Autoantibodies induced by chimeric cytokine - HIV envelope glycoprotein immunogens

Abstract

Cytokines are often used as adjuvants to increase the immunogenicity of vaccines as they can improve the immune response and/or direct it into a desired direction. As an alternative to co-delivering antigens and cytokines separately they can be fused into a composite protein, with the advantage that both moieties act on the same immune cells. The HIV-1 envelope glycoprotein (Env) spike, located on the outside of virus particles and the only relevant protein for the induction of neutralizing antibodies (NAbs), is poorly immunogenic. The induction of anti-Env Abs can be improved by coupling Env proteins to co-stimulatory molecules such as a proliferation inducing ligand (APRIL). Here, we evaluated the immunogenicity of chimeric molecules containing uncleaved Env gp140 fused to the species-matched cytokines IL-21 or GM-CSF in rabbits and mice. Each cytokine was either fused to the C-terminus of Env or embedded within Env at the position of the variable loops 1 and 2 (V1V2). The cytokine components of the chimeric Env-GM-CSF and Env-IL-21 molecules were functional in vitro, but none of the Env-cytokine fusion proteins resulted in improved Ab responses in vivo. Both the Env-GM-CSF and the Env-IL-21 molecules induced strong anti-cytokine Ab responses, in both test species. These autoimmune responses were independent of the location of the cytokine in the chimeric Env molecules; in that they were induced by cytokines inserted within the V1V2 of Env or fused to its Ct. The induction of undesired autoimmune responses should be considered when using cytokines as co-stimulatory molecules in fusion proteins.

INTRODUCTION

For many decades, vaccine immunogenicity has been enhanced by formulation with adjuvants. Proteins derived from the host immune system such as cytokines, chemokines and other co-stimulatory molecules are being evaluated as adjuvants. Such natural immune stimulators can be co-delivered with the vaccine as recombinant proteins. Alternatively, the inclusion of various cytokine genes (including IL-2, IL-12, IL-28 or GM-CSF) has been shown to improve the immunogenicity of DNA vaccines against HIV-1, influenza, HBV, HCV and cancer (1,2). Cytokines can also be fused directly to vaccine antigens, Abs or self-antigens to make chimeric proteins (3–8). Using this method, immune cells that are targeted by the antigen of interest can be simultaneously activated by the incorporated cytokine, allowing selective augmentation of the most desired immune response. One potential disadvantage of using host proteins in this way is the potential for inducing autoimmune responses, but whether anti-cytokine responses are induced is often not studied in detail.

HIV-1 recombinant envelope glycoprotein (Env) protein vaccines invariably require adjuvants. Env, the only relevant immunogen for, and target of, neutralizing antibodies (NAbs), has many structural defenses against NAbs, including the presence of hypervariable domains, multiple shielding glycans and the exposure of decoy epitopes on non-functional Env forms. Together, these factors contribute to the lack of success of HIV-1 Env-based vaccines to date (9–22). In addition to the intrinsic properties that impede the induction of broadly active NAbs (bNAbs), the half-life of the overall anti-Env Ab response is very short, typically 30–60 days (23,24). One explanation may be poor induction of long-lived plasma cells. Of note is that the N-linked oligomannose glycans present on Env can actively suppress immune cell functions (25–28).

Attempts to improve the immunogenicity of HIV-1 Env include co-delivery with molecular adjuvants (1). Alternatively, Env proteins, in most cases monomeric gp120, have been fused directly to immune modulators, including IFN-γ, TNF-α, Flt-3 ligand, CTLA4, C3d and a proliferation inducing ligand (APRIL) to make chimeric molecules (6,29–33). We have shown that fusing Env trimers to co-stimulatory molecules such as CD40L and APRIL that facilitate B cell help can be beneficial to Env immunogenicity (5,6). Furthermore, replacing Env variable loops 1 and 2 (V1V2) with GM-CSF increased both anti-gp120 binding Ab titers and T-cell responses in mice (34,35), while replacing the same loops with IL-21 enhanced Ig secretion and the formation of human plasmablast-like B cells in vitro (36). Both GM-CSF and IL-21 cytokines have a good safety profile in clinical studies. For example, GM-CSF is being used as an adjuvant in several Phase I/II trials against a variety of tumors (37,38), and IL-21 has been tested against metastatic melanoma and renal cancer (39–41).

Here, we investigated whether trimeric HIV-1 Env proteins with GM-CSF or IL-21 cytokines incorporated into the V1V2 domain or fused to the C-terminus (Ct) would increase the induction of anti-Env Ab responses in vivo. We also assessed whether autoantibodies (autoAbs) were raised against the cytokine component. The outcome was that we saw at best only a modest improvement in anti-Env responses, accompanied by strong anti-cytokine responses that were independent of the design of the Env-cytokine chimeric protein.

MATERIALS and METHODS

Constructs

All constructs in this study were based on the codon-optimized stabilized HIV-1 JR-FL (subtype B) SOSIP.R6-IZ-H8 gp140 sequence (Env_wt) that has been described in detail elsewhere (10,20,42–44). Despite the presence of the SOSIP mutations, these trimers are not native-like as we recently published for BG505 SOSIP.664 gp140 (45–47). The constructs used here were based on the earlier generation JR-FL SOSIP.R6 and the addition of C-terminal protein domains such as a GCN4-based isoleucine zipper trimerization domain (IZ) or cytokine domains render them uncleavable (5,44), and uncleaved and cleaved gp140 trimers are antigenically and structurally different (48,49). Amino acid numbering is based on the HXB2 gp160 sequence, according to convention. A list of constructs used in this study is provided in Supplemental Table I. Rabbit IL-21 (rIL-21) was amplified from rabbit PBMCs, using the Expand PCR system according to the manufacturer’s instructions (Roche, Mannheim, Germany). The PCR was performed with sense and antisense primers (5’rIL-21 [5’-CATCTGTCTGATGGTCATCTTCTT-3’] and 3’rIL-21 [5’-TTCAAGGAGATGCTGATGAATC-3’]). The PCR products were sequenced. Codon-optimized genes encoding rabbit GM-CSF (rGM-CSF), mouse IL-21 (mIL-21) or rIL-21 flanked by HindIII and BmgBI restriction sites, were synthesized (Mr. Gene, Regensburg, Germany). The V1V2 domains of Env were substituted by the sequences coding for GM-CSF or IL-21 using the HindIII and BmgBI restriction sites to generate Env_rGM-CSF, Env_mIL-21 and Env_rIL-21, similar to generation of Env_hGM-CSF, Env_mGM-CSF and Env_hIL-21 that we described previously (35,36). Codon-optimized genes of human GM-CSF (hGM-CSF), rGM-CSF or rIL-21 including a linker sequence (GGCGGCCGCGGGGGAGGGGGGTCTGGAGGAGGC, which encodes the peptide sequence GGRGGGGSGGG) flanked by Not1 and Sfu1 restriction sites were synthesized (Mr. Gene) and cloned at the Ct of SOSIP.R6-IZ (Env_hGM-CSF-Ct, Env_rGM-CSF-Ct and Env_rIL-21-Ct). To generate the His-tagged-cytokine vectors, rabbit or mouse GM-CSF or IL-21 flanked by Not1 and Sfu1 restriction sites were cloned at the Ct of a His-tag (8×histidine). The sequences of all constructs were verified by sequencing.

Reagents

Monoclonal Abs (MAbs) were obtained as gifts, or purchased, from the following sources: Polymun Scientific (2G12); AIDS Research and Reference Reagent Program (ARRRP), Division of AIDS, NIAID, NIH (pooled polyclonal Ig from HIV-1-positive patient sera (HIV-Ig)); Dennis Burton (b12, PGT121); Bill Olson, Progenics Pharmaceuticals (CD4-IgG2, soluble CD4 (sCD4) and PA1); John Mascola and Peter Kwong (VRC01); James Robinson (48d) and Hermann Katinger through the ARRRP (2F5). Recombinant human IL-21 (rhIL-21) was obtained from PeproTech (London, UK) and recombinant human GM-CSF (rhGM-CSF) for TF-1 cell stimulation was from Schering-Plough (Brussels, Belgium).

Cells and transfections

293T cells were maintained in DMEM (Invitrogen, Breda, the Netherlands) supplemented with 10% heat inactivated FCS (HyClone, Perbio, Etten-Leur, the Netherlands), MEM nonessential amino acids (0.1 mM; Invitrogen) and penicillin/streptomycin (both at 100 U/ml). The cells were transiently transfected with plasmids expressing recombinant Env using the Lipofectamine reagent according to the manufacturer's instructions (PAA, Pasching, Austria). Env-containing supernatants were harvested 48 h after transfection and frozen in aliquots. TF-1 cells (50), a gift from Paul Coffer, were cultured in RPMI 1640 (Invitrogen) with 10% FCS, supplemented with 25 U/ml rhGM-CSF (Schering-Plough). Ig secretion by B cells was quantified as described previously (36).

SDS-PAGE, Native PAGE and western blotting

SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and western blotting were performed as previously described (44). Env was detected using MAb PA1 (0.2 µg/ml) and a 1:5000 diluted secondary HRP-labeled goat-anti-mouse IgG (Kirkegaard & Perry Laboratories, Maryland, USA) followed by the Western Lightning ECL system (PerkinElmer, Groningen, The Netherlands). Native PAGE was carried out as reported elsewhere (35).

Anti-Env ELISA

Anti-gp120 Ab titers were measured by D7324-capture ELISA as described previously (35,44,47). Reactivity of Env proteins with MAbs or receptor mimics were measured by anti-trimer ELISA as described previously using Ni-NTA HisSorb 96-well plates (Qiagen, Venlo, The Netherlands)(10,34–36,47). Equal input levels of proteins were verified by SDS-PAGE followed by western blot analysis. All ELISA data are representative of at least three independent experiments, each using Env proteins derived from independent transfections. Endpoint titers were calculated using GraphPad Prism (version 5.03) by determining the serum dilution at which the optical density (OD₄₅₀) was three times above the background OD₄₅₀ signal (without animal sera).

Anti-cytokine ELISA

The anti-cytokine ELISA was performed as previously described for the anti-trimer ELISA (10,34–36,47). The wells were coated with supernatants containing His-tagged cytokines (Supplemental Table I) or mock media (or Env_wt for comparative purposes) and the plates were washed twice with Tris-buffered saline (TBS). Heat inactivated (30 min, 56°C) rabbit or mouse sera were serially diluted in TBS containing 20% sheep serum (Biotrading, Mijdrecht, Netherlands) and 2% skim milk powder, and incubated for 2 h. The subsequent steps were as previously described for the anti-Env ELISA (44).

Immunoprecipitation assays

Immunoprecipitation assays were performed as previously described (5,6), using PA1, 2G12, b12, VRC01, CD4-IgG2, 2F5, or 48d at final concentrations of 4 µg/ml and, when appropriate, also sCD4 at 10 µg/ml. The immunoprecipitates were analyzed by SDS-PAGE followed by western blotting MAb PA1.

Neutralization assays

The TZM-bl reporter cell line was obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, National Institutes of Health (John C. Kappes, Xiaoyun Wu, and Tranzyme Inc. [Durham, NC]) (51,52). All sera were heat inactivated (30 min, 56°C) before use. Neutralization assays were performed using the tier 1A virus SF162 and the tier 2 virus JR-FL, which is the homologous strain to the vaccine antigen. Single cycle infection and inhibition experiments were performed as described previously, using 3-fold serially diluted sera, with each sample tested in duplicate against each virus (10,53). The percentage of neutralization was determined by measuring the reduction of the luciferase signal by each serum dilution compared to the values measured in the absence of serum (defined as 100%). The 50% (midpoint) neutralization titers were determined using GraphPad Prism (version 5.03).

GM-CSF activity assay

Myeloid leukemia TF-1 cells (5×10⁴) (50) were seeded in a 96-well plate before addition of serially diluted supernatants from mock, Env_wt, Env_hGM-CSF or Env_hGM-CSF-Ct transfected 293T cells. rhGM-CSF (50 U/ml) diluted in mock supernatant was used as positive control. Each sample was tested in triplicate wells in at least three independent experiments using proteins derived from separate transfections. The cells were harvested on day five and the percentage of living cells was determined by FACS flow cytometry.

Gene gun DNA immunizations of animals

Plasmids encoding Env proteins were amplified using DH5α cells and isolated using an EndoFree Plasmid Giga kit (Qiagen, Venlo, the Netherlands). New Zealand White rabbits (6 or 12 per group) were immunized at weeks 0, 2, 4 and 8 with 125 μg of endotoxin-free DNA via the abdominal dermis using gene gun technology (54). Blood samples were obtained at weeks 0, 2, 4, 6 and 8, with a final bleed at week 10. Outbred NMRI mice (4 or 5 per group) were immunized at weeks 0, 2, 4, and 6 with 20 μg of DNA in the abdominal dermis using the same gene gun method. Blood samples were obtained from weeks 0, 2, 4, and 6, with a final bleed at week 6 or 8. All protocols dealing with animal manipulations were in accordance with guidelines published by FELASA (Federation of European Animal Science Association) and GV-SOLAS (German Society of Laboratory Animal Science) and were reviewed by the Harlan, MFD Diagnostics, or Davids Biotechnologie Animal Care Committees, as appropriate.

RESULTS

Generation and characterization of Env_rGM-CSF and Env_rIL-21

We have previously described the insertion of various cytokines into the V1V2 domain of Env_wt (34–36). Here, we used the same approach with the cytokines GM-CSF and IL-21, using rabbit sequences, to generate the Env_rGM-CSF and Env_rIL-21 chimeric proteins for in vivo studies in rabbits (Fig. 1A). We note that during the course of these studies a better soluble mimic of the native Env spike was identified (BG505 SOSIP.664 gp140), but we did not use it here (45–47). Both chimeric Env_rGM-CSF and Env_rIL-21 proteins were expressed by transient transfection of 293T cells, Env_rGM-CSF expression was similar to Env_wt whereas Env_rIL-21 was produced less efficiently (see Supplemental Fig. 1A) (36). The antigenic structures of Env_rGM-CSF and Env_rIL-21 were analyzed using an anti-trimer ELISA and were found to be similar to those of Env_hGM-CSF, Env_mGM-CSF, Env_hIL-21 and Env_mIL-21 (Supplemental Fig. 1B) (34–36).

FIGURE 1.

Design and functionality of the Env_wt, Env_rGM-CSF and Env_rIL-21 fusion proteins (A) Cartoon representation of Env_wt, Env_rGM-CSF and Env_rIL-21. Ig production by human B cells stimulated with Env_wt, Env_rIL-21 or rhIL-21 in vitro. IgG, IgA and IgM secretion by pan-B cells derived from human PBMCs cultured with (B) CD40L/IL-10 or (C) CD40L/IL-4/IL-10. Data represent the fold change values compared to Env_wt in samples from at least five donors, with each donor’s sample tested in duplicate. For fold-change calculations, the background levels of IgG, IgA and IgM secretion in the presence of the stimulation cocktail but in the absence of test proteins were subtracted from the test values and the fold-change was calculated compared to the background-corrected response to Env_wt. *: significantly different (p<0.05); **: significantly different (p<0.01) (one-tailed, Wilcoxon matched pairs test).

To test whether the rIL-21 moiety embedded in the V1V2 domain was functional, we treated pan-B cells with Env_wt or the chimeric proteins and examined secretion of IgG, IgA and IgM in the presence of two maturation cocktails (CD40L/IL-10 and CD40L/IL-10/IL-4), as previously described for Env_hIL-21 (36). Env_wt induced low levels of Ig secretion from human B cells, consistent with previous reports that this protein can activate B cells via glycan-dependent interactions with lectin receptors (36,55). A higher level of Ig secretion was induced by Env_rIL-21, hence rIL-21 is cross-reactive with the human IL-21 receptor and the IL-21 domain in Env_rIL-21 is functional. As previously seen with Env_hIL-21, the magnitude of Ig secretion induced by Env_rIL-21 was donor-dependent (median fold changes were 3.45, 2.54, 3.81 for IgG, IgA, IgM respectively in CD40L/IL10 and 1.68, 2.11, 2.96 for IgG, IgA, IgM respectively in CD40L/IL4/IL10 conditions) (Fig. 1B & C). We also wished to examine the function of rGM-CSF embedded in the V1V2 domain of Env. Unfortunately, the TF-1 cell based GM-CSF activity assay (50) that we used previously to show that hGM-CSF embedded in V1V2 was functional (35), was not sensitive to rGM-CSF, probably because the human GM-CSF receptor is not cross-reactive with rGM-CSF, and other reagents to validate whether rGM-CSF is functional are not available.

Rabbit immunizations with Env_wt, Env_rGM-CSF and Env_rIL-21

We immunized rabbits with DNA encoding the Env_wt (n=6) or Env_rGM-CSF (n=6) proteins using a gene gun. In a separate, but similar experiment we compared Env_wt (n=12) with Env_rIL-21 (n=12). The rabbits were immunized at weeks 0, 2, 4 and 8 as described previously (6), and bled at weeks 0, 2, 4, 6, 8 and 10 for serology studies. When the anti-gp120 mid-point IgG titers in week-10 sera were measured by ELISA, the response to Env_rGM-CSF and Env_rIL-21 were slightly lower than to Env_wt. Hence, the inserted cytokines had no beneficial effect on the anti-Env response (Fig. 2A & B). The neutralization capacity of the week-10 sera was tested in a single-cycle assay using Env-pseudotyped viruses and TZM-bl reporter cells (53). None of the test sera neutralized the resistant (tier-2) virus JR-FL (56,57), and the responses against the highly sensitive (tier-1A) virus SF162 were weak, inconsistent, and did not differ among the test groups (Tables I & II). Thus, there was no evidence that the embedded cytokines conferred an immunological benefit on the humoral response.

FIGURE 2.

Immunogenicity of Env_rGM-CSF or Env_rIL-21 in rabbits. (A, B) Endpoint anti-gp120 IgG titers for Env_rGM-CSF (n=6), Env_rIL-21 (n=12) or control-immunized rabbits at week 10 (terminal bleed), as determined by ELISA. Titration curves of anti-Env and anti-rGM-CSF Abs from rabbit immunized with Env_wt (n=6) (C) and Env_rGM-CSF (n=6) (D). Titration curves of anti-Env and anti-rIL-21 Abs from rabbit immunized with Env_wt (n=6) (E) and Env_rIL-21 (n=6) (F). Black lines represent the binding of anti-cytokine antibodies and black lines represent the anti-Env. (E & F) Anti-Env and anti-cytokine Ab endpoint titers of immunized rabbits at week 10. Control plates lacked the added cytokines, and the resulting signals were negligible (data not shown). **: significantly different (p<0.01) (one-tailed, Mann-Whitney test).

Table I.

50% neutralization titers induced in rabbits by Env_wt and Env_rGM-CSF against viruses SF162 and JR-FL.

Immunogen	Animal	SF162	JR-FL
Envwt	1	<30	<30
	2	189	<30
	3	329	<30
	4	156	<30
	5	138	38
	6	358	<30
EnvrGM-CSF	1	257	<30
	2	90	<30
	3	96	<30
	4	282	<30
	5	120	<30
	6	84	<30

Increasing titers are represented by darker shade of gray.

Table II.

50% neutralization titers induced in rabbits by Env_wt and Env_rIL-21 against viruses SF162 and JR-FL.

Immunogen	Animal	SF162	JR-FL
Envwt	1	72	<30
	2	141	<30
	3	827	<30
	4	1252	<30
	5	<30	<30
	6	93	<30
	7	169	<30
	8	37	<30
	9	<30	<30
	10	<30	<30
	11	548	<30
	12	152	<30
EnvrIL-21	1	<30	<30
	2	<30	<30
	3	<30	<30
	4	107	<30
	5	<30	<30
	6	<30	<30
	7	<30	<30
	8	33	<30
	9	<30	<30
	10	<30	<30
	11	2400	<30
	12	<30	<30

Increasing titers are represented by darker shade of gray

Induction of anti-cytokine responses in rabbits immunized with Env_rGM-CSF or Env_rIL-21

One potential concern with using self-molecules as adjuvants is the potential of inducing autoAbs. We therefore quantified the rabbit anti-cytokine Ab responses induced by the above chimeric Env proteins, using an ELISA in which His-tagged versions of Env, GM-CSF or IL-21 were immobilized on Ni-NTA coated plates (see Materials & Methods). Rabbits immunized with Env_rGM-CSF induced Abs against rGM-CSF but, as expected, the sera from the Env_wt group lacked these Abs (Fig. 2C & D). A similar result was obtained for the Env_rIL-21-immunized rabbits, which induced Abs to the IL-21 cytokine insert (Fig. 2E & F). We also determined the Ab responses to the Env components using anti-trimer ELISA, but again found no improvement in anti-Env Ab levels in the Env-chimera immunized rabbits, compared to Env_wt (Fig. 2G & H). The anti-GM-CSF responses were comparable to the anti-Env responses, but those against anti-IL-21 were ~10-fold lower compared to the anti-Env responses (and ~10-fold lower compared to the anti-GM-CSF responses). Overall, we conclude that the embedded rIL-21 or rGM-CSF domains were strongly immunogenic in that they induced autoAbs.

Induction anti-cytokine responses to Env_mGM-CSF and Env_mIL-21 molecules in mice

To test whether the anti-cytokine responses were unique to rabbits, we immunized mice with DNA encoding Env_wt and Env_mIL-21, the latter generated by the same strategy used for the human and rabbit Env-IL-21 chimeras (35,36). In addition, we tested sera from a previous mouse immunization experiment that had compared the immunogenicity of Env_wt and Env_mGM-CSF (35). The expression, antigenicity and cytokine activity of Env_mGM-CSF and Env_mIL-21 were similar to their human and rabbit counterparts (data not shown and (35,36)).

The mice were immunized with Env_wt and Env_mIL-21 DNA at weeks 0, 2 and 4, and sacrificed at week 6. The week-6 sera from mice immunized with Env_wt (n=4) and Env_mIL-21 (n=4), and the historic week-8 sera from mice immunized with Env_wt (n=5) and Env_mGM-CSF (n=5) (35), were tested for the presence of anti-mIL-21 and anti-mGM-CSF autoAbs in similar titration experiments as those shown in Fig. 2C-F (Fig. 3A & B). We note that the anti-Env and anti-cytokine Ab levels were lower in the mice than in the rabbits. This probably reflects differences in study regimen and duration. There were no anti-cytokine responses to mGM-CSF or mIL-21 in any of the Env_wt recipients, whereas anti-mGM-CSF and anti-mIL-21 autoAbs were detectable in the Env_mGM-CSF and Env_mIL-21 immunized mice, respectively (Fig. 3A & B). The anti-gp120 and anti-trimer IgG Ab titers were significantly (p=0.027) higher in the Env_mGM-CSF recipients than in Env_wt (Fig. 3A and (35)), but this was not the case for Env_mIL-21 recipients (Fig. 3B & data not shown). Thus, both mice and rabbits generate autoAbs against GM-CSF and IL-21 when these cytokines are embedded within the V1V2 domain of trimeric HIV-1 Env proteins.

FIGURE 3.

Mouse immunizations with DNA encoding Env_mGM-CSF or Env_mIL-21. (A) Endpoint titers of anti-mGM-CSF and anti-Env trimer Abs in mice immunized with Env_wt (n=5) or Env_mGM-CSF (n=5) at week 8 (terminal bleed) as determined by ELISA. (B) Endpoint titers of anti-mIL-21 and anti-Env trimer Abs in mice immunized with Env_wt (n=4) or Env_mIL-21 (n=4) at week 6 (terminal bleed). *: significantly different (p<0.05) (one-tailed, Mann-Whitney test).

Generation and functionality of chimeric HIV-1 Env molecules with GM-CSF or IL-21 fused at the C-terminus

We reasoned that the embedding of a self-molecule (i.e. the cytokine) within the body of a different molecule (i.e. Env) might exacerbate the generation of autoAbs. To investigate whether the location of the cytokine molecule within the chimera is a relevant factor, we fused the rabbit GM-CSF and IL-21 molecules to the Ct of the Env trimer, to generate the chimeric Env_rGM-CSF-Ct and Env_rIL-21-Ct proteins (Fig. 4A). These constructs are analogous to the Env_APRIL and Env_CD40L molecules we have previously studied (5,6). To minimize interactions between the Env and the cytokine, we inserted a long spacer between them (amino acid sequence GGRGGGGSGGG). The added cytokine sequences did not affect protein expression, and both Env_rGM-CSF-Ct and Env_rIL-21-Ct formed predominantly trimers, similarly to Env_wt (Supplemental Fig. 2A). Because these constructs do not have a His-tag and cannot be used in Ni-NTA ELISA (Supplemental Table I), the antigenic structure and folding of Env_rGM-CSF-Ct and Env_rIL-21-Ct was evaluated by performing immunoprecipitation experiments, with conformational NAbs or a CD4 receptor mimic as described previously (6). Env antigenicity was not influenced by the presence of cytokine molecules at the Ct (Supplemental Fig. 2B-D). Furthermore, the cytokine moieties retained their activities, as Env_hGM-CSF-Ct was active on TF-1 cells that require GM-CSF for proliferation (Supplemental Fig. 1E)(50).

FIGURE 4.

Design and immunogenicity of the Env_rGM-CSF-Ct and Env_rIL-21-Ct fusion proteins (A) Cartoon representation of Env_wt, Env_rGM-CSF-Ct and Env_rIL-21-Ct. 6 rabbits per groups were immunized with Env_wt, Env_rGM-CSF-Ct and Env_rIL-21-Ct. (B) Endpoint of anti-rGM-CSF and anti-IL-21 Abs in rabbits immunized with Env_wt, Env_rGM-CSF-Ct and Env_rIL-21-Ct at week 10 (terminal bleed). (C) Endpoint anti-Env IgG titers of Env_rGM-CSF-Ct and Env_rIL-21-Ct immunizations at week 10 determined by ELISA. **: significantly different (p<0.01) (one-tailed, Mann-Whitney test).

Induction of anti-cytokine responses by the Env_rGM-CSF-Ct and Env_rIL-21-Ct chimeras

Groups of 6 rabbits were immunized with DNA encoding Env_wt, Env_rGM-CSF-Ct or Env_rIL-21-Ct using the same protocol as before. Anti-rGM-CSF and anti-rIL-21 autoAbs were induced in the Env_rGM-CSF-Ct and Env_rIL-21-Ct recipients, but not in rabbits immunized with Env_wt (Fig. 4B). The anti-IL-21 endpoint titers were ~15-fold lower compared to the anti-GM-CSF titers. The anti-GM-CSF endpoint titers induced were slightly higher compared to the titers induced by GM-CSF embedded within the V1V2 of Env (compare Fig. 2G with Fig. 4B), whereas no such a trend was observed for IL-21 (Fig. 2H and 4B). There was no improvement in the anti-trimer or anti-gp120 IgG responses in Env_rGM-CSF-Ct and Env_rIL-21-Ct immunized rabbits, compared to Env_wt (Fig. 4C & data not shown). When the rabbit sera (week 10) were tested for SF162 and JR-FL virus neutralization, no differences were found between the three groups (data not shown). In summary, autoAbs were raised to GM-CSF and IL-21 irrespective of where the cytokine was located in the chimeric Env protein (i.e., within the V1V2 domain or at the Ct).

DISCUSSION

Adjuvants, including ones based on co-stimulatory self-molecules such as cytokines, can improve vaccine immunogenicity by various mechanisms, including by activating relevant immune cells. The use of co-stimulatory self-molecules may also help skew the immune response into the desired direction (i.e., humoral vs. cellular responses). However, a disadvantage of this approach is that undesired autoimmune responses may be generated against the self-molecule when it is presented in a non-natural context. Although the beneficial effects of including cytokines in vaccines, particularly DNA vaccines, are being studied extensively, whether autoimmunity arises is rarely investigated. Here, we studied, in both mice and rabbits, whether chimeric cytokine-Env molecules induced autoimmune responses to the cytokine moiety, and whether the location of the cytokine mattered. We found that substantial autoAb responses were raised against GM-CSF and IL-21, when the cytokine was either embedded in the body of the Env protein (i.e. within the V1V2 loop structure) or attached to the Ct (via a flexible linker sequence). We have not evaluated Ct fused cytokines in mice, but we would expect similar results. Whether induction of autoAbs accounts for the failure of the cytokines to boost the Ab responses to the Env remains to be determined. However, we note that Env_mGM-CSF did induce stronger anti-gp120 Ab and T cell responses in mice, indicating that the presence of anti-GM-CSF Abs does not necessarily abolish the adjuvant activity of the embedded cytokine (35).

AutoAbs against cytokines may have multiple adverse effects. One example involves some formulations of recombinant human erythropoietin used to treat renal anemia. Several patients developed more severe anemia, which was attributed to the unwanted induction of NAbs against erythropoietin (58). AutoAbs can be induced when recombinant IFN-α, IFN-β and GM-CSF are repeatedly administered (59–61). In a clinical trial neutropenic cancer patients were treated with rhGM-CSF. Although the patients developed anti-GM-CSF Abs no adverse affects or abnormal hematologic conditions were observed (62). Furthermore, in a rabbit immunization study, DNA co-immunization with murine GM-CSF enhanced the protective immunity to the given antigen although anti-GM-CSF Abs were detected (63). An autoAb response also arose in a subset of patients with metastatic malignant melanoma and renal cell carcinoma who were treated with rhIL-21, but the neutralizing anti-rhIL-21 Abs were not associated with any specific adverse effects (40). Abs to certain cytokines may be involved in various immune-inflammatory diseases such as rheumatoid arthritis, and also certain cancers (64–68), while anti-GM-CSF Abs have been correlated with the severity of pulmonary alveolar proteinosis (69). Abs targeting cytokines, including GM-CSF, IL-2, IL-6, IL-10, IFN-γ and IFN-α, can be detected in the sera of healthy individuals, implying that they are not necessarily harmful (59,70–74). In some cases these Abs have been suggested to have a protective role against some disease conditions (75,76), by beneficially influencing cytokine localization, half-life and activity (59,77,78). We did not observe cytokine-related adverse effects in the animals from this study.

We note that the soluble Env proteins used in this study are uncleaved gp140s that cannot adopt a native-like structure (5,6). The BG505 SOSIP.664 gp140 protein, which is a close antigenic and structural mimic of the native spike (45–47) was not available when this study was initiated. Cleaved and uncleaved soluble trimers are structurally and antigenically very dissimilar, and only cleaved trimers resemble the functional Env spikes found on virions (45–47,49). Whether fusing co-stimulatory molecules to cleaved, native-like SOSIP gp140 trimers would yield a different outcome remains to be determined. It is possible that the splayed-out nature of uncleaved trimers and their increased vulnerability to proteases (48,49) might exacerbate the induction of autoAbs against the cytokine moiety. The incorporation of cytokines in the V1V2 domain of cleaved native-like SOSIP trimers might be problematic because the wild type V1V2 domains form and stabilize the trimer apex (45,46).

In conclusion, we have investigated the immunogenicity of chimeric Env-GM-CSF and Env-IL-21 proteins. The inclusion of the cytokine moieties as intrinsic elements of the fusion proteins did not augment the immunogenicity of the Env component in rabbits or mice, even though the cytokines retained their biological activities. However, autoimmune responses were raised against the cytokines, which may limit or negate their adjuvant capabilities. In contrast, broadly similar chimeric Env molecules including the TNF superfamily members APRIL and CD40L did not induce anti-APRIL or anti-CD40L responses in rabbits (Isik and Sanders, manuscript in preparation). Hence the induction of autoimmune responses by cytokine-Env fusion proteins may not be inevitable, but should always be considered and assessed.