Two-Drug Combination Locks HIV Transcription into Durable Silence

People living with HIV face a paradox of modern medicine: antiretroviral therapy can suppress the virus in the blood to undetectable levels, yet it cannot eliminate the small population of reservoir cells where the virus persists. When treatment stops, these cells reignite viral production and the infection rebounds. A new study published in iScience by Zichong Li, Francisco J. Zapatero-Belinchón, Patrik Geleziunas, and Melanie Ott reports a strategy that goes beyond simply suppressing HIV transcription and instead pushes the provirus into a durably silent state, one that persists even after the drugs are withdrawn. The work offers a detailed mechanistic proof of concept for the so-called block-and-lock approach to an HIV cure.

The central target of the study is a step in the HIV life cycle that current antiretroviral regimens do not touch: transcription. Once HIV integrates into the host genome, its promoter, the long terminal repeat, must recruit RNA polymerase II to produce viral RNA. Polymerase initiates transcription but typically pauses around nucleotide +59, where the nascent RNA forms a stem-loop structure called TAR. Productive elongation past this pause requires the viral protein Tat, which binds TAR and recruits the positive transcription elongation factor b, or P-TEFb, a kinase complex composed of CDK9 and cyclin T1. Because P-TEFb is indispensable for generating full-length HIV RNA, inhibiting CDK9 is a powerful way to shut down viral transcription. The catch is that P-TEFb is equally essential for the expression of normal human genes, and global CDK9 inhibition produces significant toxicity.

To sidestep this problem, the researchers built on recent cancer biology findings showing that pharmacological activation of the protein phosphatase PP2A synergizes with CDK9 inhibition to repress transcription. Their own earlier CRISPR interference screens had also identified subunits of the integrator complex, a multi-protein machine that promotes premature termination of transcription, as regulators of HIV latency. The team therefore paired AZD4573, a highly selective CDK9 inhibitor, with SMAP-2, a small-molecule agonist of PP2A that activates integrator function. The logic is elegant: CDK9 drives elongation, while PP2A-associated integrator drives termination, so the two drugs attack the HIV transcription cycle from opposite directions.

The first evidence came from the J-Lat A2 cell line, a Jurkat-derived model carrying a GFP reporter under the control of a latent HIV long terminal repeat. When the researchers depleted the integrator scaffolding subunit INTS2 by RNA interference, latency was only modestly reversed. But co-depleting INTS2 together with the elongation factors FTSJ3 or PSMD3 produced synergy well beyond additive expectations, as defined by the Bliss independence model. Knockdown of FTSJ3 and PSMD3 together, by contrast, was merely additive. These genetic experiments established that elongation and termination factors cooperate to keep HIV silent, and that neither pathway alone is sufficient to enforce deep repression.

The pharmacological version of this principle proved even more striking. Screening a matrix of AZD4573 and SMAP-2 concentrations with live-cell imaging, the team identified dose pairs that synergistically suppressed HIV reactivation triggered by PMA and ionomycin without increasing annexin V staining, a marker of cell death. At the selected doses of 5 nanomolar AZD4573 and 5 micromolar SMAP-2, either drug alone reduced reactivation from roughly 74 percent of cells to 14 percent and 55 percent, respectively. The combination drove reactivation down to just 4 percent, well below the 10.4 percent predicted if the drugs’ effects were simply additive. Reverse transcription quantitative PCR then revealed the mechanistic division of labor: AZD4573 selectively reduced long, elongated HIV transcripts while sparing short initiated ones, consistent with its block on elongation, whereas SMAP-2 reduced short transcripts, indicating diminished early transcriptional output. Together the drugs suppressed both transcript classes.

Dissecting the integrator complex further, the researchers knocked down individual subunits and found that SMAP-2 lost its suppressive activity when the scaffolding subunit INTS2 or the PP2A adaptor INTS6 were depleted, but retained activity when the endonuclease subunit INTS11 was knocked down. This suggests the structural and PP2A-associated modules of integrator are the critical mediators of SMAP-2’s effect, although the authors caution that incomplete knockdown and the absence of direct nascent RNA cleavage measurements leave the role of INTS11’s catalytic activity unresolved.

The most clinically relevant experiments moved into primary human CD4+ T cells, latently infected with a single-round HIV-GFP reporter virus. After one day of pretreatment with the drug combination, activation of the T cells through CD3 and CD28 antibodies, a potent latency-reversing stimulus, was blunted by approximately 80 percent. Remarkably, the suppression endured: even after six additional days in drug-free medium, reactivation remained about 50 percent lower than in vehicle-treated controls. AZD4573 alone, by contrast, lost its effect within three days of withdrawal. This persistence is the defining feature of a true block-and-lock strategy, and it distinguishes the combination from CDK9 inhibition alone.

Chromatin immunoprecipitation explained why the silence held. SMAP-2 transiently increased recruitment of the integrator subunit INTS11 to the Nuc-1 region of the HIV long terminal repeat, but only when combined with AZD4573 did integrator occupancy persist through day seven. Sustained integrator binding correlated with prolonged reductions in total RNA polymerase II and its Ser5- and Ser2-phosphorylated forms at the promoter. At the same time, activating histone marks such as acetylated H3, acetylated H4, and H3K4me3 declined, while the repressive marks H3K9me3 and H3K27me3 accumulated. The combination thus converts the viral promoter into a repressive chromatin state, removing the paused polymerase that would otherwise act as a bookmark inviting future transcription. The authors note that the apparent coexistence of two repressive methylation marks likely reflects averaging across different cell subsets rather than co-occupancy of the same nucleosome.

Safety and specificity data supported the approach’s feasibility. The drug combination did not induce the T cell activation markers CD25 and CD69, in sharp contrast to anti-CD3/CD28 stimulation, and actually dampened activation marker induction when added alongside T cell receptor stimulation. CFSE staining showed no proliferation, and short-term exposure was tolerated, although cell death began to rise after 48 hours of continuous treatment. Transcriptome profiling of primary CD4+ T cells revealed context-dependent host effects: in resting cells, most differentially expressed genes were unannotated coding loci or long non-coding RNAs, with the canonical P-TEFb target MYC downregulated, while activated T cells showed far fewer changes, including reduced HEXIM1 and, intriguingly, upregulation of SLFN5, a transcriptional co-repressor known to silence HIV through histone methylation. SLFN5 induction may therefore contribute to the reinforced lock.

Perhaps the most translational finding concerns substance use. Cocaine and morphine are known to enhance HIV transcription through pathways involving NF-κB and P-TEFb, and recreational drug use is common among people living with HIV. When latently infected primary cells were exposed to 2 micromolar cocaine or morphine, the drug combination largely preserved both its blocking and locking effects under cocaine, while morphine partially weakened but did not abolish the locking effect. Chromatin assays showed that morphine reduced INTS11 occupancy at the viral promoter and increased polymerase loading, mirroring the effect of INTS2 knockdown, and morphine’s attenuating influence disappeared in INTS2-depleted cells. This identifies the integrator pathway as a specific node through which opioids can undermine HIV transcriptional control. The study’s limitations are acknowledged: the locking effect was monitored for only six days, drug abuse experiments used a single concentration in vitro, and donor demographics could not be assessed. Even so, the work establishes that simultaneously inhibiting P-TEFb and activating the PP2A-integrator axis can lock HIV into a silent, epigenetically repressed state that survives drug withdrawal and resists physiological challenges, laying a concrete foundation for combination silencing therapies aimed at long-term, drug-free HIV remission.

Subject of Research: Dual pharmacological targeting of P-TEFb and the PP2A-integrator axis to enforce durable silencing of HIV transcription

Article Title: “Locking” CDK9 inhibition by activating integrator function in HIV transcription

Article References: Li, Z., Zapatero-Belinchón, F. J., Geleziunas, P., & Ott, M. (2026). “Locking” CDK9 inhibition by activating integrator function in HIV transcription. iScience, 29(11), Article 117894. https://doi.org/10.1016/j.isci.2026.117894

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117894

Keywords: HIV, latent reservoir, block-and-lock, CDK9, P-TEFb, PP2A, integrator complex, AZD4573, SMAP-2, transcriptional silencing, epigenetics, Tat