A transient heritable memory regulates HIV reactivation from latency

Abstract

Reactivation of human immunodeficiency virus 1 (HIV-1) from latently infected T cells is a critical barrier to cure patients. It remains unknown whether reactivation of individual latent cells occurs stochastically in response to latency reversal agents (LRAs) or is a deterministic outcome of an underlying cell state. To characterize these single-cell responses, we leverage the classical Luria-Delbrück fluctuation test where single cells are isolated from a clonal population and exposed to LRAs after colony expansion. Data show considerable colony-to-colony fluctuations with the fraction of reactivating cells following a skewed distribution. Modeling systematic measurements of fluctuations over time uncovers a transient heritable memory that regulates HIV-1 reactivation, where single cells are in an LRA-responsive state for a few weeks before switching back to an irresponsive state. These results have enormous implications for designing therapies to purge the latent reservoir and further utilize fluctuation-based assays to uncover hidden transient cellular states underlying phenotypic heterogeneity.

Keywords: Computational Molecular Modeling; Virology.

Graphical abstract

Figure 1

Heritable versus random reactivation response of HIV-1 latency to an external stimulus (A) HIV-1 infection results in two phenotypes: active replication (red) where the infected cells produce new virions and latent infection (gray) where infected cells do not produce virions. Latently infected cells can reactivate and initiate virion production through cellular and environmental fluctuations. (B) The cell line used in our investigation, the Jurkat latency model (JLat), is latently infected with a full-length HIV-1 gene circuit, with a deletion of env reading frame and a replacement of nef reading frame with a GFP element. (C) Histogram of single-cell GFP fluorescence (measured with flow cytometry) from a JLat 9.2 bulk culture, after a 24-h tumor necrosis factor alpha (TNF-α) treatment. TNF-α is a potent activator of the HIV-1 LTR promoter and was administered at 10 ng/mL concentration. The histogram shows that there is a distinct bimodal distribution of response to TNF-α perturbation within a clonal population of JLat. (D) We propose the following Luria-Delbrück experimental design to investigate if single-cell JLat response to TNF-α perturbation is random or heritable. A clonal population of JLat 9.2 was sorted into single cells using FACS and cultured individually into colonies. The colonies were subject to 10 ng/mL of TNF-α and the percentage of reactivated cells was measured using flow cytometry after 24 h of treatment. If JLat responds to TNF-α randomly, their responsiveness to TNF-α is determined at the time of TNF-α addition and thus would show low colony-to-colony variation. In contrast, if the response is heritable, then colony behaviors would be influenced by the parent cell resulting in large variations between colonies. (E) The distribution of reactivation percentage would be centered on the unsorted average if the response is random, whereas the heritable model will result in a wide, skewed distribution for reactivation percentage across colonies. (F) Histogram of reactivation percentage of 14 different JLat 9.2 colonies after a 24-h TNF-α treatment. Colonies were cultured for 5 weeks after FACS and were grown from single cells sorted from a bulk JLat 9.2 clonal population. The resulting distribution coincides with the heritable response prediction and confirms that JLat cells possess heritable memory of responsiveness to TNF-α perturbation. Inset: The noise in percent reactivation (quantified using the coefficient of variation CV) is significantly higher than the control noise floor as measured by fluctuations in percent reactivation across unsorted JLat populations (p < 0.005). Error bars in the inset represent the standard error of CV. See also Figures S1 and S2.

Figure 2

Experimental results from long-term culturing of single-cell JLat colonies reveal a transient heritable response to TNF-α perturbation (A and B) (A) Reactivation percentage of each colony across weeks 5–13 after TNF-α exposure of 24 h, tested biweekly. Two colonies did not reach experimentally feasible concentrations at week 5 and were included starting from week 7. Black dots represent the control values, measured by exposing the unsorted culture to the same TNF-α treatment for 24 h. The colonies gradually moved closer toward the unsorted control as culturing time progressed. This trend is more clearly represented in (B) where the difference in reactivation percentage between each colony and the control, Δ% Reactivation, is visualized. The approach toward unsorted control is demonstrated by the bundling of colony-wise response around 0 at later time points. (C) Histogram of reactivation percentage of single-cell colonies from each week's measurement. Week 5 has a total of 14 colonies, whereas the remaining weeks have 16. Unsorted control average is shown as black vertical lines. Bin size for weeks 5, 7, and 9 was set to 10, whereas for weeks 11 and 13 it was set to 5 for better resolution. The histograms also show the convergence of colony-wise behavior toward the unsorted control. (D) Bar plot of the colony-wise mean reactivation percentage (blue) and colony-wise mean CV (coefficient of variation squared) of reactivation percentage (red). Mean % reactivation values showed an overall decreasing trend as culture time went on, whereas mean CV of % reactivation decreased for the first 5 weeks before reaching a plateau in weeks 11 and 13. The control value is a pooled average of all unsorted + TNF-α treatment values across the 5 time points. Bar heights and error bars represent bootstrapped mean and standard error. Levels of significance are indicated as n.s. (p ≥ 0.1), one star (∗, 0.1 > p ≥ 0.05) or three stars (∗∗∗, p < 0.01). p Values were calculated using bootstrapping. See transparent methods for details on statistics. See also Figures S5–S7 for alternative gating strategies.

Figure 3

A transient cell state governs HIV-1 exit from latency (A) Schematic of the stochastic model where prior to TNF-α exposure, individual cells reversibly switch between a TNF-α irresponsive and a responsive state. Self arrows represent proliferation of cells with the mother cell state being inherited by daughters. A sample realization of cells switching states over time is shown below. (B) The predicted colony-to-colony fluctuations in the fraction of reactivating cells (as measured by the coefficient of variation CV) increases for longer periods in the responsive state assuming a fixed 25% reactivation averaged across colonies. Plots are shown for different times of TNF-α exposure. (C) Predicted distribution of the fraction of reactivating cells across colonies after 5 weeks of growth for different average times in the responsive state. Distributions are obtained by using a beta distribution with an average 25% reactivation and noise as given by Equation (2). (D) Measured colony-to-colony fluctuations (with error bars showing standard error) starting from 5 to 13 weeks (black circles corresponding to data from Figure 2D). The fluctuations are considerably high at the initial 5-week time point (CV = 75%) and then gradually decrease to the noise floor. Note that the 11- and 13-week timepoints have been corrected for the mean as discussed in the main text. Model predicted CV as per Equation Equation (4) are shown assuming a noise floor CV_NF= 0.35 for different average times in the responsive state. See also Figures S3 and S4.