N6-methyladenosine regulates RNA abundance of SARS-CoV-2

Supplementary Fig. S1. Identification of viral mA peaks using DAMS. a Proportion of reads distribution in positive-sense and negative-sense SARS-CoV-2 RNAs. b Schematics of peak calling and definition using DAMS for SARS-CoV-2. c Sequence motif of DMAS-mA-peaks in Chlorocebus sabaeus. d Pie chart showing the fraction of the annotation in three transcript segments of Chlorocebus sabaeus for DAMS-mA-peaks (left) and MACS2-mA-peaks (right). e Metagene profiles of mA ORF6 (186 nt) ORF7ab (498 nt) ORF8 (366 nt) N (1260 nt) 3’UTR (229 nt) ORF1ab (21290 nt) S (3822 nt) ORF3a (828 nt) E (228 nt) M (669 nt) 5’UTR (265 nt) b a

The double-stranded RNA oligonucleotides (short interfering RNA, siRNA) used for RNA interference were obtained from GenePharma. Lipofectamine RNAiMax  Biosystems, KK8541) following the manufacturer's instructions.  RNA (-) were respectively extracted using Samtools (version 1.9) 8 . To minimize the rate of false positives, only uniquely mapped reads with -q ≥ 20 were kept for the subsequent analysis for each sample.

DAMS (DEW-based Analysis for MeRIP-Seq, DEW = Differential Expressed
Window). The reference sequence (both Chlorocebus sabaeus and SARS-Cov-2) was first scanned using sliding windows of 25 nt. The coverage for each window was calculated for the MeRIP and input samples using Bedtools 9 . Using the R package edgeR 10 , windows with log 2 (fold change of MeRIP/input) > 2 and FDR < 0.05 were kept and then adjacent windows were merged. Only the merged windows with length more than 100 nt were finnally identified as m 6 A peaks. Specially, compared to SARS-Cov-2, the depth of reads from Chlorocebus sabaeu is relatively low in this work so that we adjusted the cutoff as log 2 (fold change of IP/input) > 2 and p < 0.05 for its m 6 A peak calling.

MACS2.
The host Vero cells m 6 A-enriched peaks were also identified using MACS2 (version 2.1.4) 11 with the corresponding input sample as control. MACS2 was used with parameters '--nomodel, --keep-dup all and -g 2.8e9'. Peaks with FDR value < 0.05 and located in 5'UTR, CDS and 3'UTR of Chlorocebus sabaeu mRNA were kept for the following analysis. Only intersections between the peaks called by two replicates were used as final set of peak calls.

Differential m 6 A peaks analysis.
Peaks identified in SARS-Cov-2 using edgeR for control samples were used for subsequent differential analysis. The coverage for each peak was calculated for the MeRIP and input samples, which with or without siMETTL3 treatment, using Bedtools. Every peak was assigned the metric 1 (M1), 11 representing the enrichment fold change of MeRIP over input sample. The influence of RNA abundance and peak length in this method were normalized using RPKM.
Differential m 6 A peaks between control and siMETTL3 group were calculated using metric 2 (M2). Peaks were identified as METTL3-dependent in SARS-Cov-2 with ∆ < -log 2 (1.5). Permutation analysis for RRACH. To well identified whether m 6 A peaks in SARS-Cov-2 were also RRACH dependent, the summit window of each peak was first defined by window with highest enrichment, which is calculated as log 2 (fold change of MeRIP/input) using edgeR. For each peak, the center of summit window with up-and downstream 50 nt flanking sequences were scaned for RRACH. Finally, 11 m 6 A peaks were validated in total 13 m 6 A peaks along SARS-Cov-2.
Then the permutation analysis was performed to validate whether the enriched RRACH for m 6 A peaks was randomly or not. For each time of permutation, the 13 peaks were first randomly shuffled along SARS-Cov-2 reference sequence and then the proportion of RRACH contained in random peaks were calculated. After 1,000 repeats, the observed propotions from the these data were contributed as empirical probability distribution, and only 74 simulated peak pools contained at least 11 RRACH peaks.
Counting and classifying reads for junction-spanning reads. The junction-spanning reads were categorized by the position of 5' and 3' site positions 13 . For each sample, 12 different junction-spanning types are estimated by the sum of junction-spanning reads.
The junction-spanning type is considered as non-random if it contained more than 1/10 6 of total junction-spanning reads.
Based on the location of its 5' and 3' site postions, the junction-spannings were then calssified into four groups, (a) TRS-L-dependent, canonical; (b) random 3' acceptor; (c) random 5' donor; (d) random inner junction (see Fig. 1i). the other junction-spannings were then classified as random inner junction. Especially, 3'UTR inner junctions were identified as the random inner junction whose 5' site position located in 3'UTR segment.
Expression of junction-spannings. The amounts of junction-spanning were first normalized by the ERCC for each sample (used in Supplementary Fig S2c, S3a and S3b). To well compared the propotions of different junction-spannings in each samples, the expression level of junction-spannings were valued as RPM (reads per million) by normalized with the sum of total junction reads (used in Fig 1h and 1k and Supplementary Fig. S3d).
Statistical analysis. All statistical analyses of qPCR assay were performed two or three independent experimental replicates. Student's two-sided unpaired t-test was used for statistical comparisons and data were shown as mean ± S.E.M.