ZNF598 Is a Quality Control Sensor of Collided Ribosomes

Abstract

Aberrantly slow translation elicits quality control pathways initiated by the ubiquitin ligase ZNF598. How ZNF598 discriminates physiologic from pathologic translation complexes and ubiquitinates stalled ribosomes selectively is unclear. Here, we find that the minimal unit engaged by ZNF598 is the collided di-ribosome, a molecular species that arises when a trailing ribosome encounters a slower leading ribosome. The collided di-ribosome structure reveals an extensive 40S-40S interface in which the ubiquitination targets of ZNF598 reside. The paucity of 60S interactions allows for different ribosome rotation states, explaining why ZNF598 recognition is indifferent to how the leading ribosome has stalled. The use of ribosome collisions as a proxy for stalling allows the degree of tolerable slowdown to be tuned by the initiation rate on that mRNA; hence, the threshold for triggering quality control is substrate specific. These findings illustrate how higher-order ribosome architecture can be exploited by cellular factors to monitor translation status.

Graphical abstract

Figure 1

ZNF598 Engages a Sub-population of Poly(A)-Stalled Ribosomes In Vitro (A) Strategy to analyze poly(A)-stalled ribosome-nascent chain complexes produced by in vitro translation (IVT). VHP is a small autonomously folding three-helix bundle from the villin head piece. (B) Poly(A)-stalled translation complexes were affinity purified via the nascent chain and separated by sucrose gradient sedimentation. The affinity-purified products (input) and each gradient fraction were analyzed by autoradiography (to detect nascent chains) or immunoblotting for recombinant ZNF598 or ribosomal proteins uL2 and eS24. Mono- and di-ribosome fractions are indicated. (C) Poly(A)-stalled translation complexes from reactions lacking or containing 5 nM FLAG-tagged ZNF598 were immunopurified via the FLAG tag, and the nascent chains were detected by autoradiography. 1° and 2° indicate the position of nascent chains from the first and second ribosome of the stalled complexes (see diagram). (D) The input and affinity-purified samples prepared as in (C) were separated by sucrose gradient and the nascent chains detected by autoradiography. The graph below the autoradiograms depicts the distribution of nascent chains corresponding to mono-ribosomes (1°, black) or di-ribosomes (2°, red) in the input sample or the nascent chains recovered by affinity purification (”IP ZNF598” - blue). See also Figure S1.

Figure 2

Site-Specific Stalling of Native Polysomes Triggers ZNF598 Engagement (A) Strategy to site-specifically stall ribosomes at the stop codon on native polysomes translating globin mRNA using the mutant release factor eRF1^AAQ. (B) The polysomes in native reticulocyte lysate (RRL) were allowed to elongate in a translation reaction supplemented with 1 μM eRF1^AAQ, 5 nM FLAG-tagged ZNF598, and 10 μM His-tagged ubiquitin. After fractionation on a sucrose gradient, translated globin polypeptides were detected by autoradiography and the other products by immunoblotting. Ubiquitinated eS10 was detected with anti-eS10 after pull-down via the His-tag on ubiquitin. FL indicates full-length nascent chains; “trunc. NCs” are truncated nascent chains. (C) Native RRL polysomes elongated in the absence or presence of 1 μM eRF1^AAQ were incubated without or with micrococcal nuclease, ubiquitinated with ZNF598, and analyzed for eS10 by immunoblotting. (D) Native RRL polysomes stalled with eRF1^AAQ were fractionated by a high resolution sucrose gradient (see Figure S2B). Fractions enriched in mono-, di-, tri-, tetra-, and penta-ribosomes were normalized to contain an equal number of ribosomes, ubiquitinated with 10 nM ZNF598, and analyzed by immunoblotting for eS10. The stained blot verifies equal input ribosomes in each reaction. (E) Reactions corresponding to the third and fourth lanes of (C) were fractionated on a sucrose gradient and analyzed by Coomassie staining or immunoblotting. The positions of mono- and poly-ribosomes are indicated. The primary bands seen in the stained gel are ribosomal proteins. See also Figure S2.

Figure 3

Structure and Interfaces of the Collided Di-ribosome (A) Overview of the collided di-ribosome structure. The two primary interfaces between the 40S subunits of the stalled and collided ribosomes are indicated. Red and cyan balls indicate the approximate sites of ubiquitination on eS10 and uS10, respectively. The ubiquitinated residues are not modeled, so the most proximal visible residues are indicated. (B) Close-up view of interface 1. Proteins eS1, uS11, eS26, and eS28 are from the stalled ribosome and make contact with uS4 and h16 of the 18S rRNA of the collided ribosome. The mRNA in the stalled ribosome channel is shown. (C) Close-up of interface 2 showing the approximate sites of eS10 and uS10 ubiquitination as in (A). RACK1 is from the stalled ribosome, while uS3, eS10, and uS10 are from the collided ribosome. See also Figures S3 and S4 and Table S1.

Figure 4

Tolerance and Flexibility of the Collided Di-ribosome (A) Maps of canonical-state (pale cyan/yellow) and rotated-state (blue/orange) ribosomes are superimposed in the “stalled” position of the collided di-ribosome structure using the 40S subunits for alignment. The 40S interfaces are relatively invariant to rotation state and the 60S subunits do not clash with the collided ribosome (gray). (B) Multi-body refinement and principal component analysis (Figures S4E and S4F) were used to quantify the nature and extent of heterogeneity in the relative positions of the stalled and collided ribosomes in di-ribosome particles. Maximal variance of the 40S subunit of the collided ribosome relative to a fixed stalled ribosome (dark gray) is displayed as a heatmap in sausage (left) or surface (right) representation. The interfaces and areas around the mRNA channel show the least variance. See also Figures S4 and S5.

Figure 5

ZNF598 Detects Collided Ribosomes Induced by Multiple Types of Stalls In Vivo (A) Strategy to induce complete versus stochastic ribosome stalling in cells treated with translation elongation inhibitors. (B) Wild-type (WT) or ZNF598-knockout (ΔZNF598) HEK293 cells were pre-treated for 15 min with low (1.8 μM) or high (360 μM) emetine, lysed, digested with micrococcal nuclease, and separated by sucrose gradient centrifugation. The 260 nm absorbance profiles across the gradient are normalized to the 80S mono-ribosome peak. (C) HEK293 cells pre-treated for 15 min with low or high concentrations of the indicated elongation inhibitors were analyzed by immunoblotting. Ubiquitinated eS10 (Ub-eS10) was detected using more sensitive reagents than unmodified eS10. For reference, ∼10% of total eS10 is ubiquitinated in low-dose inhibitor-treated samples and ∼1%–2% in untreated samples. (D) Cytosol from HEK293 cells pre-treated for 15 min with nothing (top), high (middle), or low (bottom) concentrations of emetine were separated by sucrose gradient fractionation and immunoblotted for ZNF598 and uL2. See also Figure S6.

Figure 6

Detection of Ribosome Stalling Is Context Dependent (A) Reporter construct and expected protein products in the absence or presence of terminal stalling. FLAG-SR-X is a “stalling reporter” in which X represents either a stalling sequence (poly(A), coding for 21 Lysine codons, termed (K^AAA)₂₁) or no insert (termed (K)₀). The RFP:GFP ratio is a quantitative measure of terminal stalling. (B) Histograms of the RFP:GFP ratios quantified by flow cytometry of WT or ΔZNF598 HEK293 cells expressing the (K^AAA)₂₁ containing reporter. The cells were induced to express the reporter for 22 hr in the presence (red traces) or absence (gray traces) of 10 nM pactamycin, an inhibitor of translation initiation. (C) HEK293 cells expressing the (K^AAA)₂₁ stalling reporter (left) or control (K)₀ construct (right) were cultured for 22 hr in the absence (shaded gray trace) or presence of varying concentrations of pactamycin (colored traces). Staggered histograms of the RFP:GFP ratio measured by flow cytometry are plotted. See also Figure S7.

Figure 7

Model for Recognition of Aberrant Translation by ZNF598 (A) ZNF598 does not engage ribosomes that are translating normally or slow down at a difficult-to-translate sequence (regional slowdown). If the relative velocities of the lead and trailing ribosomes permit the trailing ribosome to close the inter-ribosomal distance (IRD) before the slowed ribosome moves 10 codons away, a collision will occur to allow recognition by ZNF598. (B) Heatmap of the collision probability for a given slowdown (x axis) as a function of the distance to the trailing ribosome (y axis). An average velocity for the trailing ribosome of 5.6 ± 2.5 codons per second was used. At the genome-wide average IRD of 66 codons (∼200 nt), appreciable collisions will not occur unless the lead ribosome slows to ∼1 codon per second (less than one-fifth the normal rate) over a 10-codon stretch. However, for very frequently initiated mRNAs that have a very short IRD (such as globin mRNA), a mere 2-fold slowdown is sufficient to begin observing collisions.