Loss-of-function mutations in TGFB2 cause a syndromic presentation of thoracic aortic aneurysm

Abstract

The TGF-β family comprises three cytokines that regulate multiple aspects of cellular behavior including, proliferation, differentiation, migration and specification of synthetic repertoire¹. Postnatally, TGF-β activity is most closely linked to wound healing, productive modulation of the immune system and multiple pathological processes, including cancer progression and tissue fibrosis². Fibrillin-1, encoded by FBN1, the gene product altered in Marfan syndrome (MFS)³, binds to the latent complex of TGFβ and regulates the release of active molecules in the extracellular environment^4,5. LDS is a syndromic presentation of aortic aneurysm that is most often caused by mutations in the genes that encode the subunits of the TGF-β receptor, TGFBR1 and TGFBR2^6,7. Mutations in SMAD3, which encodes an intracellular mediator of TGF-β signaling, have also been described in individuals with phenotypic manifestations of LDS⁸. Three TGFβ ligand isoforms exist in humans (TGFβ1, -β2, and -β3, encoded by separate genes, TGFB1, TGFB2, and TGFB3, respectively) but their relative contributions to aortic aneurysm in the context of connective tissue disorders has not been explored.

The precise role of TGFβ signaling in aneurysm progression remains controversial. On the one hand, analyses of the aortic wall obtained from patients and mouse models have consistently shown the signature of increased TGF-β signaling for MFS, LDS, cutis laxa with aneurysm, bicuspid aortic valve with aneurysm and isolated familial thoracic aortic aneurysm caused by mutations in smooth muscle cell contractile proteins^6,9–11. This signature includes increased phosphorylation and nuclear translocation of the receptor activated SMAD proteins (SMAD2 and SMAD3), increased expression of TGF-β–responsive gene products (for example, collagen, connective tissue growth factor (CTGF) and plasminogen activator inhibitor-1) and/or increased activation of noncanonical TGF-β _signaling cascades (prominently including ERK1 and ERK2 (ERK1/2))¹². In mouse models of MFS, antagonism of TGF-β signaling using either TGFβ neutralizing antibodies or angiotensin receptor blockers attenuates multisystem disease manifestations, including aortic aneurysm^11,13,14. On the other hand, an intuitive consideration of the primary consequence of many disease-associated mutations suggests the potential for loss of TGFβ signaling⁷. For example, while fibrillins may contribute to negative regulation of TGFβ signaling by sequestering ligand, they can also positively regulate signaling by concentrating cytokine at sites of intended function¹⁵. Most LDS mutations involve substitution of conserved residues in the kinase domains of TGFβ receptor subunits and recombinant expression of receptors harboring LDS mutations in cells naïve for the corresponding receptor subunit fails to support canonical (SMAD-dependent) TGFβ signaling¹⁶. At least some of the SMAD3 mutations that cause LDS are expected to confer functional haploinsufficiency by virtue of an early premature termination codon that induces accelerated decay of mutant mRNA and thus reduce signaling efficiency⁸. Furthermore lineage-specific abrogation of TGFβ signaling can impair aortic wall homeostasis^17,18. It has been our hypothesis that this apparent paradox could be reconciled if compensatory paracrine or autocrine events in response to a relative loss of TGFβ signaling potential leads to functional overshoot¹⁹. Here we describe identification and mechanistic characterization of a new gene for a syndromic aneurysm presentation within the LDS spectrum that lends validity to this pathogenic model.

We identified eight families with an autosomal dominant aortic aneurysm phenotype with variable clinical expression. Features shared with MFS and LDS include aortic aneurysm, pectus deformity, arachnodactyly, scoliosis, and skin striae. Features shared with LDS but not MFS include hypertelorism, bifid uvula, bicuspid aortic valve, arterial tortuosity, club feet, and thin skin with easy bruising (Table 1, Fig. 1, and Supplementary Table 1). Ectopia lentis was not observed. Microarray analysis in two patients with these features who also had mild developmental delay, revealed two unique, heterozygous de novo chromosomal microdeletions at 1q41 (Fig. 1&2A). The deletion in one proband measures 6.5 Mb (215.5 Mb – 222.1 Mb; GRCh37/hg19) and encompasses 20 genes, whereas the deletion in the other is only 3.5 Mb (216.6 Mb – 220.2 Mb). Both deletions include the TGFB2 gene, which encodes transforming growth factor β2 (TGFβ2), making it an obvious candidate gene for this aneurysm phenotype with MFS- and LDS-like features. We subsequently sequenced all exons and intron-exon boundaries of the TGFB2 gene in a cohort of 86 aneurysm patients (34% familial) who were negative for FBN1 and TGFBR1/2 mutations. We identified a total of six heterozygous mutations in TGFB2, including one nonsense mutation, three missense mutations and two intragenic deletions, one in-frame and one causing a frameshift (Fig. 2B and Supplementary Figure 1). The three missense mutations, p.Arg327Trp, p.Arg330Cys and p.Pro366His, resulted in the substitution of evolutionarily conserved residues (Fig. 2B) and were categorized as probably damaging by Polyphen²⁰, deleterious by SIFT²¹ and disease-causing by MutationTaster²². In addition, these mutations were not observed in the 1000 Genomes Project²³ and were not present in over 10,000 exomes in the NHLBI (USA) Exome Variant Server. All participating affected individuals tested positive for their family-specific TGFB2 mutation (Fig. 2C). Because we found two whole gene deletions in addition to two nonsense mutations predicted to lead to nonsense-mediated mRNA decay (p.Tyr99* and p.Tyr369Cysfs*26), we propose haploinsufficiency as the relevant mechanism.

Open in a separate window

Figure 1

Phenotypic characteristics of patients with TGFB2 mutation

Significant clinical features of individuals with TGFB2 mutations include mild hypertelorism (widely spaced eyes; 1-II:1, 3-III:1 and 7-III:1), malar hypoplasia (flat cheek bones;1-II:1,3-III:1, 4-II:1 and 7-III:1), retrognathia (receding chin; 1-II:1, 3-III:1, 4-II:1 and 7-III:1), arachnodactyly (long fingers; 1-II:1 and 4-II:1), pectus excavatum (7-III:1), pes planus (flat feet; 1-II:1 and 3-III:1) and hammer toes (1-II:1). Permission to publish photographs was obtained from the affected individuals or their parents.

Open in a separate window

Figure 2

Mutational Analysis of TGFB2 in aneurysm patients

(A.) Schematic representation of the microdeletions on chromosome 1q41. The TGFB2 gene is indicated in red. Pedigrees for two patients with de novo chromosomal microdeleletions completely overlapping TGFB2 (1-II:1 and 2-II:1) are shown. (+) Indicates presence of the described mutation in an associated individual while (−) indicates lack of mutation. (B-C.) TGFB2 mutations and pedigrees for families 3–8. Pedigrees document an autosomal dominant pattern of inheritance. Mutations are annotated at the nucleotide (c.) and protein level (p.; three letter code for amino acids is used; reference transcript: Ensembl ENST00000366929 or NCBI NM_001135599.2). Circle, female; square, male; open symbol, unaffected; shaded symbol, affected; diagonal line, deceased. The location of mutations in relation to the exons (numbers) of TGFB2 and the domain organization is shown (LAP; latency associated peptide; RKKRA potential furin cleavage site). Evolutionary conservation of the mutated residues in TGFB2 and related human cytokines (TGFB1/3) is shown.

In addition to aneurysm, individuals with LDS show arterial tortuosity with prominent involvement of the vertebral and carotid arteries^6,16. Individuals with TGFB2 mutations can have similar arterial tortuosity (Fig. 3a, subject 7:III-1). Aortic tissue taken at the time of surgery demonstrates elastic fiber fragmentation and increased collagen and proteoglycan deposition (Fig. 3B&C and Supplementary Figure 2), histopathologic findings reminiscent of both MFS and LDS²⁴. Immunohistochemical (IHC) analysis of aortic tissue from patients 7:III-1 and 5:II-2 demonstrated increased TGFβ signaling, as evidenced by increased nuclear activation of SMAD2 and pSMAD3 proteins and increased expression of TGFβ responsive gene products including collagen and connective tissue growth factor CTGF (Fig. 3C&D and Supplementary Figure 2). Whereas total TGF-β2 expression in IHC analyses was similar in affected individuals and in controls, expression of TGF-β1 ligand was higher in affected individuals (Fig. 3d).

Open in a separate window

Figure 3

Cardiovascular Pathology in Human Subjects with TGFB2 mutations

(A.) Multidetector computed tomography (MDCT) with 3 dimensional reconstruction of head and neck vessels demonstrating tortuosity of the distal cervical internal carotid arteries bilaterally (arrows, center panel) as well as the V1 segment of the left vertebral artery (arrows, left panel). MDCT imaging in modified sagittal view of dilated aorta at sinuses of Valsalva (arrows, right panel), Bars= 2 cm. (B.) Movat’s pentachrome staining of human aortic samples demonstrating an increase in proteoglycan deposition (Blue staining in Movat’s pentachrome) and elastic fragmentation (Black in Movat’s pentachrome) in MFS, LDS, and patient with TGFB2 mutation (7:III-1) versus control, Bar= 200µM, Enlargement Bar= 80µM. (C.) Masson’s Trichrome staining of human aortic samples with increased collagen deposition (Blue in Masson’s Trichrome) in MFS, LDS, and patient with TGFB2 mutation (7:III-1) versus control, Bar= 200µM, Enlargement Bar= 80µM. (D.) Immunocytochemical staining of the aortic media for phosphorylated Smad2 protein, CTGF, TGFB1, and TGFB2. Panels show control aorta (Control) and patient aorta with TGFB2 mutation (7:III-1). Quantification of fraction of pSmad2 positive nuclei (pSMAD2) or staining (CTGF, TGFB1, TGFB2) represents staining of three control aortas (Co) versus patients 7:III-1 and 5:II-2 (Pts), Error bars equal 2 SEM, (*p<0.05). Bar= 80µM.

We next examined the effect of Tgfb2 haploinsufficiency in gene-targeted mice. While homozygous knockout (Tgfb2^−/−) mice are known to show late embryonic lethality secondary to congenital heart disease²⁵, the phenotype of Tgfb2^+/− mice was not reported in detail. Patients with LDS develop aortic aneurysm in a characteristic anatomic distribution characterized by dilation of the aortic root at the level of the sinuses of Valsalva⁶. This pattern is similar in patients with MFS²⁶ and in MFS mice harboring a heterozygous fibrillin-1 mutation (Fbn1^+/C1039G)²⁷. By 8 months of age, Tgfb2^+/− mice showed dilation of the aortic annulus and root but the more distal ascending aortic dimensions were normal (Fig. 4A&B). These findings demonstrate that loss of function of a single allele of Tgfb2 is sufficient to cause aortic root aneurysm. To interrogate the state of TGFβ signaling in Tgfb2^+/− mice, we performed western blot analysis of protein lysates derived from proximal ascending aortic segments. Similar to the signaling perturbations seen in Fbn1^+/C1039G mice^12,28, aortas from Tgfb2^+/− mice showed increased phosphorylation of Smad2, Smad3, and Erk1/2, when compared to wild-type mice (Fig. 4C and Supplementary Figure 3).

Open in a separate window

Figure 4

Haploinsufficiency for Tgfb2 causes aortic root aneurysm in mice

(A.) Parasternal long axis echocardiographic systolic images of the aortic root of 8 month old wild type (n=10), Tgfb2^+/− (n=6), and Fbn1^+/C1039G mice (n=9). Arrows denote root dimension. Bar= 0.75 mm (B.) Echocardiographic quantification of dimensions at the aortic valve (AoV), aortic root (AoR), sinotubular junction (STJ), and ascending aorta (AscAo) in wild type, Tgfb2^+/− and Fbn1^+/C1039G mice at 8 months of age. (†p<0.005, †† p<0.001). There was no significant difference in aortic dimension between Tgfb2^+/− and Fbn1^+/C1039G mice at this age. (C.) Western blot analysis of murine ascending aortas demonstrating increased phosphorylation of Smad2, Smad3, and ERK proteins in 8 month old Tgfb2^+/− and Fbn1^+/C1039G mice. Graphs representing phosphoprotein western blot quantification standardized to GAPDH expression, Error bars equal 2 SEM, (*p<0.05, **p<0.01, ††p<0.001).

We next assessed for genetic interaction between targetedTgfb2^+/− and Fbn1^+/C1039G alleles. The latter is associated with high TGFβ signaling during periods of rapid aneurysm progression^11,12,29. Tgfb2^+/−:Fbn1^+/C1039G animals demonstrate normal body size and growth with normal blood pressure indices (Supplementary Figure 4). These double heterozygous mice show a significant increase in aortic root dimension when compared to either Fbn1^+/C1039G or Tgfb2^+/− mice, at 2 and 4 months of age (Fig. 5A). Aortic dilatation was specific to the aortic root, recapitulating the zone of predisposition seen in people with either MFS or LDS. Histological examination shows equivalent elastic fiber disorganization and increased collagen deposition within the medial compartment of the aortic wall in Tgfb2^+/− and Fbn1^+/C1039G animals; both findings were greatly accentuated in Tgfb2^+/−: Fbn1^+/C1039G mice (Fig. 5B& Supplementary Figure 5). IHC revealed a graded increase in nuclear accumulation of pSmad2 in the aortic media, with a pronounced increase in Tgfb2^+/−: Fbn1^+/C1039G compound aortas (Fig. 5C). Western blot analysis, which integrates the performance of all cell types within the aorta, showed a subtle but significant increase in pSmad2 levels in Fbn1^+/C1039G and Tgfb2^+/−: Fbn1^+/C1039G mice, but no increase in either pSmad3 or pERK1/2 at this early timepoint (4 months) (Supplementary Figure 5). Analysis of mRNA levels in the proximal aorta at 2 months of age revealed normal expression of Tgfb2 and Tgfb3 in all three mutant genotypes; Tgfb2^+/−: Fbn1^+/C1039G mice uniquely showed increased expression of Tgfb1 (Fig. 5D) recapitulating observations in the human aorta (Fig. 3C). This upregulation was ligand-specific as no significant changes in Tgfbr1 or Tgfbr2 expression were detected (Supplementary Figure 6). At 4 months of age both Tgfb2^+/− and Tgfb2^+/−: Fbn1^+/C1039G mice showed decreased Tgfb2 in the circulation when compared to wild type littermates, whereas this level was increased in Fbn1^+/C1039G animals (Supplementary Figure 7). While there was a trend for increased circulating Tgfb1 in all three mutant genotypes, high intragroup variability was observed (Data not shown).

Open in a separate window

Figure 5

Synergistic pathology in Tgfb2^+/−:Fbn1^+/C1039G double heterozygous mice

(A.) Photomicrographs and echocardiographic aortic root quantification of WT (n=8), Tgfb2^+/− (n=8), Fbn1^+/C1039G (n=7), and Tgfb2^+/−: Fbn1^+/C1039G (n=11) mice. Orange arrow demonstrates a large sinus of Valsalva aneurysm. (**p<0.01, †p<0.005, ††p<0.001) Bar = 1.5 mm. (B.) Worsened aortic phenotype from Tgfb2^+/−: Fbn1^+/C1039G double heterozygous mice. Panels of VVG (upper row) and Masson’s Trichrome (bottom row) stained aortas from 4 month old mice demonstrating elastin fragmentation and increased collagen deposition in Tgfb2^+/−: Fbn1^+/C1039G mice. Bar= 20 µm. (C.) Immunohistochemistry of phosphorylated Smad2 in aortas from 4 month old WT, Tgfb2^+/−, Fbn1^+/C1039G, and Tgfb2^+/−: Fbn1^+/C1039G mice. Bar= 20 µm (D.) Transcript analysis of ascending and descending aortas of two month old WT (n=3), Tgfb2^+/− (n=3), Fbn1^+/C1039G (n=3), and Tgfb2^+/−: Fbn1^+/C1039G (n=3) mice normalized to GAPDH expression, Error bars equal 2 SEM, (*p<0.05).

This study demonstrates that heterozygous mutations in the gene encoding TGFβ2 are sufficient to cause a syndromic presentation of thoracic aortic aneurysm in both people and mice. Given the substantial clinical and mechanistic overlap with LDS, categorizing this disorder within the LDS-spectrum should facilitate patient diagnosis and management.

The pathogenesis of the TGFβ vasculopathies seems complex. Whereas the previous finding of mutations affecting TGF-β receptor subunits and intracellular mediators underscored the primary role of this cytokine family in aneurysm initiation and/or progression, the paradoxical association of an unequivocal signature for increased TGF-β signaling in postnatal tissues with mutations that would intuitively impair signaling has engendered controversy regarding the mechanism for LDS spectrum disorders. Attainment of mechanistic insight has been slowed by the myriad of interactions and functions supported by TGFβ receptors and intracellular signaling mediators and by the extent of cross-talk with other signaling cascades. The recent demonstration that the cleft palate seen upon neural-crest specific silencing of TβRII in mice manifests a gain of p38 signaling that is mediated by TGF-β2 and a disease-specific TβR1-TβR3 receptor complex represents an overt example of this complexity and the potential inadequacy of intuitive disease model predictions³⁰.

Although our findings regarding TGFB2 haploinsufficiency and aneurysm represent yet another example of the same paradox, they may offer experimental opportunities to clarify the mechanism. In the absence of any primary perturbation of intracellular signaling machinery, model systems may be more tractable. One testable hypothesis that derives from our observations is that compensatory upregulation of TGF-β1 expression in the aorta contributes to aortic disease. Further studies will be needed to determine whether concomitant silencing of TGF-β1 can rescue the TGFβ2 deficiency state and if the apparent overshoot in compensation reflects the total level of all bioavailable TGF-β ligands or a specific detrimental consequence of upregulation of TGF-β1 in the aorta. Alternatively, low TGF-β signaling in restricted cell populations may be a critical determinant of postnatal disease progression, perhaps setting the stage for paracrine overdrive of adjacent cell types with a relative preservation of signaling potential. It is notable that increased TGF-β1 has been documented in the aorta in LDS spectrum patients with loss-of-function SMAD3 mutations⁸ and in the circulation of people and mice with MFS³¹.

Our observed deleterious genetic interaction between a Fbn1^+/C1039G mutation causing MFS and Tgfb2 haploinsufficiency causing a LDS-spectrum disorder is both novel and informative. MFS is the most comprehensively studied TGFβ vasculopathy, with clear evidence in support of an increased signaling state including cellular and tissue signatures that normalize in association with phenotypic rescue upon administration of TGFβ or ERK antagonists. In this light, pathogenic models that singularly invoke decreased TGFβ signaling for LDS-spectrum disorders would be difficult to reconcile with the worsening of disease seen in Tgfb2^+/−: Fbn1^+/C1039G animals. It seems notable that aortic root enlargement can be detected in Fbn1^+/C1039G and Tgfb2^+/−: Fbn1^+/C1039G mice at 2–4 months of age, before overt evidence for excessive Smad2/3 or ERK1/2 activation by western blot analysis. This may suggest that the average performance of all cells within an aortic segment (the parameter monitored by immunoblots) is less important than the presence of even small subpopulations of bad-acting (i.e. high signaling) cells. In keeping with this hypothesis, the specific ERK1/2 inhibitor RDEA-119 was able to completely suppress abnormal aortic growth in Fbn1^+/C1039G mice even when its use was restricted to age groups that did not yet show increased pERK1/2 in the aortic wall by western analyses¹². Full clarity on these issues will facilitate the development and testing of novel treatment strategies that may find broad application.

Online Methods

Supplementary Material

Acknowledgements

Footnotes

References