Induction of tolerance to factor VIII inhibitors by gene therapy with immunodominant A2 and C2 domains presented by B cells as Ig fusion proteins

Hemophilia A is a bleeding disorder caused by a decrease or dysfunction of blood coagulation factor VIII (fVIII). Bleeding episodes can be prevented or treated by replacement therapy using plasma-derived or recombinant fVIII. A major complication in replacement therapy is that patients can develop an inhibitory antibody response to transfused fVIII.¹  In addition to high-dose tolerance protocols (which are extremely expensive), a variety of methods to block inhibitor formation have been developed, albeit with variable success in preclinical animal models. These include using peptide decoys mimicking the anti-fVIII antibody,²  bypassing immune recognition with human/porcine fVIII hybrids,³  neutralizing fVIII-reactive CD4 T cells with anticlonotypic antibodies,⁴  attempting to induce tolerance to fVIII with the use of universal CD4 epitopes,⁴  and blocking costimulation with CTLA-4-Ig or anti-CD40L.^5-7  Nonetheless, novel approaches toward induction of specific tolerance to fVIII remain a desirable goal to treat patients with hemophilia A with inhibitors.

Our laboratory has used a gene therapy approach for tolerance in which we have engineered retroviral constructs to drive expression in B cells of different antigens in frame at the N-terminus of a murine immunoglobulin G1 (IgG1) heavy chain.^8,9  These studies were based on the well-established tolerogenicity of IgG carriers and tolerogenic antigen presentation by B cells.^10-12  We have shown that recipients of B-cell blasts, transduced with an Ig fusion of a variety of model antigens or autoimmune targets constructs, are tolerant to the protein epitopes of the expressed genes and that this therapy can lead to striking modulation of clinical disease.^8,13-16  Importantly, tolerance is more effective and of longer duration in the presence of the IgG backbone⁹  (T.C.L., Yan Su, and D.W.S., manuscript in preparation).

It is known that most inhibitory antibodies are reactive with conformational epitopes on the exposed surfaces of C2 and A2 domains of fVIII,^1,3,4,17,18  and that immunodominant T-cell epitopes can be found in the C2 region.¹⁷  In this study, we inserted the coding sequences for residues S2173-Y2332 (C2 domain) and S373-R740 (A2 domain) in frame with the IgG heavy chain backbone in a retroviral vector to induce tolerance in hemophilic mice. Herein, we show that effective suppression of immune responsiveness to fVIII can be achieved when lipopolysaccharide (LPS)-activated B-cell blasts are transduced with a fusion IgG containing the C2 or A2 domains and injected into naive hemophilic mice. Importantly, reduction of ongoing responsiveness and inhibitory antibody titers was accomplished by this treatment in fVIII primed recipients with significant anti-fVIII titers.

Factor VIII-deficient mice carrying a stop mutation in exon 16 of the fVIII gene (E16 mice)¹⁹  were used as a model for hemophilia A. These mice have been backcrossed for at least 8 generations onto a C57BL/6 background.⁵  E16 hemophilic mice were used in this study at 8 to 20 weeks of age. The genotypes of hemophilic mice were confirmed by polymerase chain reaction (PCR) analysis of genomic DNA extracted from tail segments, as described previously.⁵  All animals were housed in pathogen-free microisolator cages at the animal facilities of the Holland Laboratory, operated by the University of Maryland. Blood samples were obtained by orbital plexus bleeding, and venous blood samples were anticoagulated (9:1) with 0.105 M citrate for plasma separation. All samples were centrifuged immediately at 3600g for 10 minutes at room temperature, divided into aliquots, and frozen at -80°C until analyzed.

Molecular cloning of retroviral vectors was similar to those described previously.^13-15  Briefly, cDNAs encoding the C2 (S2173-Y2332) or A2 (S373-R740) domains of human fVIII were cloned from SIN-CMV/R/U5-FMU3-fVIII DB-SW vector (kindly provided by Dr Ali Ramezani, George Washington University, Washington, DC) using PCR. A mock control cDNA containing an unrelated antigen (SAG, arrestin) was the kind gift of Dr Wei Liang (TolerGenics, Rockville, MD). The targeted sequences were inserted at NotI/XhoI sites into the BSSK-IgG plasmid, which contains a full-length secretory murine IgG₁,⁹  and these IgG cDNA fragments were then subcloned into a bicistronic MSCV-IRES-GFP vector (obtained from Dr Kevin Bunting, Case Western Reserve University, Cleveland, OH) using the SalI and XhoI compatible sites.

To generate packaging cell lines that stably produce retrovirus, 293T cells were cotransfected with MSCV-C2-IgG (or A2-IgG)-IRES-GFP, the helper plasmid pSRα-G, and pEQPAM3e (kindly supplied by Dr Bunting) by standard calcium phosphate precipitation. The supernatant was collected 48 hours after transfection, filtered, and frozen at -80°C. To confirm transfection, the 293T cells were analyzed for green fluorescent protein (GFP) expression. Subsequently, the retroviral supernatants were used to stably transduce the GPE-86 packaging cell line.^8,9,11  The transfected cells were expanded and the GFP-positive cells were selected after fluorescence-activated cell sorting assay. The viral titer of transfected GPE-86 packaging cell lines were tested using NIH3T3 cells and the titer calculated by Poisson analysis. Preparations with a titer of more than 1 × 10⁶ colony-forming units (CFUs)/mL were used for transduction of B cells.

The transduction of splenic B lymphocytes for gene therapy was performed as described earlier.^13-15  Briefly, splenic B cells were stimulated with bacterial LPS at 10 μg/mL for 24 hours, and cocultured with irradiated (2000 rad) viral packaging cells in the presence of 6 μg/mL polybrene and 10 μg/mL LPS for an additional 24 hours. The virally infected B cells were washed and injected into syngeneic recipients at designated intervals before or after challenge, as described in “Immunologic challenge and assay methods.” On the basis of GFP expression level in B cells, the percentages of productively transduced cells were estimated to be about 50% for C2-IgG/MSCV and 25% to 35% for A2-or SAG-IgG/MSCV.

Plasma-derived human fVIII was obtained from Baxter Healthcare (Glendale, CA). Recombinant human C2 V2169-Y2332 (rC2), was generously provided by Dr Kate Pratt (University of Washington, Seattle, WA).^17,18  This was expressed in Pichia pastoris and purified by ammonium sulfate fractionation and anion exchange chromatography. The purified protein was characterized by both ion exchange and gel filtration columns as a single peak. It also was detected as a single 19-kDa band by Western blotting using monoclonal Ab ESH8. C2 protein was dissolved in 50 mM HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid; pH 7.6)/25 mM NaCl buffer and stored at -80°C, and used within 3 months. Recombinant A2 protein was provided by Dr Andrei Sarafanov (Department of Biochemistry, University of Maryland School of Medicine, Baltimore, MD).²⁰  Briefly, the sequence S373-R740 of human fVIII A2 was expressed in Bac-to Bac baculovirus expression system (Invitrogen, Carlsbad, CA) and purified by a Ni-nitrilotriacetic acid-HisSorb column via His-tags. The expressed A2 protein was detected as a single 43-kDa band by Western blotting using monoclonal Ab 413. The final A2 protein was dissolved in phosphate-buffered saline (PBS; pH 7.2)/1% bovine serum albumin and stored at -80°C and used within 3 months. Absorbance assay at 280 nm was used for the measurement of A2 or C2 concentrations, using extinction coefficients of 1.8 for C2 and 1.5 for A2 based on a standard algorithm (http://us.expasy.org/tools/protparam.html).

In our pretreatment protocol, E16 mice (5 mice per group) were injected intraperitoneally with 10 × 10⁶ to 20 × 10⁶ transduced B cells. One week later, they were immunized with fVIII using a modification of the protocol of Qian et al⁵  in which higher doses of fVIII were used to achieve more rapid and consistent immune responses. Thus, these E16 knockout mice were initially injected intravenously with fVIII (4 μg) and then boosted 3 times at 5-day intervals with 2 μg fVIII intraperitoneally to avoid anaphylactic shock. At 5 to 7 days after the last fVIII injection, mice were bled from the orbital venous plexus and serum evaluated for fVIII and C2 antibody levels using an enzyme-linked immunosorbent assay (ELISA), and plasma was analyzed for fVIII inhibitor titer (as described later in this section). Spleen cells from individual mice were collected for T-cell assays as described later in this section. Due to death of some experimental animals, the final group size was 3 to 5 mice in a given experiment.

To mimic the development of inhibitors in patients with hemophilia A, groups of 5 E16 knockout mice were primed via 4 weekly injections of therapeutic doses of fVIII, starting as described earlier in this section with an intravenous injection and followed by intraperitoneal boosts. One week after the last fVIII injection, mice were bled as described earlier in this section and serum evaluated for fVIII and C2 antibody levels using ELISA. Two weeks later, recipients were injected (10 × 10⁶-20 × 10⁶ cells/mouse, intraperitoneally) with C2-Ig plus A2-Ig-transduced B cells. A booster immunization with fVIII was given 7 days after gene therapy, and recipients bled as described earlier in this section and individual spleens harvested one week later. To test for the persistence of tolerance, this protocol was repeated but the interval after B-cell treatment extended for an additional 2 months, with a second booster injection of fVIII.

Antibody titers were measured by ELISA, as described by Qian et al,^5,6  using 5 μg/mL of rC2 or 1 μg/mL fVIII-coated plates. The concentrations of anti-fVIII were estimated from the standard curves obtained using monoclonal mouse IgG anti-A2 antibody (mAb413), obtained from Dr Eugene Saenko, University of Maryland, and anti-C2 antibody (ESH4; American Diagnostica, Stamford, CT).

T-cell proliferation analysis was carried out by standard methods as previously described.^13-15  In brief, splenic T cells from individual mice were cultured in triplicate at 5 × 10⁵/well in 96-well plates with multiple doses of C2, A2, or fVIII in RPMI1640 (Gibco, Grand Island, NY) containing 0.5% murine hemophilia (E16) serum and 2% fetal bovine serum. After 48 hours at 37°C, ³H-thymidine (Perkin-Elmer, Boston, MA) was added and the cultures harvested 16 hours later using a Packard Matrix96 harvester. The mean counts per minute (cpm) of triplicate wells from each mouse was used for statistical calculations and the data expressed as the mean plus or minus the standard error of the mean (SEM) cpm incorporated into insoluble DNA minus background.

The fVIII inhibitor titer was determined by Bethesda assay using a previously published protocol with minor modification.^21,22  In brief, one part of pooled normal human citrated plasma (purchased from DiaPharma, West Chester, OH) was mixed with one part of murine plasma from immunized animals (usually diluted from 5-fold to 40-fold with 0.05 M imidazole buffer). The mixtures were incubated at 37°C for 2 hours. A control mixture consisted of equal parts of normal pooled plasma and imidazole buffer. Factor VIII activity was measured by a chromogenic functional assay (COATEST VIII:C/4; DiaPharma). Normal pooled human plasma containing 1.0 IU fVIII/mL or 100% activity was used to generate a standard curve. Residual fVIII activity in the test mixture was divided by the fVIII activity in the control mixture to determine the percent residual fVIII activity. The inhibitor titer was established from a semilogarithmic plot of residual fVIII activity (logarithimic) and units of inhibitor activity (linear).²²  The dilution of test plasma giving a residual fVIII activity of 50% is said to contain one Bethesda unit fVIII inhibitor activity per milliliter. We also used human fVIII-deficient plasma with varying levels of fVIII inhibitor activity (purchased from George King BioMedical, Overland Park, KS) to validate the standard curve.

To remove natural T-regulatory cells, mice were injected with anti-CD25 monoclonal (1 mg intraperitoneally) or with an equivalent amount of purified normal rat IgG on days -1, +4, and +8 relative to B-cell gene therapy. The success of PC61 treatment was verified by standard flow cytometry of peripheral blood leukocytes on day 9 using control Ig-stained viable cells to set the negative gate. Animals were then immunized as above in the pretreatment protocol.

The significance of the differences for each immunologic experiment was evaluated using the Student t test with a level of P less than .05 being considered significant.

Factor VIII replacement therapy can lead to the formation of inhibitory antibodies in a significant number of patients with hemophilia A. Most of these inhibitors are directed toward the A2 or C2 domain of fVIII, and they neutralize clotting activity by either interfering with the interaction of fVIIIa with factor IXa or blocking the binding of fVIII to phospholipid surfaces.^5,17,23  In this study, we applied a gene therapy approach for tolerance induction^8,9  to block or reverse inhibitor formation in fVIII knockout mice, a model for hemophilia A.^5,19  Thus, the retroviral constructs containing the C2 or A2 domains of human fVIII in frame with a murine IgG₁ (Figure 1) were prepared and used to transduce B cells as tolerogenic antigen-presenting cells. LPS-activated B-cell blasts were transduced with these C2- and/or A2-IgG fusion constructs and injected into fVIII^-/- (E16) mice to induce specific hyporesponsiveness to immunogenic doses of fVIII. As shown in Figure 2A-B, a single injection of C2-IgG or A2-IgG-transduced LPS-activated B-cell blasts resulted in specific T-cell hyporesponsiveness to C2 or A2 protein. While only a modest reduction in the T-cell response to full-length fVIII was affected by tolerance to either domain alone, treatment with a combination of B cells expressing A2 and C2 domains led to statistically significant hyporesponsiveness to fVIII (P < .01; Figure 2C).

Similarly, specific tolerance to the C2 domain in terms of the humoral response was induced with B-cells expressing C2-Ig. Antibody responsiveness full-length fVIII was significantly lower in C2-IgG plus A2-IgG B-cell-treated recipient mice than in mock controls (P < .05, Figure 2D-E). Importantly, the combination of C2-IgG with A2-IgG transduced B cells also afforded striking tolerance in terms of inhibitory antibody titers (P < .001). Indeed, C2-Ig or A2-Ig B cells alone did lead to a significant reduction in inhibitory antibody titers (P < .01, Figure 2F). Together, these results reveal that B-cell presentation of both fVIII domains on an Ig backbone is an effective therapeutic to block the total response to fVIII in naive animals presumably because of the immunodominance of these domains in inhibitory antibodies.

Patients with hemophilia A who have received repeated infusions of fVIII are likely to be primed, regardless of their inhibitor status. To test whether B-cell-delivered retroviral gene therapy would be able to suppress inhibitor formation in fVIII-primed subjects, a clinically important goal, we infused LPS-activated B-cell blasts transduced with C2-IgG and A2-IgG or a mock control construct into E16 recipients that had been immunized multiple times with fVIII. Therefore, to more accurately mimic the state of patients with hemophilia A with inhibitor titers, E16 mice received 4 weekly injections of therapeutic doses of fVIII and anti-fVIII titers determined one week later (day 28). Mice were randomized, then rested for 2 weeks at which time they received B-cell gene therapy with C2-Ig and A2-Ig. One week later, they were boosted with another dose of fVIII. After 7 days, they were euthanized and T-cell proliferation, anti-fVIII ELISA, and inhibitory antibody titers determined.

As shown in Figure 3A-C, injection of C2-IgG plus A2-IgG-transduced B-cell blasts resulted in significant T-cell hyporesponsiveness to fVIII and its domains, compared with mice treated with the mock control. Specific antibody responses to C2 or full-length fVIII were clearly evident on day 28 (Figure 3D-E) and this titer increased upon boosting when measured at day 56. The anti-C2 and anti-fVIII antibody titers in recipients of C2-Ig plus A2-Ig-transduced blasts were reduced significantly compared with the mock control B-cell-injected mice on day 56. Indeed, the ELISA titers were actually lower than on day 28 in this treatment group (P < .05; Figure 3D-E).

Most importantly, the inhibitor titers in terms of Bethesda units (Figure 3F) were reduced more than 10-fold in the C2-Ig plus A2-Ig B-cell-treated group (P < .001). Together, these results indicate that specific tolerance induction to fVIII could be effectively achieved in fVIII-primed recipients using this retroviral gene therapy-based protocol.

To determine the persistence of the tolerance state induced by B-cell gene therapy, we modified this experimental protocol as follows. Hemophilic mice received 4 weekly injections with fVIII and were treated on day 42 with C2-Ig and A2-Ig-transduced B cells. These animals were boosted on day 49 and again on day 102 with fVIII. They were euthanized on day 109 and analyzed for immune responsiveness to fVIII. As shown in Figure 4A-B, humoral hyporesponsiveness to C2 and to fVIII persisted at least to day 109, more than 2 months after a single treatment. Inhibitory antibody titers (Figure 4C) were still more than 90% suppressed.

To gain insight to the mechanism of tolerance induction and a potential role for T-regulatory cells, E16 mice were injected with PC61 monoclonal anti-CD25 or normal rat IgG just prior to and after B-cell gene therapy with C2-Ig plus A2-Ig-transduced B cells. The results in Figure 5A indicate that PC61 treatment led to depletion of CD25⁺ T cells. However, the number of CD25⁺ cells in peripheral blood returned to normal by day 25 (data not shown). These mice were immunized with fVIII on days 10, 15, and 20. On day 25, we measured the T-cell responses (Figure 5B-C) and ELISA titers (Figure 5D-E) to C2 and to fVIII, as well as the inhibitory antibody titers (Figure 5F). Confirming the data in Figure 2, treatment with B cells expressing C2-Ig and A2-Ig led to tolerance to C2 and to fVIII by all parameters in mice given control rat IgG. Treatment of mice with anti-CD25 eliminated the tolerogenic effects of B-cell gene therapy. These data suggest that a CD25⁺ T-regulatory cell is required for the tolerogenic effect of transduced B cells, although the specificity and mode of action of these cells is unknown.

In our laboratory, we have used a gene therapy approach to induce tolerance via retroviral expression in B cells of different antigens in frame at the N-terminus of a murine IgG1 heavy chain. This is based on the well-established tolerogenicity of IgG carriers, which provided the strategy for our approach.^8,9,13-16  Our group has previously shown that recipients of B-cell blasts, transduced with an Ig fusion of a variety of model antigens or autoimmune targets constructs, are tolerant to the protein epitopes of the expressed genes and that this therapy can lead to modulation of clinical disease.^13-16  In the current study, we have extended this approach to induce tolerance to fVIII and demonstrate that tolerance to 2 important immunodominant domains is sufficient to modulate inhibitor formation in naïve and even in immunized hemophilia mice with significant inhibitory antibody titers.

Evidence has accumulated that many “inhibitors” are directed against exposed regions in the C2 and A2 domain. We and others have identified major T-cell epitopes in the C2 region of fVIII¹⁷  and mutagenesis of the A2 domain can lead to significant reduction of immunogenicity of fVIII.²³  In this study, we first created the fusion IgG construct comprised of residues S2173-Y2332 of the C2 domain and S373-R740 of the A2 domain, respectively, and then inserted them into our retroviral cassette vector. Our data demonstrate that a combination of B-cell blasts expressing these domains induced tolerance to fVIII in hemophilic mice. Importantly, this tolerant state persisted for over 2 months and withstood booster injections of fVIII.

With the present regimen, we do not need to know the precise peptide epitopes within each of these immunodominant domains to use for tolerance induction since presentation by the host's own APCs expressing the fusion IgG peptides eliminates the issue of epitope selection due to HLA polymorphism of the patients.^9,24  Our previous studies showed that it is possible to induce immune tolerance to multiple epitopes in a protein when a full-length construct was used for expression in B cells.⁹  The current findings further support the notion that full-length autoantigen or its domains can be processed and presented appropriately by B-cell blasts to induce immune tolerance.

It may be assumed that a variety of immunogenic epitopes on the surface of fVIII macromolecules are potentially recognized by the immune system when exposed to wild-type fVIII, and the resultant antibodies may or may not be inhibitory, depending on their specificity. We have not attempted to express full length or B-domain-deleted fVIII due to the large size of this protein. However, our results suggest that tolerance to epitopes contained in the C2 and A2 domains of fVIII may be sufficient. Indeed, T-cell tolerance to dominant epitopes within C2 and A2 is sufficient to block inhibitor titers of most specificities, although one might expect that other domains would also contain T-cell epitopes.

It is noteworthy that inhibitory antibody titers showed a more significant reduction than the ELISA titers of antibodies against C2 domain or full-length fVIII protein as measured by ELISA (Figure 1F). The most likely interpretation of this result is that antibody titers measured by ELISA do not reflect the level of the inhibitory response because of the presence of noninhibitory anti-fVIII antibodies.

What is the mechanism by which B-cell presentation of Ig fusion proteins leads to tolerance? We know that B-cell blasts express increased levels of costimulatory molecules, like B7.1 and B7.2, in an evanescent fashion (M. Litzinger and D.W.S., unpublished data, 2004). Indeed, our results and those of others²⁵  suggest that B7 engagement of CTLA-4 may be required for tolerance. Moreover, the apparent need for CD25⁺ T cells for tolerance induction (Figure 4), coupled with their increased presence in long-term tolerized mice,¹⁶  suggests a mechanism by which increased B7 expression may recruit T-regulatory cells bearing CTLA-4 to down-regulate the immune response. Further studies involving demonstration of active suppression by regulatory T cells in an adoptive transfer model and analysis of their specificity and cytokine profiles are needed to explore the role of these regulatory T cells in tolerance.

In sum, we have demonstrated that B-cell presentation of both fVIII domains on an Ig backbone via a retroviral gene therapy protocol is a very effective therapeutic to block the total immune response to fVIII in naive as well as in fVIII-primed recipients. This approach may offer a great promise for prevention and treatment of this serious complication of fVIII replacement therapy.

We are grateful to Dr Kathleen P. Pratt (University of Washington) for gifts of recombinant C2, and Dr Andrei Sarafanov (Department of Biochemistry, University of Maryland) for providing recombinant A2. We thank Yan Su, Linda Jin, Nadia Soukharena, and Damaris Lopez for assistance and valuable discussions, and Drs Jay Lozier, Yufei Jiang, and Biying Xu for reading the manuscript.