Abstract
The p.R482W hotspot mutation in A-type nuclear lamins causes familial partial lipodystrophy of Dunnigan-type (FPLD2), a lipodystrophic syndrome complicated by early onset atherosclerosis. Molecular mechanisms underlying endothelial cell dysfunction conferred by the lamin A mutation remain elusive. However, lamin A regulates epigenetic developmental pathways and mutations could perturb these functions. Here, we demonstrate that lamin A R482W elicits endothelial differentiation defects in a developmental model of FPLD2. Genome modeling in fibroblasts from patients with FPLD2 caused by the lamin A R482W mutation reveals repositioning of the mesodermal regulator T/Brachyury locus towards the nuclear center relative to normal fibroblasts, suggesting enhanced activation propensity of the locus in a developmental model of FPLD2. Addressing this issue, we report phenotypic and transcriptional alterations in mesodermal and endothelial differentiation of induced pluripotent stem cells we generated from a patient with R482W-associated FPLD2. Correction of the LMNA mutation ameliorates R482W-associated phenotypes and gene expression. Transcriptomics links endothelial differentiation defects to decreased Polycomb-mediated repression of the T/Brachyury locus and over-activation of T target genes. Binding of the Polycomb repressor complex 2 to T/Brachyury is impaired by the mutated lamin A network, which is unable to properly associate with the locus. This leads to a deregulation of vascular gene expression over time. By connecting a lipodystrophic hotspot lamin A mutation to a disruption of early mesodermal gene expression and defective endothelial differentiation, we propose that the mutation rewires the fate of several lineages, resulting in multi-tissue pathogenic phenotypes.
Introduction
Laminopathies represent a class of rare diseases with phenotypes clinically expressed as muscular dystrophies, cardiomyopathies, lipodystrophies, neuropathies and premature aging (1). Laminopathies are caused by mutations in the LMNA gene, which encodes nuclear lamin A/C (abbreviated here as lamin A). The hotspot heterozygous lamin A p.Arg482Trp (R482W) mutation causes familial partial lipodystrophy of Dunnigan-type (FPLD2; OMIM no. 151660). FPLD2 is manifested by lipoatrophy in the lower body, adipose tissue accumulation in the upper body, and metabolic complications (2,3). FPLD2 is also associated with early and severe atherosclerosis resulting in premature cardiovascular events such as coronary heart disease, stroke and peripheral arteritis (4–6), indicative of primary endothelial cell (EC) dysfunction (7). Moreover, expression of the lamin A R482W mutation causing FPLD2 in ECs elicits reduced nitric oxide production, increased expression of inflammation and vascular cell adhesion molecules and senescence (6). Thus, the lamin A R482W mutation links lipodystrophic laminopathies to early onset cardiovascular pathologies, but whether the multi-tissue phenotype of the disease reflects an early developmental defect remains unknown.
By interfacing with chromatin, the nuclear lamina plays an important role in orchestrating genome stability, chromatin conformation and developmental gene expression (8). At the nuclear periphery, A- and B-type lamins interact with chromatin through lamina-associated domains (LADs), which constitute large, gene-poor and mostly transcriptionally silent regions of the genome (9). Lamin A proteins are also found in the nuclear interior, where they interact with gene regulatory elements (10–12). Both pools of lamin A are believed to contribute to the regulation of progenitor cell differentiation (10,12,13). Indeed, differentiation of adipose progenitors involves developmentally regulated interactions of lamin A with adipogenic and anti-adipogenic loci (10,12), and disruption of these interactions by expressing lamin A R482W impairs differentiation (12,14). Since EC function is also primarily affected in FPLD2 (6), we raised the hypothesis that the R482W mutation would impair the regulatory role of lamin A on differentiation towards mesodermal and endothelial lineages.
The T gene encodes the Brachyury T-box (T) transcription factor, which lies at the top of signaling cascades leading to mesoderm induction (15,16) and endothelial differentiation (17). Interestingly, T/Brachyury is overexpressed in embryoid bodies derived from mouse embryonic stem cells (ESCs) with Lmna haploinsufficiency (18), suggesting that T may be under developmental control of lamin A. Moreover, knock-down of the lysine-specific demethylase LSD1 leads to T/Brachyury overexpression and defective differentiation of mesodermal derivatives, including cardiovascular and adipose lineages (19).
These observations suggest a relationship between lamin A, T expression and endothelial differentiation, which may be affected by the lamin A R482W mutation. This would be best examined in FPLD2 patient-derived cells. Fibroblasts are useful to study aspects of nuclear architecture disrupted by the mutation (20) but are not suitable for developmental studies. Patient-derived induced pluripotent stem cells (iPSCs) are relevant models to investigate disease mechanisms (21), and laminopathies have been modeled using iPSCs (21–27). However, no well-controlled iPSC-derived model of FPLD2 has been established.
Here, we derived iPSCs from a patient with FPLD2, corrected the LMNA mutation and induced endothelial differentiation of the mutant and gene-edited wild-type isogenic counterparts. We report that lamin A R482W epigenetically deregulates the central mesodermal regulator T/Brachyury, leading to precocious endothelial gene expression and impaired EC phenotypes. Disruption of the fate of mesodermal lineages by a lipodystrophic lamin A mutation suggests a view of epigenetic developmental deregulation as a contributor to multi-organ-associated clinical phenotypes in FPLD2 patients.
Results
Lamin A differentially associates with the T locus in control and FPLD2 patient fibroblasts
Gene expression or potential for expression is influenced by the positioning of genes with respect to the nuclear lamina (10,11,28–31). We have previously shown that although the distribution of chromatin domains interacting with lamin A (lamin A LADs) is overall similar in fibroblasts from control individuals and patients with lamin A R482W-linked FPLD2, discrete differences in LADs correlate with a deregulation of gene expression in these LADs (20). Thus, we first determined whether lamin A association with the T locus differed in control and FPLD2 patient fibroblasts. Genome browser representation of lamin A LADs previously mapped by chromatin immunoprecipitation-sequencing (ChIP-seq) of lamin A/C (20), shows that whereas T lies in a LAD or at a LAD border in control fibroblasts (Fig. 1A, CTL nos. 1–2), lamin A association is consistently lost in the region containing T in fibroblasts from four FPLD2 patients with the R482W mutation (FPLD2 nos. 1–4). To ascertain this finding, we also mapped here by ChIP-seq lamin A LADs in fibroblasts from another unrelated lamin A R482W FPLD2 patient. Again, we show a dissociation of lamin A from the LAD region containing T in this patient (Fig. 1A; FPLD2 no. 5). This reduced lamin A binding to the T locus was confirmed by ChIP-qPCR analysis of lamin A (Fig. 1B). These differences in lamin A association on T were not due to distinct ChIP efficiencies in these cell types because ChIP in control and FPLD2 fibroblasts with the R482W mutation revealed similar lamin A levels at other sites (Supplementary Material, Fig. S1). These results indicate that lamin A binding to the T locus is impaired in FPLD2 patient fibroblasts with the lamin A R482W mutation.
The T locus differentially associates with lamin A in fibroblasts from FPLD2 patients. (A) Browser view of lamin A LADs in fibroblasts from two controls (CTL) and five FPLD2 patients (A-LADs), and of lamin B1 LADs in FPLD2 no. 5 fibroblasts (duplicates; B1-LADs). Position of T relative to the nearest LAD in each cell type is shown. (B) ChIP-qPCR analysis of lamin A binding to the T locus in control fibroblasts and FPLD2 fibroblasts (FPLD2 no. 5) from which iPSCs were generated in this study. Control sites are shown (Lam-, UBE2B promoter). Regions examined on the T enhancer (−6 kb from the T transcription start site; TSS), promoter (−1 kb from TSS) and introns 1 and 3 are shown on top. (C) Representative full-genome and tomographic views of 3D genome models showing the position of TADs containing the T locus (red beads; highlighted more clearly in the enlargements) in the nucleus space, in control (CTL no. 1) and FPLD2 fibroblasts (FPLD2 nos. 4 and 5). (D) Median distribution of the distance of the TAD containing T to the nuclear periphery across 100 3D genome models generated of each cell type (*P = 1.95 × 10−53; two-sided Fisher’s exact test). The distance of the TAD containing T to the nuclear periphery is also significantly greater in FPLD2 no. 5 fibroblasts used in this study, compared with controls (P = 6.50 × 10−15; two-sided Fisher’s exact test; not shown).
Spatial repositioning of the T locus in FPLD2 patient fibroblasts
A modified lamin A environment around the T locus raises the possibility of alteration in positioning of the locus in the 3D nucleus space in FPLD2 fibroblasts. To test this, we generated 3D structural models of the fibroblast genome using Chrom3D, a computational 3D genome modeling platform we recently developed (20), and analyzed the 3D genome models to infer radial positioning (i.e. nuclear center-periphery) of the T locus in patient and control cells. Input data for modeling were (i) genome-wide chromosome conformation capture (Hi-C) data for human fibroblasts (32), to provide information on the position of chromosomal topologically associated domains (TADs) relative to each other, including TADs containing the T locus; and (ii) lamin A LAD data generated previously (CTL nos. 1–2 and FPLD nos. 1–4) (20) and here (FLPD2 no. 5), to ascribe TADs that contain LADs to the nuclear periphery in the models. From this information, radial placement of TADs containing T can be inferred (20).
We analyzed the distance between TADs harboring T and the nuclear periphery in 100 models generated for each of the two controls and five FPLD2 patients. The models consistently place T in the nuclear interior rather than at the periphery in control and FPLD2 fibroblasts (Fig. 1C). However, the data reveal significant repositioning of T towards the center in FPLD2 compared with controls (Fig. 1D;P = 1.95 × 10−53; two-sided Fisher’s exact test). This suggests that the R482W mutation alters the lamin A neighborhood around the T locus in patient cells, in the nucleus interior rather than at the periphery per se. Supporting this view, lamin B1 ChIP-seq analysis in FPLD2 no. 5 fibroblasts reveals T within a 1.2 megabase region between two LADs (Fig. 1A; B1-LADs). Additionally, several lines of direct (11,13,33,34) and circumstantial (10–12,35–37) evidence, including from FPLD2 fibroblasts (20), reveal distinct lamin A-associated chromatin environments in the nuclear interior. Being an early mesodermal inducer, T is repressed in fibroblasts. Nevertheless, repositioning of T towards the nuclear center and the lack of T binding to lamin A in FPLD2 cells could speculatively reflect a propensity for facilitated activation in other cell types, including in a developmental model of the disease.
FPLD2 patient-derived iPSCs show defective endothelial differentiation
To test the hypothesis that lamin A R482W would affect regulation of T expression and early EC differentiation, we generated iPSCs from the FPLD2 fibroblasts (FPLD2 no. 5) analyzed above (henceforth, ‘FPLD2 iPSCs’; Fig. 2A). We also engineered an isogenic control iPSC line using a transcription activator-like effector nuclease (TALEN) approach to correct the LMNA c.1444C > T mutation in FPLD2 iPSCs (Supplementary Material, Fig. S2A). Indeed, we reasoned that an isogenic iPSC line would be a more stringent control than iPSCs independently generated from genetically unrelated wild-type donors. Both FPLD2 and isogenic LMNA-corrected corrected FPLD2 (‘c-FPLD2’) iPSCs expressed pluripotency markers at the transcript and protein levels and stained positive for alkaline phosphatase (Supplementary Material, Fig. S2B–E).
Correction of the LMNA gene in FPLD2 patient-derived iPSCs ameliorates endothelial differentiation potential. (A) LMNA gene editing and endothelial differentiation of LMNA p.R482W FPLD2 iPSCs and LMNA-corrected c-FPLD2 iPSCs. Sequences of the mutated and TALEN-edited LMNA genes are shown. (B) Immunostaining of PECAM1 and CDH5 on D20 of endothelial differentiation. DNA is labeled with DAPI. Bars, 50 µm. (C) RT-qPCR analysis of expression of PECAM1, CDH5 and VWF in FPLD2 and c-FPLD2 iPSCs on D20 (mean ± SD; n = 4; * P < 0.05, Wilcoxon test). (D) Filling of a scratch wound area by D20 ECs 24 h after a scratch wound in culture. Bar, 100 µm. (E) Quantification of filling of the scratched area by migrating cells (mean ± SD; n = 3; *P < 0.001, 2-way ANOVA).
We next compared the endothelial differentiation potential of FPLD2 iPSCs and c-FPLD2 iPSCs, using a protocol consisting of a 4-day mesodermal induction followed by an endothelial specification step (Fig. 2A). On day 20 (D20) of differentiation, FPLD2 iPSC-derived ECs (FPLD2 ECs) show reduced immunostaining of the endothelial markers PECAM1 (CD31) and CDH5 (CD144) compared with c-FPLD2 ECs (Fig. 2B). We also corroborate the reduced expression of endothelial markers at the mRNA level (Fig. 2C), although variations in differentiation efficiency between experiments inherent to this laminopathy model tend to mask significant differences. Further, in an in vitro scratch wound assay, FPLD2 ECs display lower gap-filling capacity than c-FPLD2 ECs (Fig. 2D and E). Moreover, in experiments aiming to validate our endothelial differentiation protocol, iPSCs reprogrammed from wild-type fibroblasts (Supplementary Material, Fig. S2B–E) exhibited EC markers and similar gap-filling capacity as c-FPLD2 ECs (Supplementary Material, Fig. S3A–C). These results suggest that FPLD2 iPSCs fail to acquire a mature endothelial phenotype, and that this can be partially or fully rescued by correcting the lamin A mutation.
Lamin A R482W deregulates developmental endothelial gene expression
We then examined the expression of T, which encodes the T/Brachyury transcription factor, during the course of endothelial differentiation. As expected, T transcripts and protein are transiently induced in FPLD2 and c-FPLD2 iPSCs within the first 4 days of mesodermal induction (Fig. 3A and B). Strikingly however, T is expressed at higher levels and over a longer period in FPLD2 iPSCs (Fig. 3A and B). Accordingly, mesodermal and endothelial genes, including the T target genes MESP1, BMP4, PDGFRA, VEGFR2, EPHB2 and TIE2 are also induced in lamin A-corrected and mutant iPSCs, and either tend to be or are significantly overexpressed in FPLD2 ECs relative to their isogenic counterparts (Fig. 3A).
Lamin A (R482W) elicits overexpression of T and of T target genes in FPLD2 iPSCs. (A) RT-qPCR analysis of expression of indicated mesodermal and endothelial genes during differentiation (D0-D8) (*P < 0.05, **P < 0.01; ***P < 0.001; two-way ANOVA and Bonferoni post-test). (B) Immunoblot of T, PDGFRα and PECAM1 during endothelial differentiation.
To provide an exhaustive representation of early gene expression programs altered by lamin A R482W, we examined by RNA-sequencing (RNA-seq) the transcriptome of FPLD2 and c-FPLD2 iPSCs during the first four days of mesodermal induction (Fig. 4A). Principal component analysis discriminates D0 and D1 time-points from later stages and segregates FPLD2 from c-FPLD2 iPSC transcriptomes more markedly on D4 (Supplementary Material, Fig. S4A). Indeed, ∼80–250 genes are differentially expressed (≥ 2-fold, α = 0.05) between FPLD2 and c-FPLD2 iPSCs on any given day of differentiation (Fig. 4A;Supplementary Material, Table S1). Both cell types upregulate markers of mesodermal induction, with FPLD2 iPSCs showing precocious upregulation of a subset of these genes (Fig. 4B). This is in contrast to markers of endodermal and ectodermal differentiation, which show similar profiles in both cell types (Fig. 4B).
Exacerbated induction of mesodermal and endothelial gene and protein expression by lamin A (R482W). (A) 2D scatter plots of gene expression in FPLD2 vs. c-FPLD2 iPSCs from D0 to D4. Orange points show differentially expressed genes (≥2-fold change; α < 0.05; numbers of genes are shown). (B) Heat map of expression of indicated markers of epithelial-to-mesenchymal (EMT) transition, mesoderm, ectoderm and endoderm from D0 to D4 (means FPKM values from RNA-seq data from triplicate differentiation experiments). (C) Venn diagrams of numbers of common (intersects) and distinct upregulated genes in FPLD2 and c-FPLD2 iPSCs between time points. (D) Fractions of upregulated genes common to both cell types (Jaccard indices). (E) Enriched GO terms for indicated gene groups. (F) Proportions of T target genes (38) uniquely or commonly upregulated in c-FPLD2 and FPLD2 iPSCs between two consecutive time points.
More specifically, analysis of genes differently regulated between the two cell types at each time-point transition shows that the proportion of genes commonly upregulated in both cell types decreases over time (Fig. 4C and D). Gene ontology enrichment analysis highlights, in c-FPLD2 cells, a transient over-representation of genes involved in cell adhesion, mitogen-activated protein kinase signaling (which regulates angiogenic gene expression) and lipid metabolism (Fig. 4E). However, FPLD2 cells overexpress genes driving circulatory functions throughout mesodermal induction, again reflecting the premature induction of these genes (Fig. 4E). Notwithstanding the fact that expression of lamin A R482W leads to endothelial inflammation in vitro (6), we did not find any differentially expressed genes related to an inflammatory phenotype between FPLD2 and c-FPLD2 cells; this is likely because such genes are expressed at later developmental stages. We conclude that during mesodermal induction, FPLD2 iPSCs acquire a unique gene expression signature caused by exacerbation of vascular developmental gene expression.
Temporal overexpression of T and polycomb target genes in lamin A R482W mutant cells
We then asked to what extent deregulation of T in FPLD2 iPSCs prompted a differential expression of T target genes during mesodermal induction. To this end, we intersected our upregulated gene sets over time (Fig. 4A;Supplementary Material, Table S1) with a published dataset of T target genes identified by ChIP-seq in human ESCs differentiated toward a mesodermal lineage (38) (Supplementary Material, Table S2). We find that the proportion of upregulated T targets specific to c-FPLD2 cells increases from D0 to D2 (Fig. 4F, green quadrants), while in contrast, the fraction of upregulated T targets that are FPLD2-specific markedly increases from D2 to D3 and D4 (red quadrant). Similar trends were found using another T target gene dataset (39) (Supplementary Material, Fig. S4B). Thus on D4, most upregulated T target genes are specific to FPLD2 cells, underscoring the exacerbation of T expression in these cells. Since T overexpression impairs terminal differentiation of several mesodermal derivatives (19), our results are also consistent with the abnormal phenotype of mesodermal tissues in FPLD2 patients (4–6).
To provide a functional attribution to genes differentially expressed on D4 vs. D0 in FPLD2 or c-FPLD2 iPSCs, we carried out a gene set enrichment analysis. The results reveal enrichment of genes target of polycomb repressor complex 2 (PRC2), a protein complex involved in maintaining, through the histone H3 lysine (K) 27 methylase activity of its subunit EZH2, a repressive state at developmental loci (40) (Supplementary Material, Fig. S4C). We also find targets of EED and SUZ12, two other PRC2 subunits, and genes marked by trimethylated H3K27 [histone 3 lysine 27 trimethylated (H3K27me3); Supplementary Material, Fig. S4C and Table S1]. Ingenuity pathway analysis of the set of genes overexpressed in FPLD2 vs. c-FPLD2 ECs on D4 highlights a network of genes encoding transcription factors included in the GO term ‘vasculature development’ [GO: 0001944]; importantly, targets of EED, SUZ12 and EZH2, including the T gene, are overrepresented in this network (Fig. 5A). The network also includes GATA2 and ISL1, two inhibitors of adipogenic differentiation and activators of cardiovascular differentiation (41) overexpressed in FPLD2 ECs, and EPAS1, a regulator of VEGFR2 and TIE2, two endothelial genes also upregulated in FPLD2 ECs (see Fig. 3A). Thus, deficiency in Polycomb-mediated repression appears to play an important role in the overexpression of T and of T target genes in mesodermal derivatives of FPLD2 iPSCs.
Differential enrichment of Polycomb target genes in FPLD2 vs. c-FPLD2 iPSCs. (A) Ingenuity Pathway Analysis of transcription factors important for vascular development and upregulated on D4 vs. D0 in FPLD2 iPSCs. Data show a network of Polycomb target genes centered on T. Shaded red items are transcription factors overexpressed in FPLD2 cells, with shades corresponding to grade of overexpression on D4. PRC2 components are shown in blue since they are not overexpressed. Lines without arrows stand for ‘binding’; arrows imply ‘acts on’. (B) H3K27me3 and EZH2 ChIP-qPCR analysis on T promoter and enhancer sites on D0 and D4 (mean ± SD; biological triplicates; *P < 0.02, paired t-test).
Lamin A R482W impairs polycomb binding to, and repression of, the T locus
Lamin A has been shown to interact with the PRC2 complex (37) and is important for proper targeting of PRC2 to target genes (12,35). Thus the defect of PRC2-mediated repression on T, together with the alteration of LADs linked to lamin A R482W (20), suggests that lamin A R482W may affect PRC2 association with T, impairing transcriptional repression at this site. We find that in D0 iPSCs, the T promoter is marked by H3K27me3 at comparable levels in FPLD2 and c-FPLD2 cells (Fig. 5B), in line with absence of T expression at this stage. On D4 however, higher H3K27me3 on T in c-FPLD2 ECs concords with weaker T expression in these cells (Fig. 5B;P < 0.02, paired t-test). Levels of EZH2 are also higher on the T promoter on D0 and D4 in c-FPLD2 ECs than in FPLD2 ECs (Fig. 5B;P < 0.01, paired t-test). Lower levels of EZH2 and H3K27me3 on T in FPLD2 ECs are consistent with persistent overexpression of the gene in these cells during mesodermal induction. The lamin A R482W mutation therefore impairs PRC2 association with the T locus, favoring overexpression of T and of endothelial and vascular genes in FPLD2 patient-derived iPSCs.
The R482W mutation confers reduced association of lamin A with the T locus in iPSCs
We next examined the ability of lamin A to associate with the T locus in mutant and corrected cells. Challenges in testing this, however, are that iPSCs harbor low levels of lamin A in early stages of differentiation (Fig. 6A) (42); moreover, overexpression of lamin A in differentiating FPLD2 compared with c-FPLD2 cells (Fig. 6A) would bias an assessment of its association with chromatin. Thus, we determined whether chromatin-lamin A complexes could be isolated by ChIP from iPSCs on D0, since lamin A levels are similar (albeit low) in FPLD2 and c-FPLD2 cells at this stage.
Reduced lamin A association with the T locus in FPLD2 iPSCs. (A) Western blot of lamins A/C and B1 during endothelial differentiation. γ-tubulin was used as loading control. (B) ChIP-qPCR analysis of lamin A binding at the T locus on D0 (mean ± SD; three biological replicates; *P < 0.05, paired t-test). Enh, enhancer; Prom, promoter; Lam+ and Lam-, control sites for lamin A binding. (C) Position of the T gene relative to lamin B1 LADs mapped in human ESCs in a separate study (43). (D) Exacerbation of T expression by the lamin A R482W mutation impairs EC function. In healthy cells, T expression is modulated by PRC2 binding enabled by a wild-type lamin A network around the T locus. In FPLD2 cells expressing lamin A R482W, reduced lamin A association with T compromises PRC2 binding. This elicits T overexpression and overexpression of T target genes early during mesodermal induction.
ChIP analysis of lamin A in FPLD2 iPSCs on D0 reveals lower lamin A enrichment on T promoter and enhancer sites than in c-FPLD2 iPSCs (Fig. 6B;P < 0.05, paired t-tests), despite similar levels detected at positive and negative control sites [Lam+, a site enriched in lamin A in other cell types (10,12); Lam−]. This differential binding likely occurs in the nuclear interior rather than at the periphery, since T, as in fibroblasts (see Fig. 1A), is found between two lamin B1 LADs mapped in human ESCs in a separate study (43) (Fig. 6C). These data suggest therefore that distinct binding of PRC2 to the T locus, and overexpression of T in differentiating FPLD2 iPSCs, are linked to impairment in lamin A association with the locus.
Discussion
By differentiating FPLD2 patient-derived iPSCs and lamin A R482W-corrected isogenic controls toward the endothelial lineage, we show in an early developmental model that the lipodystrophic lamin A R482W mutation responsible for FPLD2 disturbs the acquisition of endothelial markers. Since the R482W mutation also affects mesodermal induction in our disease model, it is probable that it also impairs differentiation along other mesodermal lineages of cardiovascular relevance. Our findings raise the hypothesis of a cell-autonomous origin of the premature onset of EC and tissue dysfunctions associated with FPLD2. These are all aggravated by metabolic alterations which characterize lipodystrophies caused by lamin A mutations (4–6). Our results also imply that the mutation elicits developmental defects at stages earlier than anticipated from phenotypes previously reported in mesenchymal precursors and differentiated cells in other models of FPLD2 (12,14,44–47). These observations argue for a unified basis of the multi-tissue phenotypes of the disease.
Overexpression of lamin A in FPLD2 ECs may synergize with the dominant-negative impact of the lamin A mutation and contribute to the phenotypes observed in this study. The LMNA gene is a T target (38,39), so lamin A overexpression could at least partly be a consequence of T overexpression. Lamin A overexpression could also result from decreased proteolytic processing, which has been shown to occur in cells expressing high levels of T/Brachyury (48). Because lamins can interact with regulatory transcription factors and signaling molecules (49–53), elevated lamin A levels in FPLD2 ECs may impair cellular functions at levels other than those directly consequent to alterations in chromatin association. In addition, through its connection to the cytoskeleton (8), a nuclear lamina of altered composition may affect the migration capacity of FPLD2 ECs. Further, defective expression of ECM remodeling genes upon endothelial differentiation reported here predicts alterations in ECM composition. This may deregulate lamin A levels, which scale with cellular matrix stiffness and in turn again impact on lineage determination (54).
We provide evidence for a differential transcriptional regulation of T protein expression in wild-type and mutant iPSCs involving lamin A. Transcriptional modulation may however not be the only step regulating T protein levels, since during mesodermal induction these are not commensurate with transcript levels. An additional mechanism may operate through increased stability of T-box transcription factors (55) at these differentiation stages in FPLD2 iPSCs. While a putative deregulation of T protein level through reduced protein degradation in FPLD2 awaits further investigation, our results outline a mechanism operating at the transcriptional level, linking lamin A and the PRC2.
Precocious expression of PRC2 target genes in FPLD2 ECs supports a role of early deregulation of Polycomb in the emergence of laminopathy phenotypes. A plausible mediator of premature mesoderm induction is exacerbated Activin/Nodal signaling, a Polycomb effector pathway prominent in differentiating FPLD2 iPSCs. Moreover, inhibition of EZH2 promotes human ESC differentiation into mesoderm via a reduction of H3K27me3 (56). H3K27me3 inhibition is also important for mesodermal differentiation of human ESCs (57). Polycomb also modulates mesenchymal lineage specification by repressing WNT genes via the H3K27 methylase activity of EZH2 (58,59). Additionally, EZH2 regulates tumor-suppressor expression to prevent cellular senescence, a feature of laminopathies (3). Overexpression of developmental genes under Polycomb control in mesodermal progenitors (12) may also impair stem cell renewal in adult tissues, accounting for the delayed onset of the disease. Thus, defective PRC2 function at key developmental regulators may impair differentiation into multiple lineages and underlines an under-evaluated early developmental origin of laminopathies.
Our results support a role of the lamin A R482W mutant in mis-targeting PRC2 to target genes in an early developmental model of FPLD2. Reduced H3K27me3 shown in regulatory regions of T, a central inducer of the mesodermal lineage, could result from increased binding or activity of the histone lysine demethylases UTX/KDM6A (60) or JMJD3/KDM6B (61). Both can demethylate H3K27me3 and are important for lineage specification (60,62). Nevertheless, we show that H3K27me3 reduction on T in FPLD2 ECs parallels lower binding of EZH2. Since repairing the LMNA mutation augments PRC2 binding and H3K27me3 at the locus, we conclude that the R482W mutation impairs PRC2 targeting to the T locus. Given the importance of lamin A in PRC2 distribution in mesenchymal progenitors (12,35), the mutation probably affects PRC2 targeting to other target genes.
The R482W mutation also impairs association of lamin A with chromatin, as we show here at the T locus in FPLD2 iPSCs and in patient fibroblasts, and whether this in turns also affects binding of PRC2 to T is a possibility. Altered association of lamin A with chromatin in FPLD2 cells probably also occurs on a broad scale, as suggested by the reorganization of lamin A LADs in the nuclear interior in FPLD2 fibroblasts (20). Epigenetic and spatial rearrangements of chromatin have also been reported at the anti-adipogenic microRNA MIR335 locus in an adipose stem cell model of FPLD2 (12). Thus, the lamin A mutation affects chromatin architecture on a broad scale in several models of FPLD2, as well as in patient cells.
We propose a model where lamin A is implicated in the developmental repression of T by facilitating PRC2 tethering to the locus (Fig. 6D). In contrast, in FPLD2 cells, reduced lamin A association with T hinders PRC2 recruitment, directly or by disfavoring the formation of macromolecular complexes engaging with lamin A and PRC2. More broadly, the R482W mutation may impinge on the binding of regulatory complexes to the T locus during iPSC differentiation, leading to unscheduled endothelial gene expression and defective EC maturation. Our findings connect the lipodystrophic lamin A R482W substitution to a disruption of early mesodermal induction. Altered early developmental gene expression by the mutation may disrupt the fate of several cell lineages, resulting in multi-organ-associated clinical phenotypes.
Materials and Methods
Fibroblasts, iPSC derivation and LMNA gene editing
Studies presented here were approved by the institutional review board of Hôpital Saint Antoine, Paris, France (CPP Ile-de-France-V b-3–15 to Corinne Vigouroux). Skin fibroblasts were purchased from Lonza (cat. no. CC-2511; male, adult, age undisclosed; ‘CTL no. 1’). Fibroblasts were also obtained from a healthy donor (male, 20 years; CTL no. 2) and from five FPLD2 patients with the LMNA p.R482W heterozygous mutation (2): female, 43 years (FPLD2 no. 1), female, 37 years (FPLD2 no. 2), female 14 years (FPLD2 no. 3), male, 43 years (FPLD2 no. 4) and female, 46 years (FPLD2 no. 5). All fibroblasts were used between passages 4 and 7.
Fibroblasts from FPLD2 patient FPLD2 no. 5 were cultured in DMEM/F12 containing 10% fetal calf serum and reprogrammed at passage 6 using the CytoTune Sendai virus-iPS Reprogramming Kit (Life Technologies). Briefly, plated cells were infected with virus for 24 h and cultured in E6 medium (STEMCELL Technologies) supplemented with 10 ng/ml FGF2 (PeproTech). After 3 weeks, colonies were picked, screened and selected based on expression of pluripotency markers. Reprogrammed cells were then cultured in mTeSR (Life Technologies) on Matrigel (Corning) as described (STEMCELL Technologies, Manual 29106).
A genetically stable FLPD2 iPSC line was selected for correction of the LMNA c.1444C > T substitution using TALENs (Cellectis Bioresearch) and homologous directed recombination. Briefly, FLPD2 iPSC were co-transfected by electroporation (Human Stem Cell Nucleofector Kit 2, Nucleofector 2b, Lonza) with TALEN effectors and the exchange matrix containing the puromycin resistance cassette, and plated on matrigel. After 5 days, recombined clones were selected by puromycin (0.5 μg/ml) for 7 days. Puromycin-resistant clones were screened by PCR to identify lamin A wild-type homozygotes. Correction of the mutation was confirmed by Sanger sequencing. Wild-type BJ iPSCs have been described previously in (63).
Antibodies
The following antibodies were used for immunofluorescence: anti-PECAM1 (Santa Cruz JC70 sc-53411, 1: 200), NANOG (Cell Signaling Technology D73G4 no. 4903; 1:200), OCT4 (Biovision No. 3576; 1:200), T/Brachyury (R&D Systems AF2085, 1:100), SOX2 (Millipore AB5603; 1:200), SSEA-3/4 (Millipore MAB1435; 1:100), TRA-1–60 and TRA-1–81 (Millipore MAB4360 and MAB4381; 1:100), CDH5 (Santa Cruz C19 sc-6458, 1:100), vWF (Santa Cruz H300 sc-14014, 1:100). The following antibodies were used for Western blotting: anti-lamin A/C (Santa Cruz H102 sc-20680), lamin B1 (Abcam 16048), T (R&D Systems AF2085), PDGFRα (Cell Signaling Technology D1E1E no. 3174), γ-tubulin (Sigma-Aldrich T5326), β-actin (Sigma-Aldrich A5441). Antibodies used for ChIP were again lamin A/C (Santa Cruz sc-7292-X; 10 µg/ChIP), lamin B1 (Abcam 16048; 10 µg/ChIP), EZH2 (Diagenode pAb-039–050; 2.5 µg/ChIP) and H3K27me3 (Diagenode C15410069; 2.5 µg/ChIP).
Mesodermal induction and endothelial differentiation
Mesoderm differentiation was induced on D0 in STEMPro34 medium with 2 mM GlutaMAX (ThermoFisher Scientific), 50 µg/ml ascorbic acid (Sigma-Aldrich), 10 ng/ml BMP4 (R&D Systems), 25 ng/ml Activin A (R&D Systems) and 5 ng/ml FGF2 (Peprotech) on matrigel-coated plates (Corning). On D4.5, endothelial differentiation was induced in STEMPro34 containing 10 μg/ml VEGF (R&D Systems) and 150 ng/ml of DKK1 (R&D systems) was added on D5. On D7, endothelial progenitors were detached with TrypLE Express and replated on Matrigel-coated plates in EGM2 medium (Lonza). Medium was fully renewed daily until D20.
Immunofluorescence and Western blotting
For immunofluorescence, cells were fixed in 3.2% paraformaldehyde, blocked in PBS/3% BSA/0.1% TX100. Primary antibodies were diluted in PBS/0.01% TX100/3% BSA. Secondary antibodies coupled to Alexa 488 (ThermoFisher Scientific) were diluted 1:1000 in PBS/0.01% TX100/3% BSA. Cells were incubated in DAPI or DRAQ5 for DNA labeling and mounted in Fluoromount-G. Cells were observed under a LeicaSP2 confocal microscope. For Western blotting, proteins were resolved by SDS-PAGE, transferred to nitrocellulose and signals detected by enhanced chemiluminescence.
Scratch wound assay
For in vitro scratch wound assays, iPSC-derived ECs on D20 were detached from their support with TrypLE Express and plated on 8-well slides containing a silicon insert (Ibidi). After 24 h, cells were treated with 1 µg/ml mitomycin C (Sigma-Aldrich) for 3 h, the insert was removed (this created the scratch) and scratched areas were imaged by time-lapse microscopy (Olympus IX83) for 24 h. Images were analyzed with the Wimasis WimScratch Quantitative Wound Healing Image Analysis package (Ibidi).
Chromatin immunoprecipitation
ChIPs of lamin A and lamin B1 were done as described in (31) from 107 cells per ChIP. Briefly, cells were crosslinked in 1% formaldehyde, lysed in ChIP lysis buffer (50 mM Tris-HCl pH 7.5, 10 mM EDTA, 1% SDS, protease inhibitors and 1 mM PMSF) and sonicated 4 times 10 min in a Bioruptor (Diagenode). After sedimentation, the supernatant was diluted in RIPA buffer (140 mM NaCl, 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, protease inhibitors, 1 mM PMSF). Chromatin was incubated overnight at 4°C with anti lamin A/C antibodies (10 µg antibody/107 cells; Santa-Cruz sc-7292-X) or anti-lamin B1 antibodies (10 µg antibody/107 cells; Abcam ab16048) coupled to Dynabeads Protein A/G (Invitrogen). ChIP samples were washed four times in ice-cold RIPA, crosslinking was reversed and DNA eluted for 6 h at 68°C in 50 mM NaCl, 20 mM Tris-HCl pH 7.5, 5 mM EDTA, 1% SDS and 50 ng/µl Proteinase K. DNA was purified using phenol-chloroform isoamylalcohol, RNAse-treated and dissolved in H2O prior to Illumina library preparation and sequencing. ChIP of H3K27me3 and EZH2 was done as described (31).
Input and ChIP DNA were analyzed by qPCR using primers listed in Supplementary Material, Table S3. For sequencing analysis of lamin A and lamin B1 ChIPs, libraries were prepared as per Illumina protocol and sequenced on an Illumina HiSeq4000. Scripts were written in Perl (64) or R (65). Lamin ChIP-seq reads were mapped to hg19 using Bowtie v2.2.3 (66) with default settings. Mapped reads were used to call domains (LADs) using enriched domain detector (67). Browser files were generated by getting a ratio of ChIP/input for each of 10 kb bins with input normalized to the ratio of total ChIP reads/total input reads.
Gene expression analysis
Total RNA was isolated using Nucleospin RNA II (Macherey-Nagel) and cDNA synthesized from 1 µg RNA using the High capacity cDNA reverse transcription kit (Applied Biosystem). Quantitative PCR was done in a LightCycler 480 using SYBR Green (Roche) (95°C 10 min and 40 cycles of 95°C 10 s/58°C 5 s/72°C 10 s) and primers listed in Supplementary Material, Table S3.
RNA-seq reads were mapped to human genome build hg19 from Illumina iGenomes using Tophat2 v2.1.0 with options -G and –b2-very-sensitive (68,69). Transcript abundance was estimated using cufflinks v2.2.1 with options -G and –b (70). Differential gene expression was determined using cuffdiff v2.2.1 with options -b and -u (70). Additional scripting was done in Perl and R (65). Scatter plots were generated using FPKM values from cuffdiff and plotted using ggplot2 in R. Principle component analysis was done using cummerbund (71). Gene ontology terms were generated using Gorilla (72) from gene lists generated using cuffdiff and all human genes as background. Gene set enrichment analysis (www.broadinstitute.org/GSEA) was done from pre-ranked gene lists using 1000 gene set permutations, gene set sizes of 15–500, weighted sets and FDR q-value < 25% as significance cutoff.
Identification of T target genes
T target genes were those identified by ChIP-seq in two studies (38,39). Bound regions in promoter or intragenic regions were associated to UCSC identification numbers for the corresponding gene. When intergenic, peaks were assigned to the nearest gene within 50 kilobases upstream or downstream (38).
Transcription activator-like effector nuclease
TALEN genomic binding sites were designed and produced by Cellectis Bioresearch. TALEN genomic binding sites were 17 base-pairs (bp) (TALEN left: 5’-TGCTGACTTACTGGTTC-3’; right: 5’-ACCCTGAAGGCTGGGCA -3’) separated by a 12 bp spacer. Efficiency and specificity of TALEN-induced modification of the LMNA gene was validated by SURVEYOR assay (Transgenomic). Briefly, TALENs were transfected into HEK-293T cells and genomic DNA extracted after 72 h. LMNA was amplified by PCR using primers 5’-CCAAGAGCCTGGGTGAGCCTC-3’ and 5’-GACACTTACCCCAGCGCTCC-3’. PCR products were re-annealed, digested by Surveyor nuclease and subjected to electrophoresis to confirm TALEN-induced mutations. The homologous recombination matrix was inserted into a pSico-PGK-puro plasmid (no. 11586 Addgene) containing a Puromycin resistance cassette flanked by to LoxP sites. The plasmid was mutated to create a PacI site. The left homologous arm (600 bp) contained the intron preceding LMNA exon 8 was inserted between the ApaI and PacI sites. The right homologous arm (600 bp) contained LMNA exon 8 and 9 including the mutation correction codon (CGG) and was inserted between the HpaI and XhoI sites. To prevent cleavage of the exchange matrix, silent mutations were introduced within each TALEN binding site (TALEN left: 5’-TGTTGACCTACTGGTTC-3’; right 5’-ACCCTTAAGGCTGGCCA-3’).
Isolation of targeted clonal cell populations
FPLD2 iPSCs were co-transfected with TALENs and exchange matrix by electroporation. On day of electroporation, cultures were dissociated into single cells with 0.025% trypsin (Promocell), resuspended in 100 µl Human Stem Cell Nucleofector Kit 2 (Lonza) and electroporated with 10 μg TALEN pair (5 µg/TALEN) and 40 μg homologous recombination matrix. Electroporated cells were plated on matrigel and cultured for 5 days. Clones were selected with Puromycin for 7 days; resistant colonies were transferred onto matrigel-coated plates and DNA isolated for PCR screening after 7 days. To screen for recombined clones, the junction between Puromycin cassette and the right homologous arm containing the correction was amplified using primer 5’-GTCACCGAGCTGCAAGAACT-3’ in the cassette and primer 5’-GGAGCGCTGGGGTAAGTGTC-3’ in the right arm. PCR products were sequenced. To genotype LMNA homozygote wild-type clones both alleles were amplified using primers LMNA8F: 5’-CCAAGAGCCTGGGTGAGCCTC-3’ and LMNA8R: 5’ GACACTTACCCCAGCGCTCC-3’. Presence of both wild-type alleles was confirmed by Sanger sequencing.
3D genome modeling and visualization
One hundred models from each of two control and five FPLD2 fibroblast cell cultures (CTL no. 1–2, FPLD2 no. 1–5) were generated using Chrom3D, after Monte Carlo optimization from constraints imposed by Hi-C and lamin ChIP-seq data (20). We used Hi‐C data for IMR90 diploid human fibroblasts (downloaded from NCBI GEO GSE63525) (32) and lamin A ChIP-seq data generated by us previously (CTL no. 1–2, FPLD2 no. 1–4) (20) and here (FPLD2 no. 5). Each chromosome was modeled as a chain of beads, where each bead represents a contact domain (TAD) determined from the Hi-C data. Overlapping domains were merged, and regions not covered by a contact domain were assigned a bead of size proportional to the corresponding genomic region. Sizes of all beads in the model were scaled so that total bead volume was 15% of the volume of a 5‐μm radius modeled nucleus. Significant pairwise interactions between beads were determined from the Hi‐C data as described in (20) and yielded 4208 interactions. These interactions constrained the corresponding bead pairs towards each other by minimizing the Euclidean distance between them during the optimization. In addition, beads corresponding to regions overlapping lamin A LADs (determined from lamin A ChIP-seq data specific to each cell type) were constrained towards the nuclear periphery by minimizing the Euclidean distance of the surface of the bead to the nucleus edge. Radial distributions of the T locus were calculated from each of the 100 optimized Chrom3D models for each cell type, by selecting the contact domain where the T gene is found (chr6: 166520000–166670000). 3D images were generated using UCSF Chimera (73).
Genome data viewing
Genome browser views of LADs were prepared using Integrated Genomics Viewer (74).
Accession number
Accession number for RNA-seq and ChIP-seq data reported in this article is NCBI GEO GSE98675.
Supplementary Material
Supplementary Material is available at HMG online.
Acknowledgements
We thank Romain Morichon (UMS LUMIC platform, Sorbonne Université, Centre de Recherche Saint-Antoine, Inserm U938, Paris, France) for photonic microscopy, Jean-Pierre Siffroi (Genetics and Embryology Department, APHP Trousseau Hospital, Paris, France) for iPSC characterization, Anita Sørensen and Kristin Vekterud (University of Oslo) for laboratory assistance, and Jacqueline Capeau and Bruno Fève (Sorbonne Université, Centre de Recherche Saint-Antoine, Inserm U938, Paris, France) for facilitating the project.
Conflict of Interest statement. None declared.
Funding
This work was supported by South East Health Norway, the Research Council of Norway, The Norwegian Center for Stem Cell Research (P.C.), EU Scientia Fellowship FP7-PEOPLE-2013-COFUND no. 609020 (N.B.); Région Ile-de-France STEM-Pole (N.B.), Fondation de France (N.B.), Société Francophone du Diabète/Pierre Fabre Médicament (C.V.), INSERM, Sorbonne Université (C.V.) and a grant Investissement d’Avenir (ANR-10-IAIHU-05) to the Institute of CardioMetabolism and Nutrition (ICAN) (A.-C.G., C.V. and J.-S.H.).
References
Author notes
Equal first authors.
