Mitochondrial RNA capping: highly efficient 5’-RNA capping with NAD+ and NADH by yeast and human mitochondrial RNA polymerase

Bacterial and eukaryotic nuclear RNA polymerases (RNAPs) cap RNA with the oxidized and reduced forms of the metabolic effector nicotinamide adenine dinucleotide, NAD+ and NADH, using NAD+ and NADH as non-canonical initiating nucleotides for transcription initiation. Here, we show that mitochondrial RNAPs (mtRNAPs) cap RNA with NAD+ and NADH, and do so more efficiently than nuclear RNAPs. Direct quantitation of NAD+- and NADH-capped RNA demonstrates remarkably high levels of capping in vivo: up to ~60% NAD+ and NADH capping of yeast mitochondrial transcripts, and up to ~10% NAD+ capping of human mitochondrial transcripts. The capping efficiency is determined by promoter sequence at, and upstream of, the transcription start site and, in yeast and human cells, by intracellular NAD+ and NADH levels. Our findings indicate mtRNAPs serve as both sensors and actuators in coupling cellular metabolism to mitochondrial gene expression, sensing NAD+ and NADH levels and adjusting transcriptional outputs accordingly.

In contrast to a m 7 G cap, which is added to nascent RNA by a capping complex that associates with eukaryotic RNA polymerase II (RNAP II) (2,6,(14)(15)(16), an NAD cap is added by RNAP itself during transcription initiation, by serving as a non-canonical initiating nucleotide (NCIN) (17) [reviewed in (18,19)]. NCIN-mediated NAD capping has been demonstrated for bacterial RNAP (13,17,20,21) and eukaryotic RNAP II (17). Thus, whereas m 7 G capping occurs after transcription initiation, on formation of the ~20th RNA bond, and occurs only in organisms harboring specialized capping complexes, NAD capping occurs in transcription initiation, on formation of the first RNA bond, and because it is performed by RNAP itself, is likely to occur in most, if not all, organisms.
Jäschke and co-workers developed a method that combines click-chemistrymediated covalent capture and high-throughput sequencing, "NAD captureSeq," to detect NAD + -capped RNA (12,22). Jäschke and co-workers used this method to identify NAD + -capped RNAs in bacterial cells [Escherichia coli and Bacillus subtilis; (12,13)]. Parker, Kiledjian, and co-workers used the same method to identify NAD + -capped RNAs in eukaryotic cells [Saccharomyces cerevisiae and human cell line HEK293T; (10,11)]. Notably, the identified Saccharomyces cerevisiae NAD + -capped RNAs included not only RNAs produced by nuclear RNAPs, but also RNAs produced by mitochondrial RNAP (mtRNAP). The eukaryotic nuclear RNAPs--RNAP I, II, and III--are multi-subunit RNAPs closely related in sequence and structure to bacterial RNAP (23-26); in contrast, mtRNAPs are single-subunit RNAPs that are unrelated in sequence and structure to multi-subunit RNAPs and, instead, are related to DNA polymerases, reverse transcriptases, and DNA-dependent RNAPs from T7-like bacteriophages (27-33).
The identification of NAD + -capped mitochondrial RNAs in S. cerevisiae raises the question of whether eukaryotic single-subunit mtRNAPs--like the structurally unrelated bacterial and eukaryotic nuclear multi-subunit RNAPs--can perform NCIN-mediated capping. A recent review discussed evidence supporting the hypothesis that human mtRNAP can perform NCIN capping (18). Here, we show that single-subunit S. cerevisiae mtRNAP and human mtRNAP perform NCIN-mediated capping with NAD + and NADH in vitro, and do so substantially more efficiently than bacterial and eukaryotic multi-subunit RNAPs. Further, we show that capping efficiency is determined by promoter sequence, we demonstrate very high levels of NAD + and NADH capping--up to ~50%--of mitochondrial transcripts in vivo, and we demonstrate that the extents of capping in vivo, and distributions of NAD + capping vs. NADH capping in vivo are influenced by intracellular levels of NAD + and NADH.

S. cerevisiae and human mtRNAPs cap RNA with NAD + and NADH in vitro
To assess whether mtRNAP can cap RNA with NAD + and NADH, we performed in vitro transcription experiments (Figures 1 and S1). We analyzed S. cerevisiae mtRNAP and a DNA template carrying the S. cerevisiae mitochondrial 21S promoter (34), and, in parallel, human mtRNAP and a DNA template containing a derivative of the human mitochondrial light-strand promoter, LSPAGU (35) ( Figure 1C-D, top). We performed reactions using either ATP, NAD + , or NADH as the initiating entity and using [a 32 P]-GTP as the extending nucleotide ( Figure 1C-D, middle). We observed efficient formation of an initial RNA product in all cases ( Figure 1C-D, middle). The initial RNA products obtained with ATP, but not with NAD + or NADH, were processed by RppH, which previous work has shown to process 5'-triphosphate RNAs to 5'-monophosphate RNAs (36) ( Figure 1B), whereas the initial RNA products obtained with NAD + or NADH, but not with ATP, were processed by NudC, which previous work has shown to process 5'-NAD + -and 5'-NADH-capped RNAs to 5'-monophosphate RNAs (12, 37) ( Figure 1B).
The results establish that S. cerevisiae mtRNAP and human mtRNAP are able to generate initial RNA products using NAD + and NADH as NCINs.
We next assessed whether the initial RNA products formed using NAD + and NADH as NCINs can be extended to yield full-length RNA products ( Figure 1C-D, bottom). We performed parallel transcription experiments using either ATP, NAD + , or NADH as the initiating entity and using [a 32 P]-GTP, ATP, UTP, and 3'-deoxy-CTP ( Figure 1C, bottom) or [a 32 P]-GTP, ATP, and UTP ( Figure 1D, bottom) as extending nucleotides. We observed efficient formation of full-length RNA products in all cases, and we observed that full-length RNA products obtained with NAD + or NADH, but not with ATP, were sensitive to NudC treatment ( Figure 1C-D, bottom). Similar results were obtained in transcription experiments using [a 32 P]-ATP or [ 32 P]-NAD + as the initiating entity and using non-radiolabeled extending nucleotides ( Figure S1). The results establish that S. cerevisiae mtRNAP and human mtRNAP not only generate initial RNA products, but also generate full-length RNA products, using NAD + and NADH as NCINs.

S. cerevisiae and human mtRNAPs cap RNA with NAD + and NADH more efficiently than bacterial and nuclear RNAPs
We next determined the relative efficiencies of NCIN-mediated initiation vs. ATPmediated initiation, (kcat/KM)NCIN / (kcat/KM)ATP, for mtRNAPs (Figures 2 and S2). We performed reactions with S. cerevisiae mtRNAP and DNA templates carrying the S. cerevisiae mitochondrial 21S promoter or 15S promoter (Figures 2A and S2A), and, in parallel, with human mtRNAP and DNA templates carrying the human mitochondrial light-strand promoter (LSP) or heavy-strand promoter (HSP1) (Figures 2B and S2B).
We obtained values of (kcat/KM)NCIN / (kcat/KM)ATP of ~0.3 to ~0.4 for NCIN-mediated initiation with NAD + and NADH by S. cerevisiae mtRNAP and ~0.2 to ~0.6 for NCINmediated initiation with NAD + and NADH by human mtRNAP. These values imply that NCIN-mediated initiation with NAD + or NADH is up to 40% as efficient as initiation with ATP for S. cerevisiae mtRNAP and up to 60% as efficient as initiation with ATP for human mtRNAP.
To enable direct comparison of efficiencies of NCIN capping by mtRNAPs vs. cellular RNAPs on the same templates under identical reaction conditions, we performed transcription assays using a "fork-junction" template that bypasses the requirement for sequence-specific RNAP-DNA interactions and transcription-initiation factor-DNA interactions for transcription initiation ( Figure 2C, top). In these experiments, we observe efficiencies of NCIN-mediated initiation with NAD + and NADH by mtRNAP that are fully ~10-to ~40-fold higher than efficiencies of NCIN-mediated initiation with NAD + and NADH by E. coli RNAP and S. cerevisiae RNAP II ( Figure 2C, bottom). We conclude that S. cerevisiae mtRNAP and human mtRNAP cap RNA with NAD + and NADH more efficiently than bacterial RNAP and eukaryotic nuclear RNAP II.
We next used the same fork-junction template and reaction conditions as in assays performed with mtRNAPs to determine the efficiency of NCIN-mediated initiation with NAD + and NADH for the single-subunit RNAP of bacteriophage T7 (T7 RNAP) ( Figure S3). The efficiencies of NCIN-mediated initiation with NAD + and NADH by T7 RNAP were nearly as high as the efficiencies of NCIN-mediated initiation by mtRNAPs.
We conclude that there is a quantitative difference in the efficiency of NCIN capping between members of the single-subunit RNAP family (T7 RNAP and mtRNAPs) and members of the multi-subunit RNAP family (bacterial RNAP and eukaryotic nuclear RNAP II).

Promoter sequence determines efficiency of RNA capping by mtRNAP
In previous work, we have shown that NCIN capping with NAD + and NADH by bacterial RNAP is determined by promoter sequence, particularly at and immediately upstream of, the transcription start site (TSS) (17,21). NCIN capping by bacterial RNAP occurs only at promoters where the base pair (nontemplate-strand base:template-strand base) at the TSS is a A:T (+1A promoters), and occurs most efficiently at the subset of +1A promoters where the base pair immediately upstream of the TSS is purine:pyrimidine (-1R promoters). We have further shown that sequence determinants for NCIN capping by bacterial RNAP reside within the template strand of promoter DNA (i.e., the strand that templates incoming nucleotide substrates) (21).
To determine whether the specificity for A:T at the TSS (position +1), observed with bacterial RNAP, also is observed with mtRNAP, we assessed NAD + capping by S. cerevisiae mtRNAP using promoter derivatives having A:T or G:C at position +1 ( Figure   3A-B). We observed NAD + capping in reactions performed using the promoter derivative having A:T at position +1, but not in reactions performed using the promoter derivative having G:C at position +1 ( Figure 3B), indicating specificity for A:T at position +1. To determine whether specificity resides in the template strand for A:T at position +1, we analyzed NAD + capping with S. cerevisiae mtRNAP using template derivatives having noncomplementary nontemplate-and template-strand-nucleotides (A/C or G/T) at position +1 ( Figure 3B). We observed NAD + capping only with the promoter derivative having T as the template strand base at position +1, indicating that specificity at position +1 resides in the template strand.
To determine whether specificity for R:Y at position -1, observed with bacterial RNAP, also is observed with mtRNAP, we analyzed NAD + capping by S. cerevisiae mtRNAP using promoter derivatives having either R:Y (A:T or G:C) or Y:R (C:G or T:A) at position -1 ( Figure 3C). We observed higher efficiencies of NAD + capping with promoter derivatives having R:Y at position -1 than with promoter derivatives having Y:R ( Figure 3C). To determine whether specificity at position -1 resides in the DNA template strand, we performed experiments using promoter derivatives having Y (C or T) or R (A or G) at position -1 of the template strand and having an abasic site (*) on the nontemplate strand ( Figure 3D). We observed higher efficiencies of NAD + capping in reactions performed using promoter derivatives having Y at template-strand position -1 than with those having R. Furthermore, within error, the capping efficiencies for promoter derivatives having Y or R at template-strand position -1 matched the capping efficiencies for homoduplex promoter derivatives ( Figure 3C-D), indicating that sequence information for NAD + capping with S. cerevisiae mtRNAP resides exclusively in the template strand.
We conclude that NCIN capping with NAD + by mtRNAP is determined by the sequence at, and immediately upstream of, the TSS (positions +1 and -1, respectively).
We further conclude that the sequence and strand preferences at positions +1 and -1 for NCIN capping with NAD + by mtRNAP match the sequence and strand preferences observed for bacterial RNAP ( Figure 3C-E) (17,21), suggesting that these sequence and strand preferences may be universal determinants of NCIN capping with NAD + for all RNAPs. Consistent with this hypothesis, we find that sequence preferences for NCIN capping with NAD + by bacteriophage T7 RNAP, another member of the single-subunit RNAP family, match the sequence preferences observed for S. cerevisiae mtRNAP and bacterial RNAP (Figures 3E and S4). Further consistent with this hypothesis, structural modeling suggests the basis for these sequence and strand preferences is universal: specifically, a requirement for template-strand +1T for base pairing to the NAD + adenine moiety, and a requirement for template strand -1Y for "pseudo" base pairing to the NAD + nicotinamide moiety (17,21). We selected for analysis two S. cerevisiae mitochondrial RNAs that previously had been detected as NAD + -capped: COX2 and 21S (10). We isolated S. cerevisiae total RNA and analyzed COX2 and 21S RNAs using the procedure described in the times higher than levels of NCIN-capping in exponentially growing E. coli (less than 1% to ~20% for NAD + -capping; not previously detected for NADH-capping) (12,17,21,41).

Detection and quantitation of NAD + -and NADH-capped mitochondrial
We performed analogous experiments analyzing RNAs produced by transcription from the human mitochondrial LSP promoter ( Figure 5B, top). We isolated and analyzed total RNA from HEK293T cells. We observed an NAD + -capped species comprising ~10% of the total LSP-derived RNA pool ( Figure 5B, top). The results establish that human mitochondrial RNAs undergo NAD + capping in cells and show that human mitochondrial RNAs undergo NAD + capping at 5' ends generated by transcription initiation (as opposed 5' ends generated by RNA processing).

Detection and quantitation of NAD + -and NADH-capped mitochondrial RNA in vivo: mtRNAPs serve as both sensors and actuators in coupling cellular metabolism to mitochondrial gene expression
Mitochondria are the primary locus of metabolism and energy transformation in the eukaryotic cell, serving as the venue for the tricarboxylic acid cycle (TCA) cycle and oxidative phosphorylation. The TCA cycle reduces NAD + to NADH and oxidative phosphorylation oxidizes NADH to NAD + . Our finding that mtRNAPs perform NCIN capping with NAD + and NADH at efficiencies that vary in a simple mass-action-  Figure   5B), thereby coupling cellular metabolism to mitochondrial transcription outputs. We suggest that mtRNAPs serve as sensors through their mass-action-dependence in selecting NAD + vs. NADH vs. ATP as initiating entity during transcription initiation, and serve as actuators by incorporating NAD + vs. NADH vs. ATP at the RNA 5' end during transcription initiation.

Discussion
Our results show that S. cerevisiae and human mtRNAPs cap RNA with NAD + and NADH (Figures 1 and S1), show that S. cerevisiae and human mtRNAPs cap RNA with NAD + and NADH more efficiently than bacterial and eukaryotic nuclear RNAPs ( Figures 2 and S2), and show that capping efficiency by mtRNAPs is determined by promoter sequence (Figure 3). Our results further show that the proportions of mitochondrial RNAs that are capped with NAD + and NADH are remarkably high--up to ~50% and up to ~40%, respectively (Figures 4 and 5)--and that these proportions change in response to cellular NAD + and NADH levels ( Figure 5).
We and others previously have shown that NCIN capping by cellular RNAPs has functional consequences (11,12,17). Our results here showing that S. cerevisiae and human mitochondrial RNAs are capped at substantially higher levels than nonmitochondrial RNAs--up to ~50% for analyzed S. cerevisiae mitochondrial RNAs and up to ~10% for analyzed human mitochondrial RNAs (Figures 4 and 5)--suggest that NCIN capping in mitochondria occurs at a higher efficiency, and has a higher importance, than NCIN capping in other cellular compartments. Four other considerations support this hypothesis. First, mtRNAPs are substantially more efficient at NAD + and NADH capping than bacterial and eukaryotic nuclear RNAPs ( Figure 2C). Second, levels of NAD + and NADH in mitochondria are substantially higher than levels in other cellular compartments (47, 48). Third, all S. cerevisiae and human mitochondrial promoters are +1A promoters (18 promoters in S. cerevisiae mitochondria; 2 promoters in human mitochondria) (49-51), in contrast to bacterial and eukaryotic nuclear RNAP promoters, for which approximately half are +1A promoters (52-56). Fourth, we observe capping with both NAD + and NADH for mitochondrial RNAs in vivo (Figures 4 and 5), whereas, to date, capping has been observed with only NAD + for non-mitochondrial RNAs in vivo, raising the possibility that, in mitochondria, but not in other cellular compartments, NAD + and NADH caps dictate different RNA fates and, correspondingly, different transcription outputs.
Our results showing that levels of NAD + and NADH capping by mtRNAP correlate with changes in intracellular levels of NAD + and NADH ( Figure 5    DNA templates and representative data are shown in Figure S2.    (21), and data for T7 RNAP is from Figure S4.
T7 RNAP was prepared from E. coli strain BL21 transformed with pAR1219 using culture and inductions procedures, SP-Sephadex, CM-Sephadex and DEAE-Sephacel chromatography as described in (61).
E. coli RNAP core enzyme was prepared from E. coli strain NiCo21(DE3) transformed with plasmid pIA900 (62) using culture and induction procedures, immobilized-metal-ion affinity chromatography on Ni-NTA agarose, and affinity chromatography on Heparin HP as in (62).
S. cerevisiae RNA polymerase II core enzyme (gift of Craig Kaplan) was prepared as described in (63).
E. coli NudC was prepared from E. coli strain NiCo21(DE3) transformed with plasmid pET NudC-His (17) using metal-ion chromatography and size-exclusion chromatography as in (12). RNA 5' pyrophosphohydrolase (RppH) and T4 polynucleotide kinase (PNK) were purchased from New England Biolabs (NEB). FastAP Thermosensitive Alkaline Phosphatase was purchased from Thermo Fisher Scientific. Molar concentrations of purified proteins were determined by light absorbance at 280 nm and the calculated respective molar extinction coefficients.

Oligodeoxyribonucleotides
Sequences of the oligodeoxyribonucleotides used in this work are provided in Table S1. All oligodeoxyribonucleotides were purchased from Integrated DNA Technologies (IDT) with standard desalting purification unless otherwise specified.
Linear in vitro transcription templates used for transcription assays shown in Figures 1, 2, 3, S1, S2, S3, and S4 were generated by mixing complementary equimolar amounts of nontemplate-and template-strand DNA in 10 mM Tris HCl pH 8.0, incubating the mixture at 95°C for 5 min, and cooling the mixture by 0.5°C per minute to 25°C.
Transcription templates used to generate in vitro RNA standards for Northern analysis (Figures 4 and 5) were generated by PCR. PCR reactions contained a mixture of 5 nM template oligo, 0.5 µM forward primer, 0.5 µM reverse primer, and Phusion HF Master Mix (Thermo Scientific). Reaction products were isolated using a Monarch PCR & DNA cleanup kit (NEB).

In vitro transcription assays
Assays performed with S. cerevisiae mtRNAP were based on procedures described in (34). Assays performed with human mtRNAP were based on procedures described in (59).
A portion of the recovered initial RNA products were mixed with either 10 U of RppH or 400 nM NudC and incubated at 37°C for 30 min. Reactions were stopped by addition of 10 µl RNA loading dye. Samples were analyzed by electrophoresis on 7.5 M urea, 1x TBE, 20% polyacrylamide gels (UreaGel System; National Diagnostics), followed by storage-phosphor imaging (Typhoon 9400 variable-mode imager; GE Life Science). was used to fit the data to the equation: y = (ax) / (b+x); where y is [NCINpC / (pppApC + NCINpC)], x is ([NCIN] / [ATP]), and a and b are regression parameters. The resulting fit yields the value of x for which y = 0.5. The relative efficiency (kcat/KM)NCIN / (kcat/KM)ATP is equal to 1/x. Detection and quantitation of NAD + -and NADH-capped mitochondrial RNA in vivo: isolation of total cellular RNA from S. cerevisiae For analysis of NAD + and NADH capping during respiration, S. cerevisiae strain 246.1.1 [(64); MATa ura3 trp1 leu2 his4; gift of Andrew Vershon, Rutgers University] was grown at 30°C in 25 ml YPEG (24 g Bacto-tryptone, 20 g Bacto-yeast extract, 30 mL ethanol, 3% glycerol per liter) in 125 ml flasks (Bellco) shaken at 220 rpm. When cell density reached an OD600 ~1.8 (approximately 24 hours) the cell suspension was centrifuged to collect cells (5 min, 10,000 g at 4°C), supernatants were removed, and cell pellets were resuspended in 0.8 mL RNA extraction buffer (0.5 mM NaOAc pH 5.5, 10 mM EDTA, 0.5% SDS).
To extract RNA, an equal volume of acid phenol:chloroform (5:1, pH 4.5; Thermo Fisher Scientific) was added to cells in resuspension buffer and mixed by vortexing for 10 s. Samples were incubated at 65°C for 5 min, -80°C for 5 min, then centrifuged (15 min, 21,000 g, 4°C) to separate the aqueous and organic phases. The aqueous phase was collected and acid phenol:chloroform extraction was performed two more times on this solution. RNA transcripts were recovered by ethanol precipitation and resuspended in RNase free H2O.

Detection and quantitation of NAD + -and NADH-capped mitochondrial RNA in vivo: isolation of total cellular RNA from human cells
Human embryonic kidney HEK293T cells (obtained from ATCC) were maintained under 5% CO2 at 37°C in DMEM medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Atlanta Biologicals), 100 units/ml penicillin, and 100 µg/ml streptomycin. HEK293T cells were seeded in 100 mm tissue-culture treated plates and grown for 72 h at 37°C or seeded in 100 mm tissue-culture treated plates, grown for 24 h at 37°C, treated with 5 nM FK866 (Sigma Aldrich), and grown for an additional 48 h at 37°C. Total cellular RNA was isolated with TRIzol Reagent according to the manufacture's protocol (Thermo Fisher Scientific).