down-regulated (Fig. 3, A and B). Moreover,
genes with m^6 A-marked upstream carRNAs
attained higher increases in transcription rate
than those with non-m^6 A–marked upstream
carRNAs.Thesameistruewhensortinggenes
with their precursor mRNAs (pre-mRNAs) not
subjected to m^6 A methylation (Fig. 3, E to F,
and fig. S8, C to E), which indicates that the
reduced m^6 A methylation of these carRNAs
uponMettl3KO activates the transcription of
downstream genes.
Notably, we found that all m^6 A-dependent
genes that showed reduced upstream carRNA
methylation uponMettl3depletion (~6584
genes) exhibited increased transcription rates
(Fig. 3G). We identified a subset of these genes
that demonstrated transcription rate differ-
ences >1 uponMettl3KO (fig. S9A) and found
that these genes are mainly involved in tran-
scription regulation, chromatin modification,
and stem cell population maintenance (fig.
S9B). Hence, the reduced m^6 A methylation
of carRNAs not only promotes downstream
transcription but may activate genes involved
in chromatin opening, initiating a positive
feedback loop. We further analyzedPrdm9,
Kmt2d(encoding two H3K4me3 methyltrans-
ferases),Esrrb,andRanbp17(related with dif-
ferentiation), all of which possess upstream
carRNAs with reduced m^6 A level uponMettl3
KO (fig. S10). Consistently, the half lifetime of
these carRNAs and the transcription rate of
their downstream genes both increased upon
Mettl3KO, and these changes could be res-
cued by WT but not mutant METTL3 (fig. S11).
The interactions between super-enhancers
and their target genes are known to be affected
by transcription of exosome-regulated transcripts
in mESCs ( 23 ). We examined super-enhancers
and found that ~80% of super-enhancer RNAs
(seRNAs) contain m^6 Apeaks(Fig.3Handfig.
S12A). The m^6 A methylation level of seRNAs
decreased (fig. S12, B and C) and the m^6 A-
marked seRNAs showed a greater increase in
abundances compared with non-m^6 A–marked
ones uponMettl3KO (Fig. 3I). seRNAs that
showed reduced m^6 A level uponMettl3KO
were associated with increased transcriptionrates at downstream genes (Fig. 3J); genes
with transcription rate differences larger than
one were mainly involved in transcription reg-
ulation, chromatin modification, and stem cell
maintenance (fig. S12, D and E), consistent
with results obtained from other carRNAs
(fig. S9). Moreover, we found that genes reg-
ulated by m^6 A-marked upstream seRNAs tended
to exhibit greater increases in transcription rate
than those regulated by m^6 A-marked upstream
typical eRNAs uponMettl3KO (fig. S12F).
We next investigated chromatin state changes
affected by altered carRNA methylation. We
performed chromatin immunoprecipitation
sequencing (ChIP-seq) and observed global
increases of these two active marks, H3K4me3
and H3K27ac, uponMettl3KO(fig.S13,Aand
B), consistent with the Western blot results
(fig. S1H). Moreover, genes with m^6 A-marked
upstream carRNAs showed greater increases
in H3K4me3 and H3K27ac than genes with
non-m^6 A upstream carRNAs inMettl3KO
mESCs (Fig. 4A). Likewise,Mettl3KO mESCs
showed obvious increases in both marks atLiuet al.,Science 367 , 580–586 (2020) 31 January 2020 3of6
Fig. 2. Transcript turnover
of carRNAs is regulated by
m^6 A.(A) LC-MS/MS
quantification of the m^6 A/A
ratio in nonribosomal
(non-Rib) caRNAs (including
pre-mRNA) extracted from
WT orMettl3−/−mESCs.
n= 3 biological replicates;
error bars indicate means ±
SEM. (B)m^6 A level changes
on carRNAs were quantified
through normalizing m^6 A
sequencing results with
spike-in between WT and
Mettl3KO mESCs.n=2
biological replicates.
(C) carRNAs were divided
into methylated (m^6 A) or
nonmethylated (non-m^6 A)
groups. The boxplot shows
greater increases in
transcript abundance fold
changes of the m^6 A group
versus the non-m^6 A group upon
Mettl3KO over WT mESCs.
For (A) and (C),Pvalues
were determined by two-
tailedttest. (D) Cumulative
distribution and boxplots
(inside) of nuclear carRNA
half lifetime changes in CKO
Ythdc1and control mESCs.
(E) Cumulative distributions
and boxplots (inside) of
the half lifetime changes of carRNAs uponYthdc1CKO. carRNAs were divided into methylated (m^6 A) or nonmethylated (non-m^6 A) groups.
Depletion of YTHDC1 led to greater half lifetime increases of m^6 A-marked carRNAs than non-m^6 A–marked ones. For (D) and (E),Pvalues were
calculated by a nonparametric Wilcoxon-Mann-Whitney test. h, hours.
p = 2.09e-5p = 2.24e-50.00.10.2m6 A/A in non-Rib caRNA (%)WT
Mettl3-/--1
Mettl3-/--2Half lifetime
log 2 (Ythdc1 CKO/Control)−2−1012−2−1012−2−1012eRNA paRNA (-) RepeatsMettl3-/--1
Mettl3-/--2
Mettl3-/--1
Mettl3-/--2
Mettl3-/--1
Mettl3-/--2Transcript abundance logFC 2p = 1.01e-08
p = 1.19e-60
p = 2.13e-46
p = 1.37e-44
p = 1.69e-21
p = 0non−m^6 A
m^6 AeRNApaRNA (-)RepeatsAB DCE0.000.250.500.751.000 25 50 75 100Cumulative fractionp = 001000.000.250.500.751.000 25 50 75 100Cumulative fraction0501000.000.250.500.751.000 25 50 75 100
Half lifetime (h)
in nuclear fractionCumulative fraction05025paRNA (-)RepeatseRNAControl
Ythdc1 CKOp = 0p = 0non-m^6 A
m^6 A−20240.000.250.500.751.00010Cumulative fractionp = 1.7e−300240.000.250.500.751.00−10 0 10Cumulative fractionp = 3.5e−120240.000.250.500.751.00−5 0 5 10Cumulative fractionp = 1.9e−700.00.30.60.91.2paRNA (+)eRNA Repeats
paRNA (-)WT
Mettl3-/--1
Mettl3-/--2carRNA m6 A level−5 5 15RESEARCH | REPORT