A common cis-regulatory variant impacts normal-range and disease-associated human facial shape through regulation of PKDCC during chondrogenesis

Curation statements for this article:
  • Curated by eLife

    eLife logo

    eLife assessment

    The findings are important and would potentially have theoretical and practical implications outside the field. However the strength of evidence presented was assessed as being incomplete in several respects. Major strengths are (1) genetic factors in facial appearance are of broad interest, and the potential influence of possibly identical factors in a serious congenital disorder (cleft lip/palate) heightens that interest further; (2) proving which single nucleotide variants influence phenotypes, and by what mechanisms, is a major challenge for the field as a whole. The weakness, as assessed, was that in its present form the experimental approach was not sufficiently rigorous to support the conclusions unambiguously.

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

Genome-wide association studies (GWAS) identified thousands of genetic variants linked to phenotypic traits and disease risk. However, mechanistic understanding of how GWAS variants influence complex morphological traits and can, in certain cases, simultaneously confer normal-range phenotypic variation and disease predisposition, is still largely lacking. Here, we focus on rs6740960 , a single nucleotide polymorphism (SNP) at the 2p21 locus, which in GWAS studies has been associated both with normal-range variation in jaw shape and with an increased risk of non-syndromic orofacial clefting. Using in vitro derived embryonic cell types relevant for human facial morphogenesis, we show that this SNP resides in an enhancer that regulates chondrocytic expression of PKDCC - a gene encoding a tyrosine kinase involved in chondrogenesis and skeletal development. In agreement, we demonstrate that the rs6740960 SNP is sufficient to confer chondrocyte-specific differences in PKDCC expression. By deploying dense landmark morphometric analysis of skull elements in mice, we show that changes in Pkdcc dosage are associated with quantitative changes in the maxilla, mandible, and palatine bone shape that are concordant with the facial phenotypes and disease predisposition seen in humans. We further demonstrate that the frequency of the rs6740960 variant strongly deviated among different human populations, and that the activity of its cognate enhancer diverged in hominids. Our study provides a mechanistic explanation of how a common SNP can mediate normal-range and disease-associated morphological variation, with implications for the evolution of human facial features.

Article activity feed

  1. eLife assessment

    The findings are important and would potentially have theoretical and practical implications outside the field. However the strength of evidence presented was assessed as being incomplete in several respects. Major strengths are (1) genetic factors in facial appearance are of broad interest, and the potential influence of possibly identical factors in a serious congenital disorder (cleft lip/palate) heightens that interest further; (2) proving which single nucleotide variants influence phenotypes, and by what mechanisms, is a major challenge for the field as a whole. The weakness, as assessed, was that in its present form the experimental approach was not sufficiently rigorous to support the conclusions unambiguously.

  2. Reviewer #1 (Public Review):

    The authors begin by showing the association between rs6740960 and facial shape, specifically that protrusion of the lower jaw and zygomatic regions, and retrusion of the entire central midface, are associated with the 'T' allele. Next they show that the enhancer harboring the SNP is active in the midface of mouse embryos with lacz transgenic reporter assays. Then they show that, interestingly, while the enhancer harboring the SNP has comparable levels of H3K27Ac in hESC derived CNCC (eCNCC) and cranial chondrocytes (eCC), only in the latter there is significant level of contact between the enhancer and the promoter of PKDCC. Next, they delete the rs6740960 cognate enhancer in two heterozygous clones and demonstrate 60% decrease in PKDCC expression at the allele bearing the enhancer deletion. This is an elegant and satisfying experiment. Next, they use ChIP-qPCR to H3K27Ac in eCNCC and eCC that are heterozygous for the SNP and show an elevated level of H3K27Ac the enhancer haplotype bearing the derived "A" allele in CNCCs and even greater in CC. This is also a clear result, although because of co-operativity among enhancers, there could be another SNP in the haplotype that leads to the difference. Finally they use micro-CT and high end morphometric analysis on mice with two, one, or zero functional Pkdcc alleles, and see correlated quantitative changes in maxilla, mandible, and palatine bone shape. Strengths of the study include analysis of allele specific expression using digital PCR, quantitative H3K27Ac-HiC, showing the SNP allele correlates with the activity of the enhancer harboring it, and a deep morphometric analysis to show the subtle effect of loss of one allele of Pkdcc on craniofacial structures in mouse model. However, no experiments incisively rule out the possibility that another SNP in the haplotype cause the effects attributed to the SNP, slightly diminishing the impact of the study.

  3. Reviewer #2 (Public Review):

    The Authors demonstrate compelling genetic evidence that the region that harbors rs6740960 plays a role in both normal craniofacial development risk for craniofacial disease. They show strong evidence that the conserved element harboring this variant is tested for LacZ reporter activity in the developing mouse that is has activity in relevant tissues. They perform several assays to demonstrate a physical link between this enhancer region to a specific target gene, PKDCC, in both cranial neural crest cells and differentiated chondrocytes. Removal of a single copy of the enhancer has little effect on PKDCC expression in CNCCs but strong impacts in chondryocytes. H1 derived cells that are heterozygous at the variant above show strong bias in H3K27ac signals in chondrocytes. The researchers then go on to recharacterize a PKDCC knockout mouse to show that it has craniofacial defects. They use modern micro-CT and analysis techniques to demonstrate subtle changes in jaw and skull structure in PKDCC heterozygous mice and confirm many of the phenotypes that were described by Kinoshita et al 2009. Overall these results point to dosage of PKDCC in craniofacial development with changes in skull shape and susceptibility to orofacial clefting. However the epigenomic differences presented in Figure 2B that serve as the foundation for the rest of the work do not agree with previously published work by this group (Prescot et al 2015). The researchers claim "enrichment of the coactivator p300 and of the active chromatin mark H3K27ac at this region is higher in the chimpanzee CNCCs as compared to human, suggesting that this non-coding element may have higher regulatory activity in the chimp. However this region was not identified in the top 1000 biased enhancer regions provided in the supplement of the Prescott et al 2015 paper. The authors do not indicate any statistical significance and largely rely on signal tracks that have not been corrected for input controls to make this conclusion. The in vivo assay for enhancer activity while excellent at demonstrating where an enhancer can be active is not well suited to quantitative comparisons. Furthermore the researchers claim that the mouse orthologous sequence is not active in the assay despite strong H3K27ac and other enhancer related signals in developing mouse craniofacial tissues as available from the Mouse Encode Project. This calls into questions whether this assay is informative at all if the native sequence which shows functionally conserved activity is not active in the mouse embryo. Lastly the authors only consider this region as a potential enhancer and not any other type of regulatory sequence. GENCODE gene annotations demonstrate a potential lncRNA (LINC02898 /ENST00000378711.2) that is directly adjacent to the region marked by this variant. This could be a promoter for an RNA that regulates PKDCC in cis. Inspection of gene expression data from a recent preprint Yankee et al 2022 as well as Prescot et al data available from the recount3 database indeed indicate RNA signal from both CNCCs and primary human tissue consistent with this annotation. The Mundlos lab has demonstrated similar regulatory mechanisms through lncRNA Maenli at the En1 locus that result in limb abnormalities.

  4. Reviewer #3 (Public Review):

    Mohammed et al perform functional follow-up studies on the single nucleotide polymorphism rs6740960 located on chromosome 2p21 that was previously linked to lower jaw and chin shape variation and an increased risk of non-syndromic orofacial clefting. Through a combination of in silico multi-species alignment, in vitro enhancer marks, and finally in vivo data the team could confirm that the SNP is located in an active enhancer element driving transgene expression in the upper and lower jaw. The team tested the human and chimp orthologs in transgenic mice. Interestingly the mouse ought to look did not show any active enhancer activity in the LacZ reporter assay. Next, the authors could show a selective interaction of the enhancer element with the neighboring gene PKDCC in chondrocytes using H3K27ac HiChIP. Deletion of this enhancer in vitro led to an allele specific reduction of PKC expression. Finally, the authors aimed at evaluating the effect of rs6740960 in vivo using a mouse model. Since the enhancer sequence of the mouse did not show any positive reporter activity, the authors decided to use previously described Pkdcc full knockout mouse model (Kinoshita et al. 2009). Using sophisticated imaging technologies the authors were able to show that in mice several facial bones are Pkdcc dose sensitive.

    Overall this is an extremely exciting manuscript that addresses one of the key challenges in the post GWAS time: the functional connection of lead SNPs to their target genes and a detailed evaluation of the biological and morphological consequences.
    The manuscript is well written, and the conclusions are completely supported by the evidence provided. I really think this is a great paper, however I have several major concerns with the manuscript and its current format.

    Major comments:

    1: My main concern about the manuscript in its current format is the disconnection between the beautiful work of linking rs6740960 to Pkdcc in the first part of the manuscript and the investigation of dose sensitivity of Pkdcc itself in end of the manuscript. While I realized that this is because the enhancer itself is not conserved between humans and mice, in my opinion it still weakens the novelty of the finding of the second part of the manuscript quite significantly. The Pkdcc knockout has been well described and that the authors now present evidence that also heterozygous knockouts show a minimal phenotype in the facial bones is really not surprising. More importantly it doesn't show how the rs6740960 influences Pkdcc expression in vivo.

    A rather straightforward and very interesting experimental approach would be to replace the mouse enhancer sequence with the human or chimp enhancer carrying the risk allele or the wild type. In the last figure the authors have nicely shown that the entire experimental setup for the functional analysis of even minor changes to the facial bones caused by the SNP are available to the team. Even if the result was negative this experiment would significantly enhance the scientific impact of the paper.

    2: Another option would be to repeat the LacZ reporter essay with the human wild type and the risk allele in direct comparison. A beautiful example of such an experiment was recently shown by Yanchuset et al (A noncoding single-nucleotide polymorphism at 8q24 drives IDH1-mutant glioma formation, Yanchuset al.,Science378,68-78 2022)

    3: It is unclear how the H3K27ac HiChIP signal looks like at the Pkdcc locus in H9 ESC. What is the naïve interaction profile?