The Diagnostic Utility of Single Long Contiguous Stretches of Homozygosity in Patients without Parental Consanguinity

Abstract

We present data from our clinical department's experience with chromosomal microarray analysis (CMA) regarding the diagnostic utility of 1 or 2 long contiguous stretches of homozygosity (LCSHs) in an outbred population. The study group consisted of 2,110 consecutive patients from 2011 to 2014 for whom CMA was performed. The minimum cut-off size for defining a homozygous stretch was 5 Mb. To focus on cases with no parental consanguinity, we further studied only patients in whom the total length of homozygous stretches did not exceed 28 Mb or 1% of the autosomal genome length. We identified 6 chromosomal regions where homozygous stretches appeared in at least 3 patients and excluded these from further analysis. In 2 out of 120 patients with an isolated finding of 1 or 2 non-recurrent LCSHs, a plausible candidate gene associated with their phenotype was identified within the homozygous stretch. In both of these cases, a pathogenic mutation was detected, leading to diagnoses of pyruvate kinase deficiency and Marinesco-Sjögren syndrome. To clarify whether previously found homozygous stretches could be important for the interpretation of genome-wide sequencing data, we report 7 cases in which homozygous stretches not encompassing a clinically associated gene were first found on CMA, followed by the diagnostic whole-exome sequencing. The diagnostic utility of single LCSHs, unlikely to be caused by uniparental disomy, is discussed in detail.

Chromosomal aberrations are a well-known cause of multiple congenital anomalies, intellectual disabilities (IDs), and autism spectrum disorders. Many centers use chromosomal microarrays (CMAs) as a first-tier diagnostic test for these indications because a remarkable proportion of causative genetic aberrations are in the submicroscopic range and, therefore, cannot be detected using conventional karyotyping methods [Hochstenbach et al., 2009; Miller et al., 2010; Vissers et al., 2010; Ahn et al., 2013]. In Estonia, CMAs have been funded as a first-tier diagnostic test by the Estonian Health Insurance Fund since 2011. All the diagnostic CMA analyses in Estonia are performed in 1 diagnostic laboratory (Department of Genetics, United Laboratories, Tartu University Hospital) using single nucleotide polymorphism (SNP) genotyping technology. An overview of the Estonian experience during the first years employing CMAs in clinical diagnostics was recently published by Zilina et al. [2014]. In contrast to array comparative genomic hybridization, SNP arrays also enable the detection of copy number neutral loss of heterozygosity regions, also known as regions of homozygosity or long contiguous stretches of homozygosity (LCSHs). The most widely used cut-off for the length of an LCSH is 5 Mb, as it was shown that LCSHs <4 Mb in length are common in an outbred European population [McQuillan et al., 2008]. LCSHs can result from ancestral homozygosity, uniparental disomy (UPD), or parental consanguinity [Kearney et al., 2011]. The degree of parental consanguinity can be reliably assessed based on the total length of LCSHs relative to the total autosomal genome length [Sund et al., 2013; Wang et al., 2015]. Parental consanguinity is very rare in Estonia having a small but outbred population of 1.3 million inhabitants.

In general, if well-known UPD syndromes are excluded, LCSH as a stand-alone finding rarely results in a definite diagnosis. However, the detection of an LCSH can lead to the selection of candidate genes within the stretch that cause an autosomal recessive disorder and, therefore, to further testing [Kearney et al., 2011; Sund et al., 2013; Wang et al., 2015]. A couple of published studies have investigated the diagnostic utility of LCSHs in a clinical setting, but they concentrated on UPD [Papenhausen et al., 2011] or parental consanguinity [Sund et al., 2013]. Recently, a large study of over 14,000 consecutive samples investigated the clinical utility of LCSHs caused by both UPD and parental relatedness [Wang et al., 2015]. In 9 cases, a clinically associated candidate gene was found in an LCSH, and in 7 of them, the pathogenic mutation was confirmed [Wang et al., 2015]. Whole-exome sequencing (WES) was used in another study to clarify the role of extended homozygosity in 50 patients with parental consanguinity, and it exhibited a diagnostic yield of 36% [Makrythanasis et al., 2014].

To our knowledge, no previous studies have examined the diagnostic utility of single LCSHs in patients without parental consanguinity (second cousin or closer). Therefore, the aim of this study was to fill this gap by investigating whether the detection of a single LCSH can aid in the discovery of phenotype-associated candidate genes, thereby leading to the confirmation of pathogenic mutations in routine clinical practice.

Materials and Methods

This study was approved by the Research Ethics Committee of the University of Tartu (approval date 1/13/2015 and number 243/T-3). The primary study group consisted of 2,110 consecutive patients who were analyzed with CMAs in our department from 2011 to 2014. No prenatal or non-affected parental samples were included in this study. The main indications for testing included - but were not restricted to - IDs, multiple congenital anomalies and autism spectrum disorders. In most of the cases, a CMA analysis was used as a first-tier diagnostic genetic test. Genomic DNA was extracted from peripheral blood samples. The CMA analysis was performed on all samples using HumanCytoSNP-12 BeadChips (Illumina Inc., San Diego, Calif., USA). Genotypes were called by GenomeStudio software v2010.3 (Illumina Inc.). The cnvPartition plugin (Illumina Inc.) for GenomeStudio was used to detect LCSHs, with the minimum region size set to 5 Mb. All chromosomes of each sample were also visually analyzed for quality control by a cytogeneticist, as well as to eliminate false-positive and false-negative calls.

The inclusion criteria for the subsequent analysis was the presence of 1 or 2 LCSHs with a minimal length of 5 Mb per LCSH and a total length not exceeding 28 Mb without the co-occurrence of pathogenic, likely pathogenic, or unclearly significant copy number variants. The upper limits of LCSH number and length were used to eliminate parental consanguinity, as 28 Mb equals ∼1% of the total autosomal genome, which is below the expected percentage for parents being second cousins [Sund et al., 2013]. Additionally, it is more likely that, in the case of parental consanguinity, there are multiple LCSHs on different chromosomes. We further excluded all LCSHs with lengths >25% of the whole chromosome on which they occurred to eliminate possible UPDs, as well as recurrent LCSHs present in at least 3 unrelated patients.

All the LCSHs included in the final study group were evaluated for clinical significance by searching for encompassed, autosomal recessive disease-associated genes. This was done using the web-based Genomic Oligoarray and SNP array evaluation tool v.2.0 [Wierenga et al., 2013]. The search criterion used was ‘OMIM genes with recessive inheritance pattern’. For every patient, the list of genes and associated phenotypes found by the software were compared with the patient's clinical information based on the referral documents. If a candidate gene matching the patient's phenotype was found, WES was performed to find the causative mutation. All probable pathogenic mutations identified by WES were confirmed by Sanger sequencing.

In a secondary analysis, we included all patients from the study group with LCSHs that did not encompass a good candidate gene and for whom WES was performed previously, although not as a part of this study. In case of a homozygous pathogenic mutation, the CMA data were reanalyzed to look for a <5-Mb LCSH in the region of the mutation.

Results

Out of 2,110 patients, 172 patients (8.2%) presented with an isolated finding of 1 or 2 LCSHs when taking into account the aforementioned inclusion and exclusion criteria (other than recurrence). A total of 183 LCSHs were detected, as 11 patients had 2 LCSHs. We identified 6 different recurrent LCSHs [52 (28.4%) in total] (table 1). After excluding all recurrent LCSHs, 120 patients (5.7%) with 129 LCSHs remained in the final study sample. The main characteristics of the detected non-recurrent LCSHs are listed in table 2, and their distribution among chromosomes is presented in figure 1.

Table 1.

Location of 52 recurrent LCSHs found in this study

Chromosomal bands	Maximal coordinates among the samples	Maximal size, Mb	Minimal overlapping stretch coordinates	Minimal overlapping region, Mb	No. of LCSHs (% of a total of 183)	First report of the recurrent LCSH
3p21.33p21.1	chr3:43,805,245 – 54,093,664	10.3	chr3:48,787,219 – 50,483,437	1.7	17 (9.3%)	Wang et al., 2015
6p22.3p21.31	chr6:23,404,279 – 33,887,363	10.4	chr6:26,184,041 – 30,071,279	3.9	9 (4.9%)	this report
7q11.21q11.22	chr7:61,760,894 – 68,616,110	6.9	chr7:61,834,214 – 66,904,395	5.1	5 (2.7%)	this report
7q31.2q31.33	chr7:116,266,792 – 124,008,463	7.7	chr7:117,783,250 – 122,990,979	5.2	3 (1.6%)	this report
11p11.2q12.1	chr11:44,901,903 – 57,160,838	12.3	chr11:46,342,376 – 51,274,692	4.9	11 (6%)	Wang et al., 2015
Xq13.1q21.1	chrX:71,595,785 – 83,584,844	12.0	chrX:73,424,191 – 78,848,072	5.4	7 (3.8%)	this report

All coordinates are according to the hg19 reference genome assembly.

Table 2.

Main characteristics of non-recurrent LCSHs detected in this study

Number of patients with non-recurrent LCSH (% of a total of 2,110 patients)	120 (5.7%)
Number of patients with 1 (2) LCSHs	111 (9)
Non-recurrent LCSH count	129
Size range of single LCSHs	5 – 28 Mb
Size range of the sum of 2 LCSHs in 1 patient	11.1 – 25.1 Mb
Average size of a single LCSH	7.7 Mb
Number of genes causing a recessive disorder in an LCSH (median)	0 – 15(3)
Number of cases with a candidate recessive gene in an LCSH	2a

^aIn both cases, the mutation was confirmed.

Fig. 1.

The number of non-recurrent LCSHs and their distribution among different chromosomes (total number = 129). The plot shows that homozygous stretches in our study group were located on all chromosomes except chromosome 22.

Searching for genes associated with recessive disorders revealed a median of 3 genes per LCSH (range 0 to 15). In 2 cases, an appropriate candidate gene was discovered (table 3). First, in a 1-year-old girl with transfusion-dependent hemolytic anemia, an LCSH on chromosome 1 encompassed the PKLR gene, which is known to cause pyruvate kinase deficiency leading to hemolysis and anemia. WES revealed a novel mutation in the PKLR gene that was predicted to cause a deletion of 1 amino acid. Although an enzyme analysis on the affected child was uninformative due to recurrent transfusions, the pathogenicity of the mutation was still confirmed by the detection of decreased pyruvate kinase activity in both heterozygous parents. Second, in a 6-year-old boy with growth failure, developmental delay, ataxia, and cerebellar atrophy, an LCSH on chromosome 5 was found to encompass the SIL1 gene which is associated with Marinesco-Sjögren syndrome. WES revealed a homozygous 1-bp duplication in the SIL1 gene that was predicted to cause a translational frameshift, thereby resulting in a truncated protein. The parents were confirmed to be heterozygous carriers.

Table 3.

Confirmed pathogenic mutations found by whole-exome sequencing

Patient No.	Phenotype	LCSH	Mutated gene	Affected transcript	Protein change	Disease	First report of the mutation
Patients in whom a pathogenic mutation was confirmed inside an LCSH
1	transfusion-dependent, non-immune hemolytic anemia	arr[hg19] 1q21.2q23.2(149,698,049 – 159,667,190)×2 hmz	PKLR	NM_000298.5: c.1137_1139del	NM_000298.5: p.(Lys380del)	pyruvate kinase deficiency of red cells (OMIM 266200)	this report
2	developmental delay, ataxia, cerebellar atrophy	arr[hg19] 5q31.1q31.2(132,669,425 – 139,522,328)×2 hmz	SIL1	NM_022464.4: c.947dupT	NM_022464.4: p.(Arg317Glufs*35)	Marinesco-Sjögren syndrome (OMIM 248800)	Senderek et al., 2005
Patients in whom a pathogenic mutation was confirmed outside an LCSH
3	intellectual disability, macrocephaly	arr[hg19] 1q25.3q32.1(182,559,510 – 200,136,582)×2 hmz, 14q13.3q21.2(37,290,124 – 44,790,527)×2 hmz	–
		in smaller LCSH: arr[hg19] 19q13.32q13.33(47,035,436 – 48,534,068)×2 hmz	KPTN	NM_007059.2: c.665dupA	NM_007059.2: p.(Ser223Glnfs* 18)	mental retardation, autosomal recessive 41 (OMIM 615637)	Pajusalu et al., 2015
4	oligohydramnios, congenital renal insufficiency, hypotension, calvarial hypoplasia	arr[hg19] Xp22.1p21.2(23,811,075 – 29,602,016)×2 hmz	–
		outside LCSH in 1p32.1	REN	NM_000537.3: c.[341C>T];[255G>C]	NM_000537.3: p.[(Ser114Phe)]; [(Gln85His)]	renal tubular dysgenesis (OMIM 267430)	this report

Of the patients for whom WES had been ordered outside this study, 7 had previously been found to carry an LCSH by CMA analysis that did not encompass a good candidate gene, and 2 of them had a recurrent LCSH. Of these 7 patients, 2 received a definitive diagnosis by WES (table 3). First, WES in an adult man with ID and macrocephaly (reported in detail by us [Pajusalu et al., 2015]) revealed a homozygous 1-bp duplication in the KPTN gene (19q13.32) which is associated with such a phenotype [Baple et al., 2014]. The patient also has a sister with a similar phenotype who carried the same homozygous mutation. A previous CMA analysis found 2 LCSHs in the man (17.6 Mb in 1q25.3q32.1 and 7.5 Mb in 14q13.3q21.2), but only 1 of them (in 14q13.3q21.2) was also present in his sister. However, the KPTN gene is not located in either of these LCSHs. Reanalysis of the CMA data revealed a 1.5-Mb LCSH on chromosome 19 that encompasses the KPTN gene and which is shared by both sibs. This LCSH was not reported after the initial CMA analysis interpretation due to the 5-Mb cut-off length routinely used in our laboratory. The parents of the sibs did not report consanguinity but were born in the same parish. Second, in a girl carrying an LCSH on the X chromosome, compound heterozygous mutations were found in the REN gene (1p32.1), resulting in renal tubular dysgenesis.

Discussion

During the last decade, studies demonstrated the great utility of CMA analysis in clinical diagnostics as a first-tier diagnostic cytogenetic test for patients with ID, autism spectrum disorders and multiple congenital anomalies [Hochstenbach et al., 2009; Miller et al., 2010; Vissers et al., 2010; Ahn et al., 2013]. The primary goal of CMA analysis is to find a causative microdeletion or microduplication. SNP arrays have the advantage of also revealing genotype information, which allows the detection of copy number neutral chromosomal aberrations associated with an increased risk of autosomal recessive disorders. Multiple LCSHs on different chromosomes that indicate close parental consanguinity, as well as LCSHs that cover the majority of single chromosomes and are caused by UPD, have been studied during the last few years, and their clinical implications are, therefore, better understood [Papenhausen et al., 2011; Wang et al., 2015]. Single LCSHs in patients without parental consanguinity are not infrequently found during routine diagnostics in centers where SNP array technology is in use, but their clinical significance remains unclear in most of the cases [Zilina et al., 2014]. Reporting a variant of uncertain significance (VUCS) can cause anxiety in patients and frustration in referring doctors who are not specialists in the field of medical genetics [Coughlin et al., 2012; Zilina et al., 2014]. Therefore, it is of great importance to work toward minimizing the number of reported VUCSs.

One of the most straightforward ways to decrease the number of VUCSs is to identify recurrent LCSHs and to interpret them as benign polymorphisms, which can be left unreported depending on laboratory policies. In this study, we identified recurrent LCSHs in 28.9% of all isolated LCSH findings. Thus, we can clarify the clinical significance of nearly one-third of all LCSHs. Nevertheless, caution is still needed in classifying recurrent LCSHs as benign, because there is the possibility that a shared haplotype could be mutated in both parents. Interestingly, out of 6 chromosomal regions where we identified recurrence of an LCSH, only 2 matched with the 5 regions reported by Wang et al. [2015]. This could be attributed to different study populations, which clearly indicates the need for population-specific LCSH databases.

Although homozygosity of a chromosomal region can be intuitively attributed to an increased risk of autosomal recessive disorders, the clinical significance of single LCSHs remains unclear. In this study, only patients with 1 or 2 LCSHs, with total length not exceeding 28 Mb (i.e. 1% of the autosomal genome), were analyzed. As suspected, because only a very small proportion of the whole genome was covered, a candidate gene well-associated with a patient's phenotype was very rarely identified. Nevertheless, if a well-matched candidate gene is found, the confirmation of a pathogenic mutation is likely, since we found 2 patients with a plausible candidate gene, mutations in which were confirmed in both of them. Wang et al. [2015] reported 9 patients with good candidate genes, and a pathogenic mutation was detected in 7 of them. Although the study sample of Wang et al. [2015] was almost 7 times larger than ours, and patients with parental consanguinity were also included, we can conclude that our results are consistent with theirs due to the very small number of mutation-confirmed cases and the associated probability of a large statistical error. Despite the low success rate in our study, we can still recommend looking through detected, non-recurrent LCSHs for candidate genes, as this can lead to a molecular diagnosis. Additionally, if software tools such as the Genomic Oligoarray and SNP array evaluation tool [Wierenga et al., 2013] are used, the evaluation of LCSH regions is not very time-consuming, as the median number of genes causing a recessive disorder per LCSH was only 3 in our study group.

We also referred to a previously published case, in which a shared LCSH was discovered in an adult man and his sister, and in which subsequent WES led to the discovery of a homozygous pathogenic mutation outside the shared LCSH [Pajusalu et al., 2015]. Nevertheless, reevaluating the CMA data of the sibs revealed a small 1.5-Mb LCSH in the region of the pathogenic mutation. Therefore, if there is a strong indication for recessive inheritance and distant consanguinity or common ancestry is likely, lowering the cut-off of the LCSH size can aid in the identification of candidate genes. For example, the LCSH minimum cut-off length of 1 Mb was used in a study that focused on familial recessive IDs in outbred families [Schuurs-Hoeijmakers et al., 2011]. However, one should not presume that the finding of homozygosity is a general rule, as we also demonstrated that in a girl with a distinct phenotype and an LCSH on the X chromosome, compound heterozygous mutations causing her disease were detected outside the LCSH.

In conclusion, we have reported the results of a study based on a routine clinical diagnostic laboratory's CMA data from a 4-year period. This publication will bring more attention to single LCSHs in patients with no parental consanguinity, as more studies with large sample sizes are needed to broaden the understanding of the diagnostic utility of single LCSHs.