Acetylation of cytidine residues boosts HIV-1 gene expression by increasing viral RNA stability

Summary

Epitranscriptomic RNA modifications, including methylation of adenine and cytidine residues, are now recognized as key regulators of both cellular and viral mRNA function. Moreover, acetylation of the N⁴ position of cytidine (ac4C) was recently reported to increase the translation and stability of cellular mRNAs. Here, we show that ac4C and N-acetyltransferase 10 (NAT10), the enzyme that adds ac4C to RNAs, have been subverted by human immunodeficiency virus 1 (HIV-1) to increase viral gene expression. HIV-1 transcripts are modified with ac4C at multiple discrete sites, and silent mutagenesis of these ac4C sites led to decreased HIV-1 gene expression. Similarly, loss of ac4C from viral transcripts due to depletion of NAT10 inhibited HIV-1 replication by reducing viral RNA stability. Interestingly, the NAT10 inhibitor Remodelin could inhibit HIV-1 replication at levels that have no effect on cell viability, thus identifying ac4C addition as a potential target for antiviral drug development.

Graphical Abstract

eTOC blurb

Tsai et al. report that multiple cytidines on HIV-1 RNAs are acetylated to ac4C by cellular NAT10. Depletion of ac4C by mutagenesis or drug treatment reduces the stability of HIV-1 transcripts, causing reduced viral replication. Thus, HIV-1 has evolved to highjack a host epitranscriptomic modification enzyme to enhance viral replication.

Introduction

Previously, we and others have reported the detection and mapping of several epitranscriptomic modifications on HIV-1 transcripts (Courtney et al., 2019; Kennedy et al., 2016; Lichinchi et al., 2016; Ringeard et al., 2019). These modifications include methylation of the N⁶ position of adenosine (m⁶A), of the C⁵ position of cytidine (m⁵C) and of the ribose moiety of all four ribonucleotides (2’O-methylation, collectively N_m). All three of these epitranscriptomic modifications have now been shown to boost HIV-1 replication in cis. Specifically, m⁶A has been reported to increase viral RNA expression (Kennedy et al., 2016; Lichinchi et al., 2016), while m⁵C boosts viral mRNA translation (Courtney et al., 2019) and N_m residues increase HIV-1 replication by inhibiting activation of the cellular innate immune factor MDA5 by viral RNAs (Ringeard et al., 2019). The m⁶A modification has also been proposed to facilitate HIV-1 RNA export by increasing the binding of HIV-1 Rev to the viral Rev Response Element, though this result remains controversial (Chu et al., 2019; Lichinchi et al., 2016). While m⁶A, m⁵C and N_m have therefore all been shown to increase HIV-1 gene expression, several other epitranscriptomic mRNA modifications remain unexamined. Of particular interest is the mRNA modification ac4C, which was recently reported to enhance cellular mRNA translation and stability (Arango et al., 2018) and shown to be present on virion-derived HIV-1 genomic RNA (gRNA) (McIntyre et al., 2018). We therefore hypothesized that addition of ac4C residues to HIV-1 transcripts might also serve to boost HIV-1 gene expression and replication. We now report that HIV-1 transcripts indeed bear ac4C residues at multiple discrete sites and that the addition of ac4C to viral transcripts is mediated by cellular N-acetyltransferase 10 (NAT10). CRISPR-mediated depletion of NAT10 in the CEM T cell line resulted in a decrease in both viral RNA acetylation and gene expression, while overexpression of wild type, but not inactive mutant forms, of NAT10 increased viral gene expression. The reduction in HIV-1 gene expression in NAT10-depleted CEM T cells could be fully explained by a decrease in viral RNA stability. While HIV-1 gene expression was also reduced when silent mutagenesis was used to remove a subset of viral ac4C sites, this effect was strongly attenuated in the NAT10-depleted CEM T-cells. Lastly, we show that the previously described NAT10 inhibitor Remodelin (Larrieu et al., 2014) can inhibit HIV-1 replication at levels that have no effect on cell growth and viability. Overall these data document that HIV-1 transcripts bear multiple ac4C residues, deposited by cellular NAT10, that boost viral mRNA function in cis.

Results

We have previously used photo-assisted (PA) crosslinking of modification-specific antibodies to 4-thiouridine (4SU)-labelled RNA, followed by RNase footprinting of antibody-bound RNA and deep sequencing of bound RNA fragments, to map both m⁶A (PA-m⁶A-seq) and m⁵C (PA-m⁵C-seq) residues on the gRNAs and mRNAs encoded by HIV-1 and other viruses (Chen et al., 2015; Courtney et al., 2017; Courtney et al., 2019; Kennedy et al., 2016; Tsai et al., 2018). As an ac4C-specific antibody recently became available, we asked if a similar approach (PA-ac4C-seq) might also allow us to map ac4C modification sites on transcripts (Sinclair et al., 2017). Indeed, this technique allowed us to accurately map the previously reported ac4C site found on 18S ribosomal RNA (Ito et al., 2014; Sharma et al., 2015) (Fig. S1A) and we were also able to confirm the previous report (Arango et al, 2018) that ac4C residues on cellular mRNAs are primarily localized to coding sequences (CDS), although ac4C residues were also detected in the 3’ untranslated regions (UTRs), but not 5’UTRs, of cellular mRNAs (Fig. S1B). In Fig. 1A, we report the PA-ac4C-seq analysis of HIV-1 gRNA, isolated from HIV-1 virions generated in the CEM-SS or SupT1 T cell line, or on intracellular viral RNAs, isolated from CEM-SS cells. Despite some degree of variability in signal intensity, we were able to identify ~11 conserved sites of ac4C addition on HIV-1 transcripts that were detected across these three replicates and on additional replicates performed in HIV-1 infected CEM cells and primary CD4⁺ T cells (Fig. S1C), but not in mock infected cells. We note that the relative weakness of the ac4C sites detected on intracellular RNAs in the gag, pol and env regions is expected as these regions are removed by splicing in many intracellular viral RNAs.

Fig 1. NAT10-dependent ac4C is deposited at multiple locations on HIV-1 RNAs.

(A) ac4C sites were mapped by PA-ac4C-seq on the poly(A)⁺ fraction (mRNA) of mock or HIV-1 infected CEM-SS T cells, along with virion RNA produced by infected Sup T1 and CEM-SS cells. See also Fig. S1. (B) Schematic of HIV-1 genome organization drawn to scale (C) PA-ac4C-seq was performed on HIV-1 virion RNA produced in control (Ctrl) or ΔNAT10 CEM cells. NAT10 knock-down is validated in Fig. 3A. (D) PAR-CLIP was performed on CEM cells stably expressing FLAG-NAT10 or FLAG-GFP to identify NAT10 binding sites on HIV-1 RNA. Sequence reads were aligned to the NL4–3 genome. Consistent (across mRNA and virion) ac4C sites are highlighted in yellow and numbered above. 4SU-based CLIP methods result in T>C conversions where protein is cross-linked to 4SU residues, here shown as red-blue bars. All sequencing read pileups shown in counts per million (CPM), where read counts were normalized to the total read count of reads >15nt long with a FASTQ quality score >Q33.

If the data reported in Fig. 1A indeed map authentic ac4C addition sites, then loss of the “writer” acetyltransferase that deposits these marks should result in the loss of detectable ac4C residues on viral RNAs. Mammals are thought to express a single enzyme capable of acetylating RNA, N-acetyltransferase 10 (NAT10), and NAT10 has indeed been reported to add ac4C to several mRNAs and non-coding RNAs in human cells (Arango et al., 2018; Ito et al., 2014; Sharma et al., 2015). However, NAT10 appears to be essential in human cells and previous efforts to knock down NAT10 expression using gene editing by CRISPR/Cas have therefore focused on an exon, exon 5, that is found in the majority of, but not all, mRNAs encoding NAT10 (Arango et al., 2018). We repeated this strategy in the T cell line CEM and generated 3 clonal cell lines in which all three copies of the NAT10 gene were edited in exon 5 (Figs. S2A, B and C), resulting in a large decrease in NAT10 expression (see below). Of note, these clonal cell lines, referred to as ΔNAT10 #3, #7 and #9, did not show any decrease in growth rate when compared to control (Ctrl) CEM cells (Fig. S2D). However, reduced NAT10 expression did result in the expected strong decline in the level of ac4C residues present on not only viral transcripts but also 18S rRNA, as measured by PA-ac4C-seq, thus further validating the PA-ac4C-seq technique (Fig. 1C, Fig. S1A).

If NAT10 is indeed the writer that deposits ac4C on cellular and viral RNAs, then we reasoned it might be possible to detect NAT10 binding to these sites using the photo-assisted cross linking and immunoprecipitation (PAR-CLIP) technique, as previously described (Hafner et al., 2010; Kennedy et al., 2016). We therefore used lentiviral expression vectors encoding FLAG-tagged wild type (WT) NAT10, or FLAG-tagged green fluorescent protein (GFP), to stably express these proteins in CEM cells. As shown in Fig. 1D, we indeed detected FLAG-NAT10, but not FLAG-GFP, binding sites on viral transcripts and these were coincident with the mapped ac4C sites, as expected. Moreover, the mapped NAT10 binding sites showed a similar distribution on cellular mRNAs, with the same enrichment in CDS that was observed using PA-ac4C-seq (Fig. S1B). Of interest, analysis of not only all mapped ac4C sites, but also all mapped NAT10 binding sites, identified very similar C and U-rich consensus sequences, with a central “UCU” motif, in both uninfected and HIV-1 infected CEM cells (Fig. S1D).

If ac4C residues indeed facilitate some aspect of HIV-1 gene expression, then the reduced expression of NAT10, and concomitant reduction in ac4C addition to mRNAs, seen in the ΔNAT10 CEM cells, should result in reduced HIV-1 replication. We analyzed the level of HIV-1 Gag and NAT10 protein expression in control (Ctrl) and ΔNAT10 CEM cells 3 days after infection using WT HIV-1 isolate NL4–3. As may be observed in Fig. 2A, both NAT10 and the viral Gag proteins are expressed at a much lower level in the ΔNAT10 CEM cells, though the GAPDH loading control was unaffected. Similarly, when we infected Ctrl and ΔNAT10 CEM cells with a previously described replication competent HIV-1 that encodes the Nano luciferase (NLuc) gene in place of the dispensable nef gene (Mefferd et al., 2018), we observed a strong reduction in the level of NLuc protein expression (Fig. 2B). The observed reduction in Gag protein expression correlated with a similar reduction in the level of viral RNA expression in WT HIV-1-infected cells, as measured by qRT-PCR using an LTR-specific probe (Fig. 2C). Thus, reduced NAT10 expression indeed results in a lower level of HIV-1 gene expression.

Fig 2. NAT10 enhances the rate of HIV-1 spread in culture.

(A-C) CEM cells in which the NAT10 gene had been edited using CRISPR/Cas (ΔNAT10, see also Fig. S2), along with control CEM cells expressing Cas9 and a non-targeting guide RNA (Ctrl), were assayed for NAT10-associated viral replication phenotypes. (A) WT HIV-1 replication levels, in a spreading or mock infection of Ctrl and ΔNAT10 cells at 3 dpi, were analyzed by Western blot for the HIV-1 capsid protein p24 (B) Ctrl and ΔNAT10 cells were infected with the NL4–3NLuc reporter virus and NLuc activity determined at 3 dpi (C) HIV-1 RNA levels in the samples shown in panel A were determined using qRT-PCR. Three different ΔNAT10 single cell clones were used in panels B & C, with Ctrl cells set at 1. n=4 to 7, error bars=SD. ** 2-tailed T-test, p<0.01. (D) CEM, Ctrl, and ΔNAT10 cells were treated with the NAT10-inhibitor Remodelin. Infected CEM cells were counted at 1 dpi to determine Remodelin toxicity (shown in gray), infected Ctrl (dark blue) and ΔNAT10 (light blue) cells were harvested at 2 dpi and viral RNA levels assayed by qRT-PCR. n=3, error bars=SD. 2-tailed T-test on Ctrl cells for each condition compared to the 0 μM level, **p<0.01, *p<0.05. (E) Schematic of NAT10 functional domains and point mutations. (F) 293T cells were transfected with a plasmid expressing WT NAT10 (WT), or the K290A or G641E NAT10 mutant, or empty vector. Cells were co-transfected with plasmids expressing CD4 and firefly luciferase (FLuc). 3 days later, transfected cells were infected with the NL4-NLuc reporter virus. NLuc expression levels, assayed at 2 dpi, are shown normalized to the FLuc internal control. n=3 to 6, error bars=SD. ** 2-tailed T-test, p<0.01. (G) Western blot showing NAT10 over-expression, NAT10 probed with both FLAG and NAT10 antibodies and GAPDH probed as a loading control.

While these data address how lower NAT10 expression affects HIV-1 replication, it is also possible to inhibit NAT10 function in WT cells using a drug, called Remodelin, that has been reported to inhibit NAT10 function at concentrations that are non-toxic in culture or in mice (Balmus et al., 2018; Larrieu et al., 2014). Indeed, we observed that Remodelin reduced HIV-1 replication in WT CEM cells by up to 70%, but had little effect on HIV-1 replication in the ΔNAT10 CEM cells, at concentrations that did not reduce CEM cell growth (Fig. 2D), thus further validating NAT10 as an HIV-1 co-factor.

As reduced NAT10 expression or function led to diminished HIV-1 replication (Figs. 2A–D), we reasoned that NAT10 activity might be rate limiting for HIV-1 replication. We therefore used 293T cells to overexpress WT NAT10, or mutant forms of NAT10 lacking a functional RNA helicase domain (K290A) or acetyltransferase domain (G641E) (Fig. 2E), due to mutagenesis of residues previously shown to be required for RNA acetyltransferase function (Larrieu et al., 2014; Sharma et al., 2015). The 293T cells were also co-transfected with plasmids expressing the HIV-1 receptor CD4 and firefly luciferase (FLuc), as an internal control, and subsequently infected with the HIV-1 reporter virus encoding NLuc. All three NAT10 proteins were expressed at similar levels that were much higher than endogenous NAT10, as determined by Western blot (Fig. 2G). Importantly, we detected a highly significant ~6x higher level of virally-encoded NLuc expression in the 293T cells overexpressing WT NAT10, but not either NAT10 mutant, when compared to the control cells transfected with empty vector (Fig. 2F).

While the data presented in Fig. 2 demonstrate that NAT10, and the ac4C RNA modification, promote some aspect(s) of the HIV-1 replication cycle, they do not identify which step(s) are affected. To address this issue, we performed a single cycle HIV-1 replication assay, using WT NL4–3, in control (Ctrl) or ΔNAT10 CEM T cells and measured the efficiency of different steps in the HIV-1 replication cycle. Initially, we measured the level of viral Gag expression, which was found to be reduced by ~70% in ΔNAT10 CEM cells (Fig. 3A). Measurement of total viral RNA expression, by qRT-PCR using an LTR-specific probe, also revealed a ~70 % reduction in the ΔNAT10 CEM cells when compared to Ctrl cells, indicating an effect primarily at the RNA level (Fig. 3B). We next sought to define the step in the expression of viral transcripts that was affected when ac4C addition was inhibited. Reduced ac4C addition did not affect the subcellular location of HIV-1 transcripts (Fig 3D), or their alternative splicing (Fig. S3A). Furthermore, none of the steps from cell entry, reverse transcription to proviral integration were affected by loss of ac4C, as no difference was found in the total level of HIV-1 DNA (Fig. 3C). As predicted, analysis of the level of ribosome binding by viral mRNAs (Subtelny et al., 2014), using an assay that revealed a strong positive effect of the m⁵C modification on HIV-1 mRNA translation (Courtney et al., 2019), demonstrated that the presence or absence of ac4C had no discernable effect on the recruitment of ribosomes to viral mRNAs (Fig. S3B). However, the reduced expression of NAT10, and the concomitant loss of ac4C on viral transcripts, did result in a highly significant reduction in the stability of HIV-1 transcripts measured either by capture of nascent RNA after a pulse of 4SU incorporation (Fig. 3E) (Dolken et al., 2008; Duffy et al., 2015), or by use of the transcription inhibitor actinomycin D (Fig. S3C). The observed reduction in viral RNA stability when ac4C addition was inhibited was fully sufficient to explain the decrease in viral RNA and protein expression documented in Figs. 3A and B.

Fig 3. NAT10 depletion destabilizes HIV-1 transcripts.

Ctrl (dark blue) and ΔNAT10 (light blue) cells were infected with HIV-1, treated with the reverse transcriptase inhibitor Nevirapine (NVP) at ~16 hpi and harvested at ~48 hpi for the following single cycle infection assays. (A) Viral Gag levels assayed on Ctrl and 3 single cell clones (#9, #3, #7) of ΔNAT10 cells by Western blot. HIV-1 Gag band intensities (p24 + p55) were quantified, normalized to Ctrl levels (set to 1), and are shown in the right hand panel (Ctrl n=3, ΔNAT10 n=7). (B) Aliquots of the samples visualized in panel A were assayed for viral RNA levels by qRT-PCR. (C) Viral DNA levels in infected Ctrl and ΔNAT10 CEMs, quantified by qPCR using an LTR U3 primer set, n=4. (D) Subcellular fractionation of infected Ctrl & ΔNAT10 cells. Viral RNA in each fraction was quantified by qRT-PCR and is shown as the percentage of cytoplasmic RNA over total (cytoplasmic + nuclear) RNA, n=3. Fractionation validated by Western blot in the right panel, with Lamin A/C as the nuclear marker and GAPDH as the cytosolic marker. Statistical analyses shown in (A-D) used the two-tailed Student’s T test, error bars=SD, **p<0.01. (E) Viral RNA stability analyzed by 4SU-metabolic labeling, followed by isolation of 4SU⁺ nascent RNA at 0, 2 and 4hrs post-4SU-labeling, n=5. Slopes of regression lines compared by ANCOVA, **p=0.0008. See also Fig. S3.

The data presented in Fig 3 argue that, in the case of HIV-1 RNAs, ac4C acts to increase viral gene expression primarily by enhancing viral RNA stability. These data contrast with the previous work proposing that ac4C increases cellular mRNA gene expression by not only increasing mRNA stability but also translation by increasing the CDS decoding efficiency (Arango et al., 2018). If this is indeed the primary mechanism of action of ac4C, then only ac4C sites present in the CDS should affect mRNA function in cis. To address this question, we introduced as many silent C to U mutations as possible into conserved ac4C peaks 4 through 8 in the viral env gene region (Fig. 4A). We then transfected WT 293T cells, which lack CD4 and therefore will not support a spreading infection, with the WT and env mutant forms of HIV-1 and measured HIV-1 Gag protein expression. As shown in Figs. 4B and 4C, the ac4C site mutations introduced into the env gene, which are present exclusively in the 3’ UTR of the viral gag mRNA (Fig. 1B), nevertheless reduced Gag protein expression in cis, both in the producer cells (Fig. 4B) and in the supernatant media (Fig. 4C). The observed reduction of ~60% was not only highly significant but also only slightly less than seen in ΔNAT10 CEM cells (Fig. 3A).

Fig 4. Silent mutagenesis of ac4C sites in env diminishes viral Gag expression.

(A) ac4C sites #4–8 in the HIV-1 env CDS were mutated to remove as many ac4C sites as possible without changing the encoded amino acid. Example of silent mutations in ac4C site #7 in the lower panel. See also Fig. S4. (B) 293T cells were transfected with WT pNL4–3 HIV-1 (WT), or the mutant viral plasmid (mut.) with ac4C sites in the env gene mutated, and Gag expression determined by Western blot. (C) Virions released into the supernatant media of 293T cells transfected with WT or mut HIV-1 expression plasmids were quantified by p24 ELISA. n=4, **p=0.002 (D) WT or ac4C mut virus were normalized using the p24 levels from panel C, and used to infect Ctrl & ΔNAT10 CEM cells. Viral Gag expression from single round (NVP-treated) infections were assayed at 48 hpi by Western blot. (E) The Gag protein bands (p55+p24) from 6 repeats of panel D were quantified and the WT/mut ratio from Ctrl & ΔNAT10 CEMs plotted. (F) Similar to panel F, WT/mut. ratios of the p24 bands only. Significance determined using paired two-tailed Student’s T test, n=6, error bars=SD.

It could be argued that the inhibition of Gag expression seen with the env gene ac4C mutant (Figs 4B and C) was not due to loss of ac4C residues but rather due to disruption of some other RNA sequence element that is functionally significant (we note that none of the mapped ac4C sites in env are located in known functional RNA elements, such as the Rev Response Element, or near any splice sites). To test this idea, we collected WT and mutant virions released by the transfected 293T cells, normalized the p24 level based on Fig. 4C and then infected Control (Ctrl) and ΔNAT10 CEM cells. We then measured the level of Gag expression after a single round of replication in these cells by Western blot. A representative experiment is shown in Fig. 4D, while a compilation of data measuring total Gag protein expression (Fig. 4E) or exclusively p24 Gag expression (Fig. 4G) are also presented. We noted a bigger effect on p24 Gag expression than on total Gag protein expression and therefore present both data sets. As may be observed, the env ac4C site mutations reduced total Gag protein expression by 2.3x, and p24 Gag expression by 4.0x, in the Ctrl CEM cells. In contrast, these same mutations reduced total Gag protein expression by 1.7x, and p24 expression by 2.1x, in the ΔNAT10 CEM cells, and these differences were statistically significant. Therefore, these data demonstrate that the mutagenesis of ac4C sites located in the 3’UTR of the viral Gag mRNA indeed results in a stronger inhibitory phenotype in Ctrl CEM cells expressing WT levels of NAT10 than in the ΔNAT10 CEM cells expressing reduced levels of NAT10, as would be predicted if they indeed act via the same mechanism. Finally, we note that the enhanced translation of mRNAs bearing ac4C modifications (Arango et al, 2018) was reported to be largely due to ac4C residues located in the wobble position of codons. Analysis of the location of C residues in the mapped ac4C peaks present in HIV-1 CDS shows that ~35% are in the wobble position, which is indistinguishable from the predicted ~33% if their location was random. Moreover, comparison of the usage of codons with a C residue in the wobble position over the whole HIV-1 transcriptome versus under the mapped ac4C peaks again revealed no significant difference (Fig. S3D).

Discussion

Previously, several groups have reported that the epitranscriptomic addition of m⁶A to viral transcripts can significantly enhance the replication of a range of different viruses, including HIV-1, influenza A virus, SV40, enterovirus 71, respiratory syncytial virus, metapneumovirus and Kaposi’s sarcoma-associated herpesvirus (Courtney et al., 2017; Hao et al., 2019; Hesser et al., 2018; Kennedy et al., 2016; Lichinchi et al., 2016; Lu et al., 2020; Tsai et al., 2018; Xue et al., 2019; Ye et al., 2017). Less is known about other epitranscriptomic viral RNA modifications, though both m⁵C and N_m residues have been detected on HIV-1 transcripts at levels that are substantially higher than seen on cellular mRNAs and both m⁵C and N_m have been reported to enhance HIV-1 replication in culture (Courtney et al., 2019; Ringeard et al., 2019). Interestingly, the proposed mechanisms used by these distinct epitranscriptomic modifications appear distinct in that m⁶A has been proposed to increase viral mRNA expression levels (Kennedy et al., 2016; Lichinchi et al., 2016) while m⁵C acts primarily by boosting viral mRNA translation (Courtney et al., 2019). Finally, while N_m has been reported to increase HIV-1 replication indirectly by preventing the activation of the host innate immune factor MDA5 by viral transcripts (Ringeard et al., 2019), m⁶A residues added to viral RNAs were recently reported to inhibit the activation RIG-1 in metapneumovirus infected cells (Lu et al., 2020). Here, we extend this previous work by looking at another epitranscriptomic RNA modification, ac4C, that was recently discovered on cellular mRNAs. The ac4C modification was reported to enhance mRNA translation and stability (Arango et al., 2018), and has been detected on purified HIV-1 genomic RNA (McIntyre et al., 2018). We have mapped the ac4C residues present on HIV-1 RNAs to ~11 distinct sites and show that these are, as expected, deposited by the host acetyltransferase NAT10, as inhibition of NAT10 expression results in a loss of ac4C from viral RNAs (Fig.1B). Importantly, loss of ac4C modifications from viral transcripts results in reduced viral gene expression and replication whether caused by a reduction in NAT10 expression due to gene editing (Figs. 2A and B and Fig. 3), inhibition of NAT10 function using the drug Remodelin (Fig. 2D) or by silent mutagenesis of mapped viral ac4C sites (Fig. 4). Arango et al. (Arango et al., 2018) reported that ac4C residues in cellular CDS not only increased mRNA stability but also increased mRNA translation, by increasing decoding efficiency. In contrast, our observation that the reduced addition of ac4C observed in ΔNAT10 T cells inhibits HIV-1 RNA and protein expression to precisely the same extent (Fig. 3A and B) argues that translation of viral mRNAs is not affected. This hypothesis was confirmed by our finding that the reduced viral RNA stability observed in ΔNAT10 T cells (Figs. 3E and S3C) fully explains the reduction in both viral RNA and protein expression seen in these cells. Moreover, we observed that mutagenesis of ac4C sites in the viral env gene, which forms part of the 3’ UTR of the viral mRNA encoding Gag and therefore cannot affect Gag translation directly, nevertheless resulted in a marked drop in Gag protein expression (Figs. 4B and C). It remains unclear whether this indicates that ac4C acts differently on HIV-1 and cellular mRNAs or whether HIV-1 mRNAs are already maximally optimized for ribosome decoding and therefore this parameter cannot be further enhanced by ac4C. Regardless, these data clearly demonstrate that the cellular acetyltransferase NAT10 adds ac4C to HIV-1 transcripts at multiple discrete locations and identify ac4C as the fourth epitranscriptomic modification to enhance viral replication in cis.

STAR Methods

Lead contact and materials availability

All materials including plasmids and cell lines generated in this study are available upon request and will be fulfilled by the Lead Contact, Bryan R. Cullen (bryan.cullen@duke.edu).

Data and code availability

All deep sequencing data have been deposited at the NCBI GEO database under accession number GSE142490.

Experimental model and subject details

Cell lines used in this study include HEK293T cells (referred to as 293T), a kidney epithelial cell line of human female origin purchased from the American Type Culture Collection, and CEM, CEM-SS and SupT1 cells, human CD4+ T cell lines that were obtained from the NIH AIDS reagent program. CEM and CEM-SS cells were of female origin, while SupT1 cells were of male origin. 293T cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) with 6% FBS and 1% Antibiotic-Antimycotic. T cell lines were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% Antibiotic-Antimycotic (Gibco, 15240062).

Primary T cells were from extracted and cultured from human donor blood as follows. Donor blood leukopak was obtained from the Gulf Coast Regional Blood Center (Houston, TX), diluted 1/10 in PBS, overlaid on top of a layer of Histopaque-1077 (Sigma #10771) and spun down 400xg 30mins. The resulting buffy coat was collected, pelleted, and triple washed in PBS supplemented with 0.1% BSA. CD4+ T cells were then isolated from the collected cells using the Dynabead CD4 positive isolation kit (Invitrogen #1131D). The isolated primary T cells were then activated by culturing in RPMI supplemented with 10% FBS, IL-2 (filtered IL-2 transfected 293T cell supernatant, used at 10%), 5μg/ml Phytohemagglutinin (PHA, Roche #11082132001), and 15μg of anti-CD28/CD49d antibody (BD #347690). Primary T cells cultured for four days in activation media before being subjected to HIV-1 infection.

Recombinant virus clones used include the laboratory strain NL4–3, obtained from the NIH AIDS Reagent (Adachi et al., 1986), and the nano-luciferase reporter virus NL4–3-NLuc, in which the viral nef gene has been substituted with the NLuc indicator gene (Mefferd et al., 2018).

Method details

Generation of NAT10-depleted and overexpression CEM cells

ΔNAT10 CEM cells were produced by transducing CEM cells with a lentiviral vector, LentiCRISPRv2 (Sanjana et al., 2014), encoding Cas9 and single guide RNAs (sgRNAs) 5’-TGAGTTCATGGTCCGTAGG-3’ (as previously published (Arango et al., 2018), cloned into LentiCRISPR-NAT10–1) or 5’-GGCTAGTGGTCATCCTCCTA-3’ (GeCKOv2, guide number HGLibA_31166, cloned into LentiCRISPR-NAT10–2). Two days post-transduction, cells were subjected to 1–2 weeks of selection in 1 μg/ml puromycin, then single cell cloned by limiting dilution. Control CEM cells were produced by transduction with a lentiviral vector expressing a non-targeting sgRNA specific for GFP (5’-GTAGGTCAGGGTGGTCACGA-3’, LentiCRISPR-GFP (Courtney et al., 2017)), then puromycin selected and single cell cloned. To validate CRISPR mutations, genomic DNA from knockdown cell lines was extracted using the Zymo Quick-DNA Miniprep Plus kit (#11–397). The genomic region flanking the Cas9 target site from each ΔNAT10 cell line was PCR amplified and cloned into the XbaI/SalI sites of pGEM-3zf+ vector (Promega). 10+ bacterial cell clones of pGEM-genomic-region-plasmid from each CRISPR-knockdown cell clone were isolated for Sanger sequencing. CEM cells constitutively expressing FLAG-NAT10 or FLAG-GFP were produced using lentiviral expression vectors pLEX-FLAG-NAT10 (as described below) or pLEX-FLAG-GFP (Kennedy et al., 2016), following the same transduction and single cell selection process as above.

Expression plasmid construction

A NAT10 cDNA was cloned by PCR from CEM-SS cDNA, digested and ligated into the NotI and EcoRI sites of pK-FLAG-VP1(Tsai et al., 2018), placing NAT10 3’ to a 2xFLAG tag and replacing SV40 VP1. The NAT10 cDNA sequence was confirmed as wild type by Sanger DNA sequencing. The K290A & G641E point mutants were introduced by recombinant PCR: two complementing PCR primers were designed to overlap the mutated region, with the point mutant sequence in the middle. A first round of PCR was done to separately amplify the 5’end-to-mutation site and the mutation site-to-3’end fragments of NAT10, yielding two fragments with a region of homology around the mutation site. By running PCR a second round using two outer primers, the two fragments were joined and amplified into the full length NAT10 CDS containing the point mutation, and then ligated into the NotI and EcoRI sites of pK-FLAG as before. The lentiviral expression construct pLEX-FLAG-NAT10 was constructed by cloning the PCR amplified FLAG-NAT10 cDNA from pK-FLAG-NAT10 into the BamHI and AgeI sites of the pLEX vector (Openbiosystems). All PCR primers used are listed in Supplemental Table 1. All plasmids were prepared using ZymoPURE II plasmid maxiprep kits (Zymo D4203).

Viruses Production

The mutant NL4–3 virus with most ac4C sites in env silently mutated was cloned by replacing the SalI-NheI and NheI-BamHI segments of env with gBlocks (IDT) designed to mutate any C in identified ac4C peaks that could be mutated without changing the encoded amino acid (sequence and primers used for this mutagenesis shown in Fig. S4 and listed in Table S1). The two synthesized gBlocks DNA fragments were first separately amplified by PCR. The amplified DNA fragments were then mixed together along with PCR primers annealing to the 5’end of the first fragment and the 3’ end of the second fragment. As the two fragments were designed with 37nt overlap regions to allow annealing with each other, this second round of PCR would result in the joining of the two fragments. This joined fragment was then digested and ligated into the SalI/BamHI sites of the HIV-1 clone plasmid pNL4–3, replacing the WT segment between these two enzyme digestion sites. All viruses were produced by transfecting the viral expression plasmid into 293T cells using polyethylenimine (PEI, 2 μg plasmid for a 6-well plate, 10 μg for a 10cm plate, 20μg for a 20cm plate, PEI used at 2.5x the μg amount of plasmid DNA). 24 hours post-transfection, the media were replaced with fresh media. The supernatant media were collected at 72 hours post-infection (hpi), passed through a 0.45μm filter, then overlaid onto target cells.

PA-ac4C-seq

Harvest of cellular and virion RNA for modification mapping was performed as previously reported (Courtney et al., 2019). For infection of T cells (including primary T cells and the CD4+ T cell lines CEM, CEM-SS, and SupT1), cells were resuspended in filtered supernatant media from NL4–3-transfected 293T cells. Mock-infected cells were resuspended in filtered media from non-transfected 293T cells. The media were replaced with fresh media at 24 hpi to reduce carry over of 293T-produced virions. Cells were pulsed with 100μM 4SU at 48 hpi and incubated an additional 24 hours. At 72 hpi, cells were collected for extraction of total RNA using Trizol (Invitrogen), followed by mRNA enrichment using the Poly(A)Purist MAG Kit (Ambion). For virion RNA, the supernatant of infected cells was concentrated through Centricon Plus-70 centrifugal filters (100,000 NMWL membrane, Millipore), then the virions were pelleted by ultracentrifugation through a 20% sucrose cushion at 38,000 rpm for 90 min. The resulting virus pellet was lysed in Trizol for RNA extraction. Cellular poly(A)+ RNA and virion RNA were then subjected to ac4C site recovery following the PA-m⁶A-seq protocol (Chen et al., 2015; Courtney et al., 2019; Tsai et al., 2018). with two modifications: an ac4C-specific antibody was used, and the incubation of antibody with RNA was overnight.

PAR-CLIP

Single cell cloned CEM cells expressing FLAG-GFP or FLAG-NAT10 were used. Ten 15 cm plates of 293T cells were seeded to package virus for each (GFP+ or NAT10+) infection. Four days post pNL4–3 transfection, the virus-containing supernatant media were harvested and filtered through a 0.45μM filter. 300 million FLAG-GFP or FLAG-NAT10 expressing CEM cells were resuspended in the filtered media. At 48 hp, infected cells were pelleted and resuspended in 350 ml of fresh RPMI supplemented with 100μM 4SU. At 72 hpi, cells were collected, washed twice in PBS, then irradiated with 2500×100 μJ/cm² of 365 nm UV. The above procedure was repeated twice to obtain sufficient biomass, and PAR-CLIP was performed with a FLAG antibody, as previously described (Hafner et al., 2010; Kennedy et al., 2016).

Illumina sequencing & bioinformatic data analysis

RNA recovered from the PA-ac4C-seq and PAR-CLIP procedures were used for cDNA library preparation using the NEBNext Small RNA Library Prep Set for Illumina (NEB E7330S), then sequenced using Illumina NextSeq 500, or NovaSeq 6000 sequencers at the Duke Center for Genomic and Computational Biology (GCB) Sequencing and Genomic Technologies Shared Resource. Sequencing data analysis was done as previously described (Courtney et al., 2019; Tsai et al., 2018). Sequencing adaptors were first removed using FASTX-toolkit v0.014 or Cutadapt v1.18 (Gordon and Hannon, 2010; Martin, 2011). Then, sequencing reads >15 nt with fastq quality score >33 were aligned to the human genome (hg19) using Bowtie (Langmead et al., 2009). The human-non-aligning reads were then aligned (allowing up to 1 mismatch) to the HIV-1 NL4–3 sequence with a single copy of the long terminal repeat (LTR, U5 on the 5’ end, and U3-R on the 3’ end), essentially 551–9626 nt of GenBank AF324493.2. As UV-crosslinked 4SU results in characteristic T>C conversions, an in-house Perl script was used to discard alignments devoid of T>C mutations. After file format conversions using SAMtools (Li et al., 2009), data was visualized using Integrative Genomics Viewer (IGV) (Robinson et al., 2011). To provide HIV-1 genome maps accompanying the modification mapping figures, the genbank full annotation file (.gb file) associated with 551–9626 nt of GenBank AF324493.2 was downloaded and visualized with SnapGene Viewer (GSL Biotech LLC).

Metagene and motif analysis

For meta-gene analysis, the human-aligned PA-ac4C-seq reads were subjected to peak calling using MACS2 (Zhang et al., 2008) (parameters --nomodel --tsize=50 --extsize 32 --shift 0 --keep-dup all -g hs). Peak calling on the NAT10 PAR-CLIP data was done using PARalyzer v1.1 (Corcoran et al., 2011) (with the parameters:

• BANDWIDTH=3 CONVERSION=T>C

• MINIMUM_READ_COUNT_PER_GROUP=10

• MINIMUM_READ_COUNT_PER_CLUSTER=3

• MINIMUM_READ_COUNT_FOR_KDE=5 MINIMUM_CLUSTER_SIZE=15

• MINIMUM_CONVERSION_LOCATIONS_FOR_CLUSTER=2

• MINIMUM_CONVERSION_COUNT_FOR_CLUSTER=2

• MINIMUM_READ_COUNT_FOR_CLUSTER_INCLUSION=2

• MINIMUM_READ_LENGTH=10

• MAXIMUM_NUMBER_OF_NON_CONVERSION_MISMATCHES=1

• EXTEND_BY_READ). Metagene analysis was performed using metaPlotR (Olarerin-George and Jaffrey, 2017).

Motif analysis was performed on the sequences of called peaks using MEME following a published m⁶A-seq pipeline (Bailey et al., 2009; Dominissini et al., 2013).

Codon usage preference analysis

The codon usage frequency in the 11 conserved HIV-1 ac4C sites were analyzed using the on-line codon usage tool of the Sequence Manipulation Suite (https://www.bioinformatics.org/sms2/codon_usage.html) (Stothard, 2000). The calculated usage rate in viral ac4C sites for each codon is then divided by the codon usage rates of all 8 viral genes (also analyzed with the same codon usage tool) to calculate fold enrichment.

Viral infection of 293T cells

293T cells seeded in 6 well plates were transfected using PEI with 1.6 μg of pK-FLAG-NAT10 plasmids or empty pK vector, along with 250 ng of CD4 expression vector(Bieniasz et al., 1997), and 100 ng of firefly luciferase (FLuc) expression plasmid pcDNA3-FLuc. In parallel, 10μg of pNL43-NLuc was transfected into 293T cells in 10 cm plates. All media were changed the next day, and the NAT10/CD4/FLuc+ infected target cells split into 12 well plates two days later. NL43-NLuc virus-containing supernatant was harvested on day 3, filtered and brought up to 12mls with fresh media, and overlaid onto target cells at 1ml virus per well. At 2 or 3dpi, the supernatant media was removed from infected cells, the cells washed 3x with PBS, then lysed in passive lysis buffer (Promega, E1941). NLuc and FLuc activity was assayed using the Nano Luciferase Assay Kit and Luciferase Assay System (Promega, N1120 & E1500).

Infection assays under NAT10-inhibition with Remodelin

Control or ΔNAT10 CEM cells were seeded at 0.75 million cells per well in 1ml in a 12 well plate, and treated with Remodelin (Sigma SML1112–5MG, dissolved in DMSO to 2mM) at the needed concentration. Lower concentration sets were compensated with equal volumes of DMSO. The next day, cells were overlaid with 1 ml of NL4–3 virus. Additional drug was supplemented to compensate for the additional 1 ml volume of the virus. To compensate for potential drug decay over 3 days, 0.25x additional drug was added at 2 days post initial drug treatment. Cells were counted at 24 hpi to assess toxicity and harvested at 48 hpi for assay of viral RNA levels.

Viral infection of wild type and ΔNAT10 CEM cells

NL4–3 virus was packaged in 293T cells in 10 cm plates, transfected with 10 μg of pNL4–3 using PEI, the media were replaced the next day with 10 ml RPMI. 3 days post-transfection, CEM cells were counted and seeded at 1 million cells per well in 0.5 ml RPMI in 12 well plates. Virus-containing supernatant media harvested from 293T cells were filtered and supplemented with fresh RPMI to a total volume of 12 ml. 1 ml of this virus was then used to overlay the 1 million cells/0.5 ml in 12 well plates. For spreading infections, cells were harvested and PBS washed at 72 hpi. For single-round infections, cells were treated with 133μM of the reverse transcriptase inhibitor Nevirapine (Sigma SML0097) at 16 hpi, then harvested at 48 hpi.

Viral RNA levels in infected cells were measured using quantitative real time-PCR (qRT-PCR). Harvested cells were washed in PBS, then the RNA extracted using Trizol (Invitrogen). Total RNA was then treated with DNase I (NEB), and reverse transcribed using the Super Script III reverse transcriptase (Invitrogen). qPCR was performed with Power SYBR Green Master Mix (ABI), with primers targeting either the U3 region of HIV-1 LTR or spanning splice donor 1 and splice acceptor 1 (D1-A1). qPCR readouts were normalized to GAPDH levels using the delta-delta Ct (ΔΔCt) method. All PCR primers used are listed in Supplemental Table 1.

Viral DNA levels in infected cells were assayed by qPCR on genomic DNA extracted using the Zymo Quick-DNA Miniprep Plus Kit (Zymo #D4068). DNA qPCR was performed with Power SYBR Green Master Mix (ABI), with primers targeting the U3 region of HIV-1 LTR, normalized to the genomic DNA levels of the beta Actin gene using the ΔΔCt method. All PCR primers used are listed in Supplemental Table 1.

HIV-1 RNA splicing was assayed by Primer-ID tagged deep sequencing (Courtney et al., 2019; Emery et al., 2017). HIV-1 Gag protein expression levels were analyzed by Western blot. Western blot band intensity was quantified using Image J software (Schneider et al., 2012). Released viral particles were quantified using an HIV-1 p24 antigen capture ELISA assay (Advanced Bioscience Laboratories #5421).

Sub-cellular fractionation

Sub-cellular fractionation was done on single-round infected cells, as previously described (Malim et al., 1989; Tsai et al., 2018): 1 million infected cells were washed twice in PBS, partially lysed in 300μl NP40 lysis buffer (10 mM Tris-HCl pH 7.5, 10 mM NaCl, 3 mM MgCl₂, 0.5% (w/v) NP40), vortexed for 5secs, then left on ice for 5mins. Nuclei were spun down in a microfuge for 10secs, and the supernatant collected as the cytosolic fraction. The remaining nuclear pellet was wash in another 300μl of lysis buffer, split in two tubes and pelleted again. Each fraction was split in halves for analysis of protein and RNA levels from the same samples, by Western blot and reverse-transcription qPCR, respectively.

Quantification of ribosome-associated RNA

Ribosome association assays were performed using single-round infected cells, as previously described (Courtney et al., 2019; Subtelny et al., 2014). Virus-infected and NVP-treated cells were supplemented at ~48h post infection with 100μg/ml of cycloheximide for 10mins prior to harvest. Cells were then washed in PBS and lysed in ribosome lysis buffer (10mM Tris-HCl pH 7.4, 5mM MgCl₂, 100mM KCl, 1% Triton X, Protease inhibitor, 2mM DTT, 100μg/ml cycloheximide and RNase inhibitor). 1/10 of the lysates was isolated as the input sample. Lysates were sheared by passing through a 26G needle four times, then clarified by centrifugation at 1300xg 10min. The lysate supernatants were brought up to 10ml in lysis buffer and overlaid on a 30% sucrose cushion, also made up in lysis buffer, and centrifuged at 164,000xg for 2 hrs at 4°C. RNA was then extracted from the ribosomal pellet using TRIzol and assayed by qRT-PCR as described above.

RNA decay assays

The nascent RNA isolation method used is a combination of two protocols (Dolken et al., 2008; Duffy et al., 2015). Single-round-infected cells were pulsed at 48 hpi with 150 μM 4SU for 1.5 hours, the cells then washed and resuspended in 4SU-free fresh RPMI. Cells were collected at 0, 2, and 4 h after 4SU wash out and RNA extracted using Trizol. 500 ng of MTSEA-biotin-XX (Biotium 89139–636) dissolved in 10 μl of Dimethyl formamide (Sigma D4551) was used to biotinylate 6 μg of RNA in a 50 μl reaction mixture with 20mM HEPES pH7.4 and 1mM EDTA at room temperature for 30 min. Excess biotin was removed by two rounds of chloroform extraction followed by isopropanol precipitation of RNA. 100 μg of streptavidin magnetic beads (NEB, S1420S) were pre-blocked with glycogen, then co-incubated with the Biotinylated-4SU+ RNA at room temperature for 15 min. The resulting RNA-bead complex was washed 3x with wash buffer (10mM Tris HCl pH7.4, 100mM NaCl, 1mM EDTA, 0.005% Tween-20). Elution was done twice with 25μl of freshly made elution buffer (20mM HEPES pH7.4, 100mM DTT, 1mM EDTA, 100mM NaCl, 0.05% Tween-20), followed by RNA purification with the Zymo RNA Clean & Concentrator-5 kit (#11–326).

For the transcription stop method, single-round-infected cells were treated at 48 hpi with 5μg/ml Actinomycin D (Sigma A9415), and harvested 0, 2, 4, and 6 h later. For all RNA decay assays, viral RNA levels were assayed by qRT-PCR. Data were analyzed using the ΔΔCt method, where readouts of viral RNA at time t (D1-A1 spliced RNA specific primer set) were normalized to β-Actin levels corrected to the expected RNA level prior to decay of t hours, utilizing the published β-Actin t_1/2 in CEM cells of 13.5hrs (Ct of actin at time t is corrected by −t/13.5)(Leclerc et al., 2002). Each β-actin-normalized viral RNA value was then calculated as the fold change from that detected at time point 0. Statistical analysis of the rate of RNA decay was done on the log2 transformation of each fold change value, using GraphPad Prism 8 software, comparing slopes of linear regression lines by analysis of covariance (ANCOVA).

Quantification and statistical analysis

All statistical details are listed in the figure legends. All averaged data include error bars that denote standard deviation, with single data points shown. All statistical analysis done by 2-tailed Student’s T-test comparing subjects to control, with the exception of RNA decay studies, where the slopes of regression lines were compared by ANCOVA.

Supplementary Material

2

Table S1. PCR primers, CRISPR guides and synthesized gBlock DNA fragment sequences. Related to STAR methods. See attached Excel file.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit Anti-N4-acetylcytidine (ac4C) [EPRNCI-184–128]	Abcam	Cat# ab252215; RRID:AB_2827750
Rabbit polyclonal anti-NAT10	ProteinTech	Cat#:13365–1-AP; RRID: AB_2148944
Mouse monoclonal anti-FLAG, Clone M2	Sigma-Aldrich	Cat# F1804; RRID:AB_262044
Mouse monoclonal anti-GAPDH	Proteintech	Cat# 60004–1-Ig; RRID:AB_2107436
Mouse monoclonal anti-β-Actin	Proteintech	Cat# 66009–1-Ig; RRID:AB_2687938
Mouse monoclonal anti-Lamin A/C (E-1)	Santa Cruz	Cat# sc-376248; RRID:AB_10991536
Mouse monoclonal anti-HIV-1 p24 Gag [24–3]	NIH AIDS Reagent Program	Cat# 6458; RRID:AB_2810883
Anti-mouse-HRP	Sigma-Aldrich	Cat# A9044; RRID:AB_258431
Anti-rabbit-HRP	Sigma-Aldrich	Cat# A6154; RRID:AB_258284
Bacterial and Virus Strains
HIV-1 strain NL4–3	NIH AIDS Reagent Program	Cat#114
HIV-1 nano-luciferase reporter virus NL4–3-NLuc	Mefferd et al., 2018	N/A
HIV-1 env ac4C site mutant NL4–3	This study	N/A
Biological Samples
Human donor blood leukopak	Gulf Coast Regional Blood Center	N/A
Chemicals, Peptides, and Recombinant Proteins
Polyethylenimine (PEI)	Carbosynth	Cat# FP106457
4-thiouridine (4SU)	Carbosynth	Cat# NT06186
Remodelin	Sigma-Aldrich	Cat# SML1112
Nevirapine	Sigma-Aldrich	Cat# SML0097
MTSEA-biotin-XX	Biotium	Cat# 89139–636
Dimethyl formamide (DMF)	Sigma-Aldrich	Cat# D4551
Streptavidin magnetic beads	NEB	Cat# S1420S
Actinomycin D	Sigma-Aldrich	Cat# A9415
Critical Commercial Assays
Poly(A)Purist MAG Kit	Ambion	Cat#AM1922
NEBNext Small RNA Library Prep Set for Illumina	NEB	Cat#E7330S
Nano Luciferase Assay Kit	Promega	Cat#N1120
Luciferase Assay System	Promega	Cat#E1500
HIV-1 p24 antigen capture ELISA assay	ABL	Cat#5421
RNA Clean & Concentrator-5 kit	Zymo Research	Cat#R1013
Quick-DNA Miniprep Plus Kit	Zymo Research	Cat#D4068
ZymoPure II plasmid maxiprep kit	Zymo Research	Cat#D4203
Histopaque-1077	Sigma-Aldrich	Cat#10771
Dynabead CD4 positive isolation kit	Invitrogen	Cat #1131D
Phytohemagglutinin	Roche	Cat #11082132001
anti-CD28/CD49d	BD	Cat #347690
Deposited Data
All next generation sequencing data	This study	NCBI GEO accession #GSE142490
Experimental Models: Cell Lines
Human: HEK293T	Duke Cell Culture Facility	RRID:CVCL_0063
Human: CEM	NIH AIDS Reagent Program	Cat#117; RRID:CVCL_8178
Human: CEM-SS	NIH AIDS Reagent Program	Cat#776; RRID:CVCL_J318
Human: SupT1	NIH AIDS Reagent Program	Cat#100 ; RRID:CVCL_1714
Human: Cas9 control CEM	This study	N/A
Human: ΔNAT10 CEM clone #9	This study	N/A
Human: ΔNAT10 CEM clone #7	This study	N/A
Human: ΔNAT10 CEM clone #3	This study	N/A
Oligonucleotides
All cloning and PCR oligos (see Table S1)	This study	N/A
Recombinant DNA
Env ac4C mut gene blocks (see Table S1)	This study	N/A
pK-FLAG-NAT10	This study	N/A
pK-FLAG-NAT10-K290A	This study	N/A
pK-FLAG-NAT10-G641E	This study	N/A
pLEX-FLAG-GFP	Kennedy et al., 2016	N/A
pLEX-FLAG-NAT10	This study	N/A
LentiCRISPR-v2	Addgene	Cat#52961; RRID:Addgene_52961
LentiCRISPR-GFP	Courtney et al., 2017	N/A
LentiCRISPR-NAT10–1	This study	N/A
LentiCRISPR-NAT10–2	This study	N/A
Software and Algorithms
ImageJ	Schneider et al., 2012	https://imagej.nih.gov/ij/
Cutadapt v1.18	Martin, 2011	https://cutadapt.readthedocs.io/en/stable/index.html
FASTX-toolkit v0.014	Gordon & Hannon, 2010	http://hannonlab.cshl.edu/fastx_toolkit/index.html
Bowtie	Langmead et al., 2009	http://bowtie-bio.sourceforge.net/index.shtml RRID:SCR_005476
Samtools	Li et al., 2009	http://samtools.sourceforge.net/
MACS2	Zhang et al. 2008	https://github.com/taoliu/MACS/
PARalyzer v1.1	Corcoran et al., 2011	https://ohlerlab.mdc-berlin.de/software/PARalyzer_85/
metaPlotR	Olarerin-George & Jaffery, 2017	https://github.com/olarerin/metaPlotR
MEME	Bailey et al. 2009	http://meme-suite.org/doc/download.html
Sequence Manipulation Suite: Codon Usage	Stothard, 2000	https://www.bioinformatics.org/sms2/codon_usage.html
SnapGene Viewer	GSL Biotech LLC	https://www.snapgene.com/
Prism 8	Graph pad	https://www.graphpad.com/scientific-software/prism/
Integrative Genomics Viewer (IGV)	Robinson et al., 2011	http://software.broadinstitute.org/software/igv/

Highlights.

• Cytidines on HIV-1 RNAs are acetylated to ac4C by N-acetyltransferase 10 (NAT10)

• Deposition of ac4C residues enhances viral RNA stability

• Silent mutagenesis of ac4C sites causes a NAT10-dependent drop in HIV-1 replication

• Remodelin, a small molecule NAT10 inhibitor, can inhibit HIV-1 gene expression