Engineered deletions of HIV replicate conditionally to reduce disease in nonhuman primates

Abstract

Antiviral therapies with reduced frequencies of administration and high barriers to resistance remain a major goal. For HIV, theories have proposed that viral-deletion variants, which conditionally replicate with a basic reproductive ratio [R₀]>1 (termed “therapeutic interfering particles” or “TIPs”), could parasitize wild-type virus to constitute single-administration, escape-resistant antiviral therapies. We report the engineering of a TIP that, in rhesus macaques, reduces viremia of a highly pathogenic model of HIV by >3log₁₀ following a single intravenous injection. Animal lifespan was significantly extended, TIPs conditionally replicated and were continually detected for >6 months, and sequencing data showed no evidence of viral escape. A single TIP injection also suppressed virus replication in humanized mice and cells from persons living with HIV. These data provide proof of concept for a potential new class of single-administration antiviral therapies.

INTRODUCTION:

Rapidly evolving communicable pathogens, such as HIV, pose stubborn problems for conventional medical countermeasures. Because of substantial viral genetic diversity, there is no approved HIV vaccine. Although combination antiretroviral therapy (ART) is effective, ART is not curative and must be continually administered (discontinuation results in viral rebound), leading to major logistical challenges. Thus, despite decades of intense research, 1 to 2 million new HIV infections occur each year, with the virus overly impacting underprivileged and underserved communities. New antiviral strategies with reduced administration frequencies and high genetic barriers to the evolution of resistance are needed.

RATIONALE:

Theories proposed decades ago predicted that HIV viral-deletion variants—based on the historical concept of defective interfering particles (DIPs), which parasitize wild-type viruses—could act as therapeutics. Putative therapeutic interfering particles (TIPs), a sub-set of DIPs, if rationally engineered to conditionally replicate with a basic reproductive ratio R₀ > 1, were predicted to be single-administration, escape-resistant antiviral therapies. Notably, the R₀ > 1 conditional replication criterion imparts TIPs with the potential to act as single-dose “drive” therapeutics, thereby overcoming extant therapy barriers. In this work, we developed a TIP for HIV.

RESULTS:

Theory-driven in vitro studies showed that long-term HIV-infected bioreactors exhibited von Magnus—like oscillations (a signature of DIPs) and revealed a variant with a ~2.5-kb deletion in the HIV pol-vpr region. The deletion variant satisfied the requirements for a DIP. By introducing additional deletions to ablate splicing and eliminate expression from all HIV reading frames and repairing the HIV central polypurine tract, the DIP was engineered into an R₀ > 1 TIP that durably suppressed HIV replication in vitro and in humanized mice. The TIP also established latency and suppressed HIV rebound or outgrowth in vitro, including from patient cells, and after interruption of antiretroviral therapy. To test TIP efficacy in a physiological model, an analog TIP lentiviral vector was constructed for SIV. Infant rhesus macaques were administered the SIV TIP by a single intravenous (IV) injection, challenged with a highly pathogenic SHIV (SF162P3) through the IV route (24 hours later), and followed for ~30 weeks. TIP intervention durably suppressed SHIV viral load by 4log₁₀ in plasma and lymph tissue and resulted in significant reduction in disease. As predicted, TIPs conditionally replicated in vivo for >6 months, and TIP-treated animals exhibited significantly improved immune responses, and no evidence of increased inflammation. Full-length sequencing of SHIV revealed no evidence of increased evolution, recombination, or escape.

CONCLUSION:

Epidemiological analyses argue that each 1log reduction in HIV viral load results in significant delays in the onset of AIDS and significant reductions in transmission potential. Theories have proposed that R₀ >1 TIPs have the capacity to establish coevolutionary arms races with wild-type viruses, leading to extraordinarily high genetic barriers to the evolution of antiviral resistance. On the basis of TIP intervention in SHIV-infected rhesus macaques, mathematical modeling indicates that a single dose of TIPs could durably reduce HIV viral load to less than the World Health Organization threshold for HIV transmission of 10³ copies/ml. These data suggest that TIPs could serve as the foundation for a new class of easy-to-administer, single-administration antiviral “drive” therapies. ■

Graphical Abstract

A single-dose HIV therapy: TIPs reduce disease in nonhuman primates through conditional replication. (A) Overview of the TIP concept. (B) TIP intervention in rhesus macaques infected with a highly pathogenic SHIV results in significant protection from disease. (C and D) TIP intervention durably reduces SHIV viral load by ~4log10 (C) owing to sustained conditional replication of TIPs (D).

More than 39 million people currently live with HIV-1 or AIDS (hereafter referred to as HIV), and ~1 to 2 million new HIV infections occur each year, with the virus overly impacting underprivileged and underserved communities (1). Combination antiretroviral therapy (ART) suppresses HIV replication, but ART is not curative and must be continually administered, as discontinuation almost invariably leads to rapid viral rebound (2, 3). Epidemiological projections indicate that controlling the HIV-AIDS epidemic by using conventional approaches may be challenging owing to deployment, behavioral, and mutational barriers (4–6). Consequently, single-administration therapeutics based on defective interfering particles (DIPs) have been proposed to overcome these barriers (7–10). Originally observed for influenza virus in the 1940s by von Magnus (11), DIPs are viral deletion mutants that conditionally replicate (i.e., lack self-replication) and act as molecular parasites of the wild-type virus within infected cells, stealing essential viral proteins (e.g., replication or packaging proteins) to reduce the levels of wild-type virus. Theoretical models predicted that DIPs could be engineered into long-acting, single-administration HIV therapies (termed “therapeutic interfering particles” or “TIPs”), provided that a basic reproductive ratio [R₀]> 1 (7, 8) could be achieved (fig. S1, A and B). R₀ is a threshold value, defined as the expected number of secondary infections arising from a single infected individual or cell (12), and proof-of-concept R₀ >1 TIPs were recently shown to be therapeutic in the context of acute, self-limiting infection (13, 14). However, the concept’s ultimate therapeutic promise is for chronic infections with high genetic variability, such as HIV, for which there is no available vaccine. Unfortunately, bona fide retroviral DIPs (the requisite chassis for TIPs) have remained elusive (15–20), and it was unclear if viable R₀ > 1 TIPs could be engineered for HIV, whether they would be therapeutic in the chronic setting, or whether efficacy could be sustained given HIV’s high genetic variability.

Results

Isolation of an HIV DIP

To first understand why DIPs have been elusive for lentiviruses, despite DIPs being readily isolated for other RNA viruses (11, 21) and the abundance of defective HIV genomes (22, 23), we used established in silico models of HIV (24, 25), which predicted that spontaneously generated DIPs would not expand in conventional serial-passage cultures as a result of dilution effects. The models further predicted that 50 to 100 days of nondilutive culturing would be required for a DIP to reach detectable levels (fig. S1C). Accordingly, we developed a nondilutive continuous-culture reactor (fig. S1, D and E) in which CD4⁺ target cells were replenished exclusively by cell division for up to 100 days. The reactor was infected with a green fluorescent protein (GFP)–expressing HIV (26), von Magnus–like fluctuations (27) were observed (Fig. 1A), and after ~100 days of culture, remaining GFP⁺ cells were isolated, and the viral DNA was sequenced. Proviral DNA analysis revealed a ~2.5-kb deletion in the HIV pol-vpr region (Fig. 1A), which resulted in the complete deletion of ribonuclease H, integrase, and the truncation of reverse transcriptase. A similar deletion arose in a parallel reactor, and, in both cases, the deletions were flanked by short identity regions, suggesting recombination events (fig. S1F).

Fig. 1. Long-term selection of an HIV DIP.

(A) Detection of an HIV deletion variant in a CEM cell bioreactor. (Left) Cell concentration and frequency of GFP⁺ cells by flow cytometry in a long-term bioreactor infected with replication-competent HIV-GFP (26). (Inset) PCR-amplified HIV proviral DNA from day 71 cells (lane 3: red arrowhead, expected band; blue arrowhead, ~2.5-kb deletion). (Right) Putative DIP construct with the location of the 2.5-kb deletion. D2, splice donor D2; SA, splice acceptor; IRES, internal ribosome entry site; RRE, Rev response element; nef, negative factor. (B) Conditional packaging assay: DIP VLPs were assembled in the presence or absence of lentiviral packaging plasmid, and transduced MT4 cells were titered by flow cytometry relative to a lentiviral GFP control vector. (C) Interference assay: HIV outgrowth in the presence or absence of the putative DIP. Cells stably transduced with single integrations of DIP or a mock-transduced control were infected with HIV-BFP, and the frequency of BFP⁺ cells was measured by flow cytometry. (Inset) Filtered supernatants were harvested at 3 days postinfection (dpi) and titered by flow cytometry. *P = 0.004 (4 dpi); P = 0.001 (5 dpi); P = 0.001 (7 dpi); **P =5× 10⁻⁶; Student’s t test. (D) Transmission assay: The fraction of putative DIP or HIV genomic RNA in supernatant VLPs (left) compared with the fraction of proviral DNA integrated in downstream transduced cells (right) at 3 dpi.

To test whether the ~2.5-kb deletion generated a DIP (fig. S1B), the deletion was cloned into HIV (right, Fig. 1A); the construct also encoded a long terminal repeat (LTR)–driven GFP reporter and a nonfunctional Tat protein, necessitating HIV superinfection to transactivate its expression (fig. S2, A and B). Firstly, to test for conditional replication, we verified that the deletion variant could only produce transduction-competent viral-like particles (VLPs) conditionally upon rescue (i.e., complementation in trans) with the deleted HIV proteins (Fig. 1B). Secondly, we quantified the deletion variant’s ability to interfere with HIV production; in single-round infections, the deletion variant reduced HIV titers by 94% (P =5× 10⁻⁶) (inset, Fig. 1C, and fig. S2C). Notably, cells with an integrated copy of the deletion variant were equally as permissive to incoming HIV infection in the first round as control cells without an integrated deletion variant (fig. S2C), and control GFP-expressing HIV-derived lentiviral vectors (fig. S3, A and B) generated no interference, indicating that the data cannot be explained by resistance to infection. Interference was further verified in alternate cell lines (fig. S4A) as well as in donor-derived human primary CD4⁺ T cells (fig. S4, B and C). Thirdly, to test whether the deletion variant mobilized in the presence of replication-competent HIV (e.g., through stealing packaging proteins from HIV), RNA was quantified in extracellular VLPs. Of the extracellular RNA, 60% was from the deletion variant, indicating a packaging advantage for the variant over HIV (Fig. 1D, left). These data verify that the deletion variant constitutes a bona fide DIP by (i) conditionally replicating only in the presence of HIV, (ii) interfering with HIV outgrowth, and (iii) mobilizing from infected cells.

Engineering the DIP into an R₀ > 1 TIP

Previous theoretical modeling (7, 8) stipulated that for a DIP to constitute a TIP, it must conditionally mobilize with R₀ > 1. However, in subsequent rounds of infection, the DIP accounted for only 30% of integrated provirus (Fig. 1D, right), indicating R₀ < 1. We hypothesized that poor DIP transmission could be due to loss of the central polypurine tract (cPPT), contained within the 2.5-kb deleted region, which enhances transduction efficiency (28–32) and appears critical for efficient R₀ > 1 transmission (33). To rationally enhance DIP transmission efficiency, the cPPT was reintroduced back into the deletion variant (Fig. 2A), which generated a twofold increase in VLP transduction efficiency (Fig. 2B). Further optimization to ablate additional protein expression (tat, rev, vpu, and env) enhanced transmission further (Fig. 2B and table S1). These engineered constructs (putative TIP-1 and TIP-2; table S1) retained DIP properties, including conditional replication, packaging, and interference (Fig. 2C and fig. S5, A to C), and generated significant interference in donor-derived human primary CD4⁺ cells (Fig. 2D), thereby constituting putative TIPs.

Fig. 2. Engineering the HIV TIP.

(A) Schematics of full-length HIV, DIP, and engineered TIP constructs; the cPPT was reintroduced in “TIP-1,” and additional coding sequences were ablated in “TIP-2” to reduce extraneous HIV protein expression (table S1). KO, knockout. (B) Transduction efficiencies of DIP, TIP-1, and TIP-2 VLPs (normalized to p24 capsid protein and assayed on MT-4 cells; mean of three biological replicates shown). *P = 0.007 and **P = 0.003 from Student’s t test. (C) TIP interference: Frequency of HIV-infected cells during long-term culture in the presence or absence of TIP-1–transduced cells. Cultures were seeded at a ratio of 1:1 of untransduced and TIP-1–transduced cells (the construct used, TIP-1/Δenv, has an additional env mutation; see table S1), infected with HIV-BFP, and monitored by flow cytometry. *P < 0.0006 from Student’s t test. (D) TIP-2 interference in donor-derived human CD4⁺ primary cells; 50% of cells were transduced with TIP-2 or lentiviral GFP control (33) infected with HIV-BFP, and new activated target cells were added. HIV-BFP⁺ cells were quantified by flow cytometry (mean ± SD of two biological replicates is shown). (E) Cells from the three-color assay were analyzed by flow cytometry 3 days post coculture to quantify mobilization of TIP (or control) from “producer” cells into mCherry⁺ “target” cells conditional on HIV (see red box for dual positives). (F) Measured R₀ values (mean ± SEM of three biological replicates). (G) Relative TIP R₀ by transmission in culture. Reactor was seeded with excess (~87%) target CEMs (mCherry⁺) and ~13% TIP-2 (GFP⁺), infected with HIV-BFP, and assayed by flow cytometry for double-positive BFP-mCherry⁺ and GFP-mCherry⁺ cells. Trendlines fit to HIV or TIP-2 viral loads [i.e., $VL\left(t\right)=C{e}^{rt}$, where r represents growth rate and C represents the initial viral load], and best-fit values are r_TIP = 0.33 and r_HIV = 0.054. (H) Quantification of interference from TIP packaging: TIP-2– or TIP-2^ΔΨ–transduced cells were infected with HIV-BFP, the supernatant was collected at 2 dpi, and levels of integration-competent HIV were titered in target cells by flow cytometry (mean ± SEM of three biological replicates; ns, nonsignificant; *P < 0.01; Student’s t test).

To directly measure the TIP-specific R₀ (R₀^TIP), a three-color assay was developed (Fig. 2E and fig. S6A), wherein cells were stably transduced with a single copy of TIP (encoding a GFP reporter), infected with full-length HIV [blue fluorescent protein (BFP)–tagged] to become “producer” cells, and subsequently cocultured with an excess of susceptible mCherry⁺ “target” cells. The assay quantified single-round HIV and TIP transmission into target cells from the fractions of double-positive BFP⁺-mCherry⁺ and GFP⁺-mCherry⁺ cells, respectively (e.g., double-positive GFP⁺-mCherry⁺ cells indicated TIP transmission from producer to target cells, with the fraction of double positives allowing calculation of R₀^TIP). To test whether the assay yielded physiologically relevant R₀ values, the R₀ of HIV alone was measured (i.e., in the absence of TIP) and found to be R₀^HIV = 24, consistent with upper estimates of the HIV R₀ in vivo (25, 34, 35). In parallel, the assay yielded an R₀^TIP = 12 (Fig. 2F). To verify R₀^TIP > 1, we quantified the initial slope of TIP expansion relative to an expanding HIV infection (i.e., R₀^HIV > 1); TIPs exhibited substantial transmission into target cells, showing an exponential rate of increase ~sixfold faster than HIV over multiple rounds of infection (Fig. 2G). Notably, TIP-transduced cells did not show altered survival (either enhanced or diminished) compared with infected, mock-transduced cells (fig. S6, B and C), indicating that TIP expansion is not the result of purifying selection. Also consistent with the R₀ > 1 mechanism, in multi-week reactor cultures, the TIP significantly and durably reduced HIV viral loads, protecting cultures from exponential HIV growth for at least 40 days (Fig. 2C). In further support of the R₀ >1 mechanism of interference, deletion of the TIP packaging signal (Ψ) (36) significantly blunted the interference effect by ~30% (Fig. 2H and fig. S7, A and B), as expected from the partial penetrance of the Ψ packaging signal (37) and consistent with TIP RNAs competing intracellularly for viral Gag polyprotein precursor (7, 8, 10).

In vivo analysis of TIP efficacy

To test TIP efficacy in vivo, we first performed pilot studies in two humanized mouse models, the peripheral blood mononuclear cell (PBMC) NSG mouse model (38) and the longer-duration CD34⁺ NSG model (39), commonly used for preclinical evaluation of new HIV antiviral candidates (40, 41). In these models, immuno-deficient (NSG) mice, engrafted with either human CD4⁺ T cells or CD34⁺ progenitor cells, were infected with HIV and subsequently administered TIP lentiviral vector through direct intravenous (IV) injection 5 days following HIV challenge to evaluate TIP efficacy after a single-injection administration (fig. S8A). For these in vivo studies, we attempted to optimize the HIV TIP candidate first by ablating its Gag expression through a frameshift in gag-pol, which unfortunately blunted TIP interference (fig. S8B). Consequently, we instead incorporated two smaller deletions identified in a recent library screen (33)—one in the 5′ end of gag that allows ~100 amino acids of Gag to be expressed and the other in the 3′ end of env (fig. S8C and table S1)—both of which led to enhanced TIP efficacy (fig. S8D). In both mouse models, HIV TIP treatment generated significant reductions in HIV viral load (Fig. 3A and fig. S8E). The reduction in HIV viral load was still apparent when viremia was normalized to CD4⁺ T cell counts, indicating that the effect could not be explained by target-cell depletion (fig. S8, F and G).

Fig. 3. TIP intervention in humanized mice and rhesus macaques.

(A) Human PBMC model: NSG mice were engrafted with human cells, challenged with HIV, and then administered a single IV injection of HIV TIPs (~10⁶ TU/kg) or mock at 5 days postinfection. HIV RNA plasma viral loads in NSG mice treated with TIP (n = 8) or HIV (control; n = 4) at days 5, 12, and 19 post HIV infection. **P = 0.005. LOD, limit of detection. (B) Rhesus macaque model: Kaplan-Meier survival analysis of TIP-treated animals (blue, n = 6) compared with controls (red, n = 4) [Log-rank (Mantel-Cox) test] and historical controls (gray, n = 21). (C) Quantitative PCR of SIV viral RNA (SIV gag) in peripheral blood plasma; Blue = TIP-treated animals, red = control animals (each animal represented by a unique symbol). (D) Reversetranscribed TIP DNA in peripheral blood mononuclear cells from day 1 post TIP intervention. (E) Quantitative PCR of TIP RNA copies in peripheral blood plasma.

To more definitively evaluate in vivo efficacy, we used a highly pathogenic simian immunodeficiency virus (SIV) nonhuman primate (NHP) model (fig. S9A), where infant rhesus macaques are infected with the recombinant virus SHIV-SF162P3 (an SIVmac239 virus encoding HIV env) (42–44). This infant NHP model exhibited exceptionally rapid progression to an AIDS-like disease and clinical endpoint with a 50% survival time of ~15 weeks (Fig. 3B). As in previous immune-therapy studies in this model (42–44), animals were administered the intervention 24 hours prior to virus challenge. Notably, this intervention schedule mimics the likely clinical testing scenario—i.e., analytical treatment interruption (ATI), where antiviral candidates are administered prior to viral rebound—as pre–steady state viral kinetics during early acute infection closely approximate viral rebound kinetics following ATI (2, 3).

Testing in the NHP model required development of an analog SIV TIP based on SIVmac239 (fig. S8C and table S1). The analog SIV TIP (pseudotyped with SIV env) was first tested in vitro to verify its ability to suppress viral outgrowth in primary rhesus splenocytes (fig. S9, B to D). Then, infant rhesus macaques were administered the SIV TIP [3 ml of ~10⁷ transducing units (TU)/ml] through a single IV injection, challenged with SHIV-SF162P3 (100 IU) through the IV route (24 hours later), and followed for ~30 weeks (~7 months) or until clinical endpoint (fig. S9A). Similar to historical controls, infected (untreated) animals exhibited rapid pathogenesis (Fig. 3B). Notably, Kaplan-Meier survival analysis showed that TIP-treated animals exhibited a significant survival advantage (Fig. 3B) with no observable outward signs of simian AIDS or gross pathogenesis before the scheduled study termination at 30 weeks postinfection except for one TIP-treated outlier animal that reached clinical endpoint at 24 weeks.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis of viral loads in peripheral blood plasma (Fig. 3C) showed that almost all animals reached peak viremia of ~10⁷ copies/mL at 2 weeks postinfection. Subsequently, three of four untreated animals established set-point viral loads of ~10⁸ copies/mL, consistent with previous reports (45, 46). By contrast, five of six TIP-treated animals exhibited dramatic reductions in set-point viral load, settling at ~10⁴ SHIV RNA copies/mL, and one TIP-treated animal exhibited extremely low, intermittent viral load (i.e., a ~3log₁₀ to 5log₁₀ reduction). In these TIP-treated animals, the set-point reduction was durable and maintained for the duration of the 30-week study.

Following TIP injection, reverse-transcribed TIP DNA was robustly detected in the PBMCs of TIP-treated animals (Fig. 3D). As opposed to humans, in which longitudinal analysis of integrated proviral DNA is standard, in rhesus macaques, such analysis is exceptionally technically challenging (47), in part owing to their smaller size, which limits the number of CD4⁺ T cells that can be collected during a blood draw. However, consistent with R₀^TIP > 1, extracellular TIP RNA was robustly detected in the blood plasma of TIP-treated animals (at 10⁴ to 10⁵ copies/mL) throughout the entire 30-week study (Fig. 3E) despite the single administration of TIPs. Assay specificity was confirmed by testing control animals for TIP, and neither TIP DNA nor RNA were detected in untreated, HIV-infected animals (Fig. 3, D and E).

To physiologically assay the immune health of the animals, we used an established approach (42–44) that tests for the ability to mount an adaptive immune response (i.e., seroconversion) to the Env protein. The inability of animals to seroconvert is characteristic of loss of immune function as a result of viral pathogenesis (i.e., high viremia leading to rapid progression to AIDS). As expected, untreated animals exhibited poor seroconversion, inversely correlated with viral load, consistent with the less-developed immune systems of these infants (44). By contrast, TIP-treated animals exhibited greatly improved seroconversion (Fig. 4A), concordant with their reduced viral loads. CD4⁺ T cell levels exhibited modest declines in both TIP-treated and untreated animals (fig. S9, E and F), in agreement with previous findings that CD4⁺ T cell levels have limited correlation with disease progression (48–54).

Fig. 4. Immune response, in situ tissue analysis, and quantitative modeling of SHIV kinetics.

(A) Seroconversion ability of TIP-treated versus control animals (ELISA for HIV Env). (B) Representative images of SHIV RNA expression within lymphoid tissue by RNAscope in situ hybridization (top scale bar, 500 μm; bottom scale bar, 100 μm). (C) Correlation between qPCR of SIV gag RNA in plasma and the quantified proportion of SHIV RNA⁺ cells within lymph nodes from RNAscope. ****P = 0.0001. (D) Best fits (nonlinear least-squares regression) of mathematical models to SHIV viral load data from a representative control animal (top) and TIP-treated animal (bottom). Data from control animals were fit to the basic ODE model (gray), whereas data from TIP-treated animals fit to both the basic model and TIP model (blue). (E) Model fitting summary: Ratio of peak to steady-state SHIV viral load data (* symbol) together with mathematical model fits from Markov chain Monte Carlo sampling (○ symbol) and the best fits from Latin-Hypercube sampling (□ symbol). Gray symbols represent the basic model of viral dynamics, whereas blue symbols represent the model with TIPs incorporated.

To determine whether plasma viremia was representative of viral replication in lymphoid tissue, where the majority of viral replication occurs, we assessed SHIV RNA in lymph nodes through quantitative spatial imaging, i.e., RNAscope in situ hybridization (55, 56) (Fig. 4B and fig. S10, A to H). SHIV RNA in lymph nodes of TIP-treated animals was substantially reduced compared to untreated animals and, across all animals, there was high correlation (R² = 0.89) between SHIV levels within lymph nodes and peripheral blood plasma (Fig. 4C) as well as SHIV proviral DNA levels in lymphoid cells (fig. S10I).

To rule out alternate mechanisms of action, we examined full-length SHIV sequences and viral kinetics in the animals. To test the hypothesis that reduced SHIV viral loads could be due to evolutionary selection for an attenuated SHIV variant, full-length SHIV sequencing was performed from endpoint samples of all animals. No large deletions or rearrangements were detected in SHIV RNA (fig. S11) (GenBank accession nos. PP597405 to PP597522) or proviral DNA (GenBank accession nos. PP646066 to PP646153), and the majority of evolution (i.e., single-nucleotide polymorphisms) was observed in env, consistent with lifetime extension of TIP-treated animals enabling more viral replication as well as immune pressure from seroconversion (fig. S12). Because one animal (40471, figs. S11 and S12) exhibited a longer branch than others, we performed full-length alignment analysis with the parental molecular clone but did not observe evidence of recombination (data S1).

To test the hypothesis that TIP intervention may have reduced the effective SHIV inoculum, we analyzed viral load kinetics using an established analytic approach in which the viral peak/set point ratio (57–59) is exploited to discriminate underlying physiological mechanisms. As expected, the basic model of in vivo viral dynamics (24, 25) was sufficient to fit viral kinetics in untreated animals (Fig. 4, D and E, and fig. S13), but not TIP-treated animals (under any parameter combinations), consistent with reports (58, 59) that fitting peak/set point ratios of ~10-fold or greater require mechanisms that dynamically respond to viral load. Notably, the similar magnitude and timing of peak viremia levels between TIP-treated and untreated groups (Fig. 3C) is inconsistent with the results being explained by a reduced SHIV inoculum in the TIP-treated group. Moreover, no evidence of an increased inflammatory response, which correlates with more rapid disease progression in rhesus macaques (60) and people living with HIV who are long-term nonprogressors (61), was detected in TIP-treated animals (fig. S14), further suggesting that static innate mechanisms cannot explain the reduced viral loads in TIP-treated animals. Notably, an extension of the basic viral dynamics model incorporating TIPs (7), which dynamically respond to viral load as obligate molecular parasites, was sufficient to fit the viral kinetics of TIP-treated animals (Fig. 4, D and E, and fig. S15A). The modeling indicated that the single nonresponding TIP-treated animal exhibited substantially higher T cell turnover, consistent with previous predictions that such increased turnover could limit TIP efficacy (7). The measured in vivo viral set-point viral load matched analytically predicted steady state viremia levels, suggesting early and durable establishment of the viral set point in TIP-treated animals and that longer-term studies would yield little added viral kinetics information (fig. S15B).

Extrapolating efficacy of HIV TIPs to the human context

We next asked whether TIPs could provide any benefit in the context of viral latency, given the noncurative nature of ART. Our previous studies argued that lentiviral vectors obligately establish latency and will reverse latency when Tat is delivered in trans (62–64). Thus, we hypothesized that TIP-transduced cells would survive under ART and would reverse latency together with HIV upon interruption of ART to thereby suppress HIV outgrowth. On the basis of established latency quantification protocols (65), we developed an in vitro viral outgrowth assay (fig. S16A). We found that TIP genomes were maintained for at least 26 days under ART (Fig. 5A and fig. S16B), and upon ART cessation, TIPs rebounded with HIV significantly interfering with HIV outgrowth (Fig. 5, B and C) and protecting cells from cytopathic effects (fig. S16C).

Fig. 5. Effects of TIPs on HIV latency reversal, CD4 protection, and extrapolation to human intervention.

(A) Latency assay: qPCR quantification of integrated TIP DNA in Jurkat cells infected with HIV and/or transduced with TIP (optimized HIV-TIP, left). qPCR quantification of integrated HIV DNA (right) in cells infected with HIV and/or transduced with TIP during extended ART treatment (day 26) [normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH)]. (B) Latency reversal upon ART cessation: fluorescence-activated cell sorting (FACS) analysis of Jurkat cells following ART cessation (+TNF-α) in the HIV control (left) versus TIP-treated samples (right). A.U., arbitrary units. (C) Percentage of HIV⁺ cells reactivated spontaneously following ART cessation (background) and in presence of TNF-α in both the HIV control versus the TIP-treated sample. *P = 0.05. (D) CD4 protection assay: FACS analysis of human donor–derived primary CD4⁺ T cells infected with HIV and transduced with either TIP (optimized HIV-TIP) or a control HIV-derived lentiviral vector, both expressing GFP. (E) HIV infection spread as assayed by CD4 depletion in the HIV-TIP– and control-treated samples on days 7 and 14. (F) qPCR quantification of integrated TIP DNA in donor-derived human primary CD4⁺ T cells from days 3 to 14. (G) Predicted reduction in HIV set-point viral load following TIP intervention at physiological HIV set point values (based on model extrapolation from rhesus macaque data).

We quantified TIP antiviral efficacy in human donor–derived primary T lymphocytes (fig. S16D). Briefly, activated primary CD4⁺ T cells were transduced with TIP or a control vector, infected with HIV, and then followed for 21 days through coculture with additional T cell targets. As in previous studies (39, 66–70), HIV infection level was indirectly quantified by depletion of CD4 cells (Fig. 5, D and E), and depletion was consistent with similar humanized animal models (71, 72). HIV-TIP transduction protected the CD4 population, whereas TIPs appeared to be maintained both in terms of GFP-reporter expression (Fig. 5D) and genomic DNA (Fig. 5F). HIV-TIP transduction did not substantially affect proliferation of primary CD4⁺ T cells, and cell expansion was consistent with protection from HIV killing (fig. S16E).

We also quantified integrated TIP antiviral efficacy on samples from donors living with HIV (fig. S17A). Since the most likely clinical use scenario for TIP intervention would be to benefit individuals in resource-limited settings who cannot access ART long term, PBMCs were obtained from untreated persons living with HIV. To stimulate HIV outgrowth, PBMCs were cocultured with CD4⁺ target cells transduced with TIPs or a control GFP vector. TIP treatment significantly reduced HIV outgrowth compared with the control vector (fig. S17A), suggesting that TIP intervention can be effective against patient-derived viruses.

Because SIV and SHIV in rhesus macaques produces viral loads that are higher than those in HIV-infected humans (i.e., 10⁷ to 10⁸ copies/mL in rhesus macaques versus 10⁴ to 10⁵ copies/mL in humans), we asked what effect HIV TIPs might have given the lower HIV set point in humans. To predict the magnitude of HIV set point reduction in humans, we parametrized established mathematical models based on the NHP data. The resulting simulations indicated that TIPs could generate a 2log₁₀ to 3log₁₀ reduction in HIV set point in humans, resulting in viral loads <10³ copies/mL (Fig. 5G), which could substantially reduce viral transmission (8, 73). Overall, these results suggest that TIPs could represent a single-administration HIV therapeutic with the potential to limit viral burden and consequently reduce viral transmission below the World Health Organization–defined transmission threshold (74).

Mechanistically, how TIPs achieve R₀ >1 in vivo given that only a minority of target cells (~1%) are productively HIV infected at any given time during set point viremia (25, 75) is best explained by standard viral dynamical models and established recombination frequencies (76, 77). These mathematical models (24) indicate that the rapid turnover of HIV-infected CD4⁺ T cells exposes a substantial fraction of cells to HIV infection within a few weeks, and thus, CD4⁺ T cells with TIP integrations have a substantial probability of being superinfected by HIV (7–10). The models are consistent with observations that 5 to 10% of infected cells harbor multiple proviral integrations during chronic HIV infection (78) and reported high in vivo recombination rates observed for HIV (76, 77). Also in agreement with these models, previous gene therapy strategies that were dependent on HIV superinfection of vector-integrated T cells temporarily reduced viral load in clinical trials (16), although the therapeutic effect diminished over time owing to a lack of vector persistence (i.e., R₀ < 1) (79).

Although no immunotoxicity from TIP intervention was observed in NHPs, it is prudent to consider potential TIP genotoxicity. Historically, lentiviral gene therapies that integrate in T cells have not been associated with increases in cancers (79–81), and HIV-associated B cell malignancies are known to result from T cell depletion rather than genotoxicity (62) except in very rare cases (82). However, a recent report (83) indicates 22 secondary CD8⁺ T cell cancers in >27,000 chimeric antigen receptor (CAR) T cell–treated individuals. As that report notes, it remains unclear if these secondary cancers are due to lentiviral integration (in three cases, the CAR gene was detected in the malignant clone) or chemotherapy, and, notably, the incidence of secondary cancers following chemotherapy for lymphomas is substantially higher (84–86).

Although there have been reports of CD4⁺ T cell cancers in persons living with HIV, there are at least two speculations as to why CD4⁺ T cell cancers are rare in HIV-infected patients despite the extensive insertional mutagenesis in CD4⁺ T cells evidenced by the large reservoir of defective proviruses (22, 65): (i) These T cell cancers require the accumulation of multiple additional nonvirus induced mutations (87), or (ii) HIV’s tropism to replicate in activated CD4⁺ T cells may inherently target and control any transformed CD4⁺ T cells. Although TIPs could sample more integration sites to increase the risk of transformation, they lack Tat transactivation likely required for LTR-driven transformation (88), and so the above speculations would likely still act to limit TIP-mediated expansion of transformed cells.

Recombination between HIV and TIPs (or extant defective proviruses) is also a potential concern but was not observed in plasma virus (fig. S11), perhaps because TIP internal deletions substantially reduce recombination efficiency (10). Notably, putative HIV-TIP recombinants could only be conditionally generated in the presence of replicating HIV, and, in the improbable extrema where TIPs regain replication competence, these putative recombinants would be conceptually equivalent to conventional loss-of-efficacy mutations for therapeutics (89), i.e., consistent with equal or lower pathogenicity, assuming that the putative recombinant has equivalent tropism and equivalent drug, immune, and interferon sensitivities.

Evolutionarily, it is unclear why HIV DIPs or TIPs have not been detected in vivo despite abundant defective HIV genomes (22, 65), although the TIP deletion arrangement provides clues. Specifically, accumulation of DIP-like deletions requires R₀ > 1, but for HIV, this requires a complex series of multiple recombination-rearrangement events in which the cPPT-flanking sequences are deleted, but the cPPT itself is retained or reinserted (65).

From the translational perspective, future studies will be needed to evaluate the clinical potential of TIPs. For example, TIPs lacking GFP must be developed, though preliminary analyses indicate that such ΔGFP TIPs maintain equivalent efficacy (fig. S17B). Notably, TIPs will need to be tested in more realistic clinical intervention scenarios such as ATI studies (90) and assessed for genotoxicity, disease progression, and inflammatory signatures at later time points to determine whether TIP-treatment results in delayed disease HIV-1 long-term nonprogressors (61). It should be noted that TIPs are unlikely to be a sterilizing cure owing to their conditional replication requirement. However, as single-administration lentiviral vectors, TIPs can establish latency and reactivate upon HIV rebound (Fig. 5, A to C), thereby echoing calls for an HIV “functional cure.”

Methods summary

Putative DIPs were identified from a continuous culture reactor where CEM CD4⁺ cells were infected with replication-competent reporter virus (HIV-GFP or HIV-BFP) at multiplicity of infection (MOI) ~0.01 to 0.1 (virions/cell). Cells were followed over time through flow cytometry for GFP expression and cell viability to detect Von Magnus–like fluctuations. DNA was extracted from cells at day 71, and HIV DNA was amplified by PCR, leading to identification of the ~2.5-kb deletion. DIP VLPs were packaged on 293T cells and yield reduction, and transduction efficiency were assayed on MT4 cells.

TIP engineering was carried out, reintroducing the cPPT sequence into the DIP by restriction cloning to make TIP-1, and this construct was modified to ablate tat/rev/vpu/env expression resulting in TIP-2. Further optimization by introduction of two additional deletions (one in gag, the other in env) resulted in the optimized HIV-TIP construct (table S1). TIP transduction efficiencies were determined by titering packaged TIPs on MT4s in the presence of replication-competent HIV and normalizing titer units (TU) per ng of capsid protein.

For testing the efficacy of TIPs in CD4⁺ primary cells, CD4⁺ cells were purified from healthy donors activated with CD3/CD28 Dynabeads, then transduced with TIP and infected with HIV-BFP; cultures were supplemented with activated CD4⁺ cells if required. Cells were monitored by flow cytometry for TIP-GFP and/or HIV-BFP expression over time. Similarly, TIP efficacy was tested in cell lines (CEMs, MT4s, and Jurkats), either through coinfection with TIP and HIV or pretransducing cells with TIP followed by HIV infection. Cultures were maintained by replacing media and assayed by flow cytometry. HIV and TIP proviral integration was assayed via qPCR using HIV and/or TIP specific primers (table S2).

R₀ quantification by three-color assay utilized stable polyclonal TIP-transduced cell lines with single TIP integrations created by transducing CEM cells with TIP at low MOI. The TIP stable cell lines (“producer” cells) were infected with HIV-BFP and after a single round of infection, cocultured with “target” cells, i.e., doxycycline-inducible Tat-mCherry cells, at a ratio of 1:20. After each single round infection, cells were subjected to flow cytometry to determine the percentage of BFP⁺ cells in the “producer” cells and the BFP⁺ and GFP⁺ cells in the “target” cells. These values were then used to calculate R₀ values for TIP, as detailed in the materials and methods.

TIP in vivo efficacy was tested in two humanized mouse models: the PBMC NSG mouse model and the CD34⁺ NSG model. For both model’s mice were challenged with HIV at day 0 and treated with HIV-TIP lentiviral vector VLPs at day 5 through intravenous injection. Animals were bled weekly, and plasma viremia was assayed through qRT-PCR using HIV-1 specific primer-probes (table S2).

The SIV-TIP (table S1) was constructed from the SIVΔnef-eGFP molecular clone. Deletions were analogous to the optimized HIV-TIP and introduced using standard cloning methods. The SIV-TIP was pseudotyped with the SIV-specific envelope SIVmac251.6. Preliminary in vitro testing was carried out in rhesus macaque splenocytes transduced with SIV-TIP and then infected with SHIV SF162P3. At7 days following infection, both cells and supernatant were collected, the cells were stained for p27 and CD45 and subjected to flow analysis. The supernatant was diluted and titered on TZM-bl cells by assaying for luciferase activity.

SIV-TIP efficacy was tested in infant rhesus macaques. 6 out of 10 animals were administered TIP 24 hours prior and all animals were then challenged with SHIV-SF162P3. Animals were bled weekly and followed until clinical end point. Plasma viremia and TIP levels were determined by RT-qPCR and digital droplet PCR (ddPCR), respectively, using SIVgag primers and GFP-specific primer probes (tables S2 and S3). PBMCs were isolated from plasma at day 0 (a day post TIP administration, preinfection), and DNA was extracted and subjected to qPCR to detect TIP-GFP DNA. Viral RNA was isolated from plasma, reverse transcribed and subjected to near full-length sequencing and phylogenetic analyses using Geneious software. Plasma viral loads were fit to the mathematical models of virus dynamics (as detailed in Supplementary text section 4)

Immune response (i.e., seroconversion) was assayed by enzyme-linked immunosorbent assay (ELISA) of plasma samples using recombinant HIV-1 SF162 gp140 trimer to determine binding titers. SHIV viral RNA quantification in end-point lymphoid tissues was determined by RNAScope as detailed in the materials and methods (see supplementary materials section “Seroconversion, immunohistochemistry, RNAscope in situ hybridization, and quantitative image analysis”).

TIP efficacy in CD4⁺ cells isolated from persons living with HIV [i.e., persons who inject drugs (PWID)] was tested similar to the above efficacy testing in CD4⁺ primary cells. Patient CD4+ cells were isolated, activated and transduced with TIP, fresh activated CD4⁺ cells from PWID and pre–TIP-transduced CD4⁺ cells from healthy donors were added to propagate infection, culture supernatant was sampled over time and measured for p24 to assay HIV knockdown.

TIP latency, reversal, and associated reduction in HIV rebound was tested using a modified viral outgrowth assay: Jurkat cells were transduced with TIP, infected with HIV-BFP, and treated with Darunavir and Raltegravir until viral suppression was achieved. Cells were then washed to remove Darunavir and Raltegravir, diluted and cocultured with a fixed amount of MT4 target cells, and treated with tumor necrosis factor (TNF)–α to induce reactivation of latent HIV and TIP. Cultures were monitored daily, split as required, and analyzed by flow cytometry to measure the percentage of HIV-BFP and/or HIV-BFP TIP-GFP infected cells upon reactivation. To assay integration, cells were extracted at a given time point, the DNA was isolated, and then subjected to qPCR with HIV and/or TIP-specific primers (table S2).

For the CD4⁺ depletion or protection assay, CD4⁺ cells from healthy donors were isolated, activated, and TIP-transduced. A day later, cells were infected with HIV. Fresh primary T cell targets were added at day 7 to propagate the culture. Cells were extracted at days 3, 7, and 14 and analyzed by qPCR for TIP integration.

Data and materials availability:

The data that support the findings of this study are available from the corresponding author upon request. All biological materials are available from the corresponding author. Sequencing data has been deposited in GenBank (accession nos. PP597405 to PP597522 and PP646066 to PP646153). Custom code for mathematical modeling is available on GitHub at https://github.com/khetanneha/HIV-SIV-TIP-Modeling and archived on Zenodo along with all the experimental raw data (91). An MTA for the HIV Reagent program is available on request from the corresponding author under an agreement with J. David Gladstone Institutes.