- Email: [email protected]
‘Troy-bodies’: antibodies as vector proteins for T cell epitopes
Biomolecular Engineering 18 (2001) 109– 116 www.elsevier.com/locate/geneanabioeng
‘Troy-bodies’: antibodies as vector proteins for T cell epitopes Elin Lunde a,b, Ingunn B. Rasmussen b, Janne K. Eidem a,b, Tone F. Gregers b, Karoline H. Western a, Bjarne Bogen a, Inger Sandlie b,* b
a Institute of Immunology, The National Hospital, N-0027 Oslo, Norway Department of Biology, Di6ision of Molecular Cell Biology, P.O. Box 1050 Blindern, N-0316 Oslo, Norway
Abstract A major objective in vaccine development is the design of reagents that give a strong, specific T cell response. Targeting of antigens to antigen presenting cells (APC) results in enhanced antigen presentation and T cell activation. In this paper, we describe a novel targeting reagent denoted ‘Troy-bodies’, namely recombinant antibodies with APC-specificity and with T cell epitopes integrated in their C regions. We have made such antibodies with V regions specific for either IgD or MHC class II, and five different T cell epitopes have been tested. All epitopes could be introduced into loops of C domains without disrupting immunoglobulin (Ig) folding. Four have been tested in T cell activation studies, and all could be released and presented by APC. Furthermore, whether IgD- or MHC-specific, the molecules tested enhanced T cell stimulation compared to non-specific control antibodies in vitro as well as in vivo. Using this technology, specific reagents can be designed that target selected antigenic peptides to an APC of choice. Troy-bodies may therefore be useful for manipulation of immune responses, and in particular for vaccination purposes. © 2001 Published by Elsevier Science B.V. Keywords: Antigen processing; MHC; Antibodies; Vaccination; Molecular biology
1. The Troy strategy CD4+ T cells are central to most adaptive immune responses, and strategies that modulate their activation is therefore of great value. We have developed an antigen (Ag) targeting strategy that is based on recombinant antibodies (Ab) that are specific for APC surface molecules, and carry T cell epitopes as an integral part of their constant (C) region [1,2]. Such Ab target the integrated T cell epitopes to the type of APC for which they are specific. By doing so, they can increase the number of T cell epitopes that reach APC, and, conseAbbre6iations: aa, amino acid; Ab, antibody; Ag, antigen; APC, antigen presenting cell; BrdU, bromodeoxyuridine; C, constant; CDR, complementarity determining region; CH1, first C domain of the heavy chain; DC, dendritic cell; HA, hemagglutinin; HEL, hen egg lysozyme; Ig, immunoglobulin; MF, macrophage; MHC, major histocompatibility complex; NIP, 5-iodo-4-hydroxy-3-nitrophenacetyl; OVA, ovalbumin; PCR, polymerase chain reaction; TCR, T cell receptor; V, variable. * Corresponding author. Present address: Department of Biology, Division of Molecular Cell Biology, P.O. Box 1050 Blindern, N-0316 Oslo, Norway. Tel.: +47-22-854568; fax: + 47-22-854605. E-mail address: [email protected] (I. Sandlie).
quently, the number of peptide-major histocompatibility complex (MHC) complexes that are presented on the APC surface Fig. 1. We call these recombinant Ab Troy-bodies, since their effect is comparable to that of the Trojan horse: they enter an APC (city of Troy), and their T cell epitopes (soldiers) are released. So far, we have made peptide-Ab with epitopes derived from the l2 light chain of the M315 myeloma protein, p21ras, hemagglutinin (HA), ovalbumin (OVA), and hen egg lysozyme (HEL) [1,3]. More specifically, the T cell epitopes are introduced into loops of the first C domain of the heavy chain (CH1) by in vitro mutagenesis. Introducing peptides as integrated parts of Ig domains has several advantages. First, loops of Ig domains consist of amino acid (aa) sequences with natural sequence variation that can be accommodated by the domain framework [4]. Second, in contrast to free peptides, which have a very short half-life in vivo [5,6], Ig molecules are relatively stable. Most likely, peptides integrated in Ig molecules will share the long half-life of the carrier Ig. Furthermore, loops within an Ig domain are likely to be more protected against proteases than sequences tailing a Fab fragment [7].
1050-3862/01/$ - see front matter © 2001 Published by Elsevier Science B.V. PII: S 1 3 8 9 - 0 3 4 4 ( 0 1 ) 0 0 0 9 1 - 0
110
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
Fig. 1. Troy-bodies as targeting vehicles for delivery of T cell epitopes to APC. Troy-bodies are APC-specific Ab with T cell epitopes inserted into their C region. When they bind to APC surface molecules, they are internalized, and the integrated T cell epitopes can be released by antigen processing. The T cell epitope can then be presented on MHC class II molecules to CD4+ T cells. The T cell epitopes are introduced into C domains by in vitro mutagenesis. V regions are cloned from hybridomas producing Ab with specificity for APC, and added onto the C regions with a T cell epitope.
The CH1 domain was chosen initially, since it is also a part of Fab fragments of Ig, allowing for the production and testing of smaller molecules than complete Ig. Targeting of the integrated T cell epitopes to APC is obtained by adding variable (V) regions with specificity for APC. The V region genes were cloned from hybridomas producing APC-specific Ab by a polymerase chain reaction (PCR)-based strategy (Fig. 2c). Two different specificities have been tested so far: IgD and MHC class II. Whereas IgD allows studies of B cells as APC, class II allows comparative studies of three major types of professional APC, B cells, dendritic cells (DC) and macrophages (MF) [2,3].
2. Introduction of T cell epitopes into CH1 An Ig domain consists of two b-pleated sheets, each with three to five anti-parallel b-strands that are connected by loops Fig. 2. Some of these loops connect strands of the same sheet, whereas some pass from one sheet to the other. The V domains have nine b-strands (four in one sheet and five in the other), whereas the C domains have only seven (three+ four). As a consequence, the number of loops differs between V and C domains; the V domains have a total of eight and the C domains have a total of six. In V domains, three regions are defined by hypervariability compared to the rest of the V domain. These are the complementarity determining regions, CDR1, 2 and 3, and constitute the antigen binding site.
Fig. 2. Ig domains and localization of T cell epitopes. Structure of Ig V and C region domains. Both types of domain consist of two b-pleated sheets, and the various b-strand are connected by loops. In CH1, several different loops have been used for T cell epitope insertion. We have numbered the loops in this domain according to their order from the N terminus. The loops corresponding to the CDR 1, 2, and 3 are therefore numbered L2, L4, and L6, respectively.
The C domains also have loops in positions similar to the CDR loops, and we have used these loops for epitope insertions. We first asked if T cell epitopes could be: (1) introduced into Ig domains without disturbing Ab folding and secretion; and (2) excised from their new positions and presented to T cells.
2.1. Ab folding and secretion In our first experiments, the 91–101 l2315 epitope [8] was introduced into three different loops of the CH1 domain in human IgG3 [1] and mouse IgG2b [9] (Fig. 3). The three loops chosen to hold the epitopes were those that are similar to the V region CDR loops, and they were first numbered L1–L3, to indicate their similarity to the CDR loops. However, the loops may also be numbered according to their aa positions in the polypeptide chain. Thus, the three loops that correspond to the CDR1–CDR3 loops are now denoted L2, L4 and L6, since they are the second, fourth and sixth loops as counted from the amino terminus (Fig. 2). In both human IgG3 and mouse IgG2b, the epitopes were introduced by in vitro mutagenesis. The mutated heavy chains were expressed together with l light chains, to yield Ab specific for the hapten 5-iodo-4-hydroxy-3-nitrophenacetyl (NIP). Because misfolded proteins are retained in the endoplasmic reticulum [10], secretion of Ab is crucial when evaluating Ab folding. Further indications of a proper structure were obtained when it was verified that the secreted Ab were recognized by anti-IgG3 Ab, and that their specificity was maintained.
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
In human IgG3, we aimed at keeping loops short [1]. Thus, 7 aa of the L2 and L4 loops, and 4 aa of the L6 loop, were replaced by the 11 aa constituting the l2315 epitope. Thus, the three loops were elongated with 4, 4 or 7 aa, respectively. While replacing the original Ig loops, rather than inserting the epitope into the middle of the loop has the advantage of reducing total loop length, a disadvantage is that loop aa that may be of importance for Ab structure are removed. We found that whereas the Ab with the l2315 epitope in L4 and L6 were secreted, the L2 mutant was retained intracellularly. The normal L2 loop of CH1 has two conserved aa, a phenylalanine followed by a proline, that interact with the VH domain [11]. The l2315 peptide, however, does not have this aa combination. We concluded that the observed retention might be a result of abrogated interaction between the CH1 and VH domains. All aa from the original Ig sequence were maintained when we introduced the epitope into CH1 loops of mouse IgG2b [9]. Several different insertion mutants were made, two in L2, one in L4, and one in L6. In addition, an L2 substitution mutant similar to that made in human IgG3, and a mutant (L46) with simultaneous peptide insertions in L4 and L6, were made. We found that all the L2 mutants were secreted, whether the peptide was introduced by replacement or insertion, suggesting that the contact between CH1 and V domains is not a strict requirement for secretion. The L4 and L6 mutants were also secreted, even though the loops had been elongated by 11 aa, but the L46 mutant was not.
2.2. Effect of flanking regions It has been shown that sequences of an Ag that are close to a presented peptide can influence its presenta-
111
tion [12– 16]. Studies where the flanking regions have been altered have shown either increased [12,14] or decreased [12–16] presentation efficiency, or no effect at all [13– 15]. There are several factors related to flanking regions that could affect peptide presentation efficiency. The proteolytic enzymes that cleave Ag have, at least to some extent, substrate specificity for certain sequences [17]. Furthermore, peptides eluted from MHC class II molecules are often longer than the minimum sequences required for binding [18,19], and the aa flanking a minimal peptide could affect the binding of the peptide to the MHC [20,21], or the T cell recognition of the MHC–peptide complex [20,22]. Finally, it has been suggested that antigenic peptides located in or nearby areas of a protein that are unstable will be degraded and presented most readily [23]. In all these experiments, the epitope is located in loops in an Ig domain. Thus, the secondary structure of the epitope and its surroundings would be expected to be quite similar. Even so, the Ig-derived aa flanking the epitope in the various positions are quite different. They might influence the efficiency by which the epitopes are excised from their new positions and presented to T cells. Therefore, the various peptide-Ab were tested in antigen presentation assays. The Ab were added to cultures of APC and specific T cells. T cell activation was measured as T cell proliferation and cytokine production. We found that peptide-Ab added exogenously to APC, resulted in presentation of the epitope and T cell activation. The presentation efficiency varied, however, with the following hierarchy of potency: hL4]hL6 = mL6 \ mL4 = mL2 (h = human IgG3, m= mouse IgG2b) (Fig. 4). Notably, hL2 could not be tested in this assay, since it is not secreted. However, MHC class II+ cells transfected with the hL2 heavy chain were found to present the peptide to T cells, even though the
Fig. 3. In CH1 of human IgG3, the l2315 epitope have been introduced so as to replace the Ig derived aa of the L2 (7aa), L4 (7aa), and L6 (4aa) loops. In addition, epitopes from p21ras, HA, OVA and HEL have been introduced by similar replacements into the L6 loop. In CH1 of mouse IgG2b, the l2315 epitope has been introduced into the Ig derived loops without removing any Ig derived aa. The epitope has been inserted at two different position in the L2 loop, and in one position in each of the L4 and L6 loops.
112
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
Fig. 4. Comparison of different loops as sites for the l2315 epitope. Comparison of the L4 and L6 mutants of mouse IgG2b and human IgG3 peptide-Ab. Symbols: , mouse L4 mutant; 2, mouse L6 mutant; , human L4 mutant; and ", human L6 mutant. Reprinted from Journal of Immunological Methods, volume 245, Eidem et al., Recombinant antibodies as carrier proteins for sub-unit vaccines: influence of mode of fusion on protein production and T-cell activation, pp. 119 – 131, copyright 2000, with permission from Elsevier Science.
heavy chain was retained intracellularly. Thus, hL2 could be processed as an endogenous Ag in the secretory pathway [24–26]. Even though presentation efficiency varied, no clear pattern explaining these differences was apparent from the flanking regions.
2.3. Insertion of epitopes of different lengths and secondary structures For vaccination purposes, it is of importance to be able to introduce a variety of T cell epitopes derived from the same or different Ag. Previous studies have shown that many different T and B cell epitopes can be introduced into the CDR loops of V domains [27–35], we tested three epitopes in addition to the l2315 already described in a loop in CH1. The new T cell epitopes were: aa 323–339 from OVA [36], aa 46– 61 from HEL [37], and aa 110– 120 from influenza virus (HA) [38]. For these studies, the L6 loop of CH1 of human IgG3 was used, and all three epitopes were introduced so as to replace the loop (Fig. 3). We found that all the resulting peptide-Ab were secreted. This demonstrates that many different peptides can be introduced into a loop of CH1 without disrupting the Ab structure required for secretion. First, epitopes of different lengths (17, 16 and 11 aa, respectively) can be used. Notably, additional unpublished results have shown that p21ras peptides of upto 25 aa can be introduced without interrupting Ab secretion (Rasmussen et al., unpublished). These epitopes were introduced by a replacement mutation where four aa of the L6 loop were removed. Thus, we have shown that loop elongations of upto 21 aa can be introduced in L6 of CH1. Similar results were obtained for epitopes in CDR2 and CDR3 loops of V regions, where loop elongations of up to 17 aa have been accepted [30]. Second, epitopes of differ-
ent secondary structures may be introduced into loops. When in its original position, the 91– 101 (l2315) epitope is found as a loop in the CDR3 region of the l2315 light chain. It is therefore likely to have a conformation that facilitates formation of a loop in the CH1 domain. In contrast, the other epitopes employed have different conformations when in their original Ag (the structure of the HA epitope is not known, though). Whereas the p21ras epitope has a b-strand/loop/a-helix structure [39], the OVA epitope is mostly a b-strand [40], and the HEL epitope has three turns and two b-strands [41]. Even so, the b-strands of Ig domains build domains that accept these various sequences. Based on these results, we anticipate that the Ig C domain frameworks can accommodate peptides derived from many different secondary structures. This would also fit well with the results on the V region CDR loops, which have been used for introduction of many different epitopes [27– 35]. The peptide-Ab with HA, OVA and HEL epitopes were tested in antigen presentation assays. In all cases, the epitopes were found to be presented on MHC class II to specific CD4+ T cells [3]. In conclusion, the experiments described confirm two important requirements of our overall strategy: (1) different T cell epitopes can be introduced into loops of Ig C domains without disturbing Ab folding and secretion; and (2) the T cell epitopes can be excised from their new positions by APC and presented to T cells. Even so, we found that some loops are more permissive than others, both with regard to Ab folding (i.e. secretion) and presentation to T cells.
3. Addition of V regions with specificity for APC The peptide-Ab described above were specific for the hapten NIP. To investigate the effect of targeting, the Ab V regions were replaced by V regions with specificity for APC. These were cloned from B cell hybridomas producing Ab with two different specificities: IgD [2] and MHC class II (Lunde et al., submitted).
3.1. Targeting to IgD Previous studies have shown that targeting to IgD induces potent responses [42–44]. We chose V regions that confer specificity for the IgDa allotype only, and not IgDb [45]. Thus, normal mice or cells with the IgDb allotype could serve as negative controls, without the need for non-physiological manipulations, such as B cell depletion. The VH and VL genes from the Ig(5a)7.2 hybridoma were cloned using a PCR-based strategy developed in our laboratory that allowed amplification of complete rearranged V(D)J genes [46]. In a first PCR, upstream primers annealing to the leader se-
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
quences were used in combination with a downstream primer annealing to the C region. In this PCR, complete rearranged V(D)J genes flanked by leader and C region sequences were amplified. The PCR products were sequenced, and in the second PCR, the exact V genes were amplified using new primers annealing precisely to the ends of the V genes and introducing unique restriction enzyme sites. The VH and VL genes were then subcloned into expression vectors upstream of Cg3 and Ck genes, respectively. In principle, the procedure allows introduction of V region genes from any B cell hybridoma or phage display library, and makes possible construction of Troy-bodies with any chosen specificity. Assembled IgDa-specific Ab were isolated from cell supernatant of transfected NS0 cells. The recombinant Ab were tested and found to be specific for the IgDa allotype, and not the IgDb allotype as expected [2]. The first IgD-specific Ab tested had the l2315 epitope in the L6 loop of the CH1 domain of human IgG3. This Ab, aIgD.L6 (aIgD.L3 in the original publication), was compared to the corresponding, but NIP-specific, L6 Ab. The Ab were tested in antigen presentation assays in which titrated amounts of Ab were added to cultures of APC and specific T cells. We found that targeting to APC by use of aIgD.L6 in vitro resulted in a 1000-fold better presentation efficiency compared to the NIP-specific L6 control Ab (Fig. 5a) [2]. As expected, B cells were required for this increase, and also the presence of IgDa expression. However, when aIgD.L6 Fab fragments were tested, they were found to be less effective than complete aIgD.L6 [2]. Complete aIgD.L6 Ab was then tested for targeting to B cells in vivo. In these experiments, mice were injected i.v. with the aIgD.L6 or NIP-specific L6 Ab. After 112 h, their spleen cells
113
were isolated and tested for ability to present the l2315 epitope to T cells in vitro. Spleen cells from mice injected with aIgD.L6 were found to be 100 times better at presenting the epitope than spleen cells from mice injected with L6 (Fig. 5b) [2]. Notably, these experiments only addressed the question of whether or not the injected Ab were targeted to splenic APC in vivo. Since the T cell proliferation assays were done in vitro, processing of aIgD.L6 and presentation of the l2315 epitope could have occurred during these 2 days of in vitro culture. To address the question of in vivo T cell proliferation, aIgD.L6 or L6 was injected i.v. into T cell receptor (TCR)-transgenic mice, the T cells of which are specific for the l2315 epitope. The mice received bromodeoxyuridine (BrdU), and T cell proliferation was measured as BrdU incorporation. We found that aIgD.L6 was 20-fold more efficient than L6 at inducing in vivo T cell proliferation (Fig. 5c) [2]. These initial experiments were done using Ab with the l2315 epitope. However, we have also made IgDspecific Ab with the OVA, HA and HEL epitopes in the L6 loop of CH1 [3]. They were also 1000-fold more efficient than the corresponding peptide-Ab in in vitro experiments. Thus, the targeting strategy works equally well with T cell epitopes other than l2315. As the amount of in vivo data in the literature is growing, it becomes increasingly clear that targeting to IgD in vivo may well give heterogeneous results. Not only T cell activation [42,47,48], but also T cell non-responsiveness [49,50] or altered T cell responses [51,52], have been observed. Regardless of targeting strategy, however, a major disadvantage of using IgD as target is that the B cells involved may become activated ([42], Lunde et al.,
Fig. 5. Targeting to IgD by use of the Troy-body that has the l2315 epitope in the L6 loop of human IgG3. (A) In vitro targeting. The IgD-specific Troy-body and control Ab were titrated into cultures of irradiated BALB/c splenocytes (as a source of APC) and lymph node cells from TCR-transgenic mice (as a source of l2315 specific T cells). Antigen presentation was measured as T cell proliferation (Thd incorporation) and is shown as a function of Ab concentration in the cultures; (B) In vivo targeting. Mice were injected intravenously with the IgD-specific Troy-body or the NIP-specific control Ab. After 112 h, the spleen cells were isolated and tested for their ability to stimulate T cells in vitro. T cell proliferation is shown as a function of serum human IgG3 112 h after injection; (C) In vivo T cell proliferation. TCR-transgenic mice were injected with the IgD-specific Troy-body and the NIP-specific control Ab, and given BrdU for a 5 day period. On day 5, spleen cells were analyzed by flow cytometry to reveal proliferating T cells. The percentage of proliferating (BrdU+), l2315-specific (GB113+), CD4+ T cells in the spleen is shown as a function of amount of Ab injected. Symbols: , IgD-specific Troy-body with l2315; ", NIP-specific peptide-Ab with l2315; , IgD-specific Ab without peptide. Reprinted from Nature Biotechnology, volume 17, Lunde et al., Antibodies engineered with IgD specificity efficiently deliver integrated T-cell epitopes for antigen presentation by B cells, pp. 670 – 675, copyright 1999, Nature Biotechnology.
114
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
unpublished observations). Since all B cells expressing IgD (irrespective of specificity) can be activated by the IgD-specific Ab, the selective activation of only Ag-specific B cells is lost. Consequently, the potential for inducing autoimmunity is great, and the IgD-specific Ab should therefore be viewed as tools for testing the strategy, and not as a suggestion for a future vaccine.
3.2. Targeting to MHC class II MHC class II molecules are good targets for several reasons: first, class II molecules have, like IgD, proven to be efficient in previous targeting studies [43,44,53,54]. Second, use of class II permits targeting to all the different types of APC, allowing for comparative studies. We have compared the effect of targeting to B cells, DC and MF, respectively, and for this purpose, we have used the Ab with the l2315 epitope. We evaluated the different APC both by use of in vitro cell lines, and by use of highly purified ex vivo splenic APC, isolated by cell sorting. These experiments showed that all three types of APC could be used as target cells for MHC class II-specific Ab, and that the difference between the targeted and non-targeted Ab was approximately the same for all three cell types (Lunde et al., submitted).
3.3. Targeting to IgD or MHC class II compared to presentation of synthetic peptides or antigenic proteins The various IgD- and MHC class II-specific Troybodies have been compared not only to their respective corresponding peptide-Ab, but also to the synthetic peptides (l2315, HA and HEL) or whole antigenic proteins (l2315 and OVA). When compared on a molar basis, the targeted Ab were normally 10 000-fold better at activating T cells in vitro than their respective synthetic peptides ([3] and Lunde et al., submitted). An in vivo comparison has not been done, but we expect the difference to be even more pronounced in vivo, since Ab are relatively stable molecules compared to synthetic peptides, which are rather unstable in vivo [5,6]. Whereas the synthetic peptides tended to be somewhat less efficient than the NIP-specific peptide-Ab, whole antigenic proteins induced T cell responses at approximately the same Ag dose as the NIP-specific Ab ([3,55] Lunde et al., submitted). Thus, on a molar basis, Troy-bodies were 1000 times more efficient than whole antigenic proteins.
4. Conclusions We have developed a targeting strategy based on recombinant Ab called Troy-bodies. The Troy-bodies have C regions containing T cell epitopes and V regions
with specificity for APC. When they bind to their target APC, the T cell epitopes are released by antigen processing and presented on MHC class II molecules to CD4+ T cells. The T cell epitopes were introduced into loops of the CH1 domain, and so far, three different loops (L2, L4, L6) have been tested. These three loops correspond to the V region CDR loops. Both human IgG3 and mouse IgG2b were used, and they were first loaded with an epitope from the l2315 light chain. Two different strategies for loop introduction were employed. In the replacement mutants, the aa of the T cell epitope replaced the 4–7 aa of the Ig loop. In contrast, the insertion mutants had the aa of T cell epitope inserted into the Ig loop without removal of any Ig derived aa. In most cases, the resulting recombinant peptide-Ab appeared to be correctly folded. An exception was the L2 loop of the CH1 domain of human IgG3, which seems to be less permissive than the others. When the peptide-Ab were added to APC, they were processed and presented to specific T cells. Even the intracellularly retained L2 mutant of human IgG3 could be presented if transfected into class II+ cells. Thus, all positions seem to be permissive with respect to antigen processing and presentation. The presentation efficiency however, varied between the different mutants. One loop, the L6 loop of human IgG3, was chosen for further studies. Three more T cell epitopes (derived from HA, HEL and OVA) were introduced into this loop. In their original positions in the native proteins, the epitopes are folded into different secondary structures. Even so, they were all accepted by the CH1 domain without disruption of Ab folding and secretion. Furthermore, all three epitopes were excised and presented to T cells. To produce APC-specific Ab, V region genes were cloned by a PCR-based strategy from hybridomas producing Ab of relevant specificities. These V regions were added to C regions with epitopes. So far, two different specificities have been tested, IgD and MHC class II (I-E). In both cases, a 1000-fold increase in in vitro presentation efficiency was obtained for the APC specific peptide Ab (the Troy-body) compared to the NIP-specific peptide-Ab. When Troy-bodies were targeted to APC in vivo after injection into mice, splenic APC were 100-fold better at activating T cells in vitro than those isolated from mice injected with control Ab. The class II-specific Troy-bodies were used to target all three types of professional APC (B cells, MF and DC). All three cell types worked equally well. In addition, one of the APC types, the B cells, were tested with both IgD- and class II-specific Troy-bodies. The effect appeared to be similar ( × 1000) for Troy-bodies of the two specificities. An overview of the results is given in Table 1.
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
115
Table 1 Summary Antigen
Epitope
Carrier
Specificities
Secretion
T cells
T cell activation
l2315
aa 91–101
IgG3 IgG2b
NIP IgD I-Ea
Yes
1. T cell clones 4B2A1 and 7A10B2 2. l2315-specific TcR TG mice (B. Bogen)
Yes (in vitro and in vivo)
P21ras
aa 1–25 G12V, G12R, G13D aa 5–21 G12V, G12R, G13D
IgG3
NIP HLA-DP
Yes
T cell clones (G. Gaudernack)
N.D.
Hen egg lysozyme (HEL)
aa 46–61
IgG3
NIP IgD I-Ea
Yes
T cell hybridoma 3A9 (R. Germain)
Yes (in vitro)
Hemagglutinin (HA)
aa 110–120
IgG3
NIP IgD I-Ea
Yes
T cells from HA-specific TCR TG mice (K. Karjalainen)
Yes (in vitro)
Ovalbumin (OVA)
aa 323–339
IgG3
NIP IgD I-Ea
Yes
1. T cell hybridoma DO11.10 2. DO11.10 TcR TG mice (K. Murphy)
Yes (in vitro)
5. Future perspectives Since B cells are thought to be the least effective APC at activating naive T cells, it is a goal to develop Troy-bodies that specifically target other subsets of APC. An important advantage of using Ig C domain loops for introduction of T cell epitopes is that an Ab molecule has many different C domains, each having several loops. More specifically, each IgG heavy chain has three C domains whereas the light chain has one, and since each domain has six loops, an IgG molecule could potentially harbor 24 different peptides in its C region. In addition, the ‘lower’ loops of the V domains could possibly also be used. Thus, there are many loops that can be used for epitope insertion, and a multi epitope Ab may be created. The difference in presentation efficiency for the various positions point to a role of flanking regions. A more uniform presentation may possibly be obtained by introducing minimal epitopes with flanking regions that promote processing. Another advantage of using Ig loops for peptide grafting is that it allows production of Ig (Troy-body) variants, taking advantage of the growing knowledge on Ab-derived fragments. The Troy-body variants could then be compared in order to find the best alternative for an optimal targeting effect in vivo. Notably, the targeting strategy could be used for goals other than the activation of CD4+ T cells. For example, surface molecules that can lead to presenta-
tion of extracellular Ag on MHC class I have recently been identified on both B cells [56] and DC [57]. These could be targeted with Troy-bodies containing MHC class I-restricted peptides, hopefully leading to epitope presentation on class I and activation of CD8+ T cells. If so, Troy-bodies could be used as vaccines for diseases in which CD8+ T cell responses are cruical. Finally, for conditions such as autoimmune diseases, Troy-bodies could be used to inhibit T cell activation. For e.g. by inserting antagonist peptides [58], or directing Troy-bodies to tolerizing APC [59], T cell anergy or tolerization could possibly be obtained. Another possibility would be to target Troy-bodies to thymic epithelium or DC, thereby affecting positive or negative selection of thymocytes.
Acknowledgements The work described was supported by the Norwegian Cancer Society and the Research Council of Norway.
References [1] Lunde E, Bogen B, Sandlie I. Mol Immunol 1997;34:1167 – 76. [2] Lunde E, Munthe L, Vabø A, Sandlie I, Bogen B. Nat Biotechnol 1999;17:670 – 5. [3] Rasmussen IB, Lunde E, Michaelsen TE, Bogen B, Sandlie I. Proc Natl Acad Sci USA 2001; in press
116
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
[4] Lesk, Tramontano. Antibody Engineering, A Practical Guide. New York: W.H. Freeman and Co., 1992. pp. 1 – 38. [5] Lee VH. Crit RevTher Drug Carrier Syst 1988;5:69 –97. [6] Powell MF, Grey H, Gaeta F, Sette A, Colon S. J Pharm Sci 1992;81:731 – 5. [7] Neuberger MS, Williams GT, Fox RO. Nature 1984;312:604 – 8. [8] Bogen B, Lambris JD. EMBO J 1989;8:1947 –52. [9] Eidem JK, Rasmussen IB, Lunde E, Gregers TF, Rees AR, Bogen B, Sandlie I. J Immunol Methods 2000;245:119 –31. [10] Lippincott-Schwartz J, Bonifacino JS, Yuan LC, Klausner RD. Cell 1988;54:209 – 20. [11] Lesk AM, Chothia C. Nature 1988;335:188 –90. [12] Vacchio MS, Berzofsky JA, Krzych U, Smith JA, Hodes RJ, Finnegan A. J Immunol 1989;143:2814 –9. [13] Lowenadler B, Lycke N, Svanholm C, Svennerholm AM, Krook K, Gidlund M. Mol Immunol 1992;29:1185 – 90. [14] Eisenlohr LC, Yewdell JW, Bennink JR. J Exp Med 1992;175:481 – 7. [15] Leclerc C, Martineau P, Charbit A, Lo-Man R, Deriaud E, Hofnung M. Mol Immunol 1993;30:1561 –72. [16] Janssen R, Wauben M, van der Zee R, de Gast M, Tommassen J. Int Immunol 1994;6:1187 –93. [17] Riese RJ, Chapman HA. Curr Opin Immunol 2000;12:107 – 13. [18] Rudensky AY, Preston-Hurlburt P, Hong SC, Barlow A, Janeway CA Jr. Nature 1991;353:622 – 7. [19] Chicz RM, Urban RG, Lane WS, Gorga JC, Stern LJ, Vignali DA, Strominger JL. Nature 1992;358:764 –8. [20] Berzofsky JA, Brett SJ, Streicher HZ, Takahashi H. Immunol Rev 1988;106:5 –31. [21] Mukherjee S, Ghosh RN, Maxfield FR. Physiol Rev 1997;77:759 – 803. [22] Wauben MH, van der Zee R, Joosten I, Boog CJ, van Dijk AM, Holewijn MC, Meloen RH, Van Eden W. J Immunol 1993;150:5722 – 30. [23] Landry SJ. Immunol Today 1997;18:527 – 32. [24] Weiss S, Bogen B. Cell 1991;64:767 –76. [25] Brooks A, Hartley S, Kjer-Nielsen L, Perera J, Goodnow CC, Basten A, McCluskey J. Proc Natl Acad Sci USA 1991;88:3290 – 4. [26] Kittlesen DJ, Brown LR, Braciale VL, Sambrook JP, Gething MJ, Braciale TJ. J Exp Med 1993;177:1021 – 30. [27] Billetta R, Hollingdale MR, Zanetti M. Proc Natl Acad Sci USA 1991;88:4713 – 7. [28] Zaghouani H, Krystal M, Kuzu H, Moran T, Shah H, Kuzu Y, Schulman J, Bona C. J Immunol 1992;148:3604 – 9. [29] Zaghouani H, Steinman R, Nonacs R, Shah H, Gerhard W, Bona C. Science 1993;259:224 –7. [30] Billetta R, Filaci G, Zanetti M. Eur J Immunol 1995;25:776 – 83. [31] Cook J, Barber BH. Vaccine 1995;13:1770 – 8.
[32] Zaghouani H, Anderson SA, Sperber KE, Daian C, Kennedy RC, Mayer L, Bona CA. Proc Natl Acad Sci USA 1995;92:631 – 5. [33] Brumeanu TD, Bot A, Bona CA, Dehazya P, Wolf I, Zaghouani H. Immunotechnology 1996;2:85 – 95. [34] Xiong S, Gerloni M, Zanetti M. Nat Biotechnol 1997;15:882 –6. [35] Legge KL, Min B, Cestra AE, Pack CD, Zaghouani H. J Immunol 1998;161:106 – 11. [36] Sette A, Buus S, Colon S, Smith JA, Miles C, Grey HM. Nature 1987;328:395 – 9. [37] Allen PM, Strydom DJ, Unanue ER. Proc Natl Acad Sci USA 1984;81:2489 – 93. [38] Hackett CJ, Hurwitz JL, Dietzschold B, Gerhard W. J Immunol 1985;135:1391 – 4. [39] Krengel U, Schlichting L, Scherer A, Schumann R, Frech M, John J, Kabsch W, Pai EF, Wittinghofer A. Cell 1990;62:539 – 48. [40] Stein PE, Leslie AG, Finch JT, Carrell RW. J Mol Biol 1991;221:941 – 59. [41] Maenaka K, Matsushima M, Song H, Sunada F, Watanabe K, Kumagai I. J Mol Biol 1995;247:281 – 93. [42] Finkelman FD, Scher I, Mond JJ, Kung JT, Metcalf ES. J Immunol 1982;129:629 – 37. [43] Casten LA, Pierce SK. J Immunol 1988;140:404 – 10. [44] Snider DP, Segal DM. J Immunol 1987;139:1609 – 16. [45] Oi VT, Herzenberg LA. Mol Immunol 1979;16:1005 – 17. [46] Norderhaug L, Olafsen T, Michaelsen TE, Sandlie I. J Immunol Methods 1997;204:77 – 87. [47] Finkelman FD, Scher I, Mond JJ, Kessler S, Kung JT, Metcalf ES. J Immunol 1982;129:638 – 46. [48] Lees A, Morris SC, Thyphronitis G, Holmes JM, Inman JK, Finkelman FD. J Immunol 1990;145:3594 – 600. [49] Eynon EE, Parker DC. J Exp Med 1992;175:131 – 8. [50] Snider DP, Kaubisch A, Segal DM. J Exp Med 1990;171:1957 – 63. [51] Day MJ, Tse AG, Puklavec M, Simmonds SJ, Mason DW. J Exp Med 1992;175:655 – 9. [52] Saoudi A, Simmonds S, Huitinga I, Mason D. J Exp Med 1995;182:335 – 44. [53] Carayanniotis G, Barber BH. Nature 1987;327:59 – 61. [54] Berg SF, Mjaaland S, Fossum S. Eur J Immunol 1994;24:1262 – 8. [55] Bogen B, Malissen B, Haas W. Eur J Immunol 1986;16:1373 –8. [56] Ke Y, Kapp JA. J Exp Med 1996;184:1179 – 84. [57] Regnault A, Lankar D, Lacabanne V, Rodriguez A, Thery C, Rescigno M, Saito T, Verbeek S, Bonnerot C, Ricciardi-Castagnoli P, Amigorena S. J Exp Med 1999;189:371 – 80. [58] Sloan-Lancaster J, Evavold BD, Allen PM. Nature 1993;363:156 – 9. [59] Croft M, Joseph SB, Miner KT. J Immunol 1997;159:3257 –65.
‘Troy-bodies’: antibodies as vector proteins for T cell epitopes Elin Lunde a,b, Ingunn B. Rasmussen b, Janne K. Eidem a,b, Tone F. Gregers b, Karoline H. Western a, Bjarne Bogen a, Inger Sandlie b,* b
a Institute of Immunology, The National Hospital, N-0027 Oslo, Norway Department of Biology, Di6ision of Molecular Cell Biology, P.O. Box 1050 Blindern, N-0316 Oslo, Norway
Abstract A major objective in vaccine development is the design of reagents that give a strong, specific T cell response. Targeting of antigens to antigen presenting cells (APC) results in enhanced antigen presentation and T cell activation. In this paper, we describe a novel targeting reagent denoted ‘Troy-bodies’, namely recombinant antibodies with APC-specificity and with T cell epitopes integrated in their C regions. We have made such antibodies with V regions specific for either IgD or MHC class II, and five different T cell epitopes have been tested. All epitopes could be introduced into loops of C domains without disrupting immunoglobulin (Ig) folding. Four have been tested in T cell activation studies, and all could be released and presented by APC. Furthermore, whether IgD- or MHC-specific, the molecules tested enhanced T cell stimulation compared to non-specific control antibodies in vitro as well as in vivo. Using this technology, specific reagents can be designed that target selected antigenic peptides to an APC of choice. Troy-bodies may therefore be useful for manipulation of immune responses, and in particular for vaccination purposes. © 2001 Published by Elsevier Science B.V. Keywords: Antigen processing; MHC; Antibodies; Vaccination; Molecular biology
1. The Troy strategy CD4+ T cells are central to most adaptive immune responses, and strategies that modulate their activation is therefore of great value. We have developed an antigen (Ag) targeting strategy that is based on recombinant antibodies (Ab) that are specific for APC surface molecules, and carry T cell epitopes as an integral part of their constant (C) region [1,2]. Such Ab target the integrated T cell epitopes to the type of APC for which they are specific. By doing so, they can increase the number of T cell epitopes that reach APC, and, conseAbbre6iations: aa, amino acid; Ab, antibody; Ag, antigen; APC, antigen presenting cell; BrdU, bromodeoxyuridine; C, constant; CDR, complementarity determining region; CH1, first C domain of the heavy chain; DC, dendritic cell; HA, hemagglutinin; HEL, hen egg lysozyme; Ig, immunoglobulin; MF, macrophage; MHC, major histocompatibility complex; NIP, 5-iodo-4-hydroxy-3-nitrophenacetyl; OVA, ovalbumin; PCR, polymerase chain reaction; TCR, T cell receptor; V, variable. * Corresponding author. Present address: Department of Biology, Division of Molecular Cell Biology, P.O. Box 1050 Blindern, N-0316 Oslo, Norway. Tel.: +47-22-854568; fax: + 47-22-854605. E-mail address: [email protected] (I. Sandlie).
quently, the number of peptide-major histocompatibility complex (MHC) complexes that are presented on the APC surface Fig. 1. We call these recombinant Ab Troy-bodies, since their effect is comparable to that of the Trojan horse: they enter an APC (city of Troy), and their T cell epitopes (soldiers) are released. So far, we have made peptide-Ab with epitopes derived from the l2 light chain of the M315 myeloma protein, p21ras, hemagglutinin (HA), ovalbumin (OVA), and hen egg lysozyme (HEL) [1,3]. More specifically, the T cell epitopes are introduced into loops of the first C domain of the heavy chain (CH1) by in vitro mutagenesis. Introducing peptides as integrated parts of Ig domains has several advantages. First, loops of Ig domains consist of amino acid (aa) sequences with natural sequence variation that can be accommodated by the domain framework [4]. Second, in contrast to free peptides, which have a very short half-life in vivo [5,6], Ig molecules are relatively stable. Most likely, peptides integrated in Ig molecules will share the long half-life of the carrier Ig. Furthermore, loops within an Ig domain are likely to be more protected against proteases than sequences tailing a Fab fragment [7].
1050-3862/01/$ - see front matter © 2001 Published by Elsevier Science B.V. PII: S 1 3 8 9 - 0 3 4 4 ( 0 1 ) 0 0 0 9 1 - 0
110
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
Fig. 1. Troy-bodies as targeting vehicles for delivery of T cell epitopes to APC. Troy-bodies are APC-specific Ab with T cell epitopes inserted into their C region. When they bind to APC surface molecules, they are internalized, and the integrated T cell epitopes can be released by antigen processing. The T cell epitope can then be presented on MHC class II molecules to CD4+ T cells. The T cell epitopes are introduced into C domains by in vitro mutagenesis. V regions are cloned from hybridomas producing Ab with specificity for APC, and added onto the C regions with a T cell epitope.
The CH1 domain was chosen initially, since it is also a part of Fab fragments of Ig, allowing for the production and testing of smaller molecules than complete Ig. Targeting of the integrated T cell epitopes to APC is obtained by adding variable (V) regions with specificity for APC. The V region genes were cloned from hybridomas producing APC-specific Ab by a polymerase chain reaction (PCR)-based strategy (Fig. 2c). Two different specificities have been tested so far: IgD and MHC class II. Whereas IgD allows studies of B cells as APC, class II allows comparative studies of three major types of professional APC, B cells, dendritic cells (DC) and macrophages (MF) [2,3].
2. Introduction of T cell epitopes into CH1 An Ig domain consists of two b-pleated sheets, each with three to five anti-parallel b-strands that are connected by loops Fig. 2. Some of these loops connect strands of the same sheet, whereas some pass from one sheet to the other. The V domains have nine b-strands (four in one sheet and five in the other), whereas the C domains have only seven (three+ four). As a consequence, the number of loops differs between V and C domains; the V domains have a total of eight and the C domains have a total of six. In V domains, three regions are defined by hypervariability compared to the rest of the V domain. These are the complementarity determining regions, CDR1, 2 and 3, and constitute the antigen binding site.
Fig. 2. Ig domains and localization of T cell epitopes. Structure of Ig V and C region domains. Both types of domain consist of two b-pleated sheets, and the various b-strand are connected by loops. In CH1, several different loops have been used for T cell epitope insertion. We have numbered the loops in this domain according to their order from the N terminus. The loops corresponding to the CDR 1, 2, and 3 are therefore numbered L2, L4, and L6, respectively.
The C domains also have loops in positions similar to the CDR loops, and we have used these loops for epitope insertions. We first asked if T cell epitopes could be: (1) introduced into Ig domains without disturbing Ab folding and secretion; and (2) excised from their new positions and presented to T cells.
2.1. Ab folding and secretion In our first experiments, the 91–101 l2315 epitope [8] was introduced into three different loops of the CH1 domain in human IgG3 [1] and mouse IgG2b [9] (Fig. 3). The three loops chosen to hold the epitopes were those that are similar to the V region CDR loops, and they were first numbered L1–L3, to indicate their similarity to the CDR loops. However, the loops may also be numbered according to their aa positions in the polypeptide chain. Thus, the three loops that correspond to the CDR1–CDR3 loops are now denoted L2, L4 and L6, since they are the second, fourth and sixth loops as counted from the amino terminus (Fig. 2). In both human IgG3 and mouse IgG2b, the epitopes were introduced by in vitro mutagenesis. The mutated heavy chains were expressed together with l light chains, to yield Ab specific for the hapten 5-iodo-4-hydroxy-3-nitrophenacetyl (NIP). Because misfolded proteins are retained in the endoplasmic reticulum [10], secretion of Ab is crucial when evaluating Ab folding. Further indications of a proper structure were obtained when it was verified that the secreted Ab were recognized by anti-IgG3 Ab, and that their specificity was maintained.
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
In human IgG3, we aimed at keeping loops short [1]. Thus, 7 aa of the L2 and L4 loops, and 4 aa of the L6 loop, were replaced by the 11 aa constituting the l2315 epitope. Thus, the three loops were elongated with 4, 4 or 7 aa, respectively. While replacing the original Ig loops, rather than inserting the epitope into the middle of the loop has the advantage of reducing total loop length, a disadvantage is that loop aa that may be of importance for Ab structure are removed. We found that whereas the Ab with the l2315 epitope in L4 and L6 were secreted, the L2 mutant was retained intracellularly. The normal L2 loop of CH1 has two conserved aa, a phenylalanine followed by a proline, that interact with the VH domain [11]. The l2315 peptide, however, does not have this aa combination. We concluded that the observed retention might be a result of abrogated interaction between the CH1 and VH domains. All aa from the original Ig sequence were maintained when we introduced the epitope into CH1 loops of mouse IgG2b [9]. Several different insertion mutants were made, two in L2, one in L4, and one in L6. In addition, an L2 substitution mutant similar to that made in human IgG3, and a mutant (L46) with simultaneous peptide insertions in L4 and L6, were made. We found that all the L2 mutants were secreted, whether the peptide was introduced by replacement or insertion, suggesting that the contact between CH1 and V domains is not a strict requirement for secretion. The L4 and L6 mutants were also secreted, even though the loops had been elongated by 11 aa, but the L46 mutant was not.
2.2. Effect of flanking regions It has been shown that sequences of an Ag that are close to a presented peptide can influence its presenta-
111
tion [12– 16]. Studies where the flanking regions have been altered have shown either increased [12,14] or decreased [12–16] presentation efficiency, or no effect at all [13– 15]. There are several factors related to flanking regions that could affect peptide presentation efficiency. The proteolytic enzymes that cleave Ag have, at least to some extent, substrate specificity for certain sequences [17]. Furthermore, peptides eluted from MHC class II molecules are often longer than the minimum sequences required for binding [18,19], and the aa flanking a minimal peptide could affect the binding of the peptide to the MHC [20,21], or the T cell recognition of the MHC–peptide complex [20,22]. Finally, it has been suggested that antigenic peptides located in or nearby areas of a protein that are unstable will be degraded and presented most readily [23]. In all these experiments, the epitope is located in loops in an Ig domain. Thus, the secondary structure of the epitope and its surroundings would be expected to be quite similar. Even so, the Ig-derived aa flanking the epitope in the various positions are quite different. They might influence the efficiency by which the epitopes are excised from their new positions and presented to T cells. Therefore, the various peptide-Ab were tested in antigen presentation assays. The Ab were added to cultures of APC and specific T cells. T cell activation was measured as T cell proliferation and cytokine production. We found that peptide-Ab added exogenously to APC, resulted in presentation of the epitope and T cell activation. The presentation efficiency varied, however, with the following hierarchy of potency: hL4]hL6 = mL6 \ mL4 = mL2 (h = human IgG3, m= mouse IgG2b) (Fig. 4). Notably, hL2 could not be tested in this assay, since it is not secreted. However, MHC class II+ cells transfected with the hL2 heavy chain were found to present the peptide to T cells, even though the
Fig. 3. In CH1 of human IgG3, the l2315 epitope have been introduced so as to replace the Ig derived aa of the L2 (7aa), L4 (7aa), and L6 (4aa) loops. In addition, epitopes from p21ras, HA, OVA and HEL have been introduced by similar replacements into the L6 loop. In CH1 of mouse IgG2b, the l2315 epitope has been introduced into the Ig derived loops without removing any Ig derived aa. The epitope has been inserted at two different position in the L2 loop, and in one position in each of the L4 and L6 loops.
112
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
Fig. 4. Comparison of different loops as sites for the l2315 epitope. Comparison of the L4 and L6 mutants of mouse IgG2b and human IgG3 peptide-Ab. Symbols: , mouse L4 mutant; 2, mouse L6 mutant; , human L4 mutant; and ", human L6 mutant. Reprinted from Journal of Immunological Methods, volume 245, Eidem et al., Recombinant antibodies as carrier proteins for sub-unit vaccines: influence of mode of fusion on protein production and T-cell activation, pp. 119 – 131, copyright 2000, with permission from Elsevier Science.
heavy chain was retained intracellularly. Thus, hL2 could be processed as an endogenous Ag in the secretory pathway [24–26]. Even though presentation efficiency varied, no clear pattern explaining these differences was apparent from the flanking regions.
2.3. Insertion of epitopes of different lengths and secondary structures For vaccination purposes, it is of importance to be able to introduce a variety of T cell epitopes derived from the same or different Ag. Previous studies have shown that many different T and B cell epitopes can be introduced into the CDR loops of V domains [27–35], we tested three epitopes in addition to the l2315 already described in a loop in CH1. The new T cell epitopes were: aa 323–339 from OVA [36], aa 46– 61 from HEL [37], and aa 110– 120 from influenza virus (HA) [38]. For these studies, the L6 loop of CH1 of human IgG3 was used, and all three epitopes were introduced so as to replace the loop (Fig. 3). We found that all the resulting peptide-Ab were secreted. This demonstrates that many different peptides can be introduced into a loop of CH1 without disrupting the Ab structure required for secretion. First, epitopes of different lengths (17, 16 and 11 aa, respectively) can be used. Notably, additional unpublished results have shown that p21ras peptides of upto 25 aa can be introduced without interrupting Ab secretion (Rasmussen et al., unpublished). These epitopes were introduced by a replacement mutation where four aa of the L6 loop were removed. Thus, we have shown that loop elongations of upto 21 aa can be introduced in L6 of CH1. Similar results were obtained for epitopes in CDR2 and CDR3 loops of V regions, where loop elongations of up to 17 aa have been accepted [30]. Second, epitopes of differ-
ent secondary structures may be introduced into loops. When in its original position, the 91– 101 (l2315) epitope is found as a loop in the CDR3 region of the l2315 light chain. It is therefore likely to have a conformation that facilitates formation of a loop in the CH1 domain. In contrast, the other epitopes employed have different conformations when in their original Ag (the structure of the HA epitope is not known, though). Whereas the p21ras epitope has a b-strand/loop/a-helix structure [39], the OVA epitope is mostly a b-strand [40], and the HEL epitope has three turns and two b-strands [41]. Even so, the b-strands of Ig domains build domains that accept these various sequences. Based on these results, we anticipate that the Ig C domain frameworks can accommodate peptides derived from many different secondary structures. This would also fit well with the results on the V region CDR loops, which have been used for introduction of many different epitopes [27– 35]. The peptide-Ab with HA, OVA and HEL epitopes were tested in antigen presentation assays. In all cases, the epitopes were found to be presented on MHC class II to specific CD4+ T cells [3]. In conclusion, the experiments described confirm two important requirements of our overall strategy: (1) different T cell epitopes can be introduced into loops of Ig C domains without disturbing Ab folding and secretion; and (2) the T cell epitopes can be excised from their new positions by APC and presented to T cells. Even so, we found that some loops are more permissive than others, both with regard to Ab folding (i.e. secretion) and presentation to T cells.
3. Addition of V regions with specificity for APC The peptide-Ab described above were specific for the hapten NIP. To investigate the effect of targeting, the Ab V regions were replaced by V regions with specificity for APC. These were cloned from B cell hybridomas producing Ab with two different specificities: IgD [2] and MHC class II (Lunde et al., submitted).
3.1. Targeting to IgD Previous studies have shown that targeting to IgD induces potent responses [42–44]. We chose V regions that confer specificity for the IgDa allotype only, and not IgDb [45]. Thus, normal mice or cells with the IgDb allotype could serve as negative controls, without the need for non-physiological manipulations, such as B cell depletion. The VH and VL genes from the Ig(5a)7.2 hybridoma were cloned using a PCR-based strategy developed in our laboratory that allowed amplification of complete rearranged V(D)J genes [46]. In a first PCR, upstream primers annealing to the leader se-
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
quences were used in combination with a downstream primer annealing to the C region. In this PCR, complete rearranged V(D)J genes flanked by leader and C region sequences were amplified. The PCR products were sequenced, and in the second PCR, the exact V genes were amplified using new primers annealing precisely to the ends of the V genes and introducing unique restriction enzyme sites. The VH and VL genes were then subcloned into expression vectors upstream of Cg3 and Ck genes, respectively. In principle, the procedure allows introduction of V region genes from any B cell hybridoma or phage display library, and makes possible construction of Troy-bodies with any chosen specificity. Assembled IgDa-specific Ab were isolated from cell supernatant of transfected NS0 cells. The recombinant Ab were tested and found to be specific for the IgDa allotype, and not the IgDb allotype as expected [2]. The first IgD-specific Ab tested had the l2315 epitope in the L6 loop of the CH1 domain of human IgG3. This Ab, aIgD.L6 (aIgD.L3 in the original publication), was compared to the corresponding, but NIP-specific, L6 Ab. The Ab were tested in antigen presentation assays in which titrated amounts of Ab were added to cultures of APC and specific T cells. We found that targeting to APC by use of aIgD.L6 in vitro resulted in a 1000-fold better presentation efficiency compared to the NIP-specific L6 control Ab (Fig. 5a) [2]. As expected, B cells were required for this increase, and also the presence of IgDa expression. However, when aIgD.L6 Fab fragments were tested, they were found to be less effective than complete aIgD.L6 [2]. Complete aIgD.L6 Ab was then tested for targeting to B cells in vivo. In these experiments, mice were injected i.v. with the aIgD.L6 or NIP-specific L6 Ab. After 112 h, their spleen cells
113
were isolated and tested for ability to present the l2315 epitope to T cells in vitro. Spleen cells from mice injected with aIgD.L6 were found to be 100 times better at presenting the epitope than spleen cells from mice injected with L6 (Fig. 5b) [2]. Notably, these experiments only addressed the question of whether or not the injected Ab were targeted to splenic APC in vivo. Since the T cell proliferation assays were done in vitro, processing of aIgD.L6 and presentation of the l2315 epitope could have occurred during these 2 days of in vitro culture. To address the question of in vivo T cell proliferation, aIgD.L6 or L6 was injected i.v. into T cell receptor (TCR)-transgenic mice, the T cells of which are specific for the l2315 epitope. The mice received bromodeoxyuridine (BrdU), and T cell proliferation was measured as BrdU incorporation. We found that aIgD.L6 was 20-fold more efficient than L6 at inducing in vivo T cell proliferation (Fig. 5c) [2]. These initial experiments were done using Ab with the l2315 epitope. However, we have also made IgDspecific Ab with the OVA, HA and HEL epitopes in the L6 loop of CH1 [3]. They were also 1000-fold more efficient than the corresponding peptide-Ab in in vitro experiments. Thus, the targeting strategy works equally well with T cell epitopes other than l2315. As the amount of in vivo data in the literature is growing, it becomes increasingly clear that targeting to IgD in vivo may well give heterogeneous results. Not only T cell activation [42,47,48], but also T cell non-responsiveness [49,50] or altered T cell responses [51,52], have been observed. Regardless of targeting strategy, however, a major disadvantage of using IgD as target is that the B cells involved may become activated ([42], Lunde et al.,
Fig. 5. Targeting to IgD by use of the Troy-body that has the l2315 epitope in the L6 loop of human IgG3. (A) In vitro targeting. The IgD-specific Troy-body and control Ab were titrated into cultures of irradiated BALB/c splenocytes (as a source of APC) and lymph node cells from TCR-transgenic mice (as a source of l2315 specific T cells). Antigen presentation was measured as T cell proliferation (Thd incorporation) and is shown as a function of Ab concentration in the cultures; (B) In vivo targeting. Mice were injected intravenously with the IgD-specific Troy-body or the NIP-specific control Ab. After 112 h, the spleen cells were isolated and tested for their ability to stimulate T cells in vitro. T cell proliferation is shown as a function of serum human IgG3 112 h after injection; (C) In vivo T cell proliferation. TCR-transgenic mice were injected with the IgD-specific Troy-body and the NIP-specific control Ab, and given BrdU for a 5 day period. On day 5, spleen cells were analyzed by flow cytometry to reveal proliferating T cells. The percentage of proliferating (BrdU+), l2315-specific (GB113+), CD4+ T cells in the spleen is shown as a function of amount of Ab injected. Symbols: , IgD-specific Troy-body with l2315; ", NIP-specific peptide-Ab with l2315; , IgD-specific Ab without peptide. Reprinted from Nature Biotechnology, volume 17, Lunde et al., Antibodies engineered with IgD specificity efficiently deliver integrated T-cell epitopes for antigen presentation by B cells, pp. 670 – 675, copyright 1999, Nature Biotechnology.
114
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
unpublished observations). Since all B cells expressing IgD (irrespective of specificity) can be activated by the IgD-specific Ab, the selective activation of only Ag-specific B cells is lost. Consequently, the potential for inducing autoimmunity is great, and the IgD-specific Ab should therefore be viewed as tools for testing the strategy, and not as a suggestion for a future vaccine.
3.2. Targeting to MHC class II MHC class II molecules are good targets for several reasons: first, class II molecules have, like IgD, proven to be efficient in previous targeting studies [43,44,53,54]. Second, use of class II permits targeting to all the different types of APC, allowing for comparative studies. We have compared the effect of targeting to B cells, DC and MF, respectively, and for this purpose, we have used the Ab with the l2315 epitope. We evaluated the different APC both by use of in vitro cell lines, and by use of highly purified ex vivo splenic APC, isolated by cell sorting. These experiments showed that all three types of APC could be used as target cells for MHC class II-specific Ab, and that the difference between the targeted and non-targeted Ab was approximately the same for all three cell types (Lunde et al., submitted).
3.3. Targeting to IgD or MHC class II compared to presentation of synthetic peptides or antigenic proteins The various IgD- and MHC class II-specific Troybodies have been compared not only to their respective corresponding peptide-Ab, but also to the synthetic peptides (l2315, HA and HEL) or whole antigenic proteins (l2315 and OVA). When compared on a molar basis, the targeted Ab were normally 10 000-fold better at activating T cells in vitro than their respective synthetic peptides ([3] and Lunde et al., submitted). An in vivo comparison has not been done, but we expect the difference to be even more pronounced in vivo, since Ab are relatively stable molecules compared to synthetic peptides, which are rather unstable in vivo [5,6]. Whereas the synthetic peptides tended to be somewhat less efficient than the NIP-specific peptide-Ab, whole antigenic proteins induced T cell responses at approximately the same Ag dose as the NIP-specific Ab ([3,55] Lunde et al., submitted). Thus, on a molar basis, Troy-bodies were 1000 times more efficient than whole antigenic proteins.
4. Conclusions We have developed a targeting strategy based on recombinant Ab called Troy-bodies. The Troy-bodies have C regions containing T cell epitopes and V regions
with specificity for APC. When they bind to their target APC, the T cell epitopes are released by antigen processing and presented on MHC class II molecules to CD4+ T cells. The T cell epitopes were introduced into loops of the CH1 domain, and so far, three different loops (L2, L4, L6) have been tested. These three loops correspond to the V region CDR loops. Both human IgG3 and mouse IgG2b were used, and they were first loaded with an epitope from the l2315 light chain. Two different strategies for loop introduction were employed. In the replacement mutants, the aa of the T cell epitope replaced the 4–7 aa of the Ig loop. In contrast, the insertion mutants had the aa of T cell epitope inserted into the Ig loop without removal of any Ig derived aa. In most cases, the resulting recombinant peptide-Ab appeared to be correctly folded. An exception was the L2 loop of the CH1 domain of human IgG3, which seems to be less permissive than the others. When the peptide-Ab were added to APC, they were processed and presented to specific T cells. Even the intracellularly retained L2 mutant of human IgG3 could be presented if transfected into class II+ cells. Thus, all positions seem to be permissive with respect to antigen processing and presentation. The presentation efficiency however, varied between the different mutants. One loop, the L6 loop of human IgG3, was chosen for further studies. Three more T cell epitopes (derived from HA, HEL and OVA) were introduced into this loop. In their original positions in the native proteins, the epitopes are folded into different secondary structures. Even so, they were all accepted by the CH1 domain without disruption of Ab folding and secretion. Furthermore, all three epitopes were excised and presented to T cells. To produce APC-specific Ab, V region genes were cloned by a PCR-based strategy from hybridomas producing Ab of relevant specificities. These V regions were added to C regions with epitopes. So far, two different specificities have been tested, IgD and MHC class II (I-E). In both cases, a 1000-fold increase in in vitro presentation efficiency was obtained for the APC specific peptide Ab (the Troy-body) compared to the NIP-specific peptide-Ab. When Troy-bodies were targeted to APC in vivo after injection into mice, splenic APC were 100-fold better at activating T cells in vitro than those isolated from mice injected with control Ab. The class II-specific Troy-bodies were used to target all three types of professional APC (B cells, MF and DC). All three cell types worked equally well. In addition, one of the APC types, the B cells, were tested with both IgD- and class II-specific Troy-bodies. The effect appeared to be similar ( × 1000) for Troy-bodies of the two specificities. An overview of the results is given in Table 1.
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
115
Table 1 Summary Antigen
Epitope
Carrier
Specificities
Secretion
T cells
T cell activation
l2315
aa 91–101
IgG3 IgG2b
NIP IgD I-Ea
Yes
1. T cell clones 4B2A1 and 7A10B2 2. l2315-specific TcR TG mice (B. Bogen)
Yes (in vitro and in vivo)
P21ras
aa 1–25 G12V, G12R, G13D aa 5–21 G12V, G12R, G13D
IgG3
NIP HLA-DP
Yes
T cell clones (G. Gaudernack)
N.D.
Hen egg lysozyme (HEL)
aa 46–61
IgG3
NIP IgD I-Ea
Yes
T cell hybridoma 3A9 (R. Germain)
Yes (in vitro)
Hemagglutinin (HA)
aa 110–120
IgG3
NIP IgD I-Ea
Yes
T cells from HA-specific TCR TG mice (K. Karjalainen)
Yes (in vitro)
Ovalbumin (OVA)
aa 323–339
IgG3
NIP IgD I-Ea
Yes
1. T cell hybridoma DO11.10 2. DO11.10 TcR TG mice (K. Murphy)
Yes (in vitro)
5. Future perspectives Since B cells are thought to be the least effective APC at activating naive T cells, it is a goal to develop Troy-bodies that specifically target other subsets of APC. An important advantage of using Ig C domain loops for introduction of T cell epitopes is that an Ab molecule has many different C domains, each having several loops. More specifically, each IgG heavy chain has three C domains whereas the light chain has one, and since each domain has six loops, an IgG molecule could potentially harbor 24 different peptides in its C region. In addition, the ‘lower’ loops of the V domains could possibly also be used. Thus, there are many loops that can be used for epitope insertion, and a multi epitope Ab may be created. The difference in presentation efficiency for the various positions point to a role of flanking regions. A more uniform presentation may possibly be obtained by introducing minimal epitopes with flanking regions that promote processing. Another advantage of using Ig loops for peptide grafting is that it allows production of Ig (Troy-body) variants, taking advantage of the growing knowledge on Ab-derived fragments. The Troy-body variants could then be compared in order to find the best alternative for an optimal targeting effect in vivo. Notably, the targeting strategy could be used for goals other than the activation of CD4+ T cells. For example, surface molecules that can lead to presenta-
tion of extracellular Ag on MHC class I have recently been identified on both B cells [56] and DC [57]. These could be targeted with Troy-bodies containing MHC class I-restricted peptides, hopefully leading to epitope presentation on class I and activation of CD8+ T cells. If so, Troy-bodies could be used as vaccines for diseases in which CD8+ T cell responses are cruical. Finally, for conditions such as autoimmune diseases, Troy-bodies could be used to inhibit T cell activation. For e.g. by inserting antagonist peptides [58], or directing Troy-bodies to tolerizing APC [59], T cell anergy or tolerization could possibly be obtained. Another possibility would be to target Troy-bodies to thymic epithelium or DC, thereby affecting positive or negative selection of thymocytes.
Acknowledgements The work described was supported by the Norwegian Cancer Society and the Research Council of Norway.
References [1] Lunde E, Bogen B, Sandlie I. Mol Immunol 1997;34:1167 – 76. [2] Lunde E, Munthe L, Vabø A, Sandlie I, Bogen B. Nat Biotechnol 1999;17:670 – 5. [3] Rasmussen IB, Lunde E, Michaelsen TE, Bogen B, Sandlie I. Proc Natl Acad Sci USA 2001; in press
116
E. Lunde et al. / Biomolecular Engineering 18 (2001) 109–116
[4] Lesk, Tramontano. Antibody Engineering, A Practical Guide. New York: W.H. Freeman and Co., 1992. pp. 1 – 38. [5] Lee VH. Crit RevTher Drug Carrier Syst 1988;5:69 –97. [6] Powell MF, Grey H, Gaeta F, Sette A, Colon S. J Pharm Sci 1992;81:731 – 5. [7] Neuberger MS, Williams GT, Fox RO. Nature 1984;312:604 – 8. [8] Bogen B, Lambris JD. EMBO J 1989;8:1947 –52. [9] Eidem JK, Rasmussen IB, Lunde E, Gregers TF, Rees AR, Bogen B, Sandlie I. J Immunol Methods 2000;245:119 –31. [10] Lippincott-Schwartz J, Bonifacino JS, Yuan LC, Klausner RD. Cell 1988;54:209 – 20. [11] Lesk AM, Chothia C. Nature 1988;335:188 –90. [12] Vacchio MS, Berzofsky JA, Krzych U, Smith JA, Hodes RJ, Finnegan A. J Immunol 1989;143:2814 –9. [13] Lowenadler B, Lycke N, Svanholm C, Svennerholm AM, Krook K, Gidlund M. Mol Immunol 1992;29:1185 – 90. [14] Eisenlohr LC, Yewdell JW, Bennink JR. J Exp Med 1992;175:481 – 7. [15] Leclerc C, Martineau P, Charbit A, Lo-Man R, Deriaud E, Hofnung M. Mol Immunol 1993;30:1561 –72. [16] Janssen R, Wauben M, van der Zee R, de Gast M, Tommassen J. Int Immunol 1994;6:1187 –93. [17] Riese RJ, Chapman HA. Curr Opin Immunol 2000;12:107 – 13. [18] Rudensky AY, Preston-Hurlburt P, Hong SC, Barlow A, Janeway CA Jr. Nature 1991;353:622 – 7. [19] Chicz RM, Urban RG, Lane WS, Gorga JC, Stern LJ, Vignali DA, Strominger JL. Nature 1992;358:764 –8. [20] Berzofsky JA, Brett SJ, Streicher HZ, Takahashi H. Immunol Rev 1988;106:5 –31. [21] Mukherjee S, Ghosh RN, Maxfield FR. Physiol Rev 1997;77:759 – 803. [22] Wauben MH, van der Zee R, Joosten I, Boog CJ, van Dijk AM, Holewijn MC, Meloen RH, Van Eden W. J Immunol 1993;150:5722 – 30. [23] Landry SJ. Immunol Today 1997;18:527 – 32. [24] Weiss S, Bogen B. Cell 1991;64:767 –76. [25] Brooks A, Hartley S, Kjer-Nielsen L, Perera J, Goodnow CC, Basten A, McCluskey J. Proc Natl Acad Sci USA 1991;88:3290 – 4. [26] Kittlesen DJ, Brown LR, Braciale VL, Sambrook JP, Gething MJ, Braciale TJ. J Exp Med 1993;177:1021 – 30. [27] Billetta R, Hollingdale MR, Zanetti M. Proc Natl Acad Sci USA 1991;88:4713 – 7. [28] Zaghouani H, Krystal M, Kuzu H, Moran T, Shah H, Kuzu Y, Schulman J, Bona C. J Immunol 1992;148:3604 – 9. [29] Zaghouani H, Steinman R, Nonacs R, Shah H, Gerhard W, Bona C. Science 1993;259:224 –7. [30] Billetta R, Filaci G, Zanetti M. Eur J Immunol 1995;25:776 – 83. [31] Cook J, Barber BH. Vaccine 1995;13:1770 – 8.
[32] Zaghouani H, Anderson SA, Sperber KE, Daian C, Kennedy RC, Mayer L, Bona CA. Proc Natl Acad Sci USA 1995;92:631 – 5. [33] Brumeanu TD, Bot A, Bona CA, Dehazya P, Wolf I, Zaghouani H. Immunotechnology 1996;2:85 – 95. [34] Xiong S, Gerloni M, Zanetti M. Nat Biotechnol 1997;15:882 –6. [35] Legge KL, Min B, Cestra AE, Pack CD, Zaghouani H. J Immunol 1998;161:106 – 11. [36] Sette A, Buus S, Colon S, Smith JA, Miles C, Grey HM. Nature 1987;328:395 – 9. [37] Allen PM, Strydom DJ, Unanue ER. Proc Natl Acad Sci USA 1984;81:2489 – 93. [38] Hackett CJ, Hurwitz JL, Dietzschold B, Gerhard W. J Immunol 1985;135:1391 – 4. [39] Krengel U, Schlichting L, Scherer A, Schumann R, Frech M, John J, Kabsch W, Pai EF, Wittinghofer A. Cell 1990;62:539 – 48. [40] Stein PE, Leslie AG, Finch JT, Carrell RW. J Mol Biol 1991;221:941 – 59. [41] Maenaka K, Matsushima M, Song H, Sunada F, Watanabe K, Kumagai I. J Mol Biol 1995;247:281 – 93. [42] Finkelman FD, Scher I, Mond JJ, Kung JT, Metcalf ES. J Immunol 1982;129:629 – 37. [43] Casten LA, Pierce SK. J Immunol 1988;140:404 – 10. [44] Snider DP, Segal DM. J Immunol 1987;139:1609 – 16. [45] Oi VT, Herzenberg LA. Mol Immunol 1979;16:1005 – 17. [46] Norderhaug L, Olafsen T, Michaelsen TE, Sandlie I. J Immunol Methods 1997;204:77 – 87. [47] Finkelman FD, Scher I, Mond JJ, Kessler S, Kung JT, Metcalf ES. J Immunol 1982;129:638 – 46. [48] Lees A, Morris SC, Thyphronitis G, Holmes JM, Inman JK, Finkelman FD. J Immunol 1990;145:3594 – 600. [49] Eynon EE, Parker DC. J Exp Med 1992;175:131 – 8. [50] Snider DP, Kaubisch A, Segal DM. J Exp Med 1990;171:1957 – 63. [51] Day MJ, Tse AG, Puklavec M, Simmonds SJ, Mason DW. J Exp Med 1992;175:655 – 9. [52] Saoudi A, Simmonds S, Huitinga I, Mason D. J Exp Med 1995;182:335 – 44. [53] Carayanniotis G, Barber BH. Nature 1987;327:59 – 61. [54] Berg SF, Mjaaland S, Fossum S. Eur J Immunol 1994;24:1262 – 8. [55] Bogen B, Malissen B, Haas W. Eur J Immunol 1986;16:1373 –8. [56] Ke Y, Kapp JA. J Exp Med 1996;184:1179 – 84. [57] Regnault A, Lankar D, Lacabanne V, Rodriguez A, Thery C, Rescigno M, Saito T, Verbeek S, Bonnerot C, Ricciardi-Castagnoli P, Amigorena S. J Exp Med 1999;189:371 – 80. [58] Sloan-Lancaster J, Evavold BD, Allen PM. Nature 1993;363:156 – 9. [59] Croft M, Joseph SB, Miner KT. J Immunol 1997;159:3257 –65.