The innate immune sensor IFI16 recognizes foreign DNA in the nucleus by scanning along the duplex

The ability to recognize foreign double-stranded (ds)DNA of pathogenic origin in the intracellular environment is an essential defense mechanism of the human innate immune system. However, the molecular mechanisms underlying distinction between foreign DNA and host genomic material inside the nucleus are not understood. By combining biochemical assays and single-molecule techniques, we show that the nuclear innate immune sensor IFI16 one-dimensionally tracks long stretches of exposed foreign dsDNA to assemble into supramolecular signaling platforms. We also demonstrate that nucleosomes represent barriers that prevent IFI16 from targeting host DNA by directly interfering with these one-dimensional movements. This unique scanning-assisted assembly mechanism allows IFI16 to distinguish friend from foe and assemble into oligomers efficiently and selectively on foreign DNA. DOI: http://dx.doi.org/10.7554/eLife.11721.001


Introduction
The host innate immune system detects infection by directly recognizing molecular signatures associated with pathogens Medzhitov and Janeway, 2000). Remarkably, such signatures include universal building blocks of all life, such as DNA and RNA Orzalli and Knipe, 2014;Paludan and Bowie, 2013). In the cytoplasm, the immune system relies on the absence of endogenous DNA, and thus marks all detected DNA as 'foreign' (nonself) (Orzalli and Knipe, 2014;Paludan and Bowie, 2013). However, DNA viruses often evade the cytosolic detection machineries, as their genomes are not exposed until reaching the nucleus (Orzalli and Knipe, 2014;Paludan and Bowie, 2013). The host counters this infection strategy in the nucleus by directly assembling supramolecular signaling platforms that trigger inflammatory responses on invading foreign DNA, but not on its own genomic material (Johnson et al., 2013;Li et al., 2012;Kerur et al., 2011). Although key players that target foreign dsDNA in the host nucleus have been identified (Orzalli and Knipe, 2014;Paludan and Bowie, 2013), the molecular mechanisms by which these sensors distinguish self from nonself dsDNA remain unknown.
The interferon-inducible protein 16 (IFI16) is a key innate immune sensor that detects foreign dsDNA and uses it as a scaffold to assemble supramolecular signaling platforms in both the host nucleus and cytoplasm (Unterholzner et al., 2010;Johnson et al., 2013;Li et al., 2012;Kerur et al., 2011;Orzalli et al., 2012) ( Figure 1A). IFI16 plays a central role in defense against a number of pathogens (e.g herpes simplex virus-1) (Unterholzner et al., 2010;Johnson et al., 2013;Li et al., 2012;Kerur et al., 2011;Orzalli et al., 2012). On the other hand, persistent IFI16 signaling is associated with autoimmunity (e.g. Sjö gren's syndrome) (Mondini et al., 2007;Choubey et al., 2010;Mondini et al., 2010;Gugliesi et al., 2013;Smith and Jefferies, 2014). The molecular mechanisms by which IFI16 selectively targets foreign dsDNA remain unknown. To establish a functional signaling platform, IFI16 must overcome two challenges. First, individual IFI16 molecules must be able to locate one another on large pathogen genomes with sizes ranging from 10 5 to 10 6 base pairs (bps). Second and more importantly, this assembly mechanism can only take place on foreign dsDNA and must be inhibited on host dsDNA ( Figure 1A). Here, we report the observation of a unifying molecular mechanism that explains how IFI16 resolves these central issues in initiating its foreign-dsDNA sensing pathways.

Results and discussion
To identify the mechanisms underlying assembly of IFI16 signaling platforms on DNA, we monitored the oligomerization kinetics of FRET donor and acceptor labeled IFI16 on naked dsDNA (FRET: fluorescence resonance energy transfer; Figure 1B and Figure 1-figure supplement 1). Previous work demonstrated the existence of such oligomers and reported on their equilibrium binding properties but did not provide insights into the assembly mechanisms . Using various dsDNA fragment sizes present in excess, we observe that the assembly rate increased non-linearly and by 50-fold from 60 to 200 bps dsDNA, above which it stayed constant (up to 600 bps; Figure 1B,C). With a dsDNA-binding footprint of~15 bp for one IFI16 , our results indicate that about 4 copies are required to initiate assembly, and about 10 IFI16 molecules are required for optimal oligomeric assembly ( Figure 1C). Further, the assembly rate constants scaled linearly with the IFI16 concentration for all measured DNA lengths (Figure 1-figure supplement 1 and Supplementary file 1A), indicating that a purely cooperative assembly mechanism is unlikely. In line with this observation, previous work reported relatively small contributions of eLife digest The immune system defends us from attacks by viruses, bacteria and other microbes. One way that the immune system can identify these invaders is by detecting genetic material from the microbes. Like us, many of these microbes have genetic material made of a twostranded molecule called DNA. A protein called IFI16 is an important sensor in immune cells that can bind to foreign DNA, but not to the DNA of the host. If the ability to discriminate between host and foreign DNA is lost, then immune cells start to attack the body's own tissues, which can lead to severe "autoimmune" diseases. However, it is not clear how IFI16 and other sensor proteins are able to tell foreign and host DNA apart.
DNA in host cells is packaged in particular proteins to form structures called nucleosomes that can make it difficult for other proteins to bind to the DNA. When foreign DNA enters the cell, IFI16 promotes the formation of nucleosomes on these molecules, but the nucleosomes are further apart than those in host DNA, which leaves longer stretches of "exposed" DNA between the nucleosomes. In 2014, a group of researchers reported that multiple molecules of IFI16 associate with each other and form clusters on exposed foreign DNA. Here, Stratmann, Morrone et al. -including some of the researchers from the earlier work -investigated how IFI16 selectively binds to foreign DNA using a combination of biochemical and single-molecule techniques.
The experiments show that to start with, a few molecules of IFI16 bind to stretches of exposed foreign DNA and then move along the DNA strands. As more molecules of IFI16 bind to this DNA, they bump into each other and start to form clusters that eventually become immobile. Further experiments show that the more tightly packed nucleosomes in host DNA molecules act as a barrier to IFI16 cluster formation because they interfere with the ability of the protein to move along the DNA.
Stratmann, Morrone et al.'s findings show that IFI16 can only trigger immune responses when it binds to stretches of exposed DNA that are long enough to allow the assembly of IFI16 clusters. The next challenge will be to see whether other DNA sensors employ similar strategies to detect foreign DNA.  The time-dependent changes in the emission ratio between FRET donor and acceptor labeled IFI16 (50 nM) were monitored at 33 mg/ml of each dsDNA (e.g. sixfold higher than the dissociation constant for 39-bp dsDNA ). Lines are fits to a first-order exponential equation (see Figure 1-figure supplement 1 for 25 nM protein). All shown representative experiments were performed at least three times. (C) A plot of observed assembly rates (k assm s) vs. dsDNA-sizes (see also Figure 1-figure supplement 1). (D) 1D-diffusion assisted assembly mechanism can explain the observed assembly profile of IFI16. 1. At the same mass-concentrations, the number of individual dsDNA fragments present in each assay is inversely proportional to the length of dsDNA. 2. Individual IFI16 molecules initially bind dsDNA at random positions and diffuse one-dimensionally while searching for other respective protomers; the number of IFI16 residing on the same dsDNA fragment should be proportional to the length of dsDNA (e.g. there are four times more individual 150-bp fragments than 600-bp fragments) 3. IFI16 fails to assemble into an oligomer on dsDNA shorter than 60 bp (indicated by a red arrow pointing left). The saturating rates can be explained if the final FRET signals arise from formation of distinct optimal oligomers. DOI: 10.7554/eLife.11721.003 Figure 1 continued on next page cooperativity in oligomerization with Hill constants near 2 for DNA substrates up to 2000 bp .
Our ensemble-averaged, solution-phase observations of the faster assembly on longer dsDNA suggest a model in which IFI16 scans along dsDNA to increase the probability of encountering other IFI16 molecules ( Figure 1D). To directly test such a mechanism, we used single-molecule fluorescence imaging to track the movements of individual Cy5-labeled IFI16 molecules on stretched, double-sided attached l-phage dsDNA (ldsDNA; 48.5 kbps) ( Figure 2A and Video 1). Figure 2B shows that individual IFI16 molecules one-dimensionally (1D) diffuse on ldsDNA while bound for several seconds. The diffusion coefficient of IFI16 increased with ionic strength, indicating that IFI16 does not maintain a continuous electrostatic interaction with the dsDNA backbone, but instead moves along the ldsDNA scaffold by executing microscopically small steps (Blainey et al., 2006) ( Figure 2B and Single-molecule intensity analysis revealed that the lower limit of the number of IFI16 molecules found in immobile clusters is equivalent to eight protomers ( Figure 2D and , which also corroborate the optimal complex (ten protomers) suggested from Figure 1C. Furthermore, individual IFI16 molecules either formed new clusters or joined other clusters in a stochastic manner, and the immobile clusters formed faster with higher IFI16 concentrations ( Figure 2E). The rate of addition of IFI16 molecules to clusters is independent on the size of the existing cluster, confirming the absence of strong cooperativity in assembly ( Figure 2C; bottom panel).
The 1D diffusion of IFI16 on dsDNA explains why the assembly rates increase with the DNA length in the bulk experiments ( Figure 1C). With the longer dsDNA acting as an antenna, it allows binding of more IFI16 while 1D diffusion facilitates dynamic association ( Figure 1D). The saturation of the assembly rate ( Figure 1C) can be   explained by the square dependence of the diffusional search time on length: at a sufficiently long dsDNA length, the dissociation rate of an individual IFI16 will be faster than the time needed to scan along the entire length of the DNA. In addition, longer DNA substrates work no longer as antennae, but as traps, since individual IFI16 molecules are farther apart and thus less likely to encounter one another (Hu et al., 2006;Turkin et al., 2015;2016). Overall, the results of our bulk and single-molecule experiments are consistent with the dsDNA-size dependent binding in vitro , which also correlates with the IFI16-induced inflammatory responses in vivo (Unterholzner et al., 2010). Thus, we propose that the 1D-diffusion mediated assembly plays a key role in regulating the overall IFI16mediated immune responses.
It has long been speculated that chromatinization acts as the key feature that allows IFI16 to distinguish host from foreign DNA in the nucleus (Kerur et al., 2011;Unterholzner and Bowie, 2011;Li et al., 2012;Orzalli et al., 2012;; IFI16 oligomerizes on exposed invading foreign-dsDNA before it becomes hetero-chromatinized. Previous in vivo work demonstrated that transfected chromatinized SV40 DNA is able to evade IFI16 oligomerization and downstream responses (Orzalli et al., 2013). Nevertheless, the molecular mechanism by which IFI16 could use chromatinization to distinguish self from nonself has yet to be identified. To directly address this issue, we first used a competition-binding assay to investigate how IFI16 interacts with dsDNA fragments containing two nucleosomes with varying spacer sizes (6, 30, 50, and 70 bps; Figure 3A and Figure 3-figure supplement 1). Here, di-nucleosomes with 6-, 30-, and 50-bp spacer failed to compete against IFI16-bound FAM-labeled 70-bp dsDNA, ( Figure 3B). On the other hand, the di-nucleosome with 70-bp spacer competed similarly as 70-bp naked dsDNA, but significantly more weakly than naked 300-bp dsDNA ( Figure 3B). In FRET assembly assays, di-nucleosomes with spacers shorter than 70-bp failed to support assembly ( Figure 3C), consistent with our FRET kinetics assays using naked dsDNA ( Figure 1B). The 70-bp spacer di-nucleosome supported oligomerization of IFI16; however, the assembly kinetics was again similar to that of naked 70-bp dsDNA, but not that of naked 300-bp dsDNA ( Figure 3C). Taken together, these results show that efficient IFI16 cluster formation requires a minimal length of 50-70 base pairs of exposed dsDNA. Considering that the size of dsDNA linker between two nucleosomes is about 20 to 30 bps in mammals (McGhee et al., 1983), these results directly support the hypothesis that chromatinization is a key deterrent for preventing the assembly of IFI16 signaling platforms on self-dsDNA.
To test whether the inhibitory effect of chromatin directly arises by interfering with the 1D diffusion of IFI16, we visualized the movement of individual IFI16 molecules on DNA with nucleosomes. We introduced randomly localized nucleosomes in ldsDNA using recombinant human histone octamers and tagged nucleosome positions with fluorescent antibodies against the N-terminal tail of histone H4 (Figure 4-figure supplement 1). The application of hydrodynamic flow resulted in the single IFI16 molecules being pushed to one direction and clustering at nucleosomal sites on ldsDNA, unable to overcome the octamers by diffusion ( Figure 4A (Figure 3), and confirm that nucleosomes directly restrict the 1D diffusion of IFI16 and consequently limit the assembly of IFI16 signaling platforms on dsDNA.
The molecular mechanism by which innate immune sensors distinguish self from foreign dsDNA in the host nucleus has been a major unresolved question in innate immunology (Kerur et al., 2011;Unterholzner and Bowie, 2011;Li et al., 2012;Orzalli et al., 2012;Johnson et al., 2014). The oligomerization of IFI16 on under-chromatinized foreign DNA plays a key role not only in initiating inflammatory and antiviral responses, (Monroe et al., 2014;Orzalli et al., 2012;Kerur et al., 2011), but also in regulating the hetero-chromatinization and silencing of viral dsDNA Orzalli et al., 2013). By using time-resolved bulk and single-molecule fluorescence assays, we demonstrate here that IFI16 ID scans along exposed dsDNA to assemble into Video 2. Cy5-labeled IFI16 molecules (3 nM) moving and clustering on single-biotinylated dsl-DNA at 160 mM KCl at constant flow from left to right. This video is related to Figure 2. DOI: 10.7554/eLife.11721.011    distinct clusters and that chromatinization is sufficient to inhibit IFI16 from targeting host dsDNA for assembly. In vivo, this 1D scanning mechanism allows a limited number of IFI16 molecules to allocate each other on large genomes of invading pathogens. In combination with 3D sampling of binding sites on a collapsed DNA molecule, this process optimizes the oligomerization and downstream signaling time. While the clustering on dsDNA presents a tempting explanation for the role of IFI16 in viral gene silencing, future in vivo experiments await to test this. IFI16 belongs to the family of AIM2-like receptors, which include other nuclear and cytosolic foreign dsDNA-sensors. It will be interesting to determine whether and how these other related sensors use exposed dsDNA as a 1D 'digital ruler' to regulate their signaling platform assembly. This family of sensors is implicated in a number of autoimmune disorders (Mondini et al., 2007;Choubey et al., 2010;Gugliesi et al., 2013;Smith and Jefferies, 2014); how regulation of assembly is disrupted may provide insights into these afflictions.

Protein expression and purification
Human full-length IFI16 and IFI16HinAB were cloned and expressed using E. coli T7 express cells (NEB) as a C-terminally His6-tagged protein as described in .

DNA ligand preparation
dsDNA shorter than 90-bp were obtained from Integrated DNA Technologies (IDT) as described in . The complementary strands were dissolved and mixed in 1:1 molar ratio, melted at 95˚C for 10 min, and the temperature was lowered to 25˚C at a rate of 1˚C/ min. Ligands of greater length were obtained by polymerase-chain reaction (PCR) using the Maltose Binding Protein fusion tag cloning sequence as template and primers of appropriate sequence for a final length as indicated in the paper. Plasmids containing the Widom-601/603 sequence with indicated linker lengths were a kind gift of Dr. Gregory Bowman. The nucleosomal DNA was obtained by PCR from these constructs with appropriate primers. All substrates were gelpurified.

Fluorescent labeling
DyLight-550, DyLight-650, or Cy5 fluorophore was incorporated to IFI16 using maleimide chemistry (purchased from Thermo Scientific and Invitrogen) and was performed as described in . The label to protein ratio was~1:1. Fluorescein-labeled dsDNA72 was obtained from IDT.

Octamer refolding and nucleosome reconstitution
Lyophylized Xenopus laevis histones H1A, H2A, H3, and H4 were a kind gift of Dr. Cynthia Wolberger. Octamer refolding and nucleosome reconstitution was performed as described in Luger et al. (Luger et al., 1999), at a 2:1 molar ratio of octamer:DNA. An agarose gel of reconstituted nucleosomes is shown in Figure 3-figure supplement 1.

Bulk biochemical assays
All absorption, fluorescence anisotropy, and fluorescence excitation/emission experiments were performed in a Tecan Infinite M1000. All experiments were performed at least three times and the fits to data were generated by Kaleidagraph software (synergy).

Competition binding assays
All reactions were performed in 40 mM HEPES pH 7.4, 160 mM KCl, 5% glycerol, 1 mM EDTA, 0.1% triton-X-100, 5 mM DTT (Reaction Buffer). Here, 300 nM IFI16 and 4.5nM fluorescein-labeled dsVACV72 were incubated together at room temperature for 20 min. Increasing concentrations of competing DNA were added to the reaction to a final concentration of 100 nM IFI16 and 1.5 nM dsVACV72, and the changes in fluorescence anisotropy were recorded as indicated in .

FRET time dependence assays
All reactions were performed in reaction buffer. 66 mg/ml of each dsDNA or di-nucleosomes was placed in the plate wells, and the reaction was initiated by adding an equivalent volume of IFI16-550 and IFI16-650 (1:1 molar ratio) to the indicated final concentration. The dead time between addition of IFI16 and the first measurement was 15-20 s. The final dsDNA molar-concentrations are at least sixfold higher than their determined binding constants by fluorescence anisotropy assays described in , and the FRET ratio for each time point was calculated by dividing the acceptor emission (678 nm) by the donor emission (574 nm) .

Fluorescence microscopy assays
Microscope coverslips (Corning) were plasma-cleaned and activated with 100 mM KOH, silanized with 3-Aminopropyl-triethoxysilane in acetone and functionalized with PEG-NHS and biotin-PEG-NHS (Laysan-Bio) in sodium bicarbonate with a 1:3 mass ratio (Tanner and van Oijen, 2010). Streptavidin (Sigma Aldrich) was used for anchoring the biotinylated DNA substrates to the coverslip surface. Flow channels were constructed with custom-made PDMS chips to obtain dimensions of 10 mm length, 100 mm height, and 1 mm width, and connected to a syringe pump to allow constant flow during the measurements. Single-molecule assays were performed in 40 mM HEPES (pH 7.4), 160 mM KCl, 1 mM EDTA, 2 mM DTT, 10% glycerol, 0.1% Triton-X100, 250 mg/ml BSA, 1 mM Trolox, 40 mM glucose, 250 nM glucose oxidase, 60 nM catalase, unless otherwise stated. All assays were performed at room temperature. We applied 100 nM YoPro-1 iodide (Life Technologies) at the end of the measurements to visualize the DNA substrates.
Reactions were illuminated with a 488-nm and 641-nm laser (Coherent), and images were acquired with an EMCCD camera (Hamamatsu). We used MetaVue imaging software (Molecular Devices) for data acquisition and ImageJ and R for analysis.

l-DNA templates for microscopy
Lambda-DNA (New England Biolabs) was biotinylated at one or both ends by ligation of the respective complementary biotinylated oligos according to Tanner et al. (Tanner and van Oijen, 2010) (oligo sequences are given in Supplementary file 1B; purchased from IDT). Single-biotinylated DNA templates were stretched by constant flow (20 ml/min) throughout the experiments (IFI16 clustering and nucleosome assays). Double-biotinylated DNA templates were applied to the flow cell at high flow velocity (100 ml/min). This allowed binding of the DNA to two biotin moieties, while the DNA was stretched. Free DNA was washed out, IFI16 applied to the flow cell, and flow was then stopped for acquisition of the single molecule diffusion traces.
To reconstitute nucleosomes, recombinant histone H2A/H2B dimers and H3/H4 tetramers (New England Biolabs) were assembled on biotin-l-DNA and biotin-601 sequence (Epicypher) by salt-gradient dialysis (2 M to 0.3 mM NaCl in 5 steps, each step with an incubation time of at least 2 hr) in 10 mM Tris/HCl, pH 7.4, 0.1 mM EDTA, 0.5 mM DTT. We tested nucleosome reconstitution by an EMSA assay on digested l-DNA (Figure 4-figure supplement 1). For this, we generated DNA fragments by digestion with EcoRI (NEB), purified them (Qiagen DNA spin columns) and reconstituted nucleosomes in the concentration ratios that we also used for full-length l-DNA.
Single molecule co-localization of IFI16 with nucleosomes tagged with antibodies Antibodies against human histone H4 and H2B were chosen to target the N-terminal histone tails (Santa Cruz, sc-8657 and sc-8650), and labeled with Atto488-NHS (Life technologies) in PBS at pH 7.0. Labeled antibodies were negatively tested against bare DNA and Ifi16 clusters in the microscopy assay, and showed high specificity for nucleosomal DNA preparations only.
l-DNA templates, loaded with nucleosomes and tagged with anti-H4, were constantly stretched in the flow channel. Ifi16 co-localized strongly with the nucleosome signal (Video 3, Figure 3-figure supplement 1). In contrast, biotinylated 601-sequence prepared with nucleosomes and tagged with anti-H4-Atto488, hardly showed co-localization with Ifi16, indicating, that there is not sufficient exposed dsDNA available for binding (Figure 4-figure supplement 2).

Drift correction
For the clustering experiments, we applied a flow of 0.02 ml/min to our flow cell design of 0.1 mm height and 1 mm width, giving a volumetric flux of 0.33 cm/s. We expect the DNA molecules to be on average 0.2 mm away from the surface (Tafvizi et al., 2008), giving a velocity at this distance y of (Tafvizi et al., 2011): The Stokes drag force that acts on the DNA and bound IFI16 molecules is then described by F ¼ 6phrv 1 þ 9r 16y ¼ 1:5fN; with viscosity h, radius r and distance y. We can define this force within the diffusion coefficient D by using the drift velocity v, calculated from single-molecule trajectories (Leith et al., 2012): with v ¼ P all traj: j x j;finalÀ x j; initial P all traj: j t j;finalÀ x j; initial »0:127 m=s: In order to evaluate the effect of flow on the clustering mechanism, we implemented a 1D-random walk Monte-Carlo simulation (Figure 2-figure supplement 4). Here, we calculated the expected search time for two IFI16 molecules (with D=0.026 mm 2 /s) to allocate each other on a l-DNA molecule congested with a varying amount of other diffusing IFI16 molecules (10, 50, 100 molecules). As all particles are equal, we segmented the DNA according to the total number of molecules bound and calculated the effective distance between two particles by using the absolute distance modulo the segment length in order to take the periodic boundaries into account.
We allowed a maximum distance of 10 nm to consider two particles having met and dimerized. To simulate flow similar to our experimental conditions, to each random step the term vdt was added (The Python code is found in Source code 1).