Short-hairpin RNA against aberrant HBBIVSI-110(G>A) mRNA restores β-globin levels in a novel cell model and acts as mono- and combination therapy for β-thalassemia in primary hematopoietic stem cells

The hemoglobinopathy β-thalassemia is a common and potentially lethal monogenic disease characterized by deficient β-globin production. Standard palliative treatment comprises regular blood transfusions and iron chelation therapy. The only widely applied cure is transplantation of allogeneic hematopoietic stem and progenitor cells, which often suffers from immunological complications and is available only to few patients. Efficiency of alternative correction by gene therapy is determined by underlying mutations, calling for its stratified application.1 The severe HBB splicing mutation is common in many countries (Online Supplementary Figure S1),2 results in a 90% reduction of HBB (β-globin) and creates aberrant mRNA containing a 19-nt intronic fragment with an in-frame stop codon. Although potent targets of nonsense-mediated decay (NMD), aberrant HBB transcripts are abundant in erythroid cells of homozygous patients and disease models,43 possibly owing to saturation of NMD by high substrate levels.5 According to still limited data from preclinical studies and clinical trials, HBB appears recalcitrant to therapy by gene addition compared to other HBB mutations, suggesting aberrant transcripts as pathological trans factors likely acting by co-translational inhibition.76 Their reduction by oligonucleotide-mediated splice correction may never reach therapeutic efficiency and would require chronic application,4 whereas lentiviral delivery of short-hairpin (sh)RNAs is robust and potentially translatable for curative β-thalassemia therapy.8

Here, we evaluate shRNA-mediated therapy of HBB thalassemia, firstly in a novel murine erythroleukemia (MEL) cell line encoding a human HBB transgene (MEL-HBB), and secondly in HBB-homozygous CD34 cells, either as monotherapy or after transduction with the GLOBE HBB gene-addition vector (see Online Supplementary Methods and Supplementary Table S1 for all experimental procedures). Humanized MEL-HBB cells and MEL-HBB normal controls were created by lentiviral transduction (Figure 1A–C). After determination of vector copy number per cell (VCN),9 bulk populations (MEL-HBB at VCN 1.9 and MEL-HBB at VCN 2.0) and a clonal cell line (MEL-HBB at VCN 1) were used for further experimentation. MEL-HBB and MEL-HBB faithfully represented HBB-derived transcript expression ratios (40% aberrant mRNA compared with 46% in HBB-homozygous CD34) and mutation-specific reduction in HBB expression (MEL-HBB showing 6.3% of MEL-HBB human HBB levels)(Figure 1E and F).3 Accordingly, genedosage equivalent expression of mutant and normal transgenes in bulk MEL-HBB (VCN 1.9) and MEL-HBB (VCN 2.0) cells, respectively, suggested a HBB protein ratio of 15.8 for normal compared to mutant. We thus regarded a 15.8-fold HBB induction after treatment as target level for correction in this model.

Figure 1.Creation and characterization of humanized MEL cell lines. A. Circular amplification of the intron1-exon2 border of the GLOBE transfer vector with mutagenic primers IVSI-110_Mut_FW and IVSI-110_Mut_RV to produce a mutated amplicon for production of GLOBE^{IVSI-110(G>A)}. B. Sequence confirmation of mutagenesis for bacterial clones of the final GLOBE^{IVSI-110(G>A)} plasmid across the intron1-exon2 border. C. The provirus forms of GLOBE and GLOBE^{IVSI-110(G>A)} as held in the novel humanized cell lines MEL-HBB and MEL-HBB^IVS, respectively. Green: core sequences for hypersensitive sites (HS2, HS3) of the β-globin locus control region and the human HBB promoter (βp) as human control elements of HBB expression; Purple: exons (“x”) and introns (“i”) of HBB; Red: Rev response element (RRE), central polypurine tract (cPPT) and transcriptionally inactivated long terminal repeat (LTR) as viral control elements; Black arrows: transcription from the provirus. D. Relative abundance of aberrant and normal HBB RNA in bulk MEL-HBB^IVS (VCN 1.9) and clonal MEL-HBB^IVS (VCN 1) compared to HBB^{IVSI-110(G>A)}-homozygous CD34⁺ cells. E. Representative immunoblot showing same-gel, same-membrane Actb and HBB expression and same-gel, same-membrane Actb and Hba expression, as indicated by dashed horizontal lines, including saturated exposure (HBB (sat.)) to visualize HBB bands for MEL-HBB^IVS cell lines; red color indicates saturation. F. HBB band intensities for immunoblots of independent transductions (n=3) normalized for Hba expression and displayed as a percentage of average expression for MEL-HBB (arithmetic mean ± sample standard deviation). Results of group-wise comparison of MEL-HBB^IVS cell lines compared to MEL-HBB control by one-way ANOVA with Dunnett’s multiple comparison test: ^****P<0.0001 (calculation threshold).

In order to reduce aberrant HBB mRNA and its potential interference with HBB expression, we designed shRNAs targeting the aberrant-specific 19-nt mRNA sequence (Figure 2A). For proof of principle and in order to achieve saturating shRNA expression and effect, we employed the pLKO.1 lentiviral vector (LV)10 with its constitutive RNA-polymerase-III (RNApolIII)-dependent U6 promoter at high multiplicity of infection (MOI). Of the four HBB-specific shRNAs designed, two avoided full inclusion of a potential RNApolIII terminator, a dT₄ run central to the 19-nt sequence.11 To this end, one of the shRNA targets had overlap upstream (Up), the other overlap downstream (Down) with the dT₄ sequence. A third, central target comprising the full 19-nt sequence was represented by two shRNAs, one with perfect complementarity and loop structure (Mid), the other with a mutated dT₄ sequence (TTGT) in its passenger strand (Mid2)(Figure 2B).

Figure 2.Target sequences of HBB^{IVSI-110(G>A)}-specific shRNAs and functional analysis in humanized MEL cells. A. Double-stranded cDNA sequence of the main HBB^{IVSI-110(G>A)}-derived transcript, with a stop codon (^*) in the aberrantly retained 19-nt intronic sequence concluding the open reading frame (ORF). Target sequences for four HBB^{IVSI-110(G>A)}-specific shRNAs, Up, Mid, Mid2 and Down, are indicated relative to the 19-nt intronic sequence (yellow overlay) and its central dT₄ sequence (green overlay). B. Complete shRNA-encoding DNA sequence for all four HBB^{IVSI-110(G>A)}-specific shRNAs, excluding flanking nucleotides for cloning into pLKO.1. Rounded rectangle: mismatched nucleotide in the Mid2 sense strand, breaking up the dT₄ sequence. C. Immunoblots for MEL-HBB^IVS cells collected at day 6 of differentiation for three independent experiments (n=3). Actb, Hba: endogenous β-actin and α-globin, respectively, HBB: human HBB^IVS-transgenederived β-globin. Dashed lines separate same-gel and -membrane analyses derived from the same experiment. For raw volume measurements of band intensities, see Online Supplementary Figure S3. D. Percentage of differentiation-normalized HBB chain levels relative to the highest value for each experiment, as extracted from C; Significant results for group-wise comparisons vs. Mock: Mid or Mid2 ^****P<0.0001 (calculation threshold). E. Percentage of Actb-normalized Hba levels relative to the highest value for each experiment, as extracted from C; Significant results for group-wise comparisons vs. Mock: Up ^*P=0.0359, Mid ^*P=0.0381, Mid2 ^**P=0.0056, Down ^*P=0.0183, Scr ^**P=0.0086; Mock: Mock treatment without vector; Scr: treatment with vector encoding a scrambled shRNA, other column and lane labels specify LV-encoded shRNAs. P values reported are for group-wise comparison by one-way ANOVA with Dunnett’s multiple comparison test. Bar charts show arithmetic mean ± sample standard deviation.

All four shRNAs plus scrambled control shRNA (Up, Mid, Mid2, Down, Scr) were then transduced individually into MEL-HBB VCN-1 cells and compared with mocktransduced (Mock) controls. Functional correction of transduced cells was analyzed by immunoblots for detection of HBB-derived HBB at day 6 of differentiation, which consistently showed unchanged expression for Up and a two-fold, yet statistically insignificant, change of expression for Down. Importantly, HBB expression increased significantly for Mid and Mid2, from (9.1±6.3)% of maximum Hba (α-globin)-normalized band intensity for the mock-treated sample, to (82.6±27.7)% for Mid and (97.8±3.7)% for Mid2 (Figure 2C and 2D). Compared with the HBB protein ratio of 15.8 between MEL-HBB and MEL-HBB cells as an approximate target level for correction, shRNAs alone induced HBB expression compared to Mock 9.1-fold (58%) for Mid and 10.8-fold (69%) for Mid2. Same-sample analyses of corresponding HBB mRNA expression showed by contrast that Down, but none of the other shRNAs, significantly upregulated the ratio of normal to aberrant HBB mRNA compared to Mock (Online Supplementary Figure S2A). For total (normal+aberrant) HBB mRNA levels, no significant differences were detected at considerable variation between experiments (Online Supplementary Figure S2B). Notably, immunoblots also indicated reduced Actb-normalized Hba expression and thus reduced MEL-HBB differentiation after shRNA treatment (Figure 2E). This phenomenon is likely related to toxicity from high VCNs and unregulated shRNA expression.12

For evaluation of therapeutic activity in clinically relevant cells, Up, Mid, Mid2, Down, Scr and the GLOBE gene-addition vector were then also applied as lentiviral monotherapy to primary HBB-homozygous CD34 cells in culture and compared with Mock.13 Additionally, combination treatments of GLOBE with Mid (to exemplify both superior, central Mid and Mid2, shRNAs) and with Scr were performed. As a first key parameter of β-thalassemia pathology, erythroid differentiation was analyzed by cytocentrifugation, cell staining and treatment-blinded microscopic analysis (exemplified in Figure 3A). These analyses revealed significantly increased late-erythroid differentiation, scored as presence of orthochromatophilic erythroblasts and reticulocytes, after shRNA treatment, from 50.7% for Mock toward the 86.5% observed for normal controls (Figure 3B). All HBB-specific shRNAs and GLOBE+Mid gave significant increases over Mock (Figure 3C, black bars), with lower-level induction also by Scr, possibly owing to stress-induced erythropoiesis and HBB expression.14 Of note, at VCN 3.2±1.6, GLOBE monotherapy did not achieve significant correction of erythroid differentiation in HBB-homozygous samples, and GLOBE+Mid significantly outperformed GLOBE alone. Complementary microscopic scoring of hemoglobinization as separate measurement revealed that only the GLOBE+Mid combination treatment significantly increased hemoglobinization compared with Mock and GLOBE (Figure 3C, grey bars). As a second key parameter of β-thalassemia pathology, the ratio of β-globin to α-globin expression (HBB/HBA) was analyzed for transduced HBB-homozygous samples by reversed-phase high-performance liquid chromatography (HPLC), with the results matching those for the parameter of erythroid differentiation. Specifically, comparison of HBB/HBA with Mock showed significant increases for GLOBE+Mid by (50.81±17.85)%, Mid2 by (43.72±17.45)%, Mid by (27.72±15.61)%, GLOBE+Scr by (31.92±19.39)% and GLOBE by (25.00±13.01)% (in order of ascending P values, Figure 3D). Compared with GLOBE, HBB/HBA was significantly decreased for Mock and Scr, and was significantly increased for GLOBE+Mid. Total β-like globin chains varied considerably between experiments and were increased for all treatments, most highly for Mid2 by (42.57±23.56)% and GLOBE+Mid by (38.89±32.05%)(Online Supplementary Figure S4). Against high baseline levels of raw HBB/HBA ratios in culture (0.58 for Mock against 1.03 for Normal controls), Mid2 and GLOBE+Mid were the most effective treatments, reaching ratios of 0.82 and 0.79, respectively (Online Supplementary Figure S5). Of note, whereas GLOBE+Mid compared with GLOBE alone significantly improved lateerythroid differentiation, hemoglobinization and HBB/HBA protein ratios, GLOBE+Mid compared with Mid alone resulted in only slight and statistically insignificant increases.

Figure 3.Functional correction of primary HBB^{IVSI-110(G>A)}-homozygous CD34⁺ cells. Data from independent cultures representing four different HBB^{IVSI-110(G>A)}- homozygous patients (n=12 for Mock, n=10 for Mid and Scr, n=6 for Up, Down and each combination with GLOBE, n=5 for Mid2) for relative HBB levels, erythroid differentiation and hemoglobinization, scored 7 days after transduction. A. Representative microscopy images of cytocentrifuged histologically stained samples (scale bar 20 μm) and enlarged reference images of observed erythroid-lineage differentiation (scale bar 10 μm). Genotypes (in bold) and treatments (in italics) are indicated, B. Percentage of erythroid subpopulations across all experiments according to differential counting of cells (Mock n=4079, Up n=777, Mid n=2210, Mid2 n=452, Down n=724, Scr n=3346, GLOBE n=2873, GLOBE+Mid n=1695, GLOBE+Scr n=2051 and mock-treated normal samples n=847), with brackets combining phases for late-stage erythroid differentiation (orthochromatophilic erythroblasts, reticulocytes). C. Differential counting for late-stage erythropoiesis (black bars) and dianisidine-positive, hemoglobinized cells (grey bars), analyzed by Kruskal-Wallis test with Dunn’s multiple comparisons test. The tests were applied twice independently, once against Mock and once against GLOBE, as indicated by an arrow for the respective reference treatment and asterisks in corresponding color and shading for significantly different test samples. Mock vs.: Up ^**P=0.002046, Mid ^***P=0.0007217, Mid2 ^****P<9.0×10⁻⁶, Down ^**P=0.005604, Scr ^*P=0.01415, GLOBE+Mid ^****P<3.2×10⁻⁵, ^**P=0.006459, Normal ^****P<2.3×10⁻⁶, ^**P=0.001298; GLOBE vs.: Up ^*P=.01334, Mid ^**P=0.009452, Mid2 ^***P=0.0001369, Down ^*P=0.02607, GLOBE+Mid ^***P=0.0005801, ^**P=0.004081, Normal Mock ^****P=1.4×10⁻⁵, ^***P=0.0007011. D. HPLC results for the percentage change of HBB/HBA ratios, compared with Mock, analyzed by one-way ANOVA with Dunnett’s multiple comparison test. The tests were applied twice independently, once against Mock and once against GLOBE, as indicated by an arrow for the respective reference treatment and asterisks in corresponding color and shading for significantly different test samples. Mock vs.: Mid ^***P=.0004, Mid2 ^****P<0.0001, GLOBE ^**P=0.0085, GLOBE+Mid ^**** P<0.0001, GLOBE+Scr ^***P=0.0005; GLOBE vs.: Mock ^**P=0.0085, Scr ^*P=0.0419, GLOBE+Mid ^**P=0.0243. Red outlines: transduction with GLOBE and time-shifted co-transduction with shRNA-encoding vectors; Corresponding percentage changes for additional β-like globin chains and combinations and approximate β-like globin chain distribution for each treatment are shown in Online Supplementary Figure S4. Corresponding differentiation-corrected HBB expression levels, including additional cultured samples from normal donors, are shown in Online Supplementary Figure S5. Normal: healthy control sample; Mock: mock treatment without vector; Scr: treatment with vector encoding a scrambled shRNA; other column and lane labels specify LVs; Pro: proerythroblast; BEB: basophilic erythroblast; PEB: polychromatophilic erythroblast; OEB: orthochromatophilic erythroblast; Ret: reticulocyte.

Overall, shRNA-encoding LVs gave high VCNs with no consistent further HBB increases above VCN 5 for Mid and Mid2 and with marked cell death after transduction (Online Supplementary Figures S6 and S7), both in all likelihood because of efficient transduction with the comparably small shRNA-expressing LVs.12 For normal CD34 samples (n=2), Up, Mid and Mid2 gave variably increased HBB/HBA protein ratios, whereas Down reduced HBB/HBA (Online Supplementary Figure S8), a preliminary finding revealing that Mid and Mid2 do not interfere with HBB expression from normal loci.

This study establishes aberrant HBB mRNA as a partially dominant causative agent of disease severity in HBB thalassemia and as a potent target for mutation-specific gene therapy. High titers of shRNA-encoding vectors and constitutive expression were applied, whereas for potential clinical translation, moderate VCN and erythroid-specific shRNA expression will be required.8 The specific mechanism of shRNA-mediated HBB induction is under investigation, with findings in MEL-HBB cells for HBB RNA and protein levels, and in CD34 cells for HBB protein levels suggesting differential modes of action for Down compared with Mid and Mid2 shRNAs (see Online Supplementary Discussion). In comparison with LV HBB addition and at levels of correction in HBB-homozygous CD34 cells similar to or higher than those for GLOBE, application of the smaller shRNA-encoding vectors offers up to ten-fold higher vector yield (Online Supplementary Table S2) and may thus give more patients access to treatment. Beyond HBB and many similar thalassemia-causing mutations (see IthaGenes)2 with splice defects, hundreds of genetic diseases (see DBASS3/5)15 are associated with aberrant transcripts, whose stability and causative role in disease pathology mostly remain to be investigated. The novel approach of shRNA treatment against aberrant mRNA is thus potentially suitable for a range of disorders. Here, it proved effective as monotherapy in primary HBB- homozygous CD34 cells in culture, and in combination treatment significantly improved upon gene therapy by HBB addition.