Hemophilia A is an X-linked bleeding disorder caused by mutations in the FVIII (F8) gene. This gene encodes a 250 kDa protein, FVIII, which is a critical component of the blood coagulation cascade. Severe hemophilia A results from major deletions or inversions in the F8 gene, such that these individual have less than 1% FVIII activity; mild hemophilia can occur with missense mutations, for example, that also lead to significantly reduced clotting efficacy. These disorders can be treated with recombinant or plasma-derived FVIII replacement therapy, either prophylactically or on demand. Unfortunately, a large subset of those receiving replacement FVIII develop an antidrug antibody response because they never developed tolerance to this human protein (unlike the dog in the nighttime!) In the hematology community, these antibodies are referred to as “inhibitors” because they can inhibit or neutralize the therapeutic function of FVIII, rendering this life-saving treatment ineffective. Inhibitor formation requires CD4⁺ T cell help as evidenced originally in HIV-infected patients with inhibitors whose titers dropped when their T-cell levels diminished, but whose antibodies returned upon multi-drug therapy (32, 33). Further studies in a murine model (FVIII knockouts) verified this T-cell dependence (34, 35). Most of the inhibitory antibodies are directed at the A2 and C2 domains of the FVIII protein, which are critical for binding to partners in the cascade.

For several decades, the standard treatment in patients that develop inhibitors is repeated, high-dose FVIII therapy to reduce or eliminate titers, a process referred to clinically as “immune tolerance induction.” This is an expensive process and does not work for all inhibitor cases, being successful primarily in patients with low-titered antibodies. Thus, we have targeted the A2 and C2 domains of the FVIII protein in our approaches for inducing tolerance to FVIII (13, 22). This would be important to achieve in inhibitor positive patients or to prevent inhibitor responses, in the first place, which is of clinical importance.

Inspired by the success of engineered cytotoxic CAR T cells in blood cancers (55), our goal was to apply this approach to directly create large numbers of specific Tregs with engineered receptors. As noted above, based on our experience with retroviral transduction of Fc fusions into activated B cells, we had established a role for Tregs in the tolerance so induced. The buffy coat fractions in all of the experiments to be described below were from peripheral blood mononuclear cells (PBMC) from healthy normal adult donors from the American Red Cross or the NIH Blood Bank. CD4 fractions were isolated by magnetic cell enrichment, then labeled, and sorted based upon the following cell surface markers: naïve CD4 effector T cells were CD4⁺, CD25⁻ CD127⁺, and CD45RA⁺ and Tregs were CD4⁺, CD25^high, CD127^low (and the latter were Foxp3 and Helios positive, reflecting their status as “natural” Tregs).

In collaboration with Dr. Kate Pratt, who had obtained multiple clones of FVIII-specific T effectors from patients with hemophilia A, we determined the TCR variable (V) region genes from two of these clones, termed 17195 and 171911. In the first iteration for specific CD4 effectors and Tregs, retroviral vectors were engineered to express the 17195 or 171911 TCR variable regions in polyclonal T cells activated initially with anti-human CD3. The transduced T cells were expanded as described by Kim et al. (30) with irradiated PBMC’s as APC. Notably, Tregs were expanded but their cultures also contained random oligonucleotides (ODNs), which Kim et al. (56) had shown serve to maintain Treg properties (Foxp3 and Helios). Figure 1 illustrates the principle.

With this approach, we obtained large numbers of expanded FVIII-specific T cells expressing the 17195 or 171911 TCRs that we demonstrated were highly reactive to the FVIII peptides, albeit with different affinities based on the affinity of the initial clones (57). The transduced T effectors proliferated and produced cytokines in response to the FVIII peptide (pC2, 2191–2210) on appropriate DR1 APCs just as effectively as anti-CD3 stimulation of the donors; moreover, specific antigen led to an expansion of the cells expressing the TCR as evidenced by tetramer binding (30). Transduced and expanded Tregs also responded to peptide and displayed increased levels of Foxp3, Helios, GARP, and LAP, typical of activated Tregs, but did not produce significant levels of IL-2 and interferon gamma (IFNγ). Thus, these cells looked like and smelled like human Tregs. We next tested whether they could suppress a FVIII-specific response and found that proliferation of FVIII-specific effector T cells was suppressed even when the effector cells were cultured at an 8:1 ratio to Tregs (30).

As noted above, the antibody response to FVIII in hemophilia A patients is a major hindrance to effective therapy for bleeding. Therefore, we have tested the effect of engineered FVIII-specific human Tregs on an in vitro recall antibody response to FVIII in humanized (DR1) hemophilic knockout mice, using the approach of Hausl et al. (58). Despite being a xenogeneic system, the engineered Tregs were able to suppress the recall antibody response to FVIII (30). Interestingly, although the engineered TCR recognizes a single peptide in a large protein, the antibody response to other major epitopes of FVIII was also suppressed. This indicates that bystander suppression of other T (and B) cells had occurred in vitro. Subsequently, we demonstrated that this could also occur in vivo so it was not due to a culture artifact (31). Thus, we have engineered specificity into expanding human Tregs and shown that they can suppress the antibody response to FVIII effectively.

While these TCR-transduced Tregs were highly effective, they are MHC class II restricted, thus limiting their eventual utility to patients sharing the same MHC globally. Therefore, in collaboration with Anja Naumann Schmidt and Christoph Königs in Frankfurt, we developed a second approach to engineer specificity, namely a scFv, as shown in Figure 2. Dr. Schmidt used phage display to obtain a number of single chain antibodies that reacted with different domains of FVIII (59, 60). One of these, called ANS8, recognized the A2 domain of FVIII. This scFv was incorporated into our retroviral vector and used to transduce both CD4 effectors and Tregs. These scFv transduced cells recognized free FVIII but responded to membrane or plate bound FVIII more effectively (31), presumably reflecting the exposure of the A2 domain under these conditions.

ANS8 CAR human Tregs were generated and tested under the same conditions as the 17195 TCR Tregs. These Tregs also suppress the proliferation of FVIII-specific T effector cells, but most importantly suppressed the antibody response to FVIII both in vitro and in vivo (31). Notably, both the ANS8 CAR-transduced Tregs and 17195 (TCR)-transduced Tregs were effective in these assays at effector: target ratios with effector cells in excess (31). Suppression of the antibody response by these human Tregs in vivo lasted up to 8 weeks. When these mice were boosted with FVIII at 8 weeks post immunization, suppression was lost presumably because the human cells were rejected by the immunocompetent murine hosts. Nevertheless, these data demonstrate that both CAR- and TCR-transduced specific Tregs that recognize different B-cell and T-cell domains of FVIII can be suppressive against multiple epitopes of this large immunogenic protein. Despite this bystander effect, the response to an unrelated antigen (TNP-sheep RBC) was not affected. Thus, suppression in this model is specific.

We recently applied the principle of engineered cytotoxic CAR T cells to directly target FVIII-specific B cells. In lieu of a chimeric antibody, we engineered immunodominant B-cell domains of FVIII into both expanded cytotoxic CD8 and regulatory CD4 T cells (Figure 3). The principle hypothesis was that FVIII-specific B cells possess IgM and IgD receptors that recognize FVIII conformational epitopes. When they would encounter engineered cytotoxic T cells, for example, they would bind these epitopes to form a synapse and would receive a putative negative signal from these cytotoxic cells. This was recently demonstrated by Ellebrecht et al. (61), who used engineered cytotoxic T cells expressing a major skin target (desmoglein 3) in pemphigus vulgaris, a devastating skin disease. They showed that human cytotoxic T cells expressing desmoglein 3 could kill B-cell hybridomas specific for desmoglein. To apply this for hemophilia, we engineered immunodominant C2 or A2 domains (that are the major targets of inhibitory antibodies to FVIII into both human and mouse cytotoxic cells). These BAR expressing cytotoxic T cells were able to kill C2- and A2-specific hybridomas in vitro and in vivo. Moreover, their specificity for FVIII-specific B cells was formally demonstrated in two additional assays: elimination of naïve B cells stimulated with polyclonal B-cell activator, LPS, and inhibition of the antibody response to FVIII in vivo. Because they are domain-specific and did not display a bystander effect, both C2- and A2-BARs were needed to eliminate the response to full-length FVIII (62).

What about BAR Tregs? Theoretically, BAR-expressing Tregs could also interact with specific B cells, but we did not know whether they could directly or indirectly inhibit the B-cell response to FVIII. These data demonstrated that injections of human BAR Tregs into hemophilic mice did indeed prophylactically prevent the antibody response to FVIII (63). To examine the mechanism of this inhibition, we purified B and T cells from BAR tolerized and control mice and then performed classic mixing experiments. These results suggested that B cells may be directly targeted by BAR Tregs since they could not be “helped” by control non-tolerant T cells (64). Whether this is due to anergy or cytotoxicity of targeted B cells is under investigation.