Production of unconventional t cells

Abstract

The current invention relates to the field of in vitro culturing of T cells, more in particular of in vitro production of unconventional T cells expressing specific markers. Indeed, the present invention discloses a method to produce said unconventional T cells comprising the generation of hybrid spheroids out of combining stem cells and adherent cells and further culturing said spheroids at the air liquid interface. The latter T cells can be used to produce CAR-T cells.

Description

[0001] Production of unconventional T cells

[0002] Field of the invention

[0003] The current invention relates to the field of in vitro culturing of T cells, more in particular of in vitro production of unconventional T cells expressing specific markers. Indeed, the present invention discloses a method to produce said unconventional T cells comprising the generation of hybrid spheroids out of coculturing stem cells and adherent cells and further culturing said spheroids at the air liquid interface. The latter T cells can be used to produce CAR-T cells.

[0004] Background art of the invention

[0005] Production of T cells in vitro is notoriously cumbersome. Two well-known methods are the OP9-DL1 system and the artificial thymic organoid (ATO) cultures. OP9-DL1 cultures are flat cultures in which OP9-DL1 murine cells are cocultured with stem cells and wherein the generation of T cells is of low efficiency. ATO cultures are very cumbersome to set up and cannot be scaled up easily. In ATO, MS5-DL4 cells are cocultured as a pellet at the air liquid interphase.

[0006] There is thus still a need to design efficient and easy to manipulate culture systems which are -in addition and surprisingly- capable to produce unconventional T cells.

[0007] Description of the invention

[0008] The present invention introduces a novel in vitro cultivation system based on ThymoSpheres (TS) capable of generating large numbers of mature T cells despite the absence of conventional positive and negative selection. TS are easy to manipulate, are scalable, and cultures can be sampled several times during T cell differentiation. The mature T cells generated in TS from -for example- primary human hematopoietic progenitor cells were extensively characterized using single-cell RNA sequencing and combined TCR sequencing. The predominantly CD8 single-positive cells exhibit characteristics of an innate-like population recently described in humans. This includes the expression of hallmark genes ZNF683 and IKZF2 and expression of various NK receptors including NCR3 and KIR receptors. The TCR repertoire is polyclonal and displays autoreactive features and early rearrangements of the TCRa chain.

[0009] The cultivation method of the present invention is similar but sufficiently different from the ATO culture system. Indeed, a crucial difference is the formation of spheroids instead of organoids. However minimal these alterations are, these are sufficient to generate completely different T cells. Whereas the prior art show strong evidence that conventional T cells are generated, our method generates exclusively ‘unconventional’ T cells.

[0010] In the method of the present invention, a cell suspension of MS5 and T cell precursors is cultured. During this culture the cells spontaneously generate spheroids. These spheroids can be manipulated and 16 hrs after initiation of the culture, these spheroids can be layered on top a membrane floating on culture medium. In contrast, the method described as ‘ATO’, cultures cell pellets. A mixture of MS5 and T cell precursors is centrifuged and the resulting pellet is put as a whole on a floating membrane. As we obtain different cells, this means that the different method steps are important. In addition, our method is easily scalable: it takes just about the same time to manage cultures of thousands of precursors as it takes to manage millions of cells. This is not the case for the ATO cultures, as each individual pellet has to be manually manipulated and carefully deposited on the membrane!

[0011] The present invention relates thus to a population of T cellsof the type CD8+TCRab+ which expresses CD8alphaalpha homodimers and helios, and, which is negative for CCR7 and CD26, and, which contains variable percentages of cells expressing NK receptors of the type KIR, NKG2C and NCR3. The latter T cells are also denominated as ‘unconventional T cells’.

[0012] The term ‘unconventional’ relates to the fact that the cells are generated after agonist selection instead of after conventional selection mechanisms in the thymus. In addition, the latter unconventional T cells demonstrate an innate-like, effector phenotype which is different from the conventional, naive T cells.

[0013] The markers ‘helios’, ‘CCR7’, ‘CD26’,’KIR’,’NKG2C’ and ‘NCR3’ are well-known in the art.

[0014] The present invention thus relates to a method to produce a population of T cells as described above comprising:

[0015] -combining a source of stem cells and an adherent cell line expressing DLL4 or DLL1 in order to generate hybrid spheroids which are denominated as thymospheres

[0016] - culturing said thymospheres at the air liquid interphase until a population of T cells as stated above is generated.

[0017] In other words, the present invention relates in first instance to a method to produce a population of T cells of the type CD8+TCRab+ which expresses CD8alphaalpha homodimers and helios, and, which is negative for CCR7 and CD26, and, which contains variable percentages of cells expressing NK receptors of the type KIR, NKG2C and NCR3, comprising:

[0018] -coculturing hematopoietic stem cells isolated from any source tissue and an adherent cell line expressing DLL4 in order to generate hybrid spheroids which are denominated as thymospheres,

[0019] -layering said thymospheres on top of a membrane floating on culture medium,

[0020] - culturing said thymospheres at the air liquid interphase until said population of T cells is generated. ‘Thymospheres’ have the same composition as ATO cultures, however, cells are not pelleted but cultured overnight in microwells. The cells in each microwell spontaneously form spheroids. After 16 hrs, spheroids are harvested and cultured at the air liquid interphase. These cultures are easily manipulated.

[0021] More specifically, the present invention relates to a method as indicated above wherein said stem cells are CB CD34+ Hematopoietic precursor cells (HPCs) or pluripotent stem cells.

[0022] The present invention further relates to the usage of a population of T cells obtainable by a method as indicated above to generate CAR T cells.

[0023] Examples

[0024] Materials & Methods

[0025] Sample processing

[0026] Human CB was obtained from the Cord Blood Bank UZ Gent. Mononuclear cells were isolated using density gradient centrifugation (LymphoPrep; Axis-Shield, 1114547), after which the CD34+HPCs were selected using the human CD34 MicroBead kit (Miltenyi, 130-046-703) according to the manufacturer’s instructions. The labeled cell suspension was passed three times through the magnet to obtain a purity of >95%, which was flow cytometrically assessed via presence of CD34 and absence of CD3 expression. The CD34+HPC fractions of four different donors were combined to generate the TS cultures.

[0027] Generation of TS cultures

[0028] The non-adherent microwell-patterned agarose tissue culture plates were generated as previously described (11). In short, sterilized UltraPure Agarose (Invitrogen, 16500-500) was melted in phosphate buffered saline (PBS; Thermo Fisher Scientific, 10010015) (3% w / v) and added to a tailor-made, negative polydimethylsiloxane (PDMS) stamp in a 6-well tissue culture plate. Two types of stamps were used: a stamp creating 2865 microwells with each a diameter of 200 pm (TS-200) and a stamp creating 1585 microwells with each a diameter of 400 pm (TS-400). The plates were centrifuged (60 s, 3000 rpm, 30°C) and then allowed to solidify at room temperature for 20 minutes. After cooling, the microwell- patterned agarose was cautiously detached from the embedded PDMS stamp, an appropriate size was punched out and carefully transferred to the bottom of 12-well tissue culture plates. Following, PBS was added to every well, after which the culture plates were centrifuged (2 min, 1500 rpm, RT) to remove air bubbles. Finally, the culture plates were sealed with Parafilm and preserved at 4°C for up to 2 days until further use.

[0029] To form the spheroids in the microwells, PBS was first removed from the microwellcontaining wells. Murine-stromal cell line 5 - Delta Like Canonical Notch Ligand 4 (MS5- DLL4) cells were cultured in Dulbecco's Modified Eagle Medium (DMEM; Thermo Fisher Scientific, 41965039) supplemented with 10% fetal calf serum (Biowest, S1810), before being harvested using trypsin. 7 500 CD34+HPCs isolated from human CB and 1 920 000 MS5-DLL4 cells were resuspended in 500 pL Roswell Park Memorial Institute medium (RPMI, Thermo Fisher Scientific, 52400025) complemented with 4% B-27™ Supplement (Thermo Fisher Scientific, 17504044), 30 pM L-ascorbic acid (Sigma-Aldrich, 1713265258), 1 % penicillin / streptomycin (Thermo Fisher Scientific, 15140122), 1 % GlutaMAX™ Supplement (Thermo Fisher Scientific, 35050061 ), FLT3-L (5 ng / mL, Miltenyi, 130096480), and interleukin-7 (5 ng / mL, Miltenyi, 130095362) (ATO medium) and pipetted into the microwell-containing wells. The culture plates were then incubated for 2 hours at 37°C and 5% CO2, before adding another 500 pL of ATO medium to each well. Next, the culture plates were incubated overnight at 37°C and 5% CO2. After microscopy inspection, the spheroids were carefully transferred with a 1 -mL micropipette tip from the microwells into a 15 mL tube. After allowing to settle for 5-10 minutes, the supernatant was removed and 200 pL of fresh ATO medium was added. Finally, the spheroids were transferred onto a hydrophilic 0.4 pm Millicell insert (Millipore, PICM0RG50) in a 6-well tissue culture plate containing 1 mL of ATO medium. The plates were cultured at 37°C and 5% CO2 for several weeks, while the ATO medium was replaced every three to four days. Harvest of the TS cultures

[0030] The TS were harvested at weeks 3, 6, and 9 by pipetting MACS buffer onto the insert. The TS were disaggregated by pipetting 1 ml_ of MACS buffer vigorously on the insert. The cell suspension was subsequently passed through a 50-pm nylon cell strainer. This step was repeated 2-3 times until the insert was clear of cells.

[0031] Generation and harvest ofATO cultures

[0032] Artificial thymic organoids (ATOs) were generated and harvested by our group as previously described (5).

[0033] Flow Cytometry and Antibodies

[0034] Staining of surface markers was performed in DPBS (Lonza, 17-512F) with 1 % fetal calf serum using the antibody to cell ratio recommended by the supplier. Intranuclear stainings were performed following the supplier’s protocol using the eBioscience™ Foxp3 I Transcription Factor Staining Buffer Set (eBioscience, 00-5523-00). Flow cytometric analysis was performed on the LSR II and cell sorting on the FacsARIA Fusion (both BD Biosciences). Flow cytometry data were analyzed using FACS DIVA software (BD Biosciences) and FlowJo software (TreeStar Inc). Viable cells were gated based on propidium iodide (PI) negativity or Fixable Viability Dye (eFluor 506; Thermo Fisher Scientific, 65-0866-18) negativity, for surface and intracellular stainings respectively. The following list of anti-human monoclonal antibodies was used. Allophycocyanin (APC) / AF647-conjugated: CD27 (BioLegend, 302812), CD34 (BioLegend, 343510), Helios (Biolegend, 137221 ), TCRyS (Miltenyi, 130-113-500); APC Cy7 / APC Fire750- conjugated: CD3 (BioLegend, 300470), CD8a (BioLegend, 344746); Brilliant Violet 421- conjugated: CD1a (Biolegend, 300128), CD3 (BioLegend, 317344); Brilliant Violet 510- conjugated: CD8a (BioLegend, 301048), CD45 (BioLegend, 304036); fluorescein isothiocyanate-conjugated (FITC): CD7 (homemade), CD69 (BioLegend, 310904), TCRyS (BD Biosciences, 347903); phycoerythrin (PE)-conjugated: CD45 (Miltenyi, 130- 080-201), PD-1 (CD279, Biolegend, 367404), TCRap (BioLegend, 306708); PE Cy7- conjugated: CD3 (BioLegend, 300420), CD5 (Thermo Fisher Scientific, 25-0059-42); peridinin chlorophyll protein complex Cy5.5-conjugated: CD4 (BioLegend, 344608). Single-cell RNA and TCR sequencing analysis

[0035] Single-cell library preparation and sequencing: At week 9, multiple TS-containing inserts were harvested as described above and pooled. Subsequently, the CD3+TCRyb’ fraction was sorted. Following FACS, the cells were resuspended at an estimated final concentration of 3000 cells / pl and loaded onto a Chromium GemCode Single Cell Instrument (10x Genomics) to generate single-cell gel beads-in-emulsion (GEM) at the VIB Single Cell Core. The scRNA / TCR libraries were prepared using the GemCode Single Cell 5’ Gel Bead and Library kit, NextGEM v2 version (10x Genomics) according to the manufacturer’s instructions. Sequencing libraries were loaded on an Illumina NovaSeq flow cell at the VIB Nucleomics core with sequencing settings according to the recommendations of 10x Genomics.

[0036] Preprocessing of the scRNA-seq and scTCR data: The Cell Ranger pipeline (10x Genomics, version 3.1.0) was used to perform sample demultiplexing and to generate FASTQ files for read 1 , read 2 and the i7 sample index for the gene expression libraries. Read 2 of the gene expression libraries was mapped to the reference genome (GRCh38.99) using STAR. Quality control of the resulting count matrices was conducted using Scater (version 1.30.1) . The cells were selected based on the following parameters: number of genes per cell (701 < nFeature < 7 707), number of UMI counts per cell (1 127 < nCount < 25 325) and percentage mitochondrial genes per cell (percent. mt < 11.55). Doublets were removed using scDblFinder (version 1.14.0, default settings) (37). An additional group of 362 cells was removed as they predominantly expressed histone genes. The following steps were subsequently performed using Seurat version 5.0.1 using default settings: NormalizeData(), FindVariableFeatures(), ScaleData() while simultaneously regressing out the difference between G2M and S cell cycle phase scores, RunPCA() (12). Finally, clustering was performed (FindNeighbors(), dims = 1 :35); FindClusters(), resolution = 0.8) and a Uniform Manifold Approximation and Projection (UMAP) was generated (RunUMAP(), dims = 1 :35). Clusters DP(P)1 , DP(P)2, DP(P)3, DP(P)4, DP(P)5, DP(Q), CD8SP_TOX2, CD8SP_TNFRSF9, CD8SP_LTB,

[0037] CD8SP_ZNF683 and CD8SP_GZMM were subsetted and a new UMAP projection was generated using the parameters mentioned above. Clusters DP(Q), CD8SP_TOX2, CD8SP_TNFRSF9, CD8SP_LTB and CD8SP_ZNF683 were reclustered at a resolution = 1.1 and the new annotations were projected on the subset UMAP.

[0038] Processing the scRNA-seq data:

[0039] After the initial preprocessing steps, DEGs were determined using the Seurat function FindAIIMarkers(). A diffusion map dimensionality reduction was performed using the Destiny package (version 3.16.0). First, diffusion map embeddings were generated using the DiffusionMap() function, utilizing the first 35 principal components. Subsequently, cells were ranked based on their diffusion component 1 (DC1) values, identifying the tip cell as the cell with the earliest pseudotime ranking. Diffusion pseudotime was then computed using the DPT function, starting from the specified tip cell (39). A pseudotime analysis was performed with Monocle3 version 1.3.4 (40).

[0040] Finally, a single-cell data integration was performed of the TS data with CB and PNT samples. Four PNT samples (TTA9, TTA10, TTA12 and TTA14) and three CB samples (BVK01 , BVK02 and BVK03) were selected (3,4). The PNT and CB samples were preprocessed using the same parameters as described above. Before data integration, all samples were merged into a combined Seurat object, followed by normalization, variable features identification, scaling with regression of cell cycle genes and PCA analysis, using the functions described above. The integration was executed using the Harmony package version 1.2.0 (RunHarmonyO). Three sources of potential batch effects were taken into account: donor, sorting strategy and method (3’ or 5’).

[0041] Processing the scTCR-seq data: scTCR-seq analysis was performed using the scRepertoire package version 2.0.0 using the filtered_contig_annotations.csv output from the 10x Genomics Cell Ranger pipeline (15). Clonotypes were determined using the clonalQuant() functions. Clonal overlap was also visualized as a chord diagram using the Circlize package. Determination of the CDR3a and CDR3p apex region and cysteine usage was performed following previously described indications (30). Hydrophobic CDR3a and CDR3p doublet containing sequences were determined by calculating the percentage of sequences using any of the 175 AA doublets previously identified as promoting self-reactivity (2). The presence of cysteine or hydrophobic AA was calculated for each single-cell and the percentages were determined based on all cells containing a CDR3 sequence per cluster.

[0042] Statistical Analysis

[0043] Statistical analyses were performed in Prism version 10.1.1. (GraphPad Software, San Diego, CA, USA), using statistical tests as indicated in figure legends. Results were considered statistically significant when the p-value was less than 0.05.

[0044] Results

[0045] Efficient generation of mature CD8a+T cells in TS

[0046] Our objective was to develop an easily scalable system that facilitates the efficient in vitro differentiation and production of large quantities of mature T cells. To achieve this, we initially focused on generating a substantial number of 3D spheroids or TS. The TS were formed by allowing cells to aggregate in small microwells. Then, these TS were conveniently cultured together at the air-liquid interface using cell culture inserts. MS5- DLL1 / 4 stromal cells were used to avoid the use of thymic epithelial cells. Specifically, we generated TS by combining 7.5x103CB CD34+HPCs and 1.9x106MS5-DLL4 cells in a well containing microwell-patterned agarose forming either 2865 microwells with each a diameter of 200 pm (TS-200) or forming 1585 microwells with each a diameter of 400 pm (TS-400). Within 24 hours, the HPCs and MS5-DLL4 cells formed a dense spheroid TS, which could be visually observed in each microwell. The TS were subsequently carefully transferred onto a cell culture insert and cultured at the air-liquid interphase for up to 9 weeks. Harvesting all or part of the TS from the insert could be done efficiently by repeatedly pipetting the TS thereby mechanically dissociating the TS into a cell suspension. Starting from 7.5x103HPCs, the mean cell yield per insert of TS was significantly higher compared to an ATO insert. One insert of TS generated approximately 1x106cells from 7.5x103HPCs by week 3, reaching a steady state at more than 2x106 cells at week 6 and 9. No significant differences between the TS-200 and TS-400 cultures were observed. The CD34+population had expanded at week 3 in all cultures. Additionally, rapid commitment to the T lymphoid lineage was observed by the predominance of CD5+CD7+cells in the TS cultures, whereas CD5’CD7’ myeloid cells were virtually absent. Furthermore, a substantial number of CD3_CD4+immature single positive (ISP) and DP cells were already present at week 3, with most of the developing cells remaining CD4_CD8a_. By week 6, DP cells reached their peak, and a significantly higher percentage of CD8a+cells was detected, while CD4+cells diminished. Simultaneously, a substantial number of CD3+cells was generated. CD69, an early marker of positive-selected cells, was only expressed by a limited percentage of DPs at week 6 but increased by week 9 . The percentage of unconventional TCRyb cells generated in TS was consistently low. When focusing on the TCRap cells, phenotypically mature CD1a_CD27+T cells could be discerned at week 6, with TS-200 and TS-400 showing equal kinetics in the generation of mature cells. Mature CD1a_cells were predominantly CD8a+by week 6, generating up to 6x104CD8a+cells from one cell culture insert at week 9. In contrast, the percentage of mature CD4+cells generated in TS was negligible. Notably, the transition from immature CD1 a+DP to mature CD1 a_CD8a+cells was defined by a high expression of PD-1 and Helios, markers used to define agonist- selected cells in the human thymus and their progeny in CB (1 ,3). scRNA-seq profiling identifies mature TS T cells as unconventional T cells

[0047] To further unravel the nature of the mature ap T cells generated in the TS culture, CD3+TCRyb- cells were sorted from multiple pooled TS cultures at week 9 and processed for single-cell RNA and TCR sequencing. Using the droplet-based 10x Genomics single-cell platform, 5’ gene expression libraries were constructed. Subsequent quality control filtering steps retained 17 611 cells, which were divided into 15 clusters. Combined expression of CD4, CD8A and CD8B led to the annotation of seven CD3+DP clusters. Five proliferating DP clusters (DP(P)1 -5) were annotated based on the expression of cell cycle genes, such as CDK1, PCNA, and MKI67 . One quiescent DP cluster (DP(Q)) was observed with the pronounced peak expression of CD4 and CD8B, VDJ recombination genes RAG 1 and RAG2), immaturity markers (CD1A and CD1C). One minor DP(P / Q) cluster showed characteristics of both the proliferating and quiescent DPs, while grouping with the mature CD8SP cells.

[0048] Downregulation of CD1A and CD1C and the loss of CD4 expression defined eight mature CD8 T cell clusters . CD8B expression was generally lower compared to CD8A expression in the CD8SP clusters . In accordance with the flow cytometric analysis, no mature CD4SP cells were observed in TS at week 9 . This was also evidenced by the absence of ZBTB7B expression, while RUNX3 was expressed by the mature CD8SP cells. Besides a CD8SP_cycling cluster, the majority of the mature cells consisted of five quiescent CD8SP clusters, which were annotated based on their most prominent differentially expressed genes (DEG). The CD8SP_TOX2 cluster was characterized by a high expression of T0X2 and bridged the DP clusters and CD8SP clusters . CD27 was strongly upregulated at the CD8SP_TOX2 cluster, whereas CD69 was expressed in the subsequent CD8SP clusters. Therefore, the selection process initiating the differentiation from DP to CD8SP cells appears to be induced at the transition from the DP(Q) to the CD8SP_TOX2 cluster. Notably, hallmark genes associated with the agonist-selected UTC population in the human thymus were expressed by the CD8SP_TOX2 cluster (1). Indeed, IKZF2 (Helios) was expressed by the majority of the TS cells, with the highest expression in the CD8SP_TOX2 cluster. Furthermore, MME (CD10) and PDCD1 (PD-1) were almost exclusively expressed by the CD8SP_TOX2 and CD8SP_LTB clusters. In contrast to CTCs, CCR7 was only weakly expressed by the post-selection CD8SP clusters. Of note, marker genes for (differentiating) regulatory T cells, such as F0XP3, CTLA4 or IKZF4, were not expressed in these cultures.

[0049] The CD8SP_TNFRSF9 cluster was characterized by a high expression of TNFRSF9 (CD137), resembling the agonist-selected GNG4+CD8aa+T(l) population identified in human thymus but not found in CB (5,7). Multiple CD8SP clusters expressed ZNF683 (Hobit), with the highest expression in the CD8SP_ZNF683 cluster, which resembled the thymic agonist-selected ZNF683+CD8aa+T(ll) population (1 ,3). However, as mentioned above, PDCD1 and MME expression were mostly confined to the CD8SP_TOX2 and CD8SP_LTB clusters and were not strongly expressed by the CD8SP_TNFRSF9 or CD8SP_ZNF683. Interestingly, the CD8SP clusters expressed various transcription factors (TFs) associated with innateness, such as TBX21, HOPX, ID2, MYBL1, BHLHE40, FOSL2, NFIL3 and RUNX3 (26). Moreover, clusters CD8SP_ZNF683 and especially CD8SP_GZMM showed an elevated expression of AP1 TFs (FOS, JUNB), cytokine receptors (IL2RB), cytokines (CCL5, INFG), cytotoxic molecules (GZMK and GNLY) and NK markers including NCR3, KLRC2 (NKG2C), NKG7 and multiple killer-cell immunoglobulin-like receptors (KI Rs). These characteristics are reminiscent of the recently identified polyclonal UTCs in human CB (3).

[0050] Lastly, two minor clusters showed a pronounced expression of KLRB1 (CD161 ), while also co-expressing ZBTB16 (PLZF) . These clusters contained both double negative (DN) and CD8SP cells. Notably, different minor fractions of cluster CD8SP_GZMM also expressed low levels of KLRB1 or ZBTB16. Cluster KLRB1_17 was characterized by expression of typical MAIT cell markers such as CXCR6, CCR6, IL12RB1, and SLC4A10. Furthermore, the KLRB1_17 cluster was the only mature cluster to express RORC, analogous to NKT17 cells. Cluster KLRB1_2 only consisted of 41 cells and uniformly expressed GATA3 and IL13, analogous to NKT2 cells (2).

[0051] Comparative transcriptomic profiling in TS cultures

[0052] To determine the developmental ordering of the cells, the diffusion pseudotime of the major TS clusters was calculated. As expected, the differentiation trajectory originated in the DP(P) clusters, passed through the DP(Q) cluster and ended in the mature CD8SP clusters. Expression of MME, PDCD1 and IKZF2 along the diffusion pseudotime confirmed the results discussed above, with a peak expression at the CD8SP_TOX2 cluster. Furthermore, the expression of KIR2DL3 was only upregulated at the end of the pseudotime, in the CD8SP_GZMM cluster . Plotting cells according to their determined pseudotime, cells in the CD8SP_ZNF683 and CD8SP_TNFRSF9 clusters appeared to exhibit similar levels of maturity. Moreover, a pseudotime analysis with Monocle3 resulted in a branching point at the CD8SP_TOX2 cluster. To better comprehend this bifurcation point, the relevant clusters were reclustered at a higher resolution. As a result, the CD8SP_TOX2 cluster was divided into subset_6 and subset_9 . When analyzing the DEGs between subset_6 and subset_9, subset_9 expressed several TFs (EGR2, NR4A1, NFIL3) as well as marker genes (TNFRSF9, GNG4, XCL1) attributed to the CD8aa+T(l) population in the human thymus, while giving rise to the CD8SP_TNFSFR9 cluster according to the determined pseudotime. Subset_6, expressing higher levels of T0X2 and RASGRP2, gave rise to CD8SP_LTB cluster, which, suggested by the pseudotime, developed into CD8SP_ZNF683 and CD8SP_GZMM clusters. This observation suggests that differentiation between the CD8aa(l)-like CD8SP_TNFSFR9 and CD8aa(ll)-like CD8SP_ZNF683 populations occurs during the CD8SP stage in these TS cultures and may be associated with the expression of TFs such as T0X2, NFIL3, and EGR2.

[0053] To determine whether the mature T cells produced in TS resembled agonist-selected UTCs in humans, the TS scRNA-seq dataset was integrated with two previously published datasets of both CD3+cells from postnatal thymus (PNT) and CD3+ / |OWTCRyb- CD4- cells from CB (3,4). When focusing on the mature cells, the DP(Q) cells diverged into two main groups, in accordance with the previous studies: one containing the CTCs and one containing the UTCs from both PNT and CB (3,4). The CD8SP_TOX2 TS cluster colocalized with the apT(entry) and CD8aa+T(l) PNT clusters, with the apT(entry) reaching more towards the CTC branch and the CD8aa+T(l) and CD8SP_TOX2 clusters reaching more towards the UTC branch. The CD8SP_TNFRSF9 TS cluster was located adjacent to the CD8aa+T(l) PNT cluster and partly overlapped with the CD8aa+T(l I) PNT cluster. CD8SP_LTB was the TS clusterthat colocalized with the CD8aa+T(l I) PNT cluster and MME+UTC CB cluster (which were described as the recent thymic emigrants in CB), while also having scattered extensions between both the CD4 and CD8 CTCs from PNT. The CD8SP_ZNF683 and CD8SP_GZMM TS clusters entirely overlapped with the effector-like UTC clusters from CB, validating their similar transcriptom ic profile. Of note, the KLRB1_17 TS cluster colocalized with the NKT / MAIT CB cluster and a fraction of the CD8SP_GZMM cluster overlapped with the NK cells from PNT and CB, possibly due to its high expression of NK markers . In summary, single-cell transcriptomics suggest a similar differentiation of the polyclonal UTC lineage in TS compared to human thymus and blood. TS CD8SP T cells express a polyclonal TCR repertoire with autoreactive features

[0054] To elucidate the usage of V(D)J gene segments and TCR characteristics across the identified clusters, the TCR sequences generated in the TS cultivation system were analyzed. The DP clusters predominantly expressed high levels of beta chain transcripts, while mature T cell clusters exhibited expression of paired TCRap chains. Notably, only a few productive TCRap chains could be detected for the KLRB1 clusters, with the KLRB1_17 cluster also expressing some single TCRp chains. Noteworthy is the identification of 45 cells expressing a hybrid TRDV1-TRAJ-TRAC TCRa chain across several CD8SP clusters. These cells use the TCR 5 variable 1 (TRDV1) gene segment instead of a TRAV gene segment (7,28). Such hybrid cells were also enriched in the polyclonal UTC populations of both PNT and CB (5). Subsequently, the usage of V and J gene segments across the various TS clusters was examined, showing a diverse set of V and J gene segment usage reflecting a polyclonal TCR repertoire. Human TCRa rearrangements are known to occur sequentially, with proximal V and J segments recombining first, followed by more distal V and J segments (9). In TS cells, a biased usage of J-proximal V segments and V-proximal J segments was indeed observed for the DP clusters. Furthermore, the mature CD8SP clusters also showed a biased usage of more proximal V and J segments. Importantly, this bias is a main feature of the UTC populations of both PNT and CB, while CTCs preferentially use late distal V and J segments (1 ,3,4). This suggests that the selection process driving the CD8SP cells in TS occurs at an early DP stage. Of note, the minor KLRB1 clusters also expressed diverse V(D)J gene segments and were not restricted to typical gene segments associated with NKT or MAIT cells.

[0055] Two CDR3 self-reactivity indices have been reported for UTC populations: the presence of cysteines within two positions of the CDR3 apex (known as the cysteine index) and enrichment of hydrophobic amino acid (AA) doublets at positions 6 and 7 of the CDR3 (known as the hydrophobic index) (2,3,10). Elevated percentages of these indices were observed in mature T cell clusters compared to DP clusters, suggesting that these TCR characteristics are selected. Furthermore, the hydrophobic index of mature clusters was comparable to the PD-1+UTC population in CB, which was around 25%, and the cysteine index was even higher in these mature TS clusters, surpassing the 0.5% observed in the PD-1+UTCs in CB (3). Clonal relatedness was determined based on identical CDR3 nucleotide sequences. Within each cluster, most cells contained unique clones, and the level of clonal sharing between different clusters was limited. The most expanded clonotypes primarily contained a single CDR3p sequence and were predominantly found in the DP clusters, likely due to p-selection-induced proliferation. Only three clones were expanded to each four cells, primarily residing in the CD8SP_GZMM cluster. As suggested by the differentiation trajectory described above, this cluster represents the most differentiated stage and may have acquired the ability to proliferate in response to cytokines present in these cultures. Notably, clone CAAPNYGQNFVF_CSARSCRTEAFF was present in four different clusters (CD8SP_TNFRSF9, CD8SP_LTB, CD8SP_ZNF683, and CD8SP_GZMM) and contained both hydrophobic and cysteine AA, suggesting that the TCR specificity may not be crucially involved in the decision to differentiate to CD8SP_TNFRSF9 or CD8SP_GZMM (.

[0056] References

[0057] 1. Verstichel G, Vermijlen D, Martens L, Goetgeluk G, Brouwer M, Thiault N, et al. The checkpoint for agonist selection precedes conventional selection in human thymus. Sci Immunol. 2017 Feb 24;2(8).

[0058] 2. Daley SR, Koay HF, Dobbs K, Bosticardo M, Wirasinha RC, Paia F, et al. Cysteine and hydrophobic residues in CDR3 serve as distinct T-cell self-reactivity indices. Journal of Allergy and Clinical Immunology. 2019 Jul 1 ;144(1):333-6. Billiet L, de Cock L, Sanchez GS, Mayer RL, Goetgeluk G, de Munter S, et al. Single-cell profiling identifies a novel human polyclonal unconventional T cell lineage. J Exp Med. 2023 Jun 5;220(6). Park JE, Botting RA, Conde CD, Popescu DM, Lavaert M, Kunz DJ, et al. A cell atlas of human thymic development defines T cell repertoire formation. Science (1979) [Internet], 2020 Feb 21 [cited 2020 Sep 28];367:eaay3224. Available from: http: / / dx.doi. Pille M, Avila J, Sanchez GS, Goetgeluk G, De Munter S, Jansen H, et al. The Wiskott-Aldrich syndrome protein is required for positive selection during T-cell lineage differentiation. Front Immunol. 2023 Jun 7; 14:1188099. Gutierrez-Arcelus M, Teslovich N, Mola AR, Polidoro RB, Nathan A, Kim H, et al. Lymphocyte innateness defined by transcriptional states reflects a balance between proliferation and effector functions. Nat Commun [Internet], 2019 Feb 8 [cited 2022 Nov 23];10(1):687. Available from: https: / / www.nature.com / articles / s41467-019-08604-4 Pellicci DG, Uldrich AP, Le Nours J, Ross F, Chabrol E, Eckle SBG, et al. The molecular bases of 5 / a T cell-mediated antigen recognition. Journal of Experimental Medicine. 2014 Dec 15;211(13):2599-615. Volkmar M, Fakhr E, Zens S, Bury A, Offringa R, Gordon J, et al. Identification of TRDV-TRAJ V domains in human and mouse T-cell receptor repertoires. Front Immunol. 2023; 14. Krangel MS. Mechanics of T cell receptor gene rearrangement. Cum Opin Immunol. 2009 Apr 1 ;21 (2): 133— 9. Wirasinha RC, Singh M, Archer SK, Chan A, Harrison PF, Goodnow CC, et al. a T-cell receptors with a central CDR3 cysteine are enriched in CD8aa intraepithelial lymphocytes and their thymic precursors. Immunol Cell Biol. 2018 Jul 1 ;96(6):553-61. 11. Okkelman IA, Vercruysse C, Kondrashina A V., Borisov SM, Dmitriev Rl. Affordable Oxygen Microscopy-Assisted Biofabrication of Multicellular Spheroids. JoVE (Journal of Visualized Experiments). 2022 Apr 6;2022(182):e63403.

[0059] 12. Hao Y, Hao S, Andersen-Nissen E, Mauck WM, Zheng S, Butler A, et al. Integrated analysis of multimodal single-cell data. Cell. 2021 Jun;184(13):3573- 3587. e29.

[0060] 13 Angerer P, Haghverdi L, Buttner M, Theis FJ, Marr C, Buettner F. destiny: diffusion maps for large-scale single-cell data in R. Bioinformatics. 2016 Apr 15;32(8):1241-3.

[0061] 14. Qiu X, Mao Q, Tang Y, Wang L, Chawla R, Pliner HA, et al. Reversed graph embedding resolves complex single-cell trajectories. Nature Methods 2017 14:10. 2017 Aug 21;14(10):979-82

[0062] 15. Borcherding N, Bormann NL, Kraus G. scRepertoire: An R-based toolkit for singlecell immune receptor analysis. F1000Res. 2020;9.

Claims

Claims1 . A method to produce a population of T cells of the type CD8+TCRab+ which expresses CD8alphaalpha homodimers and helios, and, which is negative for CCR7 and CD26, and, which contains variable percentages of cells expressing NK receptors of the type KIR, NKG2C and NCR3, comprising:-coculturing hematopoietic stem cells isolated from any source tissue and an adherent cell line expressing DLL4 in order to generate hybrid spheroids which are denominated as thymospheres,-layering said thymospheres on top of a membrane floating on culture medium, - culturing said thymospheres at the air liquid interphase until said population of T cells is generated.

2. A method according to claim 2 wherein said stem cells are CB CD34+ HPCs or pluripotent stem cells.