Pacific Bioscience Sequel technology uses single‐molecule sequencing of long amplicons. Figure 1A outlines library preparation of circular templates. Each circularized DNA representing a single mRNA molecule with its full length PAT is replicated multiple times, both the sense strand and its complement (Figure 1B). Multiple ‘subreads’ generate one circular consensus sequence (CCS) read containing a unique gene-derived mRNA fragment followed by a nucleotide length-resolved PAT specific to the single mRNA molecule it represents (Figure 1B).

Figure 1 with 1 supplement see all

Download asset Open asset

Small circles produce more copies per unit time than big circles, which generates higher CCS read accuracy. Thus, limiting amplicon length with random hexamers (Figure 1A) rather than directing 5’-extended cDNA produces high accuracy reads with full length PATs and ample gene-specific mapping information.

An objective comparison of SM-PAT-seq and RNA-seq was performed. The same RNA sample was analyzed for transcriptome-wide read counts by both methods and revealed an R² correlation = 0.58 with p value of 2.2e-16 (Figure 1—figure supplement 1A). Assuming that RNA-seq is itself imperfect for quantitation and negatively impacted the R² value to some degree, the data are evidence that SM-PAT-Seq can indeed also serve to reflect relative mRNA abundances. Triplicate quantifications of GFP mRNA by SM-PAT-Seq and northern blotting produced an R² correlation = 0.96 with a p=0.000035 (next section). Finally, we note that >99% of SM-PAT-seq CCS reads reflect single molecule mRNAs as <1% CCS read identity was detected per library (Figure 1—figure supplement 1B).

SM-PAT-Seq was performed on HEK293 cells expressing LARP4 at three levels, each in triplicate. Cells received empty plasmid or plasmid expressing LARP4 at ~3X or ~11X higher than endogenous LARP4 levels (1X) (Figure 2—figure supplement 1), and each also received a uniform amount of GFP expression plasmid. Figure 2A shows effects on PAT lengths for ~13,500 nuclear-encoded mRNAs (includes poly(A)⁺ transcripts annotated as long noncoding RNAs) (Supplementary file 1). Each SM-PAT-Seq CCS read represents an mRNA molecule with a specific PAT length. LARP4 affected a large number of mRNAs distributed across a wide range of PAT lengths (Figure 2A). LARP4 expression led to shifts in mRNA PAT length distributions in three length ranges (Figure 2A), 30–60 nt, 60–90 nt, and 90- > 250 nt. At 11X, the fraction most increased was mRNAs with long PATs, 150–250 nt. The fraction shifted with 3X LARP4 was also to longer PATs, but less than with 11X. As Figure 2A shows fractions of total reads, an increase in one size range is associated with decrease in another. The intersection of the 1X and 11X tracings was at 110–90 nt.

Figure 2 with 1 supplement see all

Download asset Open asset

An interpretation of Figure 2A is that an increase in LARP4 levels leads more mRNAs to persist with long PATs at the expense of accumulation of mRNAs with short PATs. Increasing LARP4 alters the mRNA population such that more have longer PATs and less have shorter PATs. Given current models, increasing LARP4 may slow the conversion of long PAT-mRNAs to short PAT-mRNAs.

Median mRNA PAT lengths of all genes observed with a minimum of 10 CCS reads are shown as violin plots in Figure 2B. Thousands of mRNA PATs are shifted to longer lengths by increasing LARP4 levels. The median PAT length 50% cumulative read fraction was shifted from ~75 nt with 1X LARP4 to ~82 and~90 nt at 3X and 11X LARP4, respectively (Figure 2C).

The average PAT lengths for functionally related gene subsets as determined by gene ontology (GO) were almost all correlated with LARP4 expression levels, with the one clear exception being the mitochondrial DNA-encoded mRNAs, as expected (Figure 2D). However, an unexpected exception was the set of nuclear-encoded mitochondrion mRNAs. The PATs on the mRNAs that comprise this set are significantly shorter than on most other mRNAs in cells with 1X LARP4, and their lengths increased less in response to higher LARP4 levels relative to the PAT length increases in the other mRNA GO sets. This reflects a highly significant difference including by comparison to the RNA binding GO set as both are comprised of ~2000 mRNAs and represented by 35,000–50,000 CCS read counts in each LARP4 condition.

Prior analysis provided indirect evidence of PAT lengthening of GFP mRNA by LARP4 (Mattijssen et al., 2017). Here, all cells were transfected with equal amounts of GFP plasmid and triplicate samples were analyzed by northern blot and SM-PAT-Seq. This showed that GFP mRNA accumulation and PAT lengthening was dose-dependent on LARP4 (Figure 2E,F). As LARP4 levels increased, GFP mRNA median PAT lengths increased and were accompanied by increases in GFP CCS reads (Figure 2F), reflective of the trends of GFP mRNA levels and mobility shifts seen by northern blot (Figure 2E). Comparison of GFP CCS quantitative reads and GFP mRNA quantitation by northern revealed a R² value of 0.96 with p=0.000035 (Figure 2G). The general likeness of band intensity distribution in the northern blot 1X, 3X and 11X lanes, and shapes of the corresponding violin plots suggests that LARP4-mediated protection from deadenylation of long PATs specifically, confers the stable GFP mRNA with greater accumulation/stability (Discussion).

The left panel of Figure 3A shows SM-PAT-Seq analysis of ~10,500 nuclear-encoded mRNA CCS reads from triplicate MEF cell lines (Supplementary file 2). The KO MEFs have more mRNA CCS reads with short PATs (30–60 nt) than with longer PATs (75–200 nt) as compared to WT (Figure 3A, left panel). The cumulative fraction plots (right) show an overall length shift to shorter PATs in the KO MEFs. Violin plots (Figure 3B) show that the median PAT lengths of KO MEF mRNAs are shorter than in WT MEFs, for thousands of mRNAs. In contrast to nuclear DNA gene-encoded mRNAs, LARP4 expression had no apparent effect on mitochondrial DNA-encoded mRNA PATs (Figure 3—figure supplement 1).

Figure 3 with 1 supplement see all

Download asset Open asset

Occurrence of sequential peaks in the poly(A) length profiles of efficiently translated mRNAs such as those that encode ribosomal proteins is referred to as phasing (Lima et al., 2017). The phased PAT length peaks derive from fluctuations in the rates of deadenylation around sites protected by serial PABPs bound at those positions (Lima et al., 2017; Nicholson and Pasquinelli, 2019) (see Yi et al., 2018).

It was previously shown that three ribosomal protein (rp) mRNAs have shorter PATs and shorter half-lives in LARP4 KO MEF cell lines as compared to WT MEF cell lines (Mattijssen et al., 2017). Figure 3C shows SM-PAT-seq data from the rp-mRNAs as a subset of the MEF total mRNAs. The left panel shows that both profiles revealed a bimodal distribution in the ~35–70 nt size range, although the relative heights of the two peaks clearly differed in WT and KO. The longer PAT peak was higher in the KO cells and associated with a steeper transition incline from long to short PATs in the KO (Figure 3C). Interestingly, similar to Figures 2A and 3A, the transitions from long to short PATs also differed for KO and WT and overlapped in the ~110–90 nt size range f the rp-mRNAs (Discussion). We conclude that genetic deletion of LARP4 alters the PAT length phasing pattern of rp-mRNAs in MEFs in the ~70–35 nt size range, and this is associated with differences in profile shapes of longer PATs.

An ARE in LARP4 mRNA mediates decreased LARP4 levels in response to the cytokine tumor necrosis factor-α (TNFα) (Mattijssen and Maraia, 2016). The type I interferon (IFN)-stimulated genes (ISGs) are induced by a variety of molecules including some interleukins, TNFα, and cytokines that activate various extracellular receptors (Hergovits et al., 2017; Wang et al., 2016). ISGs are also induced by activation of intracellular receptors in response to viral infection and other pathogens; different types of nucleic acids activate distinct pathways to ISG induction (Atianand and Fitzgerald, 2013; Schneider et al., 2014). Triplicate RNA-seq revealed that a set of 128 ISG mRNAs were induced to ≥1.5 fold higher levels with a Padj-value <0.05 in HEK293 cells expressing 3X LARP4 relative to 1X after 1 kb AT-rich DNA transfection (Supplementary file 3). We cotransfected 293 cells with LARP4 at 3X levels together with the dsRNA poly(I:C), (dA:dT)₍₈₀₎, or 1 kb AT-rich DNA, and monitored expression of the ISG mRNAs, IFIT1 and ISG15 (Figure 4A, lanes 1–20). Triplicate data for the dsRNA poly(I:C) ISG induction are shown in Figure 4A, lanes 1–12 and quantified in Figure 4B. This verified that higher levels of ISG mRNAs accumulated with elevated LARP4 (+) expression (Figure 4B). Slower mobility was observed in the ISG15 mRNA from the cells transfected with LARP4 (+), indicative of longer PATs (Figure 4A, lanes 2 and 4, 6 and 8, 10 and 12).

Figure 4

Download asset Open asset

Transfection of LARP4 also led to higher mRNA levels after treatment with ISG-inducing DNAs, (dA:dT)₍₈₀₎, or 1 kb AT-rich DNA (Figure 4A lanes 13–16, 17–20). As noted, LARP4 levels can change in response to external stimuli, e.g., by TNFα (Mattijssen and Maraia, 2016). We tested if increasing LARP4 levels (3X) could sensitize HEK293 cells to ISG-inducing DNA (Figure 4C). For this, cells were transfected with empty plasmid (lanes 1–4) or equal amounts of LARP4 expression plasmid (lanes 5–6) and varying amounts of 1 kb AT-rich DNA, the latter as indicated by triangles above the lanes (Figure 4C). The cells expressing 3X higher levels of LARP4 produced more ISG mRNA at a lower dose of inducing DNA. Thus LARP4 expression can contribute to the levels to which ISG mRNAs accumulate after their induction (Figure 4C).

Ability to see increased ISG mRNA accumulation in response to LARP4 provides an assay for induced endogenous transcripts. Examination of different LARP4 constructs revealed that increased accumulation of IFIT1 and ISG15 mRNAs is dependent on the PABP-interaction motifs PAM2w and PBM, of LARP4, also required for GFP mRNA mobility shift indicative of PAT protection (Figure 4D; Mattijssen et al., 2017).

To examine PAT length dynamics, ISG mRNAs were induced in KO and WT MEFs by treatment with IFNα, followed by actinomycin D (ActD) to inhibit transcription, and RNAs were subjected to SM-PAT-seq at times thereafter to monitor deadenylation. Separate analyses were done on ISG mRNAs and on total mRNAs.

As part of the same duplicate experiments, northern blots of total RNA revealed a ~ 4 hr half-life for ISG15 mRNA in WT cells, consistent with prior data (Li et al., 2000), but 2.8 hr in KO cells, reflecting lower stability (Figure 5A,B). The ISG15 mRNA bands exhibited progressively faster mobility with time after ActD reflective of deadenylation (Figure 5A).

Figure 5 with 1 supplement see all

Download asset Open asset

PAT shortening after ActD treatment was apparent for the total mRNA set with clear differences between LARP4 WT and KO (Figure 5—figure supplement 1A). However, there were relatively large spans for the 95% confidence intervals represented by the whisker plots (Figure 5—figure supplement 1A).

To better resolve these data, they were plotted as the fraction of CCS reads vs. PAT length at each time point (Figure 5C). A notable feature is the transition from PAT length of ~100 nt to the peak at ~75 nt for the KO and WT tracings at time zero (Figure 5C, t = 0). Although both time zero profiles were very similar with the majority of mRNA PATs at 25–75 nt, this fraction was slightly higher in amplitude in KO than WT (panel t = 0).

A more striking feature was observed at 30 min as a significant fraction of mRNAs appeared with shorter PATs, with a peak at 35–40 nt in the KO profile, while no such peak fraction was distinguished in the WT profile (Figure 5C, panel t = 30). By 60 min the WT profile showed a fraction of mRNA PATs of ~35–40 nt similar to what appeared earlier in the KO profile.

A substantial amount of the total mRNA mass shifted to shorter PATs during the time course. The t = 0 peak at ~75 nt shrank with time as the 35–40 nt peak enlarged, as expected of a precursor-product relationship (compare t = 60, 120, 240, 480). This would appear to reflect redistribution of phased PABP protection of mRNA PATs during deadenylation that accompanies decay.

The ~75 nt peak seen in Figure 5C is consistent with mRNAs that would harbor three PABP molecules (Webster et al., 2018) (Discussion). By the t = 720 min time, a single ~38 nt peak is in the KO profile while the WT profile retains remnants of the precursor ~75 nt peak. Thus, conversion of the distinct mass of the ~75 nt PAT peak to ~38 nt PATs was initiated faster and completed more readily in LARP4 KO than in WT cells.

The time course PAT decay data in Figure 5C provide evidence that the presence of LARP4 slows deadenylation and this becomes most apparent at the point beyond which mRNAs accumulate with poly(A) lengths of ~75 nt. As discussed later, recent data suggest that the ~75 nt peak represents mRNAs to which three PABPs are bound and undergo deadenylation during the time course.

We identified a set of 194 ISG mRNAs by inducing them in MEFs (Supplementary file 4) and examined their PAT shortening at times after ActD. Each box plot in (Figure 5—figure supplement 1B) comprises >5000 CCS reads representing multiple ISG mRNA transcripts that span a range of PAT lengths that significantly differ in LARP4 WT and KO. To resolve these, the reads were sorted by PAT size into ten bins; those with the longest in the top 10% length bin and so on, with each time point also displayed in each bin (Figure 5—figure supplement 1C). By this approach, the greatest variability in PAT length sorted with the longest PATs, consistent with recent observations (Eisen et al., 2020b). PAT shortening was observed in all of the incremental size ranges (bins) in WT and KO except the last/shortest.

The fraction of total CCS reads for ISG mRNAs vs. PAT length at times after ActD was plotted (Figure 5D). These profiles differ from those of the total mRNA profiles in their overall shapes at t = 0 (Figure 5C). This is most notable in the 30–90 nt range, which is of lower amplitude in the ISG profile, although it substantially increases with time as expected for a set of induced ISGs vs. total RNAs. Reciprocally, the fraction of longer PAT reads is larger in the ISG than in the total gene set at t = 0, reflective of induced mRNAs.

At t = 0 the ISG peak at 35–40 nt for the KO tracing is slightly to the left and higher on the Y-axis than the WT peak, which is below the 0.010 fraction level, suggesting relative accelerated PAT shortening in the KO cells during the IFN treatment period (Figure 5D). At t = 30 min the KO tracing has more clearly accumulated reads in the ~38 nt peak. At t = 60 and t = 120 both the KO and WT tracings accumulate more mass above 0.010 in the 24–50 nt PAT size. By t = 240 min the ~38 nt peaks have sharply risen above the 0.015 fraction mark, with the KO tracing leading toward shorter PATs (Figure 5D).

These data revealed deadenylation kinetics of 194 ISG mRNAs and the larger transcriptome of ~10,500 mRNAs. They showed that shortening of mRNA PATs to ones that accumulate with ~38 nt PATs occurred more readily in LARP4 KO than in WT cells. Thus, by following mRNA PAT lengths over time, deadenylation was observed to occur with less impediment in the absence of LARP4 than in its presence.

Biochemical and structural studies of yeast factors indicate that Pan2-Pan3 can efficiently trim long PATs with serially bound PABP in part because of a unique architecture of PABP-PABP interactions on poly(A), whereas Pan2-Pan3 is inefficient on short poly(A)-PABP mRNPs (Schäfer et al., 2019). Biochemical studies using human factors showed that although PABP recruits the CNOT complex to promote deadenylation of shorter PATs, the CCR4 and CAF1 deadenylase subunits respond very differently (Yi et al., 2018), consistent with studies using the yeast homologs (Schäfer et al., 2019; Webster et al., 2018). The Caf1 subunit of CNOT can deadenylate unprotected poly(A) but the Ccr4 subunit is required to dislodge or peel off PABP (Webster et al., 2018; Yi et al., 2018). These and other findings help explain some aspects of irregular deadenylation noted during mRNA decay (Chen et al., 2017; Webster et al., 2018; Yamashita et al., 2005; Yi et al., 2018).

PAT length phasing is related to nonuniform deadenylation around PABP sites and was shown to be more sharply prominent in human cell lines whose CCR4 levels were decreased (Yi et al., 2018). While the ascending and descending slopes of a phased PAT peak would appear to reflect different deadenylation rates before and after PABP dissociation from a binding site, the larger overall shape of a PAT profile can reflect additional irregularity (Yi et al., 2018). The biochemistry indicates that deadenylation progressively accelerates as bound PABP decreases to two then to one molecule and thereafter (Webster et al., 2018; Yi et al., 2018). The decrease in stability of PABP complexes would appear to result in part from architectural changes as PATs shorten and from loss of PABP-PABP intermolecular contacts (Schäfer et al., 2019; Webster et al., 2018).

Although offsets in PAT peaks on different individual genes can undermine phasing in metagene profiling (Yi et al., 2018), SM-PAT-seq revealed clear differences in the phased peaks of the collective rp-mRNAs from LARP4 KO and WT cells (Figure 3C). Here we propose a model view of our PAT phasing results in Figure 6. First, we note that the rp-mRNA profiles differ most dramatically in that the KO peak that spans ~60–75 nt contains a greater fraction of reads than the WT (Figure 6A). Second, this higher peak is associated with a transition from long to short PATs that is of steeper incline in KO than WT, annotated by the upward red arrow. Third, the WT and KO transitions from long to short PATs intersect at ~110 nt (inverted triangle), a length known to be associated with transition from deadenylation of long PATs by PAN2/3 to shorter PATs by CNOT (Yamashita et al., 2005; Yi et al., 2018).

Figure 6

Download asset Open asset

We suspect that the KO 60–75 nt peak represents PATs bound by three PABPs that include peeling off/dissociation intermediates (Webster et al., 2018; Figure 6C), that are being shortened to accumulate in the peak of ~40–50 nt PATs bound by two PABPs. By extrapolation, PATs bound by two PABP-dimers would appear in the approximate 100–110 nt range in which the KO and WT profiles overlap (Figure 6A). The steeper incline of the KO transition profile might suggest more efficient formation of the 60–75 nt PATs, consistent with LARP4 impeding deadenylation by CNOT. Also consistent with this is the more irregular shape of the ascending WT profile. Likewise, different shapes of the peaks suggest they may represent different intermediates in the presence and absence of LARP4. In any case, it would appear that formation of PABP₍₃₎-PAT complexes by CNOT may occur more readily in the absence of LARP4 (see Figure 6A legend). It is also possible that differences may reflect the presence/absence of proteins other than PABP.

Kinetic profiling affords additional considerations (Figure 6B). At time zero the major PAT peak was 75 nt in KO and WT. Further deadenylation occurred sooner thereafter in KO cells as shorter peaks appeared at 30 min (arrows) that had not yet appeared in WT. The overall profile kinetics suggest that the proposed ~75 nt PABP₍₃₎ complex (Figure 6C) is less stable in LARP4 KO than in WT cells. An intermediate (dashed arrow) appeared between the 75 and 38 nt (solid arrow) peaks in the KO profile at 30 min that was not distinguished in WT (also see Figure 5C). Evidence of this is also in the 60 min KO profile more broadly (bracket) and may reflect a deadenylation intermediate in LARP4 KO cells whose formation is otherwise blocked/inhibited or masked by LARP4. Figure 6C depicts schematized PABP₍₃₎-PABP₍₂₎ PAT intermediates.

These analyses have advanced insight into the role of LARP4 action in PAT lengthening and associated mRNA stabilization. Our findings in the PAT size range in which deadenylation is known to promote mRNA decay in mammalian cells provide a mechanistic link of the activities of PAT 3’ end protection and mRNA stabilization of LARP4. It was proposed that LARP4 could protect PATs from deadenylation by RNA 3’ end binding (Mattijssen et al., 2017). The data here leave open the possibility that LARP4 may directly protect RNA 3’ ends, and also support a model in which it stabilizes PABP on short PATs to shield them from deadenylases. In this model, decreased PABP-PAT stability would be compensated by interactions with LARP4 (Cruz-Gallardo et al., 2019; Mattijssen et al., 2017; Yang et al., 2011). Such interactions are likely intricate as LARP4 can employ two motifs for PABP binding including by its unique PAM2w which can also be used to alternatively bind poly(A) (Cruz-Gallardo et al., 2019). As noted, PAM2 peptides that associate with PABP via its MLLE domain are also found on PAN3 as well as Tob1 and Tob2 which interact with the CAF1 subunit of CNOT. Thus, in the milieu of the cellular deadenylation machinery, the PAM2w would appear to contribute to an intricate network of factors in competition for poly(A) and PABP.

CNOT is an intricate multimeric complex with much regulatory potential for general and mRNA-specific deadenylation (Raisch et al., 2019), refs therein). Effects on long PATs and stable mRNA demonstrated here suggest the possibility that LARP4 can influence mRNA early in its lifespan with penetrant consequence later. Thus this work on LARP4 raises issues regarding the long and short of PAT metabolism.

PolyA-selected RNA was generated from 4 µg of total RNA using poly-T oligo attached magnetic beads from illumina TruSeq Stranded mRNA kit. An adapter was ligated to the polyadenylated 3’ end of the RNAs using T4 RNA Ligase 2 (NEB) with a double-stranded splint linker containing an overhanging stretch of 5 Ts. Following clean up, reverse transcription was performed using Superscript II (ThermoFisher). The RNA strand was digested using RNase H (NEB). The sample was cleaned up using an RNA Clean up kit (Zymo). Second strand synthesis was primed using random hexamers attached to an adapter sequence for PCR and extended using Klenow Fragment 3’−5’ exo- (NEB). The ds-cDNA was purified using 0.6X Ampure PB beads (Pacific Biosciences).

The polyA tails with adjoining 3’UTRs were amplified using primers directed against distal ends of the splint linker (polyA end) and the adapter/random hexamer oligo. The PCR reaction employed KAPA HiFi Hot Start Ready Mix (Roche) for 25 cycles. The products were again purified using 0.6X Ampure PB beads (Pacific Biosciences). Primers and adapters are in Supplementary file 5. Resulting DNA amplicons were quantified using Qubit two and fragment size determined (1700 bp avg) using a high sensitivity DNA chip on a Bioanalyzer 2100 (Agilent), and prepared for sequencing using a SMRTbell Barcoded Adapter Complete Prep kit (Pacific Biosciences). Libraries were loaded onto a SMRT Cell V2 using diffusion loading and sequenced on a PacBio Sequel using 10 hr run-time.

Subread data consisting of a minimum of 3 passes from the PacBio Sequel was converted into CCS using the SMRT Analysis command line toolset and aligned using BLAT against the GENCODE database of full length mRNA sequences (vM16 mouse, v27 human) for assignment to originating mRNA. In addition, each sequence was evaluated for occurrence of poly(A) abutting adapter sequence in either strand direction. PAT length was assigned using the observed poly(A) while allowing for any number of single base disruptions provided that non-A was bracketed by a minimum of 2 A bases on either side as part of the longer continuous poly(A) tract. Reads were accepted as properly assigned and resolved provided they had a PAT longer than 10 bp and a successful alignment to a GENCODE v23 transcript with a BLAT score exceeding a minimum threshold of 50. For the rp-mRNA PAT profiles in Fig 3C the fractional data were smoothed by a sliding +/- 5 nt length averaging prior to plotting.

Gene assigned PAT lengths based on GO annotation (The Gene Ontology Consortium, 2019; avg. length and standard error were calculated and plotted for each subset in each LARP4 expression condition. The 194 MEF ISG mRNAs were those that exhibited ≥2 fold increase in levels in uninduced vs. t = 0 after IFNα treatment, with adjusted p-value of less than 0.1 (Supplementary file 4). Binning of PAT lengths was achieved using the top_frac function on length sorted data from the R package dplyr (Wickham et al., 2019) to produce 10 bins, using all of the non-mitochondrial assigned reads.

Primary MEFs were generated from E14.5 embryos by standard methods. The LARP4 KO MEFs were described (Mattijssen et al., 2017). Each MEF cell line was derived from a different embryo, all females. The MEFs used to generate the data in Figure 3 were immortalized as described using SV40 Large-T antigen (Mattijssen et al., 2017), the experiment in Figure 2 was performed using three independent WT- and 3 LARP4 KO MEF cell lines (N = 3, biological replicates). MEFs used for the IFNα ActD experiment in Figure 5 were cultured following the 3T3 protocol for spontaneous immortalization (Todaro and Green, 1963). The experiment in Figure 5 was performed using two independent WT- and 2 LARP4 KO MEF cell lines (N = 2, biological replicates). The immortalized MEFs and HEK293 cells were cultured in DMEM plus Glutamax (Gibco) supplemented with 10% heat-inactivated FBS (Atlanta Biologicals). HEK293 cells are not commonly misidentified. Nonetheless DNA from these cells was authenticated by ATCC via STR (short tandem repeat) profiling. Standardized testing (ATCC) had verified that the cells were free of mycoplasma infection.

For SM-PAT-Seq (N = 3 biological replicates), 5.5 × 10⁵ HEK293 cells were seeded per well in a 6-well plate. The next morning the cells were ~85% confluent, media was replaced (2 ml with 10% FBS) and transfection was performed. Per well, 2.5 µg pCMV2-FLAG, pCMV2-FLAG-LARP4-WT or pCMV2-FLAG-LARP4-CS-Tyr, representing the 1X, 3X and 11X samples respectively, was co-transfected with 100 ng eGFP plasmid using 7.5 µl Lipofectamine 2000 (Invitrogen) (plasmids as described Mattijssen et al., 2017). For each condition, two wells were transfected. The next day, cells from both wells were combined; 80% of the cell suspension was passed to a 10 cm plate for RNA isolation the next day and 10% added to a well in a 6-well plate for protein isolation. One day later, 48 hr. post transfection, the cells for protein were ~80% confluent, were washed with 2 ml PBS and directly lysed into RIPA buffer (Pierce) containing protease inhibitors (Roche). The cells for total RNA isolation were washed twice with 10 ml PBS per 10 cm plate. Homogenization buffer (800 µl) containing thioglycerol (Maxwell 16 LEV simplyRNA purification kit, Promega) was added to lyse the cells. The lysate was transferred to a conical tube. The Maxwell 16 LEV simplyRNA purification kit protocol was followed using 800 µl lysis buffer and four cartridges per sample (400 µl per cartridge). The total DNase-treated RNA was eluted in 50 µl H2O per cartridge.

For LARP4 overexpression and ISG induction by poly(I:C), dA:dT(80), or AT-rich DNA, HEK293 cells were seeded at 5.5 × 10⁵ cells per well of a 6-well plate. The next morning the cells were ~85% confluent, media was replaced (2 ml with 10% FBS) and transfection was performed. Per well, 2.5 µg pCMV2 or pCMV2-LARP4-WT was co-transfected with 100 ng eGFP plasmid and 100 ng poly(I:C) (Sigma) (N = 3, biological replicates) or 500 ng dA:dT(80) (N = 1), using 7.5 µl Lipofectamine 2000. The next day, cells were passed 1:5 to 4 wells. After 24 hr, cells were washed twice with 2 ml PBS and protein and RNA was isolated as follows. For protein, one well was used; cells were lysed directly in 80 µl of RIPA (Pierce) plus protease inhibitors. For RNA, to the remaining three wells, 100 µl homogenization buffer containing thioglycerol (Maxwell 16 LEV simplyRNA purification kit, Promega) was added per well to lyse the cells. The Maxwell 16 LEV simplyRNA purification kit protocol was followed using 200 µl lysis buffer and one cartridge per sample. Total DNase-treated RNA was eluted in 60 µl H2O per cartridge.

For Figure 4A, lanes 17–20, amounts transfected were 1.25 µg empty pCMV2-FLAG or -FLAG-LARP4-WT, together with 100 ng eGFP plasmid and 3 µg pUC19 or 1 kb AT-rich DNA and 9 µl Lipofectamine2000 (N = 1).

Two LARP4 WT and two KO MEF cell lines immortalized by the 3T3 method (Todaro and Green, 1963), were seeded at a density of 2.5 × 10⁵ cells per well in 2 ml media with 10% FBS in 6-well plates; at least four wells per time point were seeded. The next day the cells were ~90% confluent. To all wells except ‘unstimulated’ was added mouse IFNα (Miltenyi) to 15000 U/ml. After 4 hr ActD was added to all wells (except t = 0, and ‘unstimulated’) to a concentration of 5 µg/ml. Total RNA was isolated at indicated time points. Cells were washed with 2 ml PBS, followed by lysis with 100 µl homogenization buffer containing thioglycerol per well (Maxwell 16 LEV simplyRNA purification kit, Promega). The lysate was transferred to an Eppendorf tube; the Maxwell 16 LEV simplyRNA purification kit protocol was followed using 200 µl lysis buffer and one cartridge per sample. Total DNase-treated RNA was eluted in 50 µl H2O per cartridge.

Northern blotting was as described (Mattijssen et al., 2017). Oligo probe sequences and hybridization temperatures can be found in Supplementary file 6.

Cell lysates as described above were sonicated 3 times for 30 s on the highest setting in a Bioruptor (Diagenode). Lysates were cleared by centrifugation at 13,000 rpm for 20 min at 4°C. Supernatant was transferred to a new tube and total protein concentration determined by BCA (Pierce). Equal amounts of total protein were separated on a NuPAGE 4–12% Bis-Tris protein gel (Invitrogen) and transferred to a nitrocellulose membrane (ThermoFisher) by wet transfer. Primary antibodies were anti-FLAG (Sigma, F1804), anti-GFP (Santa Cruz, sc-8334) and anti-actin (Thermo Scientific, PA1-16890). Secondary antibodies were from LI-COR Biosciences, conjugated to either IRDye 800CW or 680RD. Signals were scanned using the Odyssey CLx imaging system (LI-COR Biosciences).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Alan Kessler (NICHD) for helpful comments on the manuscript, as well as Yevgen Levdansky, Eugene Valkov (NCI) and Amitabh Ranjan for discussion. This work was supported by the Intramural Research Program (HD000412-31 PGD) of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health.

1. Received: June 2, 2020
2. Accepted: August 2, 2020
3. Accepted Manuscript published: August 3, 2020 (version 1)
4. Version of Record published: August 7, 2020 (version 2)

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.