show Abstracthide AbstractN6-methyladenosine (m6A) is the most abundant mRNA modification, primarily implicated in controlling mRNA stability. The distribution of m6A varies considerably between and within species, and genetic variants associated with differences in m6A levels between humans have been associated with disease. Yet, the determinants governing m6A variability are poorly understood: It is unclear whether it is driven by changes in genetic sequences ('cis') or cellular environments ('trans') and what its underlying mechanisms are. Here we dissect these determinants via interspecies hybrids in yeast and mammalian systems, allowing us to interrogate methylation at two distinguishable alleles in a shared environment. We find that m6A evolution is driven primarily in 'cis', and identify two mechanisms driving m6A changes: (1) Sequence variations leading to formation/disruption of m6A consensus motifs, and (2) Changes in local mRNA secondary structure, whereby RNA structuredness inhibits m6A formation. We demonstrate that secondary structure is causal, and that gain and loss of structure - even when driven by mutations distant from the modified position - are sufficient to abolish and acquire methylation, respectively. Using intra-species hybrids, massively parallel reporter assays and reanalyzing m6A-QTLs among 60 human individuals, we reveal the combined role of sequence and structure in shaping variability in m6A levels also across individuals from the same species. Finally, we demonstrate that differences in m6A levels between homologous genes lead to allele-specific changes in gene expression. Our findings thus define the determinants governing m6A evolution and diversity and characterize the consequences thereof on gene expression regulation. Overall design: We performed m6A-seq2 RNA sequencing of polyA-selected mRNA from multiple samples. For yeast, we sequenced Saccharomyces cerevisiae and Saccharomyces paradoxus yeast strains along with their F1 hybrid (IME4 KO and IME 4 WT), with 3 replicates for each strain (total 18 samples). For mammalian cells, we sequenced 3T3 mice cell line, BJ human cell line, JC001 cell line, JC002 cell line, and JC003 cell line, with 2 replicates for each (total 10 replicates). Additionally, we sequenced mouse embryonic stem cells originating from mating the two mouse strains house mouse and castaneous, 2 replicates.