The potential of allogeneic CAR-T lies largely in the ability to mass-produce CAR-Ts that are as efficient and potent as their autologous counterpart. One of the crucial factors in the manufacturing of allogeneic CAR-Ts lies in the source material used for the final product.

Currently, the most frequently used allogeneic cell source for CAR-T manufacturing involves using peripheral blood mononuclear cells (PBMCs) from a random healthy donor. More rarely, other cell sources are used such as umbilical cord blood (UCB) or a renewable cell source such as induced pluripotent stem cells (iPSCs).

The most frequent source for the manufacturing of allogeneic CAR-Ts is PBMCs collected from healthy donors, where T-cells are isolated and expanded. This allows for the creation of multiple vials from a single apheresis product that can be easily used in a very rapid and standardized manufacturing protocol ^[1][2][9,10]. Another manner by which CAR-T can be manufactured is via isolation of stem cells from PBMCs ^[3][11], which can be further activated and transduced into CAR T-cells. The advantage of using hematopoietic progenitor cells is their ability to self-renew; however, as their absolute numbers are limited, other strategies may be needed for their enrichment, such as CD34+ mobilization, similarly to what is performed for autologous stem cell transplantation ^[4][12]. This also allows for the generation of a bank of cells that express different human leukocyte antigen (HLA) subtypes to potentially match the donor HLA to that of the patient ^[5][13]. The selection of donors on the basis of their immune characteristics is likely to be a key factor in decreasing the heterogeneity in the final manufactured product and lower the risk of GvHD.

The use of UCB was shown to be associated with reduced incidence and severity of GvHD, making it a potentially more tolerable source material than PBMCs for allogeneic T-cells and allowing for less stringent HLA matching ^[6][14]. Furthermore, UCB is an enriched source of hematopoietic stem cells (HSCs), which are able to self-renew and can be used to differentiate into T-cells, although there is a limit to their total number ^[7][8][15,16].

Interestingly, T-cells isolated from UCB have a unique antigen-naive status which is probably linked to the decreased alloreactivity observed in UCB grafts ^[9][10][17,18]. Furthermore, UCB T-cells are characterized by impaired nuclear factor of activated T-cells (NFAT) signaling and reduced activity, which most likely further contributes to the reduced GvHD ^[11][19].

However, an obvious drawback of UCB is its limited availability compared to other cell sources.

T-cells derived from iPSCs can also be used as a source of CAR-Ts ^[12][20]. In theory, iPSCs have an unlimited capacity for self-renewal, thus allowing them to be banked and used indefinitely ^[13][21]. A bank of iPSC lines with different homozygous HLA combinations could be generated to minimize the risk of allorejection of CAR-T derived from iPSCs ^[14][22]. An advantage of using iPSCs is that CAR T-cells can be generated from a single iPSC clone with the capacity for clonal expansion, and therefore, the genetic modifications they undergo would be homogeneous in the final cell population ^[15][23]. However, the quality controls should be strict because undifferentiated proliferating iPSCs may compromise product safety, since they could induce important adverse effects such as teratomas ^[16][24].

iPSCs can be developed from different cell types, such as fibroblasts or lymphocytes, that are reprogrammed into a less differentiated cell by inducing the expression of specific factors. For example, Iriguchi et al. generated iPSCs from an antigen-specific cytotoxic T-cell clone, or from TCR-transduced iPSCs, as starting material ^[17][25]. These iPSCs can then in turn be differentiated into T-cells through the addition of several differentiation drivers and/or inhibitors (SDF1α and p58 inhibitors in the above case, for example) to enhance T-cell commitment. While the potential to create a large cell bank that covers a study cohort is appealing, the arduous tasks of T-cell differentiation and selection leading up to the commitment of a single positive T-cell is much more complex then the use of T-cells isolated from either PBMCs or UCB. However, while PBMCs and UCB both offer a heterogenous T-cell population of cells, iPSCs are clonal and thus give rise to a homogenous T-cell population both with the advantages/disadvantages of each.

The generation of allogeneic CAR-T irrelevant of the starting material faces two major hurdles. The first is the induction of GvHD and the second is the HvG response. Each T-cell expresses a T-cell receptor (TCR), where the majority of T-cells express a TCR composed of an alpha and a beta protein chain that can recognize HLA-peptide complexes on target cells through the direct pathway of allorecognition, thus leading to GvHD ^[18][19][26,27] independent of the CAR.

The use of allogeneic donor T-cells (CAR or not) that still express a functioning TCR may play a role in anti-tumor effects. This has been clearly demonstrated in leukemia in a process termed graft-versus-leukemia (GvL): after allogeneic stem cell transplantation (SCT), protection from relapse is partly due to donor T-cells that recognize leukemia-specific minor antigens ^[20][28]. This may be similar to CAR-T, although the recognition of allo-antigens will likely induce GvHD, as studies assessing both acute and chronic GvHD have clearly established a central role for αβTCR in GvHD pathogenesis ^{[21][22][23][24]}[29,30,31,32]. The application of SCT, for example, was not appreciated until T-cell-depleted grafts were assessed to eliminate GvHD ^[25][26][33,34]. These successfully decreased the occurrence of GvHD to extremely low frequencies, although the risk of opportunistic infections and relapse increased substantially ^[26][27][34,35]. While the role of αβTCR in GvHD development is not in doubt, the possible risks and/or benefits in the case of CAR-T therapy are not completely clear, and the development of GvHD may be relatively low ^[28][36].

To avoid GvHD, two main approaches exist depending on (i) T-cells that have low or non-reactive TCRs (discussed in this section) or (ii) engineering methods to avoid allorecognition (Section 4). The αβTCR repertoire is selected in the thymus and is educated based on the ability to be tolerant to self-HLA complexes. This tolerance means that the TCR recognizes the self-HLA and responds to non-self peptide. However, in the case of allorecognition, the TCR recognizes both structurally similar HLA-peptide complexes and dissimilar HLA-peptide complexes, therefore allowing for the high frequency of alloreactive T-cells (1 in 10³) ^[29][37]. It is these alloreactive αβTCRs expressed on T-cells that drive GvHD.

The HLA locus is the most polymorphic region in the human genome, thus leading to many HLA variants in each individual. There are six HLA-class-I molecules and six HLA-class-II molecules, making the matching between donor and patient a complex issue, and although decades of data from transplantation centers have shown that the most important HLAs to match are the class I HLAs A,B and class II HLA-DR ^[30][31][38,39], this still requires a vast bank of cells in order to produce the CAR-Ts, which renders the allogeneic manufacturability rather complicated.

Patients treated with allogeneic SCT can be subsequently treated with CAR-Ts generated from the same donor if they relapse. This was performed in a study by Brudno et al., where 20 patients with B-cell malignancies received CD19 CAR-Ts generated from the same donor as SCT with no chemotherapy administered before T-cell infusion. Six patients achieved complete remission and two patients achieved a partial response. No GvHD was reported ^[32][40]. These results confirmed previous observations made by other groups ^[33][34][41,42]. In a more recent study, eight r/r B-ALL patients received either HLA-matched (n = 4) or HLA-haploidentical (n = 4) CD19 CAR-Ts immediately preceding an intended SCT ^[35][43]. The haploidentical CAR-Ts induced transient or no reduction in peripheral blood leukemia cells with no significant CAR-T expansion, which suggests rejection. In contrast, patients treated with the HLA-matched CAR-Ts exhibited higher complete response rates, although more severe toxic side effects, with no GvHD observed in either group. However, only three out of eight patients reached complete response and only two of the eight patients proceeded to transplant, indicating that while HLA-matched and HLA-haploidentical allogeneic CD19 CAR-Ts are feasible in r/r B-ALL before SCT, other factors besides GvHD need to be considered in clinical applications of allogeneic CAR T-cell infusions.

T-cells with a specific memory phenotype are considered to have a TCR specificity directed to previously detected antigens, which are expected to be different from those of the patient receiving the CAR-T therapy. Interestingly, studies have shown that memory T-cells do not induce GvHD ^[36][44]. It is unclear why this is the case, but one possibility is the diversity of the TCR, which is limited in memory T-cells, thus reducing GvHD. One manner by which the T-cell memory and TCR specificity can be further specified is through selection or the development of virus-specific T-cells (VST), as has been achieved in Epstein–Barr virus (EBV)-associated malignancies. Adoptive transfer of HLA partially matched EBV-specific T-cells from healthy donors has had positive results in post-transplant lymphoproliferative disease for example, with response rates of 60–70% and low incidences of toxicity or GVHD ^[37][45]. Infusion of EBV-specific T-cells has also been used in patients with Hodgkin’s lymphoma with good tolerance and remission rates ^[38][39][46,47]. The use of viral antigens can enhance the proliferative capacity of the allogeneic CAR-Ts, making them persist longer and possibly enhance their efficacy. This has been shown with cytomegalovirus (CMV)-specific CD19-CAR-Ts that had enhanced in vivo anti-tumor activity following the administration of anti-CMV vaccination ^[40][48].

However, all these methodologies require partial matching and thus require the creation of multiple cellular banks. Next to the above-mentioned options, sub-populations of T-cells can be used for the generation of allogeneic CAR-Ts.

T-cell sub-populations comprise a relatively low percent of the circulating total T-cells (making up anywhere between 0.01 and 10% of T-cells). These sub-populations include double-negative T—cells (DNTs); invariant Natural Killer T-cells (iNKT); cytokine-induced killer (CIK) cells; mucosal-associated invariant T (MAIT)-cells; and lastly, γδT-cells.

DNTs are a rare subset of immune cells that express CD3 but not CD4, CD8, and CD1d-αGalCer ^[41][42][43][49,50,51]. DNTs comprise about 1 to 5% of human PBMCs and can be isolated and expanded ex vivo under clinically compliant conditions from the peripheral blood of healthy donors ^[44][45][52,53]. Expanded DNTs can express either γδTCR or αβTCR, where the frequency of TCR expressing DNTs can range between 60 and 90% depending on the donor origin.

In a recent study conducted by Vasic et al. the feasibility, safety, and efficacy of DNTs for the development of allogeneic CD19-CAR-T was assessed. The resulting allogeneic CD19-CAR DNTs had the properties of an off-the-shelf cellular therapy and were effective against CD19-expressing hematological and solid malignancies ^[46][54]. Pre-clinical studies have thus confirmed the feasibility of DNTs, but whether DNTs will actually yield good results clinically remains to be seen.

A phase I/IIa clinical trial using third-party-donor-derived, genetically non-modified DNTs to treat patients with relapsed/refractory acute myeloid leukemia (AML) showed that the therapy was safe and had a positive efficacy profile ^[47][55]. One major concern is regarding the cellular efficacy. Interestingly, Kang et al. have shown that one manner by which the cellular efficacy and persistence of DNTs CARs can be enhanced is through inhibition of the PI3K pathway during manufacturing, Which is something that we and others have seen in αβT-cells as well ^[48][49][56,57].

Invariant NKT-cells (iNKTs) are a subset of T-cells that share morphological and functional characteristics of both NK and T-cells. They have a restricted TCR that has a constant α-chain paired with a low-diverse β-chain. iNKTs comprise between 0.01 and 1% of the peripheral blood T-cell population and have shown not to cause GvHD in xenograft models ^[50][51][52][58,59,60]. They are restricted by CD1d, a glycolipid-presenting HLA-I-like molecule expressed on B-cells, antigen-presenting cells and some epithelial cells ^[53][54][61,62]. The fact that iNKT-cells recognize B-cell lymphomas through CD1d makes them of particular interest for B-cell malignancies ^[55][63].

CIK-cells are a heterogenous population of polyclonal effector T-cells that have functional NK-cell properties. They comprise between 0.01 and 1% of the peripheral blood T-cell population and can be expanded from PBMCs, bone marrow and UCB through a manufacturing process that involves the addition of cytokines like IFN-γ and IL-2 and TCR-activating antibodies ^[56][57][64,65]. CIK-cells have the advantage of exerting non-HLA-restricted cytotoxicity and very low alloreactivity across HLA barriers in comparison with conventional donor lymphocyte infusion ^[58][59][60][66,67,68]. This was further confirmed by preclinical and phase I/II studies, where the infusion of bulk CIK-cells population was well-tolerated ^[61][62][63][69,70,71]. In addition to the alloreactivity, the dual activity (of both NK cell receptors and TCRs) gives CIKs an added ability to mediate cytotoxicity and prevent infection, which is a major concern after CAR-T therapy. In a recent clinal trial where relapsed B-acute lymphoblastic leukemia (B-ALL) patients were treated with CD19 CAR CIK-cells, no GvHD was observed, and the cells could be detected up to 10 months after infusion ^[64][72]. The overall response rate was 61.5% (13 patients), which is in line with its autologous counterpart.

MAIT-cells are primarily localized to mucosa-rich regions, comprising a fraction of T-cells distributed throughout the pulmonary (5%), hepatic (20–40%) and intestinal (1–2%) lamina propria, as well as peripheral circulation (1–10%; ^[65][66][67][73,74,75]). MAIT-cells have a heavily restricted TCR repertoire that consists of TCR alpha variable (TRAV)1 combined with three kinds of TCRA junctionals (TRAJ; TRAJ33, TRAJ12, TRAJ20) and a limited repertoire of β chains in humans ^[68][76]. The MAIT TCR can recognize modified derivatives from the vitamin B2 synthesis pathway presented by MHC class I-related molecule MR1 on APCs. MR1 is a conserved molecule, thus making MAIT-cells incapable of inducing strong GvHD in vivo ^[69][77]. This has further been shown in clinical studies where MAIT-cells were positively correlated with improved survival and fewer allogeneic adverse events ^[70][78].

The use of MAIT-cells for CAR-T has been assessed in multiple pre-clinical studies, and while their efficacy against tumor antigens was significant (as assessed with a mesothelin and a CD19-targeting CAR), significant concerns were raised based on both cellular persistence and manufacturing due to the limited cell number ^[71][72][79,80]. These concerns imply that the use of MAIT-cells clinically may be limited.

One other subset of T-cells that is currently being used extensively in both preclinical and clinical studies are γδT-cells (reviewed elsewhere ^[73][74][75][81,82,83]), which represent 1–10% of circulating T-cells (although they are also prevalent in some epithelial tissues; ^[76][84]). The γδT-cells have a unique TCR composed of variable gamma and delta chains and recognize antigens independent of the HLA, leading to low or no risk of GvHD ^[77][78][85,86]. It is this advantage that has made them a popular starting material for the creation of allogeneic CAR-Ts, and at least a dozen trials are currently underway to assess this as a viable option ^[73][75][79][81,83,87].

Several studies have shown the safety and some efficacy of γδT-cells’ transfusion into cancer patients, thereby relying on the HLA-independent function of γδT-cells (mediated by NKG2D, for example, among others; ^[80][81][88,89]). These studies imply that the use of γδT-cells may prove beneficial as a CAR-T therapy. This observation has led to multiple CAR-T- and TCR-based strategies being employed by companies to improve the efficacy of γδT-cells for cancer immunotherapy. However, the tumor toxicity has been limited, and consistent problems with both persistence and homing in vivo has limited the translation of γδCAR-Ts.

The strategies to reduce GvHD by using partially matched allogeneic material, and/or T-cells that have low or no TCR, naturally offer good alternatives, and many CAR-Ts have shown the alloreactivity to be limited or manageable. However, in most instances, allogeneic cells are persistent for a very short amount of time, meaning that the lack of GvHD may be due in part to the lack of persistence. This lack of persistence is driven by multiple-factors, including (i) a resurgence of the host immune response (in most instances, the patients undergo lymphodepletion prior to CAR-T therapy) that in turn rejects the allogeneic cells, (ii) the immune-suppressive tumor microenvironment that may inhibit T-cell proliferation, as well as other factors ^[82][90]. This requires additional engineering to circumvent the host immune response and/or the tumor microenvironment. The different methods can be divided into gene-editing technologies and non-gene-editing technologies.

The two biggest hurdles in the use of allogeneic T-cells are GvHD and HvG. The former can be avoided by eliminating the TCR, usually through the knockout (KO) of the constant domain of one of its chains (α and/or β), or by replacing some TCR subunits, which impedes its antigen recognition function ^[83][91]. However, although this takes care of the alloreactivity, the cells would still be susceptible to HvG. The most common antigens driving HvG are the mismatched donor-HLA-I molecules on the donor cells. These are recognized by the patient αβT-cells that are CD8+ through the direct pathway of allorecognition. By knocking-out the common subunit β2-microglobulin (encoded by the B2M gene), the HLA-I molecule will not be expressed on the cell surface, thus making the cell susceptible to NK-cell lysis ^[84][92]. To avoid recognition by NK-cells, different strategies have been developed, most commonly utilizing overexpression of a non-classical HLA-I such as HLA-E or G fusion protein to avoid lysis ^[85][86][93,94].

Other strategies to avoid HvG include (i) CD47 overexpression ^[87][95] and (ii) CD52 KO ^[88][96]. CD47 is found on both healthy and malignant cells and regulates macrophage-mediated phagocytosis by sending a “don’t eat me” signal to the signal regulatory protein alpha receptor. Upon depletion of HLA-I on CAR-Ts, recognition by both macrophages and NK-cells is triggered. In a recent study by Hu et al., the overexpression of CD47 in allogeneic CD19-CAR-T negated the recognition of NK and macrophages to the absence of HLA on the cell surface, thus avoiding rejection ^[89][97]. This approach is currently under investigation in a phase I clinical trial (NCT05878184).

CD52 is protein expressed on the cell surface of many immune cells such as mature lymphocytes, NK-cells, monocytes/macrophages and others ^[88][90][96,98]. The humanized anti-CD52 monoclonal antibody (mAb), alemtuzumab, has been widely used in clinics for the treatment of transplant patients and B-cell chronic lymphocytic leukemia ^[91][92][93][99,100,101]. Alemtuzumab targets CD52+ T-cells and is capable of both complement-dependent cytotoxicity and antibody-dependent cell-mediated cytotoxicity ^[92][100]. Therefore, CD52 KO in allogeneic CAR-Ts can be combined with Alemtuzumab to enhance CAR persistence. However, this will necessitate multiple infusions and close monitoring of the immune system of each patient. This approach has been assessed in multiple clinical trials involving allogeneic CAR-Ts, most notably by Allogene, who have used this in combination with CD70 ^[94][102] and CD19 CAR-Ts ^[95][103].

Zinc finger nucleases (ZFNs)—A ZFN is an artificial endonuclease that has a zinc finger protein (ZFP) fused to the cleavage domain of the FokI restriction enzyme ^[96][107]. A ZFN is targeted to cleave a chosen genomic sequence. The FokI cleavage domain needs to be dimerized to cut DNA, and because the dimer interface is weak, a construct of two sets of fingers directed to neighboring sequences is needed. The cleavage-induced event caused by ZFN leads to a cellular repair process that mediates the efficient modification of the targeted locus. If the event is resolved via non-homologous end joining (NHEJ), it can result in small deletions or insertions, effectively leading to gene KO. If the break is resolved via a homology-directed repair (HDR), small changes or entire transgenes can be transferred into the chromosome. Because each zinc finger unit recognizes three nucleotides, three to six zinc finger units are needed to generate a specific DNA-binding domain.

TALEN—TALENs are similar to ZFNs in that they are heterodimeric nucleases that contain a fusion between the FokI restriction enzyme and a transcription activator-like effector (TALE) DNA-binding domain. The amino acid repeat variable di-residues (RVD) are two hypervariable amino acids that make up part of the sequence that mediates the binding of TALE to DNA ^[97][112]. This greatly simplifies TALEN design. The TALENs’ monomeric architectures are developed by fusing TALE domains to a sequence-specific catalytic domain derived from the homing endonuclease (HE) I-TevI, resulting in a Tev-TALE monomeric nuclease ^[98][113].

MegaTALs—MegaTALs are a short TALE domain that fused to the homing endonuclease (HE). The artificial chimeric nucleases derived from HEs can be engineered to target specific sequences within the genome ^{[99][100][101]}[117,118,119]. This fusion increases the specificity and activity of the MegaTALs ^[102][120]. Currently, to our knowledge, no clinical trials are utilizing MegaTALs for allogeneic CAR-Ts.

Clustered regularly interspaced short palindromic repeats (CRISPR)—The CRISPR system is derived from microbial adaptive immune system. It combines a nuclease and a short RNA. The specificity of the CRISPR system is not through the protein-DNA interaction (like the above) but rather RNA-DNA base pairing. A 20 nucleotide RNA that is complementary to the target DNA(termed single guide RNA; sgRNA) is responsible for the specificity. However, due to the system, off-targets are tolerated ^[103][104][121,122]. The most common nuclease is Cas9 ^[105][123]. CRISPR/Cas9 is the most widely used because it has demonstrated a remarkably low rate of off-target mutagenesis in T-cells ^[106][107][124,125]. In addition, a specific high-fidelity Cas9 mutant, called eSpCas9, did not cause any detectable off-target effect, making it an even safer technology ^[108][109][126,127].

CRISPR/other Cas—The most widely used CRISPR-Cas system is CRISPR/Cas9; however, there are multiple systems, which are generally divided into two classes (class 1 and 2), and subsequently subdivided into six types (types I through VI). Class 1 (types I, III and IV) systems use multiple Cas proteins, while class 2 systems (types II, V and VI) use a single Cas protein ^[110][134]. The class 1 CRISPR/Cas systems comprise 90% of all identified CRISPR/Cas loci. Class 2 comprises the remaining 10% and is almost exclusively in bacteria ^[111][135]. Cas9 (type II) still presents challenges, mostly due to the possibility off-targets and difficulty in delivering ribonucleoprotein particles ^[110][134]. The second most utilized Cas is Cas12a (type V). It has substantial differences in comparison with Cas9 in multiple aspects, one of which is a higher gene repression in the template strand of the target DNA than SpdCas9 ^[112][136]. It may also be easier to multiplex in comparison with Cas9 ^[113][137]. However, both Cas 9 and 12a suffer from a dependence on host cell DNA repair machinery, meaning the induction of DSB and induction of repair. Although both technologies have been used successfully to insert specific DNA into the genomic loci, their efficiency differs between cell types ^{[114][115][116]}[138,139,140]. Furthermore, DNA repair through HDR is also related to active cell division, meaning that cells that do not divide (like neurons) render the tools ineffective.

Base-pair editing—Base editing involves the use of CRISPR-Cas9 (or other Cas) together with avoidance of DNA DSBs during genetic modification. Fusing a single-strand DNA (ssDNA) deaminase enzyme to a catalytically inactive Cas9 variant leads to there being only an ssDNA cut (nick). The Cas9-mediated nicking of the genomic DNA means that a short stretch of ssDNA is exposed to the attached deaminase that can convert the selected bases within their target window ^[117][144]. Many improvements have been conducted since the first report on cytosine base editors (BE), and these have yielded novel base editors that reduce unwanted byproducts, improve the targeting scope and allow the editing of different bases ^[118][145]. Currently, four possible transition mutations can be installed: C→T, A→G, T→C and G→A.

The biggest concern with gene editing is the complexity involved in removing multiple genes (multiplexing) while keeping the safety concerns to a minimum. The researchers developed two non-gene-edited approaches: (i) The first is based on a TCR inhibitory molecule (TIM) that, upon incorporation with the T-cell DNA, competes with TCR elements rendering the TCR unresponsive ^[2][10]. This approach was used together with an NKG2D-based CAR and assessed in metastatic colorectal cancer ^[119][154]. (ii) The second uses an miRNA scaffold targeting CD3ζ, which has led to a complete abolishment of TCR from the cells ^[120][155]. This approach was assessed in a phase I clinical trial using a BCMA-targeting CAR-T in a relapse/refractory multiple myeloma patient cohort.

Another approach includes intracellular retention of TCR/HLA-I to prevent GvHD/HvG. There are multiple methods to retain components in the endoplasmic reticulum (ER), including using a peptide (such as KDEL) that is associated with the ER retention domain. Then, by combining said peptide with an scFv targeting the TCR, for example, all TCRs will be retained in the ER ^[121][156].

While the argument for the removal of the TCR is clear, it is unclear which factors govern cellular persistence in an allogeneic setting. While the usual suspects (HLA-I/II) naturally play a role, other proteins are possibly involved in HvG. Furthermore, other cellular processes such as metabolic regulation may affect cellular persistence in an allogeneic setting. Current results suggest that additional modifications are needed to achieve success in an allogeneic setting. In this regard, the ability to multiplex multiple targets simultaneously becomes a key factor. While this is complex in gene-editing approaches, t is relatively simple in a non-gene-edited approach. Multiple groups have combined either miRNA- or siRNA-like sequences in an effort to inhibit multiple target-sequences together either through a natural scaffold or a synthetic one ^{[122][123][124][125]}[157,158,159,160]. The researchers have recently developed a microRNA (miRNA)-based multiplex shRNA platform, obtained by combining highly efficient miRNA scaffolds into a chimeric cluster ^[126][161]. The researchers were able to deliver up to four shRNA-like sequences (in a plug-and-play manner) in a single vector containing the CAR and four different shRNA-like sequences targeting CD3ζ (GvHD), B2M (HLA-I/HvG) and additional combinations of either CIITA (HLA-II/HvG), CD95 (Fas receptor/inhibit apoptosis), LAG-3 (Immune-checkpoint inhibitor) and/or CD28 (co-stimulation, reduction/persistence). Interestingly, the researchers discovered that the modulation of genes rather than gene KO is essential for certain targets (such as B2M, where a clear balance exists between removal of the HLA-I and recognition by NK-cells and the minimal expression needed to avoid NK-cell lysis and/or T-cell-mediated activation), making the method a good and easy-to-use tool for certain targets.