A Clinical and Molecular Overview of the Human Osteopetroses
- First Online:
- Cite this article as:
- Balemans, W., Van Wesenbeeck, L. & Van Hul, W. Calcif Tissue Int (2005) 77: 263. doi:10.1007/s00223-005-0027-6
- 272 Downloads
The osteopetroses are a heterogeneous group of bone remodeling disorders characterized by an increase in bone density due to a defect in osteoclastic bone resorption. In humans, several types can be distinguished and a classification has been made based on their mode of inheritance, age of onset, severity, and associated clinical symptoms. The best-known forms of osteopetrosis are the malignant and intermediate autosomal recessive forms and the milder autosomal dominant subtypes. In addition to these forms, a restricted number of cases have been reported in which additional clinical features unrelated to the increased bone mass occur. During the last years, molecular genetic studies have resulted in the identification of several disease-causing gene mutations. Thus far, all genes associated with a human osteopetrosis encode proteins that participate in the functioning of the differentiated osteoclast. This contributed substantially to the understanding of osteoclast functioning and the pathogenesis of the human osteopetroses and will provide deeper insights into the molecular pathways involved in other bone pathologies, including osteoporosis.
KeywordsOsteoclast Bone resorption Osteopetrosis Human
In humans, the osteopetroses include a heterogeneous group of bone disorders characterized by increased bone density. The different subforms are classified on the basis of inheritance, age of onset, severity, and secondary clinical features into three major groups: autosomal recessive infantile malignant osteopetrosis, autosomal recessive intermediate mild osteopetrosis, and autosomal dominant adult onset benign osteopetrosis . In these forms, all clinical symptoms are caused by or are secondary to the defect in bone resorption and the consequent increase in bone mass. In addition, some families and isolated cases have been described in which osteopetrosis occurs in association with clinical symptoms unrelated to the bone-resorption defect. Genetic studies already substantially contributed to the understanding of the mechanisms of osteoclastic bone resorption by the identification of a number of genes harboring disease-causing mutations in several types of osteopetrosis.
The purpose of this review is to summarize the clinical and molecular findings of the human osteopetroses. The major part will focus on those forms of osteopetrosis in which only bone resorption is impaired. In a separate section, an overview will be given of human osteopetrotic conditions with additional clinical features unrelated to an impaired bone resorption.
Human Osteopetrotic Conditions in Which Only Bone Resorption is Impaired
Malignant Infantile Autosomal Recessive Osteopetrosis
Overview of the major forms of human osteopetrosis with radiologic and clinical characteristics, primary defect and molecular mechanism of the mutations
Type of osteopetrosis
Dense sclerotic bones, fractures, neurologic symptoms,
loss of function
bone marrow failure, infections, and early death
loss of function
loss of function
Mild osteopetrosis, short stature, fractures
Partial loss of function
Benign type Ib
Diffuse osteosclerosis, no fractures
Gain of function
Benign type II
Rugger-Jersey spine, bone-in-bone appearance, fractures
Dominant negative effect
Benign type III
Sclerosis of the distal skeleton and the skull
Severe osteopetrosis, reduced number of osteoclasts
Asymptomatic osteopetrosis, normal osteoclasts, and spontaneous resolution of radiographic abnormalities
Osteopetrosis with renal tubular acidosis
Osteopetrosis, short stature, mental retardation, renal tubular acidosis, fractures
Loss of function
Osteopetrosis with infantile neuroaxonal dystrophy
Osteopetrosis, neuroaxonal spheroids, early death
Anhidrotic ectodermal dysplasia with immunodeficiency, osteopetrosis, and lymphedema
Osteopetrosis, lymphedema, severe infections, no teeth, skin abnormalities, early death
Homozygous and compound heterozygous null mutations in the Clcn7 gene encoding chloride channel 7 (CLCN7) are a secondary, although less frequent (∼10% to 15%), cause of malignant ARO [20, 21, 22, 23]. CLCN7 resides in the ruffled border membrane of the osteoclast, where it has been suggested to act as a chloride conductor allowing the V-ATPase to acidify the extracellular resorption lacuna . In vivo knockout studies in mice indeed demonstrated that osteoclasts of Clcn7-deficient mice had lost their ability for extracellular acidification . Although in vitro studies on CD14 cells derived from nine patients with a heterozygous G215R mutation in Clcn7 causing autosomal dominant type II osteopetrosis (see subsequent text) supported this finding , similar studies in one patient with ARO who carried compound heterozygous Clcn7 mutations failed to demonstrate a defect in acid secretion . The latter result debates that CLCN7 acts in parallel with V-ATPase, but rather exerts its function via a different mechanism. The clinical spectrum of patients harboring Clcn7 mutations is broader than that of patients with TCIRG1 mutations. Kornak and colleagues  suggested that primary retinal degradation is specific to Clcn7 mutations, whereas only a secondary eye defect caused by narrowing of the optical foramina may be observed in patients with TCIRG1 mutations. In a substantial group of ARO patients described by Frattini and colleagues  carrying compound heterozygous Clcn7 mutations, central nervous system abnormalities were documented. A recent study on Clcn7 knockout mice demonstrated severe osteopetrosis and, additionally, a severe lysosomal storage phenotype associated with neurodegeneration and retinal degeneration. Histologic analysis showed degeneration of the hippocampal CA3 region with a reduction in cell density, and at later stages of life, neuronal loss was also observed in the cortex. Retinal cell death was evident within the first 2 to 4 weeks after birth. Knockout mice had a lifespan of approximately 40 days. Crosses of Clcn7 knockout mice with transgenic mice expressing Clcn7 under the control of the tartrate-resistant acid phosphatase (TRAP) promoter, which is nearly exclusively expressed in osteoclast and macrophages, revealed a rescue of the osteopetrosis phenotype; however, neurodegeneration and retinal cell death were unchanged. These data suggest that the neuronal and retinal phenotype is directly linked to the lack of CLCN7 and, therefore, not a secondary cause of the osteopetrosis phenotype .
A third gene recently discovered to be involved in malignant ARO is the grey lethal (GL) gene as it was initially found to be mutated in the grey lethal osteopetrotic mouse . This gene is currently named osteopetrosis associated transmembrane protein 1 (OSTM1) gene or GAIP interacting protein N terminus (GIPN) gene. Protein predictions revealed a single transmembrane domain and the protein has been localized in the cytoplasm where it is most likely attached to intracellular membranes. The exact role of the OSTM1 protein in bone resorption still needs to be established but functional studies already demonstrated expression in osteoclasts  and suggested a role for the protein in osteoclast maturation/functional activity, and more precisely, in cytoskeletal reorganization and development of the ruffled border . The rat ortholog has been shown to act as a E3 ubiquitin ligase suggesting a role for OSTM1 in the protein degradation machinery . Thus far, two homozygous OSTM1 mutations have been described in patients with an extremely severe form of ARO [26, 29, 30].
Currently, not all malignant ARO cases could be explained by mutations in any of the above-mentioned genes highlighting the existence of at least one other gene involved in this type of osteopetrosis.
In vitro studies of osteoclasts from ARO patients have already provided preliminary insights in some cellular abnormalities of ARO. Teti et al.  demonstrated that osteoclast-like cells from an osteopetrotic patient showed an altered cytoskeletal organization and adhesion pattern as compared to normal osteoclasts . In another study where osteoclasts are differentiated from peripheral blood of four patients with ARO, a decrease in the expression of CD44 is observed, which is normally expressed by normal human fetal and adult osteoclasts . A third study demonstrated that in osteoclasts from three ARO patients only the polarization is impaired . Defects in cell attachment of the osteoclast-like cells have been described in one osteopetrotic patient . In conclusion, only a limited number of in vitro studies on osteoclasts from ARO patients are performed, and in the majority of these cases the underlying genetic defect was not known. This makes it currently difficult to correlate the genotype with the cellular defect.
Intermediate Autosomal Recessive Osteopetrosis
In 1923, a 43-year-old woman was described with a milder form of ARO with multiple fractures, unilateral facial palsy, dental involvement, and mild anemia . Afterward, more than 40 patients have been diagnosed with this so-called intermediate form of osteopetrosis (MIM 259710) (Table 1) [35, 36, 37, 38, 39]. Patients display with typical radiologic features of osteopetrosis with a generalized increase in bone density, “bone-in-bone” appearance, and thickening and sclerosis of the calvaria; however, less severe than observed in malignant ARO patients. Recurrent fractures were also observed in the majority of cases. Clinically, osteomyelitis, dental abnormalities and short stature are important features in this type, but mild-to-moderate anemia, extramedullary hematopoiesis, mandibular prognantism, proptosis, and deafness are also present in a significant number of patients. Intermediate ARO is clinically differentiated from the malignant ARO because the outcome is less severe and life expectancy is much higher. In 2003, two independent groups reported recessive mutations in the Clcn7 gene in patients with this type of osteopetrosis (Table 1) [22, 39]. As Campos-Xavier and colleagues described, the milder phenotype observed in their patients with intermediate ARO is most likely caused by less severe Clcn7 mutations, resulting in a milder reduction in the capacity of Cl-conductance mediated by ClCN7. However, identification of additional Clcn7 mutations in intermediate ARO patients and functional studies will be necessary to make proper genotype–phenotype correlations. Whether Clcn7 is the only gene involved in intermediate autosomal recessive osteopetrosis still needs to be established.
Autosomal Dominant Osteopetrosis
The autosomal dominant forms of osteopetrosis (ADO) have a delayed phenotype and present mainly with mild symptoms and a benign prognosis. Patients are often asymptomatic, and diagnosis is frequently made by coincidental radiographic examination. In the literature, at least three different subforms of ADO have been reported (Table 1): ADOI (MIM 607634) , ADOII (MIM 166600) , and ADOIII . The frequency of ADO has been estimated at 1:100,000 to 1:500,000 , but an extended and detailed epidemiologic study carried out in Denmark revealed a frequency of >1:20,000 . ADOI is radiologically characterized by a generalized, diffuse osteosclerosis, most pronounced in the cranial vault. This subform is not associated with an increased fracture rate and is reported to be fully penetrant . We were able to identify a low-density lipoprotein receptor–related protein 5 (LRP5) gene mutation in two Danish families diagnosed with ADOI (Table 1) [40, 44, 45]. LRP5 functions as an important mediator in osteoblastic bone formation . Gain-of-function mutations in LRP5 have been associated with “high bone density” disorders [45, 47, 48, 49], whereas loss-of-function mutations were found in patients with osteoporosis pseudoglioma syndrome (OPPG) . The molecular finding in ADOI of a LRP5 missense mutation most likely activating Wnt signaling is surprising because it suggests and increased bone formation rather than a decreased bone resorption (see discussion).
Whereas in ADOI and ADOII the axial skeleton and skull are primarily affected, another variant of autosomal dominant osteopetrosis, ADOIII or “centrifugal osteopetrosis,” has been described in a Vietnamese family. This type is characterized by sclerosis of predominantly the distal, appendicular skeleton and the skull, but with minor involvement of the axial skeleton . To date, no molecular analysis in this type of ADO has been reported.
Severe Osteoclast-Poor Osteopetrosis
To date, the genes identified to cause osteopetrosis are all involved in the function of the osteoclast. These types of osteopetrosis are characterized by numerous but dysfunctional osteoclasts. However, a few osteopetrotic cases have been reported with a markedly reduced number of osteoclasts, suggesting another variant of osteopetrosis (MIM 259720) (Table 1). Flanagan et al. (2002) described two unrelated children, diagnosed with severe osteopetrosis, in whom osteoclasts were not identified in bone marrow biopsies . In vitro studies showed that neither RANKL nor CSF1 rescued the defect in osteoclast development in cultures from these children. This is highly suggestive that defective production of any of these growth factors is not responsible for the disease .
Transient Infantile Osteopetrosis
The radiographic abnormalities of one child with asymptomatic, biopsy-proven osteopetrosis have spontaneously been resolved by the age of 28 months. The unique features of this case including the lack of characteristic physical findings, the presence of normal hematopoiesis, the appearance of normal osteoclasts on bone biopsy, and the spontaneous resolution of radiographic abnormalities soon after birth differentiate this transient osteopetrosis from the malignant form (Table 1) .
Human Osteopetrotic Conditions with Additional Clinical Features Unrelated to Impaired Bone Resorption
Osteopetrosis with Renal Tubular Acidosis
Some osteopetrotic conditions have been reported in which additional clinical features unrelated to the increased bone mass occur. One form of osteopetrosis is associated with renal tubular acidosis (MIM 259730, also known as Guibaud-Vainsel syndrome or marble brain disease) (Table 1). More than 100 families with this type of osteopetrosis are described in the literature, and most of them originate from the Mediterranean region and the Middle East . The increased bone density begins during childhood but does not lead to severe bone marrow failure. Other clinical manifestations of this syndrome include short stature, cerebral calcifications, mental retardation, dental malocclusion, and fractures (reviewed in [62, 63]). Mutations in the carbonic anhydrase II (CAII) gene, producing protons necessary for the acidic environment in the resorption lacuna of the osteoclast, have been shown to be responsible for this form of osteopetrosis (Table 1) . So far, all CAII deficient patients have mutations in the coding sequence or the splice site junctions of the CAII gene . However, a considerable clinical heterogeneity is observed and, therefore, a clear genotype–phenotype correlation is difficult to make. Recently, CAII defects could be excluded in a consanguineous kindred in whom osteopetrosis and distal renal tubular acidosis (dRTA) were both present. This family is noteworthy because two separate autosomal recessive genetic disorders, each affecting a different, tissue-specific subunit of V-ATPase, created a phenocopy of CAII deficiency. A homozygous mutation in the TCIRG1 gene is responsible for the osteopetrosis, whereas the dRTA is associated with a homozygous mutation in the ATP6V1B1 gene, encoding the kidney specific isoform of the B1 subunit of V-ATPase. This provides an unusual genetic explanation for the cooccurence of osteopetrosis and dRTA .
Osteopetrosis with Neuronal Anomalies
Neurologic manifestations, mostly caused by mechanical compression of the cranial nerves, are common in the severe forms of osteopetrosis (reviewed in ). There have also been case reports of an unusual association of osteopetrosis with infantile neuroaxonal dystrophy (MIM 600329) (Table 1) [67, 68, 69, 70, 71]. In addition to an increased bone density, an accumulation of neuroaxonal spheroids in parts of the central nervous system is observed. An early death has occurred in all patients. Several explanations for this unusual combination of two autosomal recessive disorders can be suggested. With the increased risk in consanguineous matings, it is possible that these two diseases are unrelated. However, not all patients with osteopetrosis and infantile neuroaxonal dystrophy have consanguineous parents [68, 69]. This unusual phenotype could also be due to pleiotropic manifestations of a single gene, or the syndrome can be caused by a contiguous gene deletion.
Severe osteopetrosis in two affected sibs from consanguineous parents was diagnosed in utero based on the presence of increased bone density, fractures, and hydrocephaly. Neuronal loss in the cortex, intense gliosis, and numerous axonal swellings were observed, and it was suggested that these brain alterations are the result of an arrest in normal development probably related to ischemia. Bone histologic examination also showed absence of osteoclasts in one child and a severely reduced number of osteoclasts in the other child . The molecular cause of this condition is still unidentified.
Osteopetrosis with Anhidrotic Ectodermal Dysplasia, Immunodeficiency, and Lymphedema
The anhidrotic ectodermal dysplasia (EDA) syndrome is characterized by abnormal development of ectoderm-derived structures like hair, teeth, nails, and sweat glands, and may be associated with immunodeficiency (EDA-ID). Only five males have been reported presenting with EDA-ID and additional features: they also had osteopetrosis with extramedullary haematopoiesis and lymphedema of the limbs (OL-EDA-ID, MIM 300301), and they all died very young of severe infections (Table 1) [73, 74, 75]. A stopcodon mutation (X420W) in the NFκB essential modulator (NEMO) gene, also known as the IKKγ gene, encoding an essential component of the NFκB signaling pathway, has been shown to cause the OL-EDA-ID syndrome in these five males (Table 1). This mutation, resulting in the addition of 27 residues to the C terminus of the mature protein, impairs NFκB signaling, confirming the importance of this signaling pathway in osteoclast biology .
Osteopetrosis with Glanzmann’s Thrombasthenia
One report of a patient with ARO concurrently diagnosed with the hemorrhagic disorder variant type Glanzmann’s thrombasthenia has been described . This male was the second child of consanguineous parents, and his brother died because of bleeding and anemia. The pathogenesis of Glanzmann’s thrombasthenia is well known: defects in either subunit of the αIIbβ3 integrin are responsible for this bleeding disorder . The β3 subunit is also implicated in osteoclast function, as shown by the combined features of the bleeding disorder and osteopetrosis in β3−/− mice . Therefore, it is speculated that this patient might have a mutation in the β3 integrin, but genetic studies have never been performed. However, Iraqi-Jewish patients with Glanzmann’s thrombasthenia who are completely deficient in β3 integrin do not have osteopetrosis because of the upregulation of α2β1 integrin in osteoclasts, which is sufficient to enable bone resorption . Of course, the occurrence of these two rare hereditary recessive disorders in this child of consanguineous parents could also be coincidental.
Finally, there are also some rare case reports describing the association of osteopetrosis with muscular degeneration , Dandy-Walker syndrome , poikiloderma , and craniosynostosis [84, 85]. Genetic analysis of these patients needs to reveal whether a new clinical phenotype or a random association is represented.
Osteoclastic bone resorption is a highly regulated multistep process that involves differentiation of osteoclasts from hematopoetic stem cells to functional multinucleated osteoclasts, polarization, cell-surface binding, demineralization, and matrix degradation. It is highly conceivable that a defect in any of these steps in the bone resorption process can result in osteopetrosis. This is illustrated by a large number of induced and spontaneous animal models with osteopetrosis (reviewed by Van Wesenbeeck and Van Hul) . In contrast to the osteopetrotic animal models, the molecular mechanisms playing a role in human osteopetrosis are almost completely restricted to osteoclast functioning. Most mutations have been identified in genes encoding CAII, CLCN7, and the a3 subunit of the V-ATPase, all participating in the acidification of the extracellular region between the osteoclast ruffled border and the bone tissue (Fig. 1). A few mutations in the OSTM1 gene have been associated with an extremely severe form of ARO. It is currently still unclear how OSTM1 functions in osteoclastic bone resorption, but suggestions have been made for a role in cytoskeletal reorganization and development of the ruffled border .
Pycnodysostosis is another condition belonging to the group of osteoclast disorders and, therefore, often confused with osteopetrosis in the past. This rare but severe bone condition is characterized by a generalized osteosclerosis resulting in short stature; disproportionally enlarged skull; nonclosure of the anterior fontanel and cranial sutures; loss of mandibular angle; dental malocclusion; hypoplastic clavicles; and short, clubbed fingers with hypoplastic or absent nails . Positional cloning has revealed that this condition is caused by loss-of-function of the cathepsin K (CTSK) gene encoding CTSK, a cysteine proteinase responsible for degradation of collagen type I and other bone proteins , again an impairment in the function of the differentiated osteoclast.
To date, there is only one exception to the general finding that the human osteopetroses are caused by impaired functioning of the differentiated osteoclast. A single hypomorphic mutation was identified in the NEMO gene in five males in which osteopetrosis is associated with anhidrotic ectodermal dysplasia, immunodeficiency, and lymphedema . The NEMO protein modulates NFκB signaling, a critical pathway in osteoclast differentiation and activation .
The lack of human osteopetrosis gene mutations in steps of the resorption process other than acidification and matrix degradation, other than the exception in NEMO, could be explained by embryonic lethality or early death, which hampers the identification of other disease-related genes. On the other hand, the existence of human compensatory pathways, absent in rodents, might also explain some differences between species. Some types of osteopetrosis could still be caused by mutations in genes involved in osteoclast differentiation. For example in a subset of ARO patients no mutations in any of the currently known genes (TCIRG1, Clcn7, and OSTM1) have been found and also the osteoclast-poor form of osteopetrosis is unexplained at the molecular level. To date, no osteopetrosis cases have been reported in which bone resorption was completely absent. It is highly conceivable that a complete absence results in embryonic lethality. Malignant forms of ARO in which osteoclasts are present but dysfunctional are still able to resorb bone to some extent . In addition, in vitro peripheral blood mononuclear cell cultures from two patients with osteoclast-poor osteopetrosis demonstrated the presence of very low osteoclast numbers but a large number of mononuclear cells. Bone slices did show occasional areas of bone resorption with normal pit morphology .
In humans, the occurrence of TCIRG1-dependent mutations is thus far limited to malignant ARO, whereas Clcn7 mutations have been found in malignant and intermediate ARO as well as in benign ADOII, demonstrating a wide spectrum of disease. The exact pathogenic effects of the Clcn7 mutations identified to date still need to be established, but most mutations in Clcn7 found in malignant and intermediate ARO are postulated to cause loss of function of CLCN7, whereas it has been hypothesized that Clcn7 mutations found in ADOII have a dominant negative effect [21, 57]. Interestingly, parents of ARO patients are in general phenotypically normal allowing Cleiren et al.  to assume that haploinsufficiency for these gene mutations most likely does not cause clinical or radiologic complications. Highest phenotypic variability, even within families, is observed in Clcn7-associated ADOII sometimes resulting in currently unexplained, clinical and radiographical nonpenetrance .
A number of groups already reported data on the cellular abnormalities of patients with ARO and ADOII and demonstrated defects in different steps of the bone resorption process, such as cell adhesion and polarization, and acid secretion [19, 31, 32, 33]. It would be very interesting to further evaluate and compare the phenotypic characteristics of osteoclasts from patients with defects in one of the currently known osteopetrosis-causing genes (TCIRG1, Clcn7, and OSTM1) as it would undoubtedly lead to a better understanding of the functional basis of this group of bone disorders.
The molecular genetics of ADOI is a different and more complicated story. Recent findings have demonstrated the involvement of LRP5, a protein with important functions in osteoblastic bone formation, in patients from two Danish families diagnosed with ADOI . The T253I missense mutation in the first propeller domain of LRP5 found in these patients is reminiscent of the G171V mutation previously reported to cause increased bone formation in the High Bone Mass phenotype . Furthermore, we were able to identify five additional missense mutations in the same domain all in conditions with and increased bone density . On the basis of radiographic picture of these conditions, including ADOI, they could be grouped as “craniotubular hyperostoses,” as we previously suggested . For ADOI bones, ultrastructural analysis revealed osteoclastopenia . However, there are some indications that this might be a secondary effect to an increased Wnt signaling. An interesting recent study carried out by Häusler and colleagues (2004)  provided evidence for an interaction between the Wnt pathway, in which LRP5 participates, and RANKL, which is known to promote osteoclast differentiation. In addition, increased levels of osteoprotegerin have been found in transgenic mice overexpressing the G171V mutation, suggesting an effect on osteoclastogenesis . To further investigate a potential relation between increased osteoblastic bone formation, caused by increased Wnt signaling, and a decrease in osteoclast number, it would be worthwhile to evaluate bone biopsies from the above-mentioned patients carrying LRP5 mutations.
The majority of osteopetrosis cases generally manifest without clinical abnormalities other than neurologic complications caused by the increased bone density. However, several forms have been reported in which osteopetrosis is associated with other clinical phenotypes developed independent of the bone resorption defect. The best known example is osteopetrosis with renal tubular acidosis and cerebral calcification caused by CAII deficiency. Because of the small number of most of these cases and the severity often causing early death, it is difficult to identify the underlying genetic causes and to reveal whether the complete clinical picture is caused by mutations in one gene or associated with different (neighboring) genes. To conclude, large efforts to elucidate the pathogenesis of different types of human osteopetrosis have resulted in the identification of a number of important genes, most of them involved in proper functioning of the osteoclast. Consequently these breakthroughs make it possible to perform molecular diagnostics and to offer genetic counseling. An important remaining problem in this context is the malignant ARO form, as for this condition currently three genes are identified and at least one is still unidentified. In the absence of a genotype–phenotype correlation and because of the diverse nature of mutations spread over the genes, molecular diagnostics for malignant ARO implies mutation analysis of all three genes even without a guarantee to find a mutation in any of them. Therefore, further genetic, cell biological, and histologic studies will be of major importance to further unravel the pathogenesis of the osteopetroses. This current knowledge and the ongoing progress of molecular genetics to identify additional gene defects in this group of disorders will undoubtedly also contribute to better understand the complex mechanisms involved in osteoclastic bone resorption and bone homeostasis, and will also aid in elucidating the pathogenesis of other bone disorders, such as osteoporosis.
We would like to thank Geert Mortier, MD, PhD, clinical geneticist at the Department of Medical Genetics of the University of Gent (Ghent, Belgium), and Filip Vanhoenacker, MD, PhD, radiologist at the Department of Radiology of the University Hospital of Antwerp (Antwerp, Belgium) who provided us with radiographs of the patients. This study is supported by the fund for scientific research flanders (FWO) with a research project (G.0404.00). WB and LVW are postdoctoral researchers of the FWD.