In vivo haemopoietic stem cell gene therapy enabled by postnatal trafficking

Study design

Sample size in experiments with mice was chosen according to previous experience with experimental models and assays. No sample or animal was excluded from the analyses. Mice were randomly assigned to each experimental group. Investigators were blinded at the moment of LV administration to newborn ADA-SCID or OC mice. In all the other experimental settings investigators were not blinded.

Plasmid construction

Plasmids pCCLsin.cPPT.PGK.GFP, pCCLsin.cPPT.PGK.ADA, pCCLsin.cPPT.PGK.TCIRG1 and pCCLsin.cPPT.PGK.FANCA were previously described^14,19,29,30. Plasmid pCCLsin.cPPT.ET.ADA.142T was constructed by standard cloning techniques, by exchanging the human FIX cDNA⁷ with human ADA cDNA (SpeI–SpeI) into the previously described pCCLsin.cPPT.ET.coFIX-R338L.142T (NheI–SalI) after blunting the ends. Plasmid pCCLsin.cPPT.PGK.ADA.122T was constructed by standard cloning techniques, by inserting the miRNA-122 target sequence into the pCCLsin.cPPT.PGK.ADA plasmid (SalI–ScaI).

Vector production

Third-generation self-inactivating LVs were produced by calcium phosphate transient transfection into 293T cells. 293T cells were transfected with a solution containing a mix of the selected LV genome-transfer plasmid, the packaging plasmids pMDLg-pRRE and pCMV.REV, pMD2.G and pAdVantage (Promega), as previously described⁷. In brief, the medium was changed 14–16 h after transfection and the supernatant was collected 30 h after medium change. LV-containing supernatants were sterilized through a 0.22-μm filter (Millipore), pre-concentrated four times with a Vivaflow 200 Tangential Flow Filtration Cassette 100 kDa (Sartorius), according to manufacturing instructions, transferred into sterile polyallomer tubes (Beckman) and centrifuged at 20,000g for 120 min at 20 °C (Beckman Optima XL-100K Ultracentrifuge). The LV pellet was dissolved in the appropriate volume of PBS to allow 500–1,000× concentration.

LV titration

For LV titration, 1 × 10⁵ 293T cells were transduced with serial LV dilutions in the presence of polybrene (8 μg ml⁻¹). For PGK.GFP-LV, cells were analysed by flow cytometry using a FACSCanto analyser (BD Biosciences), equipped with DIVA software, 5–7 days after transduction and the infectious titre, expressed as transducing units (TU) ml⁻¹, was calculated using the formula TU ml⁻¹ = ((% GFP⁺ cells/100) × 100,000×(1/dilution factor)). For all other LVs, genomic DNA (gDNA) was extracted 14 days after transduction, using the Maxwell 48 Cell DNA Purification Kit (Promega), following the manufacturer’s instructions. The VCN was determined by ddPCR, starting from 5–20 ng of template gDNA using primers (HIV FW: 5′-TACTGACGCTCTCGCACC-3′; HIV REV: 5′-TCTCGACGCAGGACTCG-3′) and a probe (FAM 5′-ATCTCTCTCCTTCTAGCCTC-3′) designed against the primer-binding region of LV. The amount of endogenous DNA was quantified by a primers–probe set designed against the human telomerase gene (Telo FW: 5′-GGCACACGTGGCTTTTCG-3′; Telo REV: 5′-GGTGAACCTCGTAAGTTTATGCAA-3′; Telo probe: VIC 5′-TCAGGACGTCGAGTGGACACGGTG-3′ TAMRA) or the human GAPDH gene (Applied Biosystems HS00483111_cm). The PCR reaction was performed with each primer (900 nM) and the probe (250 nM; 500 nM for Telo) following the manufacturer’s instructions (Bio-Rad), read with a QX200 reader and analysed with QuantaSoft software (Bio-Rad). The infectious titre, expressed as TU ml⁻¹, was calculated using the formula TU ml⁻¹ = (VCN × 100,000×(1/dilution factor). LV physical particles were measured by HIV-1 Gag p24 antigen immunocapture assay (Perkin Elmer) following the manufacturer’s instructions. LV-specific infectivity was calculated as the ratio between infectious titre and physical particles.

VCN determination

For mouse experiments, DNA was extracted from the whole liver using the Maxwell 48 Tissue DNA Purification Kit (Promega), from a cell suspension obtained from the thymus, spleen or BM using the Maxwell 48 Cell DNA Purification Kit (Promega) or from whole PB using the Maxwell 48 Whole Blood DNA Purification Kit (Promega). DNA was extracted from lineage-negative (Lin⁻) cultured cells using the Maxwell 48 Cell DNA Purification Kit (Promega) and from single-picked colonies using Quick extract DNA (Biosearch Technologies), according to cell number. The VCN in mouse DNA was determined by ddPCR, starting from 5–20 ng of template gDNA using a primers–probe set designed against the primer-binding region of LVs (see ‘LV titration’ section). In some cases, the VCN was determined using an ad hoc ddPCR (QX200 EvaGreen Digital PCR Supermix, Bio-Rad), which selectively amplifies the reverse-transcribed vector genome (both integrated and non-integrated), discriminating it from possible plasmid contamination²⁹ (RT-LV; ΔU3 FW: 5′-TCACTCCCAACGAAGACAAGATC-3′, gag REV: 5′-GAGTCCTGCGTCGAGAGAG-3′). The amount of endogenous mouse DNA was quantified by a primers–probe set designed against the mouse Sema3a gene (Sema3a FW: 5′-ACCGATTCCAGATGATTGGC-3′; Sema3a REV: 5′-TCCATATTAATGCAGTGCTTGC-3′; Sema3a probe: HEX 5′-AGAGGCCTGTCCTGCAGCTCATGG-3′ BHQ1). The PCR reaction was performed with each primer (900 nM; 150 nM for RT-LV primers) and the probe (250 nM) following the manufacturer’s instructions (Bio-Rad), read with a QX200 reader and analysed with QuantaSoft software (Bio-Rad).

RNA extraction and ddPCR

RNA extraction was performed using a Maxwell 48 SimplyRNA Blood extraction kit (Promega, AS1380), according to the manufacturer’s instructions and reverse transcribed using the SuperScript IV VILO kit (11766050; ThermoFisher Scientific). LV gene expression was assessed by ddPCR, starting from 10–25 ng of template cDNA using a primers–probe set designed against the Woodchuck hepatitis virus post-transcriptional regulatory element (Wpre) sequence of the LV genome (Wpre FW: 5′-GGCTGTTGGGCACTGACAAT-3′; Wpre REV: 5′-ACGTCCCGCGCAGAATC-3′; Wpre probe: FAM 5′-TTTCCATGGCTGCTCGCCTGTGT-3′ MGB). Hprt was used as a reference gene for mouse samples (dMmuCPE5095493, Bio-Rad). The PCR reaction was performed with each primer (900 nM) and the probe (250 nM) following the manufacturer’s instructions (Bio-Rad), read with a QX200 reader and analysed with QuantaSoft software (Bio-Rad).

Cell cultures and in vitro experiments

The 293T cell line was maintained in Iscove’s modified Dulbecco’s medium (IMDM, Sigma) supplemented with 10% fetal bovine serum (FBS; FetalClone II, HyClone, Euroclone), 4 mM glutamine (Lonza), 100 international units (IU) ml⁻¹ penicillin and streptomycin (Lonza). Primary mouse Lin⁻ cells were FACS-sorted from cell suspensions obtained from the BM or the liver of LV-treated mice and cultured in StemSpan supplemented with 10% FBS, 4 mM glutamine, 100 IU ml⁻¹ penicillin and streptomycin, 100 ng ml⁻¹ mouse stem-cell factor (mSCF, Peprotech), 50 ng ml⁻¹ mouse thrombopoietin (mTPO, Peprotech), 100 ng ml⁻¹ mouse fms-like tyrosine kinase 3 ligand (mFLT3L, Peprotech) and 20 ng ml⁻¹ mouse interleukin-3 (mIL-3, Peprotech) for 24 h. The medium was then exchanged with IMDM supplemented with 10% FBS, 100 IU ml⁻¹ penicillin and streptomycin, 100 ng ml⁻¹ mSCF (Peprotech) and mFLT3L 100 ng ml⁻¹ (Peprotech) for 48 h. Finally, GFP expression was determined using a FACSCanto analyser (BD Biosciences), equipped with DIVA software. Some of the sorted Lin⁻ cells were plated (800 cells per plate) in classic Methocult (STEMCELL Technologies) supplemented with 100 ng ml⁻¹ mSCF (Peprotech), 50 ng ml⁻¹ mTPO (Peprotech), 100 ng ml⁻¹ mFLT3L (Peprotech) and 20 ng ml⁻¹ mIL-3 (Peprotech). All cells were maintained in a 5% CO₂ humidified atmosphere at 37 °C. Cell lines were routinely tested for mycoplasma contamination.

Mice experiments

Founder FVB,129-Adatm1Mw Tg(PLADA)4118Rkmb/J mice (referred to as ADA-SCID mice) were obtained from The Jackson Laboratory³¹ (stock 003265). Founder B6C3Fe a/a-Tcirg1oc/J mice (referred to as OC mice) were obtained from The Jackson Laboratory¹⁶ (stock 00230) and fully backcrossed on the C57BL/6 background. Founder mice with a targeted disruption in the Fanca gene³² were provided by H. J. van de Vrugt (Free University Medical Center) and backcrossed with FVB mice to obtain the Fanca^−/− FVB strain (referred to as Fanca^−/− mice). C57BL/6 and NSG mice were purchased from Charles River Laboratories. All mice were maintained in specific-pathogen–free conditions. Genotyping of ADA-SCID and OC mice was performed at 2–3 weeks of age according to the protocol available on The Jackson Laboratory website. Vector administration was carried out in adult female C57BL/6 or NSG (7–10-week–old) mice by tail-vein injection (250–300 µl per mouse), in newborns by temporal-vein injection (30 µl per mouse) and in 2-week-old mice by retro-orbital injection (100–150 µl per mouse). Mice were bled from the retro-orbital plexus using capillary tubes and blood was collected in 0.38% sodium citrate buffer, pH 7.4 or in an EDTA-coated tube (Microvette 20.1341). For transplantation experiments, 7-week-old female C57BL/6 recipient mice were conditioned with a lethal regimen of busulfan (6 mg ml⁻¹), consisting of four daily intraperitoneal (i.p.) administrations of 27 µg g⁻¹ of busulfan (prediluted 1:2 in 0.9% NaCl injectable solution). The transplant was performed by i.v. injection of total BM cells or total CD45⁺ cells FACS-sorted (BD FACSAria Fusion Cell Sorter, BD Biosciences or MoFlo Astrios EQ Cell Sorter, Beckman Coulter) from the liver or the BM of donor mice the day after the last dose of busulfan. Cells were resuspended in 200 µl of PBS. The type and number of transplanted cells are indicated in the experimental schemes. The weight of recipient mice was monitored daily during busulfan administration and weekly for 1 month after transplant. In case of transplantation experiments in NSG mice, conditioning was performed by sublethal irradiation of recipient mice (200 rad) and transplantation 3–4 h after irradiation. For the transplantation experiments in Fanca^−/− mice, conditioning was performed by lethal irradiation of recipient mice using two doses of 4.5 Gy 24 h apart. A total of 3–4 h after conditioning, 4–5 × 10⁶ BM cells from LV-treated or untreated Fanca^−/− mice were transplanted by i.v. injection in the tail vein of recipient mice. HSPC mobilization was performed in 2-week-old or adult mice by administering five i.p. doses of 250 ng g⁻¹ G-CSF per mouse 12 h apart. Then 8 h after the last dose of G-CSF, 2 µg g⁻¹ plerixafor per mouse was administered i.p. to mobilized mice. LVs were administered i.v. 2 h after plerixafor and, at the same time, haematopoietic cells were collected from the BM, PB or spleen for FACS analysis. HSPC mobilization was performed in newborn mice following the same protocol used in 2-week-old or adult mice, except for a reduced G-CSF regimen that consisted of three instead of five doses. Anti-IFNα receptor-blocking antibody (MAR-1, InVivoMAb, MAR1-5A3) or IgG control (InVivoMAb, mouse IgG1 Isotype control, MOPC-21) were administered i.v. to newborn mice 3 h before LV administration at 50 mg kg⁻¹ dose. BM failure was induced in Fanca^−/− mice by i.p. administration of two doses of 0.3 mg kg⁻¹ of mitomycin C (MilliporeSigma), spaced 7 days apart, 6 weeks after LV injection²⁰. Mice were euthanized by CO₂ inhalation at the scheduled times. Experimental procedures in Fanca^−/− mice were conducted according to European and Spanish regulations (European convention ETS 123, regarding the use and protection of vertebrate mammals used in experimentation and other scientific purposes, Directive 2010/63/UE, Spanish Law 6/2013 and Real Decreto (R.D.) 53/2013 regarding the protection and use of animals in scientific research). All animal procedures were performed according to protocols approved by the Institutional Animal Care and Use Committee.

Fractionation of liver, thymus, spleen and BM

The mouse liver was perfused (15 ml min⁻¹) through the inferior vena cava with 12.5 ml of the following solutions at subsequent steps: (1) PBS EDTA (0.5 mM); (2) HBSS (Gibco) and HEPES (10 mM); and (3) HBSS and HEPES containing 0.03% collagenase IV (Sigma). The digested mouse liver tissue was collected, passed through a 100-μm cell strainer (BD Biosciences) and processed into a single-cell suspension. This suspension was subsequently centrifuged three times (30g, 25g and 20g, for 3 min, at room temperature) to remove hepatocyte-containing pellets. The supernatant containing non-parenchymal cells was centrifuged (650g, 7 min, at room temperature) and recovered cells were subsequently incubated with the monoclonal antibodies indicated in the ‘Flow cytometry’ section. To obtain a single-cell suspension of spleen or thymus, organs were smashed on a 40-µm filter using PBS–2% FBS (FetalClone II, HyClone, Euroclone), red blood cells (RBCs) were lysed with 0.5 or 1 ml of ACK Lysing buffer (GIBCO) for 5 or 10 min on ice, respectively. The lysis was stopped by adding 10 ml of PBS–2% FBS. The cell suspension was then filtered through a 40-µm filter and cells were counted and used for gDNA extraction or flow-cytometry analysis (see ‘Flow cytometry’ section). BM cells were flushed from the tibia and femur of one or two legs, as needed, or extracted by bone crushing in the case of newborn or OC mice and then processed as described for cells isolated from the spleen.

Flow cytometry

Flow-cytometry analyses using cell suspensions obtained from BM, spleen, thymus, PB or liver were performed using a FACSCanto analyser (BD Biosciences), equipped with DIVA software. Between 500,000 and 1,000,000 cells were processed, washed with PBS–2% FBS (FetalClone II, HyClone, Euroclone) or MACS buffer (PBS pH 7.2, 0.5% bovine serum albumin (BSA), 2 mM EDTA), treated with fragment crystallizable (Fc) Receptor-Block (Miltenyi Biotec) when antibody-stained and then resuspended in the buffer used for washing. Staining was performed in MACS buffer, incubating cells with antibodies (in the proportion indicated in Supplementary Table 15) for 20 min at 4 °C in the dark. Right before analysis, 3 µl of 7-AAD (Biolegend 420404) was added to each tube to exclude dead cells. Rainbow beads (BD Biosciences) were used to calibrate the instrument detectors. Gating strategies are provided in Supplementary Figs. 1–4. A BD FACSAria Fusion Cell Sorter (BD Biosciences), equipped with DIVA software (BD Biosciences), or a MoFlo Astrios EQ Cell Sorter (Beckman Coulter), equipped with Summit software (Beckman Coulter), was used for cell sorting.

For PB samples collected from healthy human participants of different ages and mouse samples shown in Fig. 1b–i and Extended Data Fig. 1a–d, samples were analysed using two standardized multiparametric flow-cytometry assays^21,33 (through whole blood dissection). In brief, after RBC lysis with ACK lysing buffer (STEM CELL Technologies), samples were labelled with the fluorescent antibodies reported in Supplementary Table 16. Titration assays were performed to assess the best antibody concentration. Samples from mice were pre-incubated with a rat anti-mouse FcR blocking reagent (BD, dilution 1:100). After surface marking, cells were incubated with propidium iodide (PI; BioLegend) to stain dead cells. Absolute cell quantification was performed by adding Precision count beads (BioLegend) to PB samples before the staining procedure. Images of all stained samples were acquired through a BD Symphony A5 (BD Bioscience) cytofluorimeter after calibration with Rainbow beads (Spherotech) and raw data were collected using the DIVA software (BD Biosciences). Data were subsequently analysed with FlowJo software v.10.5.3 (BD Biosciences). A list of antibodies used for mouse or human whole-blood-dissection panels is reported in Supplementary Table 16.

PB samples from paediatric and adult healthy donors were collected and analysed after written informed consent was given, using protocols approved by the ethics committees of IRCCS Ospedale San Raffaele (Protocol_TIGET09 for both paediatric and adult participants). Written consent for paediatric participants was given by their parents.

Immunohistochemistry imaging

Mice were perfused through the left ventricle with PBS containing 0.5 mM EDTA to remove blood, after which organs were collected and fixed in zinc-formalin for at least 24 h. Embedding, sectioning, slide preparation, staining and image acquisition were performed at the Centro di Imaging Sperimentale facility at San Raffaele Hospital (Italy). Formalin-fixed, paraffin-embedded consecutive sections (4 µm) were dewaxed and rehydrated through a graded alcohol series before histological or immunohistochemical staining. Immunostaining was performed using the Automatic Leica BOND RX system (Leica Microsystems). Tissues were first deparaffinized and subjected to epitope retrieval using Epitope Retrieval Solution 1 (ER1 Citrate Buffer). Slides were incubated with either an anti-GFP primary antibody (Invitrogen, A11122, 1:1,000) or an anti-Ki-67 primary antibody (D3B5, Cell Signaling Technology, 1:200) for 30 min at room temperature, followed by detection using the Bond Polymer Refine Detection Kit (Leica, DS9800). Images of stained slides were acquired using an Aperio AT2 digital scanner at 20× magnification (Leica Biosystems) and analysed with ImageScope software (Leica Biosystems).

Integration-site analysis

Integration sites were retrieved by Sonication Linker Mediated (SLiM)-PCR, as previously described³⁴, with minor modifications. In brief, for each sample up to 300 ng of gDNA was processed by DNA shearing using the Covaris E220 Ultrasonicator, generating fragments with an average size of 1,000 bp. The fragmented DNA was subjected to end repair and 3′ adenylation, and then ligated to linker cassettes containing an eight-nucleotide sequence barcode used for sample identification and all the sequences required for Read 2 paired-end sequencing. The ligation products were split in three technical replicates and subjected to 35 cycles of exponential PCR using primers specific for the LV LTR and the linker cassette. A subsequent amplification with an additional ten PCR cycles was performed using a primer specific for the linker cassette and the LTR. These primers contain an 8-nucleotide barcode used for sample identification (coupled with the barcode on the linker cassette) and the sequences needed for Read 1 sequencing, plus a random 12-nucleotide sequence, enabling easier cluster recognition in the first sequencing cycles on the next-generation sequencing system. The generated SLiM-PCR products are therefore associated with a unique pair of barcodes, assembled into libraries and subjected to Illumina next-generation sequencing.

Bioinformatic analysis

Sequencing reads were processed using a custom bioinformatics pipeline³⁵ (VISPA2) that isolates genomic sequences flanking the vector LTR and maps them to the mouse genome (mm9). Because vector integration in the same genomic position in different cells is a very low probability event, identical integration sites in independent samples were considered contamination or amplification artefacts, which may occur during the technical procedure. Datasets were pruned of potential contamination and false positives for each primary mouse and of integration sites derived from unrelated secondary mice. Identical integration sites shared between mice belonging to different experimental groups were reassigned by identifying the insertion site in at least two SLiM technical replicates and through sequence counts. Downstream analyses of vector-integration sites, such as relative abundance, sharing and CIS analysis, were performed using ISAnalytics³⁶. Clonal abundance, estimated as the relative percentage of genome numbers, was determined by the R package sonicLength³⁷. Common insertion sites were calculated using the Grubbs test for outliers³⁸.

Haemocytometric analysis

Haemocytometric analysis of PB was performed using a Procyte DX Analyser (IDEXX) at the time points indicated in the experimental design schemes. Approximately 30 µl of PB was collected into EDTA-coated tubes (Microvette 20.1341) and analysed. The measured parameters were: haemoglobin, total RBC count, differential count (lymphocyte, monocyte and neutrophil (absolute count and percentage)), platelet count and total leukocyte count (white blood cell).

ADA activity

ADA activity was measured in plasma or the cellular fraction of PB as previously described³⁹. In brief, 20 μl of packed RBCs were lysed in 20 μl of lysis buffer (0.25 mol l⁻¹ HEPES, 0.025 mol l⁻¹ CHAPS and 0.025 mol l⁻¹ dithiothreitol). The cell lysate was centrifuged at 12,000g for 5 min. The enzyme activity was evaluated in a mixture containing 50 mmol l⁻¹ Tris (pH 7.2) and 0.4 mmol l⁻¹ adenosine by adding 10 μl of lysed RBC supernatant or 20 μl of plasma. The reaction was stopped by adding PCA (final concentration, 0.21 mmol l⁻¹) and neutralized with KOH (final concentration, 0.22 mmol l⁻¹). The reaction progression was analysed using aliquots at two time points with a Bio-Rad Biofocus 3000 instrument (Bio-Rad) after inosine plus hypoxanthine formation at 0 and 10 min for the cellular fraction or at 0 and 30 min for plasma. The enzyme activity was measured as mol h⁻¹ per (ml packed RBCs or ml plasma).

Osteoclast differentiation, CTX and parathyroid hormone ELISA and dentine assay

Osteoclasts were differentiated from BM cells as previously described⁴⁰ and assayed for resorption functions on dentine slides (AE8050). Dentine discs were stained with toluidine blue as previously described⁴⁰. The concentration of released CTX in the culture supernatant was measured by enzyme-linked immunosorbent assay (ELISA; CrossLaps, AC-07F1, Immunodiagnostic System) following the manufacturer’s protocol. The concentration of parathyroid hormone in the serum of LV-treated OC or WT mice was measured by ELISA (MicroVue Mouse parathyroid hormone 1-84 EIA) following the manufacturer’s protocol.

Micro-computed tomography

For micro-computed tomography analyses, vertebral columns of OC mice and their WT littermates were dissected, cleaned from connective tissues and fixed in 4% paraformaldehyde. Bones were scanned at a nominal image resolution of 7 µm using a SkyScan 1276 (Bruker) at 55 kV and 72 µA, using a 0.25-mm aluminium filter. Images were captured every 0.4° in a rotation to 360° with a 2 × 2 camera pixel binning. The reconstruction was carried out with the SkyScan NRecon 2.0.0.5 software using the InstaRecon reconstruction engine, ring artefact reduction and beam-hardening correction. Image analysis of the L5 vertebra was performed using the CT Analyser software (CTAn 1.18.8.0+). An in-house-developed semiautomatic approach was used to segment the trabecular region of the vertebral body in the coronal plane. The bone mineral density of the trabecular region was calculated by calibrating a linear extrapolation against the mean attenuation coefficients of selected regions of a 2-mm phantom rod pair containing 0.25 and 0.75 g cm⁻³ CaHA (Skyscan). A three-dimensional morphometric analysis of the trabecular region was performed after binarization of the trabeculae by an adaptive thresholding algorithm. The ratio between the bone volume and total volume (%) and the trabecular separation (mm) were assessed.

Statistical analyses

Statistical analyses were performed after consulting with professional statisticians at the San Raffaele University Center for Statistics in the Biomedical Sciences (CUSSB). When normality assumptions were not met, nonparametric statistical tests were performed. A two-tailed Mann–Whitney U-test was performed to compare two independent groups and for more than two independent groups, a Kruskal–Wallis test was used, followed by post hoc analysis (Dunn’s test for multiple comparisons against the reference control group along with Bonferroni’s correction). For paired observations, a Wilcoxon matched-pairs test was performed. Data were analysed using GraphPad (v.10). To model the dynamics of biological processes, properly accounting for dependencies induced by the repeated-measures design, LME models were estimated, specifying random-effect terms for experimental units or session. To fulfil underlying model assumptions, various transformations of the outcome variable, including logarithmic, square/cubic root and ordered quantile normalization, were considered. After the estimation of LME models, a post hoc analysis was conducted, enabling the pairwise comparison of treatment groups at a fixed time point. To account for multiplicity issues, P values were adjusted using Bonferroni’s approach. LME were estimated using R software (v.4.3.1). In all the analyses, the significance threshold was set at 0.05. Inferential techniques were used for adequate sample sizes (n ≥ 5), otherwise only descriptive statistics are reported.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.