RNA extraction and RNA-seq

Exosomal RNA was extracted using the SeraMir exosome RNA kit (SBI, Mountain View, CA, USA) according to the manufacturer's instructions. The quality and quantity of the RNA was determined using the Small RNA Chip and Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA). Samples with an RNA integrity number of >7.0 were eligible for RNA-seq. Twenty patients each in the recurrent and non-recurrent groups were matched by age, sex, race, and tumor stage. Five RNA samples from each group were pooled, and 8 were subjected to RNA-seq. The library was prepared using the NEBNext Small RNA Library Prep Set for Illumina, following the manufacturer's protocol (New England BioLabs, Ipswich, MA). The constructed libraries were quality checked using a bioanalyzer and pooled, followed by size selection in a range of 130 bp and 166 bp using a 3% agarose gel cassette on a BluePippin (Saga Science, Beverly, MA). RNA-seq was conducted on an Illumina HiSeq 2000 at the Science Park NGS facility. The quality of the sequencing data was analyzed by the bioinformatics team associated with the sequencing core. In brief, the reads were first trimmed by removing the adaptor sequences and then mapped to human miRNAs from miRBase Release 21 [32] using miRDeep2 (version 2.007) [33], which used the short read aligner Bowtie (version 1.0.0) [34] internally. The number of reads in each miRNA was enumerated by miRDeep2, and differences in expression between groups were statistically assessed by DeSeq (version 1.16.0) [35] and edgeR (version 3.6.1) [36]. A pathway analysis was conducted on the differentially expressed miRNAs with nominal significance using Ingenuity Pathway Analysis software (Ingenuity Systems, Redwood City, CA).