Article Text

Download PDFPDF

Schimke immuno-osseous dysplasia: SMARCAL1 loss-of-function and phenotypic correlation
  1. L I Elizondo1,2,
  2. K S Cho3,
  3. W Zhang4,
  4. J Yan5,
  5. C Huang5,
  6. Y Huang1,
  7. K Choi1,
  8. E A Sloan5,
  9. K Deguchi5,
  10. S Lou5,
  11. A Baradaran-Heravi1,
  12. H Takashima6,
  13. T Lücke7,
  14. F A Quiocho4,
  15. C F Boerkoel1
  1. 1
    Centre for Molecular Medicine and Therapeutics, Child and Family Research Institute, Department of Medical Genetics, University of British Columbia, Vancouver, British Columbia, Canada
  2. 2
    Interdepartmental Program in Cell and Molecular Biology, Baylor College of Medicine, Houston, Texas, USA
  3. 3
    Department of Biological Sciences, Konkuk University, Hwayang-dong, Kwangjin-gu, Seoul, Republic of Korea (South)
  4. 4
    Department of Biochemistry, Baylor College of Medicine, Houston, Texas, USA
  5. 5
    Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, Texas, USA
  6. 6
    Department of Neurology and Geriatrics, Kagoshima University Graduate, School of Medical and Dental Sciences, 8-35-1 Sakuragaoka, Kagoshima, Japan
  7. 7
    Departments of Pediatric Nephrology, Hepatology and Metabolic Diseases, Hanover Medical School, Hannover, Germany
  1. Dr C F Boerkoel, Provincial Medical Genetics Program, Department of Medical Genetics, Children’s and Women’s Health Centre of BC, 4500 Oak Street, Room C234, Vancouver, BC V6H 3N1 Canada; boerkoel{at}interchange.ubc.ca

Abstract

Background: Schimke immuno-osseous dysplasia (SIOD) is an autosomal recessive pleiotropic disorder caused by mutations in SMARCAL1. SMARCAL1 encodes an enzyme with homology to the SNF2 chromatin remodelling proteins.

Methods: To assess the affect of SMARCAL1 mutations associated with SIOD on SMARCAL1 expression and function, we characterised the effects of various mutations on mRNA and protein expression in patient tissues and cell lines, and the ATPase activity, subcellular localisation, and chromatin binding of SMARCAL1 missense mutants.

Results: The SIOD associated SMARCAL1 mutations affected SMARCAL1 protein expression, stability, subcellular localisation, chromatin binding, and enzymatic activity. Further, expressing SMARCAL1 missense mutants in Drosophila melanogaster showed that disease severity was inversely proportionate to overall SMARCAL1 activity.

Conclusion: Our results show for the first time that SMARCAL1 binds chromatin in vivo and that SIOD arises from impairment of diverse SMARCAL1 functions.

View Full Text

Statistics from Altmetric.com

Schimke immuno-osseous dysplasia (SIOD) is an autosomal recessive multisystem disorder. The prominent features of the disease are skeletal dysplasia, renal failure, and T cell immunodeficiency.14 Other features include hypothyroidism, abnormal dentition, bone marrow failure, thin hair, corneal opacities, atherosclerosis, stroke, and migraine-like headaches.2 48

SIOD has variable expressivity ranging from in utero onset with growth retardation and death within the first 5 years of life to onset of symptoms in late childhood.5 911 Among patients with severe disease, approximately half of SIOD patients have hypothyroidism, half have episodic cerebral ischaemia, and a tenth have bone marrow failure.15 12 Those with milder forms of SIOD live longer than 15 years and are not affected by hypothyroidism, recurrent infections, bone marrow failure, or cerebral ischaemia.5 9 13

Biallelic mutations of SMARCAL1 cause SIOD.14 SMARCAL1 encodes a protein homologous to the Switching defective 2 (SWI2) or Sucrose Non-Fermenting 2 (SNF2) (SWI2/SNF2) family of ATP dependent chromatin remodelling proteins.1517 SNF2 related proteins are defined by seven motifs (I, Ia, II, III, IV, V, VI) that form the nucleotide binding site and participate in nucleotide triphosphate hydrolysis and coupling of DNA binding with nucleotide triphosphate hydrolysis (fig 1).18

Figure 1 SMARCAL1 motif and mutant map. The Homo sapiens (H) SMARCAL1 amino acid sequence has 67%, 39% and 36% identity with the Mus musculus (M), Drosophila melanogaster (D) and Caenorhabditis elegans (C) proteins, respectively. Conserved motifs that define the SNF2 domain are highlighted in orange and are listed below the sequences in the grey bars. Residues highlighted in pink are additional conserved motifs proposed by Durr et al30 and are also listed below the sequences in the grey bars. The two putative nuclear localisation signals (NLS1 and NLS2), the Walker A (WA, phosphate binding) site and Walker B (WB, magnesium binding) site are conserved among species and are represented below black lines. Residues highlighted in blue represent the two HARP domains, whose functions are unknown. Red arrowheads indicate the location of SIOD-associated missense mutations studied in this report. Patient numbers are indicated above the arrowhead. Identical amino acids are indicated by a colon, gaps by a dash and stop codons by an asterisk.

We initially hypothesised that SIOD patients with milder disease carried a missense mutation on each SMARCAL1 allele, whereas patients with more severe disease carried at least one deletion, nonsense or frameshift mutation.14 However, we also identified a few patients with severe disease and biallelic missense mutations affecting conserved amino acids within the SNF2 domain (fig 1).14 From this we hypothesised that loss of SMARCAL1 function causes SIOD and that disease severity was inversely proportionate to residual SMARCAL1 activity.

We show for the first time herein that SIOD arises from loss of SMARCAL1 function. Deletion, nonsense and frameshift mutations generally cause a loss of detectable SMARCAL1 mRNA and protein, whereas missense mutations alter subcellular localisation, enzymatic activity, protein levels or chromatin binding. Finally, we find that disease severity is inversely proportionate to overall SMARCAL1 activity.

PATIENTS AND METHODS

Human subjects

Patients referred to this study gave informed consent approved by the Institutional Review Board of Baylor College of Medicine (Houston, Texas, USA) or the Hospital for Sick Children (Toronto, Ontario, Canada). The clinical data for patients were obtained from questionnaires completed by the attending physician as well as from medical records and summaries provided by that physician. Samples for immunohistochemical analyses were obtained from kidney biopsies of consenting SIOD patients. The disease severity of each patient was scored as previously described.19

Promoter analysis

Primers were designed to contain regions from 830 bp upstream of exon 1 inclusive of exons 1, 2, or 3 (supplemental table 1). After polymerase chain reaction (PCR) amplification (Platinum Pfx polymerase, Invitrogen, Carlsbad, California, USA), the products were digested with KpnI and BglII and cloned into a pGL3 basic vector (Promega, Madison, Wisconsin, USA). These plasmids, the pGL3 basic vector (negative control) and the pGL3 control plasmid (positive control) were transfected into 1×106 NIH3T3 cells with Lipofectamine 2000 (Invitrogen). Luciferase activity was assayed in three independent samples according to the manufacturer’s protocol (Luciferase Assay System, Promega, Madison, Wisconsin, USA).

Table 1 Schimke immuno-osseous dysplasia (SIOD) associated SMARCAL1 mutations tested

Reverse transcriptase-PCR

Using RNA extracted from patient derived lymphoblastoid cells, we measured the relative amount of SMARCAL1 mRNA compared to mRNA derived from unaffected lymphoblastoid cells as described by Clewing et al.19

SMARCAL1 expression plasmids

The construction of all SMARCAL1 expression plasmids used in this study is described in the supplemental methods.

Transgenic flies

Transgenic lines of Drosophila melanogaster were generated for each mutation by injecting pUAST constructs into the embryo using Genetic Services (Sudbury, Massachusetts, USA). The transgenic flies were crossed to P{GawB}BxMS1096 transgenic flies to express SMARCAL1 in the developing and mature fly wing.20 Transgenic flies expressing SMARCAL1 were scored for the presence or absence of ectopic wing veins. The populations of flies were compared using Student’s t test.

Cell culture

Patient lymphoblastoid cell lines

Isolation and culturing of lymphoblastoid cell lines were performed as described in Clewing et al.19

T-Rex-293 Flp-In System

The Flp-In T-REx-293 host cell line (Invitrogen) expresses the Tet repressor, contains a single integrated Flp recombination target (FRT) site, and is resistant to blasticidin and zeocin. Upon FRT recombination between the Flp-In T-Rex-293 host cell line and pcDNA5/FRT/TO-E/pUniD/V5-His6-SMARCAL1 constructs, the host cells lose zeocin resistance while gaining hygromycin B resistance from the pcDNA5/FRT/TO-E expression vector. Clones expressing V5/His6 tagged SMARCAL1 upon the addition of tetracycline were used in these studies. The creation of the tetracycline inducible T-Rex-293 cell lines containing the various V5/His6-tagged SMARCAL1 mutations is described in the supplemental data.

Immunofluorescence

T-Rex-293 cells were plated onto coverslips in six well plates at a density of 2×105 cells and were induced with tetracycline to express the V5/His6-tagged SMARCAL1 mutants. Immunofluorescence on the cells was performed as described by Deguchi et al.21 Rabbit anti-V5 (Sigma, St Louis, Missouri, USA) and mouse anti-tubulin were used at 1:200 and 1:500 dilutions, respectively. Alexa-555 goat anti-rabbit and alexa-488 goat anti-mouse secondary antibodies were used at a 1:500 dilution in blocking solution.

Polytene chromosomes of third instar larvae were prepared and processed as described,22 and immunostaining of the polytene chromosomes was performed according to the standard procedures.23 We used rabbit anti-V5 antibody as the primary antibody. DNA was stained with 4′,6-diamidino-2-phenylindole (1:1000, Fluka, St Louis, Missouri, USA).

Imaging

Images were acquired and analysed using a Zeiss Axiovert 200 microscope, a Zeiss AxiocamHR (brightfield) or a Zeiss AxiocamMR camera (fluorescence), and the Zeiss Axiovision 4.0 software imaging system.

ATPase assay

ATPase activity of the SMARCAL1 mutants was measured using the Kinase-Glo Luminescent Kinase Assay (Promega, Madison, Wisconsin, USA). The ATPase reaction was performed using 20 ng/μl of SMARCAL1 (extracted and purified as described in supplemental methods), 75 nM hairpin loop DNA, and ATPase Buffer (20 mM KPO4, 2 mM MgCl2, 40 mM KCl, 1 mM DTT, 100 μg/ml BSA, 100 μM ATP). The hairpin loop DNA was prepared by heating to 70°C for 3 min, followed by slow cooling to room temperature over time. Reactions were incubated at 37°C for 1 h. Kinase-Glo buffer was added to each reaction mixture and incubated for 10 min at room temperature. ATPase activity was measured by luminescence.

RESULTS

Characterisation of SMARCAL1 gene deletion

Previously we observed that two individuals (SD24 and SD31) had large deletions of the 5′ end of the SMARCAL1 gene.14 Both individuals were from the Indian subcontinent.19 To determine if SD24 and SD31 had precisely the same deletion, we cloned and sequenced the deletion breakpoints following PCR amplification across the junction. Both were found to have biallelic 14.6 kb deletions, which begin 4.3 kb upstream of exon 1 and extend into intron 5 (fig 2A). The 5′ break point lies between LTR67 and MER19A repeat elements. The 3′ break point lies within a MIR repeat. Despite the proximity of these DNA repetitive elements to the break points, BLASTn did not detect homology between the sequences flanking the breakpoints. Therefore, homologous recombination appears to be an unlikely mechanism for this deletion.

Figure 2 Analysis of SMARCAL1 deletion, nonsense and frameshift mutations. (A) Identification of the breakpoints in patients SD24 and SD31 who had a deletion of the SMARCAL1 promoter and 5′ exons. As shown by the DNA sequencing chromatograms, the deletion extends for 14.6 kb from –4.3 kb into intron 5. (B) Promoter activity assay. Following cloning into the pGL3 basic luciferase expression vector and transfection into NIH3T3 cells, two promoter fragments (promoter fragment A: −831 bp to exon 1; promoter fragment B: −831 bp to exon 2) were used to express luciferase. The pGL3 vector alone was transfected into NIH3T3 cells as a negative control, and the pGL3 luciferase expression plasmid was used as a positive control. (C) Analysis of SMARCAL1 steady state mRNA and protein levels in patient lymphoblastoid cell lines. Upper panel: mRNA expression as assayed by semiquantitative reverse transcriptase polymerase chain reaction (RT-PCR). Lower panel: protein expression as assayed by Western analysis. The value of 1 on the y axis represents wild type SMARCAL1 mRNA or protein levels, respectively. Error bars represent ±1SD. (D) Immunohistochemistry of SMARCAL1 expression in patient kidney biopsies. A kidney biopsy from a 4-year-old unaffected individual was used as a positive control to show SMARCAL1 protein expression. (E) Immunofluorescent determination of the subcellular localisation of V5/His6 tagged SMARCAL1 E848X mutant. (F) Western analysis showing expression of V5/His6 tagged wild type and E848X SMARCAL1 without (−) and with (+) tetracycline induction in T-Rex-293 cells. (G) ATPase activity of SMARCAL1 E848X mutant. The value of 1 on the y axis represents wild type SMARCAL1 ATPase activity. The positive control represents the ATPase activity of calf intestinal phosphatase (CIP). Error bars represent ±1SD.

To confirm functionally that the deleted region contains the SMARCAL1 promoter, we cloned two segments of the SMARCAL1 gene: (A) −830 bp to c.74 (exon 2); and (B) −830 bp to c.16 (exon 1) upstream of a luciferase reporter gene. We measured luciferase activity following transfection of the reporter plasmids into NIH3T3 cells. Both fragments gave comparable luciferase activity (fig 2B). In silico analysis of the 830 bp upstream of the SMARCAL1 transcription start site showed that this sequence is 89% and 98% identical to that of macaques and chimpanzees, respectively. There was significantly greater divergence between mouse and human sequences in that region with only 19% sequence identity, and while humans have two non-coding 5′ exons, mice have only one non-coding 5′ exon. In all the sequences examined, the SMARCAL1 promoter region does not contain a TATA box and thus appears to be a TATA-less promoter.24

Confirming that this deletion removes the SMARCAL1 promoter, RT-PCR and Western analysis of lymphoblastoid cell lines from individual SD31 did not detect SMARCAL1 mRNA or protein expression (fig 2C, Del Ex 1–5). Furthermore, analysis of a kidney biopsy from SD24 did not reveal any protein by immunohistochemistry (fig 2D, Del Ex 1–5).

Characterisation of SMARCAL1 nonsense and frameshift mutations

SMARCAL1 alleles encoding premature stop codons or frameshift mutations resulting in premature stop codons do not express detectable protein. SD25 was compound heterozygous for the nonsense mutations Q34X and R17X and did not express detectable protein in kidney biopsy tissue (fig 2D, Q34X/R17X). SD44 was compound heterozygous for S774X and L397fsX40 and also did not express detectable SMARCAL1 protein in kidney biopsy tissue (fig 2D, S774X/L397fsX40). With the exception of cells from patients with the E848X mutation, analysis of lymphoblastoid cell lines with SMARCAL1 nonsense or frameshift mutations affecting both alleles detected neither mRNA nor protein by RT-PCR and Western analysis (fig 2C, R774X/V641fsX51; E398X). SMARCAL1 mRNA encoding the E848X nonsense mutation is expressed at normal levels (fig 2C, E848X). This suggests that this nonsense mutation, which lies 84 bp upstream of the splice donor site in the penultimate exon, is too close to the C-terminal tail to induce nonsense mediated RNA decay. However, the protein levels are nearly undetectable (fig 2C, E848X), indicating that the E848X nonsense mutation either lacks an element enhancing protein translation or that the carboxyl terminal region is necessary for protein stability. Consistent with the latter, a stable T-Rex-293 cell line expressing V5/His6-tagged E848X SMARCAL1 from a tetracycline inducible cytomegalovirus (CMV) promoter showed prominent cytoplasmic localisation of the mutant protein despite intact nuclear localisation signals (fig 2E). Also, consistent with the protein being improperly folded, the purified V5/His6 tagged E848X SMARCAL1 protein hydrolysed ATP about half as well as wild type SMARCAL1 in the presence of hairpin DNA (fig 2F, G). The findings also show that ATPase activity is insufficient for nuclear retention of SMARCAL1.

Functional analysis of SMARCAL1 missense mutations

In contrast to our findings in tissues from most patients with biallelic nonsense or frameshift mutations, we observed SMARCAL1 mRNA and protein expression in patients with missense mutations (fig 2C, S625X/R820H, K647Q, K647T, R645C/I548N, and R586W; and fig 2D, R644W/Q568fsX3, T705I/A468P). This observation suggested to us that these missense mutations might impair specific aspects of SMARCAL1 function.

RNA and protein steady state levels

To determine if missense mutations identified in SIOD patients affect the steady state abundance of SMARCAL1 mRNA or protein levels, we analysed lymphoblastoid cells derived from human patients. We observed lower steady state levels of SMARCAL1 mRNA with the R820H, K647Q, and R586W mutations relative to wild type SMARCAL1 levels (fig 2C) and an increased ratio of protein to mRNA for mutations R820H, K647Q, and K647T (table 1). We also observed increased mRNA levels with mutations R645C/I548N with respect to wild type levels, but because this lymphoblastoid cell line carries different mutations on each allele, we cannot conclude which mutated allele causes an increase in mRNA levels.

ATPase activity

Many of the missense mutations affect conserved amino acids predicted to be in the ATPase active site or involved in the coupling of DNA binding and ATP hydrolysis (fig 1).14 Therefore, we hypothesised that some of these missense mutations might alter ATPase activity or uncouple ATP hydrolysis and DNA binding. To test this, we generated stable T-Rex-293 cell lines expressing wild type or mutant V5/His6 tagged SMARCAL1 from a tetracycline inducible CMV promoter, purified the V5/His6-tagged SMARCAL1 proteins following induction with tetracycline and tested the ability of each mutant SMARCAL1 enzyme to hydrolyse ATP in the presence and absence of hairpin DNA. Missense mutations R820H, R644W, K647T, R586W, and R645C decreased DNA dependent ATP hydrolysis (fig 3A, table 1). In contrast, missense mutations T705I, R764Q, S579L, K647Q, and I548N increased DNA dependent ATP hydrolysis (fig 3B, table 1). None of the mutations caused DNA independent ATP hydrolysis (data not shown), and ATPase activity did not correlate with steady state protein or mRNA levels measured in patient lymphocytes (table 1). This suggests that the ATPase activity does not feedback directly on SMARCAL1 mRNA or protein production or degradation.

Figure 3 ATPase activity and subcellular localisation of SMARCAL1 missense mutations. (A) ATPase activity of SMARCAL1 mutants with decreased activity. The value of 1 on the y axis represents wild type SMARCAL1 ATPase activity. The negative control represents the ATPase activity of ATPase buffer alone. The positive control represents the ATPase activity of CIP. Error bars represent ±1SD. (B) ATPase activity of SMARCAL1 mutants with increased activity. The value of 1 on the y axis represents wild type SMARCAL1 ATPase activity. The negative control represents the ATPase activity of ATPase buffer alone. The positive control represents the ATPase activity of CIP. Error bars represent ±1SD. (C) Immunofluorescent determination of the subcellular localisation of V5/His6 tagged SMARCAL1 SNF2 mutants with decreased ATPase activity. (D) Immunofluorescent determination of the subcellular localisation of V5/His6 tagged SMARCAL1 SNF2 mutants with increased ATPase activity. (E) Subcellular localisation of the green fluorescent protein (GFP) tagged nuclear localisation signal within the SNF2 domain (NLS2) of wild type and mutant SMARCAL1. (F) Western analysis showing expression of V5/His6 tagged wild type and mutant SMARCAL1 without (−) and with (+) tetracycline induction in T-Rex-293 cells. (G) Subcellular localisation of V5/His6 tagged SMARCAL1 NLS2 mutants by immunofluorescence.

Subcellular localisation

To determine whether ATPase activity was necessary for SMARCAL1 nuclear localisation and subnuclear distribution, we induced wild type or mutant V5/His6 tagged SMARCAL1 in the T-Rex-293 cell lines with tetracycline and analysed the subcellular localisation of these SMARCAL1 mutants. Generally, those mutants with increased ATPase activity relative to wild type (I548N, S579L, K647Q, T705I, R764Q) had complete nuclear localisation of SMARCAL1, whereas mutants with decreased ATPase activity (A468P, R820H, K464R) had prominent cytoplasmic localisation (fig 3C,D). However, R586W had complete nuclear localisation despite minimal ATPase activity; therefore, ATPase activity was not necessary for nuclear retention of SMARCAL1.

A subset of SMARCAL1 mutations alter a putative bipartite nuclear localisation signal (NLS) within the SNF2 domain of the SMARCAL1 protein at 644–661 amino acids (RRLKSDVLSQLPAKQRKI) (fig 1).15 Bipartite NLSs are characterised by KRXKKKK and mutations that alter lysine (K) and/or arginine (R) residues, particularly those at the N-terminus, often impair nuclear targeting.25 Therefore, we hypothesised that SMARCAL1 mutations K647Q, K647T, R644W and R645C, which alter this putative NLS, impaired nuclear targeting. To determine whether this sequence functions as an NLS, we fused amino acids 630–670 to the carboxyl terminus of green fluorescent protein (GFP) in the pcDNA3.1/NT-GFP-TOPO expression vector; these amino acids mediated transport of GFP to the nucleus (fig 3E). However, although mutations K647Q and R644W resulted in less nuclear GFP expression, none of the mutations abrogated the ability of the NLS to target GFP to the nucleus (fig 3E). We therefore tested these NLS mutations in the context of the full length SMARCAL1 protein (fig 3F). We observed that the NLS mutants K647Q and K647T accumulated in the nucleus, whereas R644W and R645C accumulated in the cytoplasm (fig 3G). Since the mutant NLS sequences retain at least some ability to target GFP to the nucleus, these results suggest that the altered localisation, which manifested only in the context of the full length SMARCAL1 protein, may result from other protein specific alterations such as errant protein folding.

Among mutant SMARCAL1 proteins entering the nucleus, our analyses also identified some differences in the subnuclear distribution (fig 3). In contrast to the diffuse nuclear staining observed with wild type SMARCAL1, mutants A468P, I548N, S579L, R586W, K647Q, T705I, R764Q, and R820H localised to or aggregated within discrete nuclear domains. Only the mutant K647T had a nuclear staining pattern similar to that observed with wild type SMARCAL1.

Polytene chromosome binding of SMARCAL1 mutant proteins

Because of the altered intranuclear localisation observed with some of the SMARCAL1 mutants, we hypothesised that these mutants might not bind chromatin effectively. To test this, we generated transgenic D melanogaster lines expressing wild type or mutant V5/His6 tagged SMARCAL1 and looked for binding to the polytene chromosomes (fig 4A,B). Using the GAL4-pUAST system to express the proteins,26 we observed that all these mutant proteins bound the fly polytene chromosomes to some degree (fig 4A). Wild type SMARCAL1 binds to the interbands; the interbands are composed primarily of open chromatin (weak Dapi staining) while the bands are predominately compact chromatin (strong Dapi staining) (fig 4A). Of the mutants with altered subnuclear localisation, A468P, S579L, R586W, T705I, R764Q, and R820H show diminished staining of the polytene chromosomes; this suggests that these mutants bind chromatin in vivo less efficiently.

Figure 4 Expression of V5/His6 tagged SMARCAL1 mutants in Drosophila melanogaster. (A) Polytene chromosomes from various transgenic D melanogaster lines expressing V5/His6 tagged SMARCAL1 mutants stained with anti-V5 antibodies (red) and Dapi (blue). (B) Western analysis showing the expression of V5/His6 tagged wild type and mutant SMARCAL1 using the GAL4-pUAST system in D melanogaster. (C) Overexpression of SMARCAL1 in D melanogaster using the GAL4-pUAST system causes the formation of an ectopic wing vein extending from the posterior cross vein in ∼70% of flies. The graph shows the proportion of wings with the ectopic vein in each transgenic SMARCAL1 fly population. Mutations associated with more severe disease induced statistically significant fewer ectopic veins than mutations associated with mild disease (**p<0.01).

Correlation of SIOD phenotype with SMARCAL1 function (supplemental table 2)

The preceding analyses suggested that the SIOD associated mutations of SMARCAL1 cause loss of function by multiple mechanisms. All patients with milder disease expressed at least one SMARCAL1 allele encoding a protein product localising within the nucleus; however, no other observations were predictive of the phenotypic features or disease course. Therefore, to assess further the relationship to patient phenotype, each mutant was expressed in D melanogaster (fig 4B). Overexpression of wild type SMARCAL1 in D melanogaster using the GAL4-pUAST system26 causes the formation of an ectopic wing vein extending from the posterior cross vein in ∼70% of flies (fig 4C). We found that the proportion of wings with the ectopic vein was inversely correlated with the severity of the patient phenotype—that is, mutations associated with more severe disease had less ability to induce the ectopic veins (fig 4C, table 1). The one exception to this association was mutation R645C, which was identified in an individual with the milder disease. This individual was a compound heterozygote for R645C and I548N, which is a mutation able to induce ectopic wing veins. Consistent with the milder phenotype arising from the I548N mutation, we have identified a severely affected patient homozygous for R645C.19 In conclusion, our in vivo assay of SMARCAL1 mutations correlates well with the patient phenotype and is suggestive of milder disease being associated with hypomorphic mutations rather than null mutations.

DISCUSSION

Among SIOD associated SMARCAL1 mutations analysed in this study, we find that a loss of SMARCAL1 function can occur by multiple mechanisms. With the exception of the E848X mutation that may impair SMARCAL1 protein localisation and stability, SIOD associated SMARCAL1 nonsense and frameshift mutations lead to loss of SMARCAL1 mRNA, which is likely mediated by nonsense mediated RNA decay.27 This notably reduced SMARCAL1 protein expression correlated with severe disease and is consistent with our prior observation that patients with biallelic deletion, nonsense or frameshift mutations had more severe disease.14

SIOD associated SMARCAL1 missense mutations impair SMARCAL1 function by altering subcellular localisation, protein stability, ATPase activity, or chromatin binding. SMARCAL1 in vitro DNA dependent ATPase activity did not reflect the nuclear localisation or chromatin binding in vivo or correlate with the severity of disease in SIOD patients. However, the overall function of SMARCAL1 was disrupted in all mutants studied as exhibited by our studies of SIOD associated SMARCAL1 mutations in Drosophila. In these studies, the overexpression of SIOD associated SMARCAL1 mutants suppressed induction of ectopic wing veins in a manner that was inversely proportionate to the severity of the patient disease, which is consistent with SIOD arising from a net loss of SMARCAL1 overall function.

As is being observed for an increasing number of diseases, our observations of SIOD patients do not confirm a direct genotype–phenotype correlation between individual SMARCAL1 mutations and SIOD pathophysiology. First, individuals with identical missense mutations can manifest different disease features.5 19 Second, siblings with the same mutant alleles can exhibit different forms of SIOD such that one has severe disease and the other mild or clinically undetectable disease.28 29 Third, only half of documented SIOD patients have biallelic mutations in SMARCAL1.19 Therefore, to establish a genotype–phenotype correlation of SIOD in a complex biological system will require a controlled genetic background and further studies in model systems such as D melanogaster should be explored. A genotype–phenotype correlation of SIOD will require further studies delineating modifiers of the disease, the precise biochemical function of SMARCAL1, the biological processes it influences, and the impact missense mutations have on these functions.

Recently, the SWI2/SNF2 domains of zebrafish Rad54 and of Sulfolobus solfataricus SSO1653 have been crystallised.30 31 Both structures revealed two RecA-like domains on either side of the minor groove of DNA, with movement controlled by a hinge region. These two RecA-like domains are made up of the SNF2 domain: motifs I, Ia, II and III make up RecA-like domain 1, while motifs IV, V, and VI make up RecA-like domain 2. Motifs Ia and IV are predicted to bind DNA while residues within the interfacial binding cleft bind ATP. After the binding of ATP, the seven helicase motifs align along the ATP binding site permitting ATP hydrolysis. The “power stroke” created upon ATP hydrolysis allows the enzyme to track along DNA one base at a time, similar to some helicases. In the model presented by Lewis et al, ATP hydrolysis occurs immediately after ATP binding30 32 with the release of ADP allowing the SNF2 domain to track along DNA.32

We have compared the SNF2 of SMARCAL1 to the SNF2 domains of zebrafish Rad54 and of S solfataricus SSO1653 and we predict a similar structure exists (WZ and FAQ, unpublished data). If these models reflect the mechanism of SMARCAL1 (fig 5A), mutations within the SMARCAL1 SNF2 domain could disrupt DNA binding, ATP binding, interactions between the two RecA-like domains, conformational changes, or other intra-domain interactions. The SMARCAL1 missense mutants with decreased in vitro DNA dependent ATPase activity can be located throughout the SNF2 domain. According to the SNF2 structure reported by Durr et al,30 mutants R644W, R645C, and K647T fall near to or within the hinge region, a segment required for SNF2 domain conformational changes. We predict that these mutants disrupt the mobility of the hinge region and prevent efficient clamping of SMARCAL1 on the DNA (fig 5B). Mutant R586W falls within the RecA-like domain 1, and since it would affect either the interface between domain 1 and DNA or between the two RecA-like domains, it likely disrupts the interaction between the SNF2 domain and DNA (fig 5C). Finally, mutant R820H falls within or just outside of the RecA-like domain 2 and we predict that it alters the SMARCAL1 structure or SMARCAL1 protein interactions in a manner that ultimately hinders ATPase activity.

Figure 5 Model for SMARCAL1 ATPase activity and its role in Schimke immuno-osseous dysplasia (SIOD). (A) In accordance with the model proposed by Durr et al,30 the SNF2 domain of SMARCAL1 binds to dsDNA. Motifs I, Ia, II and III make up RecA-like domain 1, while motifs IV, V, and VI make up RecA-like domain 2. A hinge region separates these two domains. The DNA is bound by motifs Ia of the first RecA-like domain and IV of the second RecA-like domain and ATP is bound within the binding cleft, which is immediately hydrolysed.32 When ADP is released, the SNF2 domain tracks along the DNA. We hypothesise that when SMARCAL1 is expressed at a functional level in vivo, normal gene expression allows for normal tissue development. (B) Mutations, represented by a star, within the hinge domain can alter the mobility of the two RecA-like domains and binding of DNA and ATP. This in turn would cause a decrease or loss in ATP hydrolysis and thereby a disruption of chromatin structure leading to SIOD. (C) Mutations in various motifs of the SNF2 domain facing the interfacial cleft could disrupt the interactions between the two RecA-like domains and with ATP or DNA thus leading to decreased ATP hydrolysis, which can then alter chromatin structure leading to SIOD. (D) In contrast, some mutations within the interfacial cleft could reinforce interactions between the two RecA-like domains and with ATP or DNA thus leading to increased ATP hydrolysis. Despite an increase in activity, though, these mutations would also lead to altered chromatin structure and thereby to SIOD.

Key points

  • Generally, Schimke immuno-osseous dysplasia (SIOD) associated SMARCAL1 nonsense and frameshift mutations lead to loss of SMARCAL1 mRNA.

  • SIOD associated SMARCAL1 missense mutations impair SMARCAL1 subcellular localisation, protein stability, ATPase activity, or chromatin binding.

  • SIOD disease severity is inversely proportionate to overall SMARCAL1 activity.

For the missense mutants with increased in vitro DNA dependent ATPase activity, the modelling predicts an increased affinity of SMARCAL1 for DNA and this might impede its ability to track along the DNA or release from the DNA. K647Q localises to the hinge region and may stabilise a hinge conformation that facilitates DNA dependent ATP binding and hydrolysis. Mutants S315R, I548N and S579L fall within the RecA-like domain 1 and may stabilise the interaction between RecA-like domains 1 and 2 to promote a “closed” conformation (fig 5D). Additionally, according to the structure by Durr et al,30 mutant I548N falls within a loop-helix motif that binds the 5′-3′ DNA strand; therefore, in addition to stabilising interaction of domains 1 and 2, this mutation could also stabilise the interaction with DNA. Mutants T705I and R764Q fall within the RecA-like domain 2 of the SNF2 domain which rotates via the hinge region such that all seven motifs of the SNF2 domain are located within the active interfacial cleft along the DNA.30 Consequently, T705I and R764Q could increase ATPase activity by altering the affinity of DNA binding and/or the interaction between RecA-like domains 1 and 2 (fig 5D). Final analysis of the effect of SMARCAL1 missense mutations requires crystallisation of the SMARCAL1 protein and full definition of its biochemical function. Despite this apparent gain of in vitro ATPase activity, the overall function of SMARCAL1 in vivo is lost as was demonstrated by the lack of ectopic wing vein induction when compared to wild type SMARCAL1 (fig 4C). Therefore, although it appears that some mutations cause a gain in SMARCAL1 enzymatic activity, the result remains an overall loss of SMARCAL1 activity.

In summary, this report confirms that SMARCAL1 binds chromatin in vivo and that the SMARCAL1 mutations associated with SIOD effect loss of function via affects on mRNA production and stability, protein production and stability, subcellular localisation, enzymatic activity and chromatin binding. Additionally our in vivo analyses confirm that disease severity is inversely proportionate to overall SMARCAL1 activity.

Acknowledgments

The authors thank Drs Jan M Friedman, Jennifer Northrop, Joanna Lubieniecka and Millan Patel for critical review of this manuscript. We thank Barbara A Antalffy and Pauline Grennan for preparation of tissue.

REFERENCES

View Abstract

Supplementary materials

Footnotes

  • ▸ Additional material is published online only at http://jmg.bmj.com/content/vol46/issue1

  • Funding: This work was supported in part by a Ruth L Kirschstein National Research Service Award (LIE) and grants from the March of Dimes (CFB), the Gillson Longenbaugh Foundation (CFB), the Dana Foundation (CFB) and the New Development Award, Microscopy, and Administrative Cores of the Mental Retardation and Developmental Disabilities Research Center at Baylor College of Medicine (CFB), the Burroughs Wellcome Foundation (CFB), the National Institute of Diabetes, Digestive, and Kidney Diseases, NIH (CFB), and the Association Autour D’Emeric et D’Anthony (CFB).

  • Competing interests: None.

  • Patient consent: Obtained.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.