DIAPH1 encodes human DIA1, a formin protein that elongates unbranched actin. The c.3634+1G>T DIAPH1 mutation causes autosomal dominant nonsyndromic sensorineural hearing loss, DFNA1, characterized by progressive deafness starting in childhood. The mutation occurs near the C‐terminus of the diaphanous autoregulatory domain (DAD) of DIA1, which interacts with its N‐terminal diaphanous inhibitory domain (DID), and may engender constitutive activation of DIA1. However, the underlying pathogenesis that causes DFNA1 is unclear. We describe a novel patient‐derived DIAPH1 mutation (c.3610C>T) in two unrelated families, which results in early termination prior to a basic amino acid motif (RRKR1204–1207) at the DAD C‐terminus. The mutant DIA1(R1204X) disrupted the autoinhibitory DID‐DAD interaction and was constitutively active. This unscheduled activity caused increased rates of directional actin polymerization movement and induced formation of elongated microvilli. Mice expressing FLAG‐tagged DIA1(R1204X) experienced progressive deafness and hair cell loss at the basal turn and had various morphological abnormalities in stereocilia (short, fused, elongated, sparse). Thus, the basic region of the DAD mediates DIA1 autoinhibition; disruption of the DID‐DAD interaction and consequent activation of DIA1(R1204X) causes DFNA1.
A novel clinical DIAPH1 mutation (c.3610C>T) is found in two unrelated families, resulting in a constitutively active DIA1. Patients and mice expressing the DIA1(R1204X) protein show progressive hearing loss beginning in the high‐frequency range and mimicking a novel DFNA1 subtype.
A novel DIAPH1 mutation results in early termination of DIA1 at the diaphanous autoregulatory domain (DAD), R1204X.
DIA1(R1204X) disrupts the diaphanous inhibitory domain (DID)–DAD interaction and is constitutively active.
Mice expressing DIA1(R1204X) show progressive deafness and OHC loss at the basal turn.
DIA1(R1204X) causes a novel DFNA1 subtype showing progressive hearing loss beginning in the high‐frequency range and in childhood.
In the past two decades, extensive research on the genetics of nonsyndromic hereditary deafness (NSHD) has been conducted, leading to the discovery of about 100 genes essential for hearing (http://hereditaryhearingloss.org/): about 30 of these encode proteins that interact directly or indirectly with actin (Dror & Avraham, 2010; Drummond et al, 2012). Stereocilia, which are deflected by sound stimulation and are thus the key structures for hearing, are actin‐based protrusions and exquisitely organized microvilli/filopodia composed of hundreds of parallel actin filaments with the plus ends at the distal tip; they are organized into precise rows of graded height (characteristic staircase patterns) on the apical surface of cochlear hair cells (HCs) (Frolenkov et al, 2004). Cochlear HCs are arranged in a single row of inner HCs (IHCs) and three rows of outer HCs (OHCs). Although IHCs and OHCs are believed to share common mechanotransduction machinery, they have distinct roles during sound detection: IHCs are true sensors, whereas OHCs function as amplifiers through an active process that involves stereociliary and somatic motility (Schwander et al, 2010). The high sensitivity of HCs depends on the coordinated/synchronized movement of stereocilia upon mechanical stimulation; thus, the precise organization of the length and shape of stereocilia, which is regulated by a large battery of genes, is indispensable.
Fifteen formin proteins, which nucleate and elongate unbranched/straight actin, are found in humans and can be classified into eight subfamilies (Campellone & Welch, 2010). The diaphanous‐related formin (DRF) subfamily contains proteins commonly referred to as DIAs (also known as DIAPHs and mammalian homologs of Drosophila Diaphanous), which are the best‐characterized class of formins. DIAPH1 gene encodes one of three DIA isoforms, DIA1 (DIAPH1). A mutation in DIAPH1 (c.3634+1G>T), in which the canonical splicing donor sequence is disrupted (AAGgtaagt becomes AAGttaagt), moves the splicing donor site to a cryptic site four base pairs (bp) downstream of the original site, resulting in a 4 bp (ttaa) insertion in the gene transcript. The insertion results in a frameshift near, but outside, the C‐terminal end of the diaphanous autoregulatory domain (DAD) of DIA1, hereafter referred to as DIA1(ttaa). The frameshift introduces 21 aberrant amino acids (aa) and removes 32 aa of wild‐type (WT) DIA1 (p.Ala1212ValfsX22) (Fig 1A) and causes autosomal dominant nonsyndromic sensorineural hearing loss, DFNA1 (Lynch et al, 1997). DIAs have a modular domain organization consisting of the GTPase‐binding domain (GBD), a partially overlapping N‐terminal diaphanous inhibitory domain (DID), formin homology (FH) domains FH1 and FH2, and the C‐terminal DAD (Fig 1A). Through an autoinhibitory intramolecular interaction between the DID and the DAD, which is regulated by Rho family GTPases (Amano et al, 2010), DIAs are held inactive in the resting state (Watanabe et al, 1999; Campellone & Welch, 2010; Kuhn & Geyer, 2014). DRFs including mouse Dia1 show directional movement over the distance of several micrometers in cells as they processively elongate actin filaments via their FH2 domain (Watanabe & Mitchison, 2002; Higashida et al, 2004). Given the location of the mutation, it has been speculated that disruption of the autoinhibitory intramolecular DID‐DAD interaction, and consequent activation of DIA1, might underlie DFNA1 pathology. However, it is also possible that the DIA1(ttaa) mutation might compromise other functions of DIA1 that are not related to autoregulation. Thus, the underlying pathogenesis of DFNA1 remains to be determined at the molecular level.
Patients in the original DFNA1 pedigree demonstrate mild‐ and low‐frequency deafness starting in childhood (about 10 years of age), which develops into profound deafness involving all frequencies by adulthood (Lynch et al, 1997). Although DFNA1 was thought to originate due to sensorineural defects, evaluation of a young (8‐year‐old) DFNA1 patient with mild‐ and low‐frequency deafness revealed a normal auditory brainstem response (ABR). It was therefore suggested that endolymphatic hydrops was involved in the early pathophysiology of the condition (Lalwani et al, 1998). Moreover, to date DFNA1 has only been reported in one Costa Rican family (called the “M” family because of the founder name, Monge). Thus, which of the two proposed underlying mechanisms is responsible for the development and clinical manifestations of DFNA1 remains unclear.
Here, we identified a novel DIAPH1 point mutation (c.3610C>T), which results in early termination just before four sequential basic amino acids (RRKR1204–1207) in the DAD, p.R1204X. This mutation (which we define as “R1204X”) was found in two unrelated Japanese families and showed autosomal dominant progressive deafness in the high‐frequency range that began in childhood. DIA1(R1204X) behaved as a constitutively active DIA1 mutant, as the mutation disrupted the autoinhibitory intramolecular DID‐DAD interaction. This was associated with frequent directional actin polymerization movement and an increased number of elongated microvilli in cells. Furthermore, transgenic (TG) mice expressing FLAG‐tagged DIA1(R1204X) showed progressive deafness that began in the high‐frequency range in parallel with cochlear HC loss, predominantly in OHCs at the basal turn. Stereocilia were sparse and abnormally short, elongated, and fused. Comparison of in vivo models, including the novel and clinically relevant one we present here, will provide further insight into the molecular mechanisms that underlie the development of DFNA1 and the diverse clinical symptoms with which it is associated.
By performing genomic sequencing, we identified a novel DIAPH1 point mutation (c.3610C>T) generating the DIA1(R1204X) mutant, which has an early translation termination site just before four sequential basic amino acids (RRKR1204–1207) in the C‐terminal end of the DAD (Fig 1A and B). The location of the mutation is near the 4‐bp (ttaa) insertion in DIA1 (DIAPH1) mRNA causing original DFNA1, but the ttaa insertion itself is located outside the DAD. This mutation was found in two unrelated Japanese families. Family pedigrees of both patients showed an autosomal dominant inheritance of deafness (Fig 1B).
Patient 1 had a history of hearing loss that started in her childhood. She experienced slowly progressive deafness and was referred to an otologist at the age of 48. Magnetic resonance imaging of her brain, including the inner ear and VIII nerve, showed no abnormalities. Her audiograms (both air and bone conduction) showed bilateral deafness, predominantly in the high‐frequency range, although all frequencies were affected (Fig 1B). Two years after her admission, her hearing deteriorated significantly, particularly in the low‐frequency range (Fig 1B).
Patient 2 passed newborn hearing screening. She felt progressive deafness starting around 3 years old and was referred to an otologist at the age of 6. Her audiograms (both air and bone conduction) showed bilateral and high‐tone deafness (Fig 1B).
Elongation of microvilli by DIA1(R1204X) mutant
To examine the effect of DIA1(R1204X) on stereocilia, which are an exquisitely organized subtype of microvilli/filopodia, we studied the microvilli of MDCK cells cultured on membrane inserts. Because of the difficulties in observing and comparing the length of microvilli in reconstituted lateral images obtained by a confocal laser microscope, we established MDCK cells stably expressing mCherry‐ESPIN1, which have elongated microvilli (about 2–3 μm) labeled by the mCherry red fluorescent protein. GFP‐DIA1(M1190D), a constitutively active DIA1 mutant (Lammers et al, 2005), was localized to microvilli, induced microvilli elongation and altered cell shape (Fig 2). Unlike GFP‐DIA1(M1190D), the GFP‐DIA1(R1204X) mutant was not discretely localized to microvilli, although it induced elongated and congested microvilli compared with surrounding nontransfected cells and WT GFP‐DIA1‐transfected cells (Fig 2). These results suggest that DIA1(R1204X) may be a constitutively active mutant of DIA1, albeit with weaker activity than DIA1(M1190D).
Induction of elongated microvilli by DIA1(R1204X)
To confirm that DIA1(R1204X) is a mildly active DIA1 mutant, we utilized HeLa cells, which have previously been used in studies of DIA/Dia (Watanabe et al, 1999; Sakamoto et al, 2012). Although GFP‐DIA1(M1190D) localized to the plasma membrane (PM) and induced stress fiber (SF) formation, GFP‐DIA1, GFP‐DIA1(R1204X), or GFP‐DIA1(ttaa) induced no remarkable SF formation (Fig 3A). However, GFP‐DIA1(R1204X) had a dot‐like localization adjacent to the PM (Fig 3A). Furthermore, 3D imaging revealed that GFP‐DIA1(R1204X) localized to microvilli, and induced elongated and thick microvilli on the surface (Fig 3B, Movie EV1). Nontransfected cells (Fig 3B, Movie EV1) or those expressing GFP‐DIA1 or GFP‐DIA1(ttaa) (Movies EV2 and EV3) had no effect on microvilli phenotypes. Statistical analysis revealed that the longest microvilli in GFP‐DIA1(R1204X)‐expressing cells were significantly longer than those in GFP‐DIA1‐expressing and nontransfected cells [R1204X (n = 10): 6.25 ± 0.53 μm, WT (n = 10): 3.09 ± 0.23 μm, nontransfected: 3.12 ± 0.22 (n = 10); five independent experiments; P < 0.0001 (between WT and R1204X, nontransfected and R1204X) by one‐way ANOVA followed by Bonferroni's post hoc test]. These results confirm that DIA1(R1204X) is a mildly active DIA1 that alters microvilli phenotypes, but that has no remarkable effects on SF formation.
Disrupted intramolecular DID‐DAD interaction in DIA1(R1204X)
To further confirm that DIA1(R1204X) is constitutively active due to the disrupted autoinhibitory interaction between the DID and the DAD, we performed pull‐down assays using GST‐tagged DID and biotin‐tagged DADs: DAD(WT), DAD(R1204X), and DAD(M1190D) (Fig 4A). The interaction between GST‐DID and biotin‐DAD(R1204X) was significantly weaker than that between GST‐DID and biotin‐DAD(WT), and similar to that between GST‐DID and biotin‐DAD(M1190D) (Fig 4B). These results suggest that the DIA1(R1204X) mutation disrupts the autoinhibitory intramolecular DID‐DAD interaction, thereby rendering it constitutively active. However, DIA1(R1204X) was less active than DIA(M1190D) in cells, suggesting that factors other than the DID‐DAD interaction may also regulate DIA1 activity.
Importance of electrostatic interaction for the DID‐DAD interaction
An acidic groove formed by acidic aa of the DID (E358, D361, E362, and D366 in mouse Dia1; E358, E361, E362, and D366 in human DIA1) has been identified by crystal structure analysis (Lammers et al, 2008; Nezami et al, 2010). Given that the DAD has four sequential basic aa (RRKR1204–1207 in DIA1; RRKR is conserved in Dia1) at its C‐terminus, electrostatic interactions may play an important role in the intramolecular DID‐DAD interaction (Fig EV1A) (Lammers et al, 2008; Nezami et al, 2010; Kuhn & Geyer, 2014). Based on the prediction that exchange of RRKR1204–1207 into EEEE would render DIA1 more active than DIA1(R1204X), we constructed a mutant DIA1(RRKR1204–1207/EEEEX) (Fig EV1A). GFP‐DIA1(RRKR1204–1207/EEEEX) was localized to both the PM and microvilli, and its expression induced both microvilli elongation and SF formation (Fig EV1B, Movie EV4). These data suggest that this mutant is more active than DIA1(R1204X) and instead has similar activity to DIA1(M1190D) (Fig 3A).
To further confirm that electrostatic interactions play critical roles in the DID‐DAD interaction, we made three additional mutants: DIA1(RRKR1204–1207/EEEX), DIA1(RRKR1204–1207/EEX), and DIA1(RRKR1204–1207/EX) (Fig EV2A). PM localization was used as surrogate marker of activation, because active mutants, including DIA1(M1190D) and DIA1(RRKR1204–1207/EEEEX), are localized to the PM. DIA1(RRKR1204–1207/EEEX) and DIA1(RRKR1204–1207/EEX) were localized to the PM, whereas DIA1(RRKR1204–1207/EX) and DIA1(R1204X) showed a dot‐like localization in the adjacent area to the PM. The degree to which each of the four mutants localized to the PM (or areas adjacent to the PM) depended on the number of acidic aa: DIA1(RRKR1204–1207/EEEEX) ≥ DIA1(RRKR1204–1207/EEEX) > DIA1(RRKR1204–1207/EEX) > DIA1(RRKR1204–1207/EX) > DIA1(R1204X) > WT DIA1 (Fig EV2B). These results indicate that electrostatic interactions play critical roles in the DID‐DAD interaction, and that DIA1(R1204X) is a mildly active DIA1 mutant.
We made four additional mutants: DIA1(RRKR1204–1207/EEEE), DIA1(RRKR1204–1207/EEE), DIA1(RRKR1204–1207/EE), and DIA1(RRKR1204–1207/E) (Fig EV2C), which all retain the C‐terminal 56 aa of DIA1 (1208–1263 aa). DIA1(RRKR1204–1207/E) was less strongly associated with areas adjacent to the PM when compared to DIA1(RRKR1204–1207/EX), with the degree of PM binding among this second set of mutants being DIA1(RRKR1204–1207/E) < Dia1(RRKR1204–1207/EE) < Dia1(RRKR1204–1207/EEE) < Dia1(RRKR1204–1207/EEEE) (Fig EV2D). Thus, we conclude that C‐terminal aa (1208–1263 aa) are required to suppress DIA1 activation.
Insufficient autoinhibition of DIA1(R1204X) in live cells
Previous fluorescence single‐molecule speckle microscopy (SiMS) study revealed that the FH1‐FH2 domain of mouse Dia1 shows processive movement in living cells based on its constitutive actin polymerization activity (Higashida et al, 2004). In full‐length WT Dia1, the polymerization activity of FH1‐FH2 domain is suppressed by the autoinhibitory DID‐DAD interaction, and binding of Rho family GTPases to the DID relieves this suppression (Watanabe et al, 1999). Compatible with this finding, WT Dia1 showed scarce processive movement in a SiMS study (Higashida et al, 2008), and microinjection of active RhoA(G14V) induced processive movement (Higashida et al, 2004).
We expressed GFP‐tagged DIA1 molecules in living cells in order to evaluate the actin elongation activity of DIA1(R1204X) using fluorescence SiMS (Watanabe, 2012). GFP‐DIA1(R1204X) expressed in XTC cells showed frequent processive movements (Fig 5B, Movie EV6) compared with GFP‐DIA1(WT) (Fig 5A, Movie EV5), which indicated that the R1204X mutant had insufficient autoinhibition. GFP‐DIA1(M1190D), a constitutively active mutant, also showed directional movements, which seemed more frequent than GFP‐DIA1(R1204X) (Fig 5C, Movie EV7). To further quantify the differences between these three DIA1 variants, we counted the numbers of speckles that exhibited processive movement, and normalized these values by dividing them by the total fluorescent intensity of the observed area. This revealed significant differences in behavior, with the order (in terms of the most processive variant first) being GFP‐DIA1(M1190D) > GFP‐DIA1(R1204X) > GFP‐DIA1(WT) (Fig 5D). These data suggest that DIA1(R1204X) engenders upregulated actin elongation activity by disrupting the DID‐DAD interaction, and that DIA1(R1204X) is less active than GFP‐DIA1(M1190D).
DIA1(R1204X)‐TG mice show progressive deafness and HC loss
To date, the possibility that DFNA1 caused by DIA1(ttaa) may be due to loss‐of‐function has not been completely excluded. Moreover, to the best of our knowledge, cochlear function in Dia1‐KO mice has not been experimentally evaluated in detail (Thumkeo et al, 2013). At 5 weeks of age, Dia1‐KO mice showed no HC loss or morphological abnormalities of substructures in the cochlea, including the organ of Corti (OC), spiral ganglion, spiral ligament, and stria vascularis (Fig 6A and B). The hearing function of Dia1‐KO mice evaluated by ABR was no different to that of control mice at 5 weeks of age (Fig 6C). These results further suggest that DFNA1 caused by the original DIA1(ttaa) mutant is unlikely to be due to loss‐of‐function, dominant negative effects, or haploinsufficiency.
Based on our finding that DIA1(R1204X) is a mildly active mutant, and that a loss‐of‐function effect is unlikely to be the cause of DFNA1 in the context of this variant, we decided to make TG mice expressing FLAG‐tagged DIA1(R1204X) under control of the CAG promoter (Fig 7A). First, we used quantitative PCR (qPCR) to determine the copy number of the transgene, FLAG‐DIA1(R1204X), integrated into the genome (3.86 ± 0.26; n = 7). Expression of FLAG‐DIA1(R1204X) in the inner ear was confirmed using a FLAG antibody (Ab) (Fig 7B). Finally, we assessed expression levels of FLAG‐DIA1(R1204X) protein in the inner ear, thymus, and heart by immunoblotting using a Dia1/DIA1 Ab, whose immunogen is 100% conserved between mouse Dia1 and human DIA1. The ability of the Ab to recognize the various Dia proteins with equal efficiency was confirmed by our observation that this Ab gave similar signal intensities when incubated with equal amounts of GFP‐Dia1 and GFP‐DIA1 protein obtained from transfected HEK293 cells (Fig 7C). The ratio of endogenous Dia1 + transgenic DIA1(R1204X)/endogenous Dia1 was 2.07 ± 0.33 in the inner ear and varied in other organs (for example, 4.89 ± 0.69 in the thymus, Fig 7C).
To assess cochlear function, we examined ABR with broadband click (2–4 kHz) and 8–32 kHz pure tone‐burst stimuli, using 5‐ to 25‐week‐old DIA1(R1204X)‐TG mice and control mice. At 10 weeks, DIA1(R1204X)‐TG mice showed elevated ABR thresholds at high‐frequency sounds (24 kHz and 32 kHz) compared with control mice (Fig 8A). Hearing impairment of DIA1(R1204X)‐TG mice progressively deteriorated and involved all frequencies, including low‐frequency sounds at 25 weeks compared with control mice (Figs 8A and EV3A). Furthermore, we examined sex differences in ABR using 25‐week‐old mice: No significant difference was observed between male and female DIA1(R1204X)‐TG mice (Fig EV3B).
Next, we examined the ultrastructure of the OC in DIA1(R1204X)‐TG mice using a scanning electron microscope (SEM). At 10 weeks, loss of OHCs started at the basal turn of DIA1(R1204X)‐TG cochlea; in contrast, IHC was intact in all three turns (Fig 8B). At 25 weeks, loss of OHCs at basal and middle turns became significant, and loss of IHCs at three turns was also significant in DIA1(R1204X)‐TG mice (Fig 8C). Loss of OHC function in DIA1(R1204X)‐TG was physiologically confirmed using DPOAE (Fig 8D). High magnification at 25 weeks revealed that some OHCs had sparse and short stereocilia, and some IHCs had sparse and fused stereocilia (Fig 8E).
Finally, to confirm that progressive hearing loss and OHC‐ and basal turn‐dominant HC loss was induced by the DIA1(R1204X) mutant, we bred the DIA1(R1204X) mice to homozygosity. When compared with heterozygous TG mice, the homozygotes showed exacerbated hearing impairment involving all frequencies (Fig EV4A) and exacerbated HC loss dominant in OHC and in the basal turn (Fig EV4B).
Here, we report a novel DIAPH1 mutation that generates a mildly active DIA1(R1204X) mutant due to the disrupted autoinhibitory intramolecular DID‐DAD interaction of DIA1; we also show that expression of this mutant in vivo is responsible for progressive deafness. The DIA1(R1204X) mutant has an early termination site immediately prior to a four amino acid basic sequence (RRKR1204–1207). The DAD of Dia1/DIA1 is composed of an amphipathic helix with the central consensus motif MDxLLxxL followed by an unstructured basic region, including RRKR (Lammers et al, 2005). An acidic groove formed by acidic amino acids in the DID of Dia1/DIA1 is located adjacent to the MDxLLxxL recognition site (Lammers et al, 2008; Nezami et al, 2010). While the MDxLLxxL is essential for binding to a hydrophobic surface patch at the concave side of the DID, the basic region of the DAD is thought to interact with the acidic groove of the DID (Lammers et al, 2005; Wallar et al, 2006; Kuhn & Geyer, 2014). Thus, the basic region of the C‐terminal end of the DAD is also an important contributor to the DID‐DAD interaction. This would explain why mutation around RRKR1204–1207 leads to constitutive DIA1 activation.
We previously reported inner ear HC‐specific Cdc42‐knockout (KO) mice, which show progressive deafness in parallel with HC loss with stereocilia degradation, including short, elongated, fused stereocilia (Ueyama et al, 2014). Furthermore, we found that the amount of active RhoA (GTP‐form) is increased in Cdc42‐knockdown (KD) cells (Ueyama et al, 2014). Because RhoA binds to and activates Dia1 (Watanabe et al, 1999), we suggested that the activated RhoA–Dia1 signaling axis is involved, at least to some extent, in HC loss and stereocilia degradation in Cdc42‐KO mice (Ueyama et al, 2014). In support of our hypothesis, the degenerative stereocilia phenotypes we observed in TG mice expressing the active DIA1 mutant, DIA1(R1204X), are similar to those observed in Cdc42‐KO mice. In addition to the absence of a hearing phenotype in Dia1‐KO mice in the present study, patients with loss of functional variants in DIA1 (p.Q778X, p.F923 fs, and p.R1049X) suffer from microcephalus, blindness, and seizure, but are not deaf (Al‐Maawali et al, 2016; Ercan‐Sencicek et al, 2015). Moreover, a transgenic Drosophila melanogaster strain with only one diaphanous gene and expression of a constitutive active form of diaphanous protein by removing both the N‐terminal GBD and the C‐terminal DAD showed impaired response to sound (Schoen et al, 2010). Taken together, these data indicate that the pathogenesis of DFNA1 caused by DIAPH1 mutations is likely due to gain‐of‐function rather than loss‐of‐function.
DIA1(R1204X)‐TG mice showed progressive deafness (mainly in the high‐frequency range in the early stage, but later progressing to the low‐frequency range) in parallel with OHC‐dominant HC loss. Various types of stereocilia degradation (Figs 8E and EV5), including short, sparse (reduced number per HC), fused, and elongated stereocilia, were present; however, the elongated stereocilia phenotype was less frequent than the occurrence of short stereocilia. Nonsyndromic auditory neuropathy autosomal dominant 1 (AUNA1) is due to a mutation in the promoter region of DIAPH3 (encodes DIA2), which increases the expression of both DIAPH3 mRNA and protein. The condition leads to progressive hearing loss beginning in the second decade of life (Schoen et al, 2010). In a mouse model of AUNA1, in which TG mice overexpress WT Diaph3 (encodes Dia2), animals experience progressive deafness and elongated and fused stereocilia selectively in IHCs without HC loss (Schoen et al, 2013). The different phenotypes between DFNA1 and AUNA1 both in human patients and mouse models are a strong indication that DIAPH1 (DIA1) and DIAPH3 (DIA2) have distinct expression patterns and/or function in HCs. Although accelerated actin polymerization due to Dia2 overexpression might explain the elongated stereocilia, the molecular pathology that leads to the predominance of short and sparse stereocilia in DIA1(R1204X)‐TG mice is unclear. One possibility is that DIA1(R1204X) enhances actin polymerization not only at stereocilia but also at other sites. This could include adherens junctions (AJs), which are one of the sites where Dia1 is active (Sahai & Marshall, 2002), and would lead to misregulated/unbalanced actin turnover in cells. Impairment specifically at the apical surface of the neuroepithelium was reported in Dia1/Dia3‐DKO mice, suggesting the primary functional site of Dia1 is at this site (Thumkeo et al, 2011). Transient overexpression of DIA1(R1204X) in MDCK and HeLa cells induced elongated microvilli; however, long‐term expression of the active mutant may induce misregulated/unbalanced actin polymerization not only at stereocilia but also other sites including AJs. This would ultimately lead to the phenotypic variation from enhanced (elongated stereocilia) to reduced (short and defective stereocilia) actin polymerization. Indeed, we observed the deformed cell–cell junction between HC and supporting cell (SC) in DIA1(R1204X)‐TG mice (Fig EV5).
We obtained only one line of DIA1(R1204X)‐TG mice among a total 41 founder (F0) lines, despite injecting the transgene on three separate occasions into a total of 412 fertilized eggs. qPCR analysis revealed that four transgene copies were integrated into the genome in heterozygous TG mice, which is significantly lower than the usual number (Brinster et al, 1981). Expression levels of DIA1(R1204X) protein in heterozygous TG mice varied in a tissue‐ and organ‐specific manner; for example, levels were low in the inner ear and very high in the heart. Considering previous reports and our own experience, we suggest that high level DIA1(R1204X) expression in TG mice is cytotoxic and therefore leads to embryonic lethality. In support of this speculation, overexpression of active diaphanous protein in TG flies (Schoen et al, 2010) causes lethality in the pupal stage in some cases. Dia1 plays important roles in cell polarization, migration, axonogenesis, and exocrine vesicle secretion in the apical membrane (Kuhn & Geyer, 2014). Although mice with single KO of Dia3 show no apparent phenotype (Thumkeo et al, 2013), Dia2‐KO leads to embryonic lethality (Watanabe et al, 2013). Very high levels of active DIA1 overexpression may also interfere with the function of Dia2 and Dia3. Furthermore, the HC loss and deafness phenotypes were exacerbated when heterozygous DIA1(R1204X)‐TG mice were bred to homozygosity. From these observations, we infer that that phenotypes observed in DIA1(R1204X)‐TG mice are unlikely to be due to integration site effects of the transgene, and that our DIA1(R1204X)‐TG model faithfully recapitulates phenotypes of DFNA1 patients who inherited the condition in an autosomal dominant fashion.
The original report of DFNA1 was due to the DIA1(ttaa) mutation, which caused mild‐ and low‐frequency deafness with normal ABR at the early stage of disease, and was diagnosed as endolymphatic hydrops (Lalwani et al, 1998). On the contrary, DFNA1 is associated with deafness (predominantly in the high‐frequency range) in two patients caused by the newly described DIA1(R1204X) mutation, as well as in DIA1(R1204X)‐TG mice. This suggests that the novel DFNA1 cases and the original DFNA1 are associated with distinct pathophysiologies. Moreover, patient 2 and DIA1(R1204X)‐TG mice showed deafness beginning in the high‐frequency range, and DIA1(R1204X)‐TG mice progressed to deafness involving all frequencies. Patient 1 presented with high‐frequency dominant hearing loss, but also loss of hearing at all other frequencies; this was probably since this particular clinical case was advanced, as the patient was 48 years old when diagnosed. Thus, the clinical manifestation of the novel DFNA1 subtype is likely deafness beginning with loss of the high‐frequency ranges in childhood, which slowly progresses to deafness involving all frequencies. We showed that the C‐terminal amino acid sequence after the DAD (1208–1263 aa) acts as a repressor of DIA1 activation (compare Fig EV2B and D). The DIA1(ttaa) mutant loses the basic amino acid sequence RK1214–1215, which is the next basic amino acid cluster following RRKR1204–1207, and has a shorter C‐terminal (lacking 31 aa) than WT DIA1. This suggests that DIA1(ttaa) would be a weaker constitutively active DIA1 mutant than DIA1(R1204X). Although DIA1(ttaa) elicited no apparent phenotype in the transient overexpression cell models (either MDCK or HeLa cells), long‐term observation using animal models will facilitate elucidation of the mechanisms associated with DIA1(ttaa)‐driven pathologies.
In summary, we identified a novel DFNA1 subtype caused by constitutively active DIA1 mutant, DIA1(R1204X), in two unrelated families, and confirmed its pathogenicity using cell and mouse models. Our DIA1(R1204X)‐TG mice (in which expression levels of transgenic DIA1(R1204X) mutant protein in the inner ear of heterozygous mice are probably similar to DFNA1 patients) showed progressive hearing loss beginning in the high‐frequency range in parallel with HC loss and stereocilia degradation that was dominant in OHCs and in the basal turn. Nonsense‐mediated mRNA decay is unlikely to explain the pathogenicity of DFNA1 caused by DIA1(R1204X). Studies using knock‐in mice of DIA1(R1204X) may further confirm our results and will help to provide a molecular basis for the clinical diversity associated with DFNA1.
Materials and Methods
A total of 1,120 Japanese patients with bilateral nonsyndromic sensorineural hearing loss participated in this study. In total, 269 normal hearing controls, confirmed by pure tone audiometry, were also enrolled. Informed written consent was obtained from all subjects. This study was approved by the Shinshu University Ethical Committee, and the ethics committees of all other participating institutions listed in a previous report (Nishio & Usami, 2015).
We performed massively parallel DNA sequencing (MPS) analysis using an Ion Torrent Personal Genome Machine (PGM™) system with the Ion AmpliSeq™ custom panel (Thermo Fisher) in 1,174 samples (905 hearing loss cases and 269 normal hearing controls) and using a HiSeq 2000 sequencer (Illumina) with the SureSelect target DNA enrichment kit (Agilent) in 215 cases.
Amplicon library preparation and Ion PGM platform sequencing.
Amplicon libraries of the target gene exons from 63 genes reported to cause nonsyndromic hearing loss were prepared with the Ion AmpliSeq™ panel. They were diluted to 20 pM, and the same amount of libraries from the six libraries of six patients was pooled for one sequence reaction. Emulsion polymerase chain reaction and sequencing were performed with the Ion PGM™ system using the Ion PGM™ 200 sequencing kit and the Ion 318™ chip (Thermo Fisher). A detailed protocol has been described elsewhere (Miyagawa et al, 2013; Nishio et al, 2015). The sequence data were mapped to the human genome sequence (build GRCh37/hg19), and the DNA variant regions were piled up with the Torrent Variant Caller plug‐in software ver. 4.0 (Thermo Fisher).
Targeted enrichment and HiSeq platform sequencing.
The SureSelect target DNA enrichment kit, designed for the 112 potentially deaf‐causing genes, was used in this study (Abe et al, 2003). The detailed gene list is described in our previous report (Miyagawa et al, 2013). A 3‐μg DNA aliquot was fragmented using the Covaris™ S2 System (Covaris) into fragments of ~200‐bp length. Furthermore, the target regions were enriched with a barcode adapter (Agilent). The same amount of libraries from each of 12 patients was pooled into one tube and analyzed in one lane of the HiSeq 2000 sequencer. The sequence data were processed by filtering the read quality QV = 30 as cutoff, and duplicate reads were also removed. After the filtering process, sequence reads were mapped to the human genome sequence (build GRCh37/hg19) using BWA software (Li & Durbin, 2009). After sequence mapping, the DNA variant regions were compiled with GATK software (Van der Auwera et al, 2013).
Filtering detected variants.
After their detection, the effects of the variants were analyzed using ANNOVAR software (Wang et al, 2010; Chang & Wang, 2012). We selected missense, nonsense, insertion/deletion, and splicing variants. Variants were further selected as < 1% of: (i) the 1000 Genomes Database (http://www.1000genomes.org/), (ii) the 6,500 exome variants (http://evs.gs.washington.edu/EVS/), (iii) the human genetic variation database (dataset for 1,208 Japanese exome variants) (http://www.genome.med.kyoto-u.ac.jp/SnpDB/index.html), (iv) the 269 in‐house Japanese normal hearing controls, and (v) 1,000 control data in the deafness variation database (Shearer et al, 2014) (http://deafnessvariationdatabase.org).
This study was approved by the Institutional Animal Care and Use Committees and carried out according to the Animal Experimentation Regulations of Kobe University (for DIA1‐TG mice) and Kyoto University Graduate School of Medicine (for Dia1‐KO mice). Animals were housed in specific pathogen‐free conditions using the individually ventilated cage system (Techniplast, Japan) and maintained on a 12‐h light/dark cycle. Human DIAPH1 with HindIII/NotI sites at the 5′ and 3′ ends was amplified by PCR and cloned into the HindIII/NotI site of p3xFLAG‐CMV‐10 (Sigma‐Aldrich) using an In‐Fusion HD Cloning kit (Takara Bio Inc.) and named 3xFLAG‐DIA1. 3xFLAG‐DIA1(R1204X) was made using QuickChange Lightning Site‐Directed Mutagenesis kit (Agilent Technologies). Two PCR products, 3xFLAG‐DIA1(R1204X) with a 5′ XhoI site and the SV40 polyadenylation signal (SV40 polyA) with StuI at the 3′ end, were amplified by PCR and cloned into the XhoI/StuI sites of the pCAGGS vector, which has the CAG promoter, using an In‐Fusion HD Cloning kit (Fig 7A). After confirming the identity of the plasmid by sequencing and expression of 3xFLAG‐DIA1(R1204X) in HEK293 cells by immunoblotting, a purified fragment digested with SalI and StuI that contained the CAG promoter, 3xFLAG‐DIA1(R1204X), and SV40 polyA was injected into a fertilized egg at the pronucleus stage obtained from C57BL/6 mice (Unitech Japan). Founder (F0) mice were screened by PCR using the following primer pair (named CAG/DIA1‐1): 5′‐CCTACAGCTCCTGGGCAACGTGTGCTGGTT‐3′ (in the CAG promoter) and 5′‐GTCGTTTAAGAATGTCCAATAAGGA‐3′ (in DIA1: coding nucleotides 541–565), and positive mice were then screened to determine whether they possessed the whole sequence of the injected fragment using seven primer pairs, including CAG/DIA1‐1, as listed in Table EV1. Offspring were genotyped by PCR using the CAG/DIA1‐1 primer pair. F1 and later generations of heterozygous DIA1(R1204X)‐TG mice, and F2 and later generations of homozygous DIA1(R1204X)‐TG mice were used for subsequent analyses. Dia1‐KO mice (having C57BL/6 genetic background) have been described previously (Sakata et al, 2007). Males were used in the analyses unless described (mice younger than 1 week were not differentiated based on sex). ABR measurement of DIA‐TG mice was performed using both male and female and analyzed together at the age of 5, 10 and 25 weeks and also separately at the age of 25 weeks.
Antibodies and chemicals
The following specific Abs were used (polyclonal unless indicated): monoclonal FLAG (clone M2, 1:1,000, Sigma‐Aldrich), Myo7a (25‐6790, 1:400, Proteus), and mDia1(aa 66–77) (DP4471, 1:500, ECM Biosciences), whose immunogen is 100% conserved between mouse Dia1 and human DIA1. HRP‐conjugated Abs against GST (PM013‐7, 1:5,000), α‐tubulin (PM054‐7, 1:2,000), GAPDH (M171‐7, 1:2,000), and GFP (598–7, 1:2,000) were from MBL International. Alexa568‐conjugated phalloidin (1:500) and Alexa488‐conjugated secondary Abs (1:2,000) were from Invitrogen.
Human DIAPH1 was amplified by PCR and cloned into pEGFP(C1) (Invitrogen), to give a plasmid we named GFP‐DIA1. A constitutively active mutant GFP‐DIA1(M1190D) that confers the lowest binding affinity of the DAD to the DID (Lammers et al, 2005), GFP‐DIA1(R1204X), which terminates at 1204 aa due to the 3610C>T mutation (GenBank/EMBL/DDBJ accession number: LC180357), and GFP‐DIA1(ttaa), which is the original mutant associated with DFNA1, were introduced using a QuickChange kit (Stratagene). The DID of DIA1 (137–377 aa) was amplified by PCR and cloned into the BamHI/EcoRI sites of pGEX6P‐1 and named GST‐DID. All plasmids were sequenced to confirm their identities.
MDCK and HeLa cells (RIKEN BioResource Center) were grown in DMEM supplemented with 10% FBS (Nichirei), 100 units/ml penicillin, and 100 μg/ml streptomycin (Wako), in a 5% CO2 humidified incubator at 37°C. MDCK cells with stable expression of mCherry‐ESPIN1, which induces elongation of microvilli from the normal length of 500 nm to about 2–3 μm, were obtained by the selection in the presence of 0.4 mg/ml G418 (Wako) and named MDCKmCherry‐ESPIN1 cells. Elongated microvilli were visualized and phenotypically characterized with the aid of 3D reconstruction confocal laser microscopy. Xenopus laevis XTC cells (Watanabe, 2012) were grown at 21–23°C in 70% Leibovitz's L‐15 medium (Invitrogen) with 10% FBS, without humidity control and CO2 supply.
Three‐dimensional (3D) culture of MDCK cells and confocal fluorescence imaging
To grow microvilli on the apical surface, 5 × 105 MDCKmCherry‐ESPIN1 cells were grown on 0.45‐μm polyester filter insert (12‐mm‐diameter Transwell, Corning) for 48 h to obtain complete confluency. Then, plasmids were transfected into cells using Lipofectamine LTX (Invitrogen). Twenty‐four hours after transfection, cells were fixed with 4% paraformaldehyde (PFA) in 0.1 M PBS (pH 7.4). HeLa cells were seeded in 35‐mm glass‐bottomed dishes (MatTek), 48 h before transfection, and transfected using FuGENE 6. Thirty‐two hours after transfection of plasmids, the cells were fixed using 4% PFA solution. After permeabilization with PBS containing 0.3% Triton X‐100 (PBS‐0.3T), fixed cells were stained with Alexa568‐conjugated phalloidin with/without DAPI for 1 h at 23°C. Fluorescent images (2D and 3D) were obtained using a LSM700 confocal laser‐scanning fluorescence microscope (Zeiss). For statistical analysis of the effect of DIA1 proteins on microvilli, the longest microvilli of HeLa cell from 3D reconstructed image were measured (n = 10 from five independent experiments).
Live cell imaging and fluorescence single‐molecule speckle microscopy (SiMS)
Single‐molecule speckle microscopy (SiMS) imaging was performed as described previously (Watanabe, 2012). In brief, transfected Xenopus laevis XTC cells expressing GFP‐tagged DIA1, DIA(R1204X), and DIA(M1190D) were allowed to spread on a PLL‐coated glass coverslip attached to a flow cell in 70% Leibovitz's L15 medium without serum. The flow cell was placed on the stage of a microscope (BX52; Olympus) equipped with 75‐W Xenon illumination and an EMCCD camera (Cascade II:512; Roper Scientific). Time‐lapse imaging was performed at 21–23°C using the Metamorph software (Molecular Devices) every 300 ms. Fluorescent speckle microscopy was performed by observing cells expressing a low amount of GFP‐tagged proteins using a PlanApo 100× NA 1.40 oil objective (Olympus), illuminating restricted areas near the cell edges. The number of speckles showing directional movement was counted for each cell. To roughly normalize these numbers in terms of expression level, they were divided by the values of S × (Fcell – Fbg). S is the dimension of the observed cell area (μm2). Fcell and Fbg are the fluorescence intensities of cell edge and background, respectively.
In vitro binding (pull‐down) assays
Purified GST and GST‐DID were obtained as described previously (Ueyama et al, 2007). Biotin‐labeled DAD of DIA1: DAD(WT, 1179–1215 aa), DAD(R1204X), and DAD(M1190D), was synthesized by MBL international. GST‐DID was mixed with biotin‐DAD in 300 μl binding buffer (50 nM each) (Ueyama et al, 2007). After rotation for 2 h at 4°C, 40 μl streptavidin‐coupled magnetic beads (Dynabeads M‐280 Streptavidin; Invitrogen) was added to the solution, and the mixture was agitated for 90 min at 4°C. The precipitates were washed three times using a magnetic rack, and then, the material absorbed to the beads was eluted in Laemml sample buffer; the magnetic beads were then removed using a magnetic rack. The eluents were subjected to SDS–PAGE, followed by immunoblotting using an HRP‐conjugated polyclonal Ab against GST. Bound antibodies were detected using the ECL detection system (Wako).
Dissected cochlear tissues were fixed by 4% PFA in 0.1 M PB (pH 7.4). After permeabilization with PBS containing 0.3% Triton X‐100 (PBS‐0.3T), fixed tissues were incubated with primary Ab for 2 h at 23°C in PBS‐0.03T and 0.5% fat‐free bovine serum albumin (BSA), followed by Alexa488‐conjugated secondary Ab for 2 h at 23°C, and mounted in Prolong anti‐fade (Invitrogen) with a coverslip. Immunostainings were observed under a LSM700 confocal microscope (Carl Zeiss). In the case of samples embedded in paraffin, after deparaffinization, HE staining was performed using hematoxylin solution and eosin solution (Muto Pure Chemicals). Samples were imaged under a light microscope (BX50; Olympus) with a DP26 camera (Olympus).
Dissected inner ear tissues, thymus, and heart (from P3 mice, 2–3 mice in case of the inner ear) were lysed in homogenizing buffer (Ueyama et al, 2001) by sonication in the presence of protease inhibitor cocktail (Nacalai Tesque) and 1% Triton X‐100. Total cell lysates were centrifuged at 12,000 × g for 20 min at 4°C, and the supernatants were subjected to SDS–PAGE followed by immunoblotting for 2 h at 23°C using primary Ab diluted in PBS‐0.03T containing 0.5% fat‐free BSA. The bound primary Abs were detected with secondary Ab‐HRP conjugates using the ECL detection system. For quantification, the relative expression levels of Dia1 and Dia1 + DIA1, which were detected using a mDia1(aa 66–77) Ab (DP4471, 1:500, ECM Biosciences), were normalized to that of α‐tubulin.
ABR and DPOAE measurements
To assess hearing, auditory brainstem response (ABR) and distortion product otoacoustic emission (DPOAE) were measured under anesthetization with chloral hydrate (500 mg/kg i.p., Nacalai Tesque) and on a heating pad. Blinded data analysis was performed by three otologists.
ABR was measured in DIA1(R1204X)‐TG mice and Dia1‐KO mice with their littermate control mice at the ages of 5, 10, and 25 weeks, as described previously (Kitajiri et al, 2004; Ueyama et al, 2014). ABR waveforms using sound stimuli of clicks or torn bursts at 8 kHz, 16, 24, or 32 kHz (10, 20, or 40 kHz for Dia1‐KO mice) were recorded for 12.8 ms at a sampling rate of 40,000 Hz by using 50–5,000 Hz band‐pass filter settings, and ABR waveforms from 500 stimuli were averaged. ABR thresholds (dB SPL) were defined by decreasing the sound intensity by 10 dB steps until the lowest sound intensity level resulting in a recognizable ABR wave pattern (mainly judged by recognition of wave III) was reached.
DIA1(R1204X)‐TG mice and littermate controls aged 25 weeks were measured DPOAE, as described previously (Ueyama et al, 2014). DPOAE at frequency of 2f1–f2 was elicited using two primary tone stimuli, f1 and f2, with sound pressure levels of 65 and 55 dB SPL, respectively, with f2/f1 = 1.20. DPOAE amplitudes (dB SPL) were measured at f2 frequencies of 8, 12, 16, 20 kHz and plotted after substitution with noise floor amplitude.
Mouse tail genomic DNA was extracted from using NucleoSpin Tissue (Macherey‐Nagel). Reactions were performed with negative control (no template) in a reaction mixture containing 35 ng genomic DNA, SYBR Premix Ex Taq II (Takara), and gene‐specific primers (Dia1; DIA1(R1204X); and gapdh, in Table EV1). Reactions were run at 95°C for 5 min, followed by 45 cycles of 95°C for 5 s, 60°C for 30 s, and 72°C for 20 s, using LightCycler 480 System II (Roche Diagnostics). For absolute quantification, standard curves were constructed using DIA1(R1204X) in pEGFP(C1) and Dia1 in pEGFP(C1) (Higashida et al, 2008), and the copy number of FLAG‐DIA1(R1204X) integrated into mouse genome (the ratio of FLAG‐ DIA1(R1204X)/Dia1) was calculated. The copy number of Dia1 verified by gapdh as an internal control gene using the ΔΔCT method (Livak & Schmittgen, 2001). The specificity of the amplified fragment was demonstrated by the presence of melting curves with single peaks.
Freshly dissected inner ear tissues were fixed in 2% PFA, 2.5% glutaraldehyde (GA) in 0.1 M PB. OC epithelia were dissected in the same buffer and postfixed with 1% OsO4 in H2O for 1 h. For SEM, tissues were dehydrated in an ethanol series, followed by isoamyl acetate, and dried in a freeze dryer EYELA FD‐5N (Tokyo Rikakikai). Dried samples were sputter‐coated with gold (Ion Sputter E‐1010, Hitachi) and then examined using a TM3030Plus scanning electron microscope (Hitachi High‐Technologies) at 15 kV. For statistical analysis of HC loss, the cochlea was divided three portions: basal, middle, and apical turn, as illustrated in Fig 8C.
All data are presented as the mean ± SE. Two groups were compared using unpaired two‐tailed Student's t‐test. For comparisons of more than two groups, one‐way ANOVA or two‐way ANOVA was performed and followed by Bonferroni's post hoc test of pairwise group differences. Statistical analyses were performed using Prism 6.0 software (GraphPad); P < 0.05 (*P < 0.05 and **P < 0.01) was considered statistically significant.
The paper explained
About 100 genes essential for hearing have been identified in nonsyndromic hereditary deafness: About 30 of these genes encode proteins that interact directly or indirectly with actin. DIAPH1 encodes DIA1 (DIAPH1), a formin protein that elongates unbranched straight actin and has essential roles in modeling/remodeling cytoskeletons. Autosomal dominant nonsyndromic sensorineural hearing loss, DFNA1, which is characterized by progressive deafness starting in childhood, is caused from the c.3634+1G>T mutation in DIAPH1 (p.Ala1212ValfsX22). The mutation occurs near the C‐terminus of the diaphanous autoregulatory domain (DAD), which interacts with its N‐terminal diaphanous inhibitory domain (DID), and may engender constitutive activation of DIA1 by disruption of the autoinhibitory DID‐DAD interaction. Although the first report demonstrating that mutation in DIAPH1 was associated with DFNA1 was published in 1997, the molecular mechanism underlying DFNA1 development has remained unclear.
We describe a novel patient‐derived DIAPH1 mutation (c.3610C>T, p.R1204X) in two unrelated families, which results in early termination prior to a basic amino acid motif (RRKR1204–1207) at the DAD C‐terminus. The mutant DIA1(R1204X) disrupted the autoinhibitory DID‐DAD interaction and was constitutively active. Mice expressing FLAG‐tagged DIA1(R1204X) experienced progressive deafness and outer hair cell loss at the basal turn and had various morphological abnormalities in stereocilia (such as short, fused, elongated, and sparse), which are actin‐based protrusions sensing sound stimuli.
The molecular mechanism underlying DFNA1 development has been unknown for about 20 years since the first report of DFNA1. By demonstrating that a novel mutation of DIA1 in patients and a mouse model shows progressive hearing loss beginning in the high‐frequency range, the present study provides the novel concept that constitutive activation of DIA1 leads to DFNA1. Our work also provides a conceptual framework upon which future investigations into the clinical diversity associated with DFNA1 can be built.
TU, S‐iU, and S‐iK planned the project. S‐yN, KS, HSakata, S‐iU, and S‐iK performed human studies. TU, YN, HSakaguchi, and NS performed biochemical and cellular experiments and animal studies using DIA1(R1204X)‐TG mice. TM, NW, and S‐iK performed the SiMS experiments for analysis of actin polymerization. HT, KN, DT, and S‐iK performed experiments using Dia1‐KO mice. TU, HSakaguchi, and S‐iK analyzed data, and TU wrote the manuscript.
Conflict of interest
The authors declare that they have no conflict of interest.
Expanded View Figures PDF
We thank Tatsuya Uebi, Biosignal Research Center, Kobe University, for analysis of qPCR experiments. This work was supported by JSPS KAKENHI (to TU, S‐iK, NS), the Takeda Science Foundation (to TU), the Uehara Foundation (to TU), the third step of Visionary Research Grant (a stepwise encouraging program) by Takeda Science Foundation (to S‐iK).
FundingJapan Society for the Promotion of Sciencehttp://dx.doi.org/10.13039/501100001691
This is an open access article under the terms of the Creative Commons Attribution 4.0 License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
- © 2016 The Authors. Published under the terms of the CC BY 4.0 license