reversed with increasing intracellular cholesterol [21].

IL-1 is also a critical cytokine in the pathogenesis of sterile bone inflammation in Pstpip2 deficient mice and in a newly recognized human autoinflammatory disease named deficiency of the IL-1 receptor antagonist (DIRA) [13,22,23]. DIRA is autosomal recessive disorder caused by mutations in IL1RN that presents in infancy with multifocal sterile osteitis, periostitis, pustulosis and systemic inflammatory response syndrome [13,24]. Most cases have homozygous truncation mutations, but gene deletion and compound heterozygous mutations have also been reported [25–28]. IL1RN encodes the IL-1 receptor antagonist and deficiency of this protein results in unfettered IL-1 receptor signaling [13]. Untreated DIRA has a high mortality, however, recognition of DIRA as a distinct clinical entity has resulted in the ability to diagnose this disease via DNA sequencing and to effectively treat these infants by replacing the very protein they are missing in the form of recombinant IL-1 receptor antagonist [13,24].

Using whole exome sequencing in a consanguineous family from a child with CNO, a homozygous mutation in the FBLIM1 gene was identified [12]. FBLIM1 encodes the protein Filamin-Binding LIM Protein-1 (FBLP-1) which has also been shown to be important in bone remodeling and is involved in cytoskeletal regulation [29]. In one mouse model of Fblim1 deficiency there is increased osteoclast activation and severe osteopenia [30]. Fblim1 gene expression is significantly dysregulated in bone marrow-derived macrophages from Pstpip2 mice further implicating its role in the development of sterile osteomyelitis [12]. Fblim1 is also involved in IL-10 mediated anti-inflammatory responses in macrophages as it can be regulated by STAT3 which is in keeping with its role as an anti-inflammatory protein [31]. Another unrelated individual with CNO had a deleterious mutation on one allele and a promoter polymorphism that affected gene expression in trans implicating a potential role of FBLIM1 in some sporadic CNO [12].

There are several animal models of CNO. Hypertrophic osteodystrophy (HOD) is a canine disease that predominantly affects the Weimaraner bred. Affected dogs develop multifocal sterile bone inflammation which can be accompanied by pustulosis or Crohn disease [7]. HOD clusters in litters supporting a genetic cause but a causative gene remains elusive. There are reports of non-human primates developing CNO but no genes have been identified as causative [10]. There are several mouse models of CNO and due to the ability to selectively breed mice, gene identification has been possible. The first murine model of CNO occurred as the result of a spontaneous mutation which when bred to homozygosity produced a mouse with tail kinks [8,32]. The chronic multifocal osteomyelitis (cmo) phenotype is due to homozygous (o.Ser734Pro) mutation in the proline, serine, threonine, phosphatase interacting protein 2 (Pstpip2) gene resulting in a lack of detectable protein [11,15]. Subsequently, several other Pstpip2 deficient strains of mice have been developed either through ENU mutagenesis or via gene knockout which all have a phenotype similar to the cmo mouse [15,33].

Experiments in cmo mice with additional inflammatory genes knocked out have demonstrated that disease in the cmo mouse is IL-1 dependent as cmo mice lacking an IL-1 receptor are completely protected and that IL-1β rather than IL-1α is driving the disease [22,23]. Further, cmo mice lacking either Nlrp3, ASC continue to develop the disease demonstrating that it can occur as an Nlrp3 independent process [22,23]. While caspase–1 and −8 may play redundant roles in disease [34]. Cellular studies demonstrate that neutrophils rather than macrophages predominantly hypersecrete IL-1 [22,34]. In addition to neutrophils, mast cells also play a role in disease in the cmo mouse [35]. Mast cells are present in inflamed bone in cmo mice and Mcpt1 (a mast cell protease) can be detected in their peripheral blood [35]. Cmo mice that also lack mast cells have attenuated disease. In human CNO, mast cell chymase is elevated in the peripheral blood of children with CNO and granulocytes including mast cells and neutrophils can be present in bone biopsies [35].

Another murine model of sterile multifocal osteomyelitis is the Ali18 mouse which was identified in an ENU mutagenesis screen. This mouse develops inflammation in the bones, skin and joints [36–38]. The inflammatory phenotype is due to a gain-of-function mutation in the feline Gardner-Rasheed feline sarcoma viral (V-Fgr) oncogene homolog (Fgr), a member of the Src family kinases [14]. In addition to the identification of Fgr as a murine CNO susceptibility gene, whole exome sequencing of humans with CNO (including 88 trios [child and both parents]) identified 13 patients with heterozygous protein coding variants, 2 of which (one de novo) that alter FGR kinase activity [14] supporting a role for FGR in human CNO.

Although FBLIM1 and FGR have been identified as CNO susceptibility genes in humans, mutations in these account for only a small proportion of individuals with non-syndromic CNO demonstrating the complexity of CNO genetics. And despite Pstpip2 deficient mice being very helpful in dissecting the pathogenesis of sterile osteomyelitis, no human individuals with PSTPIP2 deficiency have been identified. Golla et al. identified a susceptibility locus on chromosome 18q21.3–18q22 in a small cohort of sporadic CNO, however, no linkage to this region was present in a larger cohort genotyped with more sensitive molecular techniques [11,12].

What has been learned so far, is that innate immune system dysfunction is responsible for the chronic bone and skin inflammation seen in these disorders. Dysregulation of the IL-1 pathway is etiologic in DIRA, Majeed syndrome and murine cmo [13,18,21–23]. Disease manifestations in DIRA and Majeed improve dramatically with IL-1 blockade, implicating this cytokine in the pathogenesis in at least a subset of CNO [13,18]. The IL-1 pathway is also implicated in human CNO including the detection of increased expression of inflammasome components NLRP3 and ASC in CNO human monocytes, and site-specific DNA hypomethylation around the NLRP3 and PYCARD genes has been found [39].

There is additional evidence for immune dysregulation in sporadic forms of human CNO. Decreased IL-10 secretion from LPS-stimulated CNO monocytes accompanied by attenuated extracellular-signal regulated kinase (ERK)1/2 activity has been demonstrated in vitro [40]. Further, attenuation of Sp1 and reduced histone H3 serine-10 phosphorylation was also found, suggesting that epigenetic factors play a role in the decreased gene expression of IL-10, which disrupts the pro- and anti- inflammatory cytokine balance, resulting in increased TNF-α and IL-6 production [40–42]. There is no diagnostic test for CNO beyond DNA sequencing for the few who have mutations in CNO susceptibility genes. The identification of serum or urine biomarkers are needed to facilitate diagnosis and for disease monitoring. Serum IL-6 and CCL11/eotaxin have been identified as potentially useful biomarkers and are currently under further study [43]. Urine bone resorption marker urine N-telopeptide/urine creatinine has also been explored as a biomarker for disease activity in CNO but additional studies are needed in a larger cohort to determine its utility [44].

Clinical features

Most children experienced focal bone pain at the affected site(s) but some lesions are asymptomatic. Local swelling, warmth, and tenderness to palpation may be present. Other signs may include limp or difficulty walking if CNO affects the lower extremities. Jaw opening may be decreased if there is mandibular involvement. Urination and defecation could be affected with lesions that affect the pelvis. When the spine is involved, back pain can be vague or severe [45]. Nocturnal pain is common and may be misdiagnosed as growing pain in school-aged children; asymmetric pain, limp, objective swelling or limitation of joint range of motion in an otherwise well-appearing child should raise concern for CNO. The diagnosis is one of exclusion. Almost all the bones can be affected but most patients have lesions in the metaphyseal regions of the long bones of the lower extremities [5,46–49].

Unifocal vs. multifocal

While many patients present with only one symptomatic site, the majority of patients have multifocal lesions when assessed by whole body MRI (WBMRI). The two largest cohort studies reported that 30% of children with CNO have had a single lesion however, only 30% had whole body imaging [1,4]. Yet when followed overtime (up to 4 years of follow up), 93% of all children in one of these large cohort study developed multifocal lesions [1]. For those remaining unifocal, the clavicle is the most common site. In contrast, when WBMRI imaging is employed at initial diagnosis, only 1 out of 53 children had a single lesion whereas the remaining 52 children (98%) had 2 or more lesions [45]. Another cohort study reported that 33 of 37 (89%) children with CNO had multifocal involvement [50]. These results suggested that the multifocal cases are underestimated.

Andronikou et al. reported two major lesion distribution patterns based on data from WBMRI of 37 children with CNO [50]. The “tibio-appendicular multi-focal pattern” referred to the presence of tibial lesions (66% had symmetrical tibial lesions), which comprised of 54% of the cohort. The “claviculo-spinal pauci-focal pattern” referred to the presence of clavicular lesions, no tibial involvement in the presence or absence of lesions in the spine, which comprised of 24% of the cohort. However, these classification groups are not exclusive as an additional 14 % of individuals with clavicle involvement also had involvement of the tibia. Given the multifocality of the majority of patients with CNO, WBMRI at baseline is important to determine the extent of the disease and can be useful to guide medical decisions.

Recurrent symptoms

The majority of children with CNO suffer from one or more recurrence during their disease course [3,44,51–53]. However, in a recent study where almost 70% of children were treated with bisphosphonates, only 32% were classified as recurrent disease at the time of follow-up [54]. During a median of 6 years of observation, 57% had a painful flare, which suggested the recurrent nature of this disease in most children and underscored the importance of long-term follow up and imaging monitoring [5].

Associated conditions

There is a strong association between CNO and inflammatory bowel disease (IBD), psoriasis (both palmoplantar pustulosis [PPP] and psoriasis vulgaris), severe acne and inflammatory arthritis [1–5]. In one cohort, when any autoimmunity/autoinflammatory condition was considered, 50% of children with CNO had at least one inflammatory comorbidity [3]. From an online survey to an international parent community, 11% reported acne, 8–10% psoriasis, or IBD or PPP, 3–6% various types of inflammatory arthritis [55]. The genetic connection between CNO and associated diseases is currently being studied.

Prognosis

The long-term outcome of CNO has been reported in multiple cohorts and varies widely from 0% with active disease to 100% with active disease (follow up > 5 years for most studies) [51,56–61]. In one group (n=17) followed for a median of 12 years from the initial diagnosis [62], 50% were still on medications for CNO and 58% had active CNO lesions on WBMRI; however, selection bias may have contributed to the higher rate of active disease in this cohort. Remission off of medications has been difficult to accomplish with a rate of 13% in one North American cohort (n=70) [3] and 23% in another (n=171) [1]. Whether aggressive therapy at the initial diagnosis leads to long-term remission has not been studied. Complications such as bone deformity, vertebral fracture, leg length discrepancy, joint angulation, limb hypertrophy, and generalized growth failure can be seen as many as 20–26% of patients [1,63].

Diagnostic workup

The diagnosis of CNO is made by exclusion though several diagnostic criteria have been proposed [2,64,65]. The common features of these criteria include the presence of bone pain, typical imaging findings of CNO, multifocal pattern, or if unifocal the requirement for negative bone biopsy for infection and malignancy. These criteria have not been validated and no consensus has been developed among treating physicians. Currently an international workgroup aims to develop classification criteria for future research purposes [66].

Lab findings

Most patients have normal blood cell counts. About 40–82% patients with CNO were reported to have elevated CRP and/or ESR [1–3,5,51]. Majority of patients have negative ANA and negative HLA-B27 [1,4]. Other labs that can be useful to exclude CNO mimicks including LDH, uric acid, vitamin C, and alkaline phosphatase. Serum cytokines have been proposed as biomarkers to distinguish children with CNO from healthy children and those with other autoinflammatory diseases [67] but these results have not been validated.

Imaging findings

All children with suspected CNO should obtain a radiograph first and proceed with MRI if concerning for CNO based on abnormal findings (with the caveat that 30 % may have a normal plain film and still have CNO). Children with a normal radiograph but having persistent symptoms need further imaging such as MRI. Typical radiographic findings of CNO are mixed lytic and sclerotic lesions in common sites without concerning findings such as disorganized bone formation or bone destruction [46,47]. The sensitivity of radiographic imaging in CNO is as low as 13% when compared to MRI in a small cohort (n=13) [48]. However, clinicians must recognize that the sensitivity of plain radiographs in CNO varies from 31% (n = 486) [4] to 77% (n = 70) [3]. Other imaging studies such as regional MRI and bone scintigraphy are often used and provide additional information with the latter as a whole-body imaging modality [46,47,68]. Bone scintigraphy is considerably less sensitive than MRI [69] due to the predominant pattern of symmetrical CNO lesions near the physis. In addition, radiation from bone scintigraphy makes it less desirable as a repeated monitoring tool. At initial diagnostic stage, MRI with contrast may be needed when there are serious concerns of infection or malignancy. Otherwise, MRI without contrast focusing on short tau inverse recovery (STIR) sequence is most commonly used for screening of CNO [70]. Diffusion weighted imaging has been proposed to differentiate CNO from other mimicker diseases but needs further investigation [71]. The proposed definition of a “CNO lesion” on MRI is a lesion within bone marrow with increased signal intensity (SI) on STIR images and decreased SI lower than muscle on T1-weighted images [70,72]. From STIR sequence, hyperintensity within bone marrow is considered as signs of an active CNO lesion. Soft tissue inflammation, periosteal reaction and hyperostosis can be assessed as well as permanent damages including growth plate damage and vertebral body compression. WBMRI can be done within 40 minutes in most children. Typical WB MRI protocols include coronal images of the entire body, and sagittal images of entire spine, axial sequences of the pelvis and knees, and sagittal images of ankles and feet, acquired with STIR sequences without the need for contrast [46,72]. Scoring of disease activity based on MRI has been explored in 3 different studies [70,72,73]. Zhao et al. and Arnoldi et al. have proposed similar but slightly different approaches in determining the disease activity based on MRI findings [70,73]. Both identified key characteristics including bone edema, extramedullary inflammation (soft tissue edema and periosteal reaction), hyperostosis, presence of spinal lesion(s), and vertebral body deformation. The size of lesions was described in relative proportion by one approach [70] whereas in absolute measurements by the other [73]. A radiologic index for non-bacterial osteitis (RINBO) was proposed that gives a score which includes active disease and chronic damage and was found to correlate with clinical active disease. A new tool, ChRonic nonbacterial Osteomyelitis MRI Scoring (CROMRIS), was recently generated to report active disease and chronic damage in a systemic approach based on consensus among 11 radiologists with experience in CNO [72]. The interrater reliability was excellent in that the mean kappa of each category of bones was >0.7 with majority >0.9 among radiologists. All these scoring systems need to be validated for its sensitivity to change and validity on different longitudinal cohorts.

Bone biopsy

Bone biopsy is not always needed for diagnosis of CNO. A clinical scoring system has been developed to help clinicians in determining if a biopsy is needed [74]. Roderick et al. reported that 34 of 41 patients (83%) could have been spared from a bone biopsy by applying their proposed diagnostic criteria [65]. However, excluding alternative diagnoses, such as histocytic disorders and malignancy, is important as there is no diagnostic test for CNO. Studies investigating the rate of misdiagnosis without a biopsy in cases of presumed CNO are lacking. Based on a survey of pediatric rheumatologists, a bone biopsy is more likely to be obtained when there are constitutional symptoms, unifocal lesion or nocturnal bone pain whereas less likely obtained when there are lesions at typical CNO sites, multifocal lesions or frequently associated conditions [68]. A collaborative approach with other sub-specialists including infectious disease, oncology, orthopedics and intervention radiology is needed. When biopsy is needed, strict sterile technique should be used throughout the chain of specimen acquisition and processing to minimize the risk of contamination. For instance, mandibular lesions can be biopsied through external route [75] or intraoral but along with a piece of adjacent tissue. Pathological findings are not specific though the presence of plasma cells and lymphocytes with or without fibrosis are commonly reported [2,76].

Treatments

First-line treatment is non-steroidal anti-inflammatory drugs (NSAIDs) [1–5,68] yet there is only one prospective NSAID study performed by Beck et al. on a German cohort of 37 individuals with CNO [53]. After 12 months of treatment, 23 of 37 patients (64%) were pain-free, whereas only 10 patients (27%) were free of active lesions on MRI [53]. Surprisingly, among the nine patients with radiological lesions in the spine, seven patients were free of active radiological spine lesions and two remained unchanged after 12 months though it was unclear how many received additional treatments including sulfasalazine and glucocorticoids. Arthritis was present in 14 of 37 patients (38%) and remained active in all at 3 months, 50% at 6 months, and 21% at 12 months.

Treatment strategies for NSAID failures varies among treating physicians. Additional therapies that have been utilized in CNO include disease modifying anti-rheumatic drug (DMARDs), biologics DMARDs (e.g. anti-TNF, anti-IL-1) or bisphosphonates [77]. In the absence of prospective data to guide the choice of second-line therapy, treatment is based on treatment responses reported from retrospective case reports, case series, cohort studies and expert opinion. In general, spinal disease is treated with pamidronate due to its effect on osteoclasts and the subsequent shift toward improving bone density in order to minimized compression deformity [78–81]. Treatment with TNF inhibitors is often chosen for patients with co-morbid disorders such as psoriasis, inflammatory bowel disease or inflammatory arthritis. Side effects are possible with each of these treatment options but there is limited data on the likelihood of experiencing a side effect with each approach. The development of severe psoriasis in individuals with CNO treated with TNF inhibitors has been reported and warrants additional study [82].

Currently, the CRMO/CNO workgroup within Childhood Arthritis and Rheumatology Research Alliance (CARRA) is conducting a prospective observational study to determine the effectiveness of treatments based on developed consensus treatment plans (CTP) from CHronic nonbacterial Osteomyelitis International Registry (CHOIR). Despite recent advances, there is no information on optimal duration of treatment. Further studies are needed to determine if a personalized medicine approach to the management of CNO could improve outcomes, guide therapy and prevent medication side effects.

Conclusion

While significant progress has been made in understanding autoinflammatory bone disorders, more work remains to be done. Challenges arise from the phenotypic and genetic heterogeneity which may be reflected in variation in response to treatments. Development of specific diagnostic and monitoring test(s) for sporadic forms of CNO are needed to improve time to diagnosis and to guide controlled clinical trials.

• Sterile osteomyelitis may be accompanied by psoriasis, inflammatory bowel disease and inflammatory arthritis

• CNO is a complex disorder with genetic, epigenetic and environmental components

• Dysregulated IL-1β signaling is present in syndromic forms of CNO

• Treatment is with anti-inflammatory agents or a bisphosphonate but best treatment(s) remains to be defined