The TNF superfamily (TNFSF) includes CD40L, GITRL, RANKL, OX40L, 4-1BBL, LIGHT, CD70, and at least 12 other molecules (9
). From an immunological point of view, CD40L (also called CD154) has special importance as an initiator and promoter of the immune response (47
). CD40L is considered to be the molecular embodiment of the “help” provided by activated CD4+
T cells in that it “licenses” dendritic cells to present processed antigen to CD8+
T cells (10
) and promotes CD8+
T-cell memory responses (38
). CD40L also plays a role in antibody responses by promoting B-cell proliferation and immunoglobulin class switching (70
). These immunostimulatory functions of CD40L have led to many strategies for using it as a vaccine adjuvant.
Despite the intense interest in CD40L as a key activator of immune responses (over 6,000 papers have been published that relate to either CD40L or its receptor), there are currently no clinically accepted means for applying CD40L or most other TNFSFs in vivo. A Phase I clinical trial of a 1-trimer soluble form of CD40L (sCD40LT) (63
) was conducted in patients with cancer, but hepatotoxicity limited the feasibility of delivering this protein by the systemic route (95
). In contrast, the fusion of the extracellular domain of a TNFSF with the body of a multimerizing scaffold molecule provides a new platform for producing multimeric soluble TNFSFs that may circumvent the previous difficulties in studying this important molecular family.
TNFSFs are usually produced as type II trimeric membrane proteins but may be proteolytically cleaved from the cell surface to form soluble trimers. TNF itself was discovered as a soluble protein, which led to the assumption that all TNFSFs might be active as soluble trimers. It was soon determined, however, that soluble TNF activates only one of the two TNF receptors, whereas the multimeric membrane form of TNF activates both receptors (27
). The critical role of multimerization was proven by showing that the activities of membrane TNF could be replicated by a soluble, Flag-tagged form of TNF that was cross-linked by anti-Flag antibody (77
). Parallel studies of FasL, another TNFSF, found that soluble FasL, unlike membrane FasL, did not induce apoptosis in fresh, resting CD4+
T cells. However, a spontaneously aggregating form of soluble FasL (WX1) acted like membrane FasL, again demonstrating the importance of TNFSF multimerization for activity (87
A similar picture has arisen for CD40L, where cross-linking of the soluble form is needed to stimulate B cells to the same level as membrane CD40L. In the case of a CD8-CD40L fusion protein, cross-linking with anti-CD8 antibody was required to drive resting B cells to proliferate (46
). Similarly, membrane CD40L expressed on baculovirus-transduced SF9 insect cells strongly stimulated B cells, but small vesicles (10 to 1,500 nm) from the same cells were much less stimulatory (43
). These in vitro studies indicate that CD40L trimers are most active when presented in a multimeric form.
More recently, the multimerization requirements for the activity of soluble CD40L were studied in greater detail using multimeric soluble TNFSF fusion proteins. C1q superfamily molecules (44
) and collectin superfamily molecules (15
) are proteins that spontaneously multimerize into molecular structures with extended, trimeric, collagen-like arms. Acrp30 (adiponectin) is a C1q superfamily protein that forms a V-shaped molecule with two trimeric arms (67
), whereas pulmonary surfactant protein D (SP-D) is a collectin protein that forms a plus-sign-shaped molecule with four trimeric arms (15
). By replacing the globular head groups of Acrp30 or the carbohydrate recognition domains of SP-D with the extracellular domain of CD40L, Holler et al. (34
) and Haswell et al. (33
) produced molecules carrying two and four trimers of CD40L, respectively. Whereas single-trimer soluble CD40L did not effectively stimulate B cells, these multimeric (i.e., many-trimer) soluble CD40L fusion proteins were strongly stimulatory in vitro (33
The current study aimed at determining the multimerization requirements of soluble CD40L in a relevant in vivo setting, namely immunization. Because the large Acrp30-CD40L and SP-D-CD40L protein molecules were difficult to purify (unpublished data), a DNA vaccine approach was adopted. An antigen plasmid encoding the Gag protein of HIV-1 was mixed with plasmids encoding either native membrane CD40L or one of three forms of soluble CD40L: 1-trimer (stabilized by an isoleucine zipper), 2-trimer (Acrp30-CD40L fusion protein), and 4-trimer (SP-D-CD40L fusion protein). Remarkably, the effectiveness of soluble CD40L as an adjuvant for CD8+ T-cell responses appeared to be directly related to the valence of the trimers, with the 2- and 4-trimer CD40L constructs being significantly stronger than that of the single trimer construct.
It is worth noting, however, that the scaffold moiety might have adjuvant-promoting effects independent of its effect on the valence of the CD40L trimers. For example, the collagen-like portion of SP-D has been reported to bind to gp340, a receptor on the surface of macrophages and possibly DCs (35
). Since this portion of SP-D was included in SP-D-CD40L, binding of this fusion protein to gp340 could serve to concentrate SP-D-CD40L on the surface of antigen-presenting cells and thereby enhance its effectiveness.
The finding that membrane CD40L, pMemCD40L, did not serve as an effective DNA vaccine adjuvant stands in contrast to several previous reports (13
). In all of these cases, the antigen used in the DNA vaccine was cell associated, i.e., located either in the cytoplasm, like β-galactosidase (13
), or on the cell membrane, like HIV-1 gp160 (39
) or respiratory syncytial virus envelope (31
). In contrast, the present report and two previous reports failed to find any adjuvant effects with pMemCD40L-like plasmids when secreted antigens were studied (e.g., TSSA from Treponema cruzi
] or gp120 [88
]). Antigen localization has been previously identified as an important variable for DNA vaccination (12
) and has implications for the selection of adjuvanting molecules. When a DNA vaccine encodes nonsecreted, cell-associated antigen (either cytoplasmic or transmembrane), it may be crucial for dendritic cells to migrate to the vaccination site in order to take up antigen and transport it to the draining lymph node for presentation to T cells. Consistent with this model, chemotactic substances that attract DCs into the intramuscular vaccine site have been shown to significantly enhance the response to DNA vaccines for nonsecreted antigens (51
). While secreted antigens can be transported by DCs in the same way, lymphatic vessels can also carry these soluble proteins (exemplified by lysozyme and ovalbumin) to the draining lymph node, whereupon intranodal DCs take up and present these antigens to T cells within 3 to 6 h (40
). Applying these considerations to the data in this report, if secreted Gag generated by pScGag at the intramuscular vaccination site rapidly moved to the draining lymph node while membrane CD40L from pMemCD40L remained immobilized on myocytes at the vaccination site, this would explain why membrane CD40L (pMemCD40L) failed to adjuvant secreted Gag (pScGag). In contrast, soluble forms of CD40L such as that generated by pSP-D-CD40L can accompany this secreted antigen to the draining lymph node and thereby adjuvant the immune response more effectively.
It is also worth noting that DNA encoding a form of soluble CD40L trimer (sCD40LT) (63
) was previously reported to be an effective adjuvant for a DNA vaccine encoding a nonsecreted antigen (β-galactosidase) (28
). If sCD40LT was expressed as a single-trimer protein, this report would be inconsistent with the in vitro data described above showing that CD40L, like several other TNFSF ligands, is not very active unless multimerized beyond the 1-trimer form (33
). Previous reports that the sCD40LT protein is active on DCs in vitro could be explained if this protein were produced using a low-pH step, such as the pH 2.8 elution step used to remove Flag-tagged sCD40LT protein from an anti-Flag antibody affinity purification column (Srinivasan and Spriggs, U.S. patent 5,716,805). Under these low-pH conditions, the protein becomes disordered into a sticky “molten globule” structure that readily aggregates, thereby effectively multimerizing the purified protein into a many-trimer multimer (56
). The effectiveness of the psCD40LT plasmid as a DNA vaccine adjuvant could also be explained if it contained a Flag tag, as described for another DNA vaccine made with this construct (97
). If so, the formation of anti-Flag antibodies would be expected to multimerize the CD40L trimeric fusion protein into a high-molecular-weight, poorly diffusible immune complex (77
) that could effectively adjuvant a nonsecreted antigen. To exclude this possibility, the Flag sequence was intentionally omitted from the pTrCD40L construct studied in this report, and this 1-trimer soluble CD40L construct was significantly less active than the 2- and 4-trimer forms, consistent with the in vitro studies (33
Given the difficulties in producing a usable form of soluble CD40L, most studies have employed agonistic anti-CD40 antibodies to activate CD40-bearing cells. Agonistic anti-CD40 antibodies were used to prove the role of CD40-activated DCs in generating CD8+
T-cell responses (73
) and to induce anti-tumor responses (23
). Agonistic anti-CD40 antibodies also act as a general vaccine adjuvant (8
) and synergize with Toll-like receptor agonists to elicit profound vaccine-induced CD8+
T-cell responses (4
). However, these antibodies induce splenomegaly and shock-like symptoms unless used in small amounts in a single administration (8
), which raises concerns about using agonistic anti-CD40 antibodies in humans.
In contrast, the multimeric soluble CD40L plasmids described in this report (pAcrp30-CD40L and pSP-D-CD40L) have no apparent toxicity. No signs of distress were seen at any point in the 6-week vaccination protocol, and no inflammation was induced in the muscle tissue 48 h after injecting these plasmids. As evidence that systemic immune activation was absent, the injection of the pScGag antigen plasmid and the pSP-D-CD40L adjuvant plasmid separately into opposite legs did not induce the augmented CD8+
T-cell responses that were seen when the two plasmids were premixed and injected together. Given that DNA vaccination appears to be safe in human volunteers (19
), it is likely that plasmid DNA for multimeric soluble TNFSF fusion proteins could be added to DNA vaccines without incurring the risks associated with agonistic antibodies to TNF receptors.
While DNA vaccination may be safe, its efficacy has been called into question, particularly in humans. Although some DNA vaccines elicit strong immune responses in mice (29
), most investigators have found it necessary to add cytokines or chemokines to enhance the response to immunization (2
). Alternative or complementary approaches include electroporation (65
), gene gun delivery of DNA-coated gold particles (90
), and delivery on polymer microparticles (64
) or by polymer microencapsulation (52
). The need for such modifications is exemplified by the very weak CD8+
T-cell responses seen after DNA vaccination in clinical studies of HIV vaccines (21
). Another example in humans is the complete lack of antibodies induced by a DNA vaccine for malaria (19
Importantly, however, the pSP-D-CD40L-containing vaccine described in this report generated strong CD8+
T-cell responses. While a head-to-head comparison of a pSP-D-CD40L-adjuvanted vaccine with other vaccines has not yet been performed, it is instructive to compare these results with those of other studies in BALB/c mice that employed exactly the same CD8+
T-cell target cell (AMQMLKETI peptide-pulsed P815 cells) for either CTL assays or as a stimulator in IFN-γ ELISPOT assays. By these measures, the results in this report equaled or exceeded those reported for chimpanzee adenovirus 68-based or vaccinia virus-based Gag vaccines (22
), a flavivirus Kunjin Gag vaccine (32
), or Gag delivered intraperitoneally using live Listeria monocytogenes
as a vector (68
). Stronger CD8+
T-cell responses using these assays were reported for a vesicular stomatitis virus vector (30
) or a second-generation rabies vector (58
). However, unlike DNA vaccines, these viral and bacterial vectors generate anti-vector immune responses that could limit booster immunizations with the same agent.
Despite the enhancing effect of pSP-D-CD40L on CD8+
T-cell responses, this adjuvant plasmid did not augment the proliferative and cytokine responses of CD4+
T cells or the production of IgG antibody. Agonistic anti-CD40 antibody was previously reported to increase CD4+
T-cell responses to a modest degree (25
), but the experiments herein failed to detect any significant effects of multimeric soluble CD40L on CD4+
T-cell-proliferative responses and cytokine production. In support of the concept that CD40 stimulation is not linked to subsequent CD4+
T-cell responses, a previous study showed that lymphocytic choriomeningitis virus-exposed transgenic mice lacking either CD40 or CD40L maintained normal CD4+
T-cell-proliferative and cytokine responses to lymphocytic choriomeningitis virus antigens, indicating that the CD40L/CD40 system is not strictly necessary for CD4+
cell responses (66
). It remains possible, however, that the response to antigens that contain stronger CD4+
T-helper epitopes would be enhanced by pSP-D-CD40L (3
The lack of a consistent enhancing effect of pSP-D-CD40L on antibody responses was initially very surprising, given the known role of CD40 stimulation in B-cell activation (70
). Even more unexpected were the data indicating that pSP-D-CD40L could suppress antibody induced by the pScGag antigen plasmid alone (Fig. ). However, these data are in accord with several reports showing that strong CD40 stimulation prevented the movement of B cells into germinal centers, blocked the development of memory B cells, and impaired B-cell differentiation into antibody-secreting plasma cells (6
). Thus, the negative effect of pSP-D-CD40L on antibody production described in this report supports the concept of strong CD40 stimulation provided by multimeric soluble CD40L.
This profile of responses (strong CD8+
T-cell responses but negligible CD4+
T-cell and antibody responses) contrasts with the responses elicited by most experimental vaccines under development. While few infections could benefit from such a skewed vaccine response, this unique response profile has considerable potential for a vaccine against HIV. Unlike most microbes, HIV replicates in activated CD4+
T cells and especially in HIV-specific CD4+
T cells (18
). Indeed, when macaques were vaccinated with an simian immunodeficiency virus (SIV) envelope vaccine that elicited anti-SIV CD4+
T cells but virtually no anti-SIV CD8+
T cells, the vaccinated macaques failed to control a subsequent SIV challenge and rapidly progressed to AIDS (84
). This study revealed the need for a “CD8-focused” vaccine that would generate strong antiviral CD8+
T-cell responses but minimal CD4+
T-cell responses (24
). The multimeric soluble CD40L vaccine of the present study is perhaps the most CD8-focused vaccine yet described and thus should be considered for further testing.
Since fusion with a multimeric scaffold is a generalizable method for producing multimeric soluble TNFSF ligands (34
), this strategy provides a convenient way to test the in vivo effects of other TNFSFs. GITRL is an important TNFSF to test, given the critical role of Tregs in limiting immune responses and the characteristic presence of GITR on these cells (76
). Depletion of CD4+
T cells prior to vaccination, including DNA vaccination, has led to enhanced CD8+
T-cell responses due to the removal of Tregs (50
). Administration of agonistic anti-GITR antibodies reversed Treg suppression of CD4+
T-cell responses (16
). Membrane GITRL expressed on 293 cells completely reversed immunosuppression by Tregs and strongly costimulated antigen-responsive T cells (42
). Recently, Ji et al. described the effects of Flag-tagged soluble GITRL on CD4+
Tregs and found that cross-linking with anti-Flag antibody significantly enhanced the activity of this TNFSF protein in vitro (C. Terhorst, personal communication) (41
). This suggests that soluble GITRL, like FasL, TRAIL, and CD40L (33
), is most active as a multimeric protein.
Consistent with this hypothesis, pSP-D-GITRL was an effective DNA vaccine adjuvant. Immunization with pScGag plus pSP-D-GITRL led to strong CD8+ T-cell responses, although they were less than that with pSP-D-CD40L. Interestingly, the pSP-D-GITRL adjuvant plasmid significantly augmented proliferative responses and antibody production. It is likely that the augmentation of CD4+ T-cell responses by pSP-D-GITRL provided helper CD4+ T cells for B cells. Thus, GITRL, here used as pSP-D-GITRL, is an interesting new vaccine adjuvant, although further studies are needed to determine if it can be used safely without exacerbating autoimmunity.
In conclusion, these studies establish the adjuvant activity of two multimeric soluble TNFSF ligands in vivo when used as part of a DNA vaccine. Several other immunostimulatory TNFSFs remain to be tested in this manner, including RANKL, OX40L, 4-1BBL, CD70, and LIGHT. Using the multimeric scaffold fusion protein approach described here, these and other TNFSFs can now be tested in vivo in an expeditious manner.