Promise and problems associated with the use of recombinant AAV for the delivery of anti-HIV antibodies

Abstract

Attempts to elicit antibodies with potent neutralizing activity against a broad range of human immunodeficiency virus (HIV) isolates have so far proven unsuccessful. Long-term delivery of monoclonal antibodies (mAbs) with such activity is a creative alternative that circumvents the need for an immune response and has the potential for creating a long-lasting sterilizing barrier against HIV. This approach is made possible by an incredible array of potent broadly neutralizing antibodies (bnAbs) that have been identified over the last several years. Recombinant adeno-associated virus (rAAV) vectors are ideally suited for long-term delivery for a variety of reasons. The only products made from rAAV are derived from the transgenes that are put into it; as long as those products are not viewed as foreign, expression from muscle tissue may continue for decades. Thus, use of rAAV to achieve long-term delivery of anti-HIV mAbs with potent neutralizing activity against a broad range of HIV-1 isolates is emerging as a promising concept for the prevention or treatment of HIV-1 infection in humans. Experiments in mice and monkeys that have demonstrated protective efficacy against AIDS virus infection have raised hopes for the promise of this approach. However, all published experiments in monkeys have encountered unwanted immune responses to the AAV-delivered antibody, and these immune responses appear to limit the levels of delivered antibody that can be achieved. In this review, we highlight the promise of rAAV-mediated antibody delivery for the prevention or treatment of HIV infection in humans, but we also discuss the obstacles that will need to be understood and solved in order for the promise of this approach to be realized.

Since the first reported cases of acquired immunodeficiency syndrome (AIDS) in 1981 (ref. 1) and the identification of the AIDS-causing virus in 1983 (ref. 2), it is estimated that more than 40 million people have died from human immunodeficiency virus (HIV) infection.^3,4 About 35 years have elapsed since the first documented HIV-1 infections and no substantial progress has been made in developing a vaccine that could effectively protect against HIV infection in the vast majority of people.^5–8 Similarly, with the single exception of the “Berlin patient”,^9–11 eradication of HIV from infected individuals has also not been achievable.¹² Although the development of potent antiretroviral drugs has made it possible to vastly extend the life expectancy of HIV-infected individuals, anti-HIV drugs do not cure virus infection.^12–20 As of 2014, it was estimated that almost 37 million people were living with HIV globally, with a continuing new infection rate of 2 million per year.²¹

There are good reasons for believing that development of an effective vaccine against HIV-1 is going to be a very difficult task.^22,23 The predicted difficulties have more or less been borne out by vaccine trials in monkeys and in humans.^6–8,24 Of the six large-scale, placebo-controlled human efficacy trials of HIV vaccines, three showed no protection against acquisition and two actually showed enhanced acquisition of HIV-1 infection in the vaccine recipient.^25–37 Only one of the six vaccine trials, termed RV144 (ref. 38), appeared to show some protective effects against acquisition,^39–47 but claims regarding vaccine efficacy have not been straightforward to interpret. Furthermore, none of the six HIV efficacy trials reported a reduction of viral loads in vaccine recipients that became infected.

While attempts to develop improved vaccine strategies continue, many feel that alternate approaches that differ from conventional vaccination may be needed. One such alternate approach is the delivery of anti-HIV monoclonal antibodies (mAbs) by recombinant AAV (rAAV) gene transfer. This technology is independent of the host immune system and AAV-delivered antibodies have the potential to create a long-term sterilizing barrier against HIV. Studies that have employed rAAV vectors to deliver antibodies or antibody-like molecules have shown protective effects against simian immunodeficiency virus (SIV) in monkeys,^48,49 simian-human immunodeficiency virus (SHIV) in monkeys^50,51 and HIV in humanized mice.⁵² Although encouraging, efficacy in monkeys was limited by immune responses to the delivered transgene product.^48,49,51 AAV-mediated delivery of broadly neutralizing antibodies (bnAbs) also shows promise for inhibiting viral replication and possibly even eradicating infection in HIV-positive individuals. Passive transfer of bnAbs to HIV-infected mice,^53–55 SHIV-infected monkeys,^56–58 and HIV-infected humans^59,60 has already shown potent antiviral effects when used as a therapeutic modality. However, those inhibitory effects against virus infection were transient due to the limited bioavailability of therapeutic antibodies following passive transfer. Recombinant AAV-antibody gene transfer could eliminate the need of repeated mAb infusions to already-infected humans and create constant, long-term levels of potent bnAbs in serum.

This review discusses the promise of AAV-delivered bnAbs for the ability to protect against the acquisition of HIV infection in humans and to block virus replication in those individuals that are already infected. We also discuss the problem of immunogenicity of AAV-delivered antibodies, which appears to be a major stumbling block to effective application of this approach for use in people.

The Elusive HIV Vaccine

Soon after HIV was discovered, the scientific community was optimistic that a vaccine against the AIDS-causing virus could be developed in a timely manner. That belief has unfortunately been shattered. More than 30 years of research have shown that a vaccine against HIV will be much more difficult to develop than the successful vaccines that exist for other pathogens.^5,22,61,62 The biggest challenge in the development of an effective HIV vaccine lies in the nature of the virus itself. HIV establishes a continuous presence by the integration of its genetic information into the host genome; it is able to generate and tolerate an enormous degree of genetic variation; and it has evolved a variety of strategies for evading host immune responses.^7,63–70 Once HIV establishes the initial infection, it is able to replicate continuously and without relent despite apparently strong humoral and cellular immune responses.²² Factors that contribute to a failed immune control of HIV infection are summarized in (Figure 1).^{23,69,71–78}

Figure 1.

Difficulties associated with immune control of HIV infection. The nature of HIV and the evolution of immune evasion strategies of the virus are responsible for why a HIV vaccine has remained an elusive task. HIV preferentially infects and destroys CD4+ T cells (central mediators of the immune systems), especially in the gut-associated lymphoid tissue (GALT). The virus early establishes a reservoir in latently infected CD4+ T cells by integration of proviral DNA into the host cell genome. Recognition by cytotoxic T cells is further exacerbated by downregulation of MHC class I molecules on the surface of virus-infected cells, which is orchestrated by the viral nef gene. Sensing of the pathogenic intruder by the host innate immune system is counteracted by the HIV-1 genes vif and vpu. Antibody and CD8+ T cell responses are readily escaped by selection of antigenic escape variants facilitated by the high mutation rate of the virus. The error-prone reverse transcriptase causes an enormous sequence diversity of the envelope glycoproteins gp120 and gp41 (up to 35% among clades, 20% within clades, 10% in a single infected individual). An extensive glycan shield on the env trimer shields vulnerable targets on envelope (about 50% of the mass of gp120). Abbreviations: reverse transcriptase (RT); integrase (IN); protease (PR); capsid (CA); matrix (MA); nucleocapsid (NC); long terminal repeat (LTR); group-specific antigen (gag); the pol gene encodes RT, IN and PR; viral infectivity factor (Vif); viral protein R (Vpr); viral protein unique (Vpu); negative regulatory factor (Nef); trans-activator of transcription (tat); regulator of expression of virion proteins (rev); envelope (env) gene encodes the glycoprotein gp160 that is processed into gp120 and gp41.

Since the first HIV vaccine trial in 1987 (ref. 79), more than 270 trials have followed.⁸⁰ From these, several vaccine candidates have progressed to a total of six phase IIb or phase III efficacy trials (Table 1).^6,7,81 AIDSVAX B/B used in VAX004 (refs. 25–28) and AIDSVAX B/E used in VAX003 (ref. 29) were the first HIV vaccines to enter phase III clinical trials. The vaccine preparations consisted of combinations of HIV recombinant gp120 envelope (env) proteins. As the name implies, AIDSVAX B/B included envelope protein sequences of two clade B isolates (MN, tissue culture derived strain; GNE-8, primary isolate). AIDSVAX B/E included the sequence of a clade B isolate (MN) and the sequence of a clade E isolate (CM244, primary isolate). The goal of the two studies was to test whether the gp120-induced antibodies were capable of preventing acquisition of the virus in high-risk populations. The outcome of the trials showed that the vaccines were not effective at preventing HIV infection; the rates of infection in the vaccinated groups versus the unvaccinated groups were similar. Furthermore, the vaccines had no influence on viral loads, CD4+ T cell counts or progression to AIDS.^25–29

Table 1. HIV vaccine efficacy trials in humans.

Trial	Name of trial	Clinical trials identifier	Name of vaccine	Vaccine components	Dates	Population	Estimated enrollment	Efficacy
1	VAX 004	NCT00002441	AIDSVAX B/B	gp120 proteins (clade B)	1998–2003	Adults at risk of sexually transmitted HIV-1 infection	5,400	No
2	VAX 003	NCT00006327	AIDSVAX B/E	gp120 proteins (clades B and E)	1999–2003	Intravenous drug users	2,500	No
3	STEP study (HVTN 502)	NCT00095576	MRKAd5	HIV-1 gag/pol/nef trivalent Ad5 vector vaccine	2004–2007	Adults at risk of sexually transmitted HIV-1 infection	3,000	Enhanced acquisition
4	Phambili study (HVTN 503)	NCT00413725	MRKAd5	HIV-1 gag/pol/nef trivalent Ad5 vector vaccine	2007–2007	Adults at risk of sexually transmitted HIV-1 infection	800 (of 3,000)	Enhanced acquisition
5	RV144 (Thai trial)	NCT00223080	ALVAC-HIV and AIDSVAX B/E	Canarypox vector (HIV-1 env/gag/pro), and gp120 proteins (clades B and E)	2003–2009	Adults at risk of sexually transmitted HIV-1 infection	16,400	∗
6	HVTN 505	NCT00865566	VRC DNA/rAd5	DNA plasmid (gag/pol/nef/env), and rAd5 (gag/pol/env)	2009–2013	Men/Transgender at risk of sexually transmitted HIV-1 infection	2,500	No

∗Per-protocol analysis: no significant efficacy; intent-to-treat analysis: no significant efficacy; modified intent-to-treat analysis: 31% efficacy (P = 0.04).

These first vaccine efficacy trials were followed by two very different efficacy trials that were based on viral vectors aimed at eliciting cellular responses against HIV. Virus vectors derived from replication-defective adenovirus serotype 5 (Ad5) were utilized in the STEP study^30–34 and the Phambili study,^35,36 numbers 3 and 4 of the HIV efficacy trials. The STEP study enrolled HIV-1-negative individuals that were given either a placebo or an equal mixture of three separate recombinant Ad5 vectors from the company Merck (MRKAd5). The three vectors expressed gag from the HIV-1 strain CAM-1, pol from the HIV-1 strain IIIB and nef from the HIV-1 strain JR-FL. The MRKAd5 vaccine utilized in the STEP trial did not reduce HIV infection rates and did not decrease viral loads in individuals that became infected with HIV. On the contrary, vaccinees had a significantly higher risk for acquiring HIV as compared to the placebo group. The concomitant Phambili study aimed at testing the MRKAd5 vaccine in South Africa, where HIV-1 clade C is predominant, was stopped prematurely due to the STEP trial results. However, individuals that were already vaccinated in the Phambili study continued to be followed. As in the STEP trial, vaccinees in the Phambili trial showed significantly increased acquisition of HIV infection. Increased acquisition in the vaccine groups in both trials was restricted to subgroups of individuals with high pre-existing antibody titers to Ad5 (refs. 30,32,36). A high pre-existing antibody titer to Ad5 was associated with greater susceptibility to acquiring HIV infection when they were vaccinated with the recombinant Ad5 vaccine.³² It has been suggested that the T cell activation caused by the recombinant Ad5 vaccine in individuals with pre-existing Ad5 immunity made them more susceptible to the initial HIV infectious event.^36,82

The fifth efficacy trial, the RV144 or Thai trial,³⁸ tested a prime-boost regimen based on four injections with ALVAC-HIV⁸³ (the vCP1521 canarypox vector expressing the HIV genes env, gag, pro) followed by two injections with AIDSVAX B/E (recombinant gp120 subunit vaccine). The study enrolled over 16,000 healthy adults of which over 12,000 individuals completed all vaccination visits while remaining HIV-negative through the last scheduled vaccination. The ALVAC/AIDSVAX prime-boost vaccine did not have significant protective effects against acquisition when analyzed by a per-protocol analysis or with an intent-to-treat analysis. However, the modified intent-to-treat analysis showed 31% protection against acquisition with a p value of 0.04 (refs. 38,84). A statistical interpretation analysis published by Gilbert et al. reported less than a 78% chance of any vaccine efficacy at all.³⁹ It has been suggested that binding of IgG antibodies to variable regions of HIV-1 env proteins may have contributed to protective effects against HIV infection in the vaccine recipients.^{40,41,43,45–47} Statistically significant “sieving” effects have also been reported.⁴² However, these sieving effects included a gp120 env amino acid that was present in the vaccine being overrepresented in vaccine recipients who became infected as compared to placebo recipients who became infected. It also included selective sieving of amino acids in genes that were not even included in the vaccine.⁸⁵ There is no rational explanation for these sieving effect observations. The vaccine did not induce bnAbs, it did not elicit CD8+ cytotoxic T cell (CTL) responses^44,86 and viremia was not reduced in individuals that became infected with HIV.³⁸

HVTN 505 is the sixth efficacy trial. The goal of the vaccine approach used in this trial was to test the efficacy of a DNA prime and Ad5 vector booster immunization in high-risk male or transgender individuals. Because of the results of the STEP and Phambili trials, individuals with high pre-existing immunity to Ad5 were excluded from the study. The HVTN 505 study was prematurely stopped due to futility 24 months after initial enrollment of participants. Therefore, data analysis could only be performed on those individuals who completed the 24-month study visits, about two thirds of the intended enrollment. The vaccine induced cellular and humoral immune responses but failed at preventing HIV infection with no difference in acquisition between the vacinees and the control group. Also, the vaccine had no influence on viral loads at set point.³⁷

Vaccine Trials in Monkeys

Vaccine studies in monkeys using SIV or SHIV have been used to inform and guide the development of vaccine concepts for human clinical trials.^87–90 Results from monkey studies can be used to rank order, or select, the most promising concepts for trials in humans. The SIV strains SIVmac239 (ref. 91) and SIV251 (ref. 92), as well as the SHIV strains SHIV-SF162 (refs. 93,94) and SHIV-AD8 (refs. 95,96) have been preferentially used, but by no means exclusively.^22,88,97 The greatest protective efficacy in monkeys has been achieved by using live attenuated strains of SIV, such as those deleted of the nef gene.^97,98 Durable protection has been consistently demonstrated against homologous virus challenge in a variety of studies.^98–105 However, considerably less protection has been observed by live attenuated strains when the challenge virus was not closely matched in sequence.^105–111 This relatively unimpressive level of protection by live attenuated SIV against challenge by a heterologous AIDS virus strain is perhaps analogous to the inability of human infection with wild-type strains of HIV-1 to routinely protect against superinfection by different strains of HIV-1 (refs. 112–116).

The next most impressive degree of protection in monkey vaccine trials has been achieved with a recombinant, replication-competent herpesvirus derived from the simian cytomegalovirus (CMV).^117–119 Approximately 50% of vaccinated monkeys have shown a remarkable degree of virological control following stringent SIVmac239 challenge,¹¹⁸ and no detectable signs of virus infection after more than 1 year from the time of infectious exposure.¹¹⁹ The protective effects that were induced by the recombinant CMV vaccine have been associated with broad and unusual effector memory CD8+ T cell responses that recognize non-classical SIV epitopes including those that are restricted by class I antigen E or class II major histocompatibility complex molecules.^120,121 This type of immunogenicity has been found to be a result of the gene-deleted rhesus CMV strain 68-1 that was being used.^122,123 However, even the rhesus CMV vaccine conferred no protection against acquisition of the homologous challenge virus and 50% of the vaccinated monkeys showed no protective effects at all.

A variety of vector-based approaches are being examined in monkey testing and some have already advanced to human trials.^124–127 A replication-competent vector based on adenovirus type 26 (Ad26) has shown promise in protecting monkeys against stringent SIVmac251 infection^128,129 and a version expressing HIV-1 env protein has advanced to phase I clinical trials in humans.^130–132 Vectors based on rhesus monkey rhadinovirus, a gamma-2 herpesvirus that is closely related to human Kaposi’s sarcoma-associated herpesvirus, are also being used in monkey trials.^133,134 Other promising viral vectors that have shown significant protective effects against SIV challenge in monkeys include: modified replicating vaccinia virus Tiantan,¹³⁵ modified vaccinia Ankara virus,^136,137 live recombinant vesicular stomatitis virus, and Semliki Forest virus replicon.¹³⁸

Broadly Neutralizing Antibodies Against HIV

Following infection with HIV-1, the anti-HIV antibodies that appear over the first 3 to 6 months typically show very strain-specific neutralizing activity, specific for the sequence of the infecting strain of virus.^68,139–141 These strain-specific neutralizing antibodies target the most variable regions of the envelope protein, the so-called variable loops, principally V1 and V2 (refs. 139,142). Just as HIV can easily escape a single antiviral drug, HIV variants appear within months that resist neutralization by the early strain-specific neutralizing antibody response.^{140,141,143,144} While the B cell repertoire evolves and changes in response to the changing virus, it is a race that the B cells do not win.¹⁴⁰ On rare occasions, however, antibodies with superior neutralization potency and breadth do emerge.^145–148 These potent broadly neutralizing antibodies (bnAbs) emerge on these rare occasions over a prolonged period of years and frequently have unusual structures that allow them to target the concealed, conserved regions of the envelope protein.^149,150

Numerous attempts to induce bnAbs by vaccination in humans have not been successful.¹⁵¹ If continually replicating HIV during the long course of infection does not routinely induce bnAbs, it is easy to imagine how difficult it will be to design immunogens to do so. The long-lasting antibody-virus chase continuum that results in these rare and potent bnAbs is consequently associated with unusual characteristics, including: a highly-evolved, high degree of somatic hypermutation (SHM) that can be accompanied by insertions and deletions; very long complementarity determining regions 3 (CDR3s); unusual structures.^{149,152–156} Despite progress in areas such as reverse or structure-assisted vaccinology, it will remain an enormous challenge to those interested in antigen design and vaccine delivery to overcome these obstacles for developing a truly effective HIV vaccine.^72,157–170

Given the difficulties in eliciting antibodies with potent neutralizing activity against a broad range of HIV-1 isolates, considerable interest has emerged in the concept of directly delivering the unusual monoclonal antibodies (mAbs) with the desired properties. More than a dozen distinct, potent bnAbs have now been isolated and characterized from infected humans (Figure 2). They can be roughly categorized into at least five groups: CD4 binding site; mannose patch; the membrane-proximal external region on gp41; Apex; the gp120-gp41 interface. The reader is referred to a number of outstanding reviews on the properties of these mAbs.^{78,144,149,150,167,171,172} We may not know how to elicit such antibodies, but we already have this impressive array of potent bnAbs, they are human in origin, and they can be delivered for prevention or therapeutic purposes.

Figure 2.

Broadly neutralizing antibodies (bnAbs) to HIV-1. The HIV envelope (env) spike consists of three gp120-gp41 heterodimers that are noncovalently linked to each other. The glycoprotein gp120 harbors the CD4 receptor binding site (CD4bs) and the coreceptor binding site, but the co-receptor binding region is only fully exposed upon binding of gp120 to CD4. The glycoprotein gp41 anchors the env spike into the virus membrane and harbors the fusion machinery that facilitates entry into the target cell. The env trimer spike is considered to be unstable, and decayed or nonfunctional structures appear to be a target for binding/non-neutralizing antibodies. Neutralizing antibodies, especially bnAbs strongly bind the native or functional env trimer spike. Several vulnerable bnAb target sites have been identified and a number of bnAbs bind to at least 5 well-characterized locations on the env trimer. The high-mannose-patch is located on the outer region of gp120, centered on glycans at Asn³³² (N332). bnAbs to this site bind and penetrate the glycan shield and interact with amino acids in the variable loop 3 (V3) of gp120. Apex antibodies bind to lysine-rich regions on the V2 loop, surrounded by glycans at Asn¹⁶⁰ (N160); antibodies to this site require long penetrating heavy chain structures. CD4bs antibodies have structural features that allow binding to the env trimer similar to that of the outer domain of CD4. The CD4bs is protected by the glycan shield and variable loops. MPER-specific antibodies usually have a hydrophobic character due to their binding target that is in close proximity to the lipid bilayer, which is partly recognized by this antibody class. These antibodies are usually self-reactive. Antibodies to the gp120-gp41 interface interact with both glycoproteins and appear to be trimer-specific. Abbreviations: CD4-induced (CD4i), membrane-proximal external region (MPER).

The discovery of bnAbs can historically be divided into two phases. In the early 1990s, hybridoma and phage display methods were used to isolate the first bnAbs by adsorbing sera of HIV-infected patients with monomeric gp120 and gp41 antigens. These “first-generation” bnAbs could effectively neutralize clade B viruses at a half-maximal inhibitory concentration range (IC50) of 1 to 10 μg/ml as assessed by in vitro assays, but they were less or not effective against other global HIV isolates.¹⁴⁴ Among the first-generation bnAbs were b12, 4E10, 2F5, and 2G12 (refs. 173–182). In the year 2009, the discovery of a second wave of bnAbs began following the development of improved mAb isolation techniques and the screening of larger cohorts of HIV-infected individuals. Selective B cell sorting and B cell capture methods have facilitated the isolation of a spectacular array of potent bnAbs.^{144,149,167,171} These “second-generation” bnAbs are broader and two to three orders of magnitude more potent than the earlier generation of neutralizing antibodies.^144,149 Among the new bnAbs are PG9 and PG16 (ref. 183), VRC01 (ref. 184), 3BNC117 (refs. 185,186), PGT121 and PGT128 (refs. 187,188), 10-1074 (ref. 189), 10E8 (refs. 190,191), 35O22 (ref. 192), PGDM1400 (ref. 193), and VRC34.01 (ref. 194).

Passive transfer of first-generation bnAbs has conferred protection against SHIV infection in monkeys; protective effects seen in those experiments could be attributed to both the neutralization activity and the Fc-mediated effector functions of the utilized mAbs.^195–203 Consistent with these second-generation bnAbs exhibiting much higher potency in cell culture, they also showed a higher efficacy in vivo as compared to the first-generation bnAbs.²⁰⁴ Either 3BNC117 or 10–1074, which were given to healthy macaques, were capable of completely blocking SHIV acquisition following a single intrarectal challenge with three half-maximal animal infectious doses (AID50), as long as the infused mAb dose was above 5 mg/kg.⁵⁸ A prevention study in monkeys that was published by the same group estimated that a 1:100 neutralization titer in plasma, which was generated by passively transferred bnAbs, was sufficient to provide protection in 50% of SHIV-exposed animals.²⁰⁵

The mAb PGT121 was tested for its protective efficacy against vaginal SHIV infection. All 10 monkeys that received a PGT121 dose of ≥1 mg/kg showed sterilizing immunity against a single high-dose SHIV-SF162P3 challenge with 300 half-maximal tissue culture infectious doses (TCID50), and three of five monkeys were even protected with a mAb dose of 0.2 mg/kg.²⁰⁶ A modified version of the bnAb VRC01 with mutations in the IgG Fc portion, termed VRC01-LS, exhibited a threefold longer half-life in serum and increased translocation to mucosal tissues than unmodified VRC01 (ref. 207). The improved biochemical properties together with the overall potency of VRC01-LS provided superior protection against single high-dose rectal challenge with the strain SHIV-BaLP4 (refs. 207,208). Another study utilized the bnAbs VRC01, VRC01-LS, 3BNC117, and 10–1074 to evaluate protective efficacy against SHIV-AD8 acquisition. It was shown that monkeys that received a single mAb by passive transfer required up to 23 weekly low-dose virus challenges to become infected as compared to the control group that became infected after only a median of three challenges.²⁰⁹

Experiments in humanized mice and in monkeys have also demonstrated therapeutic potential of second-generation bnAbs. Infant rhesus macaques were infected with SHIV-SF162P3 by the oral mucosal route and treated as early as 1 day after virus infection with a mix of the bnAbs PGT121 and VRC07. Unlike the untreated animals, the mAb-treated animals were free of virus in plasma and tissue by day 14 and remained free of virus even 6 months after the infectious exposure.⁵⁶ A separate study employed monkeys that had been chronically infected with SHIV-SF162P3 for 9 months and subsequently infused with mAb cocktails containing b12, 3BNC117, and PGT121 (ref. 57). In the vast majority of animals, plasma viral loads were reduced within 7 days to undetectable levels until a median of 56 days; viremia rebounded when mAb levels decreased to sub-threshold levels. A reduction of cell-associated virus was also noted. In a parallel study, mAb treatment was employed in monkeys 3 months after SHIV-AD8 infection.⁵⁸ Monotherapy with the bnAbs 3BNC117 or 10–1074 resulted in a rapid decline in viral loads reaching undetectable levels by 4 to 7 days, followed by virus rebound that identified escape mutants to the single mAbs. A single treatment using both mAbs together suppressed viremia for 3 to 5 weeks, and readministration of the mAb combination allowed repeated transient suppression of viremia.

Monotherapy with PG16, NIH45-46^G54W, PGT128 or 10–1074 resulted in transiently reduced viral loads in humanized mice infected with the strain HIV-1_YU2 (ref. 53). Virus rebound was associated with distinct escape mutations in the envelope gene. However, a single injection of a combination of bnAbs was capable of controlling HIV infection and suppressing viremia to levels below the limit of detection.⁵³ Viral escape from one mAb is somewhat predictable, as the selective immune pressure is not sufficient to inhibit viral replication long-term. Based on in vitro neutralization assays and mathematical prediction models, it has been reported that a combination of three to four potent bnAbs is likely to provide complete or near complete protection against HIV replication.^210,211 Another study that utilized a similarly combined passive transfer regimen involving the mAbs PG16, 10–1074, and 3BNC117 confirmed suppressive effects on HIV in humanized mice, which included lowering of free virus in serum, delayed viral rebound after cessation of antiretroviral therapy (ART) and reduction of cell-associated HIV-1 DNA.⁵⁴

Although ART and multiple bnAbs are able to suppress viremia in infected mice, there are still latent reservoirs of HIV-infected cells that are refractory to those treatments. An approach, called “shock and kill”, that combines ART and inducers of viral transcription has so far failed to eradicate the latent HIV reservoir.²¹² However, a study in mice showed that a trimix of bnAbs in combination with three inducers was capable of decreasing the HIV reservoir as measured by viral rebound. Interestingly, the data also revealed that suppression of HIV by the passively transferred Abs was dependent on interaction of the IgG Fc with Fc receptors of immune cells suggesting the importance of IgG effector functions.⁵⁵ Other studies confirmed that Fc receptor-mediated effector functions of bnAbs play a substantial role in inhibiting HIV or SHIV infection.^{200,213–215} In this context, the antiviral activity of the IgG Fc is directed against both free virus and virus-infected cells. Therefore, the potency or antiviral capacity of an anti-HIV Ab is not only defined by the affinity function of its Fab but also by the effector mechanisms that are mediated by its Fc.^{72,78,144,216,217}

Some first-generation and second-generation bnAbs have also progressed to human clinical trials. When used in HIV-positive individuals as therapy, first-generation bnAbs were well tolerated and appeared to be safe at doses up to 14 g of mAb over a 4-week period.^218–220 Passive administration of 2F5 and 2G12 resulted in a transient reduction of viral loads in five of seven patients; the median decrease of RNA copies/ml in plasma was about 1 log during the treatment phase (day 0–28) while the maximum decrease was 1.5 log.²²¹ In a subsequent study, the effect of three bnAbs was tested in a human clinical trial. The goal of the experiment was to examine antibody-mediated suppression of HIV-1 rebound after cessation of ART.²²² Sequential infusions of the mAbs 2G12, 4E10, and 2F5 to HIV-infected individuals undergoing interruption of ART showed a delay in viral rebound. Passively administered antibodies showed a substantial inhibitory effect in two of eight chronically infected and in all six acutely HIV patients as compared to a control group, and viral rebound was significantly delayed in acutely infected subjects that received mAb therapy versus those that did not receive mAb therapy. The authors also noted that the bnAb 2G12 had the strongest antiviral effect of all three mAbs used, and that the loss of viremia control in 12 of the 14 immunized patients was associated with viral escape from that mAb. No escape mutants were noted for the other two mAbs, 4E10 and 2F5. Another group conducted a similar passive transfer experiment using the same three bnAbs and confirmed the previously obtained results.²²³

Phase 1 trials have now evaluated safety, pharmacokinetics and functionality of second-generation bnAbs in people as well. The bnAbs VRC01 and 3BNC117 have been among the first of these to prove their potency in humans; results with these mAbs were just published in the year 2015. Twenty-eight healthy volunteers were given intravenous infusions of the mAb VRC01 (ref. 224). VRC01 appeared to be safe and well tolerated; also, no serious adverse events and no dose-related toxicities were noted following the mAb infusions. The mean concentrations over a 28-day period were 35 μg/ml (at 20 mg/kg) and 57 μg/ml (at 40 mg/kg); readministration on day 28 increased the mean concentration in serum to 56–89 μg/ml; the half-life of VRC01 was 15 days. Furthermore, no anti-VRC01 antibody responses were detected in any volunteer at any time. In another human trial, VRC01 was given to HIV-infected individuals. ART-treated and ART-untreated HIV patients received infusions of the VRC01 mAb at a dosage of 1, 5, 20, or 40 mg/kg.⁵⁹ Two mAb infusions, conducted on day 0 and 28, did not reduce the amount of cell-associated viral DNA (also referred to as reservoir) in the ART-treated HIV patients with undetectable viral loads in plasma. However, a single infusion of VRC01 decreased the plasma viral load by 1.1–1.8 log₁₀ in six of the eight ART-untreated viremic HIV patients. Reduction of viremia was transient and viral loads returned to baseline levels within 56 days after mAb infusion due to waning mAb levels and selection for less sensitive viruses.

Another human trial that employed passive transfer was published in the same year. The bnAb 3BNC117 was tested in 12 healthy and 17 HIV-infected individuals.⁶⁰ A single infusion of the mAb appeared to be safe and well tolerated at all doses tested (1, 3, 10, 30 mg/kg); also, no serious adverse events were noted. The half-life of 3BNC117 was 17 days in healthy volunteers and 9 days in HIV-infected patients. HIV-infected individuals that received lower doses of 3BNC117 showed only small and transient reductions in viral loads followed by a rapid return to baseline levels. However, a single infusion of the mAb at higher doses (10 and 30 mg/kg) reduced the viremia up to 2.5 log₁₀ in 10 out of 11 subjects, and viral loads remained significantly reduced for 28 days. Emergence of resistant viral strains was variable among the 3BNC117 recipients. Development of increased neutralization resistance was observed in some patients that exhibited escape mutations in the CD4bs and amino acid insertions in the V5 loop of HIV env.

Further experiments were conducted to explore the antiviral capacity of the bnAb 3BNC117 (ref. 225). Suppression of viral load was attributed to clearance of free virus and reduction of virus spread by clearance of virus-infected cells; clearance of virus-infected cells was dependent on Fc-mediated effector functions of 3BNC117. Another study examined the effects of 3BNC117 monotherapy on the host’s antibody responses.²²⁶ Autologous IgG samples from day 0 and week 24 postinfusion were tested for their capacity to neutralize a panel of HIV-1 pseudoviruses and autologous viruses from day 0 and week 4 postinfusion. It was shown that autologous week 24 IgG, by which time 3BNC117 had already decayed to below detection, had an increased neutralizing activity against weeks 0 and 4 autologous viruses as compared to the neutralizing activity of autologous day 0 IgG. Therefore, patients that received a passive immunization against HIV appeared to develop stronger host antibody responses to their own HIV infection. A separate trial investigated the effects of 3BNC117 on HIV after ART interruption. The results showed that repeated mAb administrations significantly delayed virus rebound as compared to nontreated individuals; but it also revealed that virus rebounded after antibody levels waned, and that use of 3BNC117 alone led to neutralization-resistant escape mutants.²²⁷

Although it has been shown that sera from HIV-infected individuals can enhance HIV infection in vitro, there has been no clear evidence that passively transferred antibodies pose a risk to enhancement of HIV infection in vivo.^228–230 Nonetheless, antibody-dependent enhancement could theoretically represent a problem to passive immunization strategies against HIV. Despite the promise of utilizing bnAbs to prevent or treat HIV infection, reasonable risk assessments will need to be performed for each individual anti-HIV mAb to exclude the chance of increased virus acquisition or increased virus replication following passive transfer to humans.²²⁹

AAV-Mediated Delivery of Antibodies and Antibody-Like Molecules

With the availability of more than a dozen potent bnAbs, and given developments in antibody engineering that have enhanced biochemical and antiviral properties,^231–247 it is easy to imagine the potential for the effectiveness of such anti-HIV mAbs in both prevention and therapeutic scenarios. In prevention scenarios, delivery of potent bnAbs could overcome the difficult barriers to trying to induce such antibodies, with the goal of creating a long-term sterilizing barrier to infection. In therapeutic scenarios, the goals would be to greatly reduce viral replication and plasma viral loads, to eliminate the need for continuing antiviral drug therapy, and it would also hope to reduce viral reservoirs over time toward a real cure.

One issue that will need to be addressed, particularly for therapeutic scenarios, is whether some particular combinations of mAbs provide remarkably synergistic levels of protective effects. Do some combinations of potent bnAbs result in a much greater degree of virus neutralizing activity than either alone?²¹¹ Do some combinations of potent bnAbs make it much more difficult or impossible for the virus to escape the activity of the combination? Does escape from some combinations of potent bnAbs result in virus that is so poorly fit for replication that it can be easily controlled by the host? Does escape from some combinations of potent bnAbs result in virus that is so easily neutralized that it can be well controlled by the host immune responses? These questions can be readily addressed by cell culture and monkey studies.

Maintenance of effective concentrations of mAbs over prolonged periods by passive administration would require repeated, regular infusions over a prolonged period. This does not seem practically possible on a large scale for a variety of reasons. First, it would be prohibitively expensive to use on a large scale just for the antibody production, purification and quality control. Second, long-term adherence is certainly likely to be a problem, particularly in many regions of the developing world. Recombinant adeno-associated virus (rAAV) is ideally suited to achieve the goal of long-term delivery (Figure 3). AAV-based gene delivery is considered to be a safe and effective technology.^248–257 Numerous studies in monkeys^258–261 and people^262–275 have shown the successful and safe application of rAAV vectors for the treatment of various genetic diseases. The positive results of clinical trials for lipoprotein lipase deficiency have led to the first gene therapy product to achieve regulatory approval by a governmental health institute.^276–279

Figure 3.

Recombinant adeno-associated virus (rAAV) vectors for the delivery of monoclonal antibodies (mAbs). Wild-type adeno-associated virus (AAV) is a 25 nm small nonenveloped virus that packages a single-stranded DNA genome. The most prominent AAV serotype, AAV2, has a genome size of 4.7 kb and harbors two viral genes (rep and cap) that are flanked by two 145 bp inverted terminal repeats (ITRs). Four Rep proteins (Rep78, Rep68, Rep52, and Rep40) are produced from transcripts using the p5 and p19 promoters, and these proteins are important for viral replication and regulation of AAV gene expression. The virus does not encode a polymerase enzyme and relies on cellular enzymatic activities. Furthermore, AAV relies on the presence of helper viruses such as herpesvirus or adenovirus in order to undergo productive infection (replication, gene expression, and virion production). The cap gene encodes three structural capsid proteins (VP1, VP2, and VP3) from two transcripts using the p40 promoter. For generating recombinant AAV (rAAV), the entire wild-type AAV genome is replaced by a unique transgene cassette (such as for a mAb) flanked by the AAV ITRs, which are the only wild-type sequences remaining. Production of rAAV virions is achieved by triple transfection using the rAAV vector plasmid and two helper plasmids in trans, followed by CsCl purification of the replication-deficient rAAV particles. rAAV particles can be encapsidated by any of the 12 AAV serotypes and more than 100 variants that are available. The conventional single-stranded AAV (ssAAV) vector packages single-stranded DNA. The modified self-complementary AAV (scAAV) encapsidates double-stranded DNA but has only half the packaging capacity of ssAAV. scAAV vectors are produced by modification of the 5’ITR (removal of the terminal resolution site and D sequence). The size limit of the scAAV vector system requires the heavy and light chain sequences of IgG to be placed on separate rAAV vectors.

The only product that is made by rAAV derives from the transgene that was cloned into the vector.^{254,280–284} Genetically engineered AAV genomes persist in the cell in episomal form and will produce your protein of choice for the lifetime of the cell.^285,286 AAV is capable of transducing quiescent cells such as those from skeletal muscle (Figure 4); as long as the transgene product is viewed as self by the host immune system, rAAV-delivered proteins can be secreted for decades from such long-lived cells.^{48,49,251,261,286,287} Several groups have demonstrated the protective efficacy of AAV-delivered antibodies and antibody-like molecules against AIDS virus infection in monkeys and humanized mice.^{48–50,52,288–291}

Figure 4.

AAV-mediated gene transfer of anti-HIV monoclonal antibodies (mAbs). A rAAV encoding an anti-HIV mAb is injected into muscle such as the deltoid muscle. Following intramuscular inoculation, rAAV binds to a serotype-specific cellular receptor on myocytes and becomes endocytosed. Virus particles are transported to the nucleus, into which the rAAV genome is released. The single-stranded DNA (in case of ssAAV) is then converted into transcriptionally active double-stranded DNA. Double-stranded rAAV genomes appear to be stabilized by ITR-to-ITR interactions and enzymatic modifications, leading to high molecular weight (MW) rAAV genome polymers that persist in episomal form for the lifetime of the cell. Adult human muscle cells have a lifespan of more than 10 years. The extrachromosomal rAAV DNA forms maintain gene expression and produce the therapeutic mAb, which undergoes the secretory pathway and is released into the circulatory system. Depending on which AAV serotype or variant is used, the injected rAAV virus can also transcytose through multiple cell layers, leading to access to blood vessels. This will transport a proportion of intramuscularly injected rAAV particles to the liver, where cell entry and gene expression will take place. Secretion of an anti-HIV mAb from muscle and liver will potentially create a prophylactic barrier against HIV infection or fight an ongoing HIV infection in a therapeutic setting. AAV, adeno-associated virus; ITR, inverted terminal repeats.

A pioneering study conducted in rhesus macaques employed AAV-delivered single-chain fragment variable (scFv) immunoadhesins (antibody-like molecules) to protect against SIV infection.⁴⁸ The genetic material encoding the scFv immunoadhesins 4L6 and 5L7 used in that experiment was small enough to be accommodated by self-complementary AAV (scAAV) vector, a recombinant AAV variant that encapsidates double-stranded DNA.²⁹² The scAAV vector was chosen due to reports of its enhanced transduction capability and performance at achieving higher rates of transgene expression. However, scAAV is limited at packaging longer sequences such as the genetic information of both heavy and light chain sequences of a full-length immunoglobulin G (IgG).^293–295 Conventional single-stranded AAV (ssAAV) vector was used to deliver a rhesus CD4 - rhesus IgG fusion construct, termed N4. All three vectors (4L6, 5L7, and N4) had an AAV1 capsid. Following intramuscular injection of the rAAVs, immunoadhesin concentrations in serum reached up to 190 μg/ml by 4 weeks, and levels of immunoadhesins were maintained in some of the scAAV recipients above 200 μg/ml through 12 months. The nine AAV-immunized monkeys and two groups of control monkeys were challenged with a high dose of the strain SIVmac316 at 4 weeks following the AAV gene transfer. While all six control monkeys became infected by the SIV challenge, six of the nine AAV-immunized monkeys that maintained reasonable levels of immunoadhesins showed sterile protection against SIV exposure. However, the three other AAV-immunized monkeys developed immunoadhesin-specific antibody responses, which incapacitated protective efficacy and led to SIV infection.

Our group set out to build on that study by attempting to deliver authentic IgG versions of the rhesus-derived 4L6 and 5L7 (ref. 296). We wished to address the question whether the delivery of authentic full-length versions of these antibody-like molecules could obviate the anti-antibody response to them. Since both heavy and light chain sequences could not be accommodated by one scAAV, we placed them on two separate scAAV vectors. Additionally, we placed both heavy and light sequences onto one ssAAV vector by utilizing F2A peptide technology as previously described.^297,298 Two scAAV vectors or a single ssAAV vector, encapsidated by AAV1, were injected intramuscularly into 12 rhesus macaques and levels of AAV-delivered Abs were measured over time.⁴⁹ The concentration of the SIV-specific antibodies in serum ranged from 1 to 270 μg/ml through 44 weeks, regardless which AAV vector delivery system was being used. However, the conversion to authentic IgG sequences did not prevent the emergence of anti-antibody responses and this emergence limited the concentration of the SIV-specific antibodies that could be achieved. Nonetheless, we progressed with a challenge phase and conducted a repeated low dose challenge regimen using the neutralization-resistant strain SIVmac239. Although 4L6 and 5L7 IgGs showed no neutralizing activity in vitro, they exerted antiviral effects against highly pathogenic SIVmac239 challenges in vivo as assessed by the significant delay and reduction of viremia in plasma.⁴⁹

Another study conducted in monkeys utilized AAV-antibody gene transfer to protect against SHIV infection.⁵¹ Rhesus macaques were injected intramuscularly with a ssAAV8 vector expressing the potent bnAb VRC07. In four of four test animals, the serum concentration of the AAV-delivered antibody reached 8 μg/ml by week 4 and plummeted to undetectable levels by week 9. This was apparently due to a vigorous anti-VRC07 antibody response despite extensive attempts to make the VRC07 mAb as “rhesusized” as possible. A second group of monkeys then received the immunosuppressive agent cyclosporine A (CsA) prior to the AAV-antibody gene transfer. Although levels of delivered VRC07 reached serum concentrations as high as 66 μg/ml by week 3, anti-VRC07 antibody responses lowered the AAV-delivered antibody during and especially after immunosuppression. Following challenge with the strain SHIV-BaLP4, significantly more control monkeys became infected as compared to the AAV-antibody group.

One of the most potent and broad molecules capable of inhibiting AIDS virus entry is the antibody-like construct eCD4-Ig.⁵⁰ This molecule is composed of the outer two domains of CD4 (entry receptor of the AIDS virus), the Fc portion of IgG and a CCR5 (entry coreceptor of the AIDS virus) mimetic peptide that is derived from the HIV-specific antibody E51 (refs. 299–302). While some significant number of HIV-1 isolates are resistant to neutralization by potent bnAbs, eCD4-Ig has neutralized 100% of the tested neutralization-resistant strains. Furthermore, eCD4-Ig has been shown to potently neutralize HIV type 2 and the neutralization-resistant strain SIVmac239. Following AAV1 gene transfer to monkeys, levels of the antibody-like molecule eCD4-Ig ranged between 17 and 77 μg/ml by week 30. The four AAV-immunized macaques and four control macaques were then repeatedly challenged with progressively increasing doses of SHIV-AD8. Complete protection was demonstrated in the AAV-immunized animals, while all control animals became infected following the last challenge that utilized 4 AID50. Interestingly, anti-eCD4-Ig host antibody responses were low or absent, while AAV-transferred potent bnAbs used in that study elicited moderate to strong anti-antibody responses.⁵⁰

AAV-antibody gene transfer has also been used in humanized mice experiments. Notably, AAV-delivered bnAbs b12, VRC01 and 10–1074 have demonstrated protective effects against HIV acquisition and durable control of HIV in a therapeutic setting.^52,54,291 Although mouse experiments can demonstrate whether potent bnAbs have the ability to block or inhibit HIV infection in vivo, the humanized mouse model has certain limitations and may fall short when evaluating HIV pathogenesis, as well as safety and immunogenicity of AAV-delivered antibodies. Since virus challenge experiments are performed in immunocompromised mice that have been engrafted with human cells, it is difficult to translate results to immunocompetent monkeys or humans.⁸⁸

One trial is currently ongoing to evaluate safety, deliverability, and potential efficacy of rAAV-delivered potent bnAbs: PG9 in uninfected human volunteers in England.¹²⁷

Obstacles For Efficient AAV-Antibody Delivery

Several features of the rAAV vector delivery system may serve to limit the effectiveness with which the desired protein can be expressed. As with any virus, AAV can be recognized as foreign by the host immune system.^303–309 While rAAV vector does not directly express any wild-type AAV proteins, rAAV on its own may trigger innate immune responses.³⁰⁷ Furthermore, pre-existing cellular³⁰⁸ and humoral³⁰⁹ immunity to wild-type AAV may significantly limit the ability of the rAAV to “take” in the host.³¹⁰

There are 12 AAV serotypes and more than 100 variants (serovars) as specified by phylogenetic analyses.^{252,311–313} A number of studies have reported that individual AAVs can be sensed by pattern recognition receptors (PRRs), which can lead to upregulation of host defense genes and the production of proinflammatory cytokines and chemokines. This in turn will activate cells of the innate immune system and may amplify the inflammatory signal by initiating adaptive immune responses. Toll-like-receptors (TLR) such as TLR-2 and TLR-9 have been shown to be involved in innate immune responses to AAV by recognizing the AAV capsid and the AAV genome, respectively.^314–318

The prevalence of anti-AAV capsid IgG in the human population varies by AAV serotype; e.g., up to 72% of people are sero-positive for AAV2, 67% for AAV1 and 38% for AAV8. Also, antibodies against one serotype may cross-react against another serotype depending on how similar their capsid sequences are.^319–321 Neutralizing antibodies in serum at titers of more than 1:5 may already be sufficient to capture intravenously injected virus particles, and by so doing severely reduce transduction by rAAV.^{262,322–324} Furthermore, several groups have explored the possibility of AAV re-administration to muscle with little or no success when the exact same serotype was used^325–327; only in the event of immunosuppressive or immunomodulatory intervention was it possible to achieve effective uptake of the same AAV serotype.^328–330 Similarly, presence and activation of AAV capsid-specific memory CD8+ T cells can eliminate cells that have taken up rAAV particles.³³¹ Human trials that have employed rAAV to provide functional protein to individuals with hereditary disorder have reported anti-capsid responses to AAV2, AAV8 and AAV1 (refs. 263,264,269). AAV gene transfer may also elicit immune responses against the rAAV-delivered transgene product if the host has never seen that specific protein; in that context, the magnitude of the response is dependent on the degree the endogenous gene is different from the delivered gene, in particular whether the host protein may be truncated or missing entirely.^332–336 Human trials have reported transgene-specific cellular responses against rAAV-delivered α1-antitrypsin (AAT), mini-dystrophin protein and coagulation factor IX (F9).^{263,269,332,337–339} Furthermore, the magnitude and frequency of immune responses to rAAV vector and delivered transgene product is influenced by several other factors including the AAV serotype or variant that is being used, rAAV tropism for antigen-presenting cells (APCs), the rAAV dose and the route of rAAV administration.^{332,340–350}

When considering use of rAAV for delivery of mAbs, the first inclination is to assume that human antibodies are natural protein products of humans and should therefore not be viewed as foreign. However, things may not be that simple. The human B cell repertoire can create an enormous number of different antibodies with enormous sequence variation.^165,351,352 A particular antibody being made by one individual will not likely ever have been seen by another individual and will likely be less tolerized by the other individual. Furthermore, any particular antibody being made by an individual must have been accepted by a complex checkpoint system during B cell development within that host.^353–356 These considerations are further exaggerated by the highly evolved, highly mutated nature of the potent bnAbs one wants to deliver for the prevention or treatment of HIV infection.¹⁵⁰ bnAbs have undergone extensive SHM in their variable domains, which allows them to attain enhanced antiviral potency and breadth, but this may also be associated with some self-reactivity and with immunogenicity.^{154,357–360} CDR-sequence containing regions of variable domains of IgGs (idiotypic variation) may contain CD4+ T cell epitopes that induce unwanted immune responses in the mAb recipient.^361,362 Other properties of mAbs may also contribute to their immunogenicity in the recipient host: allotypic variation, misfolding, aggregation and differences in glycosylation.^363,364 Immune responses following passive transfer to humans have been reported for a number of therapeutic mAb.^365–368 Although species-specific antibodies have shown to have less immunogenic potential, immune responses to mAbs in humans have occurred independently of the nature of the transferred mAb: murine versus humanized versus fully human.^{365,369–371}

There have been five monkey trials to date where rAAV has been used to deliver antibodies or antibody-like molecules against HIV or SIV. The pioneering study by Johnson et al.⁴⁸ utilized rAAV1 to delivery the antibody-like molecules (immunoadhesins) 4L6, 5L7, and N4 as prophylaxis against SIV challenge. In contrast to the heavy and light chain coding sequences of a full-length mAb, the coding sequence of an immunoadhesin is small enough to be accommodated by scAAV; this vector type was being used for 4L6 and 5L7. Three of nine rhesus monkeys developed anti-immunoadhesin responses, and these three monkeys were not protected from SIV infection. Although 4L6 and 5L7 are composed of fully rhesus-derived sequences, humoral responses targeted these sequences. The authors found that reactivity was confined to the variable domains of these two immunoadhesins. Humoral responses were also measured against the rhesus CD4 moiety of N4, albeit modest. It is worth noting that 4L6 and 5L7 are extremely hypermutated and bear very long CDR3 sequences. Also, sequences of heavy and light chains were obtained by phage display, which might not resemble a natural pairing of these chains. The artificial fusion of variable light (VL) and variable heavy (VH) domains, as well as CD4 with the IgG Fc could have potentially created conformational epitopes that could be immunogenic.³⁷² Our group converted those immunoadhesin sequences to authentic IgG molecules to potentially avoid any unnatural structures.²⁹⁶ However, Fuchs et al.⁴⁹ and Martinez-Navio et al.³⁵⁹ found that full-length IgG versions of 4L6 and 5L7 did not prevent anti-antibody responses. Six out of six monkeys that received 4L6 IgG1 and three out of six that received 5L7 IgG1 generated anti-antibody responses. Both heavy and light chain variable regions were targeted including measured reactivity to the heavy chain CDR3 (ref. 359).

Our group has also delivered rhesusized versions of anti-HIV bnAbs (1NC9, 8ANC195, 3BNC117, 10–1074, and 10E8) to monkeys; anti-antibody responses were readily detected against all AAV-delivered antibodies in all eight animals.³⁵⁹ The levels of delivered mAbs were driven to below detection in all animals for all antibodies for which specific detection methods were available. Immunogenicity of the tested anti-HIV bnAbs correlated significantly with the degree of sequence divergence from germline.³⁵⁹ In another study, Saunders et al.⁵¹ delivered the HIV-specific bnAb VRC07 using AAV8. Four of four monkeys elicited anti-antibody responses to the mAb, and these unwanted anti-antibody responses resulted in a loss of transgene product in all animals by 9 weeks following rAAV administration.⁵¹ It is worth noting that the vigorous anti-antibody responses were mounted against VRC07, despite extensive efforts to “rhesusize” the mAb as much as possible. The bnAb VRC07 was created by pairing the light chain of the bnAb VRC01 with a heavy chain isolated from a B cell clone of the VRC01 lineage.³⁷³ This unnatural pairing of heavy and light chains, along with the 14% SHM rate of VRC01 (ref. 167) (full-length antibody sequence as compared to full-length germline sequence) and the further mutated VRC07 heavy chain sequence, may have contributed to the immunogenicity of VRC07. In a second group of animals, use of CsA did not prevent anti-VRC07 antibody responses, but humoral responses were blunted in three of six monkeys by that immunosuppressive intervention, and those three monkeys maintained measurable mAb levels through 16 weeks.⁵¹

Gardner et al.⁵⁰ delivered the broad and potent anti-HIV entry inhibitor eCD4-Ig by AAV1 to monkeys. While two of four monkeys had a weak anti-eCD4-Ig response, the other two showed no detectable anti-inhibitor reactivity. Comparably modest anti-inhibitor responses have also been observed with N4 (ref. 48). Since rhesus CD4 and rhesus IgG Fc are self proteins to rhesus monkeys, no considerable humoral responses were elicited.⁵⁰ Furthermore, no reactivity was raised against the CCR5 mimetic peptide, a CDR3-derived peptide^299,301 that was artificially fused to the IgG Fc.⁵⁰ Apparently, the amino acid sequence and the arrangement of the CCR5 mimetic peptide have not presented a major immunogenic stimulus in monkeys. The same group also tested the immunogenicity of the AAV-delivered bnAbs 3BNC117, NIH45-45, 10–1074, and PGT121 in monkeys.⁵⁰ The bnAbs elicited significantly higher anti-antibody responses as compared to eCD4-Ig. The rate of SHM among those 4 bnAbs is relatively high: 3BNC117 (36.9%), NIH45-46 (44%), 10–1074 (24.4%), and PGT121 (21.2%).¹⁶⁷

The inherent nature of an anti-HIV bnAb may be sufficient to elicit immune responses in the recipient host since the recipient likely never would have generated or experienced the specific variable domains. Human mAbs used therapeutically have been shown to elicit immune responses in a substantial fraction of humans following passive transfer.^365,366,368 To our knowledge, no side-by-side comparison has been conducted that evaluates the immunogenicity of a mAb when administered passively versus by AAV gene transfer. The anti-HIV bnAb VRC01 did not appear to elicit anti-VRC01 antibody responses in humans following one or two administrations.²²⁴ A simianized version of VRC01 elicited anti-VRC01 antibodies in two of eight macaques following four passive administrations.²⁰⁸ The simianized mAb VRC07 elicited robust anti-VRC07 antibody responses in four of four monkeys when delivered by rAAV,⁵¹ similar to the experience of Martinez-Navio et al. with a variety of AAV-delivered rhesus and rhesusized human mAbs.³⁵⁹ Again, the anti-anti responses to the AAV-delivered mAbs were directed principally or exclusively to the variable domains, i.e., they were anti-idiotypic in nature.^51,359

A number of studies have explored ways of reducing immune responses toward a variety of AAV-delivered gene products. The use of immunosuppressive agents such as CsA has shown partial success at reducing immune responses and facilitating transgene expression in monkeys.⁵¹ Temporary inhibition of CD4+ T cells has shown to be effective at preventing immune responses against AAV-mediated gene delivery, particularly in the context of AAV readministration in mice.³³⁰ A single patient case report showed that combined use of intravenous immunoglobulin (IVIG), B cell ablation and a corticosteroid has allowed for successful AAV-mediated gene transfer in the absence of immune responses towards AAV capsid and the delivered transgene product.³⁷⁴ Passive transfer of a large dose of mAb prior to recombinant AAV administration may circumvent the problem of “inverse dose-immunogenicity relationship”.^365,375 If readministration of rAAV is desired, the second AAV inoculation could employ a different serotype than the one used in the primary inoculation. Also, the use of engineered AAV capsids may help at minimizing host immune responses; AAV capsid mutations that involve Tyr, Lys, Ser, and Thr residues have shown to improve AAV transduction, and such capsid mutations could allow efficient AAV gene transfer at a lower AAV dose while potentially reducing the sensing by the innate immune system.^{343,376–378} Use of specific microRNA binding sites (miRNAbs) within the rAAV genome may prevent transgene expression in professional antigen presenting cells (APCs) and thus inhibit elicitation of immune responses.^379,380 Liver-directed AAV gene transfer may accomplish induction of tolerance toward any mAb. Expression of transgene products in liver tissue has been demonstrated to be tolerogenic by mechanisms that include but are not limited to induction of regulatory T cells (Tregs).^381–386

Summary

Wild-type AAV has never been associated with the cause of any known diseases in humans, and recombinant AAV has demonstrated its overall efficacy and safety in more than 120 clinical trials, with transient tissue inflammation as the most severe side effect.^256,387 Given the need to explore unconventional approaches against HIV, AAV-mediated delivery of potent anti-HIV bnAbs represents a promising approach for the prevention and treatment of HIV infection. Trials in monkeys have demonstrated significant efficacy of rAAV-delivered antibodies and antibody-like molecules for prevention of AIDS virus infection. Nonetheless, despite the safe and effective application that has been attributed to AAV-mediated gene transfer, immune responses to AAV-delivered antibodies remain the most significant impediment that will limit the effectiveness of this approach. This impediment needs to be better understood and overcome for the promise of the AAV-antibody approach to be effectively realized in people.

The authors declare no conflict of interest.