The universally conserved GTPase HflX is an RNA helicase that restores heat-damaged Escherichia coli ribosomes

HflX, a ribosome-binding GTPase, is classified as a member of the Obg–HflX superfamily (Verstraeten et al., 2011). The Obg-related GTPases, comprising a group of ancient GTPases of the translation factor–related GTPase class, exist in all domains of life (Leipe et al., 2002). Stimulation of the intrinsic GTPase activity of HflX upon interaction with 70S ribosome and ribosomal subunits has been well established (Noble et al., 1993; Shields et al., 2009; Huang et al., 2010; Ash et al., 2012). It has been shown that HflX preferentially binds to the 50S subunit in the presence of GTP (Blombach et al., 2011; Zhang et al., 2015). The protein shares several similarities to bacterial GTPases that interact with the ribosomal subunits, some of which are known to play key roles in ribosome biogenesis (Caldon and March, 2003; Britton, 2009). Thus, HflX has been initially implicated in ribosome biogenesis (Sato et al., 2005; Schaefer et al., 2006). Interestingly, however, although an earlier study (Shields et al., 2009) proposed that HflX functions under stress, recent studies have confirmed that Escherichia coli HflX splits 70S ribosomes (Zhang et al., 2015; Coatham et al., 2016) and acts as an antiassociation factor for the 50S subunit in the presence of GTP under heat stress (Zhang et al., 2015).

The crystal structure of Sulfolobus solfataricus HflX (Wu et al., 2010) displays two-domain architecture (N-terminal and GTP-binding domains), whereas E. coli HflX consists of three domains (Fig. 1 A): well-conserved N-terminal domain 1 (ND1) and 2 (ND2; GTPase domain) followed by an additional C-terminal domain. A fork-like helical domain (termed hereafter as linker helical domain) bridges ND1 and ND2.

Intriguingly, E. coli HflX binds and hydrolyzes not only GTP but also ATP upon ribosome binding (Dutta et al., 2009; Jain et al., 2009; Shields et al., 2009; Blombach et al., 2011). A more recent study (Jain et al., 2013) characterized ND1 to be a new ATP-binding domain in E. coli HflX, where it was reported to bind and hydrolyze ATP in the presence of 70S ribosome as well as 50S ribosomal subunit. Nevertheless, the functional role of the ATPase activity of HflX remains undefined.

In this study, we elucidate that HflX is an ATP-dependent helicase that exhibits RNA-unwinding activity. Atomic force microscopy (AFM) visualization clearly manifests its affinity and unwinding catalysis on large subunit ribosomal RNA (rRNA) in the presence of ATP. Furthermore, a cryo-EM structure of the 50S–HflX complex in the presence of ATP and GTP illuminates the mechanism of RNA-unwinding action by the protein. Our structural and functional analyses demonstrate a critical role of the linker helical domain in modulating 23S rRNA conformation. Furthermore, in vitro translation and cell survival assays provide clear evidence that HflX is capable of rescuing heat-damaged ribosomes and promoting cell survival after heat stress.

A hallmark of RNA helicases is the ability to couple free energy from ATP hydrolysis to mechanical work used for unwinding of double-stranded RNA (dsRNA). Remarkably, high-salt–washed (partially deformed; Moore et al., 2008; Pulk et al., 2010) as well as heat-shocked 50S subunits showed better ability to stimulate ATP hydrolysis on HflX compared with normal 50S subunit. The fact that ATP on HflX is activated by aberrant form of 50S subunit favors a physiological role of the protein. The combined biochemical and structural results reported in this study support a molecular model in which ND1 of the 50S-bound HflX (in conjunction with the linker helical domain) can modulate heat-inactivated 23S RNA conformation and as a consequence, help in the survival of the bacteria.

ND1 has been recognized as ATP-binding domain (Jain et al., 2013). Although a DALI search (an online server for 3D structure comparison of proteins; Holm and Rosenström, 2010) did not show significant structural similarity between ND1 of the S. solfataricus HflX crystal structure (Wu et al., 2010) and known ATP-binding domains, when we used ND1 (Fig. 1 A) of the recent cryo-EM structure (Protein Data Bank [PDB] accession number 5ADY) of E. coli HflX (bound to the 50S subunit) for our DALI search, several known ATPase domains were identified (Table S1). Intriguingly, consideration of the ND1 along with the linker helical domain for the DALI search displayed an ATP-dependent DNA helicase as one of the structural neighbors (Table S1; PDB accession number 4CGZ; Z-score, 2.3; root mean square displacement, ∼4 Å). This observation prompted us to explore the possibility of helicase activity of E. coli HflX.

Because HflX is known as a ribosome-binding protein, we analyzed its RNA helicase processivity by performing a duplex unwinding assay. A short 24-nt oligonucleotide, containing a region of self-complementary sequence of 6 bp with a single-stranded overhang of 12 nt at the 5′ end, was used as the substrate. A classical electrophoretic mobility-shift assay clearly exhibited the duplex RNA-unwinding activity of HflX (Figs. 1 B and S1 A). RNA helicases frequently function in an ATP-dependent manner. Indeed, the result shows that only ATP stimulates HflX’s unwinding activity on the dsRNA substrate, whereas GTP has no effect. Mg²⁺ ion is known to play a critical role for ATPases (Henn et al., 2008). Inhibition of helicase activity in the presence of EDTA suggests involvement of Mg²⁺ ions (Fig. S1 B).

To understand the significance of HflX’s RNA helicase activity in the context of ribosome function, we visualized the unwinding catalysis of HflX on large-subunit rRNA using high-resolution AFM imaging (Fig. 1; cryo-EM visualization was also done; Fig. S3 G). AFM imaging provided a means of directly visualizing unwinding catalysis at single-molecule resolution (Henn et al., 2001). Purified rRNA from the 50S subunit was incubated with the protein in the presence or absence of ATP and visualized by AFM. Our AFM results demonstrate that HflX can bind and unwind large-subunit rRNA in an ATP-dependent manner (Fig. 1, C–E). The rRNA molecules were seen (Fig. 1 C) in condensed form (mean height distribution of the molecules 0.8–2 nm). In the absence of ATP, HflX was apparently bound to RNA, but no unwinding catalysis was observed (Fig. 1 D). In contrast, treatment of rRNA with HflX along with ATP clearly showed single-stranded RNA intermediates (single strand height, 0.6 nm; Fig. 1 E). The AFM images showed (Fig. 1 E) several Y-shaped structures (arms comprising two single-stranded RNA regions and a base formed by dsRNA) representing intermediates trapped during unwinding process. Some loop-like structures (created by local strand separation upon unwinding) were also detected. Binding of the protein at the strand-opening junction of the RNA could be identified from an increased height value (∼1.4 nm). However, in contrast with the conventional helicases showing directional movement along a nucleic acid strand, HflX apparently functions in a localized manner.

Recent cryo-EM structure of the 50S subunit–bound E. coli HflX in the presence of GTP (GMP-PNP; Zhang et al., 2015) has revealed the binding site for HflX on the 50S subunit. To gain further insight into the mechanism of the ATP-dependent helicase action of HflX, we applied cryo-EM and single-particle reconstruction techniques for structural analysis of the 50S-bound HflX complex in the presence of both ATP and GTP (nonhydrolyzable analogues).

The HflX-bound 50S subunit complex in the presence of nonhydrolyzable nucleotides (AMP–PNP and GMP–PNP) was purified using cosedimentation, followed by density gradient separation. The 3D cryo-EM map of the 50S–HflX–ATP–GTP complex (designated in this study as the ATP–GTP complex; gold standard resolution, ∼8 Å; Fig. S3 A) was generated (Fig. 2 A). Using a flexible fitting approach, a quasiatomic model of the HflX-bound 50S subunit complex was generated (Fig. 2 B). The resolution of the map was reasonably good for molecular interpretation with the help of the atomic model (Trabuco et al., 2008).

The recent cryo-EM structure of the 50S–HflX–GMP–PNP complex (hereafter GTP complex) described interaction of different domains with the 50S subunit (Zhang et al., 2015). The overall arrangements of different domains in our ATP–GTP complex (Fig. 2 C) appeared to be very similar to that of the GTP complex, indicating that HflX’s binding site on the 50S subunit is specific. The 50S–HflX complex in the presence of only AMP–PNP was also gradient purified. The 3D cryo-EM map of that 50S–HflX–AMP–PNP complex (hereafter ATP complex) showed weaker density of the protein on the map (Fig. S3 D), indicating that GTP is likely required for tight association of the protein with the 50S subunit. Western blot analysis of the gradient-purified 50S–HflX complexes, showing higher affinity of the protein when GTP is present along with ATP, further corroborates this postulation (Fig. S1, F and G). Differential binding of HflX to the 50S subunit in the presence or absence of nucleotides also confirmed the association of nucleotides with the protein in the complexes we purified.

Close inspection of the arrangements of HflX’s different domains revealed that the ND2 and C-terminal domains remained in similar conformation in our complex (Figs. 3 A and S4), as seen in the GTP complex (PDB accession number 5ADY). In contrast, ND1 and the linker helical domain in particular showed substantial conformational changes (Fig. 3, A, D, and E) in the ATP–GTP complex when compared with the GTP complex, indicating the presence of ATP in ND1. Furthermore, a previous study has elucidated that in the presence of both nucleotides, ND2 accommodates GTP, whereas ND1 binds to ATP (Jain et al., 2013). Thus, it is conceivable that nucleotides are bound to HflX in a similar manner on our map. In the GTP complex, GTP was present on ND2 and ND1 was in apo conformation (Zhang et al., 2015). The apo form of ND1 may be considered to be in a conformation similar to the ADP-bound form (Fig. 3 C). Thus, observed ND1 conformational changes can be considered to occur because of ATP hydrolysis (Fig. 3 D). Interestingly, it is clearly visible that a stretch of density connected the linker helical domain to the P-loop region of domain V rRNA in the ATP–GTP complex (Fig. 3 B), whereas no density was seen in the corresponding region of the GTP complex (Fig. 3 C). A similar stretch of density attributable to the linker helical domain is visible in the ATP complex as well (Fig. S3 F). Although the RNA unwinding activity of HflX seems to be nonspecific, the specific binding site of HflX on the 50S subunit fosters the view that the substrate for its helicase activity is likely the region of 23S rRNA, where ND1 is placed.

We observed that the linker helical domain directly interacted with the domain V region of 23S rRNA in the ATP–GTP complex, and drastic movement of this domain was detected when compared with its conformation in the GTP complex. It appeared that the linker helical domain has a determinant role in eliciting RNA unwinding activity.

To check the RNA helicase activity of ND1 in the presence or absence of the linker helical domain, we made two ND1 constructs: (1) G-domain (ND2) and C-terminal domain deleted (ΔGC); and (2) ND1 only (ND1ΔH; Fig. 4 A). RNA unwinding activity was investigated using an electrophoretic mobility-shift assay, which clearly demonstrated that although ND1 with linker helical domain (ΔGC) showed RNA unwinding activity, ND1 without the linker helical domain (ND1ΔH) could not function as RNA helicase (Figs. 4 B and S1 C). Notably, ATP-dependent helicase activity of ΔGC further confirmed that ATP does bind to ND1.

To further elucidate the helicase activity of ND1 with the linker helical domain (ΔGC), AFM visualization was performed using domain V of 23S rRNA as the substrate. Although ensembles of double strands (height ≥1 nm) or higher-order structures (height ∼1.8 nm) were seen when only domain V was visualized (Figs. 4 C and S2 A), the AFM image manifested (Fig. S2, B and C) clear formation of single strands (height 0.4 nm) in the presence of full-length protein. Formation of Y and loop structures displayed (Fig. 4 D) a double-strand (∼0.8 nm) opening (0.4 nm) when the linker helical domain was present (ΔGC). In contrast, in the absence of linker helical domain (ND1ΔH), rRNA remained mostly in higher-order structures (Fig. 4 E). No trace of the strand opening was seen. Association of the truncated protein at the Y fork junction can be identified by the increase in height (∼1 nm). However, the helicase activity of the full protein on domain V rRNA was more prominent, suggesting that a tightly controlled crosstalk is likely present among the domains (Figs. 4 A and S2). This postulation is supported by the observation (Wu et al., 2010; Jain et al., 2013) that a salt bridge between ND1 and ND2 is crucial for regulating the function of these domains.

RNA helicases perform ubiquitous functional roles in regulating RNA biology in vivo (Jankowsky, 2011). We aimed to understand exactly what role HflX plays as an RNA helicase in a cellular context. Recent studies have implicated HflX’s function during heat stress (Zhang et al., 2015; Coatham et al., 2016). Data presented in one of the previous studies (Zhang et al., 2015) have shown that although heat stress causes damage to normal E. coli cells to some extent, it is more detrimental to cells from HflX-deleted strains. To check whether HflX’s RNA helicase activity has any additional effect under heat shock, we first sought to understand the role of HflX in cell survival during heat stress. For this assay, we took untransformed BL21 cells (as a control) and three types of plasmid-transformed BL21 cells (full-length HflX, ΔGC, and ND1ΔH in pET28∞). We observed that when the cells were subjected to heat shock (50°C for 40 min) and then plated on Luria-Bertani (LB) growth medium, the growth of untransformed BL21 was severely impaired (Fig. 5 A, row b), whereas IPTG-induced transformed cells with full length HflX survived well (Fig. 5 A, row c). More interestingly, after heat shock, although growth of ND1ΔH transformed cells was impeded (Fig. 5 A, row e), ΔGC transformed cells showed better survival potency compared with untransformed BL21 cells, although the survivability was not as good as it was for the transformed cells containing full-length protein (Fig. 5 A, row d). The effect seen for ΔGC-transformed cells suggests that the ND1, along with the linker helical domain, must have an instrumental role in rescuing heat-damaged cells.

Next, we performed in vitro translation assay (using a T7 in vitro transcription translation kit). Although addition of normal 70S ribosomes to a kit master mix overshot normal production of luciferase (measured as luminescence) by the ribosomes available in a kit master mix, when heat-shocked ribosomes (incubated at 50°C for 40 min) were added to the assay reaction set, protein availability was drastically reduced compared with normal reaction set (Fig. 5 B). Interestingly, when heat-shocked ribosomes were incubated along with HflX before addition to a kit master mix, the production of luciferase increased significantly (Fig. 5 B). The reduction of luciferase production after the addition of heat-shocked ribosomes indicates that the heat-shocked ribosomes are likely nonfunctional and that random distribution of mRNA to functional and nonfunctional ribosomes presumably reduces the efficiency of protein synthesis. However, recovery of enzyme production upon addition of heat-shocked ribosomes along with HflX suggests that HflX must have rescued heat-damaged 70S ribosomes into functional forms. This experiment was repeated using 30S and 50S subunits instead of 70S ribosomes (Fig. 5 C). Interestingly, supply of HflX along with normal 30S and heat-shocked 50S subunits resulted higher luciferase production than kit control, presumably because in this case, splitting of impaired 70S ribosomes by HflX was not required. Addition of normal 30S and heat-shocked 50S subunits produced slightly better luciferase activity. The presence of additional normal 30S subunits most likely increases the probability of normal 70S ribosome availability. It should be noted here that equal amounts of ATP and GTP were added in all reaction sets.

We observed that like many putative RNA helicases, the RNA-unwinding activity of HflX is ATP dependent (Fig. 1 B). Thus, it is logical to assume that deformed ribosome, being the substrate of the RNA helicase, would be more effective in stimulating ATP hydrolysis on HflX. In an effort to verify this presumption, we used a high-salt wash 50S subunit as a mimic of aberrant form of the 50S subunit. High-salt wash (1.5 M) treatment causes detachment of some large subunit proteins (Fig. S1 D). Ribosomal proteins are involved in stabilizing the rRNA structure (Homann and Nierhaus, 1971). It has been shown that high-salt wash results in an inactive (loss of peptidyl transferase activity) 50S subunit (Moore et al., 2008; Pulk et al., 2010). Also, isolated large-subunit rRNA as well as domain V of 23S rRNA may be considered to have improper folding.

We have performed free phosphate determination using malachite green assay (Lanzetta et al., 1979; Baykov et al., 1988) in the presence of domain V, 50S-rRNA, and high-salt wash 50S subunit. After stopping the enzyme reaction, the P_i release upon ATP hydrolysis was estimated by spectrophotometric measurement. The assay clearly showed that the rRNAs effectively stimulated ATPase activity of HflX (Fig. 5 D). However, the high-salt wash 50S subunits showed higher activity than the rRNAs as well as normal-salt wash (1 M) 50S subunits (Fig. 5 E), indicating that the aberrant conformation of domain V, when part of the 50S subunit, is likely the actual substrate. A similar phosphate release assay revealed that the heat-shocked 50S subunits catalyzed ATP hydrolysis better than normal 50S subunits (Fig. 5 F), confirming that heat treatment indeed damages the 50S subunit, which acts as better substrate for ATPase activity on HflX.

HflX has recently been described as a heat shock protein (Zhang et al., 2015). It was proposed that HflX splits 70S ribosomes and stabilizes 50S subunits by binding as an antiassociation factor during heat stress. Nevertheless, our biochemical and cell-based assays clearly show that in addition to the previously proposed function, RNA helicase activity of HflX is important in rescuing heat-damaged 50S subunit where the linker helical domain of HflX has a profound role in allosterically modulating rRNA structures. In connection to our observation, it may be noted that DNA binds to a helical domain of the Bloom syndrome helicase (Newman et al., 2015), with which structural similarity of HflX was identified by a DALI search. A recent study also suggests a prong-like helical domain in Sen1 (an RNA/DNA helicase) is a critical determinant for its duplex-unwinding activity (Leonaitė et al., 2017).

Most of the bacterial ribosome–related RNA helicases identified so far seem to work under cold stress (Owttrim, 2013; Jarmoskaite and Russell, 2014), when RNA may have a tendency to get kinetically trapped in nonnative conformations (local minima). Under heat stress, however, RNAs, fluctuating between different conformations may encounter difficulty specifying a thermodynamically favored single folded structure over other conformations and thus may need the assistance of RNA helicases (Jankowsky, 2011; Owttrim, 2013). Resolution limitation of the cryo-EM map did not allow us to pinpoint the exact rRNA location where the modulation by HflX occurs. However, in the ATP–GTP complex, the linker helical domain of HflX reaches out to a location of domain V of 23S rRNA harboring the peptidyl transferase center (PTC), and radical movement of this domain was identified when two structures were compared. It is tempting to propose that PTC rRNA is a potential substrate for the RNA helicase action of HflX. It has been known for decades that maturation of PTC is one of the last events to occur during large-subunit assembly (Nierhaus, 1982; Fuller-Pace et al., 1993; Jomaa et al., 2014; Davis et al., 2016). It is possible that this region is the most vulnerable to heat stress and thus requires the assistance of an RNA helicase to recover its active conformation for biological functions.

Based on our results, a model is proposed (Fig. 6) to explain how HflX may function as a modulator of aberrant rRNA conformation after heat shock. ND1, in ATP-bound form, is supposed to be in an active high-energy state (“on” state) and seems to act as a compressed spring when the linker helical domain remains fastened to the rRNA (as seen in the ATP–GTP complex). After ATP hydrolysis, ND1 relaxes and the thrust of the spring opening forces the linker helical domain to move ∼20 Å (to a position seen in the GTP complex). Concomitantly, the rRNA is stretched to unwind and instantly released to rearrange.

50S subunits were isolated from E. coli MRE600 cells. First, 70S ribosomes were purified from log phase (OD₆₀₀ 0.65) growth culture following standard protocols (Traub and Nomura, 1968; Ghosh and Moore, 1979; Das et al., 1996). The 70S ribosomes were split into 50S and 30S subunits in low-Mg²⁺ TMA buffer (20 mM Tris-HCl, pH 7.5, 30 mM NH₄Cl, 0.5 mM Mg(OAc)₂, and 5 mM 2-mercaptoethanol). 50S and 30S subunits mixture was centrifuged for 10 h at 140,000 relative centrifugal force (rcf) through 10–40% sucrose gradient. 50S subunits containing fractions were pooled together and sucrose was removed by buffer (20 mM Tris-Cl, pH 7.5, 30 mM NH₄Cl, 10 mM Mg(OAc)₂, and 5 mM 2-mercaptoethanol) exchange through a 100-kD cutoff Amicon-Ultra filtration unit.

rRNA (23S and 5S rRNA) of large subunits was isolated from 50S ribosomal subunits using TRIzol and a Thermo Fisher Scientific RNA isolation kit. After TRIzol-dependent RNA isolation, the RNA pellet was washed two times with ice-cold 75% ethanol at 12,000 rcf for 5 min at 4°C. After washing, the pellet was dried carefully and dissolved in nuclease-free water (Amresco). The rRNA pool was then purified again using the Thermo Fisher Scientific RNA isolation kit protocol (the 260/280 ratio was 1.99).

The domain V region of 23S rRNA sequence containing DNA (a gift from C. Barat, St. Xavier’s College [Autonomous], Kolkata, India) was cloned in pTZ57R/T after the T7 promoter site. To produce domain V rRNA, we used the Thermo Fisher Scientific in vitro transcription system. First, the construct was linearized using EcoR1 endonuclease. The in vitro transcription process was performed using T7 RNA polymerase (Thermo Fisher Scientific) and ribonucleotides (Thermo Fisher Scientific). 1 U/20 µl DNase was added to the mixture and incubated for 15 min at room temperature. Half of the volume of reaction mixture of 7 M ammonium acetate and 2.5 times the total volume of ice-cold ethanol were then added and incubated overnight at −80°C. The sample was centrifuged at 12,000 rcf for 30 min, and the supernatant was discarded. The RNA pellet was washed two times with ice-cold 75% ethanol at 12,000 rcf for 15 min at 4°C. After washing, the pellet was dried and nuclease free water (Amresco) was added to dissolve RNA (the 260/280 ratio was 2).

The E. coli HflX construct (pET28∞ plasmid with HflX gene inserted; a gift from B. Prakash, Indian Institute of Technology, Kanpur, India) was transformed to E. coli BL21 cells (Jain et al., 2009, 2013). Cell culture of log phase (OD₆₀₀ = 0.6) was induced with 0.6 mM IPTG (Sigma-Aldrich) for 10 h at 26°C. Cells were harvested and lysed in lysis buffer (20 mM Tris-HCl, pH 7.6, 500 mM NaCl, 5% glycerol, 2 mM 2-mercaptoethanol, 1 mM PMSF, and 5 mM imidazole) by ultrasonication. After centrifugation of cell lysates at 12,000 rcf for 30 min at 4°C, the supernatant was loaded onto a column containing Ni-Sephadex beads (GE Healthcare), washed with wash buffer (20 mM Tris-HCl, pH 7.6, 500 mM NaCl, 5% glycerol, 1 mM PMSF, 2 mM 2-mercaptoethanol, and 40 mM imidazole) and eluted with elution buffer (20 mM Tris-HCl, pH 7.6, 500 mM NaCl, 5% glycerol, 1 mM PMSF, 2 mM 2-mercaptoethanol, and 250 mM imidazole). HflX-containing fractions were pooled together, buffer exchanged with storage buffer (20 mM Tris-HCl, pH 7.6, 500 mM NaCl, 5% glycerol, and 5 mM 2-mercaptoethanol), and concentrated using an Amicon ultra 30 kD filtration unit. HflX proteins (confirmed by standard mass spectrometry analysis [MALDI-TOF-TOF-4800; Applied Biosystems]) were divided into aliquots and stored at −80°C.

Gene sequences of two truncated protein forms (ND1 with the linker helical domain, G and C terminal domains deletion of HflX [ΔGC], and ND1 only, without the linker helical domain [ND1ΔH]) were inserted into the pET28∞ plasmid and transformed into BL21 cells (done separately). The purification procedure for these two truncated proteins (ΔGC and ND1ΔH) was the same as that used for HflX protein, except that the pH 6.9 buffer was used for ΔGC (a list of primers is given in Table S2). The SDS-PAGE gel showed the quality of purified proteins (Fig. S1 E).

We prepared two complexes: (1) 50S–HflX–AMP–PNP–GMP–PNP and other was 50S–HflX–AMP–PNP. To make these complexes, proteins and 50S subunits were incubated at a ratio of 23:1. First, proteins were incubated with either 5 mM AMP-PNP (Sigma-Aldrich) only or both 5 mM AMP–PNP (Sigma-Aldrich) and 4 mM GMP–PNP for 20 min at 4°C (in respect to reaction volume). 50S subunits were mixed with analogue-bound proteins in binding buffer (20 mM Tris-Cl, pH 7.6, 50 mM NH₄Cl, 10 mM Mg(OAc)₂, and 1 mM DTT) for 45 min at 4°C and then loaded on the sucrose gradient (10% to 40%). Ultracentrifugation was done at 140,000 rcf for 6 h at 4°C, and the gradient was then analyzed at 260 nm. Subunits containing fractions were pooled together, and sucrose was removed by binding buffer exchanged through a 100-kD cutoff Amicon-Ultra filtration unit. The sample was then separated on SDS-PAGE gel to confirm HflX binding.

The RNA-unwinding activity of HflX, ΔGC, and ND1ΔH was performed following published protocols (Henn et al., 2001; Chandran et al., 2007) with minor modifications. A 24-nt oligoribonucleotide (Trilink) with the sequence 5′-GAAUGUACAUCAGAGUGCGCACUC-3′ was used for the helicase assay in which the underlined part of the sequence is self-complementary. The oligonucleotide was phosphorylated with T4 polynucleotide kinase (Fermentas) and gamma-[³²P]-ATP and separated from unincorporated isotopes on a Sephadex G-25 column (Roche). Labeled oligonucleotides were further purified using the TRIzol method and precipitated with an equal volume of isopropyl alcohol at −20°C overnight, and the pellet was resuspended in 50 µl hybridization buffer (20 mM Tris–HCl, pH 7.5, and 500 mM NaCl). Before the gel separation assay, oligonucleotide solution was incubated at 95°C for 4 min and then slowly cooled to room temperature. RNA-unwinding activity was performed in vitro by measuring the conversion of self-duplex RNA to single-stranded RNA. For this assay, purified protein was incubated with 6 fmol (0.1 nM) of duplex RNA in a 50-µl reaction volume containing 20 mM Tris–HCl, pH 7.5, 5 mM MgCl₂, 60 mM KCl, 5% (vol/vol) glycerol, 100 ng yeast tRNA, 50 U RNasin (Promega), and 2 mM nucleotide triphosphate (ATP or GTP). Full-length HflX along with ΔGC and ND1ΔH truncated proteins were used in this assay. In the case of ΔGC, pH 6.9 was used for the reaction buffer and 5 mM and 10 mM ATP were used to examine ATP dependency. First, RNA and protein were kept at 4°C for 20 min. The reaction mixture was then incubated at 37°C for 30 min, after which the reaction was stopped by adding 50 µl 2× stop buffer (20% glycerol, 0.2% SDS, 4 mM EDTA, 0.025% bromophenol blue, and 0.025% xylene cyanol). 20 µl reaction mixture was analyzed directly by electrophoresis on a native 12% polyacrylamide gel (29:1) in 0.5× Tris-borate-EDTA buffer. The gel was dried and visualized using a Phosphorimager. GelQuantNET software (Biochem Lab Solutions) was used to evaluate the band intensities.

For this assay, we used untransformed BL21 cells as WT cells and three types of plasmid (HflX full protein, ΔGC, and ΔH in Pet28∞)–transformed BL21 cells. Overnight-grown cells were added to new LB growth media to reach OD₆₀₀ 0.6, and then an equal amount of cells were taken for induction (0.1 mM IPTG was used for induction at 27°C for 1 h), pelleted, and washed twice. The same number of cells was used to induce heat shock. Heat shock was done at 50°C for 40 min, after which cells were plated on LB with serial dilution and kept at 30°C for 18 h.

We used the T7 in vitro transcription–translation kit to assay the role of HflX on heat-stressed ribosomes. A luciferase template (supplied by the kit) was used in every reaction set. Luciferase production was measured by activity of luciferase in a luminometer. To give heat shock to ribosomes, ribosomes were incubated at 50°C for 40 min and then immediately transferred onto ice. To repair the heat-stressed ribosomes by using HflX, 2 nM heat-stressed ribosomes (70S or 50S subunits), 0.6 mM ATP, and 0.6 mM GTP were incubated with 50 nM HflX at 4°C for 20 min and then transferred at 37°C for 30 min. After that, the reaction mixture was added to the kit master mix and incubated at 37°C for 90 min. 0.6 mM ATP and 0.6 mM GTP were maintained in all reaction sets including the control set. For heat-treated subunits (50S), 2 nM normal 30S subunits was added to the master mix. To estimate the presence of active luciferase enzyme, luciferase activity was measured in a luminometer; 50 µl luciferin (Promega) and 2.5 µl reaction mixture were mixed, and luminescence was measured using a Glomax 20/20 luminometer (the integration time was 10 s).

ATPase activity was determined using a malachite green assay (Lanzetta et al., 1979; Baykov et al., 1988; Wu et al., 2009; with slight modification). The ATP hydrolysis reaction was performed by incubating 20 µM HflX with 1 mM ATP, rRNA of 50S subunits or domain V rRNA or high-salt (1.5 M)–washed 50S subunit (100 nM for each) or heat-treated 1 M salt-washed 50S subunit (1 µM) in a buffer containing 50 mM Tris–HCl, pH 8.0, 200 mM NaCl, 1 mM 2-mercaptoethanol, and 5 mM MgCl₂ at 37°C for 60 min in 25 µl reaction volume. After 1 h, 125 µl of 1× reaction buffer was added to increase the reaction volume to 150 µl. A filtration step was performed to remove protein and RNA from reaction mixture using 30-kD cutoff filtration units (Sartorius). To determine ATP hydrolysis, a malachite green–ammonium molybdate mixture was used. Concentrated sulfuric acid (d = 1.84 g/l) was added to 300 ml double-distilled water, and 0.44 g malachite green was then mixed in (Baykov et al., 1988). After that, 2.5 ml of 7.5% ammonium molybdate and 0.2 ml of 11% Tween-20 were added to 10 ml malachite green solution and kept at room temperature for 60 min (orange). This mixture was then filtered through a 0.22-µm syringe filter just before use. The ATP hydrolysis reaction was stopped by adding the malachite green–ammonium molybdate mixture in a 4:1 ratio (sample, malachite green–ammonium molybdate mixture) and incubated for 5 min at room temperature (green). Finally, 25 µl of 34% citric acid solution (for 100 µl reaction mixture) was added (Lanzetta et al., 1979) and incubated for 2 min at room temperature, and absorbance was measured at 630 nm.

We did Western blot experiments to compare the binding efficiency of HflX protein with the 50S subunit in the presence of AMP–PNP only, both AMP–PNP and GMP–PNP, and without any nonhydrolyzing analogue. 50S subunits and a two-times molar volume of HflX protein were incubated in binding buffer (20 mM Tris-HCl, pH 7.6, 50 mM NH₄Cl, 10 mM Mg(OAc)₂, and 1 mM DTT) at 4°C for 45 min, and the reaction mixture was separated through a 10% to 40% sucrose gradient by using ultracentrifugation at 140,000 rcf for 5 h The gradient was then fractionated, absorbance was measured at 260 nm, and the 50S subunit–containing fraction pooled and filtered using 100-kD filtration units to remove sucrose. Sample concentration was measured at 260 nm, and an equal amount of sample was separated on SDS-PAGE and processed for Western blotting.

We used Western blot experiments to quantify the total HflX protein concentration including cloned HflX protein and native HflX protein in BL21 cells containing HflX construct, compared with BL21 cells without HflX construct after 50°C for 40 min of heat shock. In these experiments, OD₂₆₀ 0.2 1 ml cell culture was used for heat shock and SDS-PAGE gel separation. Cells lysis and SDS-PAGE sample preparation were done according to a published protocol (Zhang et al., 2015).

Western blot analysis was performed to quantify the bound HflX protein concentration. Blots were probed with anti–His antibody or anti–HflX antibody (Biobharati Lifescience). Binding of the secondary HRP-conjugated anti–rabbit antibodies (EMD Millipore) was analyzed using ImmunoCruz (Santa Cruz Biotechnology, Inc.). Intensities were measured using gel-quant-net software.

Freshly cleaved mica was incubated with 3-aminopropyl triethoxysilane (Sigma-Aldrich) before applying samples (Lyubchenko and Shlyakhtenko, 1997). In AFM experiments, rRNA of the 50S subunits and domain V RNA were incubated in reaction buffer A (20 mM Hepes, pH 7.5, 50 mM NH₄Cl, and 5 mM MgCl₂), heated at 95°C for 4 min, and then slowly cooled to room temperature. Either free RNA or protein–RNA solution was deposited on freshly treated mica for 5 min at room temperature, rinsed with water, and dried. HflX and domain V rRNA (1:2 ratio) were incubated in reaction buffer A containing 0.5 mM DTT for 30 min at 37°C. The final concentration of ATP was 10 mM.

We primarily followed the procedure described previously (Henn et al., 2001). AFM experiments were performed in intermittent contact mode (called “tapping” or AAC mode) to minimize tip-induced damage. AAC mode AFM was performed using a Pico plus 5500 inverted light microscope AFM (Agilent Technologies) with a Piezo scanner (maximum range 9 µm). Microfabricated silicon cantilevers 225 µm in length with a nominal spring force constant of 21–98 N/m were used (Nano Sensors). Cantilever oscillation frequency was tuned into resonance frequency. The cantilever resonance frequency was 150–300 kHz. The images (512 × 512 pixels) were captured with a scan size between 0.5 and 0.8 µm at a scan speed rate of 0.5 lines/s. Images were processed by flatten using Pico view1.1 version software (Agilent Technologies). Image manipulation was done using Pico Image Advanced version software (Agilent Technologies).

Grids for cryo-EM were prepared following standard procedures using Vitrobot Mark IV (FEI; Grassucci et al., 2007). Micrographs were recorded on a Philips FEI POLARA field emission gun electron microscope operated at 300 kV equipped with a low-dose kit and an Oxford cryo-transfer holder at a total magnification of 79,894×. Data were collected using 4,000 × 4,000 Eagle charge-coupled device Camera (FEI) with a physical pixel size of 15 µm (corresponding with a pixel size of 1.89 Å on the object scale) and defocus values ranging from 1.0–4.8 µm. All images were acquired using low-dose procedures with an estimated dose of ∼20 electrons/Ų. SPIDER (Frank et al., 1996; Shaikh et al., 2008) and EMAN2 (Tang et al., 2007) were used for image processing.

Good micrographs were selected after inspection of drift, astigmatism, and the presence of Thon rings in the power spectrum. Particles were picked by SPIDER auto picking (Rath and Frank, 2004) using an empty 50S cryo-EM density map as reference model (filtered to very low resolution), and from auto-picked particles, good particles were selected manually. Image processing using SPIDER included a 3D projection alignment procedure with correction of the contrast transfer function. Using small number of particles (∼14,000), a cryo-EM 3D map was generated, which was subsequently used as the initial model in EMAN2 3D reconstruction.

Particles were stacked and imported to EMAN2 for 3D reconstruction. Three cryo-EM density maps were generated (the control 50S subunit, the 50S–HflX–AMP–PNP–GMP–PNP complex [ATP–GTP complex], and the 50S–HflX–AMP–PNP [ATP complex]) using 57,759, 61,557, and 32,159 final sets of particles, respectively. The gold-standard resolutions of the resultant maps were 11.7 Å, 8.1 Å, and 14.8 Å, respectively. To generate a quasi-atomic model of the complex, available atomic structure was fitted into EM map. The structure of 50S-bound HflX taken from PDB accession number 5ADY (removed GTP) was fitted into the attributable cryo-EM density isolated from the 3D map of the 50S–HflX–AMP–PNP–GMP–PNP complex, followed by molecular dynamics flexible fitting (Trabuco et al., 2008). Cross-correlation coefficients were measured for the fitted domains using VMD and Chimera (Pettersen et al., 2004). Chimera and PyMol (Delano Scientific) were used for structural analysis and to prepare illustrations.

For sample preparation to visualize helicase activity in cryo–transmission EM, large-subunit rRNA was incubated with HflX protein (2:1) in reaction buffer (20 mM Hepes, pH 7.5, 50 mM NH₄Cl, 5 mM MgCl₂, 10 mM ATP, and 0.5 mM DTT) at 4°C for 20 min and then kept at 37°C for 30 min.

The cryo-EM density map of the 50S–HflX–ATP–GTP complex has been deposited in the Electron Microscopy Data Bank (EMD-6979). The fitted atomic model has been deposited in the PDB (5ZZM).

Fig. S1 shows purification of different constructs of HflX and their RNA-unwinding activity. Fig. S2 shows AFM visualizations of nucleotide-dependent unwinding of domain V of 23S rRNA by HflX. Fig. S3 presents cryo-EM maps of control and ATP complexes and resolutions of the maps. Fig. S4 shows close-up views of different domains of HflX fitted into the cryo-EM map of the ATP–GTP complex. Table S1 presents results of the DALI search. Table S2 lists the primer sequences.

Technical assistance provided by Mr. T. Muruganandan and Mr. Sandip Chakraborty during the AFM experiments and mass spectrometry analysis is gratefully acknowledged. We are grateful to Dr. Ilic Zoran (Wadsworth Center, Albany, NY) for critically checking the manuscript. We sincerely thank Mr. S. Krishnamoorthi, Ms. Priya Das, Mr. Saurav Verma, and Ms. T. Anusha for assisting with several experiments, protein purification, and preliminary cryo-EM data processing. We thank Dr. Balaji Prakash and Dr. Chandana Barat for the gifts of E. coli HflX protein and domain V constructs.

This work was supported by the Council of Scientific and Industrial Research (India) Network project UNSEEN (BSC0113), the Science and Engineering Research Board (Department of Science and Technology, India)–sponsored project GAP327, and Council of Scientific and Industrial Research – Indian Institute of Chemical Biology (Kolkata, India). Funding provided by the National Institute of Pharmaceutical Education and Research, India, for project students is gratefully acknowledged. S. Dey acknowledges University Grants Commission, India, for Senior Research Fellowship.

The authors declare no competing financial interests.

Author contributions: J. Sengupta conceived the idea. S. Dey and J. Sengupta designed the experiments. S. Dey carried out bulk of the biochemical and AFM experiments and cryo-EM 3D image processing. S. Dey and J. Sengupta analyzed the data. C. Biswas performed cryo grid preparation, cryo-EM data collection, and preliminary analysis of the dataset. J. Sengupta wrote the paper with help from S. Dey.