Version 0.3.1 ExAC data was accessed by querying “LMNA” through the ExAC Browser (Beta) (ExAC database, 2017). All data were reported as PASS per ExAC data quality standards. Data was selected for “Missense + LoF” variants and exported to CSV. The data was further narrowed to include only missense variants of canonical spliceforms of lamins A and C (Supplementary Table 1).

The full LMNA list of mutations (311 references and 2251 subjects) was accessed from the UMD-LMNA database [UMD-LMNA (laminopathy) database, 2017]. This list was then curated to exclude all non-missense mutations, and non-canonical spliceforms, yielding the 1404 variants shown in Supplementary Table 2 (comparison of ExAC and UMD-LMNA variant data).

Since we were unable to determine disease status for specific variants in the ExAC cohort, we accessed and searched LMNA variants in the Type 2 Diabetes Knowledge Portal² using a Type 2 Diabetes phenotype filter. Search results were selected for variants of “nominal significance,” defined by the T2D Knowledge database as having a p-value < 0.05. Variants significant for Type 2 Diabetes were identified solely within the 17K exome sequence analysis cohort, which comprises data from T2D-GENES (1,018 case and 1,056 control African Americans; 1,012 case and 1,153 control East Asians; 990 case and 853 control Europeans; 1,021 case and 922 control Latinos; 1,094 case and 1,123 control South Asians), GoT2D (1,369 case and 1,339 control Europeans) and SIGMA (1,817 case and 1,975 control Latinos). Detailed information about these studies can be found at http://www.type2diabetesgenetics.org/informational/data. The link between variant p.G602S and type 2 diabetes was found in three subcohorts within the T2D-GENES dataset, namely the Jackson Heart Study Candidate Gene Association Resource and the Wake Forest Study (both enriched in African American individuals), and the Singapore Indian Eye Study (enriched in South Asian individuals). All p.G602S-related statistics were obtained from the analysis reported in the T2D Knowledge Portal. This independent analysis of the T2D-GENES dataset, via the T2D Knowledge Portal, was essential to associate variant p.G602S with diabetes (even though these individuals are included in the ExAC cohort), since patient data in the T2D-GENES cohort was not accessible via ExAC.

ExAC release 0.3 has two variant call format (.vcf) files: ExAC.r0.3.sites.vep.vcf.gz, which contains the variant calls for all 60,706 individuals, and ExAC.r0.3.nonpsych.sites.vcf.gz, which contains the variant calls from 45,376 individuals from the non-psychiatric subset. To determine if any variants were enriched in the psychiatric cohort, we identified LMNA missense alleles unique to the remaining 15,330 individuals from the psychiatric subset, comprised of the Swedish Schizophrenia and Bipolar Studies cohort (12,119 individuals), Schizophrenia Trios from Taiwan (1,505 individuals), Sequencing in Suomi (948 individuals), Bulgarian Trios (461 individuals) and Tourette Syndrome Association International Consortium for Genomics (297 individuals). We used 7zip to extract the.vcf files. In a Cygwin64 terminal, header information was removed and only the chromosome number, position, reference sequence, and alternate sequence fields were retained. We generated a psychiatric-unique file by retaining lines from the “all sites” variant call file that were not present in the non-psychiatric subset file. The consequences of multi-allelic or minimal representation were handled manually in vim to ensure that all entries in the psychiatric-unique.vcf file were not also present in the non-psychiatric subset.

Shell script used to select data from psychiatric cohorts within ExAC:

#!/bin/tcsh

# Shell Script to Generate.vcf Files with Psychiatric-Unique Alleles

# remove header information

grep -v ′ˆ#′ ExAC.r0.3.sites.vep.vcf > allsites_cln.vcf

grep -v ′ˆ#′ ExAC.r0.3.nonpsych.sites.vcf > nonpsych_cln.vcf

# retain only chromosome no., position, refseq, and altseq fields

awk ′{$3 = $6 = $7 = $8″″; print $0}′ allsites_cln.vcf > allsites_final.vcf

awk ′{$3 = $6 = $7 = $8″″; print $0}′ nonpsych_cln.vcf > nonpsych_final.vcf

# subtract nonpsych subset from allsites

comm -23 allsites_final.vcf nonpsych_final.vcf > psych_final.vcf

# create file for just chromosome 1

awk ′$1== 1′ nonpsych_final.vcf > chromo1_nonpsych.vcf

awk ′$1== 1′ psych_final.vcf > chromo1_psych.vcf

# select all positions corresponding to LMNA

awk ′$2>=156082546 && $2<=156140089′ chromo1_nonpsych.vcf>LMNA_nonpsych.vcf

awk ′$2>=156082546 && $2<= 156140089′ chromo1_psych.vcf>LMNA_psych.vcf

The amino acid sequence alignments shown in Figures 3, 5 were performed manually, using Accession identifiers provided in figure legends. The human lamin A, B1 and B2 sequences used in the alignment of head and ‘neck’ domains (Figures 3A,D) were accessed through UniProtKB. We also used UniProtKB to access lamin A sequences from Camelus ferus, Dasypus novemcinctus, and Danio rerio for alignment with human lamin A (Figure 5A).

Native coiled-coil interactions between two parallel rod domains of A-type lamins were predicted using DrawCoil 1.0³. Comparative predictions were made with single missense variant sequences, as described in Figures 3B,C legends.

We compared the missense variants reported in ExAC to those identified in 1,404 distinct individuals with missense mutations in the laminopathy database (Supplementary Table 2). Among the 28 variants identified four or more times in ExAC, 15 were also represented in the laminopathy database. Note that this fact alone does not predict overt disease, since 214 of the 1,404 individuals in the laminopathy database with missense mutations were reported as asymptomatic. However, only a minority of laminopathy-associated variants are reported in asymptomatic individuals. One variant (p.R343Q, found in eight ExAC individuals) is reported in a single apparently healthy individual in the laminopathy database. By contrast, other ‘shared’ variants were nearly always reported in individuals with disease in the laminopathy database (Figure 1B and Supplementary Table 2). These included p.G523R (8 individuals in ExAC; one DCM patient in laminopathy cohort), p.S573L (11 individuals in ExAC; 10 individuals with multiple phenotypes in laminopathy cohort), p.R644H (14 individuals in ExAC; one patient with congenital muscular dystrophy in laminopathy cohort), p.G602S (29 individuals in ExAC; 3 patients with IRS and 2 patients with EDMD in laminopathy cohort), p.I299V (47 individuals in ExAC; one asymptomatic individual and 3 FPLD2 patients in laminopathy cohort), and p.R644C (145 alleles corresponding to 144 individuals in ExAC; 43 patients with multiple phenotypes including DCM and FPLD2 in laminopathy cohort) (Figure 1B). All variants were observed in heterozygosity in both the ExAC and laminopathy databases, with rare exceptions as noted. All variants in the ExAC cohort and laminopathy cohort and their locations in the prelamin A polypeptide are depicted in Figure 2. Variants found in both cohorts are shown in Figure 2A; variants unique to each cohort are shown in Figure 2B. More than 219 individuals with LMNA variants in the ExAC cohort have a mutation previously observed in patients with laminopathies— predominantly but not exclusively heart disease— including 158 individuals with either p.R644C or p.R644H, and 61 individuals with p.K117R, p.S326T, p.G523R, p.S573L, or p.G602S (Figure 2A). Because individual-level phenotypic data is unavailable for the ExAC cohort, we were unable to determine whether these variants correlated with myocardial disease or other specific phenotypes. However, since there is already strong evidence that LMNA mutations contribute to heart disease in the general population (Parks et al., 2008; Tesson et al., 2014), these variants are of particular interest for functional follow-up.

There were 145 p.R644C alleles in the ExAC cohort, representing 144 individuals (one homozygote), out of 116,680 total alleles for an allele frequency of 0.1242%. For phenotypic perspective, the laminopathy database reports 43 individuals with p.R644C, among whom 11 were asymptomatic, one had atypical progeria, one had neuropathy, six had lipodystrophy, and 23 had cardiac disease with or without muscular dystrophy. Although these discrepant phenotypes make it difficult to ascertain the true consequences or pathogenicity of this allele, these results warrant further study of the ExAC cohort to determine what percentage of individuals with p.R644C are asymptomatic and whether this variant correlates with any diseases in this population. A different variant at this same amino acid residue, p.R644H, is seen once in heterozygosity in the laminopathy cohort (a girl with congenital muscular dystrophy who died at 20 months). We note that p.R644H was identified in heterozygosity in 14 of 116,722 chromosomes in the ExAC cohort, for an allele frequency of 0.0120%. Because all individuals represented in ExAC survived childhood, the pathogenicity of this variant in isolation is dubious. Interestingly, p.R644C is ethnically clustered within ExAC. This variant is more frequent in non-Finnish Europeans (106 out of 63,724 sequenced alleles; 0.162% allele frequency) and South Asians (27 out of 16,410 sequenced alleles; 0.165% allele frequency). No obvious ethnic correlation was found for p.R644H.

The LMNA p.I299V variant was identified 47 times in the ExAC cohort (heterozygous individuals; allele frequency 0.0402%). Four related individuals in the laminopathy database have this variant: a heterozygous mother and two heterozygous daughters all with FPLD2, and a grandson homozygous for p.I299V who presented as asymptomatic (Araújo-Vilar et al., 2012). This heterogeneity is consistent with FPLD2 as a progressive and often misdiagnosed disorder with variable phenotypic manifestations and age of onset (MIM 151660; Hussain and Garg, 2016). Individuals harboring this variant in ExAC might have features of FPLD2. Alternatively, this variant may not be solely responsible for FPLD2. We conclude that p.I299V, the second most common variant in the ExAC cohort, has the potential to correlate with disease risk and warrants further genotype–phenotype analysis. Upon analyzing the ethnic sub-cohorts of ExAC, we found that p.I299V is more frequent among Latino individuals (39 of the 11,242 alleles; 0.347% allele frequency). The concentration of variant p.I299V among Latinos warrants further investigation as a potential predictor of disease risk.

One individual in the ExAC cohort was heterozygous for the LMNA variant p.R482Q. Variants p.R482Q and p.R482W are considered rare and highly penetrant mutations responsible for FPLD2, and are seen in ∼13% of all known laminopathy patients (241 of 1849 total individuals in the laminopathy database). p.R482 substitutions are implicated in 80% of FPLD2 cases, which has a disease prevalence of less than 1 in 10 million (Subramanyam et al., 2010). It is possible that the individual in ExAC who carries p.R482Q has unrecognized features of FPLD2. Alternatively, the variant might have been non-penetrant in this individual at the time of study. Further investigation in databases that provide individual genotype/phenotype data will be crucial to understanding the implications of this variant in the broader population.

The p.G602S variant was identified 29 times in heterozygosity in the ExAC cohort with an overall allele frequency of 0.0262%. The laminopathy database reports only five individuals with p.G602S: two males (relation uncertain) with EDMD (Bakay et al., 2006; Scharner et al., 2011) and three related individuals from Reunion Island (African Creole descent), including a father with Type 2 diabetes, his son with insulin-resistant diabetes, and daughter with polycystic ovarian syndrome and insulin resistance (Young et al., 2005). To explore the implications of the ExAC data, we analyzed the open-access Type 2 Diabetes Knowledge Portal (Type 2 Diabetes Knowledge Portal, 2016) and searched within 100 kb of LMNA. This search revealed 118 variants in the 17K exome sequence analysis dataset, but only one variant showed a significant correlation with disease: p.G602S was identified in 14 out of 8378 patients with Type 2 diabetes (0.17% of patients), and in 2 of 8478 controls (0.024% of controls; odds ratio 4.58; p-value 0.0200). Interestingly, 13 of the 14 individuals with Type 2 Diabetes in the 17K Exome cohort were African American and one was South Asian. This proposed link to Type 2 Diabetes is consistent with the above-discussed family in the laminopathy database in which the p.G602S variant co-segregates with certain aspects of diabetes in two generations of a family with African ancestry (Young et al., 2005). Analysis of the ExAC database using ethnic break-out information shows that 26 of the 29 individuals with p.G602S were African American; by contrast the European American, Latin American, and South Asian cohorts each included a single individual with p.G602S. The allele frequency for p.G602S within the African American cohort in ExAC was 0.297% (26 out of 8,766 total alleles). These results warrant further studies of the effects of the LMNA p.G602S variant on glucose homeostasis, and further consideration as a contributor to diabetic risk, particularly in individuals with African ancestry.

To determine whether any LMNA missense variants were enriched in individuals with psychiatric disease, we extracted variant call files from the subset of individuals in ExAC with recognized psychiatric disorders (see Materials and Methods). There are several limitations to this approach. For instance, we could not be certain that all control individuals were excluded, and individuals with psychiatric disease might also have concurrent clinical phenotypes (e.g., heart disease) attributable to LMNA variants. In addition, we could not determine whether individuals in the psychiatric cohort with the same mutation had the same psychiatric phenotype (e.g., schizophrenia vs. bipolar disorder vs. Tourette). Finally, it is possible that certain individuals in the non-psychiatric subset had undiagnosed psychiatric conditions.

With these caveats in mind, 19 LMNA variants were present only in the psychiatric cohort in ExAC. The allele frequencies of these variants among all 60,706 individuals represented in ExAC ranged from <0.001 to 0.004829%. Four such variants, identified once each, are also reported in the laminopathy database: p.S395L (one individual with IRS; Decaudain et al., 2007), p.V586M (one individual with striated muscle laminopathy; Rodriguez et al., 2008), p.R388H (two DCM patients and 5 asymptomatic individuals; Parks et al., 2008) and p.S583L (9 individuals with FPLD; Savage et al., 2004). The remaining 15 variants were all novel (Supplementary Table 3). Since most were identified in only one individual, or two individuals each for variants p.Q168H, p.R235H, and p.Q396L, their potential significance is unknown. Intriguingly, all five individuals with variant p.K108E (overall allele frequency 0.004829% in ExAC) were unique to the psychiatric cohort and were all non-Finnish Europeans. This variant has potential molecular consequences, as discussed below.

Lamins A and C have identical residues 1–566, but have different C-terminal sequences generated by alternative splicing (Luo et al., 2014). Both isoforms are widely expressed throughout most human tissues except brain, which selectively expresses lamin C (Jung et al., 2012). Lamin C has six additional unique C-terminal residues, whereas lamin A is synthesized as a precursor (prelamin A) containing additional residues 567–664. Prelamin A is post-translationally modified and finally cleaved by the protease ZMPSTE24 to generate mature lamin A (comprising residues 1–644; Sinensky et al., 1994; Wang et al., 2016). Missense mutations linked to known laminopathies are generally scattered throughout the polypeptide and can alter many aspects of lamin biology, from filament assembly to partner interactions and post-translational modifications (Simon and Wilson, 2013). This functional complexity poses challenges to understanding disease mechanisms, and motivated our search for new human LMNA variants that might provide molecular insights.

We searched for novel missense variants in lamin C-specific residues, but found none. We therefore considered all missense variants in the context of their molecular locations in the prelamin A polypeptide, which includes all but six residues of lamin C. Different regions of the prelamin A protein, diagramed in Figure 2, are considered in detail below. The globular N-terminal ‘head’ domain (residues 1–34; Figure 3A), long coiled-coil ‘rod’ domain (residues 35–383; Figures 3B,C), small flexible ‘neck’ domain (residues 384–430; Figure 3D) and Ig-fold domain (residues 431–545; Figure 4) are all identical in lamins A and C. The unstructured C-terminal region unique to prelamin A (residues 568–664) that is processed to generate mature lamin A (residues 568–646) is depicted in Figure 5.

The head domain of human lamin A is 47% identical in lamins B1 and B2 (Figure 3A). This conservation is consistent with evidence that lamin A head residues 11–30 are required for head-to-tail polymerization in vitro (Isobe et al., 2007). Residues R7, R11, and R25 are predicted to mediate electrostatic interactions with coil-2B (Strelkov et al., 2004), suggesting head-to-tail polymerization may be weakened by novel ExAC variants p.R7Q, p.R11G and p.R25L (Figure 3A). Post-translational regulation or other functions of the head domain might be perturbed by novel ExAC variants p.S5F (Ser5 is a conserved phosphorylation site; Santamaria et al., 2011; Kochin et al., 2014), and by non-conservative substitutions such as p.A14T and p.A16D (Figure 3A). Head domain variants were each identified in only 1–2 individuals (Figure 3A).

The coiled-coil ‘rod’ domain is a defining structural feature of intermediate filament proteins (Herrmann and Aebi, 2016). In lamins, the rod domain comprises three regions (coil-1A, coil-1B, and coil-2) separated by short linkers. The rod domains of two nascent lamin polypeptides associate in parallel, creating a structural coiled-coil motif (Heitlinger et al., 1992). We used DrawCoil 1.0 (developed in the Grigoryan lab at Dartmouth University) to predict the consequences of variants in the rod domain. Based on the predicted loss of conserved electrostatic ‘backbone’ interactions between rod domains, we suggest that lamin dimerization would be perturbed by novel variants p.R110H (identified in 6 individuals; allele frequency 0.00483%; Figure 3B) and p.R166W (two individuals; allele frequency 0.00384%; Figure 3C). Different amino acid substitutions at these same positions are seen in the laminopathy database: p.R110S (one DCM patient and 3 asymptomatic individuals; Botto et al., 2011) and p.R166P (3 patients with DCM; Parks et al., 2008; Saga et al., 2009). In addition to structural roles, the rod domain is post-translationally regulated and provides functional docking sites for specific partners (e.g., retinoblastoma protein binds lamin A/C residues 247–355; Mancini et al., 1994; Ozaki et al., 1994). Interestingly, novel variant p.K108E (5 individuals in the psychiatric cohort; discussed above) affects the rod domain surface, and would block two known post-translational modifications at K108: acetylation and ubiquitylation (Simon and Wilson, 2013).

Head-to-tail polymerization is further predicted to involve four Arg residues (R388, R397, R399, and R401) in the ‘neck’ region, which contact coil-1A in the partner (Strelkov et al., 2004). These interactions may be altered by ExAC variants p.R388H (one individual), p.R397C (novel; 5 individuals; allele frequency 0.004176%), p.R397H (novel; 3 individuals; allele frequency 0.002505%), p.R399C (one individual), p.R399H (5 individuals; allele frequency 0.004169%), and p.R401H (novel; 12 individuals; allele frequency ∼0.01%; Figure 3D). The positively charged nuclear localization signal (NLS; residues 417–422; Figure 3D) in the neck region is likely to be perturbed by novel variant p.R419C (one individual, allele frequency 0.0008268%), and possibly by novel variant p.R419H (8 individuals; allele frequency 0.006629%; Figure 3D). Variant p.R419C was reported in three FPLD patients, although pathogenicity was not clearly established (Haque et al., 2003). Two studies in cells with exogenously expressed p.R419C reached conflicting conclusions about its effects on nuclear localization (Yang et al., 2013; Kiel et al., 2014). Novel variants p.G411C (2 individuals; allele frequency 0.001660%), p.R427C (5 individuals; allele frequency 0.00416%), and p.S428R (2 individuals; allele frequency 0.001670%) are non-conservative substitutions that might intrinsically perturb neck flexibility or function (Figure 3D). For example, neck residues 403–425 mediate binding to nesprin-2, a component of nuclear envelope-spanning ‘LINC’ complexes (Yang et al., 2013). The ExAC cohort included single individuals with neck variants also reported in individuals with disease in the laminopathy database: p.G411D (IRS; Dutour et al., 2011), p.G413C (striated muscle disease; Rodriguez et al., 2008) and p.V415I (isolated atrial fibrillation; Brauch et al., 2009).

Novel ExAC variants in the structurally conserved Ig-fold domain were identified in single individuals (19 variants) or in two (p.S428R), three (p.A434T, p.R453Q), four (p.V442M, p.A545H) or five (p.M464V) individuals (Figure 2B and Supplementary Table 2). The ConSurf algorithm (ConSurf database, 2017), which aligns solved protein structures with all known structural homologs to assess evolutionary conservation, showed that six evolutionarily invariant residues were affected by ExAC variants: three are known to associate with disease (p.D461Y, p.R471C, and p.G523R); the others are novel non-conservative substitutions (p.S428R, p.T436I, and p.G438R) that are likely to perturb Ig-fold function.

Two slightly different atomic structures are reported for the Ig-fold domain of human lamin A. One was determined by X-ray crystallography (RCSB Protein Data Bank ID: 1IFR) for residues 432–544, but includes non-native N-terminal residues 432–434 (GSH) and a glycerol ligand (Dhe-Paganon et al., 2002). This ligand is predicted by PyMOL to form polar contacts with a backbone hydroxyl group on V513. The other structure, determined in solution by NMR (RCSB Protein Data Bank ID: 1IVT) is longer (residues 428–549), has native residues 432–434 (QHA), and no ligands (Krimm et al., 2002). This NMR structure consists of 15 frames; we used the first frame, which is representative as seen in the superimposition⁴. We used PyMOL to map the positions of wild-type residues corresponding to each novel variant (‘Unique to ExAC’; Figure 4A). We compared these to all surface-exposed residues affected by variants in the laminopathy database (‘LDB’; Figure 4A). We obtained similar results for both structures, but show only the NMR results because this structure includes affected residues S428, A434, R545, and V549.

With one exception (p.L530V), all Ig-fold variants unique to the ExAC cohort affect wild-type residues that are solvent-accessible (‘Unique to ExAC’; Figure 4A; note that M464, in a tiny cavity slightly behind P509, is not visible). By contrast, most variants linked to EDMD affect residues in the hydrophobic interior, or residues required to stabilize Ig-fold structure (e.g., R453; Burke and Stewart, 2002; Scharner et al., 2014). Surface-exposed residues linked to skeletal muscle disease include D446, R453, R455, N459, R545, and W520, roughly aligned on one side of the Ig-fold (Scharner et al., 2014), as seen in our ‘Back’ view (‘LDB’; Figure 4A). Three residues in this ‘muscle-region’ were affected in the ExAC cohort, but with substitutions more conservative (p.D446E, p.R453N, p.R545H in ExAC) than those linked to skeletal muscle disease (p.D446V, p.R453P, p.R545C; Supplementary Table 2).

Interestingly, a comparison of residues uniquely affected by ExAC variants (‘Unique to ExAC’; Figure 4A), compared to residues affected by all laminopathy variants (‘LDB’; Figure 4A), revealed that certain novel ExAC variants affect ‘new’ regions of the Ig-fold surface. In particular, ExAC variants at the evolutionarily invariant residue S428, and residue V549, affect two ‘knobs’ in the Front view (yellow labels; ‘Unique to ExAC’; Figure 4A). Different ‘knobs’ in the Bottom view are affected by ExAC variants at S463, T534, and G535 (aqua labels; Figure 4A). Seven novel ExAC variants affect residues G492, V494, T496, G503, A504, P509, and D511, best seen in the Back view (green labels; Figure 4A), adjacent to but distinct from the ‘muscle’ region (Scharner et al., 2014).

These proposed Ig-fold surface ‘features’ affected by novel ExAC variants were also distinct from three other previously recognized features. SUMO1 modification of Ig-fold residue K486, potentially relevant to FPLD2, requires an acidic ‘patch’ formed by E460, D461, E536, and E537 on the Ig-fold surface (‘SUMO’; Figure 4B; Simon et al., 2013). A different region of the Ig-fold surface is specifically affected by variants linked to progeria. In particular, atypical HGPS (AHGPS) and MADA, which have overlapping clinical phenotypes (Guénantin et al., 2014; Mehrez and Mostafa, 2014; Krawiec et al., 2016), are caused by single substitutions at residues R435, R471, R527, A529, M540, or K542, which form a ‘patch’ on the Ig-fold surface (Front views; AHGPS and MADA; Figure 4B). We exclude neighboring residue T528 as a core component of this ‘progeria patch’ because most T528 variants cause muscular dystrophy (Vytopil et al., 2003; Verstraeten et al., 2006; Astejada et al., 2007; Bertrand et al., 2011; Scharner et al., 2011). We consider residues A529, M540, and K542 as core features of the progeria patch, since variants p.A529V, p.A529T, and p.K542N each cause progeroid phenotypes in homozygosity (Plasilova et al., 2004; Garg et al., 2005; Kosho et al., 2007) and progeria also arises from compound heterozygosity for p.M540T and p.T528M (Verstraeten et al., 2006). No variants at these residues were seen in ExAC. However, the ExAC cohort included variants at three other progeria-relevant residues: R435, R471, and R527. One ExAC individual carried p.R435C (allele frequency 0.000863%); this variant is reported as heterozygous in 14 asymptomatic individuals in the laminopathy cohort and one DCM patient (Vytopil et al., 2003), and homozygously in single cases of progeria (Madej-Pilarczyk et al., 2009) or restrictive dermopathy (Youn et al., 2010). Three individuals in ExAC carried p.R471C (allele frequency 0.00256%); this variant is also reported in the laminopathy cohort in three asymptomatic heterozygous individuals, one homozygous patient with MADA/progeria (Zirn et al., 2008) and one HGPS patient in compound heterozygosity with p.R527C (Cao and Hegele, 2003). Four individuals in ExAC carried p.R527H (allele frequency 0.00683%); this variant is also reported in the laminopathy cohort in 19 asymptomatic heterozygous individuals and 18 patients with MADA due to p.R527H homozygosity (Novelli et al., 2002; Shen et al., 2003; Simha et al., 2003; Garavelli et al., 2009) or compound heterozygosity with V440M (Lombardi et al., 2007). Since ‘progeria patch’ variants appear to cause autosomal recessive disorders, the eight individuals in ExAC who carry such variants in heterozygosity may not manifest phenotypes. Although the frequency of the progeria-associated p.R527H allele (0.00683%) in the ExAC cohort might seem high compared to most other LMNA variants (only 1–2 individuals; Figure 1A), this allele was still rare (4 of ∼58,000 chromosomes), and might reflect population enrichment in ExAC. For example the p.R435C allele (not found in ExAC) is enriched among Caucasians (Starke et al., 2013), the p.R527H allele (4 individuals in ExAC) is enriched in Italians (Novelli et al., 2002) and the p.R527L allele (not found in ExAC) is enriched among northeastern Egyptians (Al-Haggar et al., 2012).

As noted above, only one individual in ExAC has an Ig-fold variant in the FPLD2 ‘hotspot’ (p.R482Q). Other variants linked to FPLD2 (Figure 4B; Burke and Stewart, 2002) or reported in metabolic syndrome patients (MetS; Figure 4B) are also somewhat clustered, affecting residues R439, G465, R482, K486, H506, or K515 on the Ig-fold surface (Chirico et al., 2014; Scharner et al., 2014). Three individuals in ExAC have variant p.R439C (allele frequency 0.00278%), implicated in FPLD2 and metabolic syndrome. Related to FPLD2, there is also a biochemically defined molecular feature (‘acidic patch’) required for SUMO1 modification of Ig-fold residue K486 (Figure 4B; Simon et al., 2013). Only one individual in ExAC had an ‘acidic patch’ variant (p.D461Y), compared to four individuals in the laminopathy database with p.D461Y (two with DCM, one with EDMD, one asymptomatic) and three with p.G465D (all FPLD2 patients). Overall, few individuals in ExAC harbored Ig-fold variants linked to FPLD2.

The C-terminal ∼100 residues of prelamin A lack defined structure, but are nonetheless conserved in evolution (Figure 5A). The ExAC cohort included disease-associated variants p.S573L, p.G602S, p.R644C, and p.R644H, which map to this region and affect residues that are invariant or conserved, along with 27 novel variants including p.G638R (17 individuals; allele frequency 0.0146%) and p.N660D (5 individuals; allele frequency 0.00412%; Figure 5A). This unstructured region has many reported post-translational modifications including phosphorylation and O-GlcNAcylation (Simon and Wilson, 2013; Simon et al., unpublished) that would be blocked by novel variants p.S612P, p.S613P, p.S628R, and p.T643I (identified in one individual each) and novel variants p.S625C and p.S625R, identified in three individuals each (Figure 5A).

Strikingly, 213 individuals in the ExAC cohort have variants located within residues 620–664 (Figure 5B). Variants in this region might affect prelamin A processing efficiency. This prediction is based on evidence that at least 23 residues upstream of Y646 in the prelamin A substrate were required for 100% efficient cleavage by ZMPSTE24; shorter prelamin A polypeptides were cleaved inefficiently (50% uncleaved) or failed to be cleaved (Figure 5B; Barrowman et al., 2012). Cleavage is also reduced or abrogated by single substitutions near Y646 (pink or purple boxes, respectively, in Figure 5B) (Barrowman et al., 2012). Remarkably p.R644C, the most frequent ExAC variant (143 heterozygous individuals, 1 homozygous individual; allele frequency 0.1243%), adjoins the ZMPSTE24-dependent cleavage site and partially blocks prelamin A cleavage in a cell-based assay (Barrowman et al., 2012). Variant p.R644C is reported in individuals with confusingly diverse phenotypes including DCM, FPLD2 and EDMD (Figure 1B; Rankin et al., 2008). Novel ExAC variants in this region, particularly p.S625C and p.S625R (3 individuals each), p.G638R (17 individuals; allele frequency 0.0146%), and p.N660D (5 individuals; allele frequency 0.00412%; Figure 5B) deserve further study to determine whether they affect prelamin A processing. Of note, 16 of 17 individuals in ExAC with the novel p.G638R allele are African American (allele frequency 0.168% among African Americans), warranting further study of its molecular and physiological impact(s).