Results

Identification of Mutations in TOP3A
In ongoing work to identify genes associated with microcephalic dwarfism, whole-exome sequencing (WES) was performed on subject P1, who had significant microcephaly and short stature (−5.7 and −4.4 standard deviations [SD], respectively). This identified the homozygous frameshift mutation c.2718del (p.Thr907LeufsTer101) (GenBank: NM_004618.4) in TOP3A in chromosomal region 17p11.2. After interrogation of other cohorts and clinical contacts made through GeneMatcher,31 we subsequently ascertained a further nine individuals from seven families affected by biallelic deleterious TOP3A variants that had also been discovered by WES (Table 1). All variants identified were validated by Sanger capillary sequencing, and all parents were confirmed to be heterozygous carriers. Aside from the c.2718del variant (minor allele frequency = 0.00041%), none of the other variants were reported in GnomAD.33 Notably, four of the families from the United Arab Emirates, Syria, and Saudi Arabia (F2, F4, F6, and F7), were homozygous for the same variant.
Table 1 Identified TOP3A and RMI1 Variants Associated with Microcephalic Dwarfism Nomenclature is according to transcript GenBank: NM_004618.4 for TOP3A and GenBank: NM_004618.4 and NM_024945.2 for RMI1. a Close to donor splice site of exon 10. b Previously reported individual with adult-onset mitochondrial disease.32

Prenatal-Onset Growth Restriction and Microcephaly in Individuals with TOP3A Mutations
Prenatal growth restriction was evident from the substantially reduced birth weight in all individuals (mean weight −2.6 ± 0.9 SD; Table 2 and Figure 1A). Postnatally, weight (mean −4.8 ± 1.8 SD), height (mean −3.9 ± 1.0 SD), and occipital-frontal circumference (mean −4.0 ± 1.2 SD) were significantly reduced. Growth parameters were similar to those previously reported for Bloom syndrome (mean birth weight −3.7 ± 1.2 SD, mean postnatal weight −3.8 ± 2.1 SD, and mean height −3.9 ± 1.1 SD).21 Although multiple café-au-lait patches were present in some individuals with TOP3A mutations, the classical erythematous malar facial rash was not apparent, and there were no reports of malignancies (although none had yet reached adulthood; Table 3 and Figure 1B).
Figure 1 Variants in TOP3A Are Associated with Prenatal-Onset Growth Retardation and Microcephaly
(A) Morphometric data for TOP3A individuals. A global reduction in growth is evident from before birth. Z scores (standard deviations from population mean for age and sex) for birth weight (Wgt), postnatal weight, height (Hgt), and occipital-frontal circumference (OFC). Dashed lines indicate the 95% confidence interval for the general population. Black circles indicate data points for individuals with TOP3A mutations. For comparison, gray bars indicate the mean value for cohorts of 89 (birth weight), 47 (current weight), and 52 (current height) subjects with Bloom syndrome.21
(B) Photographs of facial features of individuals with TOP3A mutations.
Table 2 Growth Parameters of Individuals with TOP3A and RMI1 Mutations Z scores are the standard deviation from population mean for age and sex. The following abbreviation is used: NA, not available. a Adjusted for gestation. b Growth-hormone treatment from 4 years, 3 months of age to 5 years, 11 months of age showed no response. c Previously reported individual with adult-onset mitochondrial disease (ataxia and progressive external ophthalmoplegia).32
Table 3 Clinical Phenotype of Individuals with TOP3A and RMI1 Mutations Abbreviations are as follows: CDH, Congenital dislocation of hip; CMAMMA, combined malonic and methylmalonic aciduria (MIM: 614265) due to ACSF3 (MIM: 614245) mutations; HCM, hypertrophic cardiomyopathy; NA, not available; and PEO, progressive external ophthalmoplegia. a Expressive speech delay only. b All individuals are younger than 15 years old; in Bloom syndrome, neoplasia typically manifests in early adulthood. c Recurrent otitis media and tonsillitis, leading to tonsillectomy. d Reccurent upper-respiratory-tract infections and oral thrush. e Mild left-ventricle dilatation. f Anemia due to beta-thalassemia trait.

Cardiomyopathy Is Present in Some Individuals with TOP3A Mutations
A single individual compound heterozygous for rare TOP3A variants has been reported recently with an adult-onset mitochondrial disorder (denoted as MC1 in Table 1).32 Whereas stature was normal (163 cm) in this adult female, several of the subjects reported here had clinical features that could also be attributable to mitochondrial dysfunction. P3 and P4 (family 2) had an additional diagnosis of dilated cardiomyopathy, which proved fatal in P4 at age 10 years. Another brother in this sibship died as a result of cardiomyopathy at the age of 13 years. Morphometric and other clinical information are not available for this individual. In P5 (family 3), a dilated cardiomyopathy was also reported, and skeletal muscle biopsy demonstrated 87% mitochondrial DNA depletion. P8 also had asymptomatic left ventricular dilatation noted on echocardiography, whereas P9 had hypertrophic cardiomyopathy.
Given the identification of multiple individuals with biallelic deleterious variants in TOP3A, alongside a consistent clinical phenotype across the cohort, we concluded that these mutations were likely to be pathogenic. We therefore pursued confirmatory functional studies by employing both biochemical and cell-biological approaches.

TOP3A Mutations Lead to Markedly Reduced Cellular Amounts of Enzyme TopIIIα
Most of the identified TOP3A mutations are predicted to prematurely truncate the encoded protein (Figure 2A) and hence are likely to have significant effects on cellular protein amounts. We therefore assessed the amounts of TopIIIα in primary dermal fibroblast cell lines derived from P1, P7, and P8. Immunoblotting demonstrated that the amount of full-length TopIIIα was substantially lower in total cell lysates from all three individuals than in cell lines from unrelated control individuals and parents (Figure 2B).
Figure 2 Mutations in TOP3A Markedly Reduce Amounts of TopIIIα but Do Not Affect Its Intrinsic Decatenation Activity
(A) Schematic of TopIIIα with locations of variants annotated. Colors are as follows: red, TOPRIM domain; purple, TOPA domain; and blue, GRF zinc-finger domain.
(B) TOP3A mutations markedly reduce protein amounts. An immunoblot of a polyacrylamide gel is shown and was generated from total cell lysates from dermal fibroblast lines derived from affected individuals probed with an anti-TopIIIα antibody. siRNA depletion of TOP3A in a fibroblast line from control 2 was used to confirm the specificity of the antibody. C1 and C2 are unrelated control fibroblast lines. TopIIIα antibody was raised against amino acids 652–1,001.
(C) Recombinant TopIIIαThr812LeufsTer101 (denoted TopIIIαP1) is less stable than TopIIIαWT. An immunoblot of a polyacrylamide gel is shown and was generated from purified TopIIIαWT and TopIIIαP1 proteins probed with an anti-TopIIIα antibody. Protein size markers are 250, 170, 140, 100, and 70 kDa.
(D) Schematic depicting the BTRR complex and its role in dHJ dissolution. Homologous sister chromatids are shown in red and blue.
(E and F) TopIIIαP1 is proficient for promoting dHJ dissolution with BLM in combination with RMI1 and RMI2. (E) Representative polyacrylamide gel demonstrating the enzyme-concentration dependence of the dHJ dissolution activity of TopIIIαWT and mutant (TopIIIαP1) TRR complexes with 20 nM BLM at 37°C in 30 min. TRR was included in reactions at increasing concentrations ranging from 20 pM to 10 nM. A labeled circular oligonucleotide (lane 1) and dHJ (lane 2) were included as markers. (F) Quantification of the dHJ dissolution reaction shown in (E). The reaction was repeated twice more with very similar results.
In most subjects, the frameshift mutations would be expected to result in nonsense-mediated decay of TOP3A transcripts, explaining the consequent loss of TopIIIα. However, the homozygous frameshift mutation in P1 (c.2718del [p.Thr907LeufsTer101]) is at the 3′ end of the gene and is not predicted to result in NMD; instead, it would result in a protein with a length similar to that of the WT enzyme. The frameshift did, however, result in abolition of the C-terminal zinc-finger domain, whose precise cellular function remains to be defined. We therefore expressed recombinant TopIIIαThr907LeufsTer101 in E. coli and purified it to homogeneity as a complex with the co-expressed RMI1 and RMI2 in order to characterize it further. Notably, in contrast to the TRR complex containing TopIIIαWT, the complex containing TopIIIαThr907LeufsTer101 (hereafter referred to as TopIIIαP1) exhibited reduced stability during purification; lower yields and increased amounts of degraded products were evident on SDS-PAGE (Figure 2C). Nevertheless, when used in quantities equimolar to those of TopIIIαWT, TopIIIαP1 was proficient in a biochemical assay for dHJ dissolution when combined with other components of the BTRR complex. This indicates that this TopIIIα variant retains a near-normal level of single-stranded DNA (ssDNA) decatenation activity (Figures 2D–2F). Therefore, we concluded that the major consequence of the TopIIIαP1 variant, like the other truncating variants, is severe depletion of TopIIIα enzymatic activity in cells.
P5 was the only individual we identified to have an amino acid substitution, p.Ala176Val, that was present in trans with a frameshift mutation. This amino acid substitution is absent from GnomAD, and in silico analyses (MutationTaster, CADD, and SIFT) predict it to be deleterious. It is at a highly conserved residue (Figure S1) within the TOPA domain (Figure 2A) and would therefore be expected to be highly deleterious to enzymatic function.
Given the above results and that Top3a is a developmentally essential gene in mice,34 we conclude that all of the identified mutations result in marked but most likely hypomorphic loss of function of TopIIIα as a result of a reduction in the cellular amount of the protein. Therefore, these mutations would be predicted to severely compromise the decatenation activity of the BTRR complex in dHJ dissolution in vivo. To assess this possibility, we next pursued cellular assays to assess SCE frequency.

SCEs Are Markedly Elevated in Individuals with TOP3A Mutations
dHJ can be processed by two pathways; first, dissolution by the BTRR complex yields non-crossover products only; second, resolution by endonucleases that cleave Holliday junctions generate both non-crossovers and crossover products that are visualized as SCEs (Figure 3A). To determine if dHJ dissolution is impaired in the cells of affected individuals, we assessed the frequency of SCEs. After BrdU incorporation, we performed differential sister chromatid staining on primary fibroblasts and PHA-stimulated peripheral blood leukocytes (Figures 3B and 3C). Cells from individuals with TOP3A mutations had substantially (3- to 6-fold) more SCEs, than cells from control individuals or heterozygous parents (p < 0.0001 for all affected individual cell lines; Figure 3C). Therefore, excessive crossover recombination was evident in all tested individuals with TOP3A mutations, indicating that diagnostic cytogenetic assessment of SCE levels is predictive of TOP3A mutations as well as BLM mutations.
Figure 3 Individuals with TOP3A Mutations Have More SCEs, Reflecting Chromatid Hyper-recombination
(A) Schematic of dHJ processing. dHJs are either dissolved by the BTRR complex or alternatively cleaved by structure-specific nucleases that lead to dHJ resolution.
(B and C) BrdU strand-specific labeling of sister chromatids shows that TOP3A cells have more SCEs than control and parent cells. (B) Representative images of P1 and parent fibroblast cell lines. (C) Quantification of SCEs in fibroblast cell lines (P1, P5, P7, and P8) or PHA-stimulated lymphocytes derived from peripheral-blood samples (P2 and P3). The median value was plotted with more than ten metaphase spreads counted per subject. Pairwise non-parametric Mann Whitney tests were performed against parental control cells. F1, F2, and F5 SCEs were scored in independent laboratories.

Chromosome Segregation Defects and Genome Instability Occur in Cells with TOP3A Mutations
Impaired dHJ decatenation can result in persistently entangled sister chromatids that impede chromosome segregation at mitosis.19 Therefore, we performed detailed characterization of the mitotic consequences of TopIIIα deficiency in primary fibroblasts from P1. DAPI staining demonstrated elevated amounts of chromatin bridges (control [C] = 0, parent = 0.3 ± 0.4, and P1 = 4 ± 0.6) and lagging chromatin and chromosomes (C = 0.9 ± 0.2, parent = 4 ± 0.7, and P1 = 9.2 ± 0.9) (Figures 4A and 4B), consistent with persisting chromosome entanglements. Additionally, increased numbers of ultrafine DNA bridges (UFBs) were revealed by immunostaining for PICH on DAPI-negative regions19 (C = 2.7 ± 0.3, parent = 13.1 ± 1.1, and P1 = 26.7 ± 3.8).
Figure 4 DNA Catenanes Persist into Mitosis in Cells with TOP3A Mutations, Leading to Chromosome Segregation Defects and Genome Instability
(A and B) Chromosome segregation is impaired in TOP3A-deficient primary fibroblasts. (A) Representative images of chromatin bridges, lagging DNA (DAPI), and UFBs (detected by the presence of PICH and absence of DAPI stain). (B) Quantification of chromatin bridges, lagging DNA, and PICH-positive UFBs scored in control, parental, and P1 mitotic fibroblasts staged at anaphase B (experiments ≥ 3, n > 50 cells, error bars = SEM). To enrich for mitotic cells, we treated fibroblasts with R03306 for 6 hr and released and fixed them after 45 min.
(C) TOP3A P1 fibroblasts display significantly elevated amounts of micronuclei. Top: representative picture of control and P1 fibroblasts. Bottom: quantification of micronucleus containing interphase cells (experiments ≥ 3, n > 500, error bars = SEM). To enrich for G1 cells, we released RO3306-treated cells into fresh media for 30 min and collected, re-seeded, and fixed prometaphase cells after 4–6 hr.
(D) P1 fibroblasts with TOP3A mutations display significantly elevated numbers of 53BP1 bodies in G1 nuclei. Top: representative images of 53BP1 foci (red) and DNA (DAPI). Bottom: quantification of cells with at least four 53BP1 foci in G1 nuclei (negative for cyclin A) (experiments ≥ 3, n > 500, error bars = SEM). Scale bar: 2 microns. Two-tailed t test was performed against parent cells.
We then examined the post-mitotic consequences of the observed chromosome segregation errors by enriching for G1 cells (Figures 4C and 4D). Micronuclei often arise from chromosome segregation errors,35 and analysis of DAPI-stained cells demonstrated substantially more micronucleated cells in the P1 fibroblast cell line than in control and parental cell lines (C = 0.4 ± 0.2, parent = 1.8 ± 0.5, and P1 = 7.5 ± 1.4). Elevated amounts of 53BP1 nuclear bodies were also seen in cyclin-A-negative (G1) cells, indicating the transmission of DNA damage from one cell cycle to the next36 (C = 0.8 ± 0.3, parent = 4.5 ± 0.5, and P1 = 11 ± 1.7). Altogether, we conclude that impaired dHJ decatenation in TopIIIα-deficient cells results in both abnormal recombination and mitotic errors that lead to accumulation of DNA damage. Because chromatin bridges and UFBs are also present in individuals affected by Bloom syndrome,19 these findings are concordant with a shared disease mechanism.

A Homozygous Truncating Variant in RMI1 Is Associated with Microcephalic Dwarfism
Finally, through WES we also identified the homozygous truncating variant c.1255_1259del (p.Lys419Leufs∗5) in RMI1 in two affected cousins (P11 and P12) from a consanguineous Turkish family (F8; Table 1 and Figure S3). Both individuals also had microcephalic dwarfism with a clinical phenotype and level of growth restriction similar to those of individuals with TOP3A mutations (Figure 1A and Tables 2 and 3).