Introduction

LMNA variants can cause a plethora of phenotypes that are collectively called laminopathies [1]. At least 12 different types can be distinguished [1], presenting as multi-organ systemic diseases or as tissue-specific diseases. Systemic diseases are often due to sporadic variants that lead to severe clinical symptoms such as Hutchinson Gilford Progeria Syndrome (HGPS) causing premature ageing [2]. In rare cases, compound heterozygous variants can lead to a recessive Progeria phenotype [3]. Tissue-specific laminopathies are much more common, and include Emery Dreifuss Muscular dystrophy (EDMD), Dunnigan type familial partial lipodystrophy (FPLD2), limb girdle dystrophy, and dilated cardiomyopathy (DCM). Genotype-phenotype correlations have not yet been fully resolved. On the one hand, different tissues can be affected in laminopathies arising from a single point variant, leading to overlapping syndromes [4]. Even within a single family, different phenotypes can be found. As example, patients carrying an identical LMNA variant were diagnosed as having DCM, DCM with Emery-Dreifuss muscular dystrophy (EDMD)-like symptoms and DCM with limb girdle muscular dystrophy (LGMD)-like symptoms [5].

The genetic heterogeneity of DCM in combination with phenotypic heterogeneity and variable penetrance of LMNA variants complicates LMNA variant classification. Following the guidelines of the American College of Medical Genetics and Genomics [6], LMNA variants are often classified as variant of unknown significance (VUS) in the absence of segregation and/or functional data from the literature. A VUS neither explains, nor excludes a clinical diagnosis, which poses challenges for patient counseling. In order to study the functional effect of LMNA variant, a number of groups have assessed the nuclear structure of fibroblasts from LMNA-mutated persons and linked nuclear abnormalities in dermal fibroblasts with the type of variant, and disease phenotype [7,8,9,10,11,12,13]. While most laminopathy cell cultures showed nuclear abnormalities, such as nuclear herniations (blebs) and/or irregularities in lamin staining, seen as honeycomb structures, it was difficult to correlate these nuclear abnormalities with the type of the variant, location in the protein or disease status [13]. In this paper we describe the development of a simple analysis procedure using lamin A/C immunostaining and DAPI staining to assess nuclear morphology. Combined with linear regression analysis, it can be determined whether a fibroblast culture qualifies as laminopathy and herewith this method will aid to the classification of LMNA variants based on nuclear morphology analysis of the patients’ dermal fibroblasts.

Materials and methods

Cell culture

Human dermal fibroblast cultures were obtained from dermal biopsies after written informed consent as described previously [14]. Anonymous control fibroblasts were obtained from healthy individuals or individuals with a variant in a gene that is not expressed in fibroblasts, e.g., SCN5A or MYH7. Whenever possible, fibroblasts were stained at a low passage number (mean 3.4 ± 1.1, range p2–p7). Cells were seeded onto 18 mm round glass coverslips (Menzel) without coating and grown for 48 h to a confluence of 50–75%. Cells were fixed in 100% methanol for 15 min at −20 °C and were stored at 4 °C in PBS containing 0.01% Na-azide until use.

Genetic analysis

Sanger sequencing or next-generation sequence analysis and interpretation of LMNA (NG_008692.2; NM_170707.3), SCN5A (NG_008934.1; NM_198056.2) and MYH7 (NG_007884.1; NM_00257.3); EMD (NG_08677.1; NM_000117.2) gene variants was performed as described previously [15]. A VUS or (likely) pathogenic variant in LMNA was excluded in control fibroblasts. Initial classification of identified LMNA variants was performed in line with the ACMG standards and guidelines [6]. All reported variants have been uploaded into the publically available LOVD database (https://databases.lovd.nl/shared/genes/LMNA), patient IDs: 165,050; 165,051; 165,052; 165,045; 165,042; 165,053; 165,054; 165,044; 164,979; 164,980; 164,981; 165,048; 165,003; 165,047; 164,804; 165,023; 164,810; 165,005–165,009; 165,029; 165,031; 165,010; 165,012; 165,011; 165,019; 165,046; 165,016; 181,217–181,221.

Immunofluorescence

All samples were pre-incubated in PBS containing 3% BSA (Roche Diagnostics, Mannheim, Germany) and then incubated for 60 min with primary monoclonal antibody JoL2 for Lamin A/C [16] (IgG1, kindly provided by C. Hutchison, University of Durham, UK) 1:50 diluted in PBS containing 3% BSA. Alternatively, a Lamin A mouse monoclonal antibody 133A2 (IgG3, Nordic-MUbio, Susteren, The Netherlands, diluted 1:1000); or Lamin C specific rabbit polyclonal antibody RalC, (Nordic-MUbio, Susteren, The Netherlands, diluted 1:500) were used.

After washing with PBS, FITC conjugated rabbit anti-mouse Ig antibody (Dakopatts, Glostrup, DK) diluted 1:100 in PBS/BSA was applied and incubated for 60 min. After another series of washing steps in PBS, cells were mounted in 90% glycerol, containing 20 mM Tris–HCl pH 8.0, 0.02% NaN3, 2% 1,4-di-azobicyclo-(2,2,2)-octane (DABCO; Merck, Darmstadt, Germany), and diamidino-2-phenylindole (DAPI; 0.5 μg/ml Sigma-Aldrich).

Detection of nuclear abnormalities

For every patient’s fibroblast culture at least 2 × 100 cells in different areas of the sample were evaluated using a Leica DMRBE fluorescence microscope (Leica, Mannheim, Germany), equipped with a 63x oil objective (Plan Apo, NA 1.32). Different aspect of the nuclear morphology were assessed, i.e., presence of nuclear shape abnormalities, seen as irregular lining of the nuclear membrane, forming nuclear blebs (herniations), extensive lobulations or donut-like invaginations of the nucleus. In addition, Lamin staining abnormalities were scored, including extranuclear staining, and the presence of so-called honeycombs. Also, the intensity of staining was registered as being weak, moderate or strongly positive. Part of the samples were re-counted independently by a second examiner, revealing that only a limited interobserver variation was found (variation 1.09 ± 0.25% (n = 7) (mean ± SEM). Finally, the presence of intranuclear aggregates was registered, after excluding intranuclear invaginations that are common also in normal fibroblasts.

Statistical analyses

A linear logistic regression model was constructed to classify the controls versus the laminopathy patients. A hierarchical model was built allowing for the percentage of normal cells, or cells with blebs, honeycombs, or donut cells present to be taken into account. The inference criterion used for comparing the models is their ability to predict the observed data, i.e., models are compared directly through their minimized minus log-likelihood. When the numbers of parameters in models differ, they are penalized by adding the number of estimated parameters, a form of the Akaike information criterion (AIC) [17]. All statistical analysis presented were performed using the freely available program R [18]. The formula: round(1−(1/(1 + exp(−145.1 + 1554.9 × total abnormal count/total cCount))),2) was generated in order to classify a fibroblast cell-line as “laminopathy” or “normal”.

Results

Determination of the “nuclear morphology classifier“

For morphological analysis, different LMNA antibodies detecting lamin A only; lamin A and lamin C; and lamin C only, were initially tested. As shown in Fig. 1, similar results were obtained with all three antibodies. Lamin A showed in general a relative homogeneous staining reaction in all nuclei, while the lamin C antibody showed a more heterogeneous labeling. The lamin A + C antibody shows an intermediate staining pattern. While some variations in intensity levels occurred, none of the patient samples examined showed a clear differential expression between lamin A and C. Nuclei, not showing any lamin staining were not detected. Therefore, for further development of the nuclear morphology classifier, only the JoL-2 antibody (recognizing both lamin A and lamin C) and DAPI for nuclear counterstaining were used.

Fig. 1
figure 1

Immunofluorescence labeling of control fibroblast cells with different A-type lamin antibodies. a Lamin A antibody 133A2; b lamin A + C antibody JoL2; c lamin C antibody RalC

Full size image
Table 1 Overview of published laminopathy fibroblasts used in this study
Full size table

The morphology of ≥200 nuclei per cell culture was analyzed for all 8 controls and 9 fibroblast cultures with a LMNA variant that were published previously and designated as pathogenic and laminopathy based on clinical phenotype, segregation data and/or functional analyses (Table 1). In addition to classifying nuclei as normal or abnormal, the types of nuclear malformations were noted: blebs (herniations, including micronuclei), honeycomb structures, donut-like structures, or combinations of aforementioned categories (see Fig. 2; Table 2). In some cultures, intra-nuclear aggregates occurred next to a normal lamina staining (Fig. 3). However, since these aggregates varied dramatically in size as well as in number between cells within the same culture and were even noted in some normal cells, cells with aggregates were not included in the classifier. Statistical analyses of these parameters for the healthy and laminopathy cell-lines were conducted to determine the most discriminating parameter. As shown in Table 3, the parameter “percentage of abnormal nuclei” resulted in the lowest AIC, and was therefore most discriminative and used for subsequent analyses. Using this parameter, the eight control cell-lines showed on average 4.8 ± 1.0% abnormal nuclei (range 3.3–6%) and the nine laminopathy samples showed 26.3 ± 13.6% abnormal nuclei (range 10.6–53%).

Fig. 2
figure 2

Types of scored nuclear abnormalities in laminopathy cell cultures N, normal; BL, bleb; HC, honeycombs; DO, donuts, MI, micronuclei. Note that some cells contain more than 1 nuclear abnormality

Full size image
Table 2 Results nuclear morphology analysis of control- and published laminopathy fibroblasts
Full size table
Fig. 3
figure 3

Differently sized intranuclear aggregates in laminopathy cells detected with the three lamin antibodies. a Antibody to lamin C; b antibody to lamin A; c antibody to lamin A+C

Full size image
Table 3 Quality of statistical model using different parameters
Full size table

Validation study “nuclear morphology classifier”

As noted above, the percentage of abnormal nuclei was identified as the most discriminating parameter between laminopathy and control cells. Subsequently, a validation study comprising nuclear morphology analysis of 28 fibroblast cell-lines was performed and verified by second opinion from expert (Fig. 4). These cell-lines contained a LMNA variant that was classified by molecular genetics criteria as (likely) benign, variant of unknown significance (VUS), likely pathogenic variant or pathogenic variant (Table 4). The number of normal and abnormal nuclei per cell culture was determined according to the “nuclear morphology classifier”. A cell culture with a classifier score ≥0.95 was considered as “laminopathy” and a value of ≤0.05 was considered as “normal”. If the classifier value ranged between 0.05 and 0.95, the cell-line was considered as “unclassified”. As shown in Table 4, 27 out of the 28 cell-lines (96%) could be classified with ≥95% probability as laminopathy or normal nuclear morphology. Only ID 37, containing a LMNA VUS c.1634G>A ((p.(Arg545His)), had a score of 51% and could therefore not be classified. Out of the 27 classified cell-lines, 13 were defined as laminopathy and 14 as normal. The laminopathy group contained fibroblasts with either a pathogenic variant (n = 1), a likely pathogenic variant (n = 8) or a VUS (n = 4). The group classified as normal mainly contained VUS cell cultures (n = 8), two cell cultures with a (likely) benign polymorphism, one recessive pathogenic variant, one likely pathogenic variant and two cell cultures containing a pathogenic variant. Fibroblasts containing either a (likely) benign polymorphism or recessive pathogenic LMNA c.892C>T (p.(Arg298Cys)) variant [19] (n = 3) and fibroblasts containing a (likely) pathogenic variant (n = 13) were used to calculate sensitivity and specificity of the nuclear morphology classifier. Specificity was found to be 100% (CI 29–100%) and the sensitivity was 77% (CI 46–95%). With respect to reproducibility, for two LMNA variants, nuclear morphology of two unrelated individuals was analyzed. Both fibroblast cell cultures with the LMNA c.1930C>T (p.(Arg644Cys)) variant displayed normal nuclear morphology, and nuclear morphology of two carriers of LMNA c.313_314delinsTT (p.(Glu105Leu)) both presented as laminopathy.

Fig. 4
figure 4

Distribution of the percentage abnormal nuclei identified per class LMNA variant. The percentage abnormal nuclei in controls (open circles) and laminopathy patients (black circles) were used as training set to generate the classifier. The percentage of nuclear abnormalities identified in the validation set are shown grouped based on their initial clinical genetic classification: class 1/2 ((likely) benign or recessive pathogenic variants in heterozygous state) in open squares; class 3 (VUS) light gray squares; class 4 (likely pathogenic) dark gray squares; class 5 (pathogenic) black squares. The three false negative variants are indicated by their ID number

Full size image
Table 4 Results nuclear morphology analysis in validation study
Full size table

Discussion

The aim of the present study was to assess if aberrant nuclear morphology is indicative of a pathogenic LMNA variant and can be used to aid classification of identified genetic variants in LMNA. We assessed nuclear morphology of 9 laminopathy dermal fibroblast cultures and 8 control fibroblast cultures, as training set, and demonstrated that an increased percentage of abnormal nuclei, irrespective of the type of nuclear malformation (herniation, honeycomb structure or donut shape), is the most discriminating parameter between normal and laminopathy cells. Subsequent assessment of the percentage of abnormal nuclei in validation set of fibroblast cultures containing a (likely) pathogenic LMNA variant (n = 13) or a (likely) benign LMNA variant (n = 3) demonstrated a 100% specificity (CI 29–100%) with respect to determining pathogenicity, as none of the three negative samples were classified as laminopathy. However, the confidence interval is large due to limited availability of cell cultures containing (likely) benign LMNA variants and future inclusion of more negative samples is warranted to reduce this confidence interval. From the 13 samples containing a (likely) pathogenic LMNA variant, 10 samples displayed an excessive percentage of abnormal nuclei and were classified as laminopathy, resulting in a sensitivity of 77% (CI 46–95%). Moreover, the same classification was obtained for two unrelated individuals carrying the LMNA c.313_314delinsTT (p.(Glu105Leu)) variant for two unrelated carriers of the c.1930C>T (p.(Arg644Cys)) variant, demonstrating reproducibility of the developed method. Taken together, we conclude that a laminopathy classification based on significant number of abnormal nuclei in fibroblasts is a quick and reliable method to confirm pathogenicity and aid classification of LMNA variants that predict formation of aberrant lamin A/C protein, while observation of normal nuclear morphology does not rule out pathogenicity.

To our knowledge, few attempts were made to categorize laminopathies based on their nuclear abnormalities, but this is the first and largest study undertaken (n = 45) that uses nuclear morphology assessment of fibroblasts containing an established LMNA variant or no LMNA variant in order to rank newly identified LMNA variants as laminopathy or normal based on the percentage of abnormal nuclei. The percentages of abnormal nuclei observed in the control and laminopathy group in our study are in line with the percentages reported by Decaudain et al., who showed 15–25% of dysmorphic nuclei with herniations in all six patients with a LMNA variant compared to 5% abnormal nuclei in control fibroblasts [20]. In contrast, Muchir et al. observed <1% abnormal nuclei in controls [13], this may be a consequence of various methodological differences, criteria and sensitivity. For instance, the detection of honeycomb-like structures is highly dependent on the quality of the staining and the detection method used. Honeycomb-like structures can be missed due to a weak antibody labeling, as well as by background labeling in the affected nuclei. Moreover, most of these structures are only visible with a high-resolution oil lens, by systemically focusing individual nuclei at a number of horizontal planes. In some of our cases, confocal microscopy was needed to visualize and confirm the gaps in the lamina staining.

We observed abnormal nuclear morphology in fibroblasts containing a (likely) pathogenic variant in the LMNA linker or tail domain, but also for 5 out of 7 (likely) pathogenic variants located in the coil domain of LMNA. In contrast, Muchir et al. only observed a significantly increased percentage of abnormal nuclei in the eight cell-lines from EDMD/LGMD/FPLD patients containing a variant in the LMNA head or tail domain, but in none of the five LMNA coiled-coil domain variants [13]. Our results and those of Decaudain et al. [20] demonstrate that the presence of abnormal nuclei as a consequence of a LMNA pathogenic variant is not limited to certain domains of the lamin A/C protein, even if not all LMNA variants necessarily cause nuclear malformations, which may also be the case for the three false negative results obtained in our validation study. The LMNA c.514_1995del deletion was shown to cause haploinsufficiency [12], supporting a previous hypothesis that only formation of mutated lamin A/C proteins causes nuclear malformations, rather than a reduction in lamin A/C protein quantity, as was also observed in different tissues of mice with reduced lamin A/C expression [21]. Secondly, nuclear morphology analysis in HeLa cells transfected with p.(Arg190Trp) lamin A/C by Bhattacharjee et al. showed nuclei containing aggregates, but no other nuclear abnormalities were described and they concluded that there were no notable changes compared to wild-type [22], which is in line with our classification as normal of lamin A/C (p.(Arg190Trp)) fibroblasts. Of note, nuclear foci and/or aggregates were not taken into account in our study, but have been reported in fibroblasts [13]. Also, foci seem to arise due to overexpression, e.g., after transfection of wild-type lamin A and C [23], or much more prominently after transfection with mutant lamin constructs [8, 24]. For some LMNA variants it has been reported that mutated lamin can form aggregates without interacting with wild-type lamins and thus possibly causing no harm to the nuclear shape nor integrity [25]. In our study, aggregates were only detected in c.1634G>A (p.(Arg545His)) fibroblasts, which also presented with some honeycomb and donut-like structures and a classifier score of 0.51, which we considered as “unclassified”. Future inclusion of nuclear aggregates as parameter may increase sensitivity of the classification tool, but requires further research to establish the occurrence of foci in wild-type cells and a systemic assessment of their significance. Exploring implementation of automated quantification of nuclear morphology in 2D microscopy images, as performed by Core et al. to identify dysmorphic nuclei [26], may also aid our nuclear morphology analysis procedure and classification tool further, but requires optimization as honeycomb structures are difficult to assess using 2D images and likely require 3D analysis methods. Also, automated analysis may pose challenges for implementation of the nuclear morphology analysis in routine genetic laboratories, as the required equipment may not be available. Thirdly, some LMNA pathogenic variants may not cause nuclear abnormalities at all or only cause nuclear abnormalities following stress or present in certain cell-types other than fibroblasts. This may be the case for the likely pathogenic variant c.949G>A (p.(Glu317Lys)), which has been reported in patients with AV-block and DCM [27, 28]. It is unclear if this variant can cause nuclear abnormalities at all, since it has not been functionally analyzed by other groups. Further research is required to assess pathogenicity of this variant and its mode of action.

Taken together, normal nuclear morphology does not rule out pathogenicity of LMNA variants, but detection of excessive abnormal nuclei provides functional evidence of pathogenicity and may warrant reclassification of LMNA VUS as a pathogenic variant. Implementation of this tool in our laboratory enabled reclassification of 4 of the 12 variants from VUS to likely pathogenic variant (33%). Since ~50% of all identified rare missense and splice-site variants in LMNA are being classified as VUS, implementation of this nuclear morphology analysis tool, will have considerable impact on the counseling and follow-up for patients and family members.