Vaccine-Enhanced Donor Lymphocyte Infusion (veDLI)

Primary infection by cytomegalovirus (CMV), a β-herpes-virus, is controlled by the immunocompetent immune system, and CMV subsequently assumes a latent state that may persist indefinitely or reactivate virus production at a later time. Historically, during the period of immune reconstitution following hematopoietic stem cell transplantation (HSCT), up to 80% of CMV-seropositive allogeneic recipients developed CMV reactivation, 50% had viremia and 40% developed CMV disease.³ These rates have declined in recent years, but CMV nonetheless still constitutes an important and potentially lethal post-transplant opportunistic pathogen.⁴ While controlling CMV infections is important for improving clinical outcomes following allogeneic HSCT, viral morbidity and mortality following allogeneic HSCT is not limited to CMV disease. Recipients of partially matched related donor transplantation using stringently T- and B-cell–depleted stem cell products experience uniform reactivation of latent herpes simplex virus (HSV), despite acyclovir prophylaxis, and high levels of drug resistance.⁵ All 7 cases in which the recipient and/or donor were HSV positive demonstrated post-transplant HSV reactivation, and in 5 cases acyclovir-resistant viral isolates were identified. Importantly, despite antiviral drug therapy, herpetic lesions only healed following normalization of T-cell counts,⁵ which is further evidence for the important role of cellular immunity in the control of post-transplant viral infections.

Approaches to improve cellular immunity after allo-geneic HSCT are complicated since donor T-cells exert functional antiviral immunity but also cause graft-versus-host disease (GvHD) (Figure 1; see Color Figures page 513). Thus, T-cell depletion (TCD) of the donor component, which abrogates GvHD, can also lead to impaired antiviral immunity.⁶ Like TCD of the donor graft, recipient conditioning prior to HSCT contributes to the marked immuno-suppression that exists after transplantation. Thus, through the combination of TCD and conditioning, the anti-CMV CD8⁺ cytolytic effectors (CTLs) and CD4⁺ T cells that exert definitive control over CMV infections are typically present at levels too low to suppress CMV replication, increasing the likelihood of CMV disease after transplantation. Furthermore, should GvHD occur despite alterations to the donor graft and recipient conditioning, it is usually treated with steroids. Steroid therapy, in turn, exacerbates the underlying immunosuppression and further impairs the development of functional CMV immunity in a dose-dependent manner.⁷ Thus, iatrogenic immunosuppression due to the combination of TCD of the donor graft, recipient conditioning, and use of steroids to treat GvHD significantly contributes to increased CMV disease after alloge-neic HSCT.

Ganciclovir, foscarnet or cidofovir are frequently used for anti-CMV prophylaxis or to treat diagnosed CMV infections in immunosuppressed HSCT recipients. These drugs can block CMV replication, reduce viral load, and decrease but not eliminate CMV disease (Figure 1; see Color Figures page 513).⁸ However, because HSCT recipients with low levels of anti-CMV CTLs respond relatively poorly to ganciclovir therapy,⁹ antiviral drugs are not a definitive solution to CMV infection in immunosuppressed transplant recipients. Furthermore, prolonged antiviral drug therapy can worsen immunosuppression leading to extended neutropenia, further delay in recovery of anti-CMV CTLs, and outgrowth of resistant viral mutants.¹⁰ In one study, even limited use of ganciclovir caused severe neu-tropenia in 33% (10 of 30) of treated patients, leading to increased mortality from other opportunistic pathogens including Aspergillus fumigatus, Streptococcus pneumoniae, and Pneumocystis carinii.¹¹ Up to 15% of bone marrow transplant (BMT) recipients develop late CMV disease after discontinuation of ganciclovir prophylaxis, and the mortality of late-onset disease is in the range of 50 to 70%.¹² These studies underscore the drawbacks of using anti-CMV drug therapy as the definitive approach to prevent viral disease in HSCT recipients, and the importance of developing alternative antiviral therapies, such as antiviral vaccination, that directly address the underlying problem of prolonged immunosuppression.

There is also evidence that CMV induces or increases the severity of GVHD following allogeneic transplantation (Figure 1; see Color Figures page 513).¹³ Although the mechanism(s) is unclear, it may involve heterologous immunity.¹⁴ Heterologous immunity can best be understood in the context of a normal antiviral immune response, which typically involves activation of naïve antiviral CD8⁺ T cells that subsequently differentiate into CTLs that then kill virus-infected cells.¹⁵ This is not, however, a high fidelity process since virus infection activates a broad range of CTLs with diverse T-cell receptor (TCR) specificities. In fact, some of the activated CTLs have much higher affinities for alternative targets such as alloantigens.¹⁴ The resulting heterologous immunity reshapes the immune profile of the host, modifying subsequent responses to authentic alloantigens such as might be encountered after transplantation. The model virus lymphochoriomeningitis virus (LCMV), a potent inducer of heterologous immunity, breaks transplantation tolerance and promotes rejection of allogeneic transplants including bone marrow allografts.^16,17 While CMV has not yet been shown to induce heterologous immune responses, its ability to activate alloreactive CTLs and induce or increase the severity of GvHD has not been well studied in detail in HSCT recipients.

Additionally, as shown at the bottom of Figure 1 (Color Figures page 513), evidence suggests that ongoing GvHD can accentuate CMV disease through mechanisms unrelated to steroid therapy.¹² Using a mouse model, Cray et al showed significantly higher levels of murine CMV (MCMV) DNA in the lungs of mice that received alloge-neic donor lymphocytes and developed GvHD than in those that did not.¹⁸ Thus, while the mechanisms are incompletely understood at present, the interactions shown in Figure 1 (Color Figures page 513) suggest that the relationship between CMV and GvHD is composed of feedback loops that can amplify both CMV disease and GvHD following allo-geneic HSCT. CMV is central to these pathways. Anti-CMV vaccination may be an effective alternative strategy to rapidly reconstitute anti-CMV immunity, which in turn should interdict the downstream effects of CMV disease (and antiviral drug therapy) on immunosuppression, GvHD, and other opportunistic infections, thereby reducing morbidity and mortality after allogeneic HSCT.

Passive transfer of donor antiviral CTLs is an attractive approach to augment cellular immunity. However, administration of bulk unselected allogeneic donor lymphocytes, at doses containing sufficient numbers of antiviral T-cells to suppress viral replication, results in a significant incidence of GvHD. To address this problem, Riddell et al enriched CD8⁺ CMV-specific CTLs from donor lymphocytes in vitro prior to administration. This method simultaneously addressed two issues. First, alloreactive donor CTLs that could cause GvHD were eliminated by culturing in the absence of their target antigens. Second, by growing anti-CMV CTLs in the presence of viral target antigens, the anti-viral effector cells were stimulated to divide and rapidly expand to much larger numbers than were initially isolated from the donor. Following adoptive transfer, the CMV antigen-specific CTLs did not produce GvHD or other morbidity, and could persist and retain anti-CMV cyto-toxic activity for at least 12 weeks. The effectiveness of antiviral adoptive immunotherapy for rapid reconstitution of cellular immunity and definitive control of CMV and other viral infections has now been demonstrated by a number of other groups since these pioneering studies.^20,21 Unfortunately, despite the first demonstrated success of these methods over 10 years ago, adoptive immunotherapy has still not been widely applied clinically for a number of reasons. In vitro expansion and/or selection of antiviral CTLs is technically complex, labor intensive, and expensive. Furthermore, because growing sufficient numbers of cells for therapeutic efficacy is a lengthy process (weeks or months), this approach is not applicable to acute clinical situations that require rapid availability of donor T-cells. Thus, because of these restrictions, the therapeutic potential of adoptive antiviral immunotherapy has not been fully realized.

We believe that the alternative approach of combining low-dose, unselected DLI with vaccination (veDLI) will be a simpler and more broadly applicable method for promoting rapid immune reconstitution after transplantation than traditional adoptive immunotherapy. A recently published clinical trial supports this proposition. Rapoport et al randomized multiple myeloma patients undergoing au-tologous HSCT to receive DLI (composed of in vitro stimulated PBMC) with or without combined vaccination (7-valent pneumovax).² The investigators demonstrated that a combination of pretransplant vaccination prior to PBMC collection, T cell activation in vitro, and subsequent post-transplant T-cell add-back along with vaccination was highly effective at significantly increasing CD4⁺ T cell–dependent pneumococcal antibodies. Interestingly, neither T-cell add-back nor post-transplant vaccination alone was effective at stimulating antigen-specific immunity, suggesting the importance of combining these modalities for optimal immune reconstitution in HSCT recipients. While this is an important study, it unfortunately utilized in vitro culture and activation of T cells prior to infusion, which may place this approach outside the capabilities of many centers. Future studies to investigate whether in vitro T-cell stimulation is necessary for the success of this protocol should be informative.

veDLI has a number of potential advantages over classic adoptive immunotherapy. Adoptive immunotherapy is a therapy that is difficult to apply routinely, since a subset of CTLs must be enriched and expanded in vitro from a specific donor for a selected recipient. In contrast, a relatively small number of CMV-vaccine constructs could be available off the shelf for veDLI in the vast majority of transplant recipients. For example, because the frequencies of HLA-A1, HLA-A2, and HLA-A24 antigens are 30%, 45%, and 15% in Caucasians, substantially more than half of all patients could be vaccinated with a single construct expressing CMV HLA-A1−, A2−, and A24–restricted immunodominant peptides. Thus, anti-CMV vaccines could be used, like drug therapy, with no on-site preparation required. In contrast to the expense and technical complexity of in vitro CTL expansion for adoptive immunotherapy, anti-CMV vaccination takes advantage of normal immune mechanisms to drive in vivo antiviral CTL expansion (Figure 2; see Color Figures page 513). Neither alloreactive nor other CD8⁺ T-cell subsets would be stimulated by an appropriately constructed vaccine. Importantly, vaccination in lymphopenic HSCT recipients should be extremely effective since the existing T-cell “vacuum” leads to unusually rapid proliferation among a broad array of T cells (homeostatic proliferation).²² In this setting, CD8⁺ T cells presented with strongly reactive peptides (e.g., antiviral T cells following vaccination) proliferate most rapidly, making up a greater proportion of the final lymphocyte population at the conclusion of homeostatic proliferation.²² And, because veDLI harnesses normal immune mechanisms for activation, expansion, and subsequent physiological contraction of the antiviral CTL response,¹⁵ the resulting immunity may be more durable and long-lasting than that produced by infusions of in vitro expanded effector CTLs.

Vaccination against CMV antigens could also have a number of important salutary effects on GvHD in addition to CMV disease. By repressing CMV replication, vaccination would obviate the need for ganciclovir drug therapy, eliminating ganciclovir-associated neutropenia, immuno-suppression, and associated opportunistic infections.¹⁰ Any spurious activation of alloreactive CTLs through heterologous immunity should also be reduced,¹⁴ and initiation or exacerbation of GvHD by CMV infection should be eliminated. By reducing the severity of GvHD, steroids (if needed at all) could be administered at lower doses, which should allow accelerated immune reconstitution.⁷ Furthermore, GvHD would not be expected to accentuate CMV disease, because CMV replication should be suppressed following vaccination. Given the potential advantages of post-HSCT vaccination, it is surprising that with notable exceptions² more effort has not been applied to optimize this therapy for clinical use. Boeckh and colleagues have also recently expressed this sentiment, and speculated that it may be due in part to the minimal data currently available on immune responsiveness following transplantation.⁷ We have begun to address this issue using a preclinical murine BMT model, as described below.

The ubiquitous environmental bacterium Listeria mono-cytogenes (Lm) has many characteristics of a potent vaccination vector. It is avidly phagocytized by host macrophages and other antigen-presenting cells (APCs) (Figure 3; Color Figures page 513). Some bacteria are degraded within phagolysosomes of APCs, and bacterial peptides are loaded onto MHC class II proteins, stimulating a CD4⁺ T-cell response. However, Lm can also induce phagolysosome breakdown and escape into the cytoplasm of the APC. During its intracellular phase, Lm expresses secreted proteins that can be delivered to the APC’s endogenous antigen-processing pathway and presented on MHC class I proteins to elicit a potent CD8⁺ T-cell response.^23,Lm can also activate strong innate immune responses and humoral immunity.²³ For use in vaccination, Lm can be rapidly modified to express recombinant proteins using the pPL2 plasmid integration vector. This plasmid is first engineered using standard recombinant DNA techniques to express desired antigens (such as ovalbumin or the murine CMV [MCMV] immunodominant peptide HGIRNASFI²⁴) as in-frame fusion proteins with the N-terminal region of the bacterial LLO gene. Following transformation, the plasmid becomes stably integrated into an innocuous site of the bacterial genome (adjacent to the tRNA_Arg gene), greatly reducing the chances that recombinant protein expression will be lost during bacterial passage. Engineered Lm strains can hold large recombinant DNA segments (for example, multiple CMV genes encoding several immunodominant epitopes), eliciting a broad immune response following vaccination. Infection with recombinant Lm expressing LCMV nucleoprotein antigen,²⁵ papilloma virus antigen,²⁶ or influenza NP antigen²⁷ leads to potent CD8⁺ T-cell responses against the recombinant peptides and protects experimental animals from lethal LCMV infection, papilloma virus infection, or tumors expressing NP, respectively. Additional desirable vaccination characteristics of Lm include effectiveness in recipients with preexisting antivector immunity, allowing repetitive immunizations; compatibility with simple large scale clinical-grade production; robust sensitivity to ampicillin and other antibiotics, which enhances the safety margin; and effectiveness following oral, intravenous, intraperitoneal, intradermal, intramuscular, and subcutaneous administration.²⁸ Notably, the absence of these characteristics in viral vectors, such as vaccinia and adenovirus, limit their clinical usefulness.

In order to develop a clinically useful Lm-based vaccine platform, investigators at Cerus Corporation began by deleting actA (ΔactA),²⁹ a critical virulence factor involved in cell-to-cell bacterial spread. A recent study showed that administration of similarly attenuated Lm strains to normal human volunteers was well tolerated with no untoward side effects.³⁰ Nonetheless, concern remained that these attenuated strains may still pose a small but important risk to immunocompromised patients such as HSCT recipients. Unfortunately, reports in the literature showed that extensively attenuated (i.e., heat killed) bacteria, which should be safe for administration to immunosuppressed patients, have limited efficacy and do not induce functional IFN-γ–expressing CD8⁺ effector T cells or stimulate a memory T-cell response.³¹ 

To further improve the safety of the Lm platform, while retaining its vaccination potential, a novel vaccination technology was developed. First, the investigators took advantage of proprietary DNA-crosslinking psoralens developed by Cerus. The psoralen presently in clinical use, S-59 {3-[(2-aminoethoxy)methyl]-2,5,9-trimethyl-7H-furo[3,2-g][1] benzopyran-7-one hydrochloride}, is a highly-reactive compound that in the presence of UVA light covalently crosslinks nucleic acids in an irreversible reaction, blocking DNA replication. S-59 is approved for commercial sale in Europe for pathogen inactivation of platelets, and to date S-59 photochemically treated platelet units or T cells have been safely administered to more than 1000 individuals in more than 20 clinical trials.³² Second, the investigators created Lm strains with an additional defined deletion of the ultraviolet light resistance (uvr) AB gene (uvrAB), which functions during nucleotide excision repair. The resulting Listeria ΔactA/ΔuvrAB mutants cannot repair DNA crosslinks resulting from S-59 treatment, and very few (in fact, as little as one) S-59 DNA crosslinks per bacteria render them incapable of dividing.³³ However, because the remainder of their genome is unaltered by the low frequency of crosslinks, these bacteria retain the capacity to infect APCs and express antigens that are presented via both MHC class I and class II pathways.³³ This killed but metabolically active (KBMA) Listeria platform thus defines a new class of vaccines.³³ Consistent with the theoretical advantages of this technology, we have shown (below) that KBMA Lm-MCMV (expressing the MCMV immunodominant antigen HGIRNASFI) is safe enough to administer to highly immunocompromised BMT mice on the day following transplantation without lethality and effectively induces an anti-HGIRNASFI CD8⁺ T-cell response with accompanying long-term lytic activity.

A variety of alternative vaccination methodologies are being developed for CMV, including recombinant CMVgB (which is designed to elicit neutralizing antibodies), whole killed virus, viral peptides, and plasmid-based DNA vaccines.³⁴ Among these approaches, DNA vaccination against MCMV has been extensively studied in a murine system.^35,36 Thus, it is worth considering why KBMA Lm vaccination may be a superior technology, based on safety and efficacy concerns. With respect to safety, DNA vaccines represent a well-defined immunogen contained on a cell-free plasmid vector. However, when used alone, DNA vaccines have limited efficacy and for this reason are typically combined with attenuated microbial vectors in a heterologous prime-boost strategy.³⁷ With this approach, DNA vaccination is given first to “prime” or “focus” the immune response on a defined antigen(s). Days to weeks afterwards, virus expressing the same antigen(s) is given to expand or “boost” the immune response, through a mechanism that probably depends on both higher antigenic expression from, and generation of a pro-inflammatory immune response to, the virus.³⁷ Notably, efforts to create HIV vaccines have affirmed the conclusions that DNA vaccination must be combined with a viral boost (such as from modified vaccinia virus Ankara) to raise sufficiently high levels of protective cellular immunity against HIV-like viruses in nonhuman primates.³⁸ Thus, even with DNA-based plasmid immunization, recipients must be boosted with a microbial vector, which makes this approach unsuitable for immunocompromised HSCT recipients unless extremely safe microbial vectors like KBMA Lm are used. In contrast, the use of KBMA Lm for both prime and boost is not only safe but also rapidly elicits CTL responses after transplantation (see below).

In initial studies, we have used mouse models, including a murine BMT system, to demonstrate proof of concept for veDLI (submitted for publication). Vaccination was performed using Lm engineered to secrete the MCMV H-2^b immunodominant peptide HGIRNASFI²⁴ (Lm-MCMV) in-frame with the bacterial LLO gene using the pPL2 integration vector. In wild-type (non-transplanted) C57BL/6 mice, a single inoculation of 10⁷ colony-forming units (cfu) Lm-MCMV (which is < 0.03 LD₅₀) elicited a strong HGIRNASFI-specific CD8⁺ T-cell response, as quantified by intracellu-lar cytokine and tetramer staining. An average of 8% of splenic CD8⁺ T cells were HGIRNASFI-specific 7 days after vaccination. The long-term (> 200 day) persistence and function of anti-HGIRNASFI CD8⁺ T cells was comparable between vaccinated mice and those that had survived a sublethal MCMV infection. Importantly, Lm-MCMV vaccination of wild-type mice led to marked reductions in viral load after challenging with experimental MCMV infection.

In order to assess the effects of DLI and/or Lm-MCMV vaccination in reconstitution of anti-MCMV immunity after transplantation, a syngeneic C57BL/6 → C57BL/6 BMT model was employed. Recipients were conditioned with a lethal dose of total body irradiation (11 Gy) to eliminate their T-cell function, and then transplanted with 5 × 10⁶ donor bone marrow cells that were depleted of mature T cells. These mice are immune deficient for several weeks after transplantation and are acutely sensitive to CMV lethality during this period.³⁹ Subgroups of BMT mice then received DLI (30 × 10⁶ splenocytes harvested from congenic donors that were either naïve or immunized with 10⁷ cfu Lm-MCMV 7 days previously) and/or vaccinated with 10⁷ cfu Lm-MCMV. When using the novel nonrepli-cating KBMA Lm-MCMV vaccine, immunization could be performed immediately after transplantation with no lethality over the course of multiple studies. Consistently over the course of these studies, BMT recipients who received only DLI or vaccination (or neither) had few if any detectable antiviral T-cells and minimal HGIRNASFI-specific cytolytic activity. In marked contrast, the combination of DLI and vaccination (veDLI) produced a significant expansion of HGIRNASFI-reactive CD8⁺ T cells to levels of 4-25% of total CD8⁺ cells (depending on the specific protocol). These results are very similar to the results of the clinical trial reported by Rapoport et al, where DLI and vaccination had to be combined to elicit specific pneu-mococcal immunity following HSCT.² Importantly, because KBMA-Lm-MCMV vaccination could safely be administered immediately after transplantation, antiviral T-cell expansion could be observed as early as 8 days after BMT, markedly shortening the post-transplant period of defective immunity. As with vaccination of nontransplanted mice, veDLI following BMT led to high levels of persistent antigen-specific lytic activity for at least 200 days.

The prolonged immunosuppression associated with myeloablative HSCT, which predisposes to disease recurrence and opportunistic infections, can be overcome by adoptive immunotherapy using in vitro expanded CTL lines. However, several logistical factors limit the application of this approach. The alternative approach of low-dose DLI combined with vaccination (veDLI) is supported by both clinical trials² and results from murine BMT models (submitted), and is likely to be a more broadly applicable method to reconstitute cellular immunity after HSCT and thus reduce post-transplant morbidity and mortality.