Upregulation of RNA cap methyltransferase RNMT drives ribosome biogenesis during T cell activation

Mice

Rnmt fl/fl mice with loxP sites flanking exon 3 of Rnmt were sourced from Taconic Artemis Gmbh. CD4-Cre (Tg(Cd4-cre)1Cwi) (26) mice were gifted by Doreen Cantrell. Mice were maintained on a C57B6/J background in the Biological Resource Unit at the University of Dundee using procedures approved by the University Ethical Review Committee and under the authorization of the UK Home Office Animals (Scientific Procedures) Act 1986.

Preparation of cells

Lymph nodes (inguinal, brachial, axillary, superficial cervical, mesenteric, lumbar, caudal), spleens and thymi were dissected from mice, and mashed through a 70 μm cell strainer (Falcon) to prepare cell suspensions. Blood was taken by cardiac puncture (following confirmation of death) and diluted in RPMI medium, red blood cells were lysed with ACK buffer (150 mM NH₄Cl, 10 mM KHCO₃, 0.1 mM Na₂EDTA). Cells were counted either using the BD FACSVerse (BD Biosciences) or Novocyte (Acea Biosciences) flow cytometer, by haemocytometer (Fastread counting slides from Immune systems) or using AccuCheck counting beads (Thermo Fisher Scientific) in combination with a BD LSRFortessa (BD Biosciences) flow cytometer.

Western blotting

For analysis of RPs, LARP1 and RAM in supplemental figure 1 M T cells were lysed directly in Laemmli buffer (50 mM Tris pH 6.8, 2% SDS, 10% glycerol, 100 mM DTT, bromophenol blue). For analysis of RNAPII, RNMT and RAM, T cells were lysed in F-buffer (10 mM Tris–HCl pH 7.05, 50 mM NaCl, 0.5% Triton X-10). For RNAPII analysis, lysates were sonicated 3 × 30 s on a Bioruptor (Diagenode) at medium intensity. Equivalent cell numbers were resolved by SDS-PAGE. For standard western blots, proteins were transferred onto PVDF membranes (Millipore) with Tris-glycine buffer (25 mM Tris, 190 mM glycine, 20% methanol). For analysis of ribosomal proteins the higher pH CAPs buffer (10 mM CAPs 20% methanol) was used. Membranes were incubated with primary antibodies (Supplementary Table S1), followed by the relevant HRP-conjugated secondary antibodies (Supplementary Table S1) and developed with Pierce Super signal ECL (Thermo Fisher scientific), visualised with X-ray film or using an ImageQuant LAS 4000 (GE healthcare). Western blots were quantified using NIH ImageJ software.

^m7GTP binding assay

Jurkat cells were cultured in RPMI media, supplemented with non-essential amino acids (Gibco) and sodium pyruvate (Gibco). Jurkat cell pellets were lysed using Buffer B (10 mM HEPES pH 7.4, 1 mM MgCl₂, 10 mM NaCl, 50 mM NaF, 0.5% NP-40). Lysates were pre-cleared with Protein A-Sepharose beads (Generon, PCA-125) for 30 min at 4°C, rotating. Lysate were split into equal parts, each incubated with 40 ul ^m7GTP-Agarose (Jena Bioscience) or GTP-Sepharose for 2 h at 4°C. Beads were washed with Buffer B. After washes, beads were re-suspended in LDS sample buffer (Novex) with 0.1 M DTT diluted in Buffer B. ^m7GTP and GTP pulldown samples were resolved on the SDS-PAGE gels with SDS Running Buffer (25 mM Tris, 250 mM glycine, 0.1% SDS), then transferred to PVDF membranes, which were probed with rabbit anti-LARP1 antibody (ProteinTech).

Flow cytometry

Cells were incubated with labelled antibodies (Supplementary Table S1) and FC block (anti CD16/32, Biolegend) in FACS buffer (PBS + 2% FCS (Gibco)). Dead cells were labelled with 0.1 μg/ml DAPI or Fixable Viability Dye eFluor 780 (eBioscience). For apoptosis, cells were stained with annexin V APC (Biolegend) in annexin V buffer (10 mM HEPES pH 7.4, 140 mM NaCl, 2.5 mM CaCl₂, 4 mM KCl, 0.75 mM MgCl₂). For translating ribosomes, cells were incubated with 1 ug/ml puromycin for 10 min at 37°C in T cell culture medium, stained with Fixable Viability Dye eFluor 780, fixed with 1% PFA in PBS (Santa Cruz Biotech) on ice, and permeablised and stained for puromycin and surface markers in saponin buffer (PBS+ 5% FCS + 0.1% saponin) (27). For cell cycle analysis, cells were fixed in 4% PFA, permeablised in ice cold 90% methanol 10% PBS, and stained with antibodies then 1 μg/ml DAPI in FACS buffer. Data was collected using either a BD FACSVerse, or LSR Fortessa and analyzed using FlowJo10 (BD).

T cell culture

To measure proliferation, apoptosis, puromycin incorporation and forward scatter, lymph node cells were cultured in T cell culture medium (RPMI + 10% heat inactivated FCS + pen/strep + 50 μM 2ME) with 0.5 ug/ml Ultra-LEAF purified anti-mouse CD3ϵ antibody (clone 145-2C11, Biolegend), and 0.5 ug/ml Ultra-LEAF purified anti-mouse CD28 antibody (clone 37.51, Biolegend). 20ng/ml IL-2 (proleukin, Norvatis) was added on day 2. For analysis of protein or RNA in CD4 T cells, magnet sorted CD4 T cells were cultured in T cell culture medium on anti-CD3 (5 μg/ml)/anti-CD28 (1 μg/ml) coated plates.

Preparation of CD4 T cells

To purify T cells, for most experiments CD4 T cells were magnet sorted using EasySep mouse CD4 T cell isolation kit (Stemcell Technologies), for the ribosome footprinting experiment the mouse naïve CD4 T cell kit was used, for CAP-MAP, eCLIP, and Ψ-seq the Miltenyi mouse CD4 T cell kit was used. For proteomics and naïve T cell RNA sequencing analysis, naïve CD4 T cells were FACS sorted using the BD Influx (BD Biosciences), purities were > 98%.

RNA assays

N-7 cap guanosine methylation assay was performed according to (14). The in vivo cap methylation assay was performed according to (28). More details in supplemental materials.

RNA was extracted using Tri-reagent (Sigma) or Trizol (Thermo Fisher scientific) and quantified by Nanodrop (Thermo Fisher scientific). For qPCR, cDNA synthesised using iScript (BioRad), then qPCRs performed on a BioRad CFX384 using ssoFast Eva Green supermix (Bio-rad) and primers in Supplementary Table S1. For mRNA extraction a Dynabeads mRNA DIRECT™ Purification kit was used on total RNA with two rounds of purification. mRNA was quantified using the Qubit RNA HS Assay Kit (Thermo Fisher scientific). For ³H uridine labelling of RNA, cells were cultured with 0.37 MBq [5,6–³H] uridine/ml medium, RNA and mRNA extracted and scintillation counts measured using Optiphase Hisafe3 scintillation fluid (Perkin Elmer) and a Tri-Carb 4910TR scintillation counter (Perkin Elmer).

For northern blotting, 1 μg total RNA was analysed according to the previous report (29) with methylene blue staining of 28S and 18S rRNAs for loading control. Oligonucleotides probes for pre-rRNA detection were: 5′-ITS1, 5′-GCTCCTCCACAGTCTCCCGTTTAATGATC-3′; ITS1-1b, 5′-CTCTCACCTCACTCCAGACACCTCGCTCCA-3′; ITS2-1, 5′-ACCCACCGCAGCGGGTGACGCGATTGATCG−3′ (30,31). 3′-end biotin-labelled oligonucleotides were synthesized by Eurofins Genomics.

DNA qPCR

DNA was isolated from CD4 T cells using a DNA blood and tissue kit (Qiagen). qPCR was performed as described for cDNA.

RNA sequencing

Naïve T cell transcriptome libraries were prepared by the Tayside Centre for Genomic Analysis using an Illumina Truseq kit with Ribozero Gold rRNA depletion. 20 h activated T cell cytoplasmic transcriptome and ribosome footprint libraries were prepared using the Illumina Truseq Ribo Profile (mammalian) kit. eCLIP was carried out as described in the seCLIP protocol with a few modifications (32). Pseudouridine-seq was carried out according to the Ψ-seq protocol (33). CD4 T cell RNAPII CHIPseq (RNAPII: GSM1903984, input: GSM1903982) (34), and RAM knockdown HeLa cell RNAseq (GSE87767) (16), datasets were obtained from NCBI Geo. Library preparation methods and analysis are further detailed in supplemental materials and methods.

Mass spectrometry

CAP-MAP mass spec analysis was carried out as described previously. TMT proteomics was performed as described previously (35). Ribo Mega-SEC was performed as described previously (29). Further methods and analyses are detailed in supplemental materials and methods.

Statistical analysis

Excluding ‘omics’ data, statistical analyses and charts were generated in Graphpad Prism. Where used, t-tests were double sided and unpaired.

Further information

Further detailed protocol information is available in the supplemental materials. Antibodies, primers and other key reagents are listed in Supplementary Table S1.

RNMT is upregulated following T cell activation

Following T cell activation, major increases in transcription and translation are likely to require increased production of the ^m7G cap to protect, process and translate the nascent mRNA. Consistent with this, we found that the ^m7G cap methyltransferase, RNMT, and co-factor, RAM, were upregulated downstream of T cell receptor (TCR) signalling, induced using crosslinking antibody stimulation, and maintained throughout CD4 and CD8 T cell activation in the presence of the stimulatory cytokine interleukin 2 (IL2) (Figure 1B, C, Supplementary Figure S1A–E). To assess the role of RNMT in T cells, we crossed mice with Rnmt floxed alleles to mice expressing CD4-Cre to delete Rnmt at the double positive (DP) stage of T cell development (Figure 1B, Supplementary Figure S1F). These cells are the developmental precursors to the mature T cells that are the subject of this study (26). In Rnmt^fl/fl CD4-Cre mice (Rnmt cKO, conditional knock-out), RNMT protein was gradually lost during thymic T cell development and was minimal in naïve or activated lymph node CD4 T cells compared to the equivalent cells in Rnmt^fl/fl mice (controls) (Figure 1D, Supplementary Figure S1G–J). Deletion of Rnmt exon 3 in genomic DNA and loss of RNA reads aligning to Rnmt exon 3 was confirmed in Rnmt cKO naïve CD4 T cells (Supplementary Figure S1K, L). The expression of RNMT and RAM are co-dependent in other cell lineages, and RAM was repressed in activated Rnmt cKO CD4 T cells (Supplementary Figure S1M-N) (36). ^m7G cap methyltransferase activity was undetectable in Rnmt cKO in activated (Figure 1E, F) or naïve (Supplementary Figure S1O, P) CD4 T cell extracts, confirming the loss of RNMT protein and the absence of any entirely redundant ^m7G cap methyltransferase. Rnmt cKOs had normal numbers of DP, CD4 single positive (SP) and CD8 SP in the thymus compared to controls, but ∼50% less CD4 and CD8 T cells in the peripheral lymph nodes, spleen and blood (Figure 1G-H). The proportions of central or effector memory CD4 and CD8 T cells were equivalent in Rnmt cKO mice and controls (Supplementary Figure S1Q). Thus, loss of Rnmt has a moderate impact on T cell homeostasis.

Cap guanosine N-7 methylation is reduced in Rnmt cKO T cells

Although ^m7G cap methyltransferase activity was minimal in Rnmt cKO CD4 T cell extracts (Figure 1E-F, Supplementary Figure S1O, P), it was important to determine the extent of mRNA and cap synthesis in vivo. The mRNA content per cell was unchanged in Rnmt cKO naïve CD4 T cells, but the induction of mRNA synthesis following activation was impaired suggesting that RNMT and cap formation is more critical for RNA production in activated T cells (Figure 2A, B). Nevertheless, uridine incorporation into mRNA was equivalent on a per-mRNA basis at 20 h following activation, indicating comparable net mRNA synthesis/degradation at this timepoint (Figure 2C). We determined the rate of endogenous mRNA cap methylation in activated CD4 T cells by culturing them with ³H-methionine, which is converted to the methyl donor, S-adenosyl methionine (SAM), leading to ³H-labelling of methyltransferase substrates (28). Methylation of internal nucleotides did not change in Rnmt cKO CD4 T cells compared to controls, consistent with no change in net mRNA synthesis/degradation (Figure 2D, Supplementary Figure S2A). Cap methylation, which includes both ^m7G and first nucleotide methylation, was reduced by ∼30% in Rnmt cKO CD4 T cells (Figure 2D, E). Given that we do not expect first nucleotide methylation to be affected, this indicates a significant reduction in the rate of ^m7G formation.

The cap structures present on mRNA from activated T cells were quantified by CAP-MAP mass spectrometry (37). Rnmt cKO CD4 T cells had significantly more cap structures which lack ^m7G: GpppG_m, Gppp^m6A_m, and GpppA_m, and significantly less mature ^m7G-cap structures: ^m7GpppC_m, ^m7GpppG_m and ^m7GpppA_m than controls (Figure 2F). Our protocols do not detect ^m7GpppU_m. Over 99% of control CD4 T cell mRNA had a ^m7G cap structure, whereas in Rnmt cKOs ^m7G-caps were reduced by 28% on mRNAs initiating with guanosine and by 17% on mRNAs initiating with adenosine (Figure 2G). Thus in Rnmt cKO CD4 T cells, ^m7G-caps are synthesised at a lower rate, their steady state concentration is lower, but the modification is still abundant, indicating selection for mRNA bearing a complete ^m7G cap structure. Interestingly, Rnmt cKOs had a 35% decrease in mRNA initiating with cytosine, but ∼95% of this mRNA had an ^m7G cap (Figure 2G, H). This implies that the incomplete cap generated when the first transcribed nucleotide is cytosine (GpppCm), is the least tolerated incomplete cap structure, leading to loss of RNA initiating with a C on Rnmt deletion.

Rnmt deletion has a gene-specific impact including selective loss of TOP mRNAs

To determine the impact of Rnmt deletion on RNA expression we performed RNA sequencing (seq) analysis on naïve CD4 T cells. Of 11 219 genes which passed analysis thresholds, 753 were significantly reduced and 845 were significantly increased in Rnmt cKOs compared to controls (Figure 3A, Supplementary Table S2). Pathway analysis indicated that the downregulated transcripts were enriched for ribosomal protein genes, with 58 of the 69 genes in this pathway significantly decreased in Rnmt cKOs (Figure 3A, Supplementary Figure S2B). Ribosomal protein (RP) genes were the only significantly enriched pathway among the RNMT-sensitive genes, so we focussed further analysis on this group. Differential expression of selected genes was confirmed by RT-qPCR (Supplementary Figure S2C). RP transcripts were highly abundant in naïve CD4 T cells: using the reads per kilobase per million (RPKM) as an approximation of transcript abundance they constituted ∼27% of the protein coding transcriptome, which reduced to ∼18% in Rnmt cKOs (Figure 3B, C). In HeLa cells, genes with highest RNAPII occupancy have the highest sensitivity to knockdown (KD) of RNMT-RAM (16). Analysis of published RNAPII ChIP data indicated that RP genes also have high RNAPII occupancy in naïve CD4 T cells (34), however, other transcripts with high RNAPII occupancy remain unaffected in Rnmt cKO CD4 T cells (Supplementary Figure S2D), indicating that high RNAPII is not predictive of RNMT dependence in CD4 T cells. Furthermore, unlike in HeLa cells, RNAPII CTD (C-terminal domain) Ser-5 and Ser-2 phosphorylation and net mRNA synthesis/degradation were not RNMT-dependent suggesting that transcription is not significantly regulated through RNMT in CD4 T cells (Figure 2C, Supplementary Figure S2E, F).

Transcripts encoding RPs and certain translation factors are members of a family of mRNAs initiating with cytosine followed by a terminal oligopyrimide (TOP) motif in their 5′ untranslated region (UTR). TOP RNAs bind to a specialised cap binding protein, La-related protein 1 (LARP1), which enhances their stability and represses their translation in MTOR and CDK1-regulated mechanisms that coordinate ribosome biogenesis with cell growth and circadian rhythms (24,38–45). Canonical cap binding proteins such as EIF4E of the EIF4F complex and NCBP2 of the CBC complex bind to caps on all mRNA and as such have a high specificity for the ^m7G cap, but little preference for the identity of the first transcribed nucleotide (46,47). LARP1 is a specialised cap binding protein which has a >90 fold preference for TOP RNA (in which C is the first transcribed nucleotide) over mutated TOP RNA with G as the first transcribed nucleotide (40). Therefore, we initially focussed on investigating a potential role for LARP1 in mediating the impact of RNMT on TOP-RNA expression, although other cap binding proteins are also likely to have a role in regulating their stability, export and expression (see later). Importantly, other established TOP RNAs were repressed in Rnmt cKO CD4 T cells (Figure 3A), and RP and other TOP RNAs were repressed in RNMT-RAM KD HeLa cells (Supplementary Figure S2G), and in RNMT-compromised U2OS cells, indicating a conserved mechanism linking TOP RNA abundance to RNMT activity (16,48).

LARP1 binding is predictive of RNMT dependence

LARP1 has multiple modes of RNA binding. The LA-module, made up of the LA-motif, PAM2 (PABPC1-binding) domain and RRM (RNA recognition motif), can bind to the TOP motif in a cap-independent manner (49). The LA-module also recognises the polyA tail directly and through interacting with the polyA-binding protein PABPC1 (23,24,44,49,50). The DM15 domain is the cap-binding domain and binds the cap structure and TOP motif (40,51,52). The DM15 domain binds to the RNA cap through a pi-stacking mechanism consistent with a preference for the more positively charged ^m7G-cap over a G-capped structure (40,52). Accordingly, we found that LARP1 had increased interaction with ^m7GTP over GTP in vitro (Figure 3D).

In HEK293 cells, PAR-CLIP analysis demonstrated that LARP1 is bound predominantly to the CDS and 3′UTRs (39). We sought to identify candidates for cap-dependent (5′ proximal) LARP1 binding in naïve CD4 T cells by enhanced crosslinking immunoprecipitation (eCLIP) sequencing analysis (Supplementary Figure S3A) (32). In contrast to the previous report, we found that in CD4 T cells, the majority of direct LARP1-RNA interactions occurred close to transcript start sites (TSS) in both coding (Figure 3E) and non-coding (lnc-RNA and pseudogenes) targets (Figure 3F). The TSS-adjacent binding pattern was observed with three anti-LARP1 antibodies (Figure 3G). Overall, we identified 367 LARP1 binding sites within transcript 5′ UTRs, and these had the expected pyrimidine rich motif (Supplementary Figure S3B, Figure 3H, Supplementary Table S3). These TSS-adjacent and TOP motif-containing binding sites are good candidates for cap-dependent binding through the DM15 domain. Several binding sites were also identified in long non-coding RNA (lncRNA), including snoRNA host genes, and upstream of or within RP and translation factor pseudogenes (Supplementary Figure S3B) (53). Examples of LARP1 eCLIP data alignment to individual coding and non-coding genes are shown in Figure 3I and Supplementary Figure S3C. CAGE data from the FANTOM5 project are shown as a reference (54).

LARP1 binding to 5′UTRs, but not other sites, correlated with mRNA repression in the Rnmt cKO CD4 T cells (Figure 4A). LARP1-bound lncRNA (or snoRNAs derived from them) and pseudogenes were also repressed in Rnmt cKO cells (Figure 4A). mRNAs identified as LARP1 targets included the classical TOP transcripts encoding RPs and translation factors and additional targets in other pathways including signalling, energy metabolism, splicing and nucleolar functions; amongst these functional groups, RPs, translation factors and nucleolar proteins were most consistently affected by Rnmt loss (Supplementary Table S3, Figure 4B). Since LARP1 can bind to the TOP motif via the La-module in a cap-independent manner, it is likely that some of the RNMT-insensitive LARP1 targets are bound through this mechanism, which may have distinct regulatory effects (49).

The pattern of LARP1 binding in Rnmt cKO cells was similar to controls (Figure 3E, F, I, Supplementary Figure S3C), and notably the expression of LARP1 itself is not altered in Rnmt cKO CD4 T cells (Supplementary Figure S4A, B). As discussed earlier, only fully methylated ^m7GpppC_m, and not GpppC_m, was readily detected in Rnmt cKO CD4 T cells, whereas unmethylated caps on non-TOP RNAs, GpppA_m, Gppp^m6A_m and GpppG_m, were abundant (Figure 2F, G). These data imply that only TOP-RNAs with a mature cap (i.e. ^m7GpppC_m) are stable, and GpppC_m TOP mRNA is degraded leading to the selective loss of GpppC_m transcripts in Rnmt cKO CD4 T cells. TOP RNAs with incomplete cap structures would be predicted to bind less efficiently to LARP1, which could be responsible for their reduced stability in Rnmt cKO cells. In addition, we cannot rule out the enhanced dependency of TOP RNAs on other cap binding proteins or associated factors.

Expression of small RNAs and mRNA processing is RNMT-dependent

Many small RNAs including snRNAs, scaRNAs, snoRNAs and their precursors are initially synthesised with a ^m7G cap. In addition, several snoRNAs are excised from the introns of pre-mRNAs or lncRNA hosts, of which we identified 60 as LARP1-bound (Supplementary Table S3). 50 snoRNAs were detected in the naive CD4 T cell RNAseq analysis, of which 33 were reduced in Rnmt cKO T cells (Figure 4C, Supplementary Table S4), these included both those encoded within LARP1 targets and those that were not known to be associated with LARP1. All RNAPII-transcribed snRNAs were repressed in naïve Rnmt cKO CD4 T cells compared to controls (Figure 4C, Supplementary Table S4). The snRNA U6 was not repressed, as expected, since as a RNAPIII-transcript it is synthesised without a ^m7G cap. Four of the five scaRNAs detected were also repressed in the Rnmt cKO T cells compared to controls. Interestingly, the expression of several snoRNA and snRNAs were also reduced in RNMT-RAM KD HeLa cells, indicating a conserved role for RNMT in small RNA expression (Supplementary Figure S4C) (16).

snoRNAs are involved in the cleavage, methylation and pseudouridylation of ribosomal RNA, and are required for ribosome biogenesis (see later). snRNAs are essential constituents of spliceosomes and scaRNAs guide snRNA modifications. Additionally, a major function of the RNA cap is to promote splicing of the first intron via an interaction with cap binding complex (CBC) (55–57). Therefore, we investigated splicing in naïve CD4 T cells. Rnmt cKO T cells had a mild defect in first intron splicing resulting in a relative increase in reads aligning to intron 1 across many genes (Figure 4D-E). This finding was recapitulated in cytoplasmic RNA from activated T cells (Supplementary Figure S4D-E). The intron 1 bias of the splicing defect in Rnmt cKOs is consistent with defects in CBC recruitment rather than a general effect on splicing. Following activation, Rnmt cKO CD4 T cells have less mRNA per cell than their control counterparts (Figure 2A); reduced mRNA processing linked to loss of CBC interaction is likely to contribute towards this effect.

RNMT is essential for T cell activation

Having established that RNMT has gene-specific impacts in naïve T cells, we proceeded to determine its role in T cell activation where transcription, translation and RNMT activity are considerably increased (Figure 1B,C). Rnmt cKO and control lymph node cells were activated ex vivo for 2 days by CD3 and CD28 co-receptor stimulation, followed by IL2 addition. The high affinity IL2 receptor, CD25, and type C lectin, CD69, hallmarks of TCR stimulation, were induced equivalently in Rnmt cKO and control CD4 and CD8 T cells following activation, indicating that TCR signalling is intact and induction of certain proteins is RNMT-independent (58) (Figure 5A, Supplementary Figure S5A). Control T cells proliferated rapidly following stimulation whereas Rnmt cKO T cells failed to proliferate, had delayed cell cycle entry after activation, and exhibited increased apoptosis (Figure 5B–D, Supplementary Figure S5B, C). Activated Rnmt cKO T cells had lower forward scatter, a proxy of cell size, than controls (Figure 5E, Supplementary Figure S5D). Puromycin incorporation, an indicator of the number of translating ribosomes, was decreased in Rnmt cKO T cells following activation (27) (Figure 5F, Supplementary Figure S5E). Thus, the biosynthetic pathways driving increased protein synthesis, cell growth, and cell cycle entry are RNMT-dependent.

RNMT has a gene-specific impact on transcript abundance in activated T cells

We determined the transcriptome-wide effects of Rnmt deletion in CD4 T cells following 24 h of activation: of 10,126 genes analysed, 2874 were significantly repressed and 2726 were significantly increased (Figure 5G, Supplementary Table S5). In activated CD4 T cells, the mRNA per cell was reduced in Rnmt cKOs by 30% compared to controls, reflecting the impact of RNMT on mRNA transcription, processing and stability (Figure 2B). Because of this impact on RNA levels, single gene repression will be underestimated and upregulation overestimated in RNA seq analysis. LARP1 target mRNAs were repressed in Rnmt cKO activated CD4 T cells, and pathways analysis revealed decreased expression of mRNAs encoding RPs and proteins involved in fatty acid oxidation and synthesis, the electron transport chain and TCA cycle, and DNA repair and replication (Figure 5G, Supplementary Figure S5F). Decreased expression of genes involved in these biosynthetic and energy producing pathways is consistent with the decreased proliferation and delayed cell cycle entry observed.

Activated Rnmt cKO CD4 T cells had fewer translating ribosomes (Figure 5F), and a proportion of their mRNA lacked the ^m7G cap required for efficient translation (Figure 2F, G). Thus, we investigated if there was a gene-specific reduction in translation. Ribosome footprinting detects ribosome-bound mRNA (ribosome protected fragments, RPFs), the relative abundance of which approximates translation efficiency (TE) (59). Since the number of translating mRNAs and total mRNA abundance are lower in Rnmt cKOs most mRNA will be translationally repressed compared to controls. Therefore, we cannot infer absolute differences in translation, but can compare relative changes in efficiency of translation, i.e. whether certain mRNAs are preferentially translated in Rnmt cKOs or whether all are equally repressed. For the majority of mRNAs, the fold change in RPFs matched the fold change in mRNA, indicating equivalent TE in Rnmt cKOs compared to controls (Figure 5H-I). However, in the Rnmt cKO, 732 mRNAs had a significantly decreased TE, of which the majority had increased total RNA; and 276 mRNAs had a significantly increased TE, of which the majority had decreased total RNA (Figure 5J, K, Supplementary Table S5). Despite the loss of RNMT activity (Figure 1E, F), much of the mRNA in Rnmt cKO CD4 T cells had a ^m7G cap (Figure 2F, G), however, the percentage ^m7G-capped RNA varied depending on the first transcribed nucleotide. In the Rnmt cKO cells, RNAs initiating with C, which includes TOP-RNAs, had the greatest proportion of caps with a ^m7G structure (Figure 2F, G). We would expect mRNAs with the ^m7G cap to have greater translation efficiency due to their greater affinity to the translation initiation factor EIF4E (46). Accordingly, of the 276 mRNAs with increased TE in Rnmt cKO, 93 were LARP1 target (C-capped) mRNAs. Conversely, we predict that those RNAs that are stable in the absence of a ^m7G cap will be translationally repressed mRNA in the Rnmt cKO.

Ribosome protein synthesis is RNMT-dependent in activated CD4 T cells

We investigated whether deletion of Rnmt had an impact on the proteome of naïve and activated CD4 T cells using quantitative mass spectrometry. This revealed few differences in protein expression in Rnmt cKO naïve T cells compared to controls; of 3874 proteins detected 12 were significantly increased and 12 were significantly decreased (Figure 6A, Supplementary Table S6). Although LARP1 target mRNAs were repressed in these cells, their proteins were minimally affected (Figures 3A and 6A, B). TOP mRNA is known to be translationally repressed in naïve T cells, and thus forms a pre-transcribed pool poised for expression upon T cell activation (3). Indeed, following activation, many proteins were expressed in a Rnmt-dependent manner (Figure 6C, D, Supplementary Table S7). Consistent with decreased protein synthesis and cell size, the protein content was about 15% lower in Rnmt cKO cells than controls (Figure 6C, D). From 4495 proteins detected, 838 proteins were significantly decreased and 32 were increased. Pathway analysis indicated that RPs were the most strongly repressed proteins, with translation factors, DNA replication and cholesterol biosynthesis also repressed (Supplementary Figure S6A). Reduced expression of the RPs RPS3 and RPS19 in the Rnmt cKO activated CD4 T cells was confirmed (Figure 6E, Supplementary Figure S6B).

In activated Rnmt cKO CD4 T cells, we found a correlation between LARP1-binding to the transcript 5′UTR and repression of the protein product (Figure 6D). Thus, although in naïve RNMT cKO T cells the selective repression of TOP RNAs had minimal impact on protein synthesis, these transcripts were critical for protein expression following activation. Of LARP1 target transcripts groups, the most repressed proteins were RPs, translation factors and nucleolar proteins (Figure 6F). Focussing on the nucleolar proteins which contribute to ribosome biogenesis (Figure 6G), the Rnmt cKO activated T cells had reduced expression of the RNAPI component TAF1D, RNAPI/III component POL1RC and nucleolin (NCL) all of which promote rRNA transcription (60–62). Components of the U3 snoRNP which mediates rRNA cleavage, UTP14a and IMP3 (63,64), the pseudouridine methyltransferase EMG1 and nucleophosmins 1 and 3 were also decreased (65,66). These ribosome biogenesis factors all have LARP1 binding sites.

In HeLa cells, RNMT regulates ribosome biogenesis through upregulation of c-MYC (67). Notably c-Myc mRNA was induced rather than repressed on Rnmt deletion in activated CD4 T cells and there was no evidence of repression of the Myc transcriptional program (Supplementary Figure S6C). Conversely RNMT expression is dependent on Myc in activated T cells and induction of RNMT is likely to contribute to the mechanism through which Myc induces ribosome biogenesis (Supplementary Figure S6D) (68).

Ribosome biogenesis is RNMT-dependent in activated CD4 T cells

Since the ribosomal proteins, snoRNAs and ribosome biogenesis factors were repressed following Rnmt deletion, we investigated the impact on ribosome biogenesis. Whilst the total RNA (mostly rRNA) content of naïve CD4 T cells was unchanged on Rnmt deletion, following activation the RNA per cell and net rate of total RNA synthesis/decay was reduced in Rnmt cKO T cells compared to controls (Figure 7A-C). Activated Rnmt cKO CD4 T cells had lower levels of monosomes and polysomes than controls, as assessed by Ribo Mega-SEC profiling (29) (Figure 7D, E). Thus, RNMT is critical for the induction of ribosome biogenesis following T cell activation.

Having established that ribosome biogenesis was reduced in Rnmt cKOs we investigated the impact on ribosome quality control. We characterised the protein content of ribosomes in activated Rnmt cKO CD4 T cell ribosomal fractions and found no significant differences in relative abundance of any RP in any ribosomal fraction (Supplementary Figure S7, Table S8). snoRNAs can either guide pre-rRNA cleavage, pseuouridylation or ribose 2′O-methlyation and we found examples of each class to be downregulated in Rnmt cKO CD4 T cells (Supplementary Table S4). Analysis of rRNA processing intermediates demonstrated equivalent rRNA processing in the control and Rnmt cKO activated CD4 T cells (Figure 7F, G). Pseudouridine (Ψ)-seq analysis of rRNA activated CD4 T cells detected pseudouridine in 75 positions (33), but revealed no differences in the conversion of uridine to pseudouridine in the control and Rnmt cKO (Figure 7H, Supplementary Table S9). Thus, RNMT controls the rate, but not the quality of ribosome biogenesis.

In conclusion, we report that during T cell activation, upregulation of RNMT co-ordinates mRNA capping and maturation with increased translational capacity by selectively upregulating the proteins and non-coding RNAs involved in ribosome biogenesis.