A novel truncating variant of GLI2 associated with Culler-Jones syndrome impairs Hedgehog signalling

Background GLI2 encodes for a transcription factor that controls the expression of several genes in the Hedgehog pathway. Mutations in GLI2 have been described as causative of a spectrum of clinical phenotypes, notably holoprosencephaly, hypopituitarism and postaxial polydactyl. Methods In order to identify causative genetic variant, we performed exome sequencing of a trio from an Italian family with multiple affected individuals presenting clinical phenotypes in the Culler-Jones syndrome spectrum. We performed a series of cell-based assays to test the functional properties of mutant GLI2. Results Here we report a novel deletion c.3493delC (p.P1167LfsX52) in the C-terminal activation domain of GLI2. Functional assays confirmed the pathogenicity of the identified variant and revealed a dominant-negative effect of mutant GLI2 on Hedgehog signalling. Conclusions Our results highlight the variable clinical manifestation of GLI2 mutations and emphasize the value of functional characterisation of novel gene variants to assist genetic counselling and diagnosis.


Introduction
The Hedgehog (Hh) family of secreted morphogens control cell proliferation, differentiation and patterning during embryo development [1]. In humans, gene mutations in molecular Spectrophotometer 1000 (Thermo Fisher Scientific, Wilmington, DE, USA). Library preparation was performed using Illumina Nextera Expanded Rapid Capture Enrichment Exome. Exome sequencing was carried out on Illumina HiSeq 2500 platform (Illumina, Inc. San Diego, CA, USA) using SBS chemistry. Libraries were sequenced in paired end mode, 101 nucleotides long each.
This study was approved by the Institutional Ethical Review Committee, San Raffaele Hospital, Milan, Italy (prot. RARE-DISEASE) and informed consent was obtained from all participants.
The following chain of filters was applied to the variant set: GATK VQSLOD > 0, predicted change in coding sequence, segregation according to a dominant model, rarity in the population (COMMON = 0 in dbSNP build 137), predicted to be damaging according to SIFT [30] and Polyphen [31]. Genes were prioritized using the Phenolyzer platform [32].

Quantitative real-time PCR
NIH-3T3 cells were transfected in 12-well plate format with the indicated plasmids using Lipofectamine 2000 [1 μg DNA:1.5 μl Lipofectamine per well]. After 24 hrs, cells were treated with SHH (16 nM) in DMEM containing 0.5% FBS for 40 hrs. 1 μg of total RNA extracted with Trizol (Thermo) was reverse transcribed into cDNA with M-MLV Reverse Transcriptase using random primers, and analyzed by SYBR green-based real-time quantitative PCR (SYBR Select Master Mix, Thermo) with the following primer sets:

Clinical phenotype of the patients
The proband was previously tested negative for mutations in candidate disease-linked genes POUF1, PROP1, HESX1, LHX3 and GLI3.

Exome sequencing identifies GLI2 as candidate disease gene
Exome sequencing was performed on the trio (Fig 1A) at average target coverage of 30x for each sample (S1 Table). After variant calling, 130 rare or novel variants in 125 genes were found segregating according to a dominant model and affecting coding sequences. Once filtered for putative pathogenicity, 40 variants in 40 genes were retained (S2 Table). Candidate genes were prioritized based on their association with the clinical phenotypes of the patients using Phenolyzer software. This analysis unambiguously identified GLI2 as the top-scoring gene, while other candidates (WDR34, DPAGT1, ASXL1) already known to be associated with query phenotypes were ranked significantly lower (Fig 1B). Both the proband and her father were found to carry a novel heterozygous mutation caused by the deletion c.3493delC in the GLI2 gene, leading to a frameshift and premature stop of translation at residue 1218 (p.P1167LfsX52).

The mutation p.P1167LfsX52 converts GLI2 into a dominant-negative transcriptional repressor
The frameshift mutation c.3493delC in GLI2 truncates the C-terminal portion of the transactivation domain required for transcriptional activity (p.P1167LfsX52, Fig 2A). To determine whether this truncation alters GLI2 function, we generated GFP-tagged constructs of either wild-type GLI2 or the p.P1167LfsX52 mutant (hereafter GLI2 MUT ) and examined their functional properties using cell-based assays. As predicted, GLI2 MUT revealed by western blotting in transfected HEK293 cells was smaller than the wild-type protein (~129kDa vs. 167kDa; 156kDa vs. 194 kDa after GFP fusion) (Fig 2B). However, the subcellular distribution of GLI2 MUT expressed in NIH-3T3 mouse fibroblasts, which respond to SHH, was similar to that of wild-type GLI2: both proteins were found in the cytoplasm as well as the nucleus in untreated cells and accumulated within the nucleus following stimulation with SHH (Fig 2C-2F). Nevertheless, despite normal nuclear targeting, GLI2 MUT was unable to induce expression of the transcriptional targets of SHH signaling GLI1 and Ptch1, whose mRNA levels were instead substantially higher in SHH-treated cells overexpressing wild-type GLI2 relative to mock-transfected controls (Fig 2G and 2H).
To directly investigate the effects of the p.P1167LfsX52 mutation on GLI2 transcriptional activity, we assessed the ability of GLI2 MUT to stimulate a GLI-dependent reporter construct (8xGliBS-Luc) in which a promoter containing tandem GLI responsive elements drives expression of firefly luciferase upon activation of the Hedgehog pathway [16]. In NIH-3T3 cells, overexpression of wild-type GLI2 increased reporter activity in a ligand-independent manner to an extent comparable to control cells treated with SHH. Conversely, the basal levels of reporter activity were significantly lower in cells transfected with GLI2 MUT and did not show the expected increase after SHH stimulation (Fig 2I).
A complementary set of experiments was conducted in primary cultures of neural progenitor cells derived from the chick embryo neural tube electroporated with either wild-type or mutant GLI2 together with the 8xGliBS-Luc reporter. Spinal cord progenitors depend on graded SHH signaling to acquire class-specific molecular identities during embryo development [35] and exhibit reliable dose-dependent responsiveness to SHH in culture (Fig 2J, control). Electroporation of wild-type GLI2 led to robust induction of reporter activity independent of SHH stimulation, whereas GLI2 MUT caused a considerable reduction in luciferase levels compared to control cells at all ligand concentrations tested, indicating that the mutant protein suppresses transcription mediated by endogenous GLI factors (Fig 2J).  In conclusion, the truncated mutant p.P1167LfsX52 functions as a transcriptional repressor that exerts dominant-negative effects on GLI-dependent gene expression.

Discussion and conclusions
This study expands the spectrum of GLI2 mutations reporting a novel heterozygous pathogenic variant (p.P1167LfsX52) that results in autosomal-dominant developmental abnormalities including polydactyly, hypopituitarism, GHD and hypothyroidism.
Functional studies based on cellular assays demonstrated that the frameshift mutation p. P1167LfsX52 truncates the C-terminal transactivation domain of GLI2 generating a transcriptional-repressor form that retains the ability to translocate into the nucleus in response to SHH but exhibits dominant-negative activity. As a result, we observed a significant inhibition of GLI reporter levels in cells expressing GLI2 p.P1167LfsX52, indicating that the activity of wild-type GLI proteins is suppressed by the mutant variant. Dominant-negative activity was reported for other pathogenic variants of GLI2 with deletions in the activation domain [24]. The inhibitory effect was found to require integrity of the DNA-binding and amino-terminal transcriptional repressor domains [24,36], which are intact in GLI2 p.P1167LfsX52. To inhibit positive GLI function, C-terminally truncated variants may compete with and displace wildtype GLI2 from target sites and/or form inactive complexes with the activating forms.
While GLI2 mutations were originally identified in patients with HPE and midline abnormalities, [23,24], more recently it became clear that GLI2 variants are often associated with polydactyly, pituitary deficiency and subtle midfacial facial phenotypes rather than patent HPE [20,21,37,38]. The fact that frank HPE is rare in patients with GLI2 mutations, in contrast to those with SHH variants, has suggested that other GLI proteins (GLI1, GLI3) might function redundantly to compensate in part for GLI2 deficiency, in line with studies in compound mutant mice [16,39,40]. Likewise, GLI2-null mice display normal limb patterning unless GLI1 is also ablated, and GLI2/GLI3 double heterozygous mice have a more severe polydactyly than GLI3 mutants [40]. Interestingly, polydactyly is generally present in patients with more severe GLI2 variants, including those that disrupt the zinc-finger and transactivation domains [39]. Specifically, individuals with mutations predicted to result in protein truncation are significantly more likely to present both polydactyly and pituitary insufficiency compared to those with non-truncating variants [20].There is striking variability in the phenotypic outcomes of GLI2 mutations even within the same family tree [21][22][23][24]37], ranging from unaffected carriers to patients with craniofacial abnormalities, pituitary phenotypes and polydactyly, either isolated or in combination. In the pedigree examined in this study, hormone deficiencies, but not hand and facial anomalies, were present in all individuals carrying mutant GLI2, in support of the recommendation to consider this gene as a primary candidate to screen after endocrinology testing has revealed pituitary insufficiency even in the absence of polydactyly [41]. Incomplete penetrance and variable phenotypes, as also reported in patients with autosomal dominant mutations in other loci linked to HPE or pituitary deficiencies (e.g., after treatment with SHH (16nM) for 40 hrs. All conditions are normalized to untreated control cells (mean ± SEM, n = 2). Unpaired ttest, ( �� ) p<0.01 GLI2 MUT vs. GLI2 WT either untreated or SHH-treated matching conditions. (I) Luciferase-based reporter assay with GLI-responsive construct 8x-Gli-BS-Luc in NIH-3T3 cells transfected with GFP-tagged GLI2 WT , GLI2 MUT or GFP control, before and after treatment with SHH (16nM) for 30 hrs. The expression levels of the reporter gene are measured by luciferase activity. All conditions are normalized to untreated control (mean ± SEM, n = 2). Unpaired t-test ( ��� ), p<0.001 GLI2 MUT vs. control either untreated or SHH-treated matching conditions. (NS, non-significant) p = 0.1264 GLI2 MUT untreated vs. treated. (J) Luciferase-based assay with 8x-Gli-BS-Luc reporter in chick spinal cord progenitor cells expressing GFP-tagged GLI2 WT , GLI2 MUT or GFP control, treated with increasing doses of SHH for 24 hrs. All conditions are normalized to untreated control (mean ± SEM, n = 2-4). Unpaired ttest, ( � ) p = 0.0137 GLI2 MUT vs. control, untreated; ( ��� ) p<0.001 GLI2 MUT vs. control at corresponding SHH concentrations.
https://doi.org/10.1371/journal.pone.0210097.g002 SHH, SIX3, OTX2, HESX1), suggest the contribution of additional genetic variants, epigenetic changes and environmental factors. Despite the pathogenetic role of GLI2 mutations has already been described, a small number of rare and damaging variants in the same gene can be found in public data from non-dysmorphic individuals such as the Exome Aggregation Consortium [42] or Exome Variant Server [43]. Therefore, systematic functional validation of putative pathogenic variants would be valuable for genetic counseling and patient screening.
Supporting information S1 Table. Exome sequencing statistics. Main sequencing statistics for each sample. All samples were sequenced at 30x coverage minimum. (DOCX) S2 Table. Selected variants after filtering. Genomic positions and annotation of variants selected for disease gene prioritization. Only rare, putative damaging variants segregating in the pedigree were selected. Annotation was performed on GRCh37.34 genome version. (DOCX)