Sequence quality and alignment
We assessed the quality of the sequences using FastQC (Andrews 2010), trimming of adaptors and low-quality reads was done using the RRBS default parameters (function: –rrbs) in TrimGalore! (Krueger 2016). Reads were aligned to the Kryptolebias marmoratus reference genome (NCBI ASM164957v1) (Rhee et al. 2017) prior to in-silico bisulphite conversion using Bismark v0.17.0 (Krueger & Andrews 2011), which was also used to perform cytosine methylation calls. We only considered methylation within CpG context for the downstream analysis (Feng et al. 2010) and included CpGs with a minimum coverage of ≥10 reads in each sample across the 30 individuals sequenced. Individuals were grouped into generations (parents/offspring) and environments (own/parental) (Table S2). Mapped reads were processed and compared using the R package methylKit v. 1. 10 (Akalin et al. 2012). All analyses were conducted on a local server running NEBC Bio-Linux 8.