Aberrant splicing contributes to severe α-spectrin–linked congenital hemolytic anemia

Hereditary spherocytosis (HS), the most common inherited hemolytic anemia in Northern Europeans, affects people of every ethnic background worldwide. A quarter of HS cases exhibit recessive inheritance (rHS) (1–5). Clinically, rHS patients are more severely affected than patients with typical, dominant HS, often presenting in infancy or early childhood with life-threatening, hemolytic anemia (1, 2, 5). Many of these patients are transfusion-dependent. Erythrocyte membranes from patients with rHS and the related disorder hereditary pyropoikilocytosis (HPP) are spectrin-deficient, which leads to destabilization of the lipid bilayer (1, 2, 5–7). The degree of spectrin deficiency correlates with the degree of hemolysis and the clinical response to splenectomy (6). Biochemical and genetic studies have implicated defects of α-spectrin in most rHS and HPP patients, but in most, the molecular basis of disease is unknown (1, 5, 7–9).

In normal erythroid cells, α-spectrin is synthesized in large excess, with α-spectrin synthesized in excess of β-spectrin by a 3:1 to 4:1 ratio (10–12). Thus patients with 1 normal and 1 production-defective α-spectrin allele are biochemically and clinically normal, because normal α-spectrin production remains in excess and normal amounts of αβ-spectrin heterodimers are assembled on the membrane. Patients who are homozygotes or compound heterozygotes for defects of α-spectrin production will suffer from rHS (13). Similarly, patients who are compound heterozygotes for a structurally defective α-spectrin allele and a defective α-spectrin production allele will suffer from HPP (8, 9, 14).

We have studied the genetic basis of α-spectrin–linked anemia in 24 kindreds with rHS or HPP, including the first α-spectrin–deficient rHS kindreds described by Agre et al., as well as a group of transfusion-dependent patients, using whole-exome sequencing (WES) (1, 2, 15). We identified numerous novel mutant α-spectrin (SPTA1) alleles, including null alleles and missense mutations, primarily in the αβ-spectrin self-association site. In 17 of 48 alleles, only 1 or no deleterious SPTA1 mutations were identified. Whole-genome sequencing (WGS) of selected individuals identified the α^LEPRA allele, a rare, poorly characterized variant identified in a patient with rHS associated with an elongated α-spectrin mRNA transcript, in trans to other mutations (13). Because WT erythroid cells also carry the elongated transcript described with the α^LEPRA allele, its role in the pathogenesis of α-spectrin deficiency and inherited anemia has been unclear.

In a series of in vitro minigene splicing studies and in vivo splicing assays, we demonstrate that the α^LEPRA variant changes a weak alternate branch point (BP) to a strong BP. This change leads to increased utilization of an alternate 3′ splice acceptor site, perturbing normal α-spectrin mRNA splicing and creating the elongated mRNA transcript. In vivo mRNA stability studies in erythroid cells rendered homozygous for the α^LEPRA variant via CRISPR/Cas9-based gene editing revealed that a newly created termination codon in the elongated transcript activates nonsense-mediated decay leading to spectrin deficiency. These studies indicate that a novel mechanism of human genetic disease contributes to the etiology of a third of rHS cases, facilitating disease diagnosis and treatment and identifying a new target for therapeutic manipulation.

Individuals from 24 kindreds from 3 groups with inherited hemolytic anemia were studied using WES. One group was individuals from 5 spectrin-deficient rHS kindreds from the original reports of HS by Agre et al. and 2 spectrin-deficient rHS kindreds described by Tse et al. (1–3, 15). Clinical and laboratory details of these patients have been previously described. A second group was from 7 kindreds with the clinical diagnosis of rHS or HPP referred to the Yale Center for Blood Disorders (YCBD). Clinical and laboratory details of these patients are provided in Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/JCI127195DS1 The unifying features of this group are that the patients had severe hemolytic anemia and their erythrocyte membranes were spectrin-deficient. The third group was composed of 10 patients from 9 kindreds with transfusion-dependent (TD) anemia referred to the YCBD. Clinical details of these patients are provided in Supplemental Table 2. Prenatal onset of severe hemolytic anemia or onset in the immediate neonatal period, requiring in utero or early postnatal transfusion therapy, respectively, were common. All TD patients required chelation for early onset iron overload. Two patients in this group died. One died from complications of anemia in the neonatal period, while a similarly affected sibling remains transfusion dependent. The other patient died from hepatic complications of transfusion-related iron overload at 17 months of age.

WES identified numerous mutations in α-spectrin (SPTA1), with 28 of 48 alleles carrying novel variants (Table 1). No deleterious mutations were identified in other erythrocyte membrane protein genes. Null SPTA1 alleles were frequently found in trans to missense mutations. Two TD patients had deleterious mutations in both SPTA1 alleles; one with nonsense mutations in trans died of liver failure associated with iron overload (patient 16), the other with nonsense and splicing mutations in trans remains transfusion dependent. Six missense mutations were identified, with 4 in the αβ-spectrin self-association site, critical for membrane stability. Modeling studies predicted 3 mutations (R45G, L49R, R28H) disrupt tetramer formation while a fourth, M70R, although not directly located in the tetramer interface, creates steric hindrance with the laterally associated spectrin chain in bivalent tetramers (Supplemental Figure 1 and Supplemental Figure 2). One TD kindred patient homozygous for the R45G mutation had a sibling homozygous for the same mutation die in the immediate perinatal period due to complications of anemia (patient 21).

Table 1

Patient diagnosis and genetic characteristics

In 17 of 48 alleles, no deleterious mutations were identified. All 17 alleles carried the common α-spectrin^{Bug Hill} polymorphism, (rs35948326,NM_003126.2:c.2909C>A, NP_003117.2:p.Ala970Asp, α^BH), previously associated with α-spectrin–linked rHS (Supplemental Figure 3) (3). WGS of 2 rHS patients heterozygous for the α^BH variant was performed. Comparison to samples in the 1000 Genomes database revealed a single, common SPTA1 variant, α^LEPRA, chr1:158613314C>T, present on only 4 of 4610 alleles in 1000 Genomes. In all 4, it was heterozygous and in cis to the α^BH allele. This rare intron 30 variant was previously described in a rHS patient in trans to a SPTA1 nonsense mutation (13). Analysis of α^BH haplotypes revealed 3 predominant patterns (Figure 1) with the haplotype of the rHS patients identical to the 4 heterozygous α^LEPRA individuals. All 17 mutation-negative alleles carried the α^LEPRA mutation, in trans to other mutations in 16 patients (Table 1). The proband lacking any deleterious SPTA1 alleles was homozygous for the α^LEPRA allele. Thus, a third of all mutant SPTA1 alleles associated with severe hemolytic anemia carried the α^LEPRA variant.

Figure 1

Haplotyping at the SPTA1 locus. Haplotype analysis around the SPTA1 locus of all 71 alleles with the α^BH variant, Ala970Asp, in the 1000 Genomes 2015 database and 2 rHS patients heterozygous for the α^BH variant. The α^BH allele is denoted in black and the rare intron 30 variant, the α^LEPRA allele, is denoted in blue. The HG19 reference sequence is denoted in gray and the nonreference sequence is shown in red. The SPTA1 gene is shown at the bottom. The 4 individuals in the 2015 1000 Genomes database and both rHS patients who carry the α^LEPRA allele are shown in green and red, respectively.

In the original description of the α^LEPRA variant, RT-PCR of reticulocyte RNA demonstrated the α^LEPRA allele was associated with an elongated α-spectrin mRNA transcript (13). This transcript contained 70 nt from the 3′ end of intron 30 and was predicted to lead to frameshift and premature chain termination (13). Located 29 nt upstream of the alternate acceptor splice, it was unclear if or how the α^LEPRA variant influenced α-spectrin pre-mRNA splicing. Examination of this region in transcriptomes derived from RNA-seq of primary human WT erythroid cells at varying stages of development and differentiation reveals that an alternately spliced, elongated transcript identical to that associated with α^LEPRA is produced on the background of WT SPTA1 alleles, ranging from 2.4%–6.8% of total α-spectrin transcripts (Figure 2, A and B). K562 cells, which do not carry the α^LEPRA variant, also produce this transcript. Because WT erythroid cells carry the same elongated transcript described with the α^LEPRA allele, its role in the pathogenesis of α-spectrin deficiency and inherited anemia has been unclear, with some speculating it is a single nucleotide polymorphism of no functional importance.

Figure 2

An elongated transcript of intron 30 of the SPTA1 gene. (A) Normalized RNA-seq profiles of the SPTA1 intron 30 region in human WT, primary erythroid cells, and K562 cells are shown. Estimated percentages of elongated, partial intron 30–containing mRNA transcripts based on splice junction reads are shown. (B) A Sashimi plot demonstrates utilization of splice sites in intron-spanning reads.

Attempts to quantitate the elongated α-spectrin transcript in reticulocyte RNA from affected patients led to inconsistent results among patients and controls, and among individual samples from the same patient, with elongated/WT α-spectrin mRNA ratios ranging from 3.8%–5.4% for controls and 3.5%–9.7% for heterozygous α^LEPRA patients (not shown). This variability is likely multifactorial, reflecting variability of the contribution of the mutant SPTA1 allele in trans (nonsense, splicing, other mutations) due to mRNA stability, instability of the elongated mRNA transcripts, variability in amounts and differentiation stages of circulating reticulocyte mRNA (early vs. late) in individual patients, shipping-induced changes in mRNA, and limitations of PCR-based assays.

To gain insight into potential effects of the α^LEPRA allele on mRNA processing, a series of bioinformatic studies was performed. Analysis of SPTA1 intron 30 using Sroogle, which analyzes splicing signals and the influence of mutations on splicing (16), predicted: (a) a branch point (BP1) at the expected location 31 bp 5′ of the 3′ acceptor site (3′Acc1) of exon 31; (b) an alternate upstream BP centered on an A 2 bp 3′ of the α^LEPRA mutation (BP2); (c) a novel alternate 3′ acceptor site downstream of BP2 (3′Acc2); and (d) the α^LEPRA variant substantially improves the probability of using BP2 (schematic in Figure 3, A and B). These findings predict that alternate BP usage between WT and α^LEPRA alleles influences transcript composition and abundance. Branchpointer, an algorithm that predicts and analyzes BP (17), indicated that WT BP1 and BP2 have low BP probability scores (0.51 and 0.67, respectively), whereas the α^LEPRA variant markedly improves the probability score of BP2 (0.94, Figure 3C), supporting the Sroogle predictions. These observations are true when comparing intron 30 BP probability scores to predicted BPs of 8180 introns of 1000 highly expressed erythroid genes (Figure 3D).

Figure 3

Computational analyses of intron 30 splicing of the human SPTA1 gene. A schematic of cis sequences and BP usage in WT (A) and α^LEPRA SPTA1 (B) alleles. In WT cells, BP1 is primarily utilized, producing large amounts of correctly spliced, in-frame mRNA transcripts (blue), whereas utilization of a weak, alternate BP2 produces a small amount of elongated transcript that leads to a frameshift and a novel termination codon (red). The α^LEPRA mutation, which changes BP2 from a weak alternate BP to a strong alternate BP, leads to increased utilization of BP2, producing large amounts of elongated transcript with decreased utilization of BP1. This leads to production of a small amount of correctly spliced transcript. (C) Branchpointer, an algorithm that predicts branch points, shows both BP1 and BP2-WT have low BP probability scores (0.51 and 0.67, respectively), while the α^LEPRA variant markedly improves the probability score of BP2-L (0.94). (D) These observations are true when comparing intron 30 BP probability scores to the predicted BP probability scores of the BPs of 8180 introns of 1000 highly expressed erythroid genes.

To experimentally validate the bioinformatic predictions, functional analyses were performed in minigene assays, a valuable technique for the study of splicing (18–20), and in cells manipulated via CRISPR/Cas9-based gene editing. A set of minigenes with mutation of predicted splicing elements on the background of WT and α^LEPRA alleles was analyzed in K562 cells (Figure 4, A and B). A WT minigene yielded only 0.78% of the elongated α-spectrin transcript (from use of BP2). In contrast, a minigene containing the α^LEPRA mutation yielded 42% of the elongated transcript (Figure 4C). To address the critical question of whether BP2 is a functional BP, we mutated the invariant A of BP2 to G (Figure 4C). This minigene yielded 100% WT and no elongated α-spectrin mRNA transcript. To further validate BP2 as a functional BP, we created K562 cells homozygous for G in BP2 (from central A) using gene editing. Mutation of BP2 A also eliminated the elongated α-spectrin mRNA transcript. These data support the prediction that BP2 is a functional BP.

Figure 4

Minigene studies of α-spectrin intron 30 and the α^LEPRA variant. (A) Partial sequence of intron 30 of the SPTA1 gene, showing the location of the α^LEPRA variant. (B) Each minigene construct used in minigene assays includes the ANK1 erythroid promoter, a fragment of SPTA1 genomic DNA inserted into intron 2 of the HBG1 gene, and the HBG1 3′ untranslated region and polyA signal. The hybrid HBG1-SPTA1 transcripts derived from minigenes are shown, either WT or elongated, with the locations of Taqman probes utilized to detect total spectrin (T bar) or the unique insert of the elongated transcript (E bar). (C) Minigene results. The specific sequences utilized in minigene constructs are shown. Percentages of elongated α-spectrin transcript over total α-spectrin transcript are shown in the second column from right. The adjusted P value of the difference from WT is shown on the right.

Additional aspects of α-spectrin mRNA processing were also assessed (Figure 4). Mutation of the alternate 3′ splice acceptor site led to abolition of the elongated transcript. Changing the poor BP consensus of BP1 to a U2 binding consensus sequence (21) yielded 100% WT and no elongated α-spectrin transcript. These studies indicate the α^LEPRA mutation, located upstream of an alternate 3′ splice acceptor site, changes a weak alternate BP to a strong BP (BP2) in the context of a poor primary BP (BP1). These changes lead to increased usage of the alternate 3′ splice acceptor site, which is utilized at a low level in WT cells, creating an alternate elongated α-spectrin mRNA transcript that leads to frameshift and a novel termination codon.

The novel termination codon, 87 nt upstream of an exon-exon junction, is in a position predicted to activate nonsense-mediated decay (NMD) (22, 23). To address whether NMD of the elongated transcript is the mechanism of α-spectrin deficiency, we created K562 cells homozygous for the α^LEPRA allele using gene editing (Supplemental Data) and treated them with the NMD inhibitors emetine or cycloheximide (24, 25). In homozygous α^LEPRA and to a lesser extent in WT cells, the total amount of elongated transcript increased relative to total α-spectrin transcript after NMD inhibition (Figure 5A).

Figure 5

Analyses of nonsense-mediated decay. The influence of NMD on the stability of the elongated α-spectrin transcript in WT K562 cells and in K562 cells rendered homozygous for the α^LEPRA allele. (A) After treatment with the NMD inhibitor emetine (E) or cycloheximide (C), or both, the amounts of total α-spectrin and the elongated α-spectrin transcript were determined by real-time RT-PCR with fluorescent Taqman probes. After NMD inhibition, amounts of elongated α-spectrin transcript were increased in α^LEPRA cells (adjusted P values 0.0014, 0.0027, and 0.0014 for WT vs. cycloheximide, emetine, and cycloheximide + emetine, respectively) and to a lesser extent in WT cells (adjusted P values 0.084, 0.012, and 0.0047, respectively). (B) Minigene assay of NMD. Two nucleotides were inserted into the polypyrimidine tract of the primary 3′ acceptor site of exon 30 in the minigene model to place the elongated transcript in-frame. As predicted, NMD was abrogated and the amounts of elongated α-spectrin transcript from both the WT and the α^LEPRA minigene constructs significantly increased. Three biologic samples each were tested in 3 independent experiments with the mean calculated from 9 values. Statistical significance was determined by 2-tailed Student’s t test.

To further validate these observations, we inserted 2 nt into the polypyrimidine tract of the 3′splice acceptor site of exon 30 (3′Acc1), putting the elongated transcript in-frame in our minigene model. On both WT and α^LEPRA backgrounds, the amount of the elongated α-spectrin transcript increased (Figure 5B), supporting the conclusion that NMD is responsible for low levels of α-spectrin mRNA associated with utilization of BP2.

Mutations leading to alterations in normal mRNA splicing are a major cause of genetic disease (26–28). Most are cis-acting mutations located in canonical donor or acceptor splice site sequences, with others in proteins involved in the splicing process (26, 28). Only a small number of cis mutations in BP consensus sequences have been reported, associated with mutation or deletion of the canonical A or mutation of the -2U (29, 30). With α^LEPRA, strengthening of an alternate BP in the appropriate cis context creates a series of aberrant splicing events that lead to a decreased amount of normal mRNA transcripts. As shown here, application of WGS to the study of inherited disorders is expected to increase the number of disease-associated intronic mutations influencing splicing, including abnormalities of intronic BPs (31).

These studies identify a novel mechanism of disease and reveal the etiology underlying many cases of rHS, HPP, and TD anemia. They also resolve an unanswered question of the role of the α^LEPRA allele in cases of severe anemia where its contribution to α-spectrin deficiency and disease pathogenesis had been unknown (13, 32–35). Correlating genotype with erythrocyte α-spectrin protein amounts determined by radioimmunoassay, a patient homozygous for the α^LEPRA allele exhibited 59% α-spectrin protein on the membrane and a compound heterozygote with α^LEPRA in trans to a null allele exhibited 34% to α-spectrin protein on the membrane (1, 2). This leads to speculation that α^LEPRA allele spectrin deficiency is associated with approximately 30% of normal α-spectrin protein assembled on the mature erythrocyte membrane.

These data demonstrate marked genetic heterogeneity in α-spectrin–linked hemolytic anemia, supporting the use of genomic strategies for disease diagnosis, particularly in severe and TD cases. These observations will facilitate disease diagnosis, as most diagnostic gene panels for hereditary hemolytic anemias do not include the intronic region containing the α^LEPRA variant nor do most whole-exome–sequencing capture panels.

When identified in affected patients, the findings are actionable, allowing the design of appropriate therapy. In our cohort, 3 patients became transfusion independent after splenectomy and 2 became transfusion independent after hematopoietic stem cell transplantation.

The α^LEPRA allele is a new target for therapeutic gene manipulation (26, 36–38). Therapies targeting splicing, particularly antisense oligonucleotides which recognize cis RNA regulatory elements and alter splicing or target transcripts for degradation and small molecule–based strategies that target splicing factors or RNA sequences and/or RNA structure, are showing significant promise (26, 36). The α^LEPRA allele is also an attractive candidate for therapeutic gene editing, as our data indicate alteration of the α^LEPRA allele variant itself or the alternate 3′ acceptor site will abolish production of the mutant, elongated α-spectrin mRNA transcript.

The data demonstrate how the synthesis of complementary lines of investigation, clinical, laboratory, biochemical, and genetic data can be leveraged to define and advance our understanding, diagnosis, and treatment of inherited disease.

Erythrocyte membrane preparation, quantitation of spectrin content, and limited tryptic digestion of spectrin. Erythrocyte membranes were prepared from peripheral blood as previously described (39). Membrane proteins were analyzed and spectrin content determined as described (40). Membrane proteins were separated by SDS-PAGE on 3.5%–17% gradient polyacrylamide gels and stained with Coomassie blue. The spectrin/band 3 ratio was quantified by densitometric scanning of the stained gels at 540 nm and integration of the surface area under the spectrin and band 3 peaks (41). Spectrin was extracted by incubating ghosts overnight at 4°C in low ionic strength buffer. Limited tryptic digests of spectrin extracts were prepared as described (39) and separated by 2-dimensional gel electrophoresis with isoelectric focusing (I EF) as modified by Speicher et al. (42).

Exon capture and WES. Targeted regions of genomic DNA were captured using a SeqCap EZExome V2.0 (Roche) solution-based capture system according to the manufacturer’s protocol. The captured, purified, and amplified libraries targeting exomes from patients were sequenced on a HiSeq Analyzer with paired-end sequencing at 75-bp read length. Sequencing reads were processed and analyzed as described (8).

Fastq sequence reads were aligned to the human genome (hg19build37/GRCh37) using BWA mem 0.7.9a software (9). Variant analysis was performed using Genome Analysis Toolkit (GATK) analysis software v.3.1-1 (10, 11). Sequencing reads in the region of known insertion/deletions in the 1000 Genomes database were realigned to reduce false-positive variants. Individual base quality scores were empirically recalibrated using covariate information in combination with the variation in dbSNP build 137 and from 629 human genome sequences from the 1000 Genomes project (12, 13). Base alignment quality scores were determined then recalibrated with GATK to further reduce false-positive calls (14). SNPs and insertions/deletions (indels) were called using the GATK HaplotypeCaller Bayesian algorithm for variant discovery and genotyping, which uses base qualities and allele counts to determine probabilities for called variants. Indels and novel SNPs were annotated using Annovar (43), which determines if a SNP or indel changes a protein sequence or splice site, and provides a variety of predictions, including the combined annotation dependent depletion (CADD) algorithm (44) and Mutation Taster (45), to assess the likely effect of a mutation on protein function.

Whole-genome sequencing. Genomic DNA was prepared using the TruSeq PCR-free DNA HT sample preparation kit (Illumina) with 450 bp insert size. Intact genomic DNA was sheared, followed by end-repair and bead-based size selection of fragmented molecules. Adenines were added to the 3′ ends of the DNA size–selected fragments followed by ligation of Illumina sequence adaptors ligated onto the fragments and PCR. Library quality control included a measurement of the average size of library fragments using a BioAnalyzer (Agilent), estimation of the total concentration of DNA by PicoGreen (Thermo Fisher Scientific), and a measurement of the yield and efficiency of the adaptor ligation process via a quantitative PCR assay using primers specific to the adaptor sequence.

Sequencing was performed on a HiSeq X instrument (Illumina) generating 2 × 150-bp read lengths. After alignment and duplicate removal, this equated to ×30 mean genome coverage (for the gender-specific ~2.85 gigabase mappable human genome). WGS data were processed through an automated pipeline at New York Genome Center’s high-performance computational facility. Paired-end 150-bp reads from the WGS were aligned to the GRCh37 human reference using the Burrows-Wheeler Aligner (BWA-MEM v0.78) (46) and processed using the best practices pipeline that includes marking of duplicate reads by the use of Picard tools (v1.83, http://picard.sourceforge.net), realignment around indels, and base recalibration via GATK v3.2.2 (47).

Two patients (nos. 1 and 23) were selected for WGS, both heterozygous for the α^BH variant. Analyzing WGS data, 122 variants in the SPTA1 region (chr1:158545000-158690000) present in both samples were selected for further analysis using Annovar software to annotate allele frequencies and predict functional consequences. In addition, we identified samples in the 1000 Genomes database of healthy individuals who were homozygous for α^BH and downloaded genotype data for these individuals. Variants present in the 2 patient samples that were homozygous in the 1000 Genomes samples were excluded from further analysis, leaving 16 candidate variants, of which 4 were excluded because they were present at high frequency (>1%) in the 1000 Genomes database. Eleven remaining variants were excluded because they were homopolymeric or short tandem repeat size variations that are common and often not accurately genotyped. The sole remaining variant, rs200830867, chr1:158613314 G to A, is the α^LEPRA allele (13). This variant was initially described in an rHS patient in trans to an SPTA1 nonsense mutation associated with an elongated α-spectrin mRNA transcript containing 70 nt from the 3′ end of intron 30 hypothesized to lead to frameshift and premature chain termination (13).

Mutation validation. Variants identified by WES or WGS were validated by Sanger sequencing. Sanger sequencing was performed on an 3130XL capillary sequencer (Applied Biosystems). SPTA1 variants were classified as cis or trans by study of the proband’s parents. Variants were classified as novel if not found in the 1000 Genomes or Exome Aggregation Consortium (ExAc) databases.

Molecular modeling. The effects of α-spectrin missense mutations on tetramer structure and their likely functional consequences were evaluated using the structures of the univalent α-β tetramer complex (48), closed spectrin dimers (49, 50), and the divalent tetramer complex as templates (50).

Minigene analyses. Minigene splicing assays were designed as described (18). Plasmids were constructed containing the human erythroid ANK1 promoter, the human HBG1 gene including the HBG1 ATG/Kozak consensus, and the region of the SPTA1 gene containing either WT or α^LEPRA sequence for use in minigene splicing assays. Briefly, a SmaI-BglII fragment of the ANK1 gene promoter was linked to a BglII-HindIII fragment containing exons 1–3 of the HBG1 gene and flanking 3′ region. A 1684-bp XbaI fragment of the SPTA1 gene from intron 29 to intron 32, containing either the WT or α^LEPRA sequence, was cloned into intron 2 of the HBG1 gene. These plasmids contained a termination codon and the polyadenylation signal from the HBG1 gene.

Wild-type and α^LEPRA spectrin minigene plasmids containing fragments of the SPTA1 gene from intron 29 to intron 32 were prepared and transfected into 10⁷ K562 cells (ATCC, CCL-243) using nucleofection program T-016, buffer V (Amaxa). Cells were harvested after 48 hours, RNA was prepared then reverse transcribed using oligo d(T) and random hexamers, and cDNA was amplified with primers that included sequences from flanking HBG1 and internal SPTA1 exons. These primers allow differentiation of minigene amplification products from endogenous SPTA1 transcripts.

Fluorescent Taqman probes (Applied Biosystems) corresponding to total SPTA1 and the elongated α-spectrin transcript were added to amplification reactions, allowing determination of the contribution of elongated α-spectrin transcripts after normalization. To assay the total minigene SPTA1 transcript, PCR was performed using primers 5′-CTTGGAGACTATGCCAACCTAAA-3′ (sense) and 5′-CACATTTCCCAGGAGCTGAA-3′ (antisense) with transcript quantitation by Taqman probe 5′-AATGGATCAGTGAGATGCTGCCCA-3′ (sense). To assay elongated α-spectrin transcripts, PCR was performed using primers G4833, 5′-GAGAATTCCCTGAGGTCAGATG-3′ (sense) and G44820, 5′-ATCCCAGACTCCCTCCTG-3′ (antisense) with transcript quantitation by Taqman probe 5′-AGGCTTTGATGAAGAAACGGGACGA-3′ (sense). All analyses included 3 biologic and 2 technical replicates.

CRISPR/Cas9 gene editing analyses. Four guide RNAs (gRNAs) in the regions flanking the predicted BP2/α^LEPRA region were designed, synthesized, and tested for use in gene editing experiments. gRNAs were synthesized using HiScribe T7 High Yield RNA Synthesis Kit (New England Biolabs). Testing was done using the Guide-it sgRNA Screening Kit (Takara). One gRNA was selected and used for all experiments (Target sequence: GGATTCAGAAGATATACTCA). Cas9 protein (15 μg, PNAbio) was complexed with 200 pmol gRNA. The gRNA-Cas9-RNP complex and 200 pmol donor ssDNA (BP2 A to G donor ssDNA: g(s)g(s)a(s)actctgtaccacacaagtagcccattattagatgtttctcctcctattaagttgaaacccacctccctgtaaggcatatattatttgaccctgagtatatcttctgaatccGcaaaggatacctt(s)c(s)a(s) or α^LEPRA donor ssDNA: g(s)g(s)a(s)actctgtaccacacaagtagcccattattagatgtttctcctcctattaagttgaaacccacctccctgtaaggcatatattatttgaccctgagtatatcttctgaatTcacaaaggatacctt(s)c(s)a(s)) were nucleofected (Amaxa) into K562 cells using the T-016 nucleofector program. Cells were placed into 96-well plates 48 hours after nucleofection. Clones were expanded for 10–14 days and then screened for either homozygous substitution of the BP2 A to G or homozygous alteration of the WT nucleotide α^LEPRA substitution C to T, by PCR amplification followed by Sanger nucleotide sequencing.Homozygous mutant BP2 and α^LEPRA clones were identified and used in additional studies (Supplemental Figure 4).

mRNA stability assays. The influence of NMD on the stability of WT α-spectrin and α^LEPRA mRNA was examined in WT K562 cells and in K562 cells rendered homozygous for the α^LEPRA allele. Cells were incubated at 37°C with the NMD inhibitor emetine (100 μg/ml, emetine dihydrochloride hydrate, MilliporeSigma, E2375) for 8 hours or with the NMD inhibitor cycloheximide (100 μg/ml, MilliporeSigma, C1988) for 4 hours, or with a combination of both. The amount of total α-spectrin mRNA transcripts and α^LEPRA mRNA transcripts were determined by real-time RT-PCR with fluorescent probes as described above. Analyses included 3 biologic and 3 technical replicates.

RNA-seq data. RNA-seq data sets accessed from the Gene Expression Omnibus database were GSE61566 and GSE53983 for human erythroid cells and GSM958729 for K562 cells.

Statistics. GraphPad Prism 8 software (Graph Pad Software) was used for statistical analyses. All data were reported as mean ± SD in tables or as error bars. Comparisons between 2 groups were performed using the 2-tailed Student’s t test. Multiple test correction was performed using the Benjamini and Hochberg method (51). Significance was set at P less than 0.05.

Human subjects. All human studies were approved by the Yale University Human Investigation Committee review board. Written informed consent was received from participants or their parents, as appropriate, prior to inclusion in the study.