Available online at www.sciencedirect.com
ScienceDirect A minimal fate-selection switch Leor S Weinberger To preserve fitness in unpredictable, fluctuating environments, a range of biological systems probabilistically generate variant phenotypes — a process often referred to as ‘bet-hedging’, after the financial practice of diversifying assets to minimize risk in volatile markets. The molecular mechanisms enabling bethedging have remained elusive. Here, we review how HIV makes a bet-hedging decision between active replication and proviral latency, a long-lived dormant state that is the chief barrier to an HIV cure. The discovery of a virus-encoded bethedging circuit in HIV revealed an ancient evolutionary role for latency and identified core regulatory principles, such as feedback and stochastic ‘noise’, that enable cell-fate decisions. These core principles were later extended to fate selection in stem cells and cancer, exposed new therapeutic targets for HIV, and led to a potentially broad strategy of using ‘noise modulation’ to redirect cell fate. Address Gladstone Institutes (Virology and Immunology), Department of Biochemistry & Biophysics, University of California, San Francisco, United States Corresponding author: Weinberger, Leor S (
[email protected])
Current Opinion in Cell Biology 2015, 37:111–118 This review comes from a themed issue on Differentiation and disease Edited by Scott A Armstrong and Michael Rape For a complete overview see the Issue and the Editorial Available online 21st November 2015 http://dx.doi.org/10.1016/j.ceb.2015.10.005 0955-0674/# 2015 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Introduction: ‘bet-hedging’ during fateselection decisions Biological systems use two general strategies to maintain fitness in variable, fluctuating environments. One strategy involves the evolution of sophisticated sensor-actuator networks that continually assess environmental surroundings and adapt, generating a tailored response to a given surrounding. Sensor-actuator strategies appear throughout bacterial chemotaxis systems and osmoregulation [1– 5]. An alternate strategy, proposed 50 years ago by the theoretical ecologist Dan Cohen [6], involves probabilistically generating a range of different phenotypes in any given environment. While studying desert annual plants, where reproductive success is governed by unpredictable weather patterns, Cohen noted that fitness could be enhanced if seeds varied in their germination potential, www.sciencedirect.com
allowing chance to govern each seed’s fate to germinate or not. For example, if husk thickness of each seed is allowed to stochastically vary, a fraction of seeds will randomly remain non-germinated regardless of the environment, leaving a long-lived sub-population to avoid extinction during long-term droughts. Borrowing from economic theory, Cohen proposed that biological organisms could ‘hedge their bets’ in much the same way that financial houses diversify their assets — between higherrisk (higher-yield) stocks and lower-risk (lower-yield) securities — in order to minimize risk against market crashes.
Stochastic gene expression enables ‘bethedging’: from theory to the initial HIV evidence For many years, the molecular mechanisms enabling biological systems to probabilistically hedge their bets between alternate developmental fates remained unclear. Viruses proved to be powerful model systems to define core regulatory principles underlying bet-hedging decisions. Viruses optimize fitness in variant cellular environments (i.e., target rich vs. target poor) and do so under strong genetic constraints, selecting for minimalist regulatory circuits (RNA virus genomes are typically 10 kB, thereby precluding sophisticated environmental sensing and actuation that require multiple genes and substantial genomic real estate). In the late 1990s, mathematical models of the lysis-lysogeny fate decision in the bacterial virus phage l suggested that a basic biophysical phenomenon intrinsic to life at the single-cell level — molecular fluctuations driven by Brownian diffusion — could generate the variability required for a developmental bethedging decision [7]. These diffusion-driven Brownian fluctuations in regulatory enzymes, mRNAs, and other biomolecules are unavoidable and appear sufficient to shift cells between transcriptional on and off states [8]. With some cells randomly active and others dormant, the computational result is a distribution of cell fates across a population. That is, the models showed that stochastic fluctuations in gene expression enable the phage to mediate between its two alternate developmental fates without sensor-actuator circuitry. The first experimental evidence for the theory that molecular fluctuations could drive a developmental bethedging decision was found in another virus: the human immunodeficiency virus type 1 (HIV), which probabilistically chooses between active transcription and a latent state [9]. HIV has become a useful model to experimentally elucidate core principles underlying stochastic bet-hedging, and the principles discovered in Current Opinion in Cell Biology 2015, 37:111–118
112 Differentiation and disease
HIV have been generalized to fate-selection decisions occurring in bacteria [10], stem-cell reprogramming [11,12], and cancer drug tolerance [13]. Below, these core principles of fate-selection are reviewed in the context of the HIV latency decision.
HIV latency: a bet-hedging strategy to optimize viral transmission Upon infecting CD4+ T lymphocytes, HIV either actively replicates to rapidly produce progeny virions or enters a long-lived quiescent state (proviral latency), from which it subsequently reactivates [14]. Latently infected cells form a viral reservoir, enabling life-long viral persistence and necessitating lifelong antiretroviral therapy (ART) for HIV-infected individuals. The evolutionary conundrum was how latency had been maintained over the centuries of natural lentiviral infections in non-human primates before the current ART era, given the rapid evolution of the virus. Historically, the prevailing dogma in retrovirology had been that HIV latency played no role during HIV’s natural history of infection [14,15]. In the absence of ART, latency was thought to be a deleterious phenotype, since latently infected cells produce no virus and decrease viral loads. Given latency’s reduction of lentiviral replicative fitness, latency seemed to be an evolutionary accident — an ‘epiphenomenon’ reflecting serendipitous lentiviral infection of some CD4+ T cells precisely during their transition from an activated state to a quiescentmemory state [16–18]. Latency was therefore viewed as an infrequent bystander effect that occurs only after a virus-driven adaptive immune response begins and CD4+ T lymphocytes start to form memory subsets. However, several findings were inconsistent with this dogma. First, in Rhesus monkeys, viral latency is rapidly established long before the adaptive immune response begins [19]. Second, in patient cells, reactivation of quiescent-memory cells infected with latent HIV did not necessarily reactivate the latent virus [20,21]. Third, latency occurs with high frequency in cell culture (50% of infected cells establish latency), despite the absence of immune responses or any cellular relaxation to memory [22,23]. Fourth, remarkably, HIV’s Tat feedback circuit is sufficient to generate a probabilistic bet-hedging decision in the absence of cellular relaxation, and the molecular architecture of the circuit appears optimized to generate a bet-hedging decision [9,24,25] (described in more detail below). If latency were a non-beneficial viral trait or an epiphenomenon, the latency phenotype should have been lost due to natural selection or genetic drift — given the rapid evolution of lentiviruses — and the viral gene-regulatory circuitry should have evolved away from being an intrinsic latency generator. The persistence of this Current Opinion in Cell Biology 2015, 37:111–118
virus-encoded latency circuit suggested an unknown selective advantage that counterbalances latency’s putative fitness cost (i.e., reduced long-term viral loads). To address this discrepancy, Rouzine et al. [26] examined whether latency could provide a fitness advantage during initial infection, when the virus must pass through the target-cell-poor mucosa, where >90% of HIV infections start [27]; the evolutionary precursor to HIV in humans — SIV in non-human primates — also spreads through mucosal transmission [28]. In target-poor environments (e.g., the early mucosa) high virulence quickly depletes available targets, leaving no infectable cells at the site. Consequently, if a virus is highly virulent, the mucosal infection is extinguished before the virus can spread to the target-rich environments to generate selfsustaining systemic infection (Figure 1). In fact, direct observations in macaques show that mucosal infections precipitously decay toward local extinction in the first few days after inoculation [27,29]. Rouzine et al. [26] built off patient-validated models [30] to calculate what the optimum bet-hedging frequency would be, given the inherent cost–benefit tradeoff latency generates between systemic and mucosal infection. Surprisingly, they found the optimal bet-hedging frequency was 50% latency, the same latency frequency typically found in infections of cultured T lymphocytes. Incorporating immune responses into the model enabled the authors to fit all available patient data, including a far lower latency frequency [26]. A recent clinical study in patients [31] further supports the model that latency is a bet-hedging adaptation to optimize transmission through the target-poor mucosa.
Molecular architecture of HIV’s bet-hedging circuit For many years, the absence of a viable evolutionary fitness argument bolstered the notion that HIV proviral latency was a bystander effect (i.e., epiphenomenon) ineludibly resulting from transcriptional silencing during relaxation of activated lymphocytes to a quiescent-memory state (Figure 2a). However, despite the vast literature of associative evidence linking latent HIV integration sites to silenced chromatin and correlating latency with cellular silencing [32,33], there was also evidence for an alternate model where latency was controlled by viral gene-regulatory circuitry [9,20,34] without strict dependence on cellular state (Figure 2a). Critically, the hypothesis that latency establishment was driven by cellular state had never been directly tested (i.e., lymphocytes had never been infected and tracked in real time as they underwent relaxation to memory). The first direct test between these models was only recently carried out. Razooky and Pai et al. [35] isolated primary CD4+ T lymphocytes from human www.sciencedirect.com
The HIV latency circuit Weinberger 113
Figure 1
Figure 2
Active Replication
Alternate HIV Latency Models
(a)
Autonomous Viral Program
Viral Activity
Cell-state Dependent
Viral Activity
(a)
(b)
Viral Latency Initial Infection (Target-poor)
Latencyfree Virus
Systemic Infection (Target-rich)
Cellular State
Infection
Activated
CellState
(b)
Viral Expression
Cellular State
Resting 0
4
7
10
13
Day (c)
Viral Expression Normalized Mean Intensity
Active Replication
Latencycapable Virus
1
Cell-state 0 4
Viral Latency
13
Current Opinion in Cell Biology
Current Opinion in Cell Biology
HIV latency: a bet-hedging strategy to optimize viral transmission. (a) Upon infecting CD4+ T lymphocytes, HIV can either actively replicate and rapidly destroy the cell or enter a long-lived quiescent state (proviral latency), from which it can subsequently reactivate. The reservoir of latent cells precludes curative therapy. (b) The evolutionary role of latency; schematic of two competing HIV strains during early infection — a hypothetical latency-incapable strain (upper) and the extant latencycapable strain (lower). Infection begins with viral inoculation in the mucosa and progresses — in some cases — to systemic infection in the lymphoid tissue, where >98% of CD4+ T cells reside [48]. An HIV strain incapable of entering latency would generate increased viral loads during systemic infection, transferring more virions to new hosts. However, the latencyincapable virus would rapidly destroy the small CD4+ T cell population initially present in the mucosa of the new host — reducing the probability of systemic infection. In contrast, an HIV strain capable of probabilistically entering latency (i.e., bet hedging) would generate lower viral loads during target-rich systemic infection, transferring fewer virions to new hosts, yet, the relatively few transferred virions would not destroy all mucosal target cells. By entering long-lived latency in some mucosal cells, the latencycapable strain increases its probability of surviving initial infection to establish systemic infection. Calculations of the optimal bet-hedging frequency for latency to constitute an evolutionary stable strategy match observed frequencies of latency in cell culture [26].
www.sciencedirect.com
7 10 Days post infection
HIV latency establishment is largely autonomous to cellular relaxation. (a) Two models of HIV latency regulation. The ‘epiphenomenon’ hypothesis of HIV latency regulation (left) predicted a strong correlation between cellular activation state and viral activity: as CD4+ T cells relaxed from an activated state (permissive to infection) to a resting-memory state, they would silence HIV gene expression to drive latency establishment. The alternate hypothesis (right) held that Tat positive-feedback circuitry is sufficient to enable probabilistic bet hedging and that viral gene expression is robust to changes in the cell state. This autonomous circuit hypothesis predicts that cellular relaxation does not necessarily drive establishment of latency and that both latent and active viral expression can occur irrespective of cellular state. (b) Schematic of experiment to directly test cellular dependence of latency versus autonomous viral programming of latency. CD4+ T cells, derived from human donors, were activated to enable HIV infection, infected with a single-round HIV (envdeleted and expressing a short-lived GFP), and cells were then allowed to relax to resting memory (by removal of activation stimuli). HIV expression and cellular activation levels were tracked for two weeks by flow cytometry. (c) Transitioning of primary T lymphocytes from activated to resting did not silence HIV expression as assayed by flow cytometry analysis of CD25/CD69 cell-activation markers and viral GFP expression. Although cells fully relax from activated to resting, viral gene expression does not silence (even despite extension of infected-cell lifetime due env deletion). Data shown from duplicate infections performed on cells from two donors (adapted from [35]).
Current Opinion in Cell Biology 2015, 37:111–118
114 Differentiation and disease
donors, activated the cells to enable HIV infection, and infected them with a single-round HIV (env-deleted) expressing a short-lived green fluorescent protein (GFP) (Figure 2b). The activated cells were then allowed to relax to resting memory (by removing activation
stimuli), and HIV expression and cellular activation levels were tracked by flow cytometry. Strikingly, transitioning of primary T lymphocytes from activated to resting did not silence HIV expression (Figure 2c), demonstrating that HIV expression during establishment of latency was
Figure 3
(a) Tat
Tat
5′ LTR
↑↓ Tat
5′ LTR
(c) Tat Molecules/Cell
(b)
ON
s
OFF
10–1 100 1
Time
Ta
e on
l
C
01 1
t (a 02 103 .u. )
(d) Tat Decay
5′ LTR
FKBP
Tat FKBP
No Shield-1
80 60
+ Shield-1
40 20
+ Shield-1
Fold Decline in Tat Expression
+ Shield-1
Tat
% of Max (# cells)
No Shield-1
1
100
No Shield-1
4
0 0
102
103
Tat (a.u.)
104
105
1
4
Fold Decline in Cell-state Current Opinion in Cell Biology
The architecture of HIV’s minimal bet-hedging circuit. (a) Schematic of the HIV fate-selection circuit. The HIV LTR promoter fluctuates between a transcriptional elongation-competent (ON) state and a non-elongation-competent (OFF) state [36,38]. In the ON state, Tat protein, can transactivate the LTR to enhance transcription (by promoting elongation of RNA polymerases stalled on the LTR) comprising a positive-feedback circuit that amplifies expression 100 fold, for review see [40]. (b) Stochastic simulations of this minimal circuit show that molecular fluctuations can generate a phenotypic bifurcation in cellular Tat levels where Tat levels will fluctuate to zero in some cells (i.e., trajectories) but not in other (genetically identical) cells [35]. (c) Experimental evidence (flow cytometry) that a minimal LTR-Tat-GFP feedback circuit is sufficient to generate a phenotypic bifurcation in genetically identical clones [9]. Abrogating Tat positive feedback eliminates the phenotypic bifurcation both computationally and experimentally [9,49]. (d) A synthetic Tat feedback circuit (left) validates model predictions that the minimal Tat positivefeedback circuit is sufficient to account for the circuit turning on and off (middle) and for HIV resilience to cellular silencing (right). The FKBP proteolysis domain causes Tat to rapidly degrade (abrogating positive feedback), but the small molecule Shield-1 inhibits FKBP-mediated degradation and allows Tat to assume its native half-life and complete the positive feedback loop. Cellular relaxation experiment performed as in Figure 2b but cells were transduced with the minimal-synthetic circuit and feedback was chemically modulated using Shield-1 [35].
Current Opinion in Cell Biology 2015, 37:111–118
www.sciencedirect.com
The HIV latency circuit Weinberger 115
resilient to cellular state and pointing to an intrinsic viral program controlling the establishment of latency. To determine the mechanisms that might comprise the viral program, mathematical modeling of the HIV circuit was used to define the minimal core principles that could enable a fate-selection decision and allow the fate decision to be resilient to cellular relaxation. Previous singlecell studies had determined that HIV’s long-terminal repeat (LTR) promoter, the virus’s only promoter, is exceptionally noisy [36]; the dominant noise source is large, infrequent ‘bursts’ of transcription [37,38] such that the promoter can be modeled using a two-state ‘random telegraph’ model [38]. It was also known that LTR noise is amplified by positive feedback from HIV’s Tat protein, which transactivates the LTR [39]. Thus, the viral circuit can be modeled as a two-state promoter with positive feedback (Figure 3a). Stochastic simulations of this minimal Tat-feedback circuit recapitulated a phenotypic bifurcation between a transcriptionally active and latent state (Figure 3b), which matched the original experimental observation [9] that a minimal Tat positive-feedback circuit is sufficient to generate a phenotypic bifurcation in genetically identical cells (Figure 3c). Importantly, the simulations showed that the bifurcation can occur largely independent of cell state [35]. To experimentally test the mathematical model that Tat circuitry by itself can establish HIV latency independent of cellular state, a synthetic biology approach was used [35]. A minimal synthetic Tat-feedback circuit was constructed that allowed tuning of feedback strength through chemically modulated proteolysis (Figure 3d). Then, the primary-cell cellular-relaxation experiment was repeated to determine if the minimal Tat circuit was resilient to cellular silencing, as was the full-length virus. As predicted, when Tat feedback was abrogated — by rapid proteolysis of Tat — HIV transcription was silenced as the cells became quiescent. But, when Tat proteolysis was chemically blocked by a small molecule (Shield-1), Tat feedback alone was sufficient to overcome the cellular silencing during relaxation to quiescence in the primary T cells (Figure 3d). The results demonstrate that Tat feedback is necessary and sufficient to establish HIV latency independent of cellular silencing. How does HIV achieve this paradoxical functioning? Specifically, how is HIV simultaneously resistant to the global effects of cellular silencing while sensitive enough to molecular fluctuations to allow these to drive transcriptional switching? The regulatory circuit architecture provides a mechanism. Although the LTR responds to external stimuli, the response is relatively small — strong transcriptional activators (e.g., TNF, which induces changes in NFkB) generate only 2-fold changes in LTR expression www.sciencedirect.com
[9,24,40]. In contrast, Tat transactivation increases LTR expression by >50-fold. This strong positive feedback from Tat transactivation effectively insulates the circuit from the relatively weak effects of cell-state changes and other external stimuli. The only mechanisms that can significantly influence the activated circuit are those that disrupt feedback, such as the large stochastic bursts that are intrinsic to LTR expression and that are amplified by Tat feedback. The resulting Tat fluctuations can be so extreme that Tat molecules are completely depleted, which is sufficient to drive probabilistic switching to an off state [24,40]. Essentially, during a large fluctuation the system can fluctuate into zero Tat molecules, which constitutes a ‘molecular extinction’, from which there is no recovery; this is analogous to a self-replicating system (e.g. bacterial division) where a fluctuation to a population size of zero results in extinction. For the HIV circuit, transcriptional elongation from the LTR promoter is so weak in many integration sites that the circuit approximates a self-replicating system: Tat can only be produced when Tat is present to transactivate the LTR. Overall, positive feedback is the core mechanism that enables the circuit to be flipped by stochastic transcriptional fluctuations even when deterministic processes are unable to flip the switch [35]. Two functional aspects of the circuit are worth highlighting from an evolutionary perspective. First, comparable positive-feedback architecture had been proposed on theoretical grounds to be an unreliable environmental sensor in fluctuating environments [43]. Thus, HIV’s circuit architecture is precisely the opposite of what would be expected for an environmental sensor that would respond reliably to cellular changes. Second, high-magnitude noise, as exhibited by the LTR, is generally deleterious, being selected against and filtered out of regulatory circuits [41,42]. So, the high level of noise in HIV circuitry is extraordinary given the virus’s rapid evolutionary rate, supporting the concept that expression noise is selectively beneficial for HIV’s fate decision. Hence, viral evolution appears to have selected for circuitry that both maintains strong autonomy from environmental cues and simultaneously drives probabilistic on-off decisions.
Conclusion: bet-hedging circuits as targets for therapy Stochastic noise is now considered a fundamental process driving diverse cell-fate decisions [11–13] and is recognized as a primary clinical barrier to reversing latency in patients (i.e., curing HIV) [21,44,45]. Decoding of the core fate-selection circuit in HIV recently led to development of several new therapeutic avenues. On the one hand, newly designed Tat-feedback inhibitors block latent reactivation in patient samples [46], demonstrating the critical role of the Tat circuit in latenCurrent Opinion in Cell Biology 2015, 37:111–118
116 Differentiation and disease
Figure 4
(a) Chemical Potential
Epigenetic Landscape
Catalyst
Bunsen Burner Actiivatto Activator o orr Noise Enhancer Latent
(b) LTR
Untreated TNF, PMA Noise Enhancers (85)
2.5
d2GFP
# Cells
Δ Noise
2.0
Noise Activator Enhancer
Active
1.5 1.0 0.5 0
GFP
GFP
–0.5 –2
–1
0
1
2
Log2(GFPTreated/GFPUntreated) 100
% Reactivation
(c)
90
No Reactivation Reactivation Synergy Synergy
80
–PMA +PMA
70 60
0
PMA only V12 V13 V14 V11 V4 V9 V6 V5 V3 V1 V2 V10 V7
10
Noise Enhancers
Current Opinion in Cell Biology
Controlling bet-hedging circuitry through noise modulation. (a) Conceptual basis for noise modulation. Left: chemical-reaction efficiency is enhanced not only by catalysts but also by increasing the thermal energy (i.e., kT in the Arrhenius equation). Although catalysts deterministically lower activation-energy barriers on potential-energy landscapes, amplifying thermal fluctuations (e.g., via a Bunsen burner) provides an added perturbation for crossing activation-energy barriers. Right: On a Waddington epigenetic landscape, small-molecule enhancement of stochastic gene-expression fluctuations is analogous to increasing thermal fluctuations in chemical systems. (b) Identification of small molecules that enhance noise in HIV gene expression. Left: LTR-GFP construct and schematic histograms of cells exposed to either a transcriptional activator or a noise enhancer. Right: Each point represents flow cytometry analysis of 50 000 LTR-GFP-expressing cells exposed to a compound; 85 transcriptional noise enhancers identified from among 1600 FDA-approved drugs (red) are shown as in [47]; the effect of TNF or PMA (activators) is shown in blue. (c) Noise-enhancer molecules (labeled V12, V13, among others) synergize with conventional activators such as PMA (blue) and significantly modulate HIV’s fate decision by enhancing (purple) reactivation of latent HIV in donor-derived human primary T cells [47].
Current Opinion in Cell Biology 2015, 37:111–118
www.sciencedirect.com
The HIV latency circuit Weinberger 117
cy. On the other hand, studies of LTR noise laid the foundation for an orthogonal approach of enhancing geneexpression noise to reactivate and kill latently infected cells (Figure 4a). This strategy [47] has potential to direct cellfate across a broad class of biological systems and is conceptually similar to approaches in physical and chemical systems where fluctuations are amplified to drive a system over a barrier. Indeed Dar et al. identified [47] a set of FDAapproved small molecules that modulate noise in LTR expression without altering the mean level of LTR expression (Figure 4b). Strikingly, these noise-modulating compounds — which would be overlooked in conventional screens — synergized with conventional transcriptional activators and increased reactivation of latent HIV (Figure 4c). As FDA-approved compounds, noise enhancers have potential therapeutic application to HIV and possibly other clinically relevant cell-fate decisions [11–13]. In conclusion, noise-driven cell-fate decisions are likely a primitive biological phenomenon and it appears possible to exploit the underlying noise as a clinical tool to reverse microbial persistence and potentially reprogram cell fate in general.
Acknowledgements I thank Anand Pai, Elizabeth Tanner, Noam Vardi, Brandon Razooky, Ariel Weinberger, and members of the UCSF and Gladstone community for helpful discussion and reviews of this manuscript. This work was supported by NIH awards OD017181, OD006677, AI109593 and AI109611, and by the Pew Scholars in Biomedical Sciences and the Alfred P. Sloan Foundation.
References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as: of special interest of outstanding interest 1.
Elowitz MB, Levine AJ, Siggia ED, Swain PS: Stochastic gene expression in a single cell. Science 2002, 297:1183-1186.
2.
Endres RG, Wingreen NS: Precise adaptation in bacterial chemotaxis through assistance neighborhoods. Proc Natl Acad Sci U S A 2006, 103:13040-13044.
loop: HIV-1 Tat fluctuations drive phenotypic diversity. Cell 2005, 122:169-182. 10. Suel GM, Kulkarni RP, Dworkin J, Garcia-Ojalvo J, Elowitz MB: Tunability and noise dependence in differentiation dynamics. Science 2007, 315:1716-1719. 11. Chang HH, Hemberg M, Barahona M, Ingber DE, Huang S: Transcriptome-wide noise controls lineage choice in mammalian progenitor cells. Nature 2008, 453:544-547. 12. Hanna J, Saha K, Pando B, van Zon J, Lengner CJ, Creyghton MP, van Oudenaarden A, Jaenisch R: Direct cell reprogramming is a stochastic process amenable to acceleration. Nature 2009, 462:595-601. 13. Gupta PB, Fillmore CM, Jiang G, Shapira SD, Tao K, Kuperwasser C, Lander ES: Stochastic state transitions give rise to phenotypic equilibrium in populations of cancer cells. Cell 2011, 146:633-644. 14. Siliciano RF, Greene WC: HIV latency. Cold Spring Harb Perspect Med 2011, 1:a007096. 15. Coffin J, Swanstrom R: HIV pathogenesis: dynamics and genetics of viral populations and infected cells. Cold Spring Harb Perspect Med 2013, 3:a012526. 16. Coffin J, Swanstrom R: HIV pathogenesis: dynamics and genetics of viral populations and infected cells. Cold Spring Harbor Perspect Med 2013, 3:a012526. 17. Han Y, Wind-Rotolo M, Yang HC, Siliciano JD, Siliciano RF: Experimental approaches to the study of HIV-1 latency. Nat Rev Microbiol 2007, 5:95-106. 18. Eisele E, Siliciano RF: Redefining the viral reservoirs that prevent HIV-1 eradication. Immunity 2012, 37:377-388. 19. Whitney JB, Hill A, Sanisetty S, Penaloza-MacMaster P, Liu J, Shetty M, Parenteau L, Cabral C, Shields J, Blackmore S et al.: Rapid seeding of the viral reservoir prior to SIV viraemia in rhesus monkeys. Nature 2014, 512:74-77. A study in rhesus macaques (the gold-standard model) showing that latency is established within three days of infection — long before substantial formation of resting-memory cells. During these first few days, latent infection of cells is a high probability event. These findings are difficult to reconcile with the ‘epiphenomenon’ hypothesis. 20. Ho YC, Shan L, Hosmane NN, Wang J, Laskey SB, Rosenbloom DI, Lai J, Blankson JN, Siliciano JD, Siliciano RF: Replication-competent noninduced proviruses in the latent reservoir increase barrier to HIV-1 cure. Cell 2013, 155:540-551. Shows that reactivation of latent HIV in patient cells is a probabilistic phenomenon. Maximal activation of resting-memory cells does not necessarily reactivate latent HIV in these cells. 21. Weinberger AD, Weinberger LS: Stochastic fate selection in HIVinfected patients. Cell 2013, 155:497-499.
3.
Rao CV, Kirby JR, Arkin AP: Design and diversity in bacterial chemotaxis: a comparative study in Escherichia coli and Bacillus subtilis. PLoS Biol 2004, 2:E49.
4.
Macnab RM, Koshland DE Jr: The gradient-sensing mechanism in bacterial chemotaxis. Proc Natl Acad Sci U S A 1972, 69:25092512.
22. Dahabieh MS, Ooms M, Simon V, Sadowski I: A doubly fluorescent HIV-1 reporter shows that the majority of integrated HIV-1 is latent shortly after infection. J Virol 2013, 87:4716-4727. Shows that in cell culture 50% of infections result in latency and that latency occurs rapidly, inconsistent with chromatin silencing but consistent with stochastic fluctuations in Tat driving latency establishment.
5.
Muzzey D, Gomez-Uribe CA, Mettetal JT, van Oudenaarden A: A systems-level analysis of perfect adaptation in yeast osmoregulation. Cell 2009, 138:160-171.
23. Calvanese V, Chavez L, Laurent T, Ding S, Verdin E: Dual-color HIV reporters trace a population of latently infected cells and enable their purification. Virology 2013, 446:283-292.
6.
Cohen D: Optimizing reproduction in a randomly varying environment. J Theor Biol 1966, 12:119-129.
24. Weinberger LS, Dar RD, Simpson ML: Transient-mediated fate determination in a transcriptional circuit of HIV. Nat Genet 2008, 40:466-470.
7.
Arkin A, Ross J, McAdams HH: Stochastic kinetic analysis of developmental pathway bifurcation in phage lambda-infected Escherichia coli cells. Genetics 1998, 149:1633-1648.
8.
Raj A, van Oudenaarden A: Nature, nurture, or chance: stochastic gene expression and its consequences. Cell 2008, 135:216-226.
9.
Weinberger LS, Burnett JC, Toettcher JE, Arkin AP, Schaffer DV: Stochastic gene expression in a lentiviral positive-feedback
www.sciencedirect.com
25. Burnett JC, Miller-Jensen K, Shah PS, Arkin AP, Schaffer DV: Control of stochastic gene expression by host factors at the HIV promoter. PLoS Pathogens 2009, 5:e1000260. 26. Rouzine IM, Weinberger AD, Weinberger LS: An evolutionary role for HIV latency in enhancing viral transmission. Cell 2015, 160:1002-1012. First study proposing an evolutionary fitness benefit for HIV latency; argues that latency is a stochastic bet-hedging adaptation that evolved to prevent Current Opinion in Cell Biology 2015, 37:111–118
118 Differentiation and disease
viral extinction during early infection in target-poor tissues (i.e. mucosal tissues). 27. Haase AT: Early events in sexual transmission of HIV and SIV and opportunities for interventions. Annu Rev Med 2011, 62:127-139.
37. Singh A, Razooky BS, Dar RD, Weinberger LS: Dynamics of protein noise can distinguish between alternate sources of gene-expression variability. Mol Syst Biol 2012, 8:607. 38. Dar RD, Razooky BS, Singh A, Trimeloni TV, McCollum JM, Cox CD, Simpson ML, Weinberger LS: Transcriptional burst frequency and burst size are equally modulated across the human genome. Proc Natl Acad Sci U S A 2012, 109:17454-17459.
28. Santiago ML, Range F, Keele BF, Li Y, Bailes E, Bibollet-Ruche F, Fruteau C, Noe R, Peeters M, Brookfield JF et al.: Simian immunodeficiency virus infection in free-ranging sooty mangabeys (Cercocebus atys atys) from the Tai Forest, Cote d’Ivoire: implications for the origin of epidemic human immunodeficiency virus type 2. J Virol 2005, 79:12515-12527.
39. Weinberger LS, Shenk T: An HIV feedback resistor: autoregulatory circuit deactivator and noise buffer. PLoS Biol 2007, 5:e9.
29. Zhang Z, Schuler T, Zupancic M, Wietgrefe S, Staskus KA, Reimann KA, Reinhart TA, Rogan M, Cavert W, Miller CJ et al.: Sexual transmission and propagation of SIV and HIV in resting and activated CD4+ T cells. Science 1999, 286:1353-1357.
40. Razooky BS, Weinberger LS: Mapping the architecture of the HIV-1 Tat circuit: a decision-making circuit that lacks bistability and exploits stochastic noise. Methods 2011, 53:68-77.
30. Nowak M, May R: Virus Dynamics: Mathematical Principles of Immunology and Virology. Oxford University Press; 2000.
41. Fraser HB, Hirsh AE, Giaever G, Kumm J, Eisen MB: Noise minimization in eukaryotic gene expression. PLoS Biol 2004, 2:e137.
31. Deymier MJ, Ende Z, Fenton-May AE, Dilernia DA, Kilembe W, Allen SA, Borrow P, Hunter E: Heterosexual transmission of subtype C HIV-1 selects consensus-like variants without increased replicative capacity or interferon-alpha resistance. PLoS Pathog 2015, 11:e1005154. Validates a central prediction of the latency bet-hedging model; shows that in patient isolates, HIV variants with increased replicative fitness are not transmitted more efficiently than variants with lower replication in vitro. Previously it had been assumed that the most replicative variants (i.e. lower latency frequency) would be most likely to be transmitted. This study shows the most commonly transmitted ‘founder’ variants have lower replication in vitro. 32. Tyagi M, Pearson RJ, Karn J: Establishment of HIV latency in primary CD4+ cells is due to epigenetic transcriptional silencing and P-TEFb restriction. J Virol 2010, 84:6425-6437. 33. Pearson R, Kim YK, Hokello J, Lassen K, Friedman J, Tyagi M, Karn J: Epigenetic silencing of human immunodeficiency virus (HIV) transcription by formation of restrictive chromatin structures at the viral long terminal repeat drives the progressive entry of HIV into latency. J Virol 2008, 82:1229112303. 34. Jeeninga RE, Westerhout EM, van Gerven ML, Berkhout B: HIV-1 latency in actively dividing human T cell lines. Retrovirology 2008, 5:37. 35. Razooky BS, Pai A, Aull K, Rouzine IM, Weinberger LS: A hardwired HIV latency program. Cell 2015, 160:990-1001. Shows that HIV latency is largely independent of cellular relaxation/ silencing and that Tat positive feedback is necessary and sufficient to establish latency. Suggests that Tat circuitry evolved to act as a semiautonomous latency switch. 36. Singh A, Razooky B, Cox CD, Simpson ML, Weinberger LS: Transcriptional bursting from the HIV-1 promoter is a significant source of stochastic noise in HIV-1 gene expression. Biophys J 2010, 98:L32-L34.
Current Opinion in Cell Biology 2015, 37:111–118
42. Batada NN, Hurst LD: Evolution of chromosome organization driven by selection for reduced gene expression noise. Nat Genet 2007, 39:945-949. 43. Brandman O, Ferrell JE Jr, Li R, Meyer T: Interlinked fast and slow positive feedback loops drive reliable cell decisions. Science 2005, 310:496-498. 44. Rouzine IM, Razooky BS, Weinberger LS: Stochastic variability in HIV affects viral eradication. Proc Natl Acad Sci U S A 2014, 111:13251-13252. 45. Ho YC, Shan L, Hosmane N, Wang J, Laskey S, Rosenbloom DIS, Lai J, Blankson JN, Siliciano JD, Siliciano RF: Replicationcompetent noninduced proviruses in the latent reservoir increase barrier to HIV-1 cure. Cell 2013, 155:540-551. 46. Mousseau G, Kessing CF, Fromentin R, Trautmann L, Chomont N, Valente ST: The Tat inhibitor didehydro-cortistatin A prevents HIV-1 reactivation from latency. MBio 2015, 6:e00465. Shows the integral role of Tat feedback in latency by identifying a smallmolecule inhibitor of Tat positive-feedback circuitry and demonstrating that this inhibitor prevents reactivation of latent HIV in patient cells. 47. Dar RD, Hosmane NN, Arkin MR, Siliciano RF, Weinberger LS: Screening for noise in gene expression identifies drug synergies. Science 2014, 344:1392-1396. Demonstrates that small molecules can modulate stochastic fluctuations (noise) in HIV gene expression and potentiate HIV reactivation. Exposes a new class of noise-enhancing and noise-suppressing molecules with the potential to direct cell-fate decisions. 48. Murphy K: Janeway’s Immunobiology. Eighth Edition. Garland Science; 2011. 49. Singh A, Weinberger LS: Stochastic gene expression as a molecular switch for viral latency. Curr Opin Microbiol 2009, 12:460-466.
www.sciencedirect.com