Background
Hyaline fibromatosis syndrome (HFS, MIM #22860) is characterized by the accumulation of clear (hyaline) substance in body tissues. Such noncancerous masses may grow under the skin and the gums resulting in bumps/nodules and gingival hypertrophy, respectively. Joint stiffness and deformities are frequent, and the skin covering the joints is often hyperpigmented. Villous atrophy and intestinal lymphangiectasia result in severe diarrhea and cachexia. Patients may come to clinical attention from birth to late childhood. The most common initial symptoms include extreme pain at minimal handling and progressive joint contractures [
1]. An early onset, more severe, and usually fatal form termed infantile systemic hyalinosis (ISH) had long been differentiated from a later onset and less severe form termed juvenile hyaline fibromatosis (JHF) [
2]. The finding of a shared genetic background, however, eventually resulted in the suggestion to use the umbrella term HFS along with a three-partite clinical grading scheme (mild vs. moderate vs. severe) [
3]. A refinement to four severity grades was proposed more recently [
4].
HFS is a recessive, genetically homogeneous disorder; it is caused by bi-allelic variants in
ANTXR2 [
5] [
6]. Approximately 100 genetically confirmed patients which carry a total of 46 distinct HFS-associated variants have been published to date [The Human Gene Mutation Database at
http://www.hgmd.cf.ac.uk]. Most frequent are missense alterations (
n = 19), but clearly inactivating alleles (3 x nonsense, 9 x splice site, 13 x frameshift, 2 x large deletions) collectively predominate. With the exception of a mutational hotspot at c.1072_1076, pathogenic variants are more or less equally distributed over the coding sequence [
7].
One of the two initial papers on
ANTXR2 mutations in HFS suggested that non-truncating variants which affect the protein’s C-terminus (encoded by exons 13–17) are associated with comparatively late disease onset and a rather benign disease course [
5]. Two subsequent meta-analyses found additional support for this hypothesis [
7] [
8]. The overall rarity of HFS, however, has hampered the search for further genotype-phenotype correlations.
The
ANTXR2 gene had initially been designated
CMG2 (capillary morphogenesis gene 2), and this was based on the observation of elevated expression in vein endothelial cells induced to undergo capillary formation [
9]. The subsequent finding of the encoded protein to serve as a receptor for the bacterial anthrax toxin resulted in the renaming to
ANTXR2 (anthrax toxin receptor 2) [
10]. The encoded ANTXR2 protein interacts with several components of the extracellular matrix [
9]. It has further been suggested to serve as a collagen receptor that mediates collagen transport to lysosomes [
11]. Impaired degradative processes may therefore contribute to the accumulation of hyaline material in HFS patients. The complete spectra of the physiological and pathological roles of ANTXR2, however, remain to be defined.
HFS shows considerable clinical overlap to Farber disease (FD), an autosomal recessive, infantile onset lysosomal storage disorder [
12]. FD is, in fact, the main differential diagnosis for HFS, with painful and swollen joints as well as periarticular and subcutaneous nodules being the most prominent shared symptoms [
1,
13]. FD is caused by mutations in
ASAH1, the gene encoding acid ceramidase [
14]. This lysosomal hydrolase catalyzes the breakdown of ceramides into sphingosines and fatty acids [
15]. Applying a targeted metabolomics approach, we recently identified ceramide C26:0 as a highly sensitive, blood-based biomarker for FD [
16]. Conceptually similar studies in HFS are currently lacking.
The present study is based on a large cohort of patients who presented with symptoms from the HFS-FD clinical spectrum. We set out to genetically characterize this cohort, and to utilize it in the search for novel genotype-phenotype correlations. We also applied an untargeted metabolomics approach in order to gain preliminary biochemical insight into HFS.
Discussion
HFS is a very rare disorder. Most previous clinical-genetic studies have therefore been able to present one or a few cases only. With a size of
n = 19, our cohort significantly increases the number of known independent patients with genetically confirmed HFS from 74 to 93 (compare Additional file
2: Table S2). Together with the pioneering paper by Hanks et al. [
5], in which 18 families were described, our study thereby represents the largest genetic report on HFS.
All of our patients were found to be homozygous for pathogenic
ANTXR2 variants, while 21% of previously published cases were compound heterozygous [
8]. Considering our cohort to contain many consanguineous families (Table
1), and to generally derive from regions with a high degree of consanguineous marriages (Table
2), the above observation is not surprising. Geographic origin may also serve to explain recurrent identification of certain variants (Table
2). This is probably true for c.1074delC, which was present in six of our patients and has previously been associated with a specific haplotype [
17]. There is evidence for further founder variants, but also for
ANTXR2 mutational hotspots [
5]. A more detailed investigation of this issue in our cases, however, was beyond the scope of the diagnosis-focused concept of the present study.
Three of the ten variants we observed have not been reported previously (Fig.
1a, Table
2). Our genetic findings thereby increase the number of known pathogenic
ANTXR2 variants to 49 [The Human Gene Mutation Database at
http://www.hgmd.cf.ac.uk]. All three novel variants are deletions of single nucleotides in rather 5′-situated or central exons, and are therefore predicted to trigger nonsense-mediated decay [
18]. They thus represent bona fide loss-of-function variants, supporting the hypothesis of HFS to be mediated by absence of ANTXR2 or complete functional inactivation [
19].
The phenotypes of all patients for which clinical information was available were consistent with the well-known, though wide-ranging spectrum of manifestations of HFS (Table
1) [
1]. Pertinent information, together with the comparatively large size of our cohort enabled us to analyze potential clinical correlations. The only corresponding finding from previous studies was that variants which affect the protein’s cytoplasmic tail (encoded by the terminal exons 13–17) and are predicted to not result in mRNA instability are associated with an overall milder disease and a later onset [
5,
7,
8]. As there was only one patient with such a variant in our cohort (Additional file
1: Table S1), a formal statistical analysis was not possible. However, the fact that this patient was > 20 years and alive at referral strongly supports a comparatively mild nature of the corresponding in-frame deletion (Fig.
2). We next stratified patients more generally according to the type of variant. This was based on the observation of non-truncating variants to be less detrimental than truncating variants in some genes (e.g. ref. [
20]). We did, however, not find evidence for an impact of the type of
ANTXR2 variant on age at onset of HFS (Fig.
2). When finally considering the gender of patients, we noticed considerable male predominance in our cohort, and found our male patients to be significantly younger at referral (Fig.
2). Given the geographic background of our cohort (Table
2), this observation may partially be explained by cultural factors that favor males over females in access to health care [
21]. We therefore initiated an exhaustive literature analysis. Though age-related data could not be compiled in a sufficiently uniform manner, there was a trend for females to more often being diagnosed with JFH rather than ISH (Additional file
2: Table S2), which indicates an overall milder manifestation and later onset [
3]. Together with the fact that there was no evidence for male predominance amongst previously published cases (Additional file
2: Table S2) [
8], this findings argues against a major impact of the above cultural factors. A gender-dependency of the clinical consequences of
ANTXR2 mutations may thus be real. Though very rare, the phenomenon of gender-specific disease manifestation has been reported for other autosomal genes (e.g. refs. [
22-
24]). Understanding its pathological basis in HFS may eventually result in hitherto unexplored therapeutic options.
As far as we are aware, our biochemical characterization of samples from HFS patients is the first pertinent effort published to date. It was facilitated by both the size of our HFS cohort and the availability of DBS samples. Given the lack of hypotheses about the impact of
ANTXR2 variants on certain blood metabolites, we had chosen an untargeted approach. Unsupervised analyses revealed that patient metabolomes are inherently different from control metabolomes (Fig.
3). Part of this overall difference, though, may be related to the lack of age- and gender-matching in our study. Indeed, metabolomics profiles have been shown to both change over time and differ between genders [
25-
27]. Ranges of values in corresponding studies, however, highly overlap and mean fold-changes rarely exceed 3, and this is in stark contrast to what is observed for our set of data (Fig.
4). Another factor that may conceptually affect a comparison between patient and control metabolomes is medication [
28]. For HFS, however, nonsteroidal anti-inflammatory drugs and opiates represent the only potentially shared drugs [
1], and these are not expected to have major influences. We thus considered the majority of the metabolic differences to be truly related to clinical/mutational status.
Our attempt to define potential metabolomics biomarkers for HFS resulted in a list of 181 candidate compounds that are associated with maximum discriminatory power (i.e. 100% sensitivity) for our patient vs. control cohorts. Though the inclusion of larger numbers of samples can be expected to result in a reduced list and in a drop in sensitivity, this observation of our pilot study is highly promising. In addition to the primarily diagnostic aspect addressed here, some of the compounds may eventually turn out to be of further relevance, e.g. for monitoring disease progression and drug response, for a stratification of patients, and/or for a better understanding of the underlying pathology [
29].
A phenotypic overlap of HFS and FD has long been recognized [
1], and our clinical-genetic findings (Table
1) re-inforce the notion that a primary clinical diagnosis of FD may need to be corrected to HFS upon genetic work-up (e.g. ref. [
30]). With FD resulting from an enzyme deficiency [
31] and HFS being due to inactivation of what is likely an extracellular collagen receptor [
11], additional analogies at the level of pathobiochemistry would not necessarily be expected. Our comparative analysis still suggested that the phenotypic similarity of HFS and FD extends to the blood metabolomics signatures (Fig.
5). Future studies will be needed to see whether this observation is due to a sharing of the primary cellular defect(s) and, thus, to common potential targets for therapeutic interventions.
Methods
Patients
The present study enrolled 19 unrelated patients referred for genetic diagnostic workup of presumably congenital phenotypes to (Rostock, Germany) Centogene AG. Nine of them had received an expert clinical diagnosis of HFS, and targeted
ANTXR2 sequencing was requested. For five patients, the initial genetic diagnostic request had been targeted
ASAH1 sequencing based on a clinical suspicion of FD. Whole exome sequencing (WES) was requested for the remaining five patients (Table
1). For a subset of the above index cases, samples from unaffected family members were provided, too. The most frequent region of origin was the Middle East, followed by Africa, Latin American and Asia (compare Table
2). For metabolomic profiling, eleven HFS patients five patients with genetically confirmed FD and 12 healthy controls were included [
16].
DNA preparation
Samples were provided as ready-to-use DNA, EDTA blood, or as dried blood spots (DBSs) on filter cards (CentoCard®, Centogene AG). Extraction from the blood-based samples utilized QIAsymphony instruments in combination with reagents and kits as recommended by the manufacturer (Qiagen, Hilden, Germany).
ANTXR2 variant screening
The coding sequence of ANTXR2 (NM_058172.5; NP_477520.2) including at least 50 bp of adjacent untranslated regions or intronic sequences was amplified exon-wise from genomic DNA (primers available upon request). PCR-products were extracted from agarose gels, purified according to standard procedures, and sequenced from both sides on a 3730xl sequencer (Thermo Fisher Scientific, Waltham, MA).
Three DBS punches of 3.2 mm in diameter were prepared from filtercards using a DBS puncher (Perkin Elmer LAS, Germany), and collected into 2.2 ml round bottom tubes (Eppendorf, Germany). Extraction was performed by adding 50 μL extraction solution (DMSO:H2O, 1:1) and 100 μL internal standard solution (lyso-Gb2, Matreya LLC, USA, 200 ng/mL in ethanol). After a brief vortex-mixing, the tubes were shaken (700 rpm) at 37 °C for 30 min and then sonicated at maximum power for 1 min. All liquid was subsequently transferred to an AcroPrep Filter Plate with PTFE membrane (PALL, Germany) placed on a 96 well V-shape bottom plate (VWR, Germany). To remove solid particles, samples were filtrated by centrifugation at 3.500 rpm for 5 min.
Mass spectrometric (MS) analysis was performed on a Waters Acquity i class UPLC (Waters, UK) coupled with a Vion IMS-QTof mass spectrometer (Waters, UK). Chromatographic run was performed on a Kinetex EVO C18 column (Phenomenex, Germany) with a gradient from 0 to 100% organic solvent (50 mM formic acid in acetonitrile:methanol, 1:1, v:v). Mass spectrometric acquisition was made using the following parameters: analyzer mode - sensitivity, MS mode - High definition MSE, capillary voltage - 1.2 kV, source temperature - 150 °C, desolvation temperature - 600 °C, desolvation gas - 1000 L/h, cone gag - 50 L/h, low collision energy - 6 eV, high collision energy ramp: 20–40 eV, scan mass: 50–1000 m/z, scan time - 0.5 s.
10 μL samples were injected and an HDMSE analysis method was used. The acquisition was done using the Unifi software (Waters, UK) and the results exported as a Unifi export file (.uep). The results were imported in the Progenesis QI software (Nonlinear Dynamics, UK) for statistical interpretation. From the list of identified compounds, only those with significant difference between the groups were selected for further use. Targeted mass spectrometry-based screening for the levels of ceramide C26:0 in DBSs was performed as described in detail previously [
16].
Normalization, filtering and analysis of metabolomics data
Raw abundances as detected by untargeted MS were normalized using default settings in Progenesis. Compounds with a charge of > 5 and a mass-to-charge ratio (m/z) < 179 were removed (quality filters). For comparative analyses, only compounds with a median normalized abundance of > 100 counts relative to the reference compound in at least one of the groups under consideration were retained (quantity filter).
Normalized and filtered abundances were transformed into CSV files, and uploaded into the ‘Statistical Analysis’ tool-box of MetaboAnalyst 4.0 at
http://www.metaboanalyst.ca [
32]. Principal component analysis was performed using default settings. Dendrograms were derived using distance measure ‘Spearman’ and clustering algorithm ‘Single’. Distributions and ranges for values were visualized by generating heat-maps with enforced sample grouping.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.