In flowering plants the male gametes are embedded within the cytoplasm of the growing pollen tube, which transports them to the embryo sac where double fertilization occurs. Understanding the mechanisms and conditions by which sperm cells mature and take part in fertilization are crucial goals in the study of plant reproduction, epigenetic reprogramming and genomic imprinting. Analyses in the model plant Arabidopsis thalianawere hindered because no method to isolate sperm cells was available. We used Fluorescence-activated cell sorting (FACS) to isolate sperm cells and vegetative nucleus from Arabidopsis pollen, allowing gene expression analysis at a genome-wide level by microarrays. Comparative analysis of the sperm cell transcriptome with those of representative sporophytic tissues and of entire pollen grains showed that sperm cells have a distinct and diverse transcriptional profile, functionally skewed towards DNA repair, ubiquitin-mediated proteolysis and cell cycle progression. Moreover, analysis of the small RNA and DNA methylation pathways suggested that distinct mechanisms might be involved in regulating the epigenetic state of the paternal genome. Transposable elements (TEs) are suppressed by epigenetic silencing and small interfering RNAs (siRNAs), especially in the gametes that contribute with genetic material to the next generation. In wild-type Arabidopsis pollen, TEs are reactivated and transpose, but only in the vegetative nucleus, which does not provide DNA to the fertilized zygote. TE expression coincides with down-regulation of key elements involved in heterochromatin formation, and many TE-derived siRNAs. However, 21nt siRNAs from Athilaretro-transposons are generated and accumulate in pollen and sperm, suggesting that siRNA from TEs activated in the vegetative nucleus may target silencing in gametes. We profiled DNA viii methylation of the three different types of nuclei observed through pollen maturation, from the post meiotic precursor cell to the mature pollen grain in an attempt to understand epigenetic reprogramming during plant germline development. CG and CHG methylation is reduced in the vegetative cell comparing with that of sperm and microspore, but only at specific loci, which tend to mark TEs surrounding some noted imprinted genes. This indicates that symmetric methylation in the microspore is maintained during DNA replication in the first pollen division, and actively lost at imprinted loci. Intriguingly, CHH methylation, which is tied to RNA-directed DNA methylation (RdDM), is much lower in sperm nuclei than in the vegetative nucleus, overlapping with the accumulation of correspondent siRNAs in pollen. There is a particular subset of RC/Helitron and DNA/MuDR elements that are rich in siRNAs and CHH methylation while maintaining CG and CHG methylation levels, and a second subset of elements that are devoid of siRNAs and CHH methylation, and actively lose CG and CHG. These results suggest that in the VN, DNA demethylation occurs only at TE loci that are not actively targeted by RdDM, in a mechanism that might be involved in re-establishing imprinting marks. From plants to animals, there is contradictory evidence on the importance of microRNAs (miRNAs) during germ cell development. While most of the core proteins involved in the miRNA pathway in plants have been identified in the Arabidopsis sporophyte, there is very limited understanding about potentially distinct mechanisms of post-transcriptional regulation between different cell lineages. We provide a robust comparative analysis of miRNAs identified in sperm cells and pollen by deep sequencing, using molecular tools to deplete their function along germ cell specification and beyond fertilization. In addition, we present the identification of 25 potentially novel miRNAs processed in the male gametophyte and sperm cells, as well as enriched variations in the sequence length of known ix miRNAs that might indicate sub-functionalization by association with a putative germline-specific Argonaute complex. ARGONAUTE 5 (AGO5), by close homology to AGO1 and localizing preferentially to the sperm cell cytoplasm in mature pollen, may be part of such a complex. MicroRNA targets are not misregulated in ago5-4pollen, indicating that AGO5 is not part of the canonical miRNA pathway during pollen development. In alternative and similarly to what had been previously reported for AGO10, our data suggests that AGO5 could be involved in miRNA sequestering, thus preventing post-transcriptional gene silencing in the sperm cells. Moreover, AGO5 may be additionally involved in transcriptional gene silencing by RdDM, as some genes up-regulated in the mutant are flanked by TEs and repeat sequences targeted by RdDM