Engineering T Cell Development for the Next Generation of Stem Cell-Derived Immunotherapies

Engineered T cells are at the leading edge of clinical cell therapy. T cell therapies have had a remarkable impact on patient care for a subset of hematological malignancies. This foundation has motivated the development of off-the-shelf engineered cell therapies for a broad range of devastating indications. Achieving this vision will require cost-effective manufacturing of precision cell products capable of addressing multiple process and clinical-design challenges. Pluripotent stem cell (PSC)-derived engineered T cells are emerging as a solution of choice. To unleash the full potential of PSC-derived T cell therapies, the field will require technologies capable of robustly orchestrating the complex series of time- and dose-dependent signaling events needed to recreate functional T cell development in the laboratory. In this article, we review the current state of allogenic T cell therapies, focusing on strategies to generate engineered lymphoid cells from PSCs. We highlight exciting recent progress in this field and outline timely opportunities for advancement with an emphasis on niche engineering and synthetic biology.


F
or the past decade, there has been tremendous progress in treating cancer with cellular immunotherapies.In particular, chimeric antigen receptors (CARs) that redirect T cells to recognize CD19 or BCMA have shown remarkable efficacy against certain types of leukemia, lymphoma, and multiple myeloma. 1 So far, six CAR-T therapies have been approved by the U.S. Food and Drug Administration (FDA) since 2017, with several hundred more products at various stages of clinical development. 2][5][6][7] Despite their impressive and expanding resume, several technical hurdles must be overcome before T cell therapies realize their full potential.Questions of safety, efficacy, antigen escape, and translation to solid tumors are important challenges that have been extensively reviewed elsewhere. 8Another major bottleneck is the manufacturing process itself.The predominant method of manufacturing T cell therapies today is to source T cells from the patient themselves, genetically engineer them to recognize malignant cells using bespoke viral transduction pipelines and infuse the modified cells back into the patient. 1,9his personalized manufacturing process adds significant cost 10,11 and variability 12,13 to T cell therapies.Building more ro-bust, scalable, and reproducible manufacturing workflows for T cell therapies will help improve product safety and efficacy and will expand access to these lifesaving therapies. 9n this review, we examine the current state of allogenic T cell therapies, paying particular attention on strategies to generate engineered lymphoid cells from pluripotent stem cells (PSCs).We also discuss recent progress in the field and consider opportunities for advancement, especially with regard to niche engineering and synthetic biology.

Challenges of autologous CAR-T therapies
Unlike small molecules and biologics, which are produced in scalable, controllable, and fully defined environments inside chemical or biological reactors (Domokos, Nagy, 2021), T cell therapy products are living medicines.Currently, for clinical use, T cell therapies are predominantly produced in single doses, sourcing the unmodified T cells from the patient themselves. 9,14hese autologous T cell therapies face several significant manufacturing challenges.One is cost.For allogenic products, each time a patient is prescribed CAR-T, a personalized multistep workflow is initiated. 9The patient donates peripheral blood and their T cells are isolated.Next, the T cells are stimulated ex vivo and begin to undergo massive expansion.Once the T cells have been activated, they are transduced with a viral vector comprising the CAR transgene.Finally, the CAR-T product undergoes release testing for quality assurance before being infused into the patient.
This entire workflow, including production of the CAR viral vectors, is carried out in compliance with good manufacturing practices.Because of this complex personalized production process, the price of CAR-T is remarkably high compared with other types of therapies, with one dose priced on the order of $500,000. 10,11,15Whether the benefits of CAR-T therapies to patients can justify their current price tag is difficult to quantify and a matter of ongoing debate. 11,16Not only do excessive costs limit the number of patients who are able to access currently approved therapies, but they also restrict the scope of future indications for which CAR-T cells are contemplated.
A second challenge is variability in product quality.This variability exists in multiple aspects.There is donor-to-donor variability in the number and quality of T cells available for producing the CAR-T product. 12,13,17Beyond the biological diversity that exists among individuals in the healthy population, this problem is exacerbated by the impact of cancer on a patient's T cell compartment.9][20] Consequently, a meaningful percentage of patients are not able to produce a graft of good enough quality to proceed to infusion. 21n addition to donor-to-donor variability, there is also heterogeneity within an individual graft.Depending on the specific manufacturing process, the isolated and expanded T cells comprise a mixture of T cell subtypes at various states of polarization and differentiation along the effector-memory axis. 22Furthermore, not all T cells in the graft will successfully be transduced with the CAR, leading to a mixture of modified and unmodified cells.Even among modified cells, the semirandom nature of retroviral vector integration leads to heterogeneity in expression level that impacts functionality and carries the risk that important loci are disrupted. 23As grafts comprise billions of cells, even rare oncogenic events must be considered.

Alternatives to autologous T cell therapies
Cost, variability, and product quality control are widely recognized barriers to expanding on the success of CAR-T therapy.Hereunder, we discuss how gene-edited CARs and allogeneic T cell products can partially address these challenges.We also explore the advantages and limitations of in vivo CAR transduction as a low-cost alternative to ex vivo manufacturing.We then make the case for PSC-derived T cell products as a compelling longterm solution for accessible high-performance immunotherapies.
One way to avoid the gene expression heterogeneity caused by semirandom viral integration is to target the CAR to a defined locus using precision gene editing such as CRISPR-Cas9.In a preclinical model of B cell leukemia, introducing a CD19 CAR into the T cell receptor alpha chain constant (TRAC) locus with CRISPR-Cas9 reduced both median CAR expression and expression variability compared with a virally engineered control. 23RAC knock-in also significantly improved tumor clearance compared with virally modified controls.Virally introduced CARs are typically expressed at much higher levels than endogenous T cell receptors.High constitutive CAR expression can drive T cells into an exhausted state and impair their functionality. 23n contrast, knocking a CAR into the TRAC locus leads to a more T cell receptor (TCR)-like expression level, mitigating exhaustion and improving efficacy. 23ene-edited CARs may provide a more consistent and effective product.However, further improvements are required.Although recent designer fusion proteins can improve CRISPR-Cas9 knock-in efficiencies, integration rates for large multikilobase payloads remain low in primary T cells. 24CRISPR is also prone to off-targets mutagenesis, meaning edited products will still comprise a mix of correctly modified, incorrectly modified, and unmodified cells. 25Precise editing workflows, especially ones that require both nucleofection and AAV homology donors will add cost to the manufacturing process, although this can be mitigated using nonviral editing. 26,27Gene-edited CAR-T cells will not solve the problem of intradonor heterogeneity.
Primary donor-derived allogeneic T cell therapies can partially address the challenge of donor-donor variability.By sourcing T cells from healthy donors, it is possible to circumvent the negative impact that cancer and cancer therapies have on a patient's T cell compartment.Although autologous CAR-T products often fail release testing, 28 in recent trials of primary allogeneic CAR-T therapy, all enrolled patients were able to receive treatmentnone were excluded because of poor-quality T cell products. 29,30llogeneic therapies may also simplify the manufacturing process as cell therapy products can be cryopreserved and utilized on demand, reducing the time delay required for autologous production. 30espite their advantages, primary allogeneic therapies must overcome immunological barriers between the donor and recipient.Graft versus host reactivity is an important safety concern and avoidance of graft rejection to achieve a durable therapeutic benefit is an added challenge. 29Although healthy donors can provide a larger number of functional T cells compared with cancer patients, there is still sizable heterogeneity in T cell abundance and functionality across healthy individuals. 31Beyond the brief overview provided earlier, the advantages and limitations of primary allogeneic therapy are more thoroughly reviewed elsewhere. 32ene-edited T cells and allogeneic donors can theoretically reduce variability in CAR-T outcomes, but neither approach will eliminate the costly ex vivo T cell manufacturing process.A potential cost-saving solution is to perform a rapid ex vivo transduction step and allow the lengthy expansion process to occur within the patient.In one ongoing clinical trial of this approach, an extremely rapid ex vivo manufacturing step (*2 days) has shown promising early results. 33eyond shortening and automating the CAR-T manufacturing processes, an elegant alternative is to eliminate the ex vivo cell therapy manufacturing step altogether by delivering the CAR to T cells within a patient's body.This has previously been achieved by infusing mice with CAR expression cassettes using delivery systems such as nanoparticles loaded with DNA or mRNA, 34,35 or viral vectors. 36,37A recently proposed hybrid approach involves taking patient T cells, seeding them in an ENGINEERING ALLOGENEIC T CELL THERAPIES alginate scaffold along with viral particles and T cell activating reagents, and implanting the biomaterial back into the patient where the cells become transduced and expand in vivo. 38hese in vivo and hybrid manufacturing processes should reduce costs, time, and labor associated with CAR-T manufacturing if they prove to be effective in humans.Nevertheless, these approaches carry additional risks, including the possibility of transducing off-target cell types.
Primary allogeneic CAR-T cells can help address product variability, and in vivo CAR delivery may reduce manufacturing costs.PSCs are an extremely attractive alternative approach because they can simultaneously address both challenges.Rather than obtaining T cells from individual donors, they can be differentiated from PSCs (PSC allogeneic CAR-T).PSCs are amenable to clonal expansion and unlimited self-renewal. 39This means CARs, or other therapeutic augmentations, can be easily be introduced at targeted loci through gene editing. 40fter rigorous quality control, clones that contain the desired modification and are free from off-target mutations and chromosomal abnormalities can be expanded at scale in bioreactors for subsequent in vitro differentiation into T cells.The resulting cell therapy product can be cryopreserved for on-demand offthe-shelf therapy. 41PSCs have the potential to serve as a lowcost source of high-quality clonally engineered T cell therapies.To realize this vision and enable large-scale manufacturing of T cells for therapy, researchers have worked to develop and improve PSC-to-T cell differentiation processes.

Efforts to Make PSC-Derived T Cells
PSC differentiation protocols are often designed using insights from embryogenesis; therefore, a thorough understanding of T cell development as it occurs in vivo is critical to the success of efforts to generate PSC-derived T cells.In this section, we briefly outline the process of T cell development, then highlight how our understanding of this process has shaped the development of in vitro T cell differentiation protocols.[44][45][46][47]

T cell development in vivo
All cells of the hematopoietic system are derived from a subset of mesoderm.During early embryogenesis, pluripotent cells ingress through a transient structure called the primitive streak, where they receive signals that prime them toward the hematopoietic lineage, forming the so-called hematopoietic mesoderm compartment. 48The hematopoietic mesoderm gives rise to at least two distinct waves of hematopoietic progenitors, which are distinguished in vivo based on where and when they emerge, as well as the differentiated cell types they produce.The ''primitive'' wave is the first to emerge, and is first detected at E7 in mouse 49 and day 18.5 of gestation in human. 50Progenitors of the primitive wave are derived from ''blood islands,'' or clusters of hematopoietic and endothelial cells in the yolk sac.This early wave of hematopoiesis is transient, and gives rise to a small subset of hematopoietic cell types, including erythrocytes, megakaryocytes, and macrophages. 49portantly, the primitive waves do not produce any cells of the lymphoid lineage, including T cells.In contrast, the ''definitive'' wave of hematopoiesis emerges later in developmental time, at E11.5 in mouse 51 and CS14 in human. 52Definitive hematopoietic progenitors are derived from endothelial cells with hematopoietic potential, or hemogenic endothelial (HE) cells, within the embryo proper, predominantly in the aorta-gonadmesonephros region, which undergo endothelial to hematopoietic transition (EHT) to generate hematopoietic cells.In contrast to the apparently restricted potential of primitive progenitors, the definitive wave gives rise to hematopoietic stem cells (HSCs), or single cells that can reconstitute the entire blood system upon transplantation into a new host. 51n recent years, the definitions of primitive versus definitive hematopoiesis have become blurred.For instance, studies in mouse have identified yolk sac-derived hematopoietic progenitors with lymphoid potential, [53][54][55][56][57] and corresponding populations have been described in studies using human PSCs. 580][61][62] Although researchers have noted that certain T cell subsets are preferentially derived from progenitors that emerge before HSC development, 63 functional differences between the T cells derived from yolk sac progenitors, intraembryonic progenitors, and HSCs have not yet been characterized.
Regardless of developmental origin, T-competent progenitors initially seed the thymus, the primary organ responsible for T cell development 64 (Fig. 1A).The thymus is organized into various anatomical zones with distinct signaling environments; these different signals promote transition to successive developmental stages as T cell progenitors, or thymocytes, migrate through the thymus. 65The earliest thymus-seeding progenitors are termed double negative (DN), due to lack of expression of the co-receptors CD4 and CD8.When T-competent progenitors first enter the thymus, they retain the ability to produce myeloid and natural killer (NK) cells. 66][70] After T-lineage commitment, DN cells proceed to the immature single positive (iSP) stage, characterized by upregulation of CD4. 71During this stage, thymocytes become biased toward one of two possible lineages: TCRab+ ''conventional'' T cells, or TCRcd ''innate-like'' T cells. 72The majority of thymocytes will begin rearranging the TCRb locus.Thymocytes that produce an in-frame TCRb rearrangement will express a functional TCRb protein, which complexes with a surrogate pre-TCRa protein and CD3 to form the pre-TCR complex. 73Thymocytes that fail to produce an in-frame TCRb rearrangement are eliminated through a process termed beta selection.Although pre-TCR expression biases iSPs toward the ab-T cell lineage, a small subset of iSPs productively rearrange the TCRc and TCRd loci, leading to expression of a functional cd-TCR that initiates commitment to the cd-T cell lineage. 72After pre-TCR expression, ab-T cell precursors upregulate CD8 to become CD4+CD8+ ''double positive'' (DP) cells.Before exiting the thymus cortex, DP thymocytes rearrange their TCRa locus to produce a mature TCRa gene, resulting in expression of the TCRab complex on their surface. 74Positive selection of thymocytes expressing TCRs that can bind with moderate affinity to self-peptide-human leukocyte antigen (HLA) complexes is critical for further differentiation, as nonreactive TCR clones are eliminated in the cortex through nonselection.During this selection process, DP thymocytes also select a CD8+ cytotoxic phenotype or a CD4+ helper phenotype based on whether the TCR interacts with HLA-I molecules through the CD8 co-receptor, or HLA-II molecules through the CD4 co-receptor. 44Next, thymocytes travel to the medulla, where clones expressing TCRs with highly reactivity to selfantigens are eliminated through negative selection. 45ature T cells exit the thymus and enter the circulation as naive T cells, where they scan throughout the body for their cognate antigen.When a naive CD4 or CD8 T cell encounters its target antigen, it undergoes rapid clonal expansion and differentiates into an effector phenotype.CD8 T cells differentiate into cytotoxic T lymphocytes (CTLs), which can directly kill cells presenting the target antigen.CD4 T cells differentiate into a range of ''helper'' phenotypes, which secrete subsets of cytokines to support the function of CTLs, as well as other innate immune cells.Once the initial infection is cleared, the majority of differentiated effector cells die; however, a subset of effector cells are maintained as memory T cells.Memory T cells are of particular clinical interest, as they are relatively long-lived, but can mount an immune response to their target antigen much more rapidly than a naive T cell. 75arning from human development to make T cells in vitro with stromal feeder cells Motivated by therapeutic and research applications, scientists have long sought to recreate T cell development in vitro (Fig. 1B).Early attempts to differentiate T cells have focused on recapitulating key features of the thymus. 76Among the multitude of molecules produced in the thymus, Notch signals are crucial for T cell development and, in combination with the cytokines FLT3L and IL-7, are sufficient to support T cell development from primary hematopoietic stem and progenitor cells (HSPCs) in vitro. 77Immortalized stromal cell lines engineered to express Notch ligands, such as OP9-DL1 and OP9-DL4 systems, have been used successfully for decades. 77Additional feeder-based systems have emerged in recent years, including the artificial thymic organoid (ATO) platform, which more accurately phenocopies some aspects of three-dimensional thymic structure. 78,79][81] But these human-donor-derived cell sources suffer similar limitations as primary T cells in terms of scale and intradonor variability.PSCs can theoretically provide a vastly more scalable, cheaper, and genetically consistent source of T cells.A longstanding goal has thus been to produce HSPCs with T lineage potential from PSCs as starting material.
As blood progenitors are derived from the mesoderm, the majority of established PSC-to-T cell differentiation protocols involve a mesoderm induction step.3][84] This is predominantly due to the addition of exogenous Activin A without Wnt supplementation during the mesoderm induction step, which promotes the development of primitive-biased mesoderm. 85,868][89] These PSC-derived HSPCs could develop into mature T cells in vitro on stromal feeder layers.
Note that these cells still lack some key attributes of true HSCs as they have thus far failed to engraft and reconstitute immunodeficient mice without extensive genetic manipulation. 90This could reflect the fact that these cells more closely resemble an HSC-independent hematopoietic progenitor, or that they are similar to true HSCs but require further maturation in a fetal liver-like environment before they attain engraftment potential.
Despite the remaining hurdle of engraftment, protocol advances in both HSPC differentiation and T cell differentiation have had a profound impact across scientific disciplines.Not only have they served as powerful platforms for studying cellular development and modeling disease, but they also have demonstrated the feasibility of PSC-derived T cell immunotherapies.In a landmark study, Themeli et al. demonstrated that induced PSCs (iPSCs) reprogrammed from a T cell clone could be differentiated into cytotoxic CAR T cells in vitro. 91Although these PSCderived T cells had functional properties more akin to cr, as opposed to ab-T cells, they were capable of eradicating CD19+ tumor cells in vitro and in a xenograft leukemia model. 91

Chemically defined T cell differentiation systems
Despite this exciting progress, the aforementioned PSC-to-T cell differentiation protocols all require animal derivatives, including serum, immortalized stroma, or basement-membranes extracted from tumor cultures.Undefined animal derivatives, including sera and extracts, introduce lot-to-lot variability, 92,93 pose a risk of contamination, 94 and mask the underlying signals that drive developmental processes.Furthermore, engineered stromal cell lines such as OP9-DL1 present a constant level of Notch ligand over time. 77These systems are not ideal for supporting differentiation processes where dynamic changes in signaling strength are required.Collectively these are significant barriers to clinical translation of xenogeneic differentiation processes.
To enable safer, tunable, and robust differentiation protocols, our group and others have focused on building chemically defined methods that avoid feeder cells, serum, and membrane extracts.A major milestone was hit in 2017 when Shukla et al. showed that primary human HSPCs derived from umbilical CB could be directed to become T cell progenitors using recombinant Notch ligands rather than OP9 stroma. 95Shukla et al. discovered that the cell adhesion molecule VCAM1 can be used to increase the strength of Notch signaling. 95Although immobilizing recombinant DLL4 alone to a culture surface was insufficient to promote T cell progenitor development, the addition of immobilized recombinant VCAM1 synergized with DLL4 to unlock T lineage differentiation potential. 95We subsequently extended this method to take CB-derived HSPCs all the way to mature T cell phenotypes using a multistage model-guided media optimization process. 96he next key challenge was to extend these chemically defined differentiation protocols to make use of PSCs, rather than CB as starting material.In 2020, Motazedian et al. demonstrated differentiation of CD4+, CD8+ DP, and CD3-cells from iPSCs in feeder-free conditions using an air-liquid-interface culture system. 97These cells could subsequently mature into CD3+, TCR+ cells on OP9-DL4 cells. 97The cells produced in their culture system express the T-lineage associated recombinase gene surprisingly early in their development. 97The authors suggested that these cells represent an HSC-independent T cell progenitor that emerges directly from hemogenic endothelium. 97riguchi et al. were the first group to differentiate PSCs all the way into mature T cells without the use of immortalized stroma. 98The authors reprogrammed iPSC lines from T cells and applied a modified version of previously established chemically defined blood differentiation process to make HSPCs. 87,89They transferred these HSPCs on to culture vessels coated with recombinant DLL4 and retronectin, a fragment of the fibronectin protein that can bind to a4b1, the same integrin that interacts with VCAM1. 98This process yielded CD8 single positive (CD8SP) cells that expressed the original TCR from the parent iPSC clone.Impressively, these cells were capable of expansion and produced effector cytokines in response to stimulation, demonstrating important functional properties of primary T cells. 98They were also capable of specifically killing target cells expressing the cognate peptide antigen of the TCR both in vitro and in vivo. 98lthough most of the results presented in the Iriguchi et al. study were generated using a serum-based media, the authors also reported a successful, although less efficient, serum-free process. 98This study is an important milestone for clinical translation of PSC-derived T cells, but the method does not appear to be effective for iPSC lines where the TCR locus is in the un-rearranged germline configuration. 98This limitation indicates that this protocol fails to capture some key aspect of T cell development.
Less than a year after the Iriguchi et al. study, Trotman-Grant et al. were the first group to report defined differentiation of abTCR+ DP T cell progenitors starting from PSCs that did not already comprise a rearranged TCR. 99These authors generated CD34+ HE cells (CD34 marks both HE cells and HSPCs) and then transitioned them to a downstream T cell differentiation culture system where Notch signals were provided by microbeads functionalized with DLL4. 99This process produced TCR+, CD3+, and DP cells at very high purities, with >80% of cells co-expressing CD4 and CD8b and *75% of cells co-expressing abTCR and CD3 by day 42 of differentiation. 99A key advantage of this process is the ability to control the timing and dose of Notch signaling by the provision of engineered beads, in a system that may not be limited by two-dimensional surface area.
Our group sought to build on these impressive results and address some important remaining limitations for chemically defined PSC-to-T cell differentiation cell differentiation. 100We aimed to overcome the low efficiency of converting PSC-derived CD34+ HSPCs into lymphoid cells and to establish defined differentiation media that could take cells all the way to a abTCR+, CD8 SP T cell state, the phenotype of functional effector T cells that leave the thymus at the end of development.
The CD34+ cells used as input into downstream T cell differentiation in the Trotman-Grant et al. study have an immunophenotype more consistent with HE, rather than HSPC. 99During ontogeny, HE cells first undergo EHT to become HSPC before they enter the thymus.HE cells likely cannot appropriately respond to differentiation conditions meant to support T cell development.To address this missing step, we established a dedicated chemically defined EHT culture system and showed that this improved the efficiency of T-lineage specification by more than an order of magnitude. 100e used a multistage model-guided optimization to build an open media formula that efficiently supports CD3+, abTCR+, CD4+, and CD8+ DP differentiation from PSC-derived HSPCs.We also used aCD3, aCD28, and aCD2 antibody complexes to mimic the TCR:HLA engagement event that drives positive selection in the thymus, allowing us to mature our DP cells into to CD8SPs.These CD8SPs had a conventional immunophenotype and possessed the ability to expand and secrete cytokines in response to stimulation.
Jing et al. have also demonstrated defined differentiation of PSC-derived functional T cells using a similar process. 101

Jing et al. and Michaels et al. provide independent validation that
HSPCs that emerge in the presence of recombinant DLL4 and VCAM1 can mature into functional T cells. 100,101The Jing et al. study also showed that knocking down the histone methyltransferase EZH1 can significantly improve the efficiency of T cell differentiation. 101As EZH1 is a positive regulator of Notch signaling, this study further emphasizes the importance of carefully titrated Notch activation during T cell development. 101fter decades of enabling foundational work, the past 2 years have seen a burst of publications in the area of defined T cell differentiation from PSCs, moving this paradigm ever closer to clinical translation.

Mastering T Cell Development by Controlling the Magnitude and Timing of Key Signaling Programs
Recent advances in chemically defined T cell differentiation systems will facilitate clinical translation of PSC-derived immunotherapies.In addition to reducing the risk of contamination and improving consistency, defined systems are inherently more tunable.This makes them easier to iteratively optimize.
3][104][105][106] In vivo studies and in vitro differentiation experiments have collectively shown that successful T cell differentiation, and the functional properties of the resulting T cells, depend on the timing and magnitude of a small set of key regulatory programs, including Notch-, TCR-, and cytokinesignaling.Defined differentiation platforms afford the unique and timely opportunity to precisely manipulate these key pathways, setting the stage for an era of rapid progress in T cell manufacturing. 96,100,101

ENGINEERING ALLOGENEIC T CELL THERAPIES
We envision that two foundational approaches will work in concert to realize this vision-bioengineering and synthetic biology.Hereunder, we provide a perspective on some of the most compelling opportunities to advance PSC-derived T cell production by controlling the timing and strength of key signaling programs by engineering the cell-extrinsic niche and cellintrinsic genetic programs.
Engineering cytokine signaling to improve differentiation efficiency and reduce cost T cell development from HSPCs requires activation of a collection of cytokine receptors, initiating a cascade of signal-transduction events that ultimately activate and repress a collection of transcription factor programs that drive cells forward through development.IL-7 and FLT3L are both essential for T cell development. 67,77Additional cytokines can enhance or accelerate this process. 96The signal transduction cascades and gene expression programs that mediate the effect of these cytokines are both context dependent and share overlapping nodes with other cellular signaling processes. 107This makes it extremely difficult to assess the impact of altering the dosage and timing of exogenous cytokines.Getting the dynamic levels of cytokines just right has major implications for the clinical success of T cell differentiation pipelines.The wrong cytokine dosages can drive cells toward alternate fates or completely block differentiation. 96

Niche engineering of cytokine signaling
To address this challenge, we and others have successfully employed statistical design of experiments to rapidly and comprehensively explore the space of cytokine combinations 96,108 (Fig. 2A).Importantly, we encourage the use of models that can account for changing cytokine requirements over time.By understanding the ancestor-progeny relationships during differentiation, it is possible to apply multistage statistical learning to maximize the yields of intermediate and terminal cell types over the course of a long developmental process. 96 typical approach is to use one of several statistical models to choose a limited set of cytokine dosage combinations for a fixed time period in the differentiation process.Cytokines of interest could be manually selected from the literature or identified using high-throughput screens of receptor knockouts or cell-cell signaling interactions. 109Yields of a target population are quantified, often using flow cytometry for a set of cellstate defining surface markers.Next, a model is constructed based on the sampled cytokine combinations and extrapolated to the untested parameter space. 96Optimal cytokine concentrations are predicted and then experimentally validated for their ability to increase the yield of the target immunophenotype at each interval. 96n the future, we propose screening for conditions that maximize the yield of cells with desired functional properties such as target cell killing, cytotoxic cytokine secretion, proliferation, and persistence, as opposed to cell surface phenotypes (Fig. 2A).Ultimately, the goal of T cell immunotherapy is to safely and effectively eradicate cancer.Producing PSC-derived cells with a surface marker expression profile that matches primary T cells is an important first step and making a ''T-cell like'' product will ease the path to regulatory approval.But it is not clear whether

A B
FIG. 2. Engineering cytokine signals to enhance in vitro T cell differentiation.
(A) Schematic overview of a general model-guided approach to optimizing the dosage and timing of cytokine exposure required to support functional T cell differentiation.
(B) Schematic overview of a synthetic biology approach where cytokines are manufactured by the differentiating cells themselves.DP, double positive T cell progenitor; pro T, T cell progenitor.

MICHAELS ET AL.
PSC-derived cells that phenotypical resemble conventional abT cells are actually the best population for immunotherapy.1][112] It is conceivable that the most useful cells might not even show clear resemblance to any primary equivalent.As our clinical knowledge grows, we will find better phenotypical and functional correlates with therapeutic success.These new markers can serve as the goal posts for optimizing in vitro differentiation conditions.

Cell engineering of cytokine signaling
To complement this dynamic media development strategy, synthetic biology approaches could be used to write differentiation programs into the stem cell genome. 113Although synthetic biology strategies have not been applied to T cell differentiation extensively as of yet, strategies of interest include ''forward programming''-the controlled expression of transcription factors mimicking developmental trajectories; 114 or engineering the differentiating cells to express and secrete factors that can engage with the cells' endogenous receptor signaling programs to promote differentiation (Fig. 2B).
Because cytokines need to be dosed and timed appropriately for successful differentiation, 96 crude cytokine overexpression techniques have thus achieved limited success.][117][118] We recently demonstrated the feasibility of this strategy using a PSC model of gastrulation.
We engineered stem cells to express and secrete BMP4, a factor that is typically added exogenously to promote germ-layer differentiation. 119By fine-tuning BMP4 expression levels using synthetic microRNA target sites, 118 we were able to precisely control the differentiation outcome and completely obviated the need for exogenous BMP4. 119As the field makes strides to address the ongoing challenges of large payload delivery, 120 off-target editing, and transcriptional silencing, 121 synthetic control over cellular differentiation becomes an increasingly plausible manufacturing strategy.

Making the right T cells by controlling the magnitude and timing of TCR signaling
Recent progress has enabled efficient differentiation of functional CD8+, effector T cells from PSCs with TCR loci in the germline configuration.However, when the starting PSCs comprise a fully recombined TCR or constitutively express a CAR, these differentiation protocols tend to produce cells with innate-like functional properties instead of conventional abT cells. 40,91A related limitation is that chemically defined differentiation protocols can make CD8+ effector T cells 98,100 but defined differentiation of functional CD4+ helper T cells has not yet been demonstrated.Both of these issues stem from a common underlying problem-failure to provide the correct timing and magnitude of TCR signaling in vitro.

Overcoming premature TCR-like signals to make conventional T cells
A recent study made the interesting observation that premature TCR signaling (that can be imparted by a pre-rearranged TCR or CAR) can dampen Notch pathway activity. 40This increases the threshold of Notch input signals required to promote conventional T cell development.This model also helps to explain the known phenomena that cdT cells, which produce a fully formed TCR earlier in development, 122 require higher levels of Notch input signals to successfully differentiate. 123Thus, carefully managing the complex interplay between Notch and TCR signaling is likely to be very important for efficient in vitro differentiation of T cells harboring an engineered TCR or CAR.
Recognizing this phenomenon, the authors elegantly showed that targeting the CAR into the TRAC locus, which delays CAR expression until after beta selection, provides the right timing of complete TCR expression, and makes cells more similar to ab T cells when coupled with an appropriately strong Notch ligand. 40This can further be enhanced by modifying the costimulatory domains to get the magnitude of CAR signaling closer to a physiologically relevant range. 40Although this is an important step, the resulting T cells still displayed an unconventional phenotype, producing significantly lower levels of effector cytokines compared with primary T cells. 40inding a way to precisely juggle TCR and Notch signaling over developmental time could enable production of conventional cells from iPSCs engineered to express CARs and TCRs.On the side of the TCR signal, this can be achieved by inserting bona fide TCRs in the TRAC locus, rather than CARs, matching physiological signaling levels.The timing of CAR activation can be regulated using split-CAR designs; for instance, separating the scFV from the intracellular domains 124 or separating different modules of the intracellular domain. 125These different CAR architectures can be induced to dimerize by the user, enabling specific activation of the CAR at a timepoint of interest during differentiation.
On the side of controlling Notch signaling, stronger Notch ligands can be developed through in vitro protein evolution, as was recently demonstrated by Gonzalez-Perez et al., 126 or through the use of alternate high-affinity binders such as Notch-activating antibodies 127 tethered to beads or biomaterials (Fig. 3A).Alternately, artificial Notch signals could be provided using synthetic biology approaches that bypass mechanical activation at the cell membrane.For example, overexpressing a constitutively active form of Notch, 67 or fusing Notch's DNA-binding partner RBPJ to a generic transcriptional activator like VP64 may provide sufficiently strong Notch signaling to counteract the repressive effects of a CAR (Fig. 3A).

Unlocking CD4 T cell development with niche engineering and synthetic biology
Defined T cell differentiation protocols have thus far failed to generate functional CD4+ helper T cells.CD4+ cells improve the efficacy of CAR T therapy 128,129 and have applications for treating infection, autoimmunity, and transplant rejection.Thus, producing CD4+ T cells from PSCs has substantial translational potential.Below, we briefly outline the biology underlying CD4 versus CD8 lineage choice, then suggest several strategies by which these biological mechanisms could be leveraged to produce CD4+ T cells in vitro.

ENGINEERING ALLOGENEIC T CELL THERAPIES
The decision to adopt a CD4 helper fate versus CD8 effector fate occurs when a TCR expressed by a CD4+, CD8+ DP progenitor engages with a class I or class II HLA molecule.Cells that successfully engage with HLA-I-expressing cells are directed to become CD8SP, whereas TCR engagement with HLA-II promotes differentiation to CD4SP. 130CD8 and CD4 are co-stimulatory molecules that enhance downstream TCR signaling.TCR activation downregulates CD8 expression which interrupts signaling for HLA-I restricted cells.In contrast, CD4 expression persists after TCR engagement, leading to an enduring signal in HLA-II restricted T cells. 130he duration of TCR signaling during this selection event ultimately dictates whether DP cells become CD8 cytotoxic, or CD4 helper cells. 130In current in vitro T cell differentiation protocols, positive selection is triggered either by developing cells selecting off of one another, 79 or using antibodies against the TCR complex. 98,100In humans, HLA-II expression is largely restricted to antigen-presenting cells and thymic epithelial cells, whereas HLA-I is expressed broadly across cell types.Thus, cells undergoing in vitro differentiation likely have access to HLA-I on other developing T cell progenitors but lack a sufficient source of HLA-II offering one explanation for the lack of CD4+ T cells in current protocols.
One way to provide an appropriate TCR signal to facilitate CD4 differentiation would be to coculture the developing T cell progenitors with primary or immortalized antigenpresenting cells.This is an important advantage of technologies such as the ATO platform, 79 or differentiation systems that utilize PSC-derived thymic epithelial cells. 131ather than using living material to trigger TCR:HLA-II interactions, recombinant proteins provide a scalable and clinically translatable alternative (Fig. 3B).Soluble or plate-bound recombinant peptide-HLA-II complexes may be sufficient to drive CD4+ fate specification from DP progenitors.But it is not clear how the choice of peptide and HLA-II allele will shape the repertoire of TCRs that can successfully undergo positive selection.A more generalizable approach that would be agnostic to the TCR itself is to use plate-bound antibodies that simultaneously ligate the TCR-constant region and CD4 130 (Fig. 3B).
Longer-term cell engineering could be used to directly activate downstream TCR signaling.When the TCR engages with MHC, protein tyrosine kinases including Lck are recruited to TCR/CD3 complex. 132Lck phosphorylates immunoreceptor tyrosine-based activation motifs (ITAMs) on the cytosolic domain of CD3, which triggers a complex downstream signaling  ENGINEERING ALLOGENEIC T CELL THERAPIES cascade. 132It may be possible to construct an inducible protein dimerization system whereby an exogenous small molecule triggers Lck recruitment to the CD3 ITAMs (Fig. 3B).A similar approach was applied to make CAR T activation dependent on a small molecule inducer. 133oiding unwanted by-products to increase T cell yield and purity One compelling opportunity to reduce the cost of in vitro T cell manufacturing is to increase the purity and yield of desired target populations at each stage of differentiation.En route to making T cells, protocols first specify PSCs to the mesoderm lineage, then through a HE intermediate, then an HSPC state, next to become T cell progenitors and finally mature T cells. 88At every stage, unwanted off-target populations are produced. 88hese cellular by-products consume resources and limit the maximum yield of the target population per-input PSC.Off-target populations may also compete directly with the desired intermediate. 134A proven solution is to enrich for the target population at key transitions in the culture.Exemplifying this approach, we and others routinely select for CD34+ cells early during differentiation by antibody-mediated magnetic separation. 99,100Additional enrichment steps, such as at the CD7+ T cell progenitor stage, would likely increase purity and efficiency but would also further complicate the manufacturing workflow and may not reduce the net product cost (Fig. 4).
As an alternative approach, a cellular engineering solution is to block access to unwanted cell fate trajectories by knocking out genes that are required for their development (Fig. 4).Pruning the developmental tree in this manner should allow cell culture resources to be directed efficiently towards supporting T cell development.There is already a strong precedent that knocking out an important fate-specifying gene can improve the efficiency of in vitro T cell differentiation.By knocking down EZH1, Jing et al. were able to substantially increase the yield of CD3+ T cell production from PSCs by *threefold compared with control conditions. 101n addition to blocking access to unwanted developmental paths, cellular engineering can also be applied to enhance or accelerate progression along the target trajectory.6][137][138][139][140] Although this approach has progressed at a modest pace for the past decade, recent advances in high-throughput perturbation screening technologies are poised to drastically accelerate progress in this space. 141

Summary and Conclusion
T cell therapies are disrupting the way we treat disease, but widespread adoption of this modality is limited by our ability to cost-effectively manufacture the right cell types at scale.PSCs are a compelling solution to this challenge that benefit from decades of fundamental research into T cell development using feeder-based differentiation protocols.
Presently, the field is in transitioning to chemically defined bioengineered differentiation systems.Not only are these contemporary approaches more readily translatable to the clinic, but they also afford much more precise control over the levels and timing of signaling inputs during differentiation.T cell development from PSC follows a complex multistage process that is dependent on a carefully balanced orchestra of signals and bioengineering is well suited to accurately reproduce this process in vitro.
We anticipate that synthetic biology will drive the next major shift in T cell manufacturing from PSCs.Parallel improvements in our fundamental understanding of T cell differentiation and our ability to synthetically control cell function are setting the stage for more efficient and cost-effective cell therapies.Ultimately, our ability to generate multiple T cell subtypes at scale will enable new therapeutic strategies in oncology and beyond.

BFIG. 1 .
FIG. 1.Using developmental biology to guide in vitro T cell differentiation.(A)Top: Schematic outline of T cell development in the human thymus.Bottom: Relative importance of Notch and TCR signaling during T cell development.(B)Feeder-based and chemically defined T cell differentiation protocols.PSC-derived CD34+ HE cells or HSPCs are cultured on immortalized stromal cells expressing Notch ligands to promote T cell differentiation in monolayer cultures.PSC-derived CD34+ HSPC cells are incorporated into spheroids with stromal cells expressing Notch ligands to promote T cell differentiation in artificial organoid cultures.In chemically defined stepwise differentiation protocols, PSC-derived CD34+ cells are cultured on plates coated with DLL4 and VCAM1, with different chemically defined media to induce EHT, Pro-T, and DP T cell development, respectively.DP cells are then exposed to nonspecific TCR stimulation to promote further maturation to the CD8SP stage.CD8SP, CD8 single positive; DP, double positive; EHT, endothelial to hematopoietic transition; HE, hemogenic endothelium; HSPC, hematopoietic stem and progenitor cell; PSC, pluripotent stem cell; TCR, T cell receptor; TSP, thymus-seeding progenitors.
FIG.4.Avoiding unwanted cell populations through affinity-based targeted enrichment and gene editing.
Left: Cell populations of interest can be enriched with antibodies through fluorescence-activated cell sorting or magnetic purification.Right: Gene editing can be deployed to block access to undesired cell fates.
Methods for manipulating Notch signaling pathway activity.Niche engineering approaches may involve coating plates with immobilized Notch ligands such as DLL4, or anti-Notch receptor antibodies to activate the Notch pathway in developing T cells.Cellular engineering approaches may involve expressing the Notch intracellular domain from a promoter that can be activated by a small-molecule inducer.
(A) (B) Methods for manipulating TCR pathway activity.Niche engineering approaches may involve coating plates with recombinant peptide-MHC complexes for antigen-specific TCR stimulation, or anti-TCR antibodies for nonspecific TCR stimulation.Cellular engineering approaches may involve fusing CD3 and LCK to domains such as FKBP and FRB, which dimerize in response to the chemical inducer Rapamycin, enabling user-regulated fusion of LCK and CD3 to activate TCR signaling.