Incomplete penetrance for isolated congenital asplenia in humans with mutations in translated and untranslated RPSA exons

A Model for the Genetic Architecture of ICA Caused by RPSA Mutations.

The field of human genetics has benefited from the recent release of large databases reporting allele frequencies for variants observed in the exomes (123,000) and genomes (75,000) of about 200,000 individuals (11). These new tools can be used to rule variants out as the cause of a disease on the basis of their allele frequency in the general population. However, several parameters (prevalence, inheritance, penetrance, and genetic and allelic heterogeneity) must be taken into account before defining the highest allele frequency in these databases considered plausible for an ICA-causing variant. Based on an estimate of about one in a million children being admitted to the hospital for ICA (2), and the fact that about 40% of individuals with ICA were never admitted to the hospital during childhood (10 of the 25 patients that were not on prophylactic antibiotics during childhood, in our cohort) (Fig. 1A and SI Appendix, Table S1) (5), we estimate the global prevalence of ICA to be about 1 in 600,000. We found mutations in the protein-coding sequence of RPSA to display complete penetrance for ICA in our previous study, but AD disorders by haploinsufficiency generally display incomplete penetrance (12, 13). We therefore chose to apply a model of high penetrance but not complete to be conservative and selected 75% penetrance. As variants of RPSA were shown to be the genetic cause of ICA in 8 of 23 kindreds in our previous study, we estimated the RPSA gene to be responsible for about 30% of ICA cases (genetic heterogeneity of 0.3). We observed the same mutation twice in the eight ICA kindreds with RPSA mutations. We therefore hypothesized that a single mutation could explain up to 25% of the ICA kindreds (allelic heterogeneity of 0.25). Using these estimates, we were able to calculate a maximum plausible allele frequency for an ICA-causing variant in a monoallelic dominant model, as follows: $AF\max \_mono=\left(1/600,000\right)\hspace{0.17em}\times \hspace{0.17em}0.25\hspace{0.17em}\times \hspace{0.17em}0.3\hspace{0.17em}\times \left(1/0.75\right)\times \left(1/2\right)$ = 8.33 × 10⁻⁸ (14). Therefore, in our monoallelic model, the maximum tolerated allele count for an ICA-causing variant in the two combined and complete databases (∼200,000 individuals) would be 0 (95% CI). The maximum tolerated allele count for an ICA-causing variant in these databases restricted to whole-genome sequencing (WGS) (75,000 individuals) would also be 0. The seven nonsynonymous mutations identified in our 2013 study (5) are absent from the gnomAD and Bravo databases (accessed in February 2018) and, therefore, meet these criteria.

Identification of Previously Unknown Protein-Coding Mutations in RPSA.

We therefore hypothesized that nonsynonymous variants of RPSA absent from gnomAD and Bravo would be present in about a third of the newly recruited kindreds (SI Appendix, Fig. S1). Sanger sequencing of the protein-coding region of RPSA was performed for the 33 newly recruited kindreds. We identified 11 heterozygous coding mutations, in 12 kindreds, none of which was previously reported in ICA patients. None of these mutations is present in any of the public databases, or in our in-house cohort of 4,500 exomes from patients with various infectious diseases (as of February 2018). The mutations can be grouped as (i) six missense mutations (p.A21P, p.G26S, p.M34V, p.[I65I;V66F], p.Q84R, and p.R180Q), (ii) one in-frame deletion (p.L122del), (iii) three small frameshift insertions or deletions (p.S43Kfs*2, p.D73Vfs*16, and p.V197Sfs*26), and (iv) one nonsense mutation (p.W231*) (Fig. 1 B, C, and E). All known ICA patients in the new five multiplex families for whom DNA was available carried a missense mutation of RPSA (Fig. 1 B and C). No genomic DNA (gDNA) was available for one of the deceased ICA patients from family ICA-AQ, for another deceased individual who probably had ICA from family ICA-AG, and for one ICA patient from family ICA-BV who was recently recruited to participate in the study. The mutations in kindreds ICA-AN, ICA-AT, ICA-BH, and ICA-BU were found to have occurred de novo, after microsatellite analysis (MSA) to validate parent–child relationships of the DNA samples we sequenced. The mutation observed in ICA-AV was found to have occurred de novo in the father of the index case (first-generation relative) (SI Appendix, Fig. S3). Mutation p.R180Q is recurring in two families, but we were not able to analyze the full haplotype around the mutation in family ICA-BO. Finally, we performed MLPA in the remaining kindreds, to detect CNVs overlapping with the protein-coding region of RPSA. No CNVs were identified (Methods). The overall proportion of ICA kindreds carrying protein-coding mutations of RPSA (12 of the 33 new kindreds) is consistent with both our previous report (8 of the 23 previous kindreds) and our model for the genetic architecture of RPSA mutations underlying ICA.

The ICA Cohort Is Enriched in RPSA Protein-Coding Mutations.

We then tested the hypothesis that these very rare nonsynonymous mutations of RPSA caused ICA, by comparing their frequency in the ICA cohort and the general population (15). The gnomAD and Bravo databases were considered the most suitable for such an RPSA-burden test as, together, they contain sequencing information for about 375,000 chromosomes, and the exome and genome sequences they contain provide good coverage of the protein-coding region of RPSA. We restricted our search to mutations of RPSA absent from the combined gnomAD and Bravo data, based on our previous estimates. Thus, for the calculation of RPSA burden, we considered only mutations that passed the quality filter and were present on no more than one chromosome in gnomAD and Bravo for our original comparison. We did not restrict our analysis to a specific population, as our cohort is diverse and the variants compared are extremely rare. We found that the newly recruited ICA kindreds, and the entire ICA cohort, were significantly enriched in very rare nonsynonymous variants, relative to the general population (P = 2.44 × 10⁻³¹and P = 9.95 × 10⁻⁵¹, respectively) (Table 1). Restriction of the analysis to very rare missense variants and very rare in-frame indels revealed the newly recruited ICA kindreds, and the entire ICA cohort, to be significantly enriched in these variants (P = 9.89 × 10⁻²¹ and P = 3.44 × 10⁻³⁵, respectively) (Table 1). Finally, we compared the nonsense, frameshift, and essential splicing mutations (in coding exons) in the ICA cohort with those present in the public databases, regardless of frequency. On the 375,000 chromosomes of gnomAD and Bravo, only two essential splice variants and three rare potentially damaging frameshift or nonsense variants (p.W176*, p.E235Vfs*60, and p.Q283*) were identified in the canonical ENST00000301821 transcript of RPSA (burden test, P = 2.85 × 10⁻¹⁹) (Table 1 and SI Appendix, Fig. S4 and Tables S2 and S3). The end of the RPSA sequence is less strongly conserved than the rest, so the frameshift of the last 60 aa and the p.Q283* mutation removing the last 12 aa may have little effect. Overall, these results confirm our original observation that very rare nonsynonymous mutations of RPSA cause ICA in about a third of the ICA kindreds.

Incomplete Penetrance of Some Novel Protein-Coding RPSA Mutations.

Unlike that of the original RPSA protein-coding mutations reported in eight ICA families (5), the penetrance for ICA of the mutations in 6 of the 12 new kindreds (50%) with new protein-coding mutations of RPSA was incomplete (Fig. 1C). Fifteen individuals carrying an ICA-causing mutation were found to have structurally and functionally normal spleens based on abdominal US or CT scans and/or the absence of Howell–Jolly bodies on blood smears. The distribution of these 15 asymptomatic carriers of RPSA mutation was as follows: p.A21P (1 of 3 carriers), p.G26S (1 of 3), p.M34V (1 of 2), p.[I65I; V66F] (9 of 11), p.Q84R (1 of 2), and p.V197Sfs*26 (2 of 3). We hypothesized that the nature of the RPSA mutation (e.g., its biochemical impact) contributes to the penetrance of ICA. We compared the Combined Annotation Dependent Depletion (CADD) scores of the mutations with their penetrance for ICA. CADD scores take into account information about evolutionary conservation, gene regulation, and transcription, to calculate the extent to which a nucleotide change is likely to be deleterious (Methods) (16). There was no significant difference between the mean CADD of incompletely penetrant compared with fully penetrant RPSA mutations (mean CADD 25.8 vs. 29.4, P = 0.58) (Fig. 2A and SI Appendix, Fig. S4). Similarly, other deleteriousness predictors did not show significant differences between the two sets of variants (SI Appendix, Fig. S5). We then mapped positions of the human mutations onto the human crystal structure of the small ribosomal subunit (Fig. 2B) (17). All five missense mutations with incomplete penetrance were found to be located outside of the hydrophobic core of RPSA lined with two α-helices (Fig. 2 C and E). Amino acid 180, a mutational hotspot in ICA modified in 5 of 14 families with completely penetrant RPSA mutations, contacts the nearby α-helix and seems to stabilize the core (Fig. 2 C and D). Moreover, all seven missense mutations with complete penetrance caused a change in the polarity or charge of the amino acid side-chain group, whereas p.M34V and p.V66F did not. These findings suggest that the incomplete penetrance of nonsynonymous mutations of RPSA may be due to the residual function of the protein, but further studies are required to understand the difference in penetrance of different RPSA mutations.

Identification of Mutations in the Non–Protein-Coding Region of RPSA.

We hypothesized that some cases of ICA may result from mutations in the noncoding region of RPSA exons. We searched for mutations in the 5′-UTR and 3′-UTR of RPSA, which are poorly covered by most WES kits. The 5′- and 3′-UTRs were subjected to Sanger sequencing in all 36 ICA kindreds of our cohort without very rare nonsynonymous mutations of RPSA. As for the candidate ICA-causing protein-coding mutations, we considered only mutations not present in gnomAD and Bravo. We identified two mutations predicted to affect the splicing of the 5′-UTR of RPSA that were private to our ICA cohort in three kindreds (Fig. 1D). There were no very rare or private 3′-UTR mutations. One mutation, c.-34+5G>C (canonical ENST00000301821 transcript), was recurrent. It was identified in both a multiplex kindred (ICA-G) and a patient with sporadic disease (ICA-BD). Another mutation, c.-39_-34+3del (ENST00000301821), was identified in a patient with sporadic disease (ICA-AO). These two mutations were absent from the 150,000 chromosomes sequenced by WGS in gnomAD and Bravo corresponding to the subset of samples with good coverage of exon 1 of RPSA. Both 5′-UTR splice mutations were heterozygous. As for some of the protein-coding mutations, penetrance was incomplete in ICA kindreds carrying a mutation in the 5′-UTR of RPSA: c.-34+5G>C (one in four of the individuals in ICA-G, and two of three in ICA-BD had a normal spleen) and c.-39_-34+3del (two of three). Overall, 3 of 36 ICA kindreds carried rare noncoding mutations in RPSA exons, whereas the overall proportion of ICA kindreds carrying a rare coding or noncoding mutation in RPSA exons was 23 of 56.

Enrichment of the ICA Cohort in RPSA Noncoding Mutations.

We tested the hypothesis that these very rare noncoding mutations of RPSA cause ICA, by comparing their frequency in the ICA cohort and the general population. In total, 26 chromosomes carried a very rare RPSA 5′-UTR variant (one chromosome in gnomAD WGS and Bravo; SI Appendix, Tables S4 and S5). Fisher’s exact tests indicated that very rare mutations in the 5′-UTR or 3′-UTR of RPSA were more frequent in the ICA cohort than in the general population (P = 9.75 × 10⁻⁰⁶; Table 2). These results suggest that the very rare noncoding mutations identified in ICA patients may cause ICA. We also identified a 5′-UTR variant (c.-47C>T) in three unrelated Western European families that was absent from the gnomAD database but present in Bravo, with an allele frequency of 3.9 × 10⁻⁰⁵. This variant was not retained as potentially disease-causing in our study, due to its allele frequency, but it might have a modifying effect (Discussion). The two kindreds carrying the c.-34+5G>C mutation were both from the United Kingdom. This mutation is not present in the UK10K dataset (18) corresponding to high-quality WGS data for 3,781 samples. We investigated whether this very rare variant was recurrent due to a founder effect or a mutational hotspot. We defined the haplotypes encompassing this mutation in both kindreds. We identified an 850-kb region with a haplotype of 250 SNPs common to the two kindreds. The length of this common haplotype strongly suggests that the mutation was inherited through a founder effect in both kindreds. Using ESTIAGE (19), we estimated the most recent common ancestor (MRCA) to have occurred 163 generations ago (95% CI [52–743]). Assuming a generation time of 25 y, the MRCA was estimated to have lived 4,075 years ago (95% CI [1,300–18,575]). The ICA cohort is, thus, enriched in private 5′-UTR mutations, one of which is present in two distantly related families, with incomplete penetrance.

Functional Impact of the RPSA 5′-UTR Mutations.

We investigated whether these 5′-UTR mutations were responsible for ICA, by studying their functional impact. The 5′-UTR of RPSA is not generally well conserved (SI Appendix, Fig. S5). However, the two mutations found in our ICA cohort are located at positions chr3:39,406,769 (hg38) and chr3:39,406,759-39,406,767, which are highly conserved between 100 vertebrate species (SI Appendix, Fig. S6) (18). This suggests that the 5′-UTR mutations identified in ICA patients may have a functional impact (SI Appendix, Fig. S4). We then used computational tools to construct hypotheses concerning the potential functional impact of these 5′-UTR mutations. The c.-34+5G>C mutation was predicted to impair splicing at the exon1/intron1 junction by disrupting the acceptor site [method: NNsplice9.0 (20)]. The c.-39_-34+3del mutation eliminates the exon1/intron1 splicing junction and would, therefore, also be predicted to impair the splicing of intron 1. NNsplice9.0 predicted that the next splicing site would be 70 bp further downstream. We also estimated the potential impact on splicing of the other 5′-UTR variants present in gnomAD or Bravo. Only two of these 52 variants were predicted to impair splicing, both disrupting the acceptor splicing site for the intron1/exon2 junction (SI Appendix, Table S6). We then experimentally tested the hypothesis that these ICA-associated 5′-UTR mutations impair splicing. RNA-sequencing (RNAseq) analysis confirmed the predictions made: both the c.-34+5G>C and c.-39_-34+3del mutations resulted in the retention of the same 70 bp (67 bp in the case of c.-39_-34+3del given that three nucleotides are deleted) (Fig. 3A). The two 5′-UTR mutations lead to a 50% WT:50% mutant allelic ratio (Fig. 3B). These 70 or 67 extra nucleotides contain two potential ATG codons (SI Appendix, Fig. S7). We also performed TOPO-TA cloning on cDNAs generated from peripheral blood mononuclear cells (PBMCs) from kindred ICA-G (c.-34+5G>C) (SI Appendix, Fig. S8). Again, we found that the c.-34+5G>C mutation prevented correct splicing at the exon 1/intron 1 junction, creating an insertion of 70 bp at the end of exon 1. Both splice mutations of the 5′-UTR thus impair the structure but not the stability of the mRNA, contrasting with the completely penetrant p.D73Vfs*16 mutation, for which no RPSA transcript carrying the mutation (0/67) was identified by TOPO-TA cloning, likely due to nonsense-mediated decay. Overall, these results suggest that rare mutations with a significant impact on the 5′-UTR of RPSA probably cause ICA, with incomplete penetrance.

Patients.

The publication of our paper identifying RPSA variants as a genetic etiology of ICA led to additional referrals. Nine families, ICA-AG, ICA-AL, ICA-AM, ICA-AO, ICA-AQ, ICA-AR, ICA-BH, ICA-BP, and ICA-BO, approached us directly. Patients were recruited on the basis of three diagnostic criteria for ICA: (i) absence of a spleen on US or CT, or absence/presence of a splenic remnant on autopsy, (ii) presence of Howell–Jolly bodies on blood smear, and (iii) absence of heart abnormalities. After weighing up the pros and cons, we decided not to consider the members of ICA families with an accessory spleen as ICA patients. However, it seems plausible that a mutation leading to ICA may affect normal spleen development and results in the development of an accessory spleen. This would provide an argument for both individuals with no spleen and those with an accessory spleen being included in the “case” group for familial segregation analyses. However, accessory spleens are present in about 10% of the population, a frequency several orders of magnitude greater than that of total asplenia (35). As we have more than 50 kindreds enrolled in our cohort, we would expect there to be at least a few kindreds including individuals with an accessory spleen. We therefore decided not to consider these individuals as cases during familial segregation studies, and the absence of an RPSA mutation in these individuals would therefore not rule out the involvement of the mutation in ICA in other family members. In addition to studying the newly recruited families, we also updated the information for families reported in 2013. First, a newborn in family A was diagnosed with ICA. This baby carries the same missense mutation, p.R180G, as his relatives with ICA. Second, one patient, ICA-M-01, WT for RPSA has recently been diagnosed with Fanconi anemia, and we have identified compound heterozygous mutations of FANCA. Third, a sporadic boy (ICA-AH) has been found to carry a mutation in ATRX, which has been associated with asplenia as part of a broader syndrome in the past. He also suffers from mental retardation. We have retained these two kindreds in all of the calculations in this paper to ensure consistency with our first report. Finally, the reader may note several gaps in the “numbering” of the kindreds. For example, there is no kindred ICA-BA or ICA-BB, but there is an ICA-BC kindred. These gaps correspond to families or patients that have contacted us but that were not included in this study, for diverse reasons such as a lack of availability of DNA samples or of a signed informed consent form.

All experiments involving human subjects were conducted in accordance with local, national, and international regulations and were approved by the French ethics committee, ANSM (the French National Agency for the Safety of Medicines and Health Products) and the French Ministry of Research (protocol no. C10-16), and the Rockefeller University IRB (protocol no. JCA-0689). Informed consent was obtained from all participants included in this study.

Sanger Sequencing of the Seven Exons of RPSA.

In addition to the primers used for sequencing exons 2–7, as described in our previous paper (5), we designed primers to sequence the 5′-UTR/exon 1 of RPSA. We also included an M13 sequence in all primers (exons 1–7), to decrease the sequencing time. Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/) was used to check that the primers would not amplify another sequence from the human genome. Following the isolation of gDNA from EBV-B cells or PBMCs from patients and their relatives, each of the seven exons of RPSA was sequenced with these primers (listed in Table 3).

Allele Frequencies.

The allele frequencies of RPSA variants present in the general population were retrieved from the exome and genome information in two public databases (last accessed February 2018). The gnomAD database contains 123,136 exome sequences and 15,496 whole-genome sequences from unrelated individuals sequenced as part of various disease-specific and population genetic studies. Information about RPSA can be obtained from the following URL: gnomad.broadinstitute.org/gene/ENSG00000168028, and the Bravo database lists genomic information for 62,784 possibly related individuals sequenced as part of NHLBI's TOPMed program. Information for RPSA can be obtained from https://bravo.sph.umich.edu/freeze5/hg38/gene/ENSG00000168028. We also interrogated the UK10K dataset (https://www.uk10k.org/data.html) to determine whether the variant identified in two kindreds from the UK was present in the general population of the United Kingdom. The UK10K dataset contains 3,781 low-depth whole-genome sequences and aims to characterize genetic variation exhaustively, down to a minor allele frequency of 0.1% in the British population. We also sequenced RPSA in 29 Japanese controls from the Centre de l’Étude de Polymorphisme Humain (CEPH) panel because of the Japanese origin of one side of the ICA-AG family. None of the 29 Japanese controls carried the p.[I65I;V66F] mutation.

Estimation of Allele Frequency for ICA.

Maximum tolerated reference allele counts were calculated with the Frequency Filter available from https://cardiodb.org/allelefrequencyapp/. Based on the findings published in our first report, we predicted that ICA-RPSA would display monogenic inheritance, an incidence of about 1/600,000, an allelic heterozygosity of 0.25, a genetic heterozygosity of 0.3, and a penetrance of 0.75. Confidence was set at 0.95 and the reference population at 400,000 alleles.

Gene-Burden Tests.

All of the data from public databases were downloaded on February 20, 2018. All variants with an impact on the canonical transcript ENST00000301821 were analyzed. Bravo and gnomAD differ in their representation of ENST00000301821, particularly at the end of exon 3. We used the gnomAD representation, which is identical to the Ensembl description. We also used the LiftOver tool (https://genome.ucsc.edu/cgi-bin/hgLiftOver) to facilitate comparisons between the two databases. We converted the GnomAD hg19 coordinates into hg38 coordinates and performed manual checks to ensure that this process did not introduce any errors.

Based on our model, we focused principally on variants present on only one chromosome in the public databases. However, about 3,000 whole-genome sequences are present in both gnomAD and Bravo. We therefore decided to retain a variant that was present on one chromosome in Bravo and one chromosome in gnomAD. However, we counted it as occurring on only one chromosome in this case (1 + 1 = 1). Other variants present on two chromosomes in either database were systematically filtered out (2 + 0 = 0).

Nonsynonymous variants were defined as nonsense, stop-loss, frameshift, missense, essential splice, and splice variants affecting exons 2–7. Splice variants affecting only exon 1 were not taken into account for the calculation concerning nonsynonymous variants, because exon 1 is noncoding, but they were analyzed in the UTR-burden test. Moreover, the whole-exome sequences of the gnomAD database do not cover this region. The c.*3_25delTGTTCTTGCATAGGCTCTTAAGC mutation was not taken into account for the loss-of-function variant-burden test, as the deletion results in the same 295 aa and termination codon. All of the variants used in the gene-burden tests are presented in supplementary tables (SI Appendix, Tables S2–S5).

We restricted our RPSA-burden tests to variants that passed the quality filter in gnomAD and Bravo, as most of the low-quality variants are false positives (36). Fisher exact tests were performed with R software:

• input2.df = matrix(c(a,b,c,d), nrow = 2)

• p_val = fisher.test(input2.df)$p.value.

MLPA for CNV Detection.

We also searched for losses of one copy of RPSA as a cause of haploinsufficiency. We used the MLPA method, which has proved successful for the detection of CNVs in DBA (37). We designed probes mapping at least partly to the intronic region, because RPSA has several pseudogenes. As controls, we used probes binding to chromosomes 18, X and Y. We also designed a probe binding upstream from exon 1 to cover a region including a 21-bp deletion SNP (rs199844419) that would impair the binding of the probe, as a proxy for a control for the loss of one copy of RPSA. Three individuals in our cohort are heterozygous carriers of this SNP. No kindred presented a loss of one copy of one RPSA probe except for the probe upstream of exon 1 in the individuals carrying the 21-bp deletion. We also searched the ExAC database (exac.broadinstitute.org/gene/ENSG00000168028) for the presence of CNVs in RPSA: None have been reported to date (last accessed January 2018). At the time of submission (March 2018), CNVs have not yet been included in the gnomAD or Bravo databases.

The MLPA protocol was performed using the reagent from MRC-Holland. Primers are listed in Table 4.

• • The left and right probes (LPO and RPO) are mixed together to a final concentration of 200 nM each in 1× Tris-EDTA (10 mM Tris, pH 8.0 and 1 mM EDTA).

• • All of the probes are then mixed and diluted using the volumes in Table 4 to prepare the Synthetic Basic Probemix.

• • gDNA from individuals to be studied (100–150 ng) was resuspended in a total of 5 µL Tris-EDTA and denatured (98 °C, 5 min; 25 °C, 5 min).

• • Synthetic Basic Probemix (1.5 µL) and MLPA buffer (1.5 µL) were added to the denatured gDNA and incubated at 95 °C for 1 min followed by 60 °C for 16–20 h for hybridization. Thirty-two microliters of ligase mix buffer was then added to the mix for a 15-min incubation at 54 °C, followed by an inactivation step (98 °C, 5 min). Amplification of the ligated products is performed by adding 10 µL of PCR mix (SALSA PCR) followed by PCR amplification for 35 cycles [95 °C, 30 s; (60 °C, 30 s, 72 °C, 1 min) ×35; 72 °C, 20 min; 15 °C].

• • The separation of fragments by capillary electrophoresis were performed mixing 1 µL of undiluted PCR product to 9.0 µL formamide and 0.2 µL LIZ standard and then denatured at 85 °C for 2 min and held at 4 °C for a minimum of 5 min before analyzing using a ABI-3730 DNA analyzer.

Table 4.

MLPA probe for CNV detection in RPSA

Open in a new tab

^*

Noted that all of the RPO primers have a 5′-phopshorylated nucleotides. Sequences in blue indicate universal primer sequence and sequence in green indicate specific RPSA sequence. qs, quantum stasis.

^†

Solution of 200 nM of each LPO and RPO probe.

RPSA Haplotypes Investigated by MSA (de Novo).

We determined whether the samples sequenced in kindreds in which RPSA mutations appeared de novo were of the expected origin (i.e., the sample labeled as the father indeed corresponded to DNA from the father, and that labeled as the mother indeed corresponded to the DNA of the mother; in other words, we checked that the parent–child relationships between the subjects were as indicated by the family and that there was no mixing up of the tubes at any point during the study). This was necessary to determine whether RPSA mutations truly occurred de novo in the ICA-AN, ICA-AT, ICA-AV, ICA-BH, and ICA-BU families. We thus analyzed microsatellites throughout the genome for the samples sequenced (proband, father and mother). MSA was performed with the Identifiler Plus kit (Thermo Fisher Scientific) according to the manufacturer’s protocol, optimized for use on an ABI Prism 3730 DNA analyzer (Applied Biosystems). PCR was first performed on a Veriti 96-well thermal cycler, with 10 μL AmpFlSTR Identifiler Plus Master Mix, 5 μL AmpFlSTR Identifiler Plus Primer Set, and 10 μL of a 0.1 ng/μL solution of DNA from the subject to be tested. The samples were then prepared for electrophoresis, using 1 μL of the PCR products or an allelic ladder and 9 μL of a 1:9 mixture of GeneScanTM 600 LIZ Size Standard and Hi-Di Formamide. Finally, data were analyzed with GeneMapper 4.0. The results are shown in SI Appendix, Fig. S3.

Haplotype Analysis for c.-34+5G>C.

SNPs were genotyped with an Affymetrix 6.0 chip. After quality control (including a call rate of 100%), 909,622 SNPs were retained. Genotypes were phased with SHAPEIT v2 (38) software, with the 1000 Genomes Project samples as a reference panel. Familial information was used to improve the phasing and to identify the paternal haplotypes in which the mutation was located. We then determined the shared haplotype around the mutation and applied the likelihood-based ESTIAGE method (19) to estimate the age of the MRCA for the mutation. Recombination rates and haplotype frequencies were provided by the HapMap Project.

Prediction of the Impact of 5′UTR Mutations on mRNA Structure.

We used the NNsplice 9.0 tool (www.fruitfly.org/seq_tools/splice.html) (20) to assess the potential impact on splicing of the rare variants at the exon1/intron1 splice junction identified in three kindreds with ICA. This tool has been shown to perform well for predicting the impact of variants leading to a splicing defect (https://github.com/counsylresearch/posters/raw/gh-pages/2017/ASHG/Q506C_Tran_Intronic_Variants_V1_Final_PRINT.pdf).

RPSA Structure and Modeling of the Mutations.

The structural data were obtained by mapping the positions of the human mutations onto the human crystal structure of the small ribosomal subunit [Protein Data Bank (PDB) ID code 6EK0] (17). All structural analyses and illustrations were made using PyMOL Molecular Graphics System, v.1.8 (Schrödinger).

RNAseq Preparation.

Total cellular RNA was isolated from PBMCs with the RNeasy Mini kit (Qiagen). Multiplexed RNA libraries were prepared with the Truseq RNA sample prep kit (Illumina). In brief, poly(A)-containing mRNAs were captured with oligodT beads, fragmented, and reverse-transcribed, and the cDNA was ligated to Illumina adapters containing indexing barcodes. Libraries were quantified with KAPA Library Quant kits (KAPA Biosystems), and run on a HiSEq 2000 Sequencing System (Illumina) to produce 150-bp single-end reads. Reads were then mapped with STAR v2.5.3a. Sashimi plots were generated with IGV (Broad Institute), using a minimum threshold of 20 reads for a splice event to be taken into account. Accession ID: PRJNA477701.

TOPO-TA Cloning.

RNA extracted from PBMCs was reverse-transcribed with the SuperScript III First-Strand synthesis system for RT-PCR kit (18080-051), following the manufacturer’s protocol, with random hexamers. The TOPO TA cloning kit for sequencing (450030; Invitrogen) was then used for subsequent cloning. The products were used to transduce Oneshot TOP 10 competent cells. Finally, we performed PCR on these colonies and sequenced them with the cRPSA-WT-I1-R/cRPSA-WT-R6 primers (exons 1–4).

Prediction Software.

We used 12 software tools for the prediction of the deleteriousness of missense or indel-frameshift mutations in ICA (CADD, Condel, Envision, FATHMM, Grantham, Mutation Assessor, MCAP, MPC, Polyphen2, Provean, Revel, and Sift). See SI Appendix, Fig. S5 for data and references.