Multi-infarct dementia of Swedish type is caused by a 3’UTR mutation of COL4A1

Sir,

Cerebral small vessel diseases (cSVDs) often present as sporadic conditions but several monogenic families have also been reported (Hagel et al., 2004; Herve et al., 2012). In 1977, Sourander and Wålinder described a family with an autosomal dominant cerebrovascular disease manifesting with transient ischaemic attacks/strokes, neuropsychiatric symptoms and progressive cognitive decline. Thirty years later it was proved that this family did not have mutations in NOTCH3, excluding the initially suspected diagnosis of cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL). Consequently, it was concluded that this family presented a new cSVD, which was named hereditary multi-infarct dementia (hMID) of the Swedish type (Low et al., 2007).

To identify the genetic cause of disease in this Swedish hMID family we performed whole-exome sequencing and genetic linkage analysis. Twenty-one family members participated in this study: 10 were affected, 10 were unaffected and one participant was an unrelated spouse used as a technical control (Fig. 1A).

Figure 1.

Pedigree of the Swedish hMID family and luciferase assay of wild-type and mutant transfected cells. (A) Pedigree of the Swedish hMID family. Family members included in the study are marked with B-DNA (blood-derived DNA) or FFPE (formalin fixed paraffin embedded tissue) according to the sample type available. Black symbols represent affected individuals and white symbols represent healthy family members. Diagonal lines indicate deceased individuals. The arrowhead indicates the proband (Subject II:5). (B) Luciferase assay. HEK-293T cells were transfected with pMIR-REPORT luciferase wild-type (WT) or mutant (mut). Cells were co-transfected either with miR-29b-3p (black bars) or negative control microRNA (white bars). Normalized luciferase activity of cells transfected with the wild-type construct was significantly decreased by miR-29b-3p, compared to cells transfected with a mimic-negative control. Luciferase activity was not altered in cells transfected with a mutated construct (^***P < 0.001, two-sided Student t-test). Error bars indicate mean standard deviation.

Blood samples (n = 17) were collected and DNA was extracted by standard methods after written informed consent was provided by all family members taking part in this study. For four patients only formalin-fixed paraffin-embedded (FFPE) tissues were available.

Four patients (Patients III:1c, IV:7, IV:16 and V:3) and two unaffected family members (Subjects IV:5 and IV:14) were selected for whole-exome sequencing. Exomes were prepared using the SeqCap EZ Human Exome Library version 2.0 (Roche Nimblegen Inc) and sequencing runs were performed on HiSeq 2000 (Illumina). Sequencing reads were aligned to GRCh37/hg19 using BWA (Li and Durbin, 2010) and variants were called according to GATK’s standard best practices v3 (McKenna et al., 2010; DePristo et al., 2011). Following variant calling, annotation was performed using SnpEff (Cingolani et al., 2012). For the linkage study, whole genome genotyping was performed for 12 blood-derived DNA samples using HumanOmniExpress Bead chips (Illumina). Parametric multipoint linkage analysis was performed using Allegro (Gudbjartsson et al., 2000) under a fully penetrant autosomal dominant model.

Data analyses were based on an autosomal dominant mode of inheritance of the disease and the hypothesis that the underlying mutation was not present in neurologically healthy control subjects or in the general population (Table 1). Validation of variants found by whole-exome sequencing was done using Sanger sequencing with BigDye® Terminator version 3.1 chemistry (Applied Biosystems).

Table 1.

Linkage regions with LOD > 2 and the variants identified by whole-exome sequencing located in these chromosomal areas

Chr	Chr location	dbSNP	LOD	Variants	Gmaf
10	7349891–76372030	rs10823837-rs4746209	2.352	SPOCK2 NM_014767.2:c.*11G >A	0.379
11	47929846–49000550	rs6485795-rs11040198	2.294	No variants identified
12	85165879–87281210	rs11116595-rs7316774	2.348	No variants identified
13	109327788–111067000	rs9284246‐rs10851243	2.407	COL4A1 NM_001845.5:c.*32G >A	–
				COL4A1 NM_001845.5:c.4470C >T	0.376

Chr = chromosome; Chr location = chromosomal location (hg19); dbSNP = dbSNP accession numbers; LOD = logarithm of the odds score; Variants = variants identified by whole-exome sequencing in the region; Gmaf = global minor allele frequency in the Exome Aggregation Consortium (ExAC). Minor allele frequencies of 0.4 in the general population for both SPOCK2 NM_014767.2:c.*11G >A and COL4A1 NM_001845.5:c.4470C >T were considered to be too high for a mutation causative of a rare disease as Swedish hMID.

Parametric multipoint linkage analysis identified four peaks on chromosomes 10, 11, 12 and 13 achieving LOD scores >2. When these regions were compared to the whole-exome sequencing data we identified three variants present in affected family members and absent in healthy relatives: one in SPOCK2 and two variants in COL4A1 (Table 1).

Only one of these variants (COL4A1:c.*32G>A) was found to segregate with the disease in the extended family and was absent from population databases. Although the variant is located in the 3’UTR of COL4A1 both gnomAD and ExAC databases report variants in this locus in a minimum of 117 613 and 44 384 individuals, respectively (Lek et al., 2016). The segregation of COL4A1 c^*32G>A with the disease was confirmed using Sanger sequencing. All affected cases had the variant and none of the older unaffected cases (age > 40 years) carried it. One younger, currently unaffected, family member also carried the COL4A1 c.^*32G > A.

The c.^*32G > A mutation is located in the 3’ UTR region of COL4A1, and may affect the binding site of miR-29, located in this region. To test this hypothesis, we performed a microRNA transfection study combined with luciferase reporter assay. A 197-bp fragment of wild-type and mutated target site was amplified by PCR from patients’ genomic DNA. The inserts were validated by sequencing. The amplified target region was digested with Hind III/Spe1, cloned into the pMIR REPORT Luciferase plasmid (Applied Biosystems). HEK293T cells [Dulbecco’s modified Eagle medium, 10% foetal calf serum with penicillin (100 U/ml) and streptomycin (100 µg/ml) in humidified air containing 5% CO₂ at 37°C] were plated in 24-well plates. At 80% confluence, 100 ng of empty, wild-type, or mutated plasmids were cotransfected with 25 pmol of either miRIDIAN™ hsa-miR-29b-3p or negative control (Dhmarcon), using DharmaFECT Duo 2.5 μl in each well (Dharmacon). The triplicate samples were lysed with 1% NP40, 150 mM NaCl and 25 mM Tris, pH 7.6, and firefly luciferase activities were measured 36 h after transfection using the 1000 Assay System (Promega) and analysed with BioTek, Cytostation 5. The results suggested that the COL4A1 c.^*32G > A mutation affects miR-29 binding and hence leads to upregulation of COL4A1 (Fig. 1B).

COL4A1 mutations have been reported as the cause of a wide variety of autosomal dominant diseases being associated with variable phenotypes (Lemmens et al., 2013). These include: porencephaly 1 (OMIM #175780); small vessel disease of the brain with or without ocular anomalies (BSVD, OMIM #607595); retinal arterial tortuosity (RATOR, OMIM #180000); hereditary angiopathy with nephropathy, aneurysms and muscle cramp (HANAC, OMIM #611773); Walker-Warburg syndrome (Labelle-Dumais et al., 2011) and pontine autosomal dominant microangiopathy with leukoencephalopathy (PADMAL) (Verdura et al., 2016).

Swedish hMID is a cSVD characterized by multifocal impaired cerebral blood flow resulting in multiple infarctions. Clinically and pathologically it fits within the expanding phenotypic group of COL4A1-related disorders, most closely resembling PADMAL with lacunar infarcts in the subcortical and pontine areas (Sourander and Walinder, 1977; Hagel et al., 2004; Verdura et al., 2016).

A recent publication by Verdura and colleagues (2016) identified mutations in COL4A1 3’UTR as the cause of cSVD in six families, including PADMAL cases. The mutations identified also affected the binding site for miR-29 micro-RNA located within the 3’UTR of COL4A1, and were shown to lead to upregulation of COL4A1 mRNA expression (Verdura et al., 2016). Although the variant found in this Swedish hMID family is novel, it disrupts the same miR-29 binding site, adding support to the pathogenicity of the mutation and suggesting that COL4A1 upregulation is a central pathogenic mechanism both in Swedish hMID and PADMAL. The similarities at clinical and pathological levels also support this view: both diseases are characterized by fibrohyalinosis and elastosis of small arterioles with atrophy of media and proliferation of the intima. These changes result in multiple lacunar infarcts in the basal ganglia, thalamus, periventricular white matter and pons, and in cortical and white matter atrophy. At the electron microscopy level, thickening of the basement membrane is observed in both diseases (Sourander and Walinder, 1977; Hagel et al., 2004; Low et al., 2007; Verdura et al., 2016), and the clinical findings include cognitive impairment and progressive dementia, strokes, as well as mood and gait disturbances (Hagel et al., 2004; Low et al., 2007; Craggs et al., 2013). In our previous study, immunostaining of COL4 revealed increased levels of staining in walls of small cerebral arteries both in PADMAL and Swedish hMID cases. However, the investigation of the sclerotic index showed some regional differences between the diseases: PADMAL seemed to affect the vessels of the frontal region more than those of the basal ganglia, whereas hMID cases showed the opposite effect (Craggs et al., 2013). Furthermore, no haemorrhages have been described in subjects with PADMAL, while one hMID subject with anticoagulative treatment was reported to suffer from a massive haemorrhage (Sourander and Walinder, 1977).

As previously proposed, perturbations of the cerebrovascular matrisome (the group of proteins both constituting and associated with the extracellular matrix) can represent a convergent pathologic pathway in monogenic small vessel diseases (Joutel et al., 2016). Still, further studies are needed to clarify the detailed pathogenic molecular mechanisms behind these diseases and to understand the phenotypic differences arising from mutations in the same micro-RNA binding site.