Loss-of-function in RBBP5 results in a syndromic neurodevelopmental disorder associated with microcephaly

Abstract

Purpose:

Epigenetic dysregulation has been associated with many inherited disorders. RBBP5 (HGNC:9888) encodes a core member of the protein complex that methylates histone 3 lysine-4 and has not been implicated in human disease.

Methods:

We identify 5 unrelated individuals with de novo heterozygous variants in RBBP5. Three nonsense/frameshift and 2 missense variants were identified in probands with neurodevelopmental symptoms, including global developmental delay, intellectual disability, microcephaly, and short stature. Here, we investigate the pathogenicity of the variants through protein structural analysis and transgenic Drosophila models.

Results:

Both missense p.(T232I) and p.(E296D) variants affect evolutionarily conserved amino acids located at the interface between RBBP5 and the nucleosome. In Drosophila, overexpression analysis identifies partial loss-of-function mechanisms when the variants are expressed using the fly Rbbp5 or human RBBP5 cDNA. Loss of Rbbp5 leads to a reduction in brain size. The human reference or variant transgenes fail to rescue this loss and expression of either missense variant in an Rbbp5 null background results in a less severe microcephaly phenotype than the human reference, indicating both missense variants are partial loss-of-function alleles.

Conclusion:

Haploinsufficiency of RBBP5 observed through de novo null and hypomorphic loss-of-function variants is associated with a syndromic neurodevelopmental disorder.

Introduction

The epigenetic machinery has an essential role in the spatiotemporal regulation of gene expression. One of the main epigenetic regulatory mechanisms is the post-translational modifications of histones, including methylation, acetylation, phosphorylation, and ubiquitylation.¹ These histone modifications allow precise and dynamic regulation of the accessibility of genomic regions to DNA-dependent processes, including transcription, replication, DNA repair, and recombination.² The methylation of histone 3 lysine-4 (H3K4) is an evolutionarily conserved chromatin mark that is typically found in active transcription sites and is considered a marker for gene activation.³ H3K4 methylation is predominantly mediated by the complex of proteins associated with SET1 (COMPASS) protein complex, which includes one of the 6 SET1 domain-containing methyltransferases (KMT2A-F) and 3 other core members, WDR5, ASH2L, and RBBP5, to modulate the catalytic activity of methyltransferases.⁴

The number of Mendelian disorders caused by disruption of epigenetic machinery have greatly expanded in the past decade.⁵ The functional classification divides the genes into 4 groups: writer, eraser, reader, and remodeler.⁵ Various histone methylation writers, including the SET1 domain-containing methyltransferases in COMPASS, have been associated with genetic disorders, such as Kabuki syndrome (OMIM#147920).⁶ However, the nonmethyltransferase core members of COMPASS have not been linked to a human disorder to date. A heterozygous pathogenic variant in KMT2D (HGNC:7133) accounts for 50% to 70% cases with Kabuki syndrome, and 20% to 30% of patients with a clinical diagnosis of Kabuki syndrome have no identified known variant that causes the disease.⁶ It has been hypothesized that a pathogenic mutation in other core members of the COMPASS protein complex could contribute to cases with Kabuki-like phenotypes.⁷

In this study, we report 5 unrelated patients with de novo heterozygous variants in RBBP5. These individuals present with neurodevelopmental features, including intellectual disability, developmental delay, microcephaly, and short stature. We provide evidence to support the pathogenicity of the variants through bioinformatic analysis, protein structure modeling, and functional studies in Drosophila. The difficulty in diagnosing individual 1 led to his enrollment in the Undiagnosed Diseases Network (UDN). The goal of the UDN is to facilitate collaboration between clinicians and researchers to improve diagnosis and care.^8–10 In addition to taking advantage of state-of-the-art phenotyping and genotyping tools, the UDN uses model organisms such as Drosophila melanogaster to perform functional assays on rare genetic variants identified in rare disease patients.^11,12 Fly researchers have generated an extensive transgenic toolkit for Drosophila, allowing for rapid functional characterization of candidate pathological variants.^13–16 Functional validation of genetic variants is a critical step toward confirming diagnosis and efforts to characterize novel disease genes improve diagnostic success of additional patients in the future.^12,17–20 Elucidating the mechanism for rare and undiagnosed diseases leads to improved diagnostic rates, earlier intervention, possible targeted therapeutics, and ultimately improved quality of life.²⁰

Materials and Methods

Identification of individuals

Individual 1 was referred to the UDN at UCLA by a local physician. Other individuals were identified in Gene-Matcher,²¹ and clinical information were collected through collaborators. In addition, a thorough search for candidate RBBP5 variants was conducted in the clinical exome/genome database at the Baylor Genetics Laboratories. All individuals provided written consent for participation of research and publication, including consent to publish patient photos. This study has been approved by the institution review board at UCLA. RBBP5 variants were identified by either exome or genome sequencing using genome build GRCh37/hg19. All individuals except 1 had both biological parents as comparators in the sequencing, and the variants were confirmed to be de novo because of absence in the parental sequences. Sanger sequencing was not performed because of high quality of variant calling.

Structural analysis

Structural analysis of human MLL3-ubNCP complex (PDB: 6KIW) was carried out with PyMOL (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC, https://pymol.org/2/).

Drosophila melanogaster

Fly lines were obtained from the Bloomington Drosophila stock center and the Kyoto Drosophila stock center: daughterless-GAL4(3) (daughterless^GAL4) (Bellen lab), w[*]; P{w[mC]=UAS-mCherry.NLS}3 (BDSC: 38424) (UAS-mCherry.NLS), y[1]w[*]; P{w[+mC]=Act5C-GAL 4}25Fo1/Cyo, y[+] (BDSC:4414) (Actin^GAL4), y[1]w[1118]; P{w[+mC] ey3.5-GAL4.Exel}2 (BDSC: 8220) (eyeless^GAL4), P{w[+mC] GAL4-elav.L}2/Cyo (BDSC: 8765) (elav^GAL4), w[1118]; P{w[+m*] GAL4}repo/TM3, Sb (BDSC: 7415) (repo^GAL4), UAS-lacZ (Bellen lab) w[*];Cyo, P{w[+mC] Tb[1]Cpr[Cyo-A]/sna[Sco] (BDSC: 36335) (Cyo, Tb), w[1118]; Df(3L)BSC447/TM6C, Sb[1] cu[1] (BDSC:24951) (Df (3L)BSC447), y¹ w*; PBac {y^+mDint2 w^+mC UAS-hsp-Hsap\RBBP5.HA.1}VK00037 / CyO, P{ry^+t7.2 sevRas1.V12}FK1 (DGRC: 305207) (RBBP5-HA). Stocks were held at 25°C, and all experiments were carried out at 25°C unless specified.

Rbbp5^{Kozak GAL4} transgenic line generation

The Rbbp5^{Kozak GAL4} was generated as previously described.²² Briefly, the Rbbp5 coding sequence was replaced with a Kozak sequence-GAL4-polyA-FRT-3XP3EGFP-polyA-FRT (KozakGAL4) cassette using CRISPR-mediated homologous recombination. The homology donor intermediate construct is prepared by synthesis of sgRNA targeting the 5′ and 3′ untranslated regions (AAAATGAATTTGGAGCTACTAGG and TTATTTCTTGGTACGTCCGGCGG, respectively) and short (200 bp) homology arms in pUC57_Kan_gw_OK2 vector. The KozakGAL4-polyA-FRT-3XP3EGFP-polyA-FRT cassette is subcloned in the homology donor intermediate to generate the homology donor construct. Homology donor construct is injected (250 ng/μl) in embryos containing Cas9 in their germline (y¹w*; attP40(y+){nos-Cas9(v+)}; iso5) and the resulting G0 progeny are crossed to y¹w* flies and lines are screened for 3XP3-EGFP (enhanced green fluorescent protein)-driven expression of EGFP. The correct integration of the KozakGAL4 cassette in the proper locus is verified by polymerase chain reaction using forward and reverse primers flanking the homology arms and construct specific forward and reverse primers as described in Kanca et al.²² The RBBP5^{Kozak GAL4} transgenic line was submitted to the Bloomington Drosophila Stock Center (BDSC: 97331).

Human RBBP5 and Drosophila Rbbp5 construct generation

Human pcDNA3-FLAG-RBBP5 plasmid used in the protein expression experiments was obtained from Addgene (Cat# 15550). Candidate variants were introduced using Quik-Change II Site-directed mutagenesis kit according to manufacturer’s protocol (Agilent). The mutations were confirmed by Sanger sequencing. For Drosophila experiments, Human cDNA for RBBP5 (NM_005057.4) was obtained from the collection of the late Dr Kenneth Scott at Baylor College of Medicine (clone: IOH28957). Q5 site-directed mutagenesis was completed to create the variant sequence from the reference c.695C>T p.(T232I), c.888A>T p.(E296D). Constructs were transformed using high efficiency E. coli competent cells (New England Biolabs, Cat # C2987H) from the pDONR 221 entry vector to the pGW-HA.attB destination vector using Gateway cloning and the sequences were confirmed by Sanger sequencing. Vectors were injected into embryos to create P{UASt-RBBP5-Ref}VK37, P{UASt-RBBP5-T232I}VK37, and P{UASt-RBBP5-E296D}VK37. For fly cDNA construct generation, Rbbp5 (NM_140952.3) wild type and variant (p.T231I and p.E295D) lines were obtained (clone OFa19095D, GenScript USA, Inc). Constructs were transformed using high efficiency E. coli competent cells from the pGenDONR vector to the Gateway compatible pDONR 223 entry vector and subcloned to the pGW-HA.attB destination vector. Sequences were confirmed using Sanger sequencing, and vectors were injected into embryos to create P{UASt-Rbbp5}VK37, P{UASt-Rbbp5-T231I} VK37, and P{UASt-RBBP5-E295D}VK37. Transgenic males are crossed to y¹ w* stocks, and construct integration is confirmed through selection for the mini white gene.

RBBP5 protein expression and western blot

Flag-tagged wild type and mutated human RBBP5 plasmid were transfected to HEK293 cells by lipofectamine 3000 according to manufacturer’s protocol (Invitrogen). Cells were lysed by RIPA buffer, and 20 mg of whole-cell lysate were used to assess protein expression in western blot. ANTI-FLAG M2 antibody from Sigma and anti-beta-actin from Santa Cruz Biotechnology were used to detect FLAG-tagged RBBP5 and Beta-actin, respectively. For Drosophila experiments, histone extraction was performed (Abcam, Cat# Ab113476) with 5 whole 3rd instar larvae (n = 6 replicates) using the manufacturer’s protocol. Protein (10 μL) was loaded of each sample and RBBP5 (Cell Signaling Technology [CST], Cat#12766), H3K4me3 (CST, Cat# 9733), and histone H3 (CST, Cat# 9715) primary antibodies were used with goat anti-rabbit horse-radish peroxidase and imaged on a Bio-Rad Chemidoc MP imaging system. H3K4me3 normalization to total H3 was completed using ImageJ.

RBBP5/Rbbp5 overexpression viability and morphology evaluation in Drosophila

Heterozygous or homozygous (in the case of larval and pupal lethal crosses) male UAS-cDNA stocks were crossed to female GAL4 ubiquitous (Actin^GAL4 and daughterless^GAL4) and tissue-specific (eyeless^GAL4, elav^GAL4, and repo^GAL4) lines. The resulting progeny were counted by genotype (n > 50). Viability was calculated by comparing the observed number progeny with the expected (o/e ratio) based on Mendelian inheritance patterns. A normalized o/e ratio greater than 0.8 is defined as viable, semi-lethality is 0.8 to 0.15, and lethality is less than 0.15, and the latest developmental stage observed is reported as the lethal point. For da^GAL4 overexpression, only latest developmental stage reached is reported because the da^GAL4 stock is not balanced. Overexpression-based phenotypes are scored in greater than 5 larvae or adult flies.

RBBP5 human cDNA rescue evaluation

In human cDNA rescue experiments, Rbbp5^{Kozak GAL4} lethality is reported as the latest developmental stage reached. Nonbalanced dead pupae (n = 19/341) were observed in self-crosses of Rbbp5^{Kozak GAL4}/TM6B,Sb,Tb stocks, and dead larvae were also observed in the crosses. Therefore, there is the possibility that Rbbp5^{Kozak GAL4} is lethal at the embryo, larval, and pupal stages; however, for the purposes of this study, rescue was evaluated as the ability of the human cDNA to rescue to the adult developmental stage in (n > 40) progeny.

Drosophila developmental staging

The following characteristics were selected to stage animals: early L3 larvae have branched but not extruding spiracles and late L3 larvae have extruding spiracles, visible gut clearance, and exhibit wandering behavior.

Brain immunostaining and brain lobe quantification

For larval counterselection, homozygous da^GAL4 lines were used and Actin^GAL4 was crossed to Cyo, Tb to create Actin-GAL4/Cyo,Tb. These lines were then crossed to homozygous RBBP5^Ref, RBBP5^T232I and RBBP5^E296D stocks. Third instar larvae (gut clearance, branched spiracles, and wandering behavior) were dissected in ice-cold PBS. Larval brain preps were fixed in 4% PFA/PBS/2% Triton over-night. Brains were blocked in PBS/2%Triton/5% normal donkey serum for 1 hour. Primary antibodies were incubated overnight; rat anti-Deadpan (Abcam, Cat# ab195173, 1:250) and mouse anti-Prospero (Developmental Studies Hybridoma Bank, Cat# MR1A, 1:1000). Secondary antibodies were incubated for 2 hours at room temperature; rat anti-GFP and mouse anti-Cy3 (1:250). Brains were mounted and imaged using a Zeiss 710 confocal microscope (Neurovisualization core, Baylor College of Medicine) with 1 mm sections. The area of 1 brain lobe was quantified using the area tool in Image J.

Eye area quantification

Whole heads were imaged using a Leica KL1500 LCD microscope using 10× magnification. The area of one eye (n = 5) was quantified using the area tool in ImageJ.

Statistical analysis

Statistical analysis was completed using GraphPad prism (Version 9.0.0). Continuous analysis was completed by ordinary one-way ANOVA in which differences between groups were quantified and a P value less than .05 is considered significant.

Results

Identification of individuals and characterization of clinical features

Five unrelated individuals were included in this study. Individual 1 was enrolled in the UDN, subsequently individuals 2 to 5 were identified in unrelated families through GeneMatcher.²¹ The 5 individuals presented with neurodevelopmental features, including developmental delay, intellectual disability, and microcephaly. In addition, short stature, musculoskeletal abnormalities, and dysmorphic facies were the common phenotypes observed in this cohort of individuals (Figure 1). Four out of 5 patients were reported to have short stature and microcephaly. Intellectual disability and global developmental delay were seen in all individuals except the youngest. Sensorineural hearing loss and seizure were reported in 2 different individuals. Abnormalities in fingers and toes were observed in all 5 individuals. All individuals presented with dysmorphic facies, but these dysmorphic features did not exhibit as a recognizable pattern. Clinical features of the individuals are summarized in Table 1.

Figure 1. Human subjects with RBBP5 de novo variants exhibit a range of clinical features.

Dysmorphic features in individual 1, including hypertelorism, high arched eyebrow, long eyelashes, synophrys, and board nasal tip as shown in (A); retrognathia, large ear, and a preauricular ear tag as shown in (B); bilateral 5th finger clinodactyly and prominent fingertip pad in (C). Dysmorphic features in individual 3 as shown in (D) with midface hypoplasia and cupped ears, clinodacyly in (E), and supernumerary teeth in (F). Dysmorphic features in individual 4 as shown in (G) with short and upslanting palpebral fissures, high forehead, anteverted nostrils, and sparse eyebrows. (H) and (I) shows the dysmorphic facial features, including sparse eyebrows, short nose, long philtrum, small and squared ears, and small mouth with thin lips, in individual 5. Common phenotypes are illustrated in (J).

Table 1.

Summary of clinical features in affected individuals

	Individual 1	Individual 2	Individual 3	Individual 4	Individual 5
Age (year)	16	16	10	5	3
Sex	Male	Female	Male	Female	Male
Variant	c.695C>T p.(T232I)	c.762G>A p.(W254*)	c.888A>T p.(E296D)	c.919C>T p.(R307*)	c.729del p.(K244Nfs*6)
Inheritance	De novo	Unknowna	De novo	De novo	De novo
Gestational age	40 week	40 week	38 week	30 week	37 week
Abnormal prenatal findings	Small for gestational age	None	Echogenic intracardiac focus	Premature birth due to maternal complicationsb	Left club foot
Growth parameters	Height: 152 cm, −2.4 SD Weight: 38.6 kg, −2.7 SD HC: 53.5 cm, −1.0 SD	Height: 132 cm, −4.8SD Weight: 23.9 kg, −10.8 SD HC: 47.5 cm, −4.8 SD	Height: 117 cm, −3.0 SD Weight: 27 kg, −1.3 SD HC: 50 cm, −2.1 SD	Height: 107.5 cm, −0.7 SD Weight: 16 kg, −1.1 SD HC: 47 cm, −2.8 SD	Height: 66.5 cm, −2.0 SD Weight: 6.5 kg, −2.0 SD HC: 44.5 cm, −2.0 SD
Failure to thrive	Yes	Yes	No	No	Yes
Developmental delay	Yes	Yes	Yes	Yes	No
Intellectual disability	Yes, severe	Yes, severe	Yes, severe	Yes, mild	No
Dysmorphic features	Hypertelorism, high arched eyebrows, long eyelashes, synophrys, broad nasal tip, and retrognathia, right ear tag	Enophthalmia, long columella, mild retrognathia, right face fuller than left, microdontia	Plagiocephaly, dolichocephaly, midface hypoplasia, cupped ears, large and supernumerary teeth	Enophthalmia, short and upslanting palpebral fissures, high forehead, attached earlobe, anteverted nostrils, sparse eyebrows	Dolichocephaly, hypotelorism, sparse eyebrows, short nose, long philtrum, small and squared ears with convoluted helix, small mouth with thin lips, retrognathia
Auditory	Bilateral SNHL	Bilateral SNHL	No	No	No
Neurological	Hypotonia, dysautonomia, reduced sensation to pain and temperature	Seizure, dystonia	Generalized tonic-clonic seizures from 9.5 years of age	Hypotonia	Hypotonia
Cardiac	Two SVC	No	No	No	No
Gastrointestinal	Chronic vomiting and constipation, GERD	No	No	Anal stenosis	No
Urogenital	Cryptorchidism	No	Hypospadias	No	No
Immunological	Recurrent infection, low IgG	No	No	No	No
Musculoskeletal	Clinodactyly, prominent fingertip pad, nail hypoplasia, hyperextensible joints, delayed bone age	2nd and 3rd toe syndactyly, left pes planus	Clinodactyly brachydactyly, short 4th toe bilateral, sandal gap, pes planus, forefoot adduction	Clinodactyly	Clinodactyly, nail hypoplasia, left club foot, and hip dysplasia
Skin	Cutis marmorata	No	No	No	No
Behavior or mental	Autism, ADHD	No	Autism	N/A	No
Other clinical features	Lacrimal duct stenosis, gluteal cleft	Retinal dystrophy, low visual acuity, night blindness		Prematurity, necrotizing enterocolitis	Gluteal cleft

ADHD, attention-deficit/hyperactivity disorder; GERD, gastroesophageal reflux disease; HC, head circumference; SNHL, sensorineural hearing loss; SVC, superior vena cava.

^aThe variant was not maternally inherited in the duo exome sequencing. Paternal sample was not available.

^bMaternal complications included pulmonary arterial hypertension, preeclampsia, and gestational diabetes mellitus.

Identification and analysis of variants

All 5 de novo heterozygous variants in RBBP5 (NM_005057.4) were identified through trio exome or genome sequencing, including 2 missense variants p.(T232I) and p.(E296D) and 3 nonsense/frameshifting variants c.762G>A p.(W254*), c.729del p.(K244Nfs*6), and c.919C>T p.(R307*). RBBP5 has a pLI score of 1 and missense z-score of 4.64,^23,24 suggesting intolerance to haploinsufficiency and missense changes, respectively. In addition, all 5 variants had not been observed in the gno-mAD database (v4.1.0).² In silico variant analysis results were inconsistent in their pathogenicity predictions for the missense variants (Supplemental Table 1). Interestingly, all 5 variants are located in a small region between WD40 repeat domains 4 and 6 in RBBP5. This region is predicted to be intolerant to changes based on MetaDome analysis, indicating a potential hotspot for pathogenic variants (Figure 2A).²⁵ The missense variants affect amino acids that are well-conserved across species from human to Drosophila (Figure 2B). Expression of the RBBP5 protein with missense variants in HEK293T cells resulted in the full-length protein expressed at a similar level compared with that with a wild-type construct based on western blotting (Figure 2C). However, the single-nucleotide deletion at nucleotide 729 in RBBP5 p.(K244Nfs*6) produced a frameshift product that terminates prematurely and showed no detectable protein expression. Similarly, the C>T change at nucleotide 919 led to premature termination without protein expression (Figure 2C) confirming that these are complete loss-of-function (null) alleles. In structural analysis of the missense variants, we found that the threonine 232 residue is located in one of the WD40 repeats that mediates the interaction between RBBP5 and the ubiquitinated histone 2B lysine-120 (H2BK120), which has been known to promote the catalytic activity of methyltransferase in H3K4.²⁶ The p.(T232I) missense variant is predicted to alter the interaction between RBBP5 and the ubiquitinated H2BK120 (Figure 2D and E). Similarly, the E296 residue is located at a crucial position in the loop 2 of RBBP5 WD40 domain, which has been shown to mediate the direct interaction between RBBP5 and the nucleosome.²⁶ The p.(E296D) variant is expected to affect the conformation of loop 2, which could interfere the binding of RBBP5 to the histones (Figure 2F). Because the missense p.(T232I) and p.(E296D) variants do not affect the protein expression of RBBP5 and structural analysis alone is insufficient to prove the pathogenicity of missense variants, we developed transgenic Drosophila models to investigate the mechanism of these variants in vivo.

Figure 2. Bioinformatic and structural variant analysis.

The functional domains of RBBP5 and position of variants in MetaDome are shown in (A). The evolutionarily conserved residues affected by variants are shown in (B). The protein expression of FLAG-tagged human RBBP5 reference and variants were shown in (C). Structure analysis of T232 and E296 were performed in RBBP5^WD40 of the cryo-EM structure of MLL3-ubNCP complex. The overall structure of RBBP5^WD40 complexed with a nucleosome core particle mono-ubiquitinated at the Lys 120 of histone H2B (ubNCP). The RBBP5 WD40 repeat 4 is sandwiched between ubiquitin and core histones. The RBBP5^WD40 is shown in orange and ubiquitin in blue (D). Detailed view of the recognition interface of RBBP5^WD40-ubiquitin. Residue p.(T232I), which is located on the α-helix-containing loop of RBBP5^WD40 blade 5, lies close to residues L8, T9, and H68 of ubiquitin (E). All these residues are shown in stick model. Detailed view of the interaction interface between RBBP5^WD40 and histone H2B-H4. Two loops (loop 1 and loop 2), which connect the WD40 propeller blades 5, 6, and 7, interact with nucleosome directly. Residue p.(E296D) is located on loop 2 and shown in stick model (F).

Overexpression of human RBBP5 in Drosophila results in microcephaly

For this study, we performed overexpression and rescue experiments with the human RBBP5 cDNA to determine if the variants have a functional consequence in vivo. The human RBBP5 is orthologous to Rbbp5 in Drosophila with a high-sequence identity (340/505 amino acids, 67%) and similarity (398/505, 78%) (DIOPT score: 14/16).²⁷ In flies, Rbbp5 is a member of the trithorax complex, which is required for differentiation of neural lineages.^28,29 Importantly, neuronal fate determination is similar between humans and flies.³⁰ Previous research has reported complete loss of H3K4me3 in Rbbp5 mutant clones.²⁹ The resulting Rbbp5 loss-of-function phenotype includes an inability to maintain type II neuroblast identity that gives rise to intermediate progenitor cells. Rbbp5 null neuroblasts instead express markers for type I neuroblasts.²⁹ Intermediate progenitor cells undergo several rounds of self-renewal before being committed to a ganglion mother cell that will terminally differentiate into 2 neurons. Critically, type I neuroblasts are not capable of generating intermediate progenitor cells. This study also demonstrated that neuronal identity and H3K4me3 levels can be restored by the expression of the full-length Rbbp5.²⁹ A shift from type II to type I neuroblast identity could result in a decrease in the overall number of neurons inducing a microcephaly phenotype in flies, similar to the phenotype of the individuals included in this study.

We first determined whether overexpression of the human reference RBBP5 cDNA (NM_005057.4) (RBBP5^Ref) or variants p.(T232I) (RBBP5^T232I) and p.(E296D) (RBBP5^E296D) by ubiquitous or tissue-specific GAL4 drivers cause phenotypes in the fly. In the GAL4-UAS system, a GAL4 transcriptional activator protein is fused to the promoter of a gene of interest. When paired with a construct containing an upstream activation sequence (UAS), this drives expression of the construct based on the spatial and temporal expression pattern of the gene.³¹ All UAS-RBBP5/Rbbp5 lines used in this study are integrated into the same location in the genome to avoid a position effect on expression level.³² In human protein overexpression paradigms, the endogenous Drosophila gene is unaffected, thus enabling the detection of gain- or loss-of-function mechanisms. Gain-of-function variant phenotypes would appear more severe than wild type and complete loss-of-function variants would appear wild type because the fly gene is expressed in the genetic background.³³ Gain-of-function alleles include hyper-, anti-, and neo-morphic alleles in which the variant results in overactive wild-type function or induces dominant negative or novel function in the protein, respectively, as described in previous studies.^34–36 Loss-of-function alleles include hypo- and a-morphic mechanisms that result in a partial or complete loss of protein function, which has been well described in literature.^37–40

UDN MOSC investigation of candidate variants in the fly has yielded the identification of 59 novel disease genes during phase I of investigation (September 2015-August 2018).²⁰ We have reported numerous instances in which the overexpression of the human protein results in phenotypes (ie, small eye and wing patterning defects) and confirmed loss-of-function mechanisms as the failure of the variant cDNA to induce the same phenotypes and observed partial loss-of-function as a milder yet still apparent morphological defect.^33,41,42 We have furthermore identified “complex phenotypes” when overexpressing neurodevelopmental risk genes in the fly in which a gain-of-function mechanism is observed in one tissue and a loss-of-function mechanism is observed in another, highlighting the importance of thorough investigation of variant mechanisms.³³ RBBP5 is located at 1q32.1 and 1q duplication has been identified as a driver of tumorigenesis in breast cancer.⁴³ Another study identified that RBBP5 has both high haploinsufficiency and triploinsufficiency scores, which supports a dosage-sensitive mechanism for RBBP5.⁴⁴ As such, any deviation from wild-type Rbbp5 (as in the fly) or reference RBBP5 (as in the human) gene function could be deleterious.

We overexpressed RBBP5^Ref, RBBP5^T232I, or RBBP5^E296D using Actin- (Act-, strong ubiquitous), daughterless- (da-, weak ubiquitous) eyeless- (ey-, developing visual system and parts of the head and brain), elav- (neuron), and repo- (glia) GAL4 lines for tissue-specific expression. We then scored lethality by developmental stage in which the Drosophila life cycle is 10 days long at 25 °C. Overexpression of RBBP5^Ref or either missense variant with Actin^GAL4 is lethal in the third instar larval (L3) stage (Figure 3A). Flies undergo embryo, larval, pupal, and adult stages, and larval development is marked by progression through L1-L3 stages, with L3 spanning 3 days before pupation.⁴⁵ Expression of RBBP5^Ref with da^GAL4 is again L3 lethal; however, expression of RBBP5^T232I or RBBP5^E296D is pupal lethal (Figure 3A). We observe no decrease in viability with ey-, elav-, and repo-GAL4 drivers (Figure 3A). Both Actin and daughterless are expressed early in development, and these data support previous findings that Rbbp5 has a critical role early in development. Disruption later in development with elav- or repo^GAL4 induces no phenotype because critical functions have already been carried out when neuronal and glial specific genes are expressed. Importantly, because the fly Rbbp5 is being expressed in the genetic background in these human cDNA overexpression experiments, observing a loss of larval lethality that is induced by overexpression of the RBBP5^Ref suggests that both RBBP5^T232I and RBBP5^E296D are partial loss-of-function alleles.

Figure 3. Ubiquitous expression of RBBP5 in flies induces lethality and results in microcephaly in the larval developmental stage.

The Drosophila life cycle is approximately 10 days long at 25°C. Expression of the human RBBP5 or either missense variant with a strong ubiquitous driver (Actin^GAL4) is larval lethal (Actin^GAL4 / RBBP5^Ref observed in 0/128 F1 progeny, o/e 0.0; Actin^GAL4/RBBP5^T232I observed in 0/144 F1 progeny, o/e 0.0; Actin^GAL4/RBBP5^E296D observed in 0/277, o/e 0.0). With a weak ubiquitous driver (da^GAL4), expression of the human reference is larval lethal, but expression of p.(T232I) or p.(E296D) is pupal lethal. Expression with tissue specific (ey-, elav-, or repo^GAL4) drivers does not affect viability in (A). Representative developmentally staged control late L3 brains (UAS-lacZ; da^GAL4) with Deadpan staining of neuroblasts and intermediate progenitor cells in green and Prospero staining of neural progenitors in red in (B). Experimental RBBP5 reference (RBBP5^Ref; da^GAL4) brain shown in (C), RBBP5^T232I (RBBP5^T232I; da^GAL4) in (D) and RBBP5^E296D (RBBP5^E296D; da^GAL4) in (E). Quantification of ubiquitous overexpression (da^GAL4) of RBBP5^Ref, RBBP5^T232I, and RBBP5^E296D compared with UAS-lacZ (one-way ANOVA, ns, P > .05, *P < .05, **P < .01, ***P < .001, ****P < .0001) in (F). Created with Biorender.com.

We investigated brain development during the larval stage because daughterless overexpression of RBBP5^Ref is L3 lethal, whereas the expression of RBBP5^T232I or RBBP5^E296D is pupal lethal. We dissected L3 brains using da^GAL4 to express a neutral cDNA (da^GAL4; UAS-lacZ) as an overexpression control and immunostained for markers of progenitor lineages, Deadpan (neuroblasts and intermediate progenitor cells) and Prospero (differentiating neurons)²⁹ (Figure 3B). Deadpan-positive intermediate progenitor cells are present in the optic lobes of control larvae; however, when RBBP5^Ref is expressed, L3 stage-matched brains are severely reduced in size and appear approximately the size of L2 brains with a loss of intermediate progenitor cells in the optic lobes of the brain (Figure 3C and F). Interestingly, although overexpression of either missense variant is pupal lethal, RBBP5^T232I and RBBP5^E296D L3 brains also fail to develop to control lobe size and are not significantly larger than RBBP5^Ref brains (Figure 3D-F) despite the observed difference in lethality staging. These results indicate a discordance between the developmental stage reached (pupa) with the developmental stage of the brain (late L2-early L3) for variant expressing animals and suggest that RBBP5 could have a pleiotropic effect on growth.

We therefore dissected L3 brains using the strong ubiquitous driver Actin^GAL4, which induces L3 lethality with RBBP5^Ref, RBBP5^T232I, or RBBP5^E296D. We again performed immunostainings to label neural progenitor lineages with Deadpan and Prospero antibodies.⁴⁶ Early and late L3 control (Actin^GAL4; UAS-lacZ) larvae were used to identify neural development occurring throughout the L3 stage (Supplemental Figure 1A and B). Late L3 brains that express RBBP5^Ref are less developed in size compared with control stage-matched wandering L3 larva (Supplemental Figure 1C, F). Late L3 brains that express RBBP5^T232I or RBBP5^E296D also display a small brain (Supplemental Figure 1D-F). Ubiquitous expression of RBBP5 leads to dramatic reduction in brain size with no significant differences in size between RBBP5^Ref, RBBP5^T232I, or RBBP5^E296D (Supplemental Figure 1F). Overall, these data suggest that the expression of the human RBBP5 interrupts the function of the fly Rbbp5 protein and results in a strong microcephaly phenotype in the fly.

We also observed changes in overall larval size and development upon ubiquitous overexpression of RBBP5. Larval developmental progression is a highly stereotyped pattern and a failure to reach developmental stages indicates possible dysregulation of factors that direct development.⁴⁷ Late L3 larvae ubiquitously expressing RBBP5^Ref, RBBP5^T232I, or RBBP5^E296D with Actin^GAL4 exhibit reduced overall body size compared with control larvae expressing UAS-lacZ (Supplemental Figure 2A). Control wandering L3 (Supplemental Figure 2B) develop mature anterior and posterior spiracles indicating that the late L3 stage has been reached.⁴⁸ Severe growth phenotypes are seen upon expression of the human reference cDNA. Expression of either RBBP5^Ref (Supplemental Figure 2C) or RBBP5^T232I (Supplemental Figure 2D) results mature posterior spiracle formation but failure of the posterior spiracles to develop. In RBBP5^E296D-expressing larvae, however, anterior and posterior spiracles successfully develop similar to controls (Supplemental Figure 2E). These results indicate an inability of p.(E296D) to induce the developmental phenotype seen in reference expressing larvae again, suggesting a partial loss-of-function mechanism for p.(E296D). To investigate the effect of our RBBP5 variants on trimethylation in the L3 developmental stage, we confirmed RBBP5 expression with Actin^GAL4 and quantified H3K4me3 compared with total H3 (Supplemental Figure 2F and G). A significant reduction in H3K4me3 is observed compared with UAS-lacZ control larvae, whereas no significant change in RBBP5 protein level is observed between RBBP5^Ref, RBBP5^T232I, or RBBP5^E296D (Supplemental Figure 2G). These data confirm that the human RBBP5 can interact with the fly trithorax complex members and that expression of human alleles can affect H3K4me3 levels in the fly. These results also suggest that dysregulation of Rbbp5 results in the failure to express developmental genes that are critical for the progression through the L3 developmental stage.

Tissue-specific RBBP5 expression in Drosophila results in a small eye phenotype

To assess RBBP5 function in the eye, we drove the variants using the eyeless^GAL4 and quantified eye size in the adult stage. The eyes of RBBP5^Ref or RBBP5^E296D flies are smaller than UAS-lacZ or RBBP5^T232I (Figure 4A-E). These data suggest that in the eye, there is a toxic effect of over-expression of RBBP5^Ref, resulting in a small eye phenotype, and RBBP5^T232I fails to induce this. Indeed, eye size for RBBP5^T232I is not significantly smaller than that of the overexpression control (Figure 4E). These results indicate a complete loss-of-function mechanism for p.(T232I) when expressed in a tissue-specific manner in the eye. Overall, these experiments conclude that overexpression of the human RBBP5 is toxic and results in growth phenotypes, including microcephaly in the brain and reduced size in the body and eye. To summarize these data, we observe both partial and complete loss-of-function mechanisms using overexpression of the human p.(T232I) and p.(E396D) compared with the RBBP5 reference cDNA. We found partial loss-of-function using ubiquitous overexpression for both missense variants and a complete loss-of-function for p.(T232I) using tissue-specific expression in the eye, suggesting that p.(T232I) could be a stronger hypomorphic allele.

Figure 4. Human RBBP5 expression induces a small eye phenotype not recapitulated by the p.(T232I) variant.

Overexpression with eyeless-GAL4 (ey^GAL4) in a control line (ey^GAL4 / UAS-lacZ) as shown in (A), ey^GAL4 / RBBP5^Ref in (B), ey^GAL4 / RBBP5^T232I in (C), and ey^GAL4 / RBBP5^E296D in (D). Expression of RBBP5^Ref and RBBP5^E296D results in a small eye phenotype compared with UAS-lacZ. Expression of RBBP5^T232I does not induce a small eye phenotype and eye size is not significantly different than controls (one-way ANOVA, ns P > .05, *P < .05, **P < .01, ***P < .001, ****P < .0001) in (E).

Overexpression of the fly Rbbp5 induces wing patterning defects

Next, we compared human and fly cDNA overexpression. We generated the orthologous fly Rbbp5 constructs and created the p.(T232I) and p.(E296D) homologous variants, p.(T231I) (Rbbp5^T231I) and p.(E295D) (Rbbp5^E295D), respectively. We compared this with an HA-tagged version of RBBP5 (RBBP5-HA). Ubiquitous expression of Rbbp5, Rbbp5^T231I, or Rbbp5^E295D with Actin^GAL4 or da^GAL4 is viable, but Actin^GAL4 induces ectopic wing vein formation that is not fully penetrant at 25°C (Supplemental Figure 3A). Because the GAL4-UAS system is temperature dependent, we increased the culture temperature, and upon expression of Rbbp5, ectopic wing vein formation is fully penetrant at 29°C, but the phenotypes of Rbbp5^T231I or Rbbp5^E295D are not (Supplemental Figure 3B-E). Expression of RBBP5-HA is pupal lethal with Actin^GAL4 and viable with da^GAL4, but ectopic wing vein formation is present (Supplemental Figure 3A). Ectopic wing vein formation is observed with both da^GAL4 and Actin^GAL4 compared with the laboratory control strain Canton S; however, Actin^GAL4 escaper flies are rarely observed (Supplemental Figure 3F-H). The ectopic wing vein formation seen upon overexpression of RBBP5-HA is similar to the phenotypes observed with Rbbp5, suggesting a disruption of factors that direct wing development when the fly or human cDNA is overexpressed. Furthermore, expression of the wild-type fly cDNA induces fully penetrant wing patterning defects, whereas either missense variant p.(T231I) or p.(E295D) results in wing phenotypes that are not fully penetrant, again indicating a partial loss-of-function mechanism in the context of the fly variants.

Rbbp5 is expressed in neurons and glia in the fly brain

To determine the Rbbp5 expression pattern and further explore the function of the variants, we replaced the open reading frame of Rbbp5 with the Kozak GAL4 sequence (Kozak sequence-GAL4-polyA-FRT-3XP3EGFP-polyA-FRT).³⁰ This effectively removes Rbbp5 and leads to GAL4 expression in a similar spatial and temporal expression pattern (hereafter Rbbp5^{Kozak GAL4}). The GAL4 transcriptional activator protein will bind UAS-containing constructs to drive expression of a reporter protein or human cDNA to determine if the human protein can rescue loss of the Drosophila protein and whether that ability is impaired by the variant.³⁰ We used the Rbbp5^{Kozak GAL4} allele to determine the expression pattern of Rbbp5 in the developing nervous system. We crossed the Rbbp5^{Kozak GAL4} to a UAS-mCherry NLS reporter line and determined that Rbbp5 is expressed in a subset of Elav-positive neurons (Figure 5A-C) and Repo-positive glia (Figure 5D-F) in the optic lobes and ventral nerve cord in the larval brain. Confirming Rbbp5 expression in both neurons and glia supports the canonical function of Rbbp5 to direct neuronal fate because type II neuroblasts give rise to both neurons and glia.

Figure 5. Rbbp5 is expressed in a subset of neurons and glia in the Drosophila brain.

Rbbp5^{Kozak GAL4}/UAS mCherry.NLS expression pattern in the dorsal larval brain shown in (A). Elav expression in (B), and merge in (C) with co-localization in the ventral nerve cord and optic lobes of the central brain. Rbbp5^{Kozak GAL4}/UAS mCherry.NLS expression pattern in the ventral larval brain shown in (D). Repo expression in (E), and merge in (F) with co-localization in the ventral nerve cord and optic lobes.

RBBP5 missense variants induce a less severe microcephaly phenotype than the human reference cDNA in rescue experiments in an Rbbp5 null genetic background

Next, we assessed the phenotypes associated with loss of Rbbp5 function. We dissected L3 brains of Rbbp5^{Kozak GAL4} heterozygous mutant animals and a laboratory control strain (y¹ w*) and again immunostained for markers of progenitor lineages, Deadpan and Prospero.³³ Deadpan-positive intermediate progenitor cells are present in the optic lobes Rbbp5^{Kozak GAL4} heterozygous animals (Figure 6A), and there is no difference in brain size between Rbbp5^{Kozak GAL4} /+ and y¹ w* controls (Figure 6F). However, in Rbbp5 null animals (Rbbp5^{Kozak GAL4} / Df(3L)BSC447) (Df(3L)BSC447 is a 125 kB deficiency that encompasses the Rbbp5 locus), loss of Rbbp5 is pupal lethal. Furthermore, development of intermediate progenitor cells is severely impaired, and brain size is reduced in the L3 developmental stage (Figure 6B and F). When we attempt to rescue this microcephaly phenotype with Rbbp5^{Kozak GAL4}-driven expression of RBBP5^Ref, it fails to rescue the loss of the fly Rbbp5, moreover there is a further reduction in brain size and more severely affected intermediate progenitor cell population (Figure 6C and F). Therefore the RBBP5^Ref transgene not only fails to rescue but also exacerbates the loss-of-function phenotypes. Next, we attempted the rescue experiment with RBBP5^T232I (Figure 6D) or RBBP5^E296D (Figure 6E), and these transgenes also do not rescue, but they do not produce the more severe phenotypes observed with RBBP5^Ref (Figure 6C). These results indicate that p.(T232I) or p.(E296D) cannot rescue loss of Rbbp5 but impair development less severely than the reference cDNA (Figure 6F). This again indicates that overexpression of the RBBP5^Ref is toxic and that RBBP5^T232I and RBBP5^E296D are less toxic even when expressed under the control of endogenous Rbbp5 promoter. In summary, total brain lobe area was reduced in the Rbbp5 mutants and was not rescued by the human reference nor the p.(T232I) or p.(E296D) transgenes. However, RBBP5^Ref produces even stronger lobe size reduction than Rbbp5 null larvae. Furthermore, there is no significant difference in brain lobe area between either RBBP5^T232I or RBBP5^E296D in a null genetic background and RBBP5 null lobe size, again supporting the finding that p.(T232I) and p.(E296D) are loss-of-function alleles (Figure 6F). This is consistent with the phenotypes observed in overexpression experiments with RBBP5^T232I and RBBP5^E296D failing to induce the toxic effects of RBBP5^Ref. Ubiquitous expression of RBBP5^Ref induces earlier lethality than expression of RBBP5^T232I or RBBP5^E296D, and impaired growth phenotypes are present upon expression of the human RBBP5 reference that are not present when the missense variants are expressed indicating a hypomorphic loss-of-function mechanism for the p.(T232I) and p.(E296D) variants. Experimental findings supporting a loss-of-function mechanism for p.(T232I) and p.(E296D) are summarized in Supplemental Table 2.

Figure 6. RBBP5 human transgenes fail to rescue loss of Drosophila Rbbp5.

Heterozygous L3 Rbbp5 loss-of-function controls (Rbbp5^{Kozak GAL4}/+) with Deadpan-positive neuroblasts and intermediate progenitor cells in green and Prospero-positive neural progenitor cells in red in (A). Homozygous loss-of-function with Rbbp5^{Kozak GAL4} crossed to a deficiency line Df(3L)BSC447 that includes the Rbbp5 locus (Rbbp5^{Kozak GAL4} /Df(3L)BSC447) in (B). Attempted rescue with the human RBBP5^Ref (RBBP5^Ref; Rbbp5^{Kozak GAL4}/Df(3L)BSC447) in (C), RBBP5^T232I (RBBP5^T232I; Rbbp5^{Koza kGAL4}/Df(3L)BSC447) in (D), RBBP5^E296D (RBBP5^E296D; Rbbp5^{Kozak GAL4}/Df(3L)BSC447) in (E). Quantification of the microcephaly phenotype (brain area) by genotype (one-way ANOVA, ns P > .05, *P < .05, **P < .01, ***P < .001, ****P < .0001) in (F). RBBP5^Ref, RBBP5^T232I, and RBBP5^E296D fail to rescue loss of the Drosophila Rbbp5. RBBP5^Ref expression induces a significantly more severe microcephaly phenotype than Rbbp5^{Kozak GAL4}/Df(3L)BSC447, and brain size of Rbbp5^T232I; Rbbp5^KozakGAL4/Df(3L)BSC447 or Rbbp5^E296D; Rbbp5^KozakGAL4/Df(3L)BSC447 larvae is not significantly different than Rbbp5^KozakGAL4/Df(3L)BSC447.

Discussion

In this study, we identified 5 affected individuals with de novo heterozygous variants in one of the COMPASS core members RBBP5. We propose that haploinsufficient loss of RBBP5 is responsible for the neurodevelopmental disorder presented here. This is consistent with the fact that RBBP5 has a pLI score of 1, and other disorders of epigenetic machinery classically are due to haploinsufficiency.⁵ We confirm that the frameshifting/nonsense variants are loss-of-function null alleles. Additionally, the p.(T232I) and p.(E296D) missense variants also meet the criteria for a pathogenic variant based on the variant interpretation guidelines from a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.⁴⁹ These criteria include confirmed de novo variants, in vitro or in vivo functional studies supportive of a damaging effect, the absence of these variants in the population database, and the low rate of benign missense variation in RBBP5 (Supplemental Table 1). We have also conducted a search in a large clinical exome/genome database at Baylor Genetics for candidate RBBP5 variants, which identified 2 likely benign de novo variants. These 2 cases were considered as likely benign because of inconsistent phenotype and benign in silico analysis (Supplemental Tables 1 and 3).

The COMPASS protein complex consists of 4 core members: RBBP5, WDR5, ASH2L, and 1 of the 6 methyltransferases.⁵⁰ The methyltransferases have the enzymatic SET1 domain to methylate H3K4, whereas RBBP5 functions to modulate the activity of the complex and mediate the interaction between the nucleosome and the complex.^4,51 Our structural analysis showed both the T232 and E296 residues are located at critical positions of the interface between RBBP5 and the histones, which has been known to be involved in the recruitment of COMPASS to the nucleosome and promoting the methyltransferase activity in H3K4.^26,52,53 The mutations in T232 and E296 are likely to interrupt the interaction between RBBP5 and histones, resulting in the dysregulation of downstream target genes.

Kabuki syndrome is one of the most common disorders in the epigenetic machinery.⁶ KMT2D, one of the methyltransferases in the COMPASS, is the major disease gene for the Kabuki syndrome. Nevertheless, about 20% to 30% of clinically diagnosed Kabuki syndrome patients have negative genetic testing.⁶ It has been hypothesized that a pathogenic variant in other COMPASS members could result in a disorder that phenotypically resembles Kabuki syndrome.⁷ We observe some striking similarities in phenotypes between our probands and those typically seen in patients with Kabuki syndrome, such as the neurodevelopmental features, microcephaly, short stature, hypotonia, sensorineural hearing loss, and seizure.⁵⁴ However, there are phenotypes such as congenital cardiac defects and the characteristic dysmorphic facial features in the Kabuki syndrome that are absent in our probands. Given KMT2D is only 1 of the 6 methyltransferases to which RBBP5 binds, it is reasonable to expect differences in the clinical spectrum between KMT2D- and RBBP5-related disorders.

The Drosophila model has previously been used to confirm the functional mechanism of variants in SET domain-containing methyltransferases.⁵⁵ The fly Rbbp5 interacts with trithorax proteins, a coactivator complex that maintains gene activation through H3K4 methylation.²⁸ The Kabuki syndrome implicated methyltransferase KMT2D is homologous to trithorax related (trr), a gene that is important for eye development and hormone responsive development.⁵⁶ In addition to RBBP5, another COMPASS member ASH2L homolog, ash2, also interacts with Trr and the ecdysone receptor (EcR) to direct molting and metamorphosis.⁵⁷ Notably, ash2 mutants also exhibit neural and optic lobe developmental defects.⁵⁸

We present evidence that the RBBP5 p.(T232I) and p.(E296D) variants are hypomorphic loss-of-function alleles. We observed earlier lethality upon ubiquitous expression of the human reference than either missense variant (Figure 3). We also observed variant-specific loss-of-function phenotypes using ubiquitous and tissue-specific overexpression (Figure 4, Supplemental Figure 2). We identified that Rbbp5 is expressed in both neurons and glia in the developing Drosophila brain (Figure 5). We found that loss of Rbbp5 results in microcephaly in the larval stage and confirmed that this loss is lethal (Figure 6). Unfortunately, the human RBBP5 is unable to rescue loss of the Drosophila Rbbp5 gene. Failure to rescue with the human cDNA is observed in 30% of cases in our experience. The human protein may bind different targets or could have less specificity at the lower body temperature of flies or other reasons stemming from evolutionary divergence. This could also be influenced by the precise expression of neural genes that is required for brain development. Indeed, we observe consistent microcephaly phenotypes across all genotypes, even when overall body size is not as severely affected (Supplemental Figures 1 and 2). However, we did find variant-specific differences when the RBBP5 transgenes are expressed. A similar microcephaly phenotype is induced upon co-expression of either p.(T232I) or p.(E296D) in an Rbbp5 null background that is not as severe as co-expression of the human reference. Moreover, expression of p.(T232I) or p.(E296D) in a null background induces the same microcephaly phenotype as seen in Rbbp5 null animals confirming that both missense variants are loss-of-function alleles (see Figure 6). In addition, expression of the fly and human cDNA leads to wing patterning defects, and expression of the fly p.(T231I) or p.(E295D) variants fail to induce fully penetrant wing patterning defects as seen upon overexpression of the fly Rbbp5 (Supplemental Figure 3). We observe both complete and partial loss-of-function phenotypes through analysis of the p.(T232I) and p.(E296D) variants using the human cDNA (Supplemental Table 2). Thus, a conservative evaluation is that they are partial loss-of-function alleles. Because we do not observe milder phenotypes in clinical symptoms between missense and truncating variants, we cannot exclude that these variants could be complete loss-of-function alleles in the human system. From these cumulative data, we conclude that both missense variants investigated in this study are partial loss-of-function hypomorphic alleles.

We have shown that H3K4 trimethylation is disrupted in RBBP5 expressing animals, suggesting an inability to activate expression of key developmental genes. We observed variant-specific developmental abnormalities in larvae ubiquitously overexpressing RBBP5, including microcephaly and overall growth phenotypes. Therefore, the p.(T232I) and p.(E296D) variants disrupt the function of the COMPASS complex possibly because of the substitution of critical residues, resulting in an inability to trimethylate H3K4 to direct downstream transcriptional activation. Because we observe unique tissue-specific loss-of-function phenotypes between the missense variants, it is possible that downstream gene expression is dysregulated in a variant-specific manner. Future transcriptomics studies could identify the critical genes dysregulated by these and additional RBBP5 variants. Furthermore, inclusion of variants in additional members of the COMPASS complex in transcriptomics studies could begin to identify the target genes responsible for the overlapping and distinct phenotypes involved in the spectrum of observed clinical symptoms.

In summary, we have provided the first evidence for a syndromic neurodevelopmental disorder that is associated with pathogenic variants in RBBP5. This study provides a new perspective to the disorders of the epigenetic machinery.

Data Availability

The authors confirm that the data supporting the findings of this research are available within the manuscript or available upon request.