Research ArticleDEVELOPMENTAL BIOLOGY

PRDM15 loss of function links NOTCH and WNT/PCP signaling to patterning defects in holoprosencephaly

See allHide authors and affiliations

Science Advances  10 Jan 2020:
Vol. 6, no. 2, eaax9852
DOI: 10.1126/sciadv.aax9852

Abstract

Holoprosencephaly (HPE) is a congenital forebrain defect often associated with embryonic lethality and lifelong disabilities. Currently, therapeutic and diagnostic options are limited by lack of knowledge of potential disease-causing mutations. We have identified a new mutation in the PRDM15 gene (C844Y) associated with a syndromic form of HPE in multiple families. We demonstrate that C844Y is a loss-of-function mutation impairing PRDM15 transcriptional activity. Genetic deletion of murine Prdm15 causes anterior/posterior (A/P) patterning defects and recapitulates the brain malformations observed in patients. Mechanistically, PRDM15 regulates the transcription of key effectors of the NOTCH and WNT/PCP pathways to preserve early midline structures in the developing embryo. Analysis of a large cohort of patients with HPE revealed potentially damaging mutations in several regulators of both pathways. Our findings uncover an unexpected link between NOTCH and WNT/PCP signaling and A/P patterning and set the stage for the identification of new HPE candidate genes.

INTRODUCTION

Congenital defects are a leading cause of morbidity worldwide, accounting for the deaths of 330,000 newborns every year. Brain malformations, including microcephaly and holoprosencephaly (HPE), are the most common congenital anomalies and place a heavy burden on the affected individuals and the health care system (13). HPE is a structural anomaly of the developing forebrain affecting 1:250 embryos and 1:16,000 live-born infants. Clinically, HPE encompasses a continuum of brain malformations and is accompanied with a spectrum of craniofacial defects in 80% of the cases; microcephaly and eye defects are among the most common features in affected individuals (4). In the majority of cases, the underlying cause remains uncertain due to the high complexity and the multigenic origin of these anomalies (5, 6). Lately, it has become clear that HPE is caused by a malfunction in key signaling pathways in the early embryo, leading to developmental defects in the organizing centers and midline structures (7). The defects involve a sequence of developmental steps that begin with Nodal signaling to establish the midline progenitors in the developing primitive streak (PS). It then continues with the proper positioning of the forming prechordal plate beneath the neuroectoderm and activation of midline Hedgehog signals to maintain the anterior identity of the forebrain. However, the restriction of HPE genetic determinants to a handful of NODAL and Sonic hedgehog (SHH) pathway regulators stems from our limited understanding of the molecular events governing specification of early and late midline structures. Expansion of this genetic repertoire has become a necessity to develop therapeutic options and improve molecular diagnosis of HPE.

Genes encoding transcription factors (TFs) and epigenetic regulators are relevant etiological candidates given their central role in integrating signaling cascades and orchestrating multiple biological processes. Deficiency in their function can disturb entire transcriptional programs, involving numerous genes and molecular pathways, leading to a complex pathological outcome. Consistent with this hypothesis, we have recently identified a loss-of-function (LOF) mutation in the transcriptional regulator PRDM15 in patients with a syndromic form of HPE. Here, we combine mouse genetics and epigenomic approaches to uncover the role of this TF in congenital brain malformations. Our findings establish PRDM15 as a key regulator of NOTCH and WNT/PCP pathways in the developing embryo, implicating them in regulation of anterior/posterior (A/P) patterning and forebrain development. In addition, we uncover new genetic variants in key components of these signaling pathways in patients with HPE. Collectively, our findings refine the molecular mechanisms governing forebrain development and set the stage for the identification of new HPE candidate genes.

RESULTS

The C844Y zinc finger mutation impairs PRDM15 transcriptional activity

Homozygosity mapping and whole-exome sequencing on patients with steroid resistant nephrotic syndrome (SRNS) identified three recessive mutations in PRDM15 (NM_001040424.2). These mutations are located in the sequences coding for the PR domain (c.461T>A; p.Met154Lys-M154K and c.568G>A; p.Glu190Lys-E190K) and the 15th zinc finger (c.2531G>A; p.Cys844Tyr-C844Y), respectively (Fig. 1A). Of particular interest, in four consanguineous families that have the variant encoding PRDM15 C844Y, the affected probands exhibited a syndromic form of SRNS consistent with the Galloway-Mowat syndrome (8). Besides renal defects, the patients displayed facial (narrow forehead, microcephaly, abnormal cerebral gyration, and ophthalmic abnormalities) and extracranial defects (heart malformations and postaxial polydactyly) (9).

Fig. 1 PRDM15 mutations in patients with SRNS and microcephaly/HPE.

(A) Schematic representation of the PRDM15 mutation positions and the affected domains. (B) Alkaline phosphatase (AP) staining of ESCs; the respective genotypes are indicated in the lower panel. Data are average of four independent cell cultures (n = 4) ± SD. Statistical tests were applied on differences observed in the percentage of completely undifferentiated colonies. Student’s t test (two sided) was used to determine significance. (C) Heat map of differentially expressed genes in ESCs upon the indicated genetic manipulations. (D) mRNA levels of Rspo1 in ESCs; the respective genotypes are indicated by color code. Expression levels were normalized to Ubiquitin (Ubb), and Prdm15fl/fl (empty vector) was used as reference. Data shown are from three independent experiments (n = 3). (E) Enrichment of PRDM15 binding on promoter regions of the target gene (Rspo1) in ESCs—respective genotypes are indicated by color code—as measured by ChIP-qPCR. Depicted is the average enrichment [data from three independent cell cultures (n = 3)] over percent of input. In (B) to (E), the endogenous mouse Prdm15 has been deleted by the addition of OHT (50 nM) after ectopic expression of WT or mutant human PRDM15 (hPR15). In (D) and (E), center values, mean; error bars, SD. Student’s t test (two sided) was used to determine significance.

We have recently demonstrated that PRDM15 regulates the transcription of Rpso1 and Spry1, two key components of the MAPK (mitogen-activated protein kinase)/ERK (extracellular signal–regulated kinase) and WNT pathways, to maintain naïve pluripotency of mouse embryonic stem cells (mESCs) (10). To evaluate the effects of these mutations on PRDM15 function, we ectopically expressed the three identified human variants in Prdm15-deficient embryonic stem cells (ESCs) (Prdm15/). Only hPR15-C844Y, which is associated with brain defects in humans, failed to restore ESC self-renewal (Fig. 1B), and most importantly, the global changes in gene expression, induced by loss of endogenous PRDM15 (Fig. 1C and table S1 (A to E)]. These data strongly suggest that hPR15-C844Y is a LOF mutation. While hPR15-M154K and hPR15-E190K rescued Rspo1 expression at levels comparable to the wild-type (WT) human PRDM15 (hPR15-WT), hPR15-C844Y failed to restore its transcript levels [quantitative polymerase chain reaction (qPCR)] and to activate its transcription in a luciferase reporter assay (Fig. 1D and fig. S1A).

To gain further insights into the impact of these mutations on PRDM15 function, we tested the stability of the encoded proteins and their cellular localization. Immunofluorescence staining, in a Prdm15/ background, showed that none of the mutations affected the nuclear localization of PRDM15 (fig. S1B). On the other hand, all three mutants encoded less stable proteins (fig. S1C). We have previously shown that the zinc finger domains are required for DNA binding and transcriptional activity of PRDM15 (10). Thus, we sought to test the ability of the various mutants to bind to chromatin. Consistent with an LOF of the zinc finger mutant, chromatin immunoprecipitation (ChIP)–qPCR analysis revealed reduced enrichment of hPR15-C844Y at the promoter region of Rspo1 (Fig. 1E), a result compatible with its inability to promote its transcription (Fig. 1D and fig. S1A).

Genetic deletion of Prdm15 leads to brain malformations and midgestation lethality in mice

To gain molecular insights on the effects of PRDM15 LOF during mammalian development, we intercrossed Prdm15lacZ/+ heterozygous mice, which are healthy and fertile. A description of all the Prdm15 alleles and deleter strains used in this study is summarized in fig. S2A. Consistent with a fundamental role of PRDM15 during embryonic development, we obtained no homozygous mutant [Prdm15lazZ/lacZ knockout (KO)] pups (Fig. 2A), while of the hundreds Prdm15lacZ/+ embryos that were dissected at various stages of development, none showed any defects. Timed matings revealed the embryonic lethality of Prdm15lazZ/lacZ (KO) embryos occurs between embryonic days 12.5 (E12.5) and E14.5 (Fig. 2A). Notably, at E12.5, KO embryos were smaller and showed a spectrum of brain malformations affecting predominantly the anteriormost structures of the head, including the eyes (Fig. 2B), consistent with the brain and facial features observed in patients with the C844Y mutation. Coronal sections of the brain at this stage confirmed that the lateral and medial ganglionic eminences were underdeveloped. Furthermore, we noted an abnormal separation of the cerebral hemispheres, reminiscent of HPE (Fig. 2C). Classic HPE encompasses a continuum of brain anomalies caused by neural tube patterning defects that affect the anteriormost structures and is often accompanied by craniofacial defects involving the eyes (4, 11, 12).

Fig. 2 Genetic deletion of Prdm15 leads to brain malformations and midgestation lethality in mice.

(A) Genetic distribution of embryos from Prdm15+/LacZ intercrosses, indicating lethality between E12.5 and E14.5. (B) Phenotypic continuum of brain defects in E12.5 Prdm15lacZ/lacZ KO embryos. (C) Hematoxylin and eosin (H&E) staining of serial coronal sections of E12.5 brains from Prdm15+/+ WT (upper panel) and Prdm15lacZ/lacZ KO (lower panel) embryos. The mutants lack the complex organization of the anterior forebrain, including the lateral (LGE) and medial ganglionic eminences (MGE), the epithalamic and dorsal thalamic neuropeithelium (NE), and eyes. (D) Nestin-Cre–meditated deletion of Prdm15 in neuronal precursors does not affect brain development. Representative images are shown in (B) to (D). LGE/MGE, lateral and medial ganglionic eminences; NE, neuropeithelium; NCX, neocortex; E, eye; LV, lateral ventricle; V, ventricle; TOT, total. (B and D) Photo credit: Messerschmidt and Mzoughi.

These results prompted us to delete Prdm15 specifically in the developing brain by crossing Prdm15fl/fl mice to the Nestin-Cre deleter strain. This Cre recombinase is active at ~E11 in neural stem cells/progenitors and would reveal whether PRDM15 is essential for the process of neurogenesis. The resulting Prdm15/::Nestin-CRE embryos did not show any apparent defects at E12.5 (Fig. 2D), were born at the expected Mendelian ratios, and developed into healthy adults (fig. S2B). This suggests that PRDM15 is required at earlier time points of forebrain specification.

PRDM15 controls the AME specification and A/P patterning of the neural plate

Defects in Prdm15 KO embryos are apparent before the onset of neurulation, as mutants were markedly smaller and had an abnormal morphogenesis by E7.5 (fig. S3A). Between E6.5 and E7.5, two signaling centers act sequentially to pattern the forebrain in the mouse embryo (Fig. 3A) [reviewed in (1315)]. The first resides within the extraembryonic lineages and is called the anterior visceral endoderm (AVE). The AVE imparts anterior identity to the underlying epiblast, thereby restricting the site of gastrulation—the PS—to the posterior epiblast. During gastrulation, a second specialized population of cells, known as the AME, emerges from the anterior PS (APS). These cells migrate anteriorly, giving rise to the anterior definitive endoderm and prechordal plate mesoderm. Their role is to produce secondary inductive cues that reinforce anterior identity in the overlying neural plate (Fig. 3A).

Fig. 3 PRDM15 controls the AME specification and A/P patterning of the neural plate.

(A) Schematic of the signaling centers governing A/P patterning in the mouse embryo. (B) At E6.5, Foxa2 is expressed in the AVE (red line) and APS (red asterisk). At E7.5, Lhx1 transcripts label the visceral endoderm (VE) overlying the epiblast including the AVE as well nascent mesoderm and midline axial mesendoderm. In Prdm15 mutants (mut), Foxa2 expression is confined to the distal VE, with little enrichment in the prospective AVE. Lhx1 is detected in the VE and mesoderm of the middle Prdm15 mutant, but only in the VE of the one on the right. (C) Expression of T, Lefty2, Foxa2, Chordin, and Shh in WT and Prdm15lacZ/lacZ embryos at E7.5. In Prdm15 mutants, T is expressed normally in the PS; Lefty2 transcripts are down-regulated in nascent mesoderm; Foxa2 and Chordin expression remains high distally in the region of the APS (angled black-dashed line) but does not extend anteriorly in the midline axial mesendoderm (am); and Shh expression is similarly weak in the anterior midline (asterisk). n, node. (D) Expression of Six3/Shh or Otx2/Shh in WT (upper) and Prdm15lacZ/lacZ KO (lower) embryos at E8.5. Six3 and Otx2 expression highlights the reduction in anterior forebrain (fb) development (angled black dashed lines) in Prdm15lacZ/lacZ KO mutants. no, notochord; mb, midbrain; DVE, Distal Visceral Endoderm. Representative images are shown in (B) to (D). (C and D) Photo credit: Dun and Ong.

We reasoned that loss of PRDM15 might impair forebrain specification during the earliest events of anterior patterning and therefore examined the expression of a panel of marker genes diagnostic for defects in either the AVE or AME in Prdm15 KO embryos. Foxa2 is a marker of both, AVE and APS, in early PS stage embryos at E6.5. In Prdm15 KO embryos, in situ labeling shows expression in the distal visceral endoderm overlying the epiblast in a pattern typically observed 1 day earlier in WT embryos (Fig. 3B) (16). We conclude that Prdm15 KO embryos are developmentally delayed even before gastrulation. At E7.5, Lhx1 is expressed in the nascent mesoderm and anterior midline mesendoderm. In the smaller, delayed Prdm15 KO littermate embryos, Lhx1 is expressed normally throughout the visceral endoderm, including the AVE, as well as in the nascent mesoderm (Fig. 3B) (17, 18). Both FOXA2 and LHX1 are required for the formation and function of the AVE, and their activation provides evidence that the initial specification of the primary anterior-posterior axis by the AVE is normal in Prdm15 KO embryos.

We next examined the expression of PS (T and Lefty2) and AME (Foxa2, Chordin, and Shh) marker genes. By E7.5, Prdm15 KO embryos are easily recognizable due to a characteristic ruffling in the extraembryonic visceral endoderm, with a fully extended PS that expresses both T and Lefty2 (Fig. 3C). At this stage, Foxa2 is expressed in the node, which marks the anterior end of the PS, and the AME that extends rostrally in WT embryos. In contrast, in Prdm15 KO embryos, Foxa2 transcripts are present distally but do not extend anteriorly (Fig. 3C). A similar pattern is observed with Chordin, which also labels the node and AME in WT embryos but is confined to the APS in Prdm15 KOs (Fig. 3C). Shh expression is also diagnostic for the node and AME, but in KO embryos, only a few Shh-positive cells are observed along the anterior midline (Fig. 3C). Together, these results show that loss of PRDM15 specifically affects the production of the anterior AME. Consequently, the crucial refining signals produced by these cells that orchestrate the continued patterning and morphogenesis of the anterior neuroectoderm are lost, resulting in anterior truncations that are evident by diminished forebrain expression of Six3 and Otx2 in Prdm15 KO mutant embryos at E8.5 (Fig. 3D). To further corroborate these findings, we deleted Prdm15 specifically in the epiblast, using the Sox2-Cre transgene (fig. S3B) (19), while maintaining WT extraembryonic tissues. Consistent with an essential role for PRDM15 in the PS-derived AME and not AVE specification, Prdm15/::Sox2-CRE embryos died in utero starting at E12.5 (fig. S3C) and exhibited a spectrum of brain defects similar to those observed in Prdm15 KO embryos (fig. S3D).

PRDM15 orchestrates transcriptional programs governing A/P patterning and brain development in the mouse embryo

To examine the impact of PRDM15 depletion on early embryonic processes, namely, A/P patterning, we sequenced the transcriptome of WT versus Prdm15 KO E6.5 embryos. We reasoned this could be the most critical stage for AME specification as AME cells emerge less than 24 hours later. Unbiased clustering of global gene expression separated WT versus Prdm15 KO embryos into distinct groups, indicating marked transcriptional differences (Fig. 4A and table S1F). Gene ontology (GO) analysis of the significantly down-regulated genes identified “Pattern specification process,” “Head development,” and “Neural tube development” among the enriched terms. Among these genes, several are important regulators of forebrain development and A/P patterning (Fig. 4B and fig. S4A, and table S1, G to H). We noted a striking reduction in the expression of key components of three signaling pathways: WNT, NOTCH, and SHH (Fig. 4C, fig. S4B, and table S1, I and J).

Fig. 4 PRDM15 orchestrates transcriptional programs governing A/P patterning and brain development in the mouse embryo.

(A) Unbiased clustering heat map of the entire transcriptome in WT (n = 8) versus Prdm15lacz/lacz KO (n = 10) E6.5 embryos, analyzed by RNA sequencing. Heat maps of differentially expressed genes from the indicated GO categories (B) and KEGG pathway (C) identified as top hits in the RNA sequencing. Light and dark blue rectangles on the right side indicate genes whose promoter region is directly bound by PRDM15 in ESCs only or both in ESCs and E6.5 embryos, respectively. (D) Snapshots of representative PRDM15 ChIP tracks (UCSC genome browser). Examples of conserved target genes (binding sites) between E6.5 embryos (blue) and ESCs (orange) are shown.

We have recently shown that PRDM15 recognizes a defined DNA motif present at promoters or enhancers of target genes (10). To define the set of direct PRDM15 transcriptional targets, we performed ChIP sequencing (ChIP-seq) on mESCs and WT E6.5 embryos (table S1, K and L). Despite the limited biological material available from the pre-gastrula embryos, we identified 58 high-confidence promoter-bound targets, the majority of which (~84%) were also bound by PRDM15 in ESCs (Fig. 4D, fig. S4C, and table S1M). In addition, identification of the same PRDM15 consensus binding motif in both systems implies a conservation of its targets. We therefore chose to consider PRDM15-bound promoters identified in ESCs as relevant for our follow-up analyses. Among these, a handful of PRDM15 targets, including Rbpj, Notch3, Maml3 (NOTCH), Vangl2, Wnt5b, Gpc6, Nphp4 (noncanonical WNT), and Gas1 (SHH), were of particular interest as they are significantly down-regulated in the mutant embryos (fig. S4D). Collectively, these data indicate that lack of PRDM15 leads to transcriptional down-regulation of key regulators of developmentally important signaling pathways (NOTCH, noncanonical WNT, and SHH).

PRDM15 is a master transcriptional regulator of the NOTCH and WNT/PCP pathways, and its targets are mutated in a large cohort of patients with HPE

These results prompted us to perform a targeted analysis of the down-regulated PRDM15 target genes in a large cohort of patients with HPE (132 trios and 188 singletons). We found heterozygous variants in 99 genes, ~17% of them were likely to be damaging (table S2A). To gain insights on potential functional interactions between these genes, we generated functional protein association networks using STRING. Although the majority of the proteins did not seem to be functionally related, two main networks representing NOTCH and WNT/PCP signaling formed (Fig. 5A and table S2B), supporting their potential involvement in HPE pathobiology.

Fig. 5 PRDM15 is a master transcriptional regulator of the NOTCH and WNT/PCP pathways, and its targets are mutated in a large cohort of patients with HPE.

(A) Functional groups identified by protein association network analysis of PRDM15 target genes mutated in patients with HPE using STRING. (B) mRNA levels of the indicated genes in ESCs; the respective genotypes are indicated in the lower panel. Expression levels were normalized to Ubiquitin (Ubb), and Prdm15fl/fl (empty vector) was used as reference. Rspo1 expression levels were used as positive control in Fig. 1D. Data shown are from three independent experiments (n = 3). (C) Enrichment of PRDM15 binding on promoter regions of the indicated target genes in ESCs—respective genotypes are indicated in the lower panel—as measured by ChIP-qPCR. ChIP on the Rspo1 promoter was used as a positive control for PRDM15 binding. Depicted is the average enrichment [data from three independent cell cultures (n = 3)] over percent of input. In (B) and (C), the endogenous mouse Prdm15 has been deleted by the addition of OHT (50 nM) after ectopic expression of WT or mutant human PRDM15. In (B) and (C), center values, mean; error bars, SD. Student’s t test (two sided) was used to determine significance.

To assess the ability of the PRDM15 mutants to regulate the expression of critical components of both pathways, we took two approaches. First, we performed rescue experiments in Prdm15/ ESCs by reintroducing WT or mutant PRDM15 expression constructs. While hPR15-M154K and hPR15-E190K restored the expression of target genes at levels comparable to the WT human PRDM15 (hPR15-WT), none were significantly rescued by hPR15-C844Y (Fig. 5B and fig. S5A). In addition, ChIP-qPCR analysis confirmed a reduced enrichment of hPR15-C844Y at the promoter regions of these target genes (Fig. 5C and fig. S5B), which is consistent with the failure to promote their transcription (Fig. 5B). Second, to confirm that the C844Y mutation in humans is indeed an LOF mutation, we introduced the corresponding homozygous mutation (C842Y) in mESCs using CRISPR-Cas9 technology (fig. S5, C to E). Although the C842Y knock-in allele did not affect Prdm15 transcript levels, the resulting protein was unstable and less abundant (fig. S6, A and B). qPCR confirmed that Prdm15C842Y cells express PRDM15 target genes (i.e., Rbpj, Notch3, Vangl2, etc.) at lower levels compared with WT (fig. S6C) and that endogenous PRDM15C842Y protein is unable to bind (ChIP-qPCR) to its target promoters (fig. S6D).

Our findings call for a future functional characterization of the NOTCH and PCP gene variants and should motivate targeted genetic mapping for new HPE candidates in regulators of both pathways.

DISCUSSION

LOF mutation in RPDM15 is associated with a syndromic form of HPE

We have identified new mutations in the PRDM15 gene in patients with SNRS. Although the mutations affecting the PR domain of the protein (M154K and E190K) are associated with isolated SRNS cases only, the zinc finger mutation (C844Y) causes a syndromic form of HPE. In our in vitro ESC system, these PR domain mutations reduced the stability of the encoded protein but rescued considerably the phenotypic and molecular changes induced by loss of the endogenous protein. This is consistent with the fact that these mutations in humans cause isolated SRNS only and could imply a context-dependent requirement for the PR domain. Alternatively, the differential impact of the PR versus ZNF mutations on protein stability may support a threshold model, where different levels of PRDM15 expression are required for the development of specific organ systems. On the other hand, the ZNF mutation (C844Y) had marked effects on PRDM15 function in both settings, which we attribute here to impaired binding of the mutant protein to regulatory regions of its transcriptional targets.

Genetic deletion of Prdm15 in mice recapitulates the brain malformations reported in patients with HPE

Similar to the LOF mutation in humans, genetic deletion of Prdm15 in mice leads to a broad spectrum of brain defects, affecting predominantly the anteriormost structures including the eyes. Such phenotypic continua are commonly assigned to allelism, polygenic origin, and the action of modifier genes. Yet, here we report that perturbation of a single transcriptional regulator can indeed affect an entire transcriptional network, relevant to both normal development and pathological manifestations.

PRDM15 regulates the transcription of NOTCH and WNT/PCP to orchestrate forebrain development

Our findings show that PRDM15 promotes transcription of several regulators of the NOTCH and WNT/PCP pathways to orchestrate formation of midline structures. Perturbation of these transcriptional programs, upon PRDM15 depletion, disrupts forebrain development due to impaired AME specification and lack of SHH signaling, consistent with the sequence of developmental defects associated with HPE pathobiology (7).

Of note are the prominent phenotypic similarities between Prdm15 null embryos and genetic (or microsurgical) modulation of the Nodal signaling pathway in mouse. That is, Nodal hypomorphic alleles, assorted combinations of mutations in Smad2 and Smad3, as well as the mutations in the downstream effectors Foxh1 and Foxa2, all result in embryos with defective AME production and compromised anterior forebrain development (2023).

On the other hand, the characteristic ruffling of the visceral endoderm observed in Prdm15 KO embryos at E7.5 has been observed in other mutants where extraembryonic mesoderm (ExMeso) production during gastrulation is impaired, such as in loss of Smad1 (24), combined loss of Smad2 and Smad3 in the epiblast (21), or Otx2 (chimeric analysis) (25). It is, however, important to emphasize that epiblast-specific deletion of Prdm15 (Prdm15/::Sox2-CRE embryos) equally results in smaller embryos with defects in the formation of anterior structures (fig. S3). It is additionally possible that the developmental delay we observed in Prdm15 KO embryos disproportionally affects some parts of the gastrulating embryo, rather than an overall delay in epiblast proliferation before gastrulation.

On the basis of our molecular analysis, we conclude that like modulation of the Nodal signaling pathway, loss of Prdm15 specifically affects AME specification. Given the requirement of this critical signaling center in providing reinforcing anterior patterning signals, we favor a model in which its lack or dysfunction underlies the Prdm15 phenotype, rather than a paucity of mes(endo)derm produced during gastrulation by a mutant embryo experiencing developmental delay.

The restriction of HPE genetic determinants to a handful of NODAL and SHH pathway regulators stems from our limited understanding of the molecular events governing specification of early and late midline structures. Recent studies have implicated components of the WNT/PCP pathway in regulating polarity of the node along the A/P axis and linked their deregulation to structural anomalies of this critical organizing center (2629). Thus, it is not unexpected that perturbation of the WNT/PCP pathway affects the specification of APS derivatives, namely, the AME and node (29). In addition, while the links between mutations in PCP signaling and neural tube defects are well established (6, 3032), their involvement in HPE remains uncharted. NOTCH signaling, on the other hand, has been implicated in HPE only recently (33). Besides its established neurogenic role in the developing mouse telencephalon, growing evidence supports the involvement of key NOTCH regulators (for example Dll1 and Rbpj) in node morphogenesis and midline truncations (34, 35).

Identification of new potential determinants of HPE in the NOTCH and WNT/PCP pathways

Our findings prompted us to perform a targeted search for mutations in a large cohort of patients with HPE. Our analysis of exome sequencing data from 132 trios and 188 singletons revealed multiple rare heterozygous variants in PRDM15 transcriptional targets involved in forebrain development. In silico protein association network analysis of these variants identified two major functional groups regulating the NOTCH and WNT/PCP pathways. We expect that our findings will encourage validation of the reported variants/mutations as well as further mining for additional HPE candidates in both pathways.

METHODS

Mouse strains

PRDM15 KO-first mice that harbor the Prdm15lacZ allele were obtained from the European Conditional Mouse Mutagenesis Program. Hemizygous (Prdm15lacZ/+) animal intercrossings were performed to obtain homozygous (Prdm15lacZ/lacZ) embryos. Further details on these animals and the conditional Prdm15fl/fl strain can be found in (10). To generate epiblast-specific Prdm15/ embryos, Prdm15fl/fl mice were first crossed to heterozygous Sox2-CRE transgenic animals [B6.Cg-Edil3Tg(Sox2-cre)1Amc/J; JAX Laboratory] (36). The resulting males (Prdm15/+::Sox2-CRE) were then crossed again to Prdm15fl/fl females. In this generation, a quarter of the progeny is expected to be Prdm15/::Sox2-CRE. The Sox2-CRE transgene was always propagated through male animals. A similar breeding strategy, using Nestin-CRE [B6.Cg-Tg(Nes-cre)1Kln/J; JAX Laboratory] transgenics, was followed to generate Prdm15/:: Nestin-CRE mice. All mice-related procedures were approved by the local Institutional Animal Care and Use Committee (IACUC) and performed in compliance with the respective guidelines (IACUC nos. 151042 and 18/10EGDM/90).

Hematoxylin and eosin slide preparation

E12.5 embryos were fixed in 4% PFA (paraformaldehyde) for 48 hours before being mounted in OCT (Optimal Cutting Temperature) embedding compounds. Then, serial coronal sections of the brains (anterior-posterior) were made using a cryostat and immediately thaw mounted on poly-l-lysine–coated histological slides for hematoxylin and eosin staining.

Isolation and culture of ESCs

Prdm15fl/fl; ROSA26-CreERT2 ESCs have been described in (10). For all experiments, ESCs were cultured in the conventional [serum + Lif (Leukemia Inhibitory Factor) (SL)] medium unless otherwise stated. OHT (4-Hydroxytamoxifen) (50 nM; SIGMA-H7904) was added to the culture medium overnight (O/N) to generate Prdm15/ cells.

Whole-mount in situ hybridization assays

Embryos were isolated between E6.5 and 8.5, genotyped, then processed for whole-mount in situ hybridization as described in (37) with the following probes: Foxa2, Lhx1, T, Lefty2, Chordin, Shh, Otx2, and Six3.

Site-directed mutagenesis

Full-length human PRDM15 cDNA (NM_001040424.2) was subcloned into the PiggyBac vector (DNA2.0, PJ549). Clones encoding the various PRDM15 mutations were generated using the QuickChange II XL Site-directed Mutagenesis Kit (Agilent Technologies). The sequence of primers used can be found in table S3.

Knock-in of c.2525G>A p.Cys842Tyr in mESCs

To introduce the hC844Y/mC842Y point mutation, mESCs were transfected with PX458 [pSpCas9 (BB)-2A-GFP] vector expressing a guide RNA targeting the site to be mutated, along with a single-stranded oligonucleotide containing the target point mutation, to serve as a DNA repair template. Additional eight silent mutations have been introduced to avoid editing of the template by the CAS9 protein. Single clones were sorted and expanded in 2i medium. Genomic DNA was used for screening by digestion with XMN I restriction enzyme. DNA from potential mutants was cloned into the pCR 4-TOPO TA vector following the manufacturer’s instructions, and 5 to 10 colonies were sequenced. Details of the strategy and the sequence of the oligonucleotides used are described in fig. S5 and table S3.

Cycloheximide chase assay and Western blotting

To assess protein stability, Prdm15/ ESCs expressing either wild or mutant PRDM15 were treated with cycloheximide (CHX; 150 μg/ml) (Sigma, no. C-7698), and then collected at different time points (2, 4, and 6 hours) for protein extraction and analysis by Western blotting. Samples collected immediately before treatment with CHX (t = 0) served as reference. Antibodies and dilutions used were PRDM15 (in house, 1:3500) and TUBA (Alpha-TUBULIN) (Sigma T5168, 1:10,000).

Alkaline phosphatase staining

To assess ESC self-renewal/differentiation, cells were stained with alkaline phosphatase staining solution (AP detection kit, Millipore, SCR 004). In brief, 500 cells per well (12-well plates) were seeded in triplicates and cultured for 5 days with daily change of medium before being stained as per the supplier’s recommendations.

Immunofluorescence staining

ESCs were seeded on gelatin-coated eight-well glass slides (Millipore, PEZGS0816), at 3 × 103 per well, and cultured in 2i medium. Three days later, cells were fixed in 4% PFA at room temperature, permeabilized with 0.5% Triton X-100, and then blocked using 2% bovine serum albumin (BSA) for 1 hour at room temperature before O/N staining with anti- PRDM15 (in house, 1:100) at 4°C. The next day, slides were washed with phosphate-buffered saline (PBS) (three times) and stained with Alexa Fluor–conjugated secondary rabbit antibody at 37°C (30 min). Last, slides were washed with PBS (three times) before they were mounted with a DAPI (4′,6-diamidino-2-phenylindole)–containing mounting medium (VECTASHIELD, Vector Laboratory H1200).

Quantitative real-time PCR

Total RNA from cells was isolated using PureLink RNA Mini Kit (Ambion, 1283-018A) according to the manufacturer’s instructions. RNA was retrotranscribed into cDNA using Maxima First Strand cDNA Synthesis Kit (Thermo Scientific, K1642) and subjected to quantitative real-time PCR (qRT-PCR) on an ABI PRISM 7500 machine. qPCRs (20 μl) contained 10 μl of SYBR Green PCR supermix (2×), 4 μl of a forward and reverse primer mix (final concentration, 200 nM), and 6 μl of cDNA (20 ng). Primers sequences are listed in table S4.

Chromatin immunoprecipitation

The detailed procedure for ChIP experiments has been described previously (38); all steps were performed at 4°C and protease inhibitor was added, unless stated otherwise. In brief, 20 to 40 million ESCs were fixed in 1% formaldehyde for 10 min at room temperature before quenching with 0.125 M glycine (5 min at room temperature). Cells were then washed in PBS and harvested in lysis buffer before freezing at −20°C O/N. The following day, cells were pelleted by centrifugation, resuspended in ice-cold ChIP buffer, and sonicated for six cycles (30-s ON/30-s OFF) using a BRANSON Digital Sonifier (no. S540D). Lysates were then precleared for 2 hours in Sepharose A beads (blocked in 5 mg/ml BSA) before O/N incubation with PRDM15 antibody (4°C). The next day, Protein A beads were added for 4 hours before washing then de-cross-linking in 1% SDS and 0.1 M NaHCO3 (65°C, O/N). Last, DNA was eluted in T-buffer (pH 8) using QIAquick PCR Purification Kit, QIAGEN. Sequences of primers used in ChIP-qPCR are listed in table S4. For the E6.5 ChIP, approximately 40 to 50 embryos per experiment were pooled together and fixed immediately after isolation.

ChIP-seq and bioinformatics analysis

TruSeq ChIP Sample Prep Kit (IP-202-1012) was used for DNA library preparation. Sequencing was performed in the Illumina HiSeq 2000 and NextSeq 500 at the Genome Institute Singapore. Details of the bioinformatics analysis can be found in (10). In brief, the sequenced reads were aligned to the mm9 genome assembly using bowtie version 2. Peak calling was done using MACS 2.1.1 (https://github.com/taoliu/MACS). Peaks were then annotated using the ChIPpeakAnno package in R—promoters were defined to be 5 kb upstream and 1 kb downstream of the transcription start site. Motif discovery was done using MEME-ChIP in the MEME Suite (http://meme-suite.org).

RNA sequencing and bioinformatics analysis

For E6.5 embryo transcriptome analysis, RNA was extracted from 8 WT and 10 Prdm15lacZ/lacZ littermates. RNA from ESCs was collected 3 days after ethanol/OHT treatment. Library preparation was performed following the TruSeq RNA Sample preparation v2 guide (Illumina). The sequenced reads were mapped to mm9 build of the mouse genome using STAR version 2.4.2a. Differential expression analysis was performed using the DESeq2 package in R. Only genes with an average FPKM (Fragment Per Kilobase Million) >1 are considered expressed. Enriched GO terms and KEGG pathway were identified using Metascape. Genes used for GO analysis were filtered based on statistical significance (P < 0.05) and fold change (log2 fold change of ±0.322) in E6.5 embryo RNA sequencing. Heatmaps of gene expressions (FPKM) were generated with in-house scripts with R.

HPE database mining and protein association network analysis

To identify potential new candidate genes associated with HPE, we searched for genetic variants in genes/proteins acting downstream of PRDM15. Exome sequencing data from a cohort of 320 patients with HPE (132 trios and 188 singletons) were evaluated. Filter criteria are as follows: allele frequency <0.0001 in ExAC database (39) de novo (if trio available); synonymous changes were omitted; and benign changes by ACMG 2015 (40) criteria were removed. To identify protein networks and functional groups, genes with potential HPE variants were subjected to protein association network analysis using STRING database (https://string-db.org).

Statistical analysis

All experiments were repeated at least three times with similar results. Each biological repeat was done in at least two to four technical replicates/independent cell cultures, where applicable. Normal distribution was assumed for all statistical analyses. Unpaired Student’s t test (two sided) was applied using GraphPad Prism (version 7.0) to determine the statistical significance of the observed differences. Changes were considered statistically significant when P < 0.05.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/6/2/eaax9852/DC1

Fig. S1. Functional characterization of PRDM15 mutations associated with SRNS and microcephaly/HPE.

Fig. S2. Constitutive and conditional Prdm15 KO alleles.

Fig. S3. Sox2-Cre epiblast-specific deletion of Prdm15 recapitulates the Prdm15Lacz phenotype.

Fig. S4. Identification of PRDM15 transcriptional targets in the pre-gastrula mouse embryo.

Fig. S5. CRISPR-Cas9–mediated knock-in of the mouse C842 mutation in ESCs.

Fig. S6. mC842Y/hC844Y substitution is an LOF mutation.

Table S1. ChIP-sequencing and RNA-sequencing analysis.

Table S2. Genetic variants of PRDM15 target genes associated with HPE.

Table S3. Oligonucleotides used for site-directed mutagenesis and mutation knock-in.

Table S4. qRT-PCR primers and ChIP-qPCR primers.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We thank the BRC shared facilities for technical support. We are grateful to the GIS Genome Sequencing Team for help with the Solexa high-throughput sequencing and to the entire EG laboratory for the critical discussions. Funding: This work was supported by AGA-SINGA (SINgapore Graduate Award) fellowship to S.M. E.G. is supported by NMRC/OFIRG/0032/2017 and NRF-CRP17-2017-06 grants. Author contributions: S.M., N.R.D., D.M.M., and E.G. designed the study. S.M. performed most of the experiments with support from F.D.T., C.-M.M., H.W., and C.M.W. D.H.P.L. performed computational analysis. N.R.D. and S.L.M.O. performed whole-mount in situ hybridization. P.K., M.M., and F.H. identified human PRDM15 mutations. N.R.D., D.M.M., and E.G. provided general supervision. S.M. and E.G. prepared the manuscript with help from all coauthors. Competing interests: E.G. has received research funding from Eli-Lilly and Prelude Therapeutics, has served as a consultant for Prelude Therapeutics and SKBP, and has served on the advisory board for LION TCR and Janssen. E.G. is a cofounder of ImmuNOA Pte. Ltd. The authors declare that they have no other competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. All transcriptome (RNA sequencing) and cistrome (ChiP-seq) data supporting the findings of this study have been deposited in the NCBI Gene Expression Omnibus under accession GSE125752. Additional data related to this paper may be requested from the authors. MEME software, http://meme-suite.org/doc/meme-chip.html; binding motif prediction tool for ZNF proteins, http://zf.princeton.edu. Sequencing datasets were deposited into the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) under accession GSE125752: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE125752.
View Abstract

Stay Connected to Science Advances

Navigate This Article