Realtime 2-5A kinetics suggests interferons β and λ evade global arrest of translation by RNase L

Cells of all mammals recognize double-stranded RNA (dsRNA) as a foreign material. In response, they release interferons (IFNs) and activate a ubiquitously expressed pseudokinase/endoribonuclease RNase L. RNase L executes regulated RNA decay and halts global translation. Here we developed a biosensor for 2’,5’-oligoadenylate (2-5A), the natural activator of RNase L. We found that 2-5A was acutely synthesized by cells in response to dsRNA sensing, which immediately triggered cellular RNA cleavage by RNase L and arrested host protein synthesis. However, translation-arrested cells still transcribed IFN-stimulated genes (ISGs) and secreted IFNs of types I and III (IFN-β and IFN-λ). Our data suggests that IFNs escape from the action of RNase L on translation. We propose that 2-5A/RNase L pathway serves to rapidly and accurately suppress basal protein synthesis, preserving privileged production of defense proteins of the innate immune system. Significance RNase L is a mammalian enzyme that can stop global protein synthesis during interferon response. Cells must balance the need to make interferons (which are proteins) with the risk to lose cell-wide translation due to RNase L. This balance can most simply be achieved if RNase L was activated late in the interferon response. However, we show by engineering a biosensor for the RNase L pathway, that on the contrary, RNase L activation precedes interferon synthesis. Further, translation of interferons evades the action of RNase L. Our data suggest that RNase L facilitates a switch of protein synthesis from homeostasis to specific needs of innate immune signaling.


Introduction
Interferons IFNs of type I ( and β) and type III (1, 2 and 3) are cytokines secreted by cells after exposure to pathogens or internal damage. Both types of IFNs activate strong transcriptional programs in surrounding tissues and have a central role in the innate immune system. Production of the IFNs requires protein synthesis.
However, during innate immune response to double-stranded RNA (dsRNA), IFN production is universally accompanied by translational arrest. Inhibition of protein synthesis arises in part due to activation of the dsRNA-dependent protein kinase R (PKR), and in part due to signaling by conserved small RNAs that contain 2',5'-linked oligoadenylates (2-5A) 1-4 . The action of 2-5A is sufficient for arrest of translation, independent of PKR, and at least in some cell lines 2-5A is the main cause of translational arrest 5 . In human cells, 2-5A is synthesized by three enzymes: oligoadenylate synthetase 1 (OAS1), OAS2 and OAS3 (OASs), which function as cytosolic dsRNA sensors using dsRNA binding for activation 6,7 . The activity of the OASs is normally low to allow housekeeping protein synthesis, but it increases in the presence of viral or host dsRNA molecules 8 . 2-5A has a strong antiviral effect, against which many viruses have evolved 2-5A antagonist genes that are essential for infection [9][10][11][12][13][14][15][16] .
The 2-5A system is also a surveillance pathway for endogenous dsRNAs from mammalian genomes. Cells with adenosine deaminase 1 (ADAR1) deficiency accumulate self-dsRNA that promotes 2-5A-driven apoptosis 8 . 2-5A synthesis in the presence of low amounts of endogenous dsRNAs does not cause cell death, but functions as a suppressor of adhesion, proliferation, migration, and prostate cancer metastasis 17,18 . Additionally, activation of the 2-5A system also blocks secretion of milk proteins and stops lactation, presumably as a mechanism to prevent passing infection via breast feeding 19 .
All of the effects of 2-5A arise from the action of a single mammalian 2-5A receptor, pseudokinase-endoribonuclease L (RNase L) 20 . 2-5A binds to the ankyrinrepeat (ANK) domain of RNase L and promotes its oligomerization and formation of a dimeric endoribonuclease active site [21][22][23] . This dimer further assembles into high-order oligomers 24 that cleave viral RNAs 16,18 and all components of the translation apparatus, including mRNAs 25 , tRNAs 26 and 28S/18S rRNAs 27,28 . The resulting action of RNase L inhibits global translation, which puts all proteins, including IFNs, at risk of arrest during a cellular response to dsRNA. The impact of translational shutdown by RNase L on IFN synthesis and paracrine IFN signaling is unknown. Measurements of 2-5A/RNase L activity have been limited by the need for biochemical analysis, which are incompatible with live cells. To address this challenge, we developed a realtime 2-5A biosensor and used it to elucidate the kinetics of 2-5A-mediated RNase L activation and translational arrest taking place during the cellular response to immuno-stimulatory dsRNAs. Our biosensor can detect in situ 2-5A synthesis in mammalian cells and thereby it establishes a heretofore missing platform for cell-based applications. These applications can range from live cell screens for modulators of innate immune responses to mechanistic analysis of dsRNA sensing, which we describe in our present work.

Realtime 2-5A dynamics in live cells
In the absence of methods to monitor 2-5A without cell disruption, the cellular dynamics of this second messenger are poorly understood. Here we developed a biosensor for continuous and non-invasive 2-5A monitoring in live cells. The biosensor was designed based on the crystal structures of the cellular 2-5A receptor RNase L 22,24 ( Fig. 1A), which indicate that the N-terminal ANK domain of RNase L is sufficient for 2-5A sensing and provides a minimal dimerization module 24 . In a single ANK protomer, the N-and C-termini in cis are separated, but upon dimerization of two ANK domains, which bind notably head-to-tail, the C-/N-termini become positioned in trans next to each other (Fig. 1B). We employed this in cis vs in trans N-/C-distance decrease to drive a dual-split-luciferase sensor.
The combinatorially optimized sensor has two halves, which consist of the core ANK domain fused to a modified split firefly luciferase (Fluc) 29 (Fig. 1B). We modified Fluc by replacing the published overlapping split junction N416/C415 with a nonoverlapping junction ( Fig. 2A). Due to the head-to-tail structure, the sensor was encoded as one polypeptide, which simplified its use compared to usual two-protein split systems. We optimized the reporter by engineering the linker regions and selected variant V6 with the highest luminescence response to 2-5A for further work ( Fig. 2A).
To determine whether V6 was sufficiently bright and stable in live cells, we expressed FLAG-tagged V6 in HeLa cells and stimulated these reporter cells with poly-IC. Poly-IC treatment conditions were selected to produce cleavage of 28S rRNA in the conventional end-point assay using cell disruption 22 . 28S rRNA cleavage was noticeable after one hour of poly-IC treatment and further increased after three and four hours ( Further evidence for specific 2-5A detection was obtained in cells with 2-5A synthetases knocked out. It has been shown that knockout of OAS3 is sufficient to inhibit 2-5A synthesis in human cells 11 . We found that biosensor exhibited a robust response to poly-IC in OAS1-KO and OAS2-KO cells. In contrast, the response was lost in OAS3-KO cells ( fig. S2), providing a confirmation that 2-5A synthesis gave rise to the reporter activity. RNase L may be activated by local production of 2-5A 30 , at secrete sites inside cells (the cytosol). Notably, OASs are present not only in the cytosol, but also in the nucleus [31][32][33] , which further supports that 2-5A accumulation could be nonuniform with subcellular spaces. The availability of live cell 2-5A sensor offers an opportunity to evaluate 2-5A accumulation in individual cellular compartments in situ.
Toward this end, we engineered tagged versions of V6 with nuclear localization signal (NLS) and nuclear export signal (NES). Both variants localized to the expected sites ( Fig. 3C). In the presence of poly-IC, NLS-V6 and NES-V6 produced nearly identical luminescence profiles, which are similar to the trace obtained with untagged V6 (Fig.   3D). This similarity is most simply reconciled with a model of rapid 2-5A equilibration between the two compartments due to diffusion. Indeed, 2-5A produced at the center of a HeLa cell, may take only several seconds to diffuse across the nucleus, through the nuclear pores, and to become evenly distributed between the nucleus and the cytosol To determine whether cellular 2-5A dynamics corresponds to a rapid arrest of translation by RNase L, we measured nascent protein synthesis by puromycin pulse labeling in WT and RNase L -/-A549 cells 5 . Treatment of WT, but not RNase L -/cells with poly-IC halted global translation before ISG induction ( Fig. 4F and 4B). RNase L -/cells exhibited a delayed and incomplete translational attenuation, presumably due to PKR. Translational arrest ahead of ISG induction was also present in HeLa cells (Fig.   S4, B and C). Disengagement of basal protein synthesis before the IFN response was further confirmed using metabolic labeling of nascent proteome with 35 S (Fig. 4G).
IFNs β and  escape the translational shutoff caused by 2-5A 2-5A rapidly stops cell-wide protein synthesis. To examine IFN protein production under these conditions, we treated WT and RNase L -/cells with poly-IC and assayed the media for IFN activity (Fig. 5A). These tests To test whether IFN arises from actively ongoing translation rather than from other potential mechanisms (e.g. delayed secretion of pre-translated IFN stores), we used pulse-treatment with a translation inhibitor, anisomycin (Fig. 5E). In this setting, cells were first treated with poly-IC for three hours, which stopped protein synthesis but did not yet activate a strong transcriptional IFN response. Next, anisomycin was added to arrest all protein synthesis and the cells were kept for three additional hours. Control cells were kept for the same duration without anisomycin. During the last hour, media was changed to remove poly-IC, but anisomycin treatment was continued to keep the cells translationally arrested. When IFN activity in the media was assayed, we found that anisomycin treatment after the 2-5A-induced global translational inhibition ( fig.   S8C), but before the IFN response, blocked IFN production (Fig. 5E). A control experiment showed that anisomycin was compatible with IFN sensing by naïve cells (Fig. 5F). Of note, anisomycin had a mild stimulatory effect on ISG mRNAs due to an unknown mechanism; this effect acted in the opposite direction from blocking IFNs and thus did not affect the suitability of anisomycin as a control in our tests. Together, our experiments indicated that IFNs are indeed translated when the bulk of protein synthesis remains silenced by 2-5A.
A549 cells treated with poly-IC express both type I and type III interferons ( Fig.   6A). To determine which of these IFNs escape RNase L in our experiments, we employed hamster CHO reporter cell lines developed previously for specific detection of human IFNs of type I and type III. Hamster cells do not respond to human IFNs, however the reporter cells are rendered sensitive via expression of chimeric type I and type III human IFN receptors fused to a potent STAT1 docking domain 41,42 . The reporter cells analysis, based on readout of phospho-STAT, indicates the presence of type I and type III IFNs, exhibiting strongest p-STAT response to IFNs- ( Fig. 6B and fig. S10).

Discussion
We have developed a biosensor for 2-5A and determined that this second messenger is synthesized without a delay and mediates immediate dsRNA sensing. 2-5A activates RNase L, which suppresses protein synthesis 43 . This mechanism is potent and attenuates basal cell-wide translation by more than 1,000-fold ( Normally, RNase L is activated as a part of the innate immune system. However, RNase L activation by small molecules could be explored for developing adjuvants and anticancer therapeutics with some of the beneficial effects of mTOR blockers, and with an added advantage of maintaining the protein synthesis activity of the innate immune system. The search for such on demand activators can be facilitated by biochemical, cell-based as well as in vivo applications of the 2-5A biosensor we describe here.

Tissue culture
Cells were grown using ATCC (American Type Culture Collection) or provider recommended conditions in MEM media + 10% FBS (HeLa) or RPMI media + 10% FBS (A549), or DMEM media + 10% FBS (293T). All media were purchased from Gibco, Life Technologies. HeLa and 293T were a gift from Yibin Kang (Princeton University, Princeton). WT, RNase L KO, and OAS KO A549 were generated by the laboratory of Susan Weiss (University of Pennsylvania, Philadelphia). Luminescence assays in live cells were carried out in a plate reader or in 12-well plates at 37 °C.

2-5A extraction from human cells and RNase L cleavage measurements
HeLa cells in 10 cm plates at 80% confluency were treated with 1 μg/mL poly-IC for 6 hours and lysed with 1 ml RIPA buffer (Thermo Fisher Scientific) supplemented with 0.5 mM PMSF protease inhibitor. The lysate was spun down at 12,000 g for 10 minutes at 4°C. The supernatant was passed through a 3 kDa centrifugal filter to collect the fraction of small nucleic acids. Equivalent amounts of small RNA from mock and poly-IC treated samples were used in the luminescence assays. 2-5A activity was tested using 32 p-5'-radiolabeled RNA substrate and recombinant human RNase L as described previously 22 .

Western blots
The V6 and V6-Y312A reporters were subcloned into pCDNA4.TO using primers

qRT-PCR analysis
Cells were harvested in 350 μL RLT buffer (Qiagen) and RNA was purified according to the RNeasy protocol (Qiagen). cDNA was prepared using oligo-dT and a High Capacity RNA to cDNA kit (Applied Biosystems). qPCR was performed using the Power SYBR green PCR mix in a 96 well format on StepOnePlus qPCR instrument (Life Technologies). qPCR primers used in this work were from (IDT). Primer sequences are listed in Table S1.

Ribopuromycilation and 35 S labeling to monitor nascent translation
To generate puromycin-tagged nascent peptides, human cells were treated with 0.1-1 µg/mL of poly-IC for times specified on the figures, after which the growth media was supplemented with 10 µg/mL puromycin (Invitrogen). Puromycin pulse lasted for 5 min. Cells were trypsinized and harvested in NuPAGE LDS sample buffer for western blot analyses. Proteins were separated by 10% BisTris PAGE (NuPAGE), and transferred on PVDF membranes (Life Technologies). The membrane was stained with Ponceau to normalize for sample loading, then washed and blocked with 5% non-fat dry milk in TBST buffer. The membranes were probed with 1:1000 mouse anti-puromycin antibody (EMD Millipore) that binds to de novo synthesized proteins, followed by horseradish peroxidase-conjugated goat anti-mouse secondary antibody (1:10,000, Jackson ImmunoResearch).
Metabolic labeling with 35 S was conducted using the following procedure. After cell treatment with poly-IC, media was changed to methionine-free RPMI + 10% FBS supplemented with 11 μCi EasyTag™ EXPRESS35S Protein Labeling Mix (Perkin Elmer). Cellular proteins were resolved by 10% BisTris PAGE (NuPAGE) and analyzed by phosphorimaging.

Secreted IFN detection by qPCR in conditioned media
A549 WT and RNase L KO cells in 12-well plates were transfected with 1 μg/mL poly-IC and lipofectamine 2000 for 0.5, 2 or 6 hours. At the end of the time course, cells were washed four times with 1 mL of growth medium (RPMI+10% FBS) to remove residual poly-IC. After the washes, fresh 1 mL medium was added and the cells were kept in fresh media for one hour to allow for protein secretion.
Treatments of naive cells with conditioned media were done in a separate 12well plate and using cells seeded 1 day before the analysis. Naïve cell media was replaced with 1 mL of the conditioned medium from above. Cells were grown in the conditioned media for 16 hours and harvested in 300 μL RLT (Qiagen). For pulse-chase experiments in Fig. 5C, 10 μM anisomycin was added 3 hours after poly-IC addition.
Anisomycin level was additionally maintained during the last hour allocated for IFN secretion into fresh media. To exclude a possible inhibitory effects of anisomycin on IFN response (Fig. 5D), WT cells were treated for 16 hours with 1000 U/mL recombinant IFN-β in plan or conditioned media.

RNA-seq
Poly-A + RNA sequencing was conducted and processed as described previously in our work 18,46 . The datasets (Fig. 5C) were deposited to GEO database under accession number GSE120355.

Reporter assay to detect type-I and type-III IFNs
Previously developed reporter cell lines, which are selectively sensitive to either human type I 41 or type III 42 IFNs, were used to detect the presence of IFNs in the media.
Briefly, 4x10 6 cells were treated with each sample and incubated at 37 °C for 20 min.

Diffusion calculations
First, we estimate diffusive relaxation rate in a simplistic cell without the nucleus. This will define how fast a non-uniform concentration of 2-5A should relax in a sphere the size of a cell. The slowest mode to relax will be the spherically symmetric one with a single peak at the cell center, and zero gradient at the cell membrane (reflecting boundary conditions). The diffusion equation with no sources or sinks is: ∂c/∂t = D·∇ 2 c, which in the case of spherical symmetry simplifies to ∂c/∂t = D·(1/r 2 )·∂/∂r(r 2 ·∂c/∂r). This simplifies further if we make the substitution c = u/r, and use ∂c/∂r = (1/r)·∂u/∂r -(1/r 2 )·u and ∂c/∂t = (1/r)·∂u/∂t, yielding (1/r)·∂u/∂t = D·[(1/r 2 )·∂/∂r(r·∂u/∂r -u)], and finally ∂u/∂t = D·∂ 2 u/∂r 2 . We can expand u in eigenfunctions of the right hand side, which are just sines and cosines, and the general solution for u(r,t) will be of the form: u(r,t) = Σu k ·sin(k·r)·exp(- k ·t) plus similar terms for cosines. Rather than solving for complete generality, we note that we want the slowest decaying mode for which c is finite at r=0 (eliminating cosines), and has a no-flux boundary condition at the sphere's radius R. No flux at R implies ∂c/∂r = 0, which implies that ∂u/∂r·(1/u) = 1/r at r = R. Therefore, k·cos(k·R)/sin(k·R) = 1/R, i.e. tan(k·R) = k·R, and we need to solve this equation to obtain k. The lowest-k solution (slowest mode of diffusion) is k·R = 4.49. After defining k* = 4.49/R, we find the corresponding relaxation rate of this mode from the diffusion equation: (- k* )·u = -D·k* 2 ·u, so that  k* = D·k* 2 .
Up to this point, we have neglected contributions to the solution for u(r,t) that are zero when acted on by ∇ 2 , so we can add terms c 0 (t)·r or c 1 (t) to u(r,t), or equivalently terms c 0 (t) or c 1 (t)/r to c(r,t). Only the constant term satisfies continuity at the center of the sphere for c(r,t), as well as zero flux at the boundary, and this constant term does not decay in time. So the solution for c(r,t) at long times has the form: c(r,t) = c 0 + (c*/r)·exp(- k* ·t)·sin(k*·r). It approaches the constant c 0 at very long times, with the leading spatially non-uniform term decaying at a rate:  k* = D·k* 2 = D·(4.49/R) 2 .