Human iPSCs were generated from healthy human newborn foreskin fibroblasts (NuFFs) C.3 using the CytoTune iPS Sendai Reprogramming kit (Thermo Fisher Scientific) (19). IPSCs were cultured on Vitronectin XF™ (STEMCELL Technologies) in StemMACS™ iPSC Brew XF (Miltenyi) with daily medium change.

Murine stroma cells OP9-DL1 were used as feeders for the differentiation into NK cells. OP9-DL1 were cultured in alpha-MEM medium (without nucleosides) supplemented with 20% heat-inactivated (HI) (56°C for 30 min) fetal bovine serum (FBS) (Biowest™), 1% non-essential amino acids (NEAAs), 1% Glutamax, and 100 U of penicillin and streptomycin (P/S) (Thermo Fisher Scientific).

The murine mastocytoma FcR⁺ cell line P815 was used for functional NK cell studies. P815 was cultured in RPMI supplemented with 20% HI FBS (PAN™ serum), 1% NEAA, 1% Glutamax, and 100 U/ml of P/S.

The human K562 cell line (DMSZ, ACC Nr. 10) was used as target cells for cytotoxicity assays. K562 was cultured with RPMI supplemented with 20% HI FBS (PAN™), 1% NEAA, 1% Glutamax, and 100 U/ml of P/S.

A RAG-fate mapping reporter was introduced into the AAVS1 locus of human iPSC using CRISPR/Cas9 gene editing. Subsequently, iPSCs were investigated for bi- (RSS–EGFP^+/+) or monoallelic (RSS–EGFP^+/−) integrations of the reporter construct. We used the vector system developed in the Zhang laboratory (20, 21) (Addgene^® 80945 AAVS1-Pur_CAG-EGFP) and replaced the EGFP cassette with an inverted EGFP cassette flanked by two RSS sequences (18, 22). Cloning was performed using the Fusion™ HD Transformation Kit Eco Dry (Takara). Targeting of the inverted EGFP cassette by the nucleases RAG1 and RAG2 results in flipping of the cassette into sense orientation ( Figure 1A ). CRISPR/Cas9 protein and gRNA were transfected into iPSCs by nucleofection (Amaxa Kit V, Roche). Transfected iPSC clones were selected using 1 ng/µl of puromycin (Thermo Fisher Scientific). Stable vector integration was confirmed by PCR ( Supplementary Figure S1A ) and Sanger sequencing, and the number of integrations was determined by RT-PCR ( Supplementary Figure S1B ). Pluripotency and karyotypic integrity were confirmed in generated reporter iPSC lines ( Supplementary Figures S1C–E ) as described in the supplementary methods.

iPSCs were cultured in six-well plates coated with Vitronectin XF™ (STEMCELL Technologies) to 60%–100% confluency. Colonies were cut into squares using the StemPro™ EZPassage™ Disposable Stem Cell Passaging Tool (Thermo Fisher Scientific), transferred into one well of a six-well ultra-low attachment plate, and incubated overnight. Cell clusters developed into EBs of different sizes. To induce differentiation toward HSPCs, the STEMdiff™ Hematopoietic Kit (STEMCELL Technologies) was used according to manufacturer’s instructions. A maximum of five to seven EBs were transferred into a well of a 12-well plate coated with Vitronectin XF™ and containing 1 ml of differentiation medium A (1 ml/12 well). On day 12, the suspension cells were harvested and passed through a 40-μm cell strainer. No positive selection of HSPCs was performed prior to NK cell differentiation.

HSPCs were differentiated into NK cells on OP9-DL1 feeder cells as described before (19). OP9-DL1 feeder cells/well (1 × 10⁵) were seeded into 12-well culture plates and treated with 10 mg/µl of mitomycin C (MMC) (STEMCELL Technologies) for 3 h on the following day. HSPCs (1 × 10⁴) were plated on feeder cells and cultured in 2 ml of NK cell differentiation medium [56.6% DMEM high glucose (Thermo Fisher Scientific), 28.3% Ham’s F-12 Nutrient Mix (Thermo Fisher Scientific), 15% heat-inactivated human AB serum (Valley Biomedical), 2 mM L-glutamine (Thermo Fisher Scientific), 25 μM β-mercaptoethanol (Thermo Fisher Scientific), 5 ng/ml of sodium selenite (Sigma-Aldrich), 50 μM ethanolamine (Sigma-Aldrich), 20 mg/l of L-ascorbic acid (Sigma-Aldrich), 1% NEAA, 1% P/S supplemented with 10 ng/ml of recombinant human stem cell factor (SCF), 10 ng/ml of recombinant human IL-7, 10 ng/ml of recombinant human IL-15, 10 ng/ml of recombinant human Flt3 ligand, and 10 ng/ml of recombinant human IL-3 (PeproTech). IL-3 was only used until day 9. Half medium changes were performed on days 3, 6, 9, 13, and 16. On days 6 and 13, progenitor cells were replated on fresh feeder cells. On days 6 [week 1 (w1)], 13 [week 2 (w2)], and 20 [week 3 (w3)], cells were analyzed for surface marker expression using flow cytometry. OP-DL1 stroma cells were separated from NK suspension cells by filtering through a 40-µm strainer.

For flow cytometry, 1 × 10⁴–1 × 10⁵ cells were collected, washed twice with PBS, and stained with antibody mixes at concentrations of 1:100 in 100 µl of staining buffer (PBS, 1% FCS) for 15 to 30 min at room temperature (RT) in the dark. Antibodies and isotypes used are listed in Supplementary Table S1 . For intracellular staining of perforin, granzyme B, and interferon gamma, Cytofix/Cytoperm (BD Biosciences) was used for fixation and permeabilization according to the manufacturers’ instructions. The FoxP3 Staining Buffer Set (Miltenyi Biotec) was used for permeabilization of iPSCs characterized for pluripotency and trilineage differentiation. Cells were subsequently stained with antibody mixes at concentrations up to 1:100 in 100 µl of Perm Wash or FoxP3 Staining Buffer, respectively, for 30–60 min at 4°C in the dark. After two washes, cells were resuspended in 50–100 µl of PBS or staining buffer and analyzed on a BD FACSAria I™ or BD FACSAria III™ Cell Sorter (BD Biosciences), respectively.

To assess degranulation, cytokine production, and antibody-dependent cellular cytotoxicity (ADCC), NK cells generated on OP9-DL1 cells were harvested after week 3 and co-incubated with human K562 erythroleukemia tumor cells (E:T of 3:1 and 6:1), or mouse P815 mastocytoma cells (E:T of 3:1 and 6:1), respectively. The latter were either uncoated or coated with anti FCyRIIIa (a-CD16, 10 µg/µl; Biolegend) or isotype control (10 µg/µl) for 1 h. Prior to effector cell co-incubation, target cells were labeled using the CellTrace™ Far Red Cell Proliferation Kit (1 µg/ml, Thermo Fisher Scientific) for 20 min.

DNA damage was induced in NK cells obtained at weeks 1, 2, and 3 using 2 Gy of ionizing radiation from a Cs134 source. Cells were fixed and permeabilized after 2, 4, 8, and 24 h using solution A of Fix & PERM (Thermo Fisher Scientific) mixed 1:1 (w/w) with culture medium. After 3 min of incubation at room temperature, 2 ml of ice-cold methanol was added. In addition to samples obtained at indicated time points, an unirradiated control and an isotype control obtained 2 h after radiation were prepared. Subsequently, cells were stained with APC anti-H2A.X-phosphorylated (Ser139) (Biolegend), or APC mouse IgG1, K isotype control (Biolegend) for 1 h in the dark.

NK cells harvested at week 3 of differentiation were screened for V(D)J recombination events at the IGH, IGK, IGL, TCRB, TCRG, and TCRD loci using a multiplex PCR reported before (23). Genomic DNA was isolated from sorted GFP⁺ and GFP⁻ NK progenitor cells using the QIAamp DNA Micro kit (Qiagen).

To study V(D)J joints in the IGH locus of NK progenitor cells by next-generation sequencing (NGS), 1,000 GFP⁺ NK cells (derived from RSS–EGFP^+/+ iPSC) were sorted from differentiation cultures at weeks 1, 2, and 3, and gDNA was isolated. Three replicates were prepared for each sample that was obtained from three independent experiments. NGS of the IGH repertoire was performed by Adaptive Biotechnologies, and analysis tools provided by Adaptive were used. Heat maps were generated using the open-access webtool Morpheus provided by the Broad Institute, Cambridge, MA, USA (Morpheus, https://software.broadinstitute.org/morpheus).

For RNA sequencing analysis, 500 GFP⁺ and GFP⁻ NK progenitor cells (derived from RSS–EGFP^+/+ iPSC) were sorted on days 19 (week 1), 26 (week 2), and 33 (week 3) into 25 µl of NEBNext^® cell lysis buffer (New England Biolabs). Three replicates from three independent experiments were analyzed for each sample and time point.

Synthesis of cDNA, library preparation, and sequencing were performed by the c.ATG genomics core facility at the University Hospital Tuebingen, Tuebingen, Germany, and bioinformatic data analysis was performed with the bioinformatics core facility QBiC (University of Tuebingen, Tuebingen, Germany). RNA sequencing and the bioinformatic analysis are described in detail in the supplementary section.

GO enrichment analysis was performed using the Gene Ontology Resource (https://geneontology.org/) (GO Ontology database DOI: 10.5281/zenodo.12173881 Released 2024-06-17), PANTHER™ (version 19, released 2024-06-20), and Reactome (version 86, released 2023-09-07) (24–26). Results can be retrieved from Supplementary Table S2 . Genes associated with B (GO_0030183) and NK lymphocyte (GO_0001779) development were identified using Ensembl BioMart (Ensembl 111) (https://www.ensembl.org/info/data/biomart/index.html).

Each experiment was repeated at least three times, and means were used for statistical evaluation. Data were analyzed using GraphPad Prism 9.0 Software. The type of statistical analysis performed and the subset of the study population are listed in the figure legends. Statistical significance is indicated by p-values (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). The web-based platforms Galaxy (27) and SRplot (28) were used for visualization of transcriptomic data. Dot plots were generated using the MATLAB^® Engine API for Python^®.

To study the expression of RAG recombinase in NK cell development, we introduced a reporter cassette consisting of an inverted EGFP sequence flanked by two recombination signal sequences (RSS) (18) into the AAVS1 locus of human induced pluripotent stem cells (iPSCs) using CRISPR/Cas9 (20, 21) ( Figure 1A ). Efficient targeting of RSS by RAG1 and RAG2 results in flipping of the EGFP cassette into sense orientation and permanently marks the ontogeny of RAG expression. Cells with successfully integrated reporter constructs were selected by culture with puromycin (1 ng/µl) and subsequently screened for stable reporter integration using PCR ( Supplementary Figure S1A ), Sanger sequencing, and copy number variation (CNV) determined by qPCR ( Supplementary Figure S1B ). iPSC lines were characterized for their karyotypic integrity and retained pluripotency by expression of pluripotency-associated genes and proteins ( Supplementary Figures S1C, D ). In addition, iPSCs were differentiated into all three germ layers confirmed by surface expression of SOX2 and PAX6 (ectoderm), SOX17 and CD184 (CXCR4) (endoderm), and CD144 (VE-cadherin) and CD140b (mesoderm) ( Supplementary Figure S1E ). To validate the efficiency of the reporter construct, RAG1 and RAG2 mRNAs were introduced into iPSCs and HSPCs using nucleofection. As reported before (29), iPSCs were not able to perform V(D)J recombination, but efficient rearrangement of the reporter cassette was observed in HSPCs ( Supplementary Figure S1F ). Based on these investigations, iPSC lines containing bi- (RSS-EGFP^+/+) and monoallelic (RSS-EGFP^+/−) integrations of the reporter construct were selected and further differentiated into HSPCs and NK cells on OP1-DL1 stroma cells in vitro. Expression of GFP and NK lineage markers determining the process of differentiation was monitored weekly using flow cytometry ( Supplementary Figure S1G ). For simplification, we designated cells obtained after completion of the hematopoietic differentiation protocol “HSPCs” and cells in the process of differentiation toward the NK lymphocyte lineage “NK cell progenitors” or “NK cells.”

Importantly, two populations of CD45- and CD56-expressing cells could be identified that were discriminated as CD45^bright and CD45^dim, and CD56^bright and CD56^dim, respectively ( Figure 1B ).

GFP expression could be detected in up to 14% of NK cell progenitors from day 3 of differentiation on OP9-DL1 stroma cells in fluctuating intensities ( Figure 1C ). Of note, no GFP expression could be observed in HSPCs. Additional transduction of RAG1 and RAG2 mRNAs at the stage of HSPCs increased the fraction of GFP⁺ cells in NK cell precursors (14%–16.1% at week 1, 10.5%–21.4% at week 2, and 7.3%–20% at week 3) ( Figure 1C ) in RSS–EGFP^+/+ and RSS–EGFP^+/− reporter iPSC lines, respectively. Most NK cells obtained in vitro expressed GFP⁻CD45^brightCD56^bright, but GFP⁺ NK cells were found in up to 90% of the CD45^dim NK cell population at all stages of differentiation ( Figure 1D , Supplementary Figure S2A ). These populations could be further discriminated by CD56 expression, whereby CD45^bright NK cells predominantly expressed CD56^bright, and CD45^dim NK cells, CD56^dim ( Figure 1E , Supplementary Figure S2B ).

The observation of CD45^bright and CD45^dim populations in GFP⁺ and GFP⁻ NK progenitors was further investigated by studying the CD45RA, CD45RB, CD45RC, and CD45RO isotype expression using flow cytometry ( Supplementary Figure S2C ). CD45RB and CD45RC expression was analyzed in CD45RA⁺, CD45RO⁺, CD45RA⁺RO⁺, and CD45RA⁻RO⁻ populations.

Peripheral blood NK cells isolated from three healthy buffy coat donors stained predominantly CD45^bright and expressed CD45RA in co-expression with CD45RB and CD45RC (CD45RA/RC/RB and CD45RA/RC) isotypes ( Supplementary Figures S2D, E ). Since different CD45 isotypes can be co-expressed (30), we analyzed all possible combinations in CD45^bright and CD45^dim HSPCs, and GFP⁻/GFP⁺ NK cell progenitors obtained at weeks 1, 2, and 3, respectively ( Supplementary Figures S3A, B ).

IPSC-derived HSPCs that were all GFP⁻ due to lack of recombination at this stage, expressed predominantly all CD45 isotypes (RA/RO/RB/RC) and CD45RA/RC/RB. On CD45^dim HSPCs, up to 16% CD45RB, or CD45RB/RC isotypes could be detected. No specific CD45 isotype could be identified in up to 20% of HSPCs suggesting a lack of lymphocyte commitment ( Supplementary Figures S3A–D ).

CD45^bright NK cells were characterized by CD45RA/RB/RC and CD45RA/RC in GFP⁻ and CD45RA/RO/RB/RC, and CD45RA/RO/RB in GFP⁺ NK cells, respectively. In contrast, CD45^dim NK cells obtained at weeks 1–3 expressed predominantly CD45RO, whereas co-expression of CD45RB could be identified on GFP⁺ cells ( Supplementary Figures S3A–D ).

In summary, CD45 isotype distribution differs in NK progenitors with and without RAG-fate ontogeny with a preference for CD45RO and CD45RB in GFP⁺ and CD45RA in GFP⁻ NK cells. The different CD45 isotype distribution results in the observation of CD45^dim and CD45^bright NK progenitors.

For technical reasons, the following studies were performed on CD45^bright and CD45^dim NK cells without consideration of the CD45 isotypes.

The maturation process of differentiating NK cells was monitored weekly for expression of NK lineage marker using flow cytometry ( Supplementary Figure S1G ). HSPCs were harvested after 12 days of hematopoietic differentiation and analyzed for CD34, CD43 (leucosialin), CD45, and CD15 surface marker expression ( Supplementary Figure S4A ). No GFP expression, but CD45^dim and CD45^bright populations, were detected at the stage of HSPCs that expressed mainly CD45^bright and up to 25% CD45^dim. An increased fraction of CD34⁺CD43⁻ cells was found in CD45^dim compared to that in CD45^bright HSPCs, which predominantly stained CD34⁺CD43⁺ ( Supplementary Figures S4B, C ).

From weeks 1–3 of the NK cell differentiation, expression of CD117 (c-kit), CD7, CD161, CD94, CD335 (NKp46), CD56, and CD16 was assessed on CD45^bright and CD45^dim NK cell progenitors ( Supplementary Figure S5A ). Expression of NK cell markers was compared between RAG-fate-mapped (GFP⁺) and non-mapped (GFP⁻) CD45^bright, CD45^dim, and total CD45⁺ precursors ( Figures 2A, B , Supplementary Figures S5B–D ). CD56 expression could be observed from week 2 with a phenotype of predominantly CD56^bright in GFP⁻ and CD56^dim in GFP⁺ NK cell precursors. Although CD117 expression was similar in both GFP⁻ and GFP⁺ precursors increasing from week 1 (mean 54% ± 9.4% in GFP⁻, mean 51.5% ± 16.3% in GFP⁺) to weeks 2 and 3 (mean 78.5% ± 4.7% in GFP⁻, mean 68.7% ± 6.1% in GFP⁺), there were significant differences between GFP^-CD45^bright and GFP⁺CD45^dim NK cells ( Supplementary Figures S5B–D ). CD161 and CD94, the latter of which participates in the heterodimeric complex of NK2G receptors, were both increasingly expressed from week 2 to 3, and to a significantly higher extent in GFP⁻ NK cells (RSS–EGFP^+/−). The natural cytotoxicity receptor (NCR) NKp46 (CD335) was observed from week 2 and expressed by up to 78% (±13.4%) GFP⁻ and 85% (±26.9%) GFP⁺ NK cells at week 3. Expression of CD16 accelerated from week 2 to 3 and could be detected on 8%–30% (±8.5%) of GFP⁻CD45^bright NK cells compared to 5%–19% (±8.0%) of GFP⁺CD45^dim cells at week 3. Of note, GFP⁻CD45^dim cells stained negative for NK lineage markers investigated in this study ( Figure 2B , Supplementary Figure S5C, D ).

NK lineage marker were further analyzed on CD56⁺ NK cells obtained at week 3, which were additionally evaluated for markers of activation (CD69, DNAM1, CD62L), maturity (CD57, CXCR1), activating cytotoxicity receptors (NCRs) NKp46 (NCR1, CD335), NKp44 (NCR2, CD336), NKp30 (NCR3, CD337), as well as activating (NKG2C, NKG2D) and inhibiting (NKG2A) transmembrane NKG2/CD94 receptors, and killer cell immunoglobulin-like receptors (KIRs) KIRDL2/L3 and KIR3DL1 ( Figures 2C–E , Supplementary Figure S6A–C ).

In contrast to GFP⁻ cells, CD56^dim RAG-fate-mapped NK cells were characterized by predominant expression of the senescence marker CD57. In addition, NKG2C and KIRs could be found on GFP⁺ NK cells, whereas these markers of terminal differentiation were completely absent on GFP⁻ cells. The NCRs NKp30, NKp44, and NKp46 were differentially expressed, with a predominance of NKp30 and NKp46 in GFP⁻, and NKp44 in GFP⁺ cells, respectively. This suggests a higher activation status in the latter, whereas NKp30 and NKp46 are expressed on resting cells (31). In addition, GFP⁺ NK cells were characterized by an elevated fraction of DNAM1⁺ NK cells ( Figures 2D, E , Supplementary Figure S6A–C ). Expression of most NK lineage markers was reduced in GFP⁺CD56^bright compared to that in GFP⁻CD56^bright cells and almost absent in GFP⁻CD56^dim NK cells ( Figure 2E , Supplementary Figures S6B, C ). Similar results were obtained for both reporter cell lines, RSS–EGFP^+/− and RSS–EGFP^+/+.

These results suggest accelerated differentiation and a more mature phenotype in GFP⁺ versus GFP⁻ NK precursor cells.

To address the question whether the phenotype observed in GFP⁺ NK cells can be attributed to targeting of RAG endonucleases, RAG1 and RAG2 mRNA was transfected into HSPCs derived from reporter iPSCs. The increased targeting rate of the reporter construct raised the percentage of GFP⁺ NK cells up to 21.4% ( Figure 1C ). As in cells mapped by endogenous RAG expression, CD45^dim cells were predominantly found in GFP⁺ NK cells ( Supplementary Figure S7A ). Furthermore, CD45^dim NK cells expressed mostly CD56^dim, whereas CD56^bright expression was reduced in RAG1-/RAG2-transfected GFP⁺ NK cells ( Supplementary Figures S7B, C ).

The phenotype of NK cells with induced RAG1/RAG2 expression was investigated regarding NK lineage marker expression and maturity at weeks 1–3 of the NK cell differentiation. Interestingly, phenotypic differences between GFP⁻ and GFP⁺, and CD45^bright and CD45^dim NK cells resembled NK cells mapped by endogenous RAG1/RAG2 expression ( Supplementary Figures S7D–F ). In contrast to RAG-fate-mapped NK cells, expression of CD94, NKp46, and CD56^bright accelerated in GFP⁻ cells from week 1–3.

Following 3 weeks of differentiation, NK lineage surface markers were investigated on CD56⁺ cells ( Supplementary Figures S8A–G ). The expression of DNAM1, CD57, CXCR1, and NKp44 was significantly increased in GFP⁺ compared to that in GFP⁻ NK cells, whereas elevated fractions of NKp30⁺ and NKp46⁺ were detected in the GFP⁻ population ( Supplementary Figures S8E, F ). NKG2C and KIR2DL2/3 could only be detected on GFP⁺ NK cells. Though expressed at low levels on both populations, CD16 was predominantly found on GFP⁻ NK cells.

In summary, these data show that additional transfection of RAG1 and RAG2 mRNA results in increased targeting of the reporter construct, which contrasts the differential phenotype toward increased maturity observed in RAG-fate-mapped compared to non-mapped NK cells.

Degranulation capacity quantified by expression of CD107a, as well as intracellular perforin, granzyme B, and interferon gamma (IFNγ), was investigated in mature iPSC-derived NK cells harvested at week 3 and cocultured with K562 cells at defined effector:target (ET) ratios using flow cytometry. To assess cytotoxicity and cell lysis of target cells, the latter were labeled using a cell tracer dye. Compared to GFP⁻ populations, GFP⁺CD56⁺CD16^+/− mature NK cells displayed increased degranulation capacity shown by CD107a expression in response to K562 coculture ( Figure 3A ). Perforin, but not granzyme B, expression differed moderately between GFP⁻ and GFP⁺ NK cells at 3:1 and 6:1 effector:target ratios ( Figures 3B, C ). Interestingly, GFP⁺ NK cells produced more IFNγ in response to stimulation with PMA/ionomycin compared to GFP⁻ NK cells ( Figure 3D ). The redirected antibody-dependent cytotoxicity (ADCC) was measured using anti-CD16-coated murine FcR⁺ P815 cells as target for iPSC-derived NK cells (32) ( Figure 3E ). Degranulation in response to coated P815 cells was increased in GFP⁺ compared to GFP⁻ CD56⁺CD16^+/− NK cell populations ( Figure 3F ).

These results demonstrate that iPSC-derived NK lymphocytes are functional, and RAG-fate-mapped NK cells have an increased potential for both degranulation and cytokine production.

NK progenitor cells obtained at weeks 1, 2, and 3 were irradiated with 2 Gy and fixed after 2, 4, 8, and 24 h. Mean fluorescence intensities (MFIs) of the DDR marker γH2AX were measured using flow cytometry. Phosphorylation of the histone protein H2AX (γH2AX) occurred shortly after induction of DNA damage in GFP⁻ NK cells and was followed by subsequent downregulation of γH2AX foci upon repair. However, γH2AX level remained constantly increased in GFP⁺ populations following IR ( Figures 3G , Supplementary Figure S9A ). This finding suggests persistent DNA damage signaling in RAG-fate-mapped NK cells due to senescence. Accordingly, cellular survival after IR was severely reduced in GFP⁺ in contrast to that in GFP⁻ NK cells ( Figure 3H , Supplementary Figure S9B ).

To verify targeting by RAG recombinases, we searched for rearrangements in heavy chains and kappa and lambda light chains of the immunoglobulin receptor, and rearrangements in the TCR beta, gamma, and delta locus in sorted GFP⁺ NK cells obtained after 3 weeks of differentiation (23) ( Supplementary Figure S10A ). Since rearrangements were observed on the IGH, but not on the TCR locus, the genomic IGH repertoire was studied in gDNA extracted from sorted GFP⁺ NK cells at weeks 1, 2, and 3 using NGS analysis. Three independent experiments were performed, and results were calculated on three replicates of each sample.

Up to 17 unique sequences and up to 6,825 total reads per sample were obtained for NK cells harvested at weeks 1–3 of differentiation. The mean ratios between unique and total sequences are shown for each time point in Figure 4A . As shown by the frequency of the top five clones ( Figure 4B ), the IGH repertoire was extremely clonal due to a few unique sequences. Further investigation of the diversity using indices, such as sample clonality, Shannon entropy, and Simpson clonality ( Figures 4C–E ), confirmed a clonal repertoire but increasing diversity in weeks 2 and 3. Importantly, none of the rearrangements found was productive. This was due to a high range of pseudogene usage ( Figure 4F ), or VJ and DJ rearrangements leading to out of frame sequences or stop codons within the open reading frame ( Figures 4G, H ). Clonal expansion is further indicated by the frequency of all VJ and DJ rearrangements as shown by heat maps ( Figure 4I , Supplementary Figure S10B ). The number of resolved V and D rearrangements found was only slightly higher than the number of unresolved sequences ( Supplementary Figure S10C ). There were no differences between the three time points regarding nucleotide losses at the joints ( Supplementary Figure S10D ) and N nucleotide additions ( Supplementary Figures S10E–G ). Despite low diversity, increased numbers of N1 and N2 nucleotide additions (>16 nt) were found in all samples ( Supplementary Figures S10F, G ), and a shifted distribution of the CDR3 length toward larger sequences ( Supplementary Figure S10H ) was observed.

We conclude from these findings that all IGH rearrangements found in NK cells and NK precursor cells are unproductive due to either pseudogene usage or disruption of the reading frame. Interestingly, rearrangements can be observed early (week 1) at differentiation, and the diversity of clones increases toward more mature NK cells.

Transcriptome analyses were performed using RNA-Seq on sorted GFP⁻ and GFP⁺ RSS–cEGFP^+/+ NK cell precursors obtained at weeks 1, 2, and 3 of differentiation. Due to potential contamination with gDNA, mapped sequences were filtered to contain gapped alignments and exon–exon boundaries only ( Supplementary Figures S11A–C ). Principal component analysis (PCA) of the filtered RNA-Seq data on GFP⁻ versus GFP⁺ NK cells from weeks 1, 2, and 3 showed two different clusters that further separated in the process of differentiation ( Figure 5A ). In addition, a similarity analysis between all samples revealed a high distance between NK progenitors at weeks 1 and 3 ( Supplementary Figure S11D ). Accordingly, the number of differentially expressed (DE) genes increased over time from 220 at week 1, to 395 at week 2, and to 439 at week 3 (p_adj < 0.05). The normalized gene counts for GFP⁻ and GFP⁺ NK cells of the top 10 DE genes of each time point are shown in Figure 5B . The distribution of DE genes obtained from weeks 1, 2, and 3 is represented by volcano plots in Figure 5C indicating the top 10 DE genes.

There was a substantial overlap of DE genes at weeks 1, 2, and 3 ( Figure 5D ). In total, 94 genes that were differentially regulated at all time points could be identified. Gene expression was analyzed in this data set, and DE genes relevant for NK cell differentiation were identified ( Figure 5E ). The distribution of significantly up- and downregulated genes at all time points (p_adj < 0.05) is shown in Figure 5F , in which the top 10 DE genes are indicated. Of note, most of the DE genes were associated with transcription, translation, and protein processing.

The expression profiles of the top 100 DE genes observed at the individual time points (weeks 1–3) are shown in Supplementary Figures S12A–C .

Gene ontology (GO) analysis of biological processes performed on DE genes of weeks 1–3 (FDR < 0.05) revealed differential regulation of gene expression, translation, and metabolic processes ( Figure 5G ). The GO terms associated with gene expression and translation, metabolism, and cellular function found to be enriched in NK progenitors obtained from week 1 to 3 are shown in Supplementary Figures S12D–F .

Pathway analysis using Reactome showed differences between RAG-fate-mapped and non-mapped NK cells regarding hematopoietic differentiation, DDR, and immune function ( Figures 6A–C ). Differentially regulated pathways associated with hematopoietic differentiation were concerned with NOTCH signaling and Wnt signaling ( Figure 6A ), pathways associated with DDR were predominantly related to p53 expression and apoptosis ( Figure 6B ), and pathways involved in immune function were associated with TCR, BCR, and cytokine signaling ( Figure 6C ). Pathways associated with hematopoietic differentiation and immune function in NK progenitors were increasingly upregulated in the process of differentiation from week 1 to 3 ( Figures 6A, C , Supplementary Table S2 ).

Six NK cell populations characterized by differential maturity and functional properties were recently classified using high-dimensional single-cell RNA sequencing (33). Gene clusters assigned to specific NK cell subgroups were not differentially regulated in RAG-fate-mapped (GFP⁺) and non-mapped (GFP⁻) NK cells. However, some genes related to the NK1B population were upregulated in GFP⁻ NK cells, whereas GFP⁺ NK cells expressed genes associated with the more mature NK1C, NK intermediates (NKint), and NK2 and NK3 populations ( Figures 6D, E ).

Expression of genes associated with NK cell differentiation and function were selectively analyzed in the dataset. Means of normalized gene counts obtained from GFP⁻ and GFP⁺ NK cells at weeks 1–3 are visualized by heatmap, and grouped into categories of genes encoding activating and inhibiting receptors, proteins related to activation, maturity, cytotoxicity, cytokines, chemokines, as well as cytokine and chemokine receptors ( Figure 7A ). Expression of most genes increased during the time of differentiation from week 1–3, which was particularly observed in genes encoding for NK lymphocyte and cytokine receptors.

Differential expression is shown for selected genes in Figures 7B ; Supplementary Figure S12A . In contrast to GZMA (granzyme A), GZMB (granzyme B), and PRF1 (perforin) that mediate cellular cytotoxicity by forming pores in the target cell membrane, FASLG and TNFSF10 are involved in receptor-mediated apoptosis induced by NK cells. However, only GZMA, GZMB, and TNFSF10 were found to be significantly upregulated in GFP⁺ NK cells ( Figures 7A, B , Supplementary Figure S13A ). In contrast to the expression of genes coding for several chemokines and chemokine receptors that were downregulated in RAG-fate-mapped NK cells, significantly upregulated gene expression could be observed for cytokines such as CCL4 and CCL5.

Although the differentiation potential of the NK cell subsets investigated in this study is heterogeneous, RAG-fate-mapped NK cells show a tendency toward increased maturation and cytotoxicity.

Genes associated with B (GO_0030183) and NK lymphocyte (GO_0001779) development were identified using Ensembl Biomart, and gene expression was analyzed in GFP⁻ and GFP⁺ NK progenitors form week 1–3 ( Supplementary Figure S13B ). In total, 134 genes could be identified in this dataset, of which 13 were associated with NK-cell and 121 with B-cell differentiation. Differential expression is shown for the top 20 highly expressed genes ( Figure 7C ), including 7 DE genes (PTPN6, DOCK11, PRKDC, ID2, DOCK10, LAMTOR2, IL2RG). Highly significant were the differential expressions of PRKDC and DOCK11. PRKDC encodes for DNA-dependent protein kinase, catalytic subunit (DNA-PKcs), a key regulator of NHEJ and V(D)J recombination, and DDR processes (34). DOCK11 is associated with early B-cell development in the bone marrow and germinal center formation with some redundancy with DOCK10 (35, 36). None of the genes have been associated with NK-cell differentiation before.

In summary, expression of genes associated with B-cell development can be detected in iPSC-derived NK cells, of which PRKDC is differentially expressed in RAG-fate-mapped NK cells.