Background
Adoptive cell therapies represent a paradigm shift from conventional drug treatments and offer the potential of treating diseases with greater precision and less toxicity. In particular, T-cell therapies have shown many impressive results. For example, T-cells expressing chimeric antigen receptors (CARs) targeting CD19 have been very effective against B-cell malignancies and virus specific T-cells (VSTs) have shown great promise in treating viral infections in hematopoietic stem cell transplant (HSCT) recipients and patients with Epstein–Barr virus (EBV) associated lymphomas [
1‐
4]. To avoid graft rejection and/or graft-vs-host disease (GVHD), most T-cell therapies are derived directly from the patient or from the stem cell donor in HSCT recipients. This highly personalized strategy has limited wider application due to the time and cost of generating a customized product as well as difficulties in generating therapeutic cell products from some patients with genetic disorders or cancer [
5]. If graft rejection and GVHD could be overcome, highly characterized “off-the-shelf” cell products could be derived from healthy donors and dramatically improve the feasibility and availability of cell therapies [
6].
The main cellular mediators of both graft rejection and GVHD are alloreactive T-cells that recognize non-self human leukocyte antigen (HLA) molecules on allogeneic cells. To protect allogeneic T-cells from rejection, several groups have eliminated HLA class I expression by knocking out either individual HLA molecules or beta-2 microglobulin (B2M), a universal component of all HLA class I molecules [
7,
8]. Although this strategy minimizes T-cell mediated rejection, loss of HLA antigens increases susceptibility to killing by natural killer (NK) cells [
9]. The direct elimination of alloreactive T-cells is an alternative approach to graft protection. In 1980, Miller introduced the concept of a “veto cell” that can specifically eliminate a cognate alloreactive T-cell [
10,
11]. Although several different cell types, including dendritic cells, NK cells, and T-cells, can demonstrate veto activity, cytotoxic CD8 T-cells are thought to exhibit the strongest effect [
12,
13]. Reisner et al. have shown that in murine T-cells this veto effect is independent of T-cell receptor (TCR) ligation and instead is mediated by a Fas–FasL dependent mechanism in which FasL expressed on veto cells binds to Fas on alloreactive T-cells, inducing apoptotic cell death [
14‐
16].
While T-cells may possess an inherent ability to veto alloreactive T-cells without TCR ligation, engagement of the veto cell TCR could initiate a more potent cytolytic effect since it would recruit the more rapidly acting perforin/granzyme pathway [
17,
18]. To this end, Margalit et al. constructed a chimera of B2M and the cytolytic domain of the TCR zeta chain [
19]. Theoretically, this chimeric B2M/CD3-zeta protein can complex with any HLA class I molecule, and when expressed on an allogeneic T-cell could mediate killing of any engaged alloreactive T-cell. The initial study showed that a murine T-cell hybridoma expressing the B2M/CD3-zeta protein could produce IL-2 when bound by an antibody specific to the murine major histocompatibility complex (MHC) on the T-cell hybridoma, but did not demonstrate veto mediated killing [
19]. Subsequent studies focused on autoimmune disease and showed that murine T-cells expressing the B2M/CD3-zeta protein and presenting insulin peptides could reduce progression of diabetes in mice by targeting insulin-specific diabetogenic T-cells [
20,
21]. Thus far, however, no studies have evaluated the ability of the B2M/CD3-zeta protein to eliminate alloreactive T-cells or tested this approach in human T-cells.
Preventing allo-rejection overcomes one barrier to off-the-shelf therapy, however, since allogeneic T-cell products may contain alloreactive T-cells that could attack recipient tissues, avoiding GVHD is also essential. To this end, several groups have knocked out the endogenous TCR in T-cells [
7,
22,
23], however, complete depletion of TCR positive T-cells may not be feasible and patients infused with less than 1% residual TCR positive T-cells can still develop GVHD [
24]. VSTs, by contrast, rarely produce GVHD when infused into allogeneic recipients [
25]. Furthermore, allogeneic VSTs that have been banked for use as off-the-shelf therapy have proved safe and effective in treating viral infections in HSCT recipients [
26,
27]. Therefore we have used VSTs in our study to avoid the problem of GVHD.
To determine if human T-cells can be engineered to eliminate human alloreactive T-cells, we generated a human version of the B2M/CD3-zeta protein termed the Chimeric HLA Accessory Receptor (CHAR). We found the CHAR could complex with endogenous human HLA class I molecules and carry them to the cell surface. When expressed in VSTs, the CHAR could eliminate alloreactive T-cells in co-cultures with allogeneic peripheral blood mononuclear cells (PBMC) without eliminating pathogen-specific T-cells, and in contrast to unmodified VSTs, were protected from allo-specific elimination. By eliminating alloreactive T-cells, CHAR expressing VSTs could prevent the rejection of allogeneic cell therapy products increasing the persistence of off-the-shelf cell therapies. This strategy could have widespread impact not only on the use of allogeneic cells for the treatment of viral infections and cancer but also on other fields such as regenerative medicine.
Methods
Generation of retroviral constructs
The codon optimized CHAR construct was synthesized by GeneArt (Invitrogen, Carlsbad, CA) and cloned into the gamma retroviral vector SFG [
28] using In-Fusion cloning (Takara Bio USA, Mountain View, CA). The CHAR sequence consisted of the entire human B2M sequence including the signal peptide, a portion of human HLA-A2 (Uniprot: AA 296-308) to bridge the physical distance between B2M and the cell membrane surface, the transmembrane domain of human CD8 alpha (AA 183-210), and the signaling endodomain of human CD3 zeta chain (AA 52-164). To allow expression of two genes from a single mRNA, a 2A peptide sequence derived from porcine teschovirus-1 with a GSG linker was placed downstream of the CHAR [
29]. Downstream of the 2A peptide is the Q8 marker gene that contains a small compact epitope of human CD34 that is recognized by the clinical grade monoclonal antibody QBend10 [
30]. This epitope is attached to a human CD8 alpha (CD8a) stalk and transmembrane region (AA 134-222). To limit homologous recombination between the CD8a regions in both Q8 and CHAR constructs, the CD8a region in Q8 was substituted for wildtype CD8a while the CD8a region in the CHAR was codon optimized.
To generate an inducible CHAR we used the Tet-One system from Takara Bio USA (Mountain View, CA) that expresses the two components of the system, the transactivator protein (Tet-On 3G) and the tet-responsive promoter (TRE3GS) in a single plasmid [
31]. We inserted our gene product that includes the CHAR, 2A, and Q8 into the Tet-One plasmid downstream of the TRE3GS (Additional file
1: Fig. S1A). As seen in Additional file
1: Fig. S1B, expression of the CHAR from the Tet-One construct resulted in low transduction efficiency in VSTs, consistent with previous reports of low Tet-One transduction in primary T lymphocytes [
32]. To improve the transduction efficiency, we made several modified Tet-One constructs (data not shown) and found that inversion of the entire coding sequence between the 5′ LTR and 3′ LTRs (shown in Fig.
2c) resulted in higher transduction (Fig.
2d) compared to the original Tet-One construct (Additional file
1: Fig. S1B). This construct was used for the rest of the study and will be referred to as the inducible CHAR (iCHAR).
Cell lines
The Daudi cell line was obtained from American Type Culture Collection (ATCC) (Manassas, VA). Daudi cells were maintained in RPMI 1640 media (GE Healthcare Life Sciences, Pittsburgh, PA) supplemented with 10% FBS (GE Healthcare Life Sciences) and 1% GlutaMAX (Thermo Fisher Scientific, Waltham, MA). Cell were grown at 37o C in a humidified atmosphere containing 5% carbon dioxide.
Generation of T-cells
Peripheral blood mononuclear cells (PBMCs) were isolated from healthy donors after obtaining informed consent under the Institutional Review Board of Baylor College of Medicine and in accordance with the guidelines established by the Declaration of Helsinki. Activated T-cells (ATCs) were generated by plating PBMCs on 24-well plates coated with 1 mg/ml anti-CD3 (OKT3) (ATCC, Manassas, VA) and 1 mg/ml anti-CD28 (BD Biosciences, San Jose, CA). ATCs were maintained in medium with IL-2 (NIH, Bethesda, VA) at 40 IU/ml. Virus specific T-cells (VSTs) were generated from PBMC devoid of CD4 T-cells and NK cells by magnetic column depletion using CD4 and CD56 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). Pepmix peptide pools to pp65 (JPT Peptide Technologies, Berlin, Germany) were added to depleted PBMCs (10 ng per 1 × 106 PBMCs) to generate CMV-specific T-cells (CMVSTs). CMVSTs were grown in IL-7 at 10 ng/ml and IL-15 at 10 ng/ml (PeproTech, Rocky Hill, NJ). ATCs and CMVSTs were maintained in medium consisting of a 1:1 mix of RPMI 1640 (GE Healthcare Life Sciences) and Click’s Media (Irvine Scientific, Santa Ana, CA) supplemented with 10% FBS (GE Healthcare Life Sciences) and 1% Glutamax (Thermo Fisher Scientific). Every 2–3 days, T-cells were fed with fresh media containing the respective cytokines. For experiments in which the inducible CHAR was used, certified Tet-Free FBS (Takara Bio USA) was used in place of conventional FBS. Doxycycline (Sigma-Aldrich, St. Louis, MO) was used at 100 ng/ml to induce express of the CHAR.
Retrovirus production and T-cell transduction
Retroviral supernatants were produced as previously described [
33] and plated on non-tissue culture treated 24-well plates pre-coated with RetroNectin (Takara Bio USA). After centrifugation at 2000×
g for 90 min, retroviral supernatant was removed and CMVSTs from day 4–5 were plated at 0.5 × 10
6/well. On day 9, CMVSTs were restimulated using a combination of pepmix-pulsed ATCs and a HLA negative costimulatory cell line, K562CS (gift from Carl June), as previously described [
34].
Flow cytometry
The following fluorochrome-conjugated monoclonal antibodies were used in this study: CD3, CD4, CD8, CD19, and IFNγ from Beckman Coulter (Indianapolis, IN); CD71, HLA-A2, and HLA-A, B, C from BioLegend (San Diego, CA); CD95 (Fas) and CD107a from BD Biosciences (San Jose, CA); and CD34 (QBEnd-10) from Abnova (Taipei, Taiwan). Cell viability was assessed using 7-amino actinomycin D (7-AAD) (BD Biosciences) staining. We used the Gallios Flow Cytometer (Beckman Coulter) to acquire flow cytometric data and Kaluza Analysis Software (Beckman Coulter) to analyze data and for graphical representation.
Co-culture of CMVSTs and allogeneic PBMC
CMVSTs were co-cultured with allogeneic CD56-depleted PBMCs at a 1:2 ratio. Discrimination between CMVSTs and allogeneic PBMCs was determined by HLA-A2 expression. Media contained IL-2 at 20 IU/ml and doxycycline at 100 ng/ml. On days 0, 4 and 8 co-cultures were harvested, stained with antibodies and analyzed by flow cytometry. Countbright Beads (Life Technologies, Carlsbad, CA) were used to assess cell numbers.
CellTrace Violet proliferation assay
On day 5 after the primary (1st) mixed lymphocyte reaction (MLR) of CMVSTs and allogeneic PBMCs was initiated, all cells within the 1st MLR were stained with CellTrace Violet (Thermo Fisher Scientific, Waltham, MA) at 2.5 μM and then mixed at a 1:1 ratio with PBMCs derived either from the CMVST donor or allogeneic PBMC donor to generate a secondary (2nd) MLR. The 2nd MLR was then harvested after 4–5 days and analyzed by flow cytometry.
Co-culture of CMVSTs with autologous PBMCs or ATCs
After isolating PBMCs to generate CMVSTs, a portion was cryopreserved for subsequent co-cultures. Thawed autologous PBMCs were rested overnight, and then either these non-activated PBMCs or ATCs, generated by plating the PBMCs on anti-CD3 (ATCC) and anti-CD28 (BD Biosciences) coated plates for 4 h, were labeled with CellTrace Violet (3 μM) and mixed with CMVSTs at a 1:1 ratio. After 4 days in media containing IL-2 at 40 IU/ml, co-cultures were stained with antibodies and analyzed by flow cytometry.
Knockout of Fas in PBMC and co-culture with CMVST
To knockout Fas in PBMC we used a previously optimized protocol developed by Seki and Rutz [
35]. Briefly, we combined 2ul TrueCut Cas9 Protein v2 (Thermo Fisher Scientific, Waltham, MA) with 1 μl each of three single guide RNAs (sgRNA) synthetized by in vitro transcription. CRISPR sgRNA for Fas were designed using the online tool CRISPRscan (
https://www.crisprscan.org) and recognized the following target site sequences: GGATTGCTCAACAACCATGCTGG, GATTGCTCAACAACCATGCTGGG, GTGACTGACATCAACTCCAAGGG. The Cas9 and sgRNA complexes were then combined with 2–4 × 10
6 CD56-depleted PBMC resuspended in 20 μl of buffer solution from the P2 Primary Cell 4D-Nucleofector X Kit S (Lonza, Basel, Switzerland) and put into Nucleofection cuvette strips. Cells were electroporated with the 4D Nucleofector system (4D-Nucleofector Core Unit, AAF-1002B; 4D-Nucleofector X Unit, AAF-1002X) from Lonza (Basel, Switzerland) using the pulse code EH100. After PBMCs were rested overnight at 37 °C, they were mixed at a 10:1 ratio with CMVSTs in media containing IL-2 at 20 IU/ml and doxycycline at 100 ng/ml. Co-cultures were harvested on day 8, stained with antibodies and analyzed by flow cytometry.
CD107a degranulation assay
CMVSTs, pretreated with doxycycline 1 day prior to induce CHAR expression, were co-cultured with autologous sTCR-ATCs on day 8. Prior to the co-culture, CMVSTs were labeled with CellTrace Violet (0.05 μM) and either pulsed with LML peptide (10 ng per 1 × 106 cells) or a DMSO vehicle control for 1 h. After co-cultures were incubated for 4–5 h in media containing GolgiStop (BD Biosciences) at 1 μl/ml and CD107a antibody (BD Biosciences) at 10 μl/ml, they were stained with additional antibodies and analyzed by flow cytometry.
Intracellular cytokine staining
CD56-depleted PBMCs were co-cultured with allogeneic iCHAR CMVSTs at a 2:1 ratio or cultured alone for 5 days in media containing IL-2 at 20 IU/ml and doxycycline at 100 ng/ml. Afterward, cells were washed, rested for several hours, and then incubated with pepmixes (500 ng/ml) for adenovirus (Hexon and Penton), CMV (IE1) or EBV (EBNA1, BZLF1, LMP1, and LMP2) overnight in media containing GolgiPlug (BD Biosciences) at 1 μl/ml. The survivin pepmix was used as an irrelevant control. Cells were then labeled with cell surface antibodies, fixed, permeabilized, stained for intracellular IFNγ, and analyzed by flow cytometry. Discrimination between PBMCs and allogeneic CMVSTs was determined by HLA-A2 expression.
Statistical analysis
Data are presented as mean ± standard error of the mean (SEM) and statistical analysis was performed using GraphPad Prism 5 software (GraphPad Software, Inc.). Paired two-tailed Student t-tests were used for comparisons between two groups.
Discussion
In this study, we show that human VSTs engineered to express a humanized iCHAR can limit the activation and expansion of alloreactive T-cells via the perforin/granzyme pathway and notably do not affect pathogen-specific memory T-cells, such as VSTs, which should spare recipients from general immune dysfunction. To the best of our knowledge, this is the first study to demonstrate that primary human T-cells can be engineered to enhance their veto activity, potentially allowing iCHAR VSTs to be used as a platform for future off-the-shelf allogeneic T-cell therapies.
Unexpectedly, the human CHAR was not exclusively specific for alloreactive HLA class I-restricted CD8 T-cells but also targeted alloreactive CD4 T cells and irrelevant activated T-cells, a characteristic that likely contributed to the observed fratricide of CHAR VSTs. The mechanisms underlying these findings still remain unclear. We excluded a role for the Fas-FasL pathway, since iCHAR CMVSTs were still able to eliminate activated CD4 alloreactive T-cells lacking Fas. The exact mechanism responsible for these effects is an interesting avenue for future studies, and may result from phenotypic changes that occur in response to T-cell activation [
47,
48]. For example, upregulation of killer cell immunoglobulin-like receptors (KIRs) that bind HLA class I molecules [
49] may render activated T-cells susceptible to killing by the CHAR. We are currently investigating this and other possible mechanisms that can explain the killing of activated T-cells as well as CHAR-mediated fratricide.
This unanticipated targeting of activated T-cells by CHAR VSTs was fortuitous, since it resulted in killing of alloreactive CD4 T-cells that can also mediate rejection of allografts [
50,
51]. However, targeting all activated T-cells could be detrimental if left unchecked, since pathogen-specific T-cell responses could also be impaired. Fortunately, the iCHAR is drug inducible, so that withdrawal of drug after the elimination of alloreactive T-cells should downregulate the iCHAR in VSTs allowing them to respond normally to viral infections and provide undisturbed protective immunity.
A barrier to the clinical use of inducible expression systems in cell therapy is their frequent reliance on xenogeneic and hence immunogenic components, such as our bacterial derived Tet transactivator protein, which could induce immune mediated elimination of engineered therapeutic cells [
32,
52]. As our iCHAR VSTs have veto ability, any T-cell that recognized Tet transactivator-derived epitopes would be eliminated. Thus, combining the veto ability of our CHAR VSTs with an inducible expression system improves the safety of CHAR VSTs while negating the immunogenicity of the inducible system. This ability to eliminate T-cells that recognize immunogenic proteins could have widespread implications in the field of synthetic biology in which cells can be endowed with sophisticated capabilities but often use components derived from viruses and bacteria to achieve high specificity and potency [
53‐
55]. Such a strategy to tolerize patients to immunogenic foreign proteins could pave the way for clinical translation of synthetic biology.
Reisner and colleagues have shown that unmodified murine T-cells can veto/eliminate alloreactive T-cells leading to allografts acceptance in murine models [
56,
57]. To our knowledge however, effective translation of this work to humans has yet to be demonstrated. In our hands, unmodified activated human T-cells did not prevent the expansion of human alloreactive T-cells and only after CHAR transduction did they develop significant veto activity. These discrepancies in veto efficacy of unmodified T-cells may be due to differences in the veto cell types examined, the assays used to assess veto activity, or to inherent disparities in veto mechanisms between mouse and human cells. Given that the immune systems of mice and humans can differ in significant ways [
58,
59], characterizing alloreactivity and tolerance mechanisms in human cells may be more appropriate for the evaluation of therapies that can translate effectively to the clinic.
Elimination of alloreactive T-cells by iCHAR VSTs could tolerize recipients to allow protection not only of iCHAR VSTs but also other cell therapy products that are matched to the iCHAR VST donor. As direct effectors, rejection resistant allogeneic iCHAR VSTs could be used to treat viral infections [
26,
60] and malignancies [
61]. Alternatively, iCHAR VSTs transduced with CARs [
62,
63] could be used as off-the-shelf products for a range of malignancies. However, since iCHAR VSTs may not optimally perform the function of a veto cell and effector cell at the same time, it may be more effective to use iCHAR VSTs to protect subsequently infused therapeutic T-cell products derived from the same donor.
Protecting allogeneic cell therapy products beyond T-cells could significantly impact many other fields such as regenerative medicine in which immune rejection could be a critical barrier to long-term therapy [
64,
65]. Similarly, in transplantation biology, eliminating alloreactive T-cells to induce lasting tolerance could prevent rejection of both solid organ and stem cell transplants without the use of immunosuppressants that are associated with long-term toxicities [
12,
66]. Since alloreactive T-cells also mediate GVHD [
67], iCHAR VSTs derived from the patient could be used to prevent or treat GVHD by eliminating anti-host alloreactive T-cells present in the graft. In addition, patient-derived iCHAR VSTs could potentially prevent or treat autoimmune diseases such as type 1 diabetes if they could be further modified to eliminate self-reactive T-cells like diabetogenic T-cells, as demonstrated in mice by Wong and colleagues [
20,
41].
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.