A cardiovascular, craniofacial, and neurodevelopmental disorder caused by loss-of-function variants in the eIF3 complex component genes EIF3A and EIF3B

Summary

Syndromic cardiac malformations can result in morbidity, yet their genetic etiology is only understood for a subset of individuals. Genome sequencing efforts in congenital anomaly cohorts may identify disease-associated variants in previously unrecognized genes. Through international matchmaking efforts, we identified eighteen individuals in total with de novo or loss-of-function variants in EIF3A (n = 4) or EIF3B (n = 14). The clinical phenotype varied but predominantly included cardiac defects, craniofacial dysmorphisms, mild developmental delays, and behavioral abnormalities. These genes encode core subunits of the eukaryotic initiation factor 3 (eIF3) complex, which plays a critical role in binding mRNA transcripts to the 40S ribosomal subunit during translation initiation. Both genes are highly constrained against loss of function, and animal models have demonstrated that disruptions in the eIF3 complex result in a range of developmental defects, including cardiovascular malformations. Additionally, EIF3B is located within the minimally overlapping region implicated in cardiac anomalies associated with 7p22.3 microdeletions. We sought to further study the role of these genes in syndromic congenital heart disease. To explore their functional impact, we generated zebrafish models with mutations in the orthologous eif3s10 and eif3ba genes, which resulted in developmental abnormalities, including thin heart tubes, lack of craniofacial cartilage, and embryonic lethality. We propose that pathogenic variants in EIF3A, as well as pathogenic variants or microdeletions involving EIF3B, cause a distinct autosomal-dominant neurodevelopmental syndrome characterized by cardiovascular and craniofacial manifestations.

Loss-of-function variants in EIF3A and EIF3B can lead to a variable multisystem congenital disorder, primarily involving cardiac anomalies, craniofacial differences, and mild neurodevelopmental features. These findings are supported by an international clinical cohort and zebrafish mutant models.

Introduction

Congenital heart disease (CHD) is common, affecting ∼1% of live births, and is often associated with significant morbidity and mortality.^1,2 While the genetic basis of CHD is increasingly recognized, an identifiable genetic cause is found in only ∼35% of affected individuals (aneuploidy, copy-number variant, or monogenic variant), with a higher diagnostic yield in individuals with extracardiac abnormalities (i.e., syndromic CHD).^3,4,5,6 Genome and exome sequencing continue to reveal disease-causing variants in genes associated with syndromic and nonsyndromic CHD, including those with well-established associations, as well as those for which evidence is still emerging. As part of a broader effort to sequence the genomes of individuals with CHD at the Hospital for Sick Children (SickKids), the Cardiac Genome Clinic initially identified two unrelated individuals with de novo loss-of-function (LoF) variants in genes encoding core subunits of the eukaryotic initiation factor 3 (eIF3) complex: EIF3A (MIM: 602039) and EIF3B (MIM: 603917). Through international matchmaking efforts, we assembled a cohort of eighteen individuals with de novo or LoF variants in EIF3A (n = 4) or EIF3B (n = 14).

EIF3A and EIF3B encode two of thirteen subunits of the eIF3 complex, the largest translation initiation factor complex, which plays a critical role in the initiation of protein synthesis.^7,8,9 The eIF3 complex binds to the 40S small ribosomal subunit, preventing its premature association with the 60S ribosomal subunit and facilitating mRNA transcript binding to the ribosome.¹⁰ Recent evidence suggests that eIF3A and eIF3B can directly bind to mRNA and may provide transcript specificity during translation.⁸ Both subunits function as scaffold proteins that are essential for proper assembly of the eIF3 complex, a role that is highly conserved across eukaryotic species.^11,12 eIF3a, the largest subunit of the eIF3 complex, is critical for the stabilization of the complex on the 40S ribosomal subunit^7,13 and can regulate the selective translation of specific mRNAs.^14,15 The eIF3b subunit is essential for anchoring the eIF3 complex to the small ribosomal subunit and facilitating the assembly of the translation initiation machinery.¹⁶ EIF3A and EIF3B are ubiquitously expressed across tissues and developmental stages.^12,17 Additionally, their pLOEUF (putative LoF observed/expected upper fraction) scores in the Genome Aggregation Database (gnomAD), 0.18 for EIF3A and 0.11 for EIF3B, indicate a high intolerance to LoF variants for both genes.¹⁸

EIF3B is located within the boundary of microdeletions reported at chromosome region 7p22.3 in individuals with developmental delay, intellectual disability, craniofacial dysmorphisms, and CHD, particularly tetralogy of Fallot (TOF).^19,20,21 Notably, EIF3B is one of only two genes within the smallest region of overlap among individuals affected by CHD.^{22,23,24,25,26,27,28} Regarding EIF3A, a missense variant, c.1145A>G (GenBank: NM_003750.4) (p.Tyr382Cys), was found to segregate with left-ventricular noncompaction (LVNC) in two related individuals.²⁹ Given the strong constraint against LoF in these genes and their observed associations with disease phenotypes, we investigated EIF3A and EIF3B variants as potential contributors to syndromic cardiac malformations.

In this study, we report the phenotype and genotype of fourteen unrelated individuals with heterozygous LoF and/or de novo heterozygous variants in EIF3B, as well as four unrelated individuals with de novo heterozygous LoF variants in EIF3A. These individuals present with a variable phenotype that includes CHD, craniofacial dysmorphisms, and mild neurodevelopmental abnormalities. We further validated the link between the loss of eIF3a or eIF3b and the development of CHD and craniofacial dysmorphism using zebrafish as an animal model. Zebrafish mutants exhibited severe developmental defects, including hypoplastic heart tubes, absent craniofacial cartilage, coloboma, reduced overall size, and embryonic lethality. Integrating the findings from our human cohort and zebrafish models, we propose that haploinsufficiency of either EIF3A or EIF3B causes an eIF3-related cardiovascular, craniofacial, and neurodevelopmental disorder.

Subjects, material, and methods

Subject recruitment

Two initial individuals were identified through the Cardiac Genome Clinic and subjected to a whole-genome short-read sequencing study at the Hospital for Sick Children (SickKids) and the Ted Rogers Centre for Heart Research.³⁰ The remaining sixteen individuals were identified via GeneDx, GeneMatcher,^31,32 and the Genomics England database.³³ Detailed clinical data for each proband were collected from collaborating institutions. Informed consent was obtained from all individuals and their families through the respective institution’s research ethics board (REB)/institutional review board (IRB). This study was approved by the REB at the Hospital for Sick Children (SickKids), Toronto, ON, Canada (REB #1000053844).

Identification and evaluation of the variants

Variants in EIF3B and EIF3A were identified in probands using massively parallel sequencing (next-generation sequencing [NGS]) technologies, including exome and genome sequencing, in clinical diagnostic or research settings. Parental testing for the identified variant was performed when possible (n = 17). Candidates from Genomics England were identified by querying for pLoF variants using the Interactive Variant Analysis (IVA) tool within the Genomics England research environment.³³ Clinical collaboration requests were sent to providers of all rare disease cohort participants with Integrative Genomics Viewer-confirmed pLoF variants as of April 2024, inviting them to participate in the clinical series. To assess the presence of the variants in control populations, we used gnomAD (v.4.1.0).¹⁸ All variants are described using the GenBank: NM_003750.4; NP_003741.1 MANE select transcript of EIF3A and the GenBank: NM_001037283.2; NP_001032360.1 MANE select transcript of EIF3B.³⁴

Zebrafish husbandry

Adult TLAB zebrafish (Danio rerio) were maintained in accordance with the Canadian Council on Animal Care (CCAC) and the Hospital for Sick Children Animal Services (LAS) guidelines. Embryos were raised in embryo medium at 28.5°C, as described in “The Zebrafish Book.”³⁵ Stable transgenic lines were used to characterize defects in mutants, including Tg(myl7:EGFP)^twu3436 and Tg(gata1:dsRed)^sd2.³⁷

Zebrafish mutant generation and genotyping

The CRISPR-Cas9 system was used to target early exons in eif3s10 and eif3ba. Guide RNAs (gRNAs) were designed using Benchling (www.benchling.com) and synthesized with the HiScribe T7 Quick High Yield RNA Synthesis Kit (NEB), followed by isopropanol/acetate precipitation.³⁸ The Cas9 protein was purified as described by Gagnon et al.,³⁹ with two modifications: cultures were induced with 0.3 mM IPTG (instead of auto-induction) and the dialysis buffer contained 10% glycerol. gRNAs and Cas9 were mixed in a 1:1 ratio (4 μM each) with a final concentration of 300 mM KCl to keep Cas9 in the solution⁴⁰ and then incubated at 37°C for 5 min before injection. 1 nL of the gRNA/Cas9 solution was injected into the yolk at the one-cell stage. Once the injected embryos reached adulthood, they were outcrossed to identify founders with germline mutations. Mutants carrying the desired allele were then outcrossed for at least three generations (F3) before being incrossed for phenotyping.

The eif3s10 guide sequence is 5′-GAAGACCGAGACCGCTAAAG-3′. Two alleles were isolated for eif3s10: a −1 allele (hsc223) or a +5 A>G allele (hsc224), both located in exon 3 and leading to a premature stop codon early in exon 4 (Figure S7A). As both alleles exhibited the same phenotype (Figure S7B), allele hsc224 was used for the remainder of the study. Genotyping was performed using high-resolution melt (HRM) analysis,⁴¹ with the following primers: eif3s10 HRM F: 5′-TCGTCCGTGCTTACCTCAAAC-3′ and eif3s10 HRM R: 5′-ATCCTCAATGTCCAGAACCATCT-3′. To confirm the specific insertion or deletion (indel) and generation of a premature stop codon, the region surrounding the indel was PCR amplified using the primers eif3s10 seq F: 5′-GGAAGACAACCTCTCTCTCTGGTT-3′ and eif3s10 seq R: 5′-TCGAGCAGCAAAATGAAGCTGAA-3′. Sanger sequencing was performed by the Centre for Applied Genomics (TCAG) at SickKids.

The eif3ba guide sequences are g1, 5′-GCCTTCGTTCAGTGATCCGG-3′ and g2, 5′-GGAGGCTGACGGCATCGACT-3′. One allele was isolated using guide 1 (g1): a +4 C>T in exon 1 (hsc225) (Figure S7A). A second allele was isolated using guide 2 (g2): a −4 deletion in exon 2 (hsc226) (Figure S7A). Both alleles resulted in a premature stop codon and exhibited the same phenotype (Figure S7B); therefore, the allele hsc225 was used for the remainder of the study. Genotyping was performed using HRM analysis with the following primers (for g1 and g2, respectively): eif3ba HRM 1F: 5′-AGGCCTCTCTCTCACCCTGAT-3′ and eif3ba HRM 1R: 5′-ATGGAGGCCGAAATGGAGTATG-3′, or eif3ba HRM 2F: 5′-CCGACCTGAGGCACATTATCAA-3′ and eif3ba HRM 2R: 5′-GATGTTCTCAGGGACAAACCG-3′. Indels were then confirmed by Sanger sequencing of PCR products surrounding the cut sites, using the following primers (g1 and g2, respectively): eif3ba seq g1F: 5′-CATGGTACGGTCCTTCTCTCG-3′ and eif3ba seq g1R: 5′-GGCATTAACTGTGAAACAGCGT-3′, or eif3ba seq g2F: 5′-GCAGCTGTTATGCTTTTGGCTT-3′ and eif3ba seq g2R: 5′-GCAGACCCAAACATACCCCTT-3′. Sequencing was performed by the TCAG at SickKids.

Imaging

Bright-field and fluorescent images and videos were captured using a Zeiss AXIO Zoom V16 microscope with ZEN software (Carl Zeiss AG). Image processing was performed in Fiji,⁴² where images were cropped and the brightness/contrast were slightly adjusted when necessary. For heart rate analysis with PyHeart4Fish,⁴³ 500 frames were captured (exposure: 40 milliseconds, frame rate: 25 frames per second). Embryos were treated with 1-phenyl-2-thiourea (PTU) 1 day post-fertilization (dpf) to prevent the development of pigmentation, which would hinder video analysis. Still images of live embryos were obtained after anesthetizing them with tricaine and immobilizing them in 3% methyl cellulose.

Immunofluorescence

Immunofluorescence staining was performed as described by Rosenthal et al.,⁴⁴ with minor modifications. Briefly, embryos were fixed at 3 dpf in 4% paraformaldehyde (PFA) overnight, followed by washes in phosphate-buffered saline (PBS) + 0.1% Tween 20 (PBSTw) and dehydration through a methanol gradient. Embryos were stored in 100% methanol at −20°C until staining. Prior to staining, embryos were rehydrated (reverse methanol gradient) and bleached with a solution of 0.8% KOH, 0.9% H₂O₂, and 0.1% Tween 20 in sterile water until the eyes appeared clear.⁴⁵ Next, the embryos were permeabilized with ice-cold acetone and blocked with 10% normal goat serum + 5% bovine serum albumin in PBSTx (PBS with 1% Triton X-100), as described by Rosenthal et al.⁴⁴ The embryos were incubated overnight at 4°C with mouse anti-HuC/HuD monoclonal primary antibody (clone 16A11, #A-21271, Invitrogen) at a 1:500 dilution in blocking buffer. After PBSTx washes, goat anti-rabbit immunoglobulin (Ig)G (H+L) cross-adsorbed secondary antibody conjugated to Alexa Fluor 488 (#A-11008, Invitrogen) was added at a 1:200 dilution in blocking buffer and incubated overnight at 4°C. Following the final PBSTw washes, embryos were mounted in methyl cellulose for imaging and subsequently genotyped via HRM.

Alcian blue staining

Alcian blue staining for cartilage was performed as outlined in the referenced protocol (see web resources), with modifications to the bleaching procedure. The embryos were fixed at 5 dpf in 4% PFA overnight and then washed in PBSTw and dehydrated through a methanol gradient (50% MeOH/PBSTw and 100% MeOH). After rehydration, the embryos were bleached in the dark as described for immunofluorescence and then rinsed with PBSTw before staining overnight at room temperature in an Alcian blue solution (1% concentrated hydrochloric acid, 70% ethanol, and 0.1% Alcian blue). The embryos were rinsed with acidic ethanol (70% EtOH + 5% HCl) and washed through a gradient of acidic ethanol/water (75%, 50%, and 25%) before final storage in 100% water. For imaging, the embryos were mounted in 80% glycerol-KOH, followed by genotyping via HRM.

Acridine orange staining

1 or 2 dpf embryos were manually dechorionated and then treated with 3 μg/mL acridine orange (in embryo medium) for 1 h in the dark at 28.5°C. Embryos were then washed twice for 5 min with embryo medium, anesthetized with tricaine, and mounted in 1% low melt agarose for imaging. Fiji⁴¹ was used to identify fluorescent foci.

Results

Summary of clinical findings

Through the Cardiac Genome Clinic at SickKids,³⁰ GeneDx, GeneMatcher,^31,32 and Genomics England,³³ we studied fourteen unrelated individuals with heterozygous damaging EIF3B variants. This included ten individuals with LoF variants, one individual with a de novo microdeletion overlapping EIF3B and adjacent genes, one individual with a de novo splice acceptor variant, one individual with a de novo intronic variant, and one individual with a de novo missense variant (Figure 1). Additionally, four unrelated individuals with heterozygous de novo LoF variants in EIF3A were studied (Figure 1). Additional findings from genome or exome sequencing are detailed in Table S3.

Figure 1.

Schematic of EIF3A and EIF3B domains, highlighting the proteasome component domain in EIF3A, and the RNA recognition motif in EIF3B

Variants identified in the human cohort are shown in orange. The EIF3B full-gene deletion, hg38 (chr7:2314321–2554246)x1 also encompasses exon 1 of SNX8, CHST12, GRIFIN, LFNG, and several exons in the 3ʹ end of BRAT1.

EIF3B cohort

In the EIF3B cohort, eleven of fourteen individuals presented with CHD, including four individuals (probands #1, #2, #3, and #10) with TOF (Tables 1, 2, 3, and S1). One individual (proband #6) presented with a large perimembranous ventricular septal defect (VSD), small secundum atrial septal defect (ASD), pulmonary stenosis, and aortic dilation. Proband #8 presented with subaortic membrane stenosis, conoventricular VSD, aortic insufficiency, and an anomalous muscle bundle in the right ventricle, while proband #14 presented with pulmonary atresia with patent septum type 2. Other cardiac lesions included septal defects, bicuspid aortic valve (BAV), patent ductus arteriosus (PDA), and persistent left vena cava superior draining into the coronary sinus. Neurodevelopmental phenotypes were reported in eight individuals, including developmental delay, speech and language delay, intellectual disability, and mild or specific learning disabilities. Behavioral abnormalities, including attention-deficit hyperactivity disorder and autism spectrum disorder, were also noted. Facial differences were observed in eleven individuals (Tables 3 and S1; Figure 2). While features varied, several individuals exhibited differences in the eye region that overlapped with those previously reported in individuals with 7q22.3 microdeletions, including ptosis, arched eyebrows, downslanting palpebral fissures, hypertelorism, and epicanthal folds. Ptosis required intervention in proband #8. Cleft lip and palate were seen in four individuals in this study. One individual (proband #9) had a transverse terminal deficiency of the left hand (Figure S4; Table S1). Four probands were reported to have hearing loss and/or inner and middle ear malformations (Table S1).

Table 1.

Summary of clinical characteristics in the EIF3B and EIF3A cohort and previously reported 7p22.3 microdeletion cases overlapping EIF3B

Clinical characteristics	EIF3B microdeletions (n = 10)a	EIF3B probands (n = 14)	EIF3A probands (n = 4)	Total (N = 28)
Developmental delay	7	6	1	14 (50%)
Intellectual disability/cognitive impairment	4	1	0	5 (18%)
Speech and language delay	5	5	2b	12 (43%)
Learning difficulties	N/A	2	1c	3 (11%)
Autism spectrum disorder	2	3	0	5 (18%)
Attention-deficit disorder	1	5	0	6 (21%)
Congenital heart defects	7	11	4	22 (79%)
Tetralogy of Fallot	3	4	2	9 (32%)
Craniofacial dysmorphisms	8	11	4	23 (82%)

N/A, not applicable.

^aPublished 7p22.3 microdeletions overlapping with EIF3B (Gallego et al.,²² Richards et al.,²³ Silversides et al.,²⁴ Yu et al.,²⁶ and Skvortsova et al.²⁸) (Table S2).

^bOne individual was previously reported to have speech and language delay but is currently within normal limits. The other individual was reported to have mild articulation problems.

^cQuery of a learning difficulty.

Table 2.

Genetic information for probands with de novo or loss-of-function variants in EIF3B and EIF3A

Demographics and family information				Genetic information
Proband	Sex	Age at last assessment	Family history	Gene	Variant	gnomAD frequency (v.4.1.0)	Inheritance pattern	Detection method
1	F	1 y, 1 mo	mother with hip dysplasia	EIF3B	c.459del (p.Phe153Leufs∗11)	absent	maternal	GS
2	M	TOP at 22 6/7 weeks	none	EIF3B	c.2182T>C (p.Ser728Pro)	absent	de novo	ES
3	F	7 y, 8 mo	unknown; adopted	EIF3B	c.415G>T (p.Glu139∗)	absent	unknown	ES
4	M	11 y	none	EIF3B	c.2342−2_2342−1del (p.?)	0.000063% (1 allele)	de novo	GS
5	F	9 y	none	EIF3B	c.923_924del (p.Val308Aspfs∗3)	absent	de novo	ES
6	M	10 y, 8 mo	none	EIF3B	c.983C>G (p.Ser328∗)	absent	de novo	GS
7	F	11 y, 11 mo	none	EIF3B	GRCh38/hg38 7p22.3 (chr7:2314321–2554246)x1 240 kbp	N/A	de novo	ES
8	M	9 y	none	EIF3B	c.1085dup (p.Glu363Argfs∗18)	0.00012% (2 alleles)	de novo	ES
9	M	15 y	familial short stature	EIF3B	c.1545C>A (p.Cys515∗)	absent	de novo	ES
10	M	9 mo	none	EIF3B	c.2233G>T (p.Glu745∗)	absent	de novo	GS
11	F	2 mo	none	EIF3B	c.1995C>A (p.Tyr665∗)	absent	de novo	ES
12	F	16 y, 4 mo	cleft palate	EIF3B	c.2097_2098dup (p.Cys700Serfs∗14)	absent	unknown, not maternal	GS
13	M	3 mo	none	EIF3B	c.1289+2dup (p.?)	absent	de novo	ES panel
14	M	8 y	mother unaffected	EIF3B	c.918_919del (p.Tyr306∗)	absent	maternal	GS
15	M	6 y	mother with ventricular septal defect	EIF3A	GRCh38/hg38 10q26.11 (chr10:119051299–119053879)x1 2.58 kbp	N/A	de novo	GS
16	F	6 y, 10 mo	family history of autism, ADHD, and heart problems; mother unaffected	EIF3A	c.295_296del (p.Glu99Lysfs∗3)	absent	maternal	GS
17	F	1 y, 1 mo	mother had 1 early pregnancy loss	EIF3A	c.1209_1210del (p.Cys404∗)	absent	de novo	ES
18	M	6 y	father with downslanting palpebral fissures	EIF3A	c.3088C>T (p.Arg1030∗)	absent	de novo	ES

All EIF3A variants are mapped to GenBank: NM_003750.4; NP_003741.1, and all EIF3B variants are mapped to GenBank: NM_001037283.2; NP_001032360.1. ADHD, attention-deficit hyperactivity disorder; ES, exome sequencing; GS, genome sequencing; kbp, kilobase pairs; mo, months; N/A, not applicable; y, year(s); TOP, termination of pregnancy.

Table 3.

Detailed clinical characteristics for probands with de novo or loss-of-function variants in EIF3B and EIF3A

Proband	Congenital heart defects	Craniofacial dysmorphisms	Behavioral and neurodevelopmental features	Seizure
1	TOF	none	none	–
2	TOF	large head, square face, microretrognathia	N/A	–
3	TOF	N/A	ADHD, aggression, DD, memory retention concerns	–
4	interatrial communication (spontaneously closed)	bilateral cleft lip and palate	delay in verbal language without intellectual disability, autism spectrum disorder	+
5	none	downslanting palpebral fissures, short neck, Noonan-like facies	ADHD, learning difficulties	–
6	pulmonary stenosis, ASD, VSD, aortic dilation	hypertelorism with telecanthus, arched eyebrows, eyelid ptosis, downslanting palpebral fissures, broad nasal root, posteriorly rotated ears, broad face, flattened and thin philtrum, Cupid’s bow mouth	early DD, ADHD, autism spectrum disorder, dyslexia	–
7	none	cleft lip and palate	none; speech delay due to hearing loss	–
8	VSD, subaortic membrane stenosis, aortic insufficiency, anomalous muscle bundle of the right ventricle	ptosis	none	–
9	BAV	severe congenital ptosis OS > OD, dysplastic ears, mild malar hypoplasia, microcephaly	DD without cognitive impairment	–
10	TOF	bilateral cleft lip and palate, no teeth, large crus	N/A	–
11	VSD and ASD	microcephaly, dysmorphic features: broad forehead, temporal balding, arched eyebrows, broad nasal bridge, small nose, midface hypoplasia, flat philtrum, short neck	possible ADHD, mild language delay	–
12	none	none	mild learning disability	–
13	PDA, three small muscular VSD, perimembranous VSD, small ASD, persistent left vena cava superior draining in the coronary sinus	progressive microcephaly, high forehead with slightly prominent metopic suture, bilateral epicanthus, hypertelorism, prominent nasal tip, small mouth, retrognathia	none	–
14	pulmonary atresia with patent septum type 2	cleft lip and palate	intellectual disability, walked at 18 months, no language, hyperactivity, aggression, autism spectrum disorder	–
15	TOF	thin upper lip vermillion, retrognathia, epicanthus, anteverted nares, deeply set eyes	delayed speech and language now within normal limits	+
16	TOF with right-sided aortic arch	low-set posteriorly rotated ears, long philtrum, flat nasal bridge	met developmental milestones; 1:1 supervision at school for visual difficulties	–
17	VSD, ASD, PFO	thin upper vermillion, quite flat midface, downturned mouth corners	mild articulation problems	–
18	VSD, right-sided aortic arch, vascular ring	acrocephaly, downslanting palpebral fissures, epicanthal folds, telecanthus, deeply set eyes, maxillary hypoplasia, malar hypoplasia, micrognathia, triangular facies, asymmetry, tented lips, high arched palate, small flat teeth	DD	–

Please see Table S2 for the full table. ADHD, attention-deficit hyperactivity disorder; ASD, atrial septal defect; BAV, bicuspid aortic valve; DD, developmental delay; ECHO, echocardiogram; HC, head circumference; N/A, not applicable; NDD, neurodevelopmental delay; NGS, next-generations sequencing; OS > OD, more pronounced in the left eye (oculus sinister) compared to the right (oculus dexter); PDA, patent ductus arteriosus; PFO, patent foramen ovale; TOF, tetralogy of Fallot; VSD, ventricular septal defect; +, present; −, absent.

Figure 2.

Photographs of probands with de novo variants in EIF3B or EIF3A

(A–C) Clinical images of proband #4, with a canonical acceptor splice site variant in EIF3B; (D–F) proband #9 with a loss-of-function variant in EIF3B; (G–H) proband #15, with a loss-of-function variant EIF3A; and (I and J) proband #18, with a loss-of-function variant in EIF3A.

EIF3A cohort

Within the EIF3A cohort, two individuals (probands #15 and #16) presented with TOF. A third individual (proband #17) had a perimembranous VSD, ASD, and patent foramen ovale. The fourth individual (proband #18) presented with VSD, right-sided aortic arch, and a vascular ring (Tables 1, 2, 3, and S1). One individual (proband #15) had a history of speech and language delays but is currently within normal limits. Proband #16 was suspected of having a learning difficulty, although all developmental milestones were met. Proband #17 exhibited mild articulation issues, and proband #18 was reported to have a developmental delay. Seizures were reported in one of these individuals. All four individuals exhibited nonspecific facial differences (Tables 3 and S1; Figure 2). Additional features observed included hearing loss, bilateral coloboma, and symphalangism of the left thumb in one individual (proband #16, Table S1).

Spectrum of the EIF3B LoF variants

Fourteen distinct EIF3B variants were identified (Tables 2 and S1; Figure 1). Ten were LoF sequence variants predicted to undergo nonsense-mediated decay (probands #1: Figure S1, family 1 #II-1; #3: Figure S1, family 3 #II-1; #5: Figure S1, family 5 #II-1; #6: Figure S1, family 6 #II-1; #8–#12: Figure S1, families 8–12 individuals #II-1; and #14: Figure S1, family 14 #II-1). The variant identified in proband #1,c.459del (GenBank: NM_001037283.2) (p.Phe153Leufs^∗11), was found to be maternally inherited, and the proband’s mom presented with hip dysplasia. The variant identified in proband #14, c.918_919del (GenBank: NM_001037283.2) (p.Tyr306^∗), was also maternally inherited, and the mother was not reported to be clinically affected. Of the eight remaining truncating variants, seven were reported to be de novo when parental testing was available. The variant identified in proband #7 (Figure S1, family 7 #II-1), chr7:2314321–2554246 (GRCh38]) chr7:2353956–2593880 (GRCh37), is a de novo 240 kbp microdeletion at chromosome region 7p22.3 that overlaps the 5′ end of SNX8 (MIM: 614905), EIF3B, CHST12 (MIM: 610129), GRIFIN (MIM: 619187), and LFNG (MIM: 602576) and several exons in the 3′ end of BRAT1 (MIM: 614506). The genes SNX8, CHS12, and GRIFIN do not have an established OMIM Morbid association, while LFNG and BRAT1 are associated with autosomal-recessive phenotypes in OMIM (MIM: 609813, 618056, and 614498). Similar microdeletions overlapping EIF3B and additional adjacent genes have previously been reported in the literature in individuals with a variable phenotype that includes CHD, craniofacial malformations, and neurodevelopmental phenotypes^{22,23,24,26,28} (Figure S3; Table S2). The de novo c.2342−2_2342−1del (GenBank: NM_001037283.2) (p.?) variant in proband #4 (Figure S1, family 4 #II-1) is located at the canonical acceptor splice site of the last coding exon (exon 18). RNA sequencing confirmed that this variant results in intron 17 retention and disruption of the protein starting from amino acid 782. However, it is not expected to undergo nonsense-mediated decay (Figure S5). The de novo variant in proband #13 (Figure S1, family 13 #II-1), c.1289+2dup (GenBank: NM_001037283.2) (p.?), is located within the splicing region of exon 7. The in silico splicing tools SpliceSiteFinder-Like,⁴⁶ MaxEnt,⁴⁷ and NNSPLICE⁴⁸ predict that it disrupts splicing by disrupting the canonical donor site at exon 7. RNA functional studies have not been performed for this variant. The de novo variant in proband #2 (Figure S1, family 2 #II-1), c.2182T>C (GenBank: NM_001037283.2) (p.Ser728Pro), is a missense change with in silico prediction scores of CADD: 25.8⁴⁹ and REVEL: 0.359.⁵⁰

Spectrum of the EIF3A LoF variants

Four LoF variants were identified (Tables 2 and S1; Figure 1). A de novo 2.58 kbp intragenic deletion, chr10:119051299–119053879 (GRCh38) chr10:120810811–120813391 (GRCh37) at chromosome region 10q26.11, was identified in proband #15 (Figure S2, family 15 #II-1). The distal breakpoint of this deletion is located in intron 14 of EIF3A, while the proximal breakpoint extends 23 bp into the 5ʹ end of exon 15, disrupting its canonical acceptor splice site. Sanger sequencing was used to confirm the deletion result and further define its exact genomic coordinates (Figure S6). Additionally, de novo nonsense variants, c.1209_1210del (GenBank: NM_003750.4) (p.Cys404^∗) and c.3088C>T (GenBank: NM_003750.4) (p.Arg1030^∗), were found in proband #17 (Figure S2, family 17 #II-1) and proband #18 (Figure S2, family 18 #II-1), respectively. A maternally inherited nonsense variant, c.295_296del (GenBank: NM_003750.4) (p.Glu99Lysfs^∗3), was also identified in another individual (proband #16; Figure S2, family 16 #II-1), though the mother reportedly did not exhibit any related features.

Loss of orthologous eif3s10 and eif3ba genes in zebrafish causes a spectrum of defects similar to those seen in the human cohort

To further investigate the link between EIF3A and EIF3B variants and the individuals’ phenotypes described above, zebrafish mutant models were created using CRISPR-Cas9 to target early exons in the orthologous genes. EIF3A has a single ortholog, eif3s10, while EIF3B has two orthologs: eif3ba and eif3bb. The presence of two orthologs is common in zebrafish due to a whole-genome duplication event in the teleost lineage.⁵¹ We chose to pursue eif3ba for this study, as targeting eif3bb did not cause a phenotype (data not shown), which may suggest divergence of its function. For eif3s10 and eif3ba, the generated indels resulted in an early stop codon occurring upstream of or inside the essential protein domains (the proteasome component [PCI] domain in eif3s10 or the RNA recognition motif [RRM] in eif3ba) (Figure S7A). Homozygous eif3s10 and eif3ba mutants exhibited similar, fully penetrant phenotypes, while heterozygous loss of either gene caused no overt defects. Additionally, transheterozygote (eif3ba^+/−; eif3s10^+/−) mutants displayed no phenotype. In homozygous eif3s10^−/− or eif3ba^−/− mutants, minor yolk sac edema first became visible at approximately 26–28 h post-fertilization (hpf). By 30 hpf, mutants exhibited reduced eye size and worsening edema. The phenotype became extremely apparent by 2 days post-fertilization (dpf), characterized by severe pericardial edema, significantly reduced eye and head size, reduced pigmentation (particularly in the eye), and trunk curvature (Figure 3A). Embryo body length was also significantly reduced (Figure 3B). Many of the characteristics (small head, small eyes, short body length, and reduced pigmentation) point to an overall growth and developmental delay in mutant embryos. Additionally, mutants exhibited a significantly reduced head angle, which is routinely used to determine the age of embryos,⁵² further providing evidence for a growth delay (Figure 3B). The reduced size may also be due to an increase in apoptosis in mutant embryos. To assess this, apoptotic cells were stained with the vital dye acridine orange at 1 and 2 dpf. At 1 dpf, there was a subtle increase in apoptosis for eif3ba^−/− mutant embryos, while there was no significant change for eif3s10^−/− mutant embryos (Figure 3C). However, by 2 dpf, there was a significant uptick in apoptosis compared to wild-type (WT) or heterozygous siblings (Figure 3C). This increase in cell death, combined with the overall growth delay, likely accounts for the reduced size and delayed morphological characteristics. Finally, mutant embryos showed no swimming activity and failed to exit their chorions spontaneously. While touch responsiveness was present at 24–30 hpf, it was lost by 3 dpf as the phenotype progressed. All mutants were uniformly dead by 5 dpf. Altogether, these findings demonstrate the essential roles of eif3s10 and eif3ba during early development in zebrafish, aligning with the phenotypes observed in affected individuals.

Figure 3.

Loss of eif3a and eif3b in zebrafish causes a spectrum of defects similar to those observed in affected individuals

(A) Bright-field images of eif3s10^−/− and eif3ba^−/− embryos compared to heterozygous siblings at 2 and 3 dpf. Scale bars: 500 μm. Mutant embryos exhibit reduced size, delayed pigmentation, trunk curvature, and pericardial edema.

(B) Body length is significantly reduced at 3 dpf for both eif3s10^−/− and eif3ba^−/− mutant embryos. Additionally, head angle is significantly reduced at 3 dpf, indicative of a growth delay. Significance was determined by one-way ANOVA with Tukey multiple comparisons test; ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, and ^∗∗∗p ≤ 0.001. Each data point is one embryo, body length n = 2–7, and head angle n = 4–9.

(C) Acridine orange staining for apoptotic cells indicates a slight increase in apoptosis for eif3ba^−/− embryos at 1 dpf, while there is no change for eif3s10^−/− embryos. At 2 dpf, there is a significant increase in apoptosis for both eif3s10^−/− and eif3ba^−/− embryos when compared to their heterozygous or WT siblings. The orange asterisk (anal opening) and orange arrowhead (retina) indicate regions that have high levels of apoptosis during normal development; therefore, a high concentration of acridine orange is expected even in WT embryos. Significance was determined by one-way ANOVA with Tukey multiple comparisons test; ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, and ^∗∗∗p ≤ 0.001. Scale bars: 100 μm. Each data point is one embryo, 1 dpf n = 5–12 embryos, and 2 dpf n = 5–11 embryos.

eif3s10 and eif3ba LoF mutants have underdeveloped hearts and significantly impaired cardiac function

Given that many of the individuals exhibit TOF or other cardiac abnormalities, we closely investigated the cardiac phenotype in zebrafish eif3s10 and eif3ba mutants. At 2 dpf, while WT and heterozygous siblings developed a looped, two-chambered heart, homozygous mutant embryos maintained a more linear heart tube (Figure 4A). By 3 dpf, these mutants exhibited hypoplastic heart tubes with little to no looping and severe pericardial edema. Their heartbeats were slow, weak, irregular, and insufficient for proper blood circulation (Videos S1, S2, and S3). To determine whether edema contributed to the phenotype, embryos were raised in 175 mM mannitol salts to balance osmolarity⁵³; however, this treatment did not alleviate the stretched heart phenotype (Figure S7C), indicating a primary defect in heart development.

Figure 4.

eif3s10 and eif3ba loss-of-function mutants exhibit underdeveloped hearts and significantly impaired cardiac function

(A) At 2 dpf, homozygous mutants have stretched out hearts compared to WT. This worsens by 3 dpf, with hypoplastic heart tubes and severe pericardial edema. Scale bars: 100 μm.

(B) A greater proportion of homozygous mutants show “severe”-looking hearts at 3 dpf compared to heterozygous or WT siblings from the same clutch. Categories: 1 = WT-like, fully looped heart, no pericardial edema; 2 = heart is mostly looped, small pericardial edema; 3 = little to no looping, heart is stretched out, small chambers, large pericardial edema; and 4 = thin heart tube with no clear chambers or looping, severe pericardial edema that extends beyond the yolk (note: this image is lateral, the others are ventral). The gray dashed line in example images indicates the edge of pericardial edema. 3 clutches were scored per gene. eif3s10 n = 133, 71, and 99 embryos. eif3ba n = 88, 169, and 112 embryos. Proportions of each phenotypic category per genotype were averaged prior to plotting.

(C) Mutant heart rate (beats per minute [bpm]), measured via PyHeart4Fish video analysis, was reduced compared to heterozygous and WT siblings. Significance was determined by one-way ANOVA with Tukey multiple comparisons test; ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, and ^∗∗∗p ≤ 0.001. Each data point is one embryo, and n = 3–15 embryos.

(D) Arrhythmia score at 2 dpf, measured by PyHeart4Fish, reveals irregular heartbeats in eif3s10 and eif3ba mutants. A lower score indicates more regular heartbeats, with 0.7 considered the cutoff for arrhythmia. Significance was determined by one-way ANOVA with Tukey multiple comparisons test; ^∗p ≤ 0.05. Each data point is one embryo, and n = 3–7 embryos.

(E) Sample heartbeat traces at 2 and 3 dpf (measured by PyHeart4Fish) demonstrate irregular heartbeats for mutants vs. heterozygotes (top). Blue lines indicate atrial beats, while orange lines indicate ventricular beats.

Video S1. At 3 dpf, WT zebrafish exhibit a strong heartbeat that moves blood throughout the trunk

Blood is visualized by a Tg(gata1:dsRed) reporter.

While 100% of the homozygous mutants exhibited some degree of cardiac abnormality at 3 dpf (and 100% were dead by 5 dpf), there was a range in severity. This was classified into four groups, ranging from “WT-like” (1) to most severe (4), based on the extent of the pericardial edema, size of chambers, and degree of looping (Figure 4B). Entire clutches of eif3s10 or eif3ba embryos were scored at 3 dpf before genotyping, revealing that severe cardiac phenotypes (classes 3 and 4) were strongly associated with eif3s10 or eif3ba loss compared to heterozygous or WT siblings. Heart rate and cardiac function of embryos were analyzed at 2 and 3 dpf via videomicroscopy using PyHeart4Fish⁴³ (Videos S4, S5, S6, S7, S8, S9, S10, and S11). Consistent with the phenotyping bins, these metrics were measured prior to genotyping. Heart rate was reduced in homozygous mutants compared to their heterozygous or WT siblings at both 2 and 3 dpf (Figure 4C). Additionally, mutants exhibited irregular heartbeats, with high arrhythmia scores (>0.7 indicates arrhythmia) at 2 dpf (Figure 4D). Heartbeat traces revealed further dysfunction, including simultaneous atrial and ventricular contractions (instead of sequential beating), prolonged pauses between beats, and missing atrial or ventricular contractions, resulting in disorganized traces (Figure 4E). These findings suggest defects in cardiac conduction system function. Altogether, these results demonstrate that eif3s10 and eif3ba loss severely compromised both heart morphology and function in zebrafish mutants.

eif3s10 and eif3ba LoF mutants exhibit extracardiac abnormalities including reduced brain size, coloboma, and craniofacial cartilage defects

Beyond cardiac phenotypes, craniofacial dysmorphisms and mild neurodevelopmental abnormalities were among the most common characteristics observed in the human cohort. To investigate these features in our zebrafish mutants, we examined brain and craniofacial development. At 3 dpf, immunostaining for a pan-neuronal marker (HuC/HuD) revealed that homozygous mutants had significantly smaller brains compared to heterozygous or WT siblings (Figure 5A). Alcian blue staining showed a complete absence of craniofacial cartilage at 5 dpf in homozygous mutants, including the jaw and pharyngeal arches, retaining only minor cartilage deposition in the otic vesicle (Figure 5B). Mutant embryos also exhibited significantly smaller eyes when compared to their heterozygous or WT siblings, which is associated with microcephaly in zebrafish⁵⁴ (Figure 5C). Notably, mutants also exhibited coloboma (Figure 5C), a feature reported in one EIF3A individual (proband #16). Taken together, these findings demonstrate that loss of eif3s10 or eif3ba in zebrafish results in a broad spectrum of developmental defects, consistent with the phenotypes observed in affected individuals.

Figure 5.

eif3s10 and eif3ba loss-of-function mutants exhibit extracardiac abnormalities including reduced brain size, coloboma, reduced eye size, and craniofacial cartilage defects

(A) Immunofluorescent staining for a pan-neuronal marker (HuC/HuD) reveals reduced brain size in mutants compared to WT siblings at 3 dpf. Quantification revealed a significant decrease in brain area for both eif3s10 and eif3ba mutants. Significance was determined by one-way ANOVA with Tukey multiple comparisons test; ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001, and ^∗∗∗∗p ≤ 0.0001. Each data point is one embryo, and n = 5–8 embryos. All scale bars: 100 μm.

(B) Dorsal view of Alcian blue-stained embryos reveals a lack of facial cartilage (including jaw and pharyngeal arches) in mutant embryos compared to WT at 5 dpf. White arrowheads indicate otic vesicles, and white arrows indicate pharyngeal arches. All scale bars: 100 μm.

(C) Mutant embryos have small eyes and coloboma at 3 dpf, with white arrowheads marking incomplete closure of the optic fissure. Quantification revealed a significant decrease in eye area and minimum Feret's diameter. All scale bars: 100 μm. Significance was determined by one-way ANOVA with Tukey multiple comparisons test; ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001, and ^∗∗∗∗p ≤ 0.0001. Each data point is one embryo (right eye measured only), and n = 3–7 embryos.

Discussion

This study presents a cohort of individuals with heterozygous damaging variants in EIF3A and EIF3B, exhibiting a heterogeneous clinical phenotype primarily characterized by craniofacial abnormalities, CHD (most commonly TOF), and mild neurodevelopmental features. Among the fourteen damaging variants found in EIF3B and the four LoF variants found in EIF3A, thirteen were found to be de novo, two had unknown inheritance, and three were found to be maternally inherited (probands #1, #14, and #16). The mothers of these individuals were not reported to exhibit any relevant phenotypic features. Moreover, the variability observed in cardiac and craniofacial phenotypes is further suggestive of reduced penetrance and variable expressivity associated with clinically relevant variants in EIF3A and EIF3B. This clinical association is further supported by zebrafish models with null mutations in the orthologous genes, which largely match the range of phenotypes observed in affected individuals. Most variants in our cohort are LoF variants expected to result in nonsense-mediated decay, supporting haploinsufficiency as the mechanism of disease. The missense variant c.2182T>C (p.Ser728Pro) and the truncating variant c.2342−2_2342−1del in EIF3B do not directly impact known functional domains. More functional work is needed to determine how these variants impact protein function.

This study further supports haploinsufficiency of EIF3B as a contributing factor in 7p22.3 deletion syndrome, given that the gene is encompassed by the deleted region. Gallego et al.²² described a male individual with a 2,880 kbp deletion containing EIF3B, presenting with VSD, developmental delay, cleft palate, failure to thrive, and hypotonia. His mother, who carried the same deletion, had mild intellectual disability, facial dysmorphisms, and velopharyngeal insufficiency. Richards et al.²³ reported an individual with a de novo ∼400 kb deletion overlapping MAD1L1 (MIM: 602686), MRM2 (MIM: 606906), NUDT1 (MIM: 600312), SNX8, and the first exon of EIF3B. This individual presented with TOF, language delay, facial dysmorphisms, and failure to thrive.²³ Similarly, Silversides et al.²⁴ reported a male individual with a 180 kb deletion spanning SNX8, EIF3B, and CHST12, who also had TOF. More recent studies have identified six additional individuals with 7p22.3 microdeletions associated with developmental delay, facial dysmorphisms, and CHDs, including TOF, ASD, double aortic arch, and VSD.^26,28 While SNX8 has been considered as a candidate gene for the variable TOF and other CHDs in these individuals, neither SNX8 nor other nearby genes within the deletion breakpoints are constrained against LoF (Table S4). Moreover, two microdeletions (<200 kbp) overlapping SNX8 but not EIF3B have been reported in individuals without cardiac anomalies.^25,27 Overall, haploinsufficiency of EIF3B should be considered the primary cause for the cardiac phenotypes observed in individuals with this deletion and is also likely involved in the observed neurodevelopmental and craniofacial features, given the phenotypic overlap with our cohort of individuals with EIF3A or EIF3B pathogenic variants.

While some large-scale CHD genomic studies have not found damaging variants in EIF3A or EIF3B, one recent exome sequencing study reported high-confidence LoF EIF3B variants in two individuals with CHD (2/3,876) and none in control subjects (0/45,082),⁵⁵ further supporting the association between EIF3B and CHD. A previously reported missense variant in EIF3A (c.1145G>A [p.Tyr382Cys]) was identified in a family with LVNC and shown to affect cardiomyocyte proliferation, differentiation, and apoptosis.²⁹ While in vitro functional studies suggest that this variant may impact cardiomyocyte function, the association between these findings and the pathogenesis of LVNC has not been determined. The reported variant is present in 16 alleles in gnomAD v.4.1.1, and the phenotype differs from those associated with EIF3A LoF variants in our study. Therefore, the clinical significance of this variant remains unclear.

In humans, a different gene coding for a subunit in the eIF3 complex, EIF3F (MIM: 603914), is associated with autosomal-recessive intellectual developmental disorder 67 (MIM: 618295). This EIF3F-related syndrome is characterized by intellectual disability, developmental delay, behavioral abnormalities, hypertonia and/or hypotonia, hearing loss, and short stature.^56,57 Notably, delayed or absent speech development is a common feature within these reported cohorts. Additional findings include mild facial dysmorphisms, microcephaly, and ophthalmological manifestations.⁵⁶ Structural defects such as TOF, cleft lip and palate, and coloboma have also been observed in individuals with EIF3F-related disorder.⁵⁶ The presence of another disease-associated gene encoding an eIF3 subunit and the similar features observed in individuals with EIF3F-related disorder and those reported in our study provide further evidence of the critical role of the eIF3 complex and how disruption of a single subunit can contribute to disease. In addition to EIF3F, other subunits of the eIF complexes are associated with disease; for example, EIF2B1 (MIM: 606686), EIF2B2 (MIM: 606454), EIF2B3 (MIM: 606273), EIF2B4 (MIM: 606687), and EIF2B5 (MIM: 603945) are associated with autosomal-recessive leukodystrophy phenotypes (MIM: 603896, 620312, 620313, 620314, and 620315), highlighting the vital role of proper eIF activity in postnatal brain function.

Our eif3s10 and eif3ba zebrafish mutant models provide further evidence that disruption of human EIF3A or EIF3B can lead to a spectrum of developmental defects, including cardiovascular and craniofacial abnormalities. This study presents a comprehensive eif3s10 zebrafish model, and while a previously published eif3ba mutant (generated via retroviral insertion) exists, we offer a more detailed characterization of the cardiac phenotype.⁵⁸ Our model largely recapitulates previous findings, including microcephaly, coloboma, delayed pigmentation, elongated hearts, and absent craniofacial cartilage. While additional mechanistic studies are required, it is possible that many of these defects stem, at least in part, from disruptions in neural crest cell development.⁵⁸ Neural crest cells contribute to multiple affected structures, including cardiac tissue (myocardium and outflow tract), craniofacial cartilage, melanocytes (pigmentation), and neurons.⁵⁹ Disruptions to cardiac neural crest cells have been implicated in the pathogenesis of several human cardiocraniofacial syndromes and cardiac outflow tract defects, including TOF.^5,6 We acknowledge that our zebrafish models are homozygous mutants, whereas the variants seen in the human cohort are heterozygous. It is a common occurrence in disease modeling for heterozygous variants that cause disease in humans to be tolerated in animal models. This is possibly due to the controlled lab setting, which prevents stressful environmental insults during development and thus hides any mild fitness defects. Additionally, in human individuals, other disease modifiers may be at play, which could further explain the phenotypic range. This phenomenon is exemplified for other CHD genes that are haploinsufficient in humans but require homozygous loss in animal models, including TBX1 (MIM: 602054) (DiGeorge syndrome [DGS] [MIM: 188400]),^60,61 TBX20 (MIM: 606061),^62,63 and GATA4 (MIM: 600576).^64,65 Overall, our findings support the inclusion of EIF3A and EIF3B in genetic testing for CHD and other neurocristopathies.

Interestingly, despite their broad expression and essential cellular functions, EIF3A and EIF3B appear to cause a distinct developmental phenotype when disrupted. This may be due to heightened dosage sensitivity in the affected tissues or the presence of compensatory mechanisms in unaffected tissues. Several other “general” genes with essential cellular roles have been implicated in genetic disorders. For example, mutations in genes responsible for chromosomal structure maintenance cause Cornelia de Lange syndrome (CDLS [MIM: 122470]), which includes microcephaly, distinct facial features, intellectual deficits, and heart defects.⁶⁶ Similarly, mutations in CHD7 (MIM: 608892) (encoding a chromatin remodeler) cause CHARGE syndrome (MIM: 214800),⁶⁷ while mutations in PTPN11 (MIM: 176876) (encoding a protein tyrosine phosphatase involved in cell signaling) lead to Noonan syndrome (NS1 [MIM: 163950]).⁶⁸ Further investigation is required to uncover why certain organ systems, including neural crest-derived tissues, are more susceptible to disruptions in these broadly expressed genes.

Recent studies have highlighted the transcript selectivity of the eIF3 complex, suggesting that it plays a regulatory role at the level of translation. This may further explain why the loss of eIF3 function results in tissue-specific developmental defects. For example, Lee et al. profiled eIF3 binding targets in a human cell line and identified nearly 500 specific mRNA transcripts bound via their 5′ UTRs.⁸ Notably, four of the thirteen eIF3 subunits, including eIF3a and eIF3b, directly crosslinked to RNA. These eIF3-bound transcripts are involved in key biological processes such as cell proliferation, differentiation, and apoptosis. A similar phenomenon was observed in vivo, where eIF3 was found to bind 5′ UTRs, specifically at a 20-nt UC-rich motif.⁶⁹ Gene Ontology (GO) analysis of these transcripts revealed enrichment for pathways related to cell migration, tissue morphogenesis, and differentiation, further supporting that eIF3 exerts some transcript selectivity during translation. Another eIF3 subunit, eIF3d, has been demonstrated to directly bind the 5′ cap of processed mRNA transcripts, providing additional evidence for transcript-specific regulation of translation by the eIF3 complex.⁹ Moreover, eIF3 has been detected at the 3′ UTRs of highly translated transcripts during stem cell differentiation (i.e., at a time of drastic proteome changes), which may indicate another type of translational regulation by eIF3.⁷⁰ Altogether, the results show that transcript selectivity may contribute to the distinct phenotype observed in EIF3A and EIF3B haploinsufficiency. Further studies investigating the broader role of the eIF3 complex in translational regulation during development will help to better understand its pathological mechanism of action.

We have provided evidence for an eIF3-related cardiovascular, craniofacial, and neurodevelopmental disorder. This cohort of fourteen individuals with EIF3B variants and four with EIF3A variants depicts the variable phenotypic spectrum associated with EIF3B- and EIF3A-deficiency. Our targeted zebrafish models recapitulate key phenotypic features observed in affected individuals, further providing evidence that loss of EIF3A or EIF3B contributes to congenital anomalies. These findings underscore the importance of evaluating EIF3A and EIF3B in CHD cohorts. Expanding investigations into other clinical populations would help to better characterize the full phenotypic spectrum of EIF3A and EIF3B haploinsufficiency.

Data and code availability

There are restrictions to the availability of genome sequencing data due to privacy considerations, and institution-specific regulatory requirements. Anonymized data may be available upon reasonable request. Requests should be directed to and will be fulfilled by the lead contacts Dr. Rebekah Jobling (rebekah.jobling@sickkids.ca) and Dr. Ian Scott (ian.scott@sickkids.ca).

Acknowledgments

This research was made possible through access to data in the National Genomic Research Library, which is managed by Genomics England Limited (a wholly owned company of the Department of Health and Social Care). The National Genomic Research Library holds data provided by patients and collected by the NHS as part of their care and data collected as part of their participation in research. The National Genomic Research Library is funded by the National Institute for Health Research and NHS England. The Wellcome Trust, Cancer Research UK, and the Medical Research Council have also funded research infrastructure. We thank all families and their clinicians for partaking in this study. Zebrafish studies were supported by Canadian Institutes of Health Research (CIHR) funding (CIHR PJT 178155) and the Azrieli PCHP Catalyst Program (SickKids/Ste. Justine) provided to I.C.S. E.E. was supported by a CIHR Vanier Scholarship and a SickKids Restracomp Scholarship. We thank the zebrafish technicians at SickKids for animal husbandry. Thank you to Nathan Stutt for assistance with videomicroscopy of zebrafish hearts and Simon Monis for troubleshooting the PyHeart4Fish software. We thank the following labs for sharing reagents: Dr. James Ellis, Dr. Brian Ciruna, Dr. Madeline Hayes, and Dr. James Dowling. Boxplots in Figures 3, 4, and 5 were created with BioRender. Research reported in this publication was supported by the Ted Rogers Centre for Heart Research (R.K.J.), the Plan France Médecine Génomique 2025 (V.P., C.Q., and C.D.), the Nationwide Foundation Innovation Fund (D.C.K.), the ChildCare Foundation (S.E.A.), and the University of Zurich Clinical Research Priority Program (A.R.).

Author contributions

E.E. and C. Somerville wrote the paper, with input from all authors. I.C.S. and R.K.J. provided supervision. L.M. generated the zebrafish CRISPR mutants. I.C.S. and E.E. devised the zebrafish phenotyping experiments, and E.E. performed and analyzed the experiments. A.-S.R. and M.-T.S. assisted with zebrafish experiments. C. Somerville, Q.D., X.C., and R.M. were involved in genome analysis for probands #1 and #15. O.M.M. and R.K.J. provided genetic consultation for probands #1 and #15. M.L.B.S., X.C., O.M.M., R.H.K., I.C.S., R.K.J., and S.E.A. provided valuable direction in the development of the manuscript. All other authors contributed clinical/genetic analysis and provided clinical and genotypic data.

Declaration of interests

S.E.A. and E.R. are cofounders of Medigenome, Swiss Institute of Genomic Medicine. M.J.G.S. is an employee and may own stock in GeneDx.

Published: September 30, 2025

Web resources

• CADD, https://cadd.gs.washington.edu/

• ClinVar, https://www.ncbi.nlm.nih.gov/clinvar/

• Ensembl, https://useast.ensembl.org/index.html

• GenBank, https://www.ncbi.nlm.nih.gov/genbank

• GeneMatcher, https://genematcher.org/

• Gene Ontology, https://geneontology.org/

• Genome Aggregation Database, https://gnomad.broadinstitute.org/

• Genomics England, https://www.genomicsengland.co.uk/

• OMIM, https://omim.org/

• SDB Protocol: Staining and observation of cartilaginous structures in zebrafish complete larvae, https://www.sdbonline.org/sites/2008ShortCourse/Zebrafish_AlcianBlue.pdf

• UCSC Genome Browser, https://genome.ucsc.edu/

• UniProt, https://www.uniprot.org/