The evolution of gonochorism from hermaphroditism is linked with the formation of sex chromosomes, as well as the evolution of sex-biased and sex-specific gene expression to allow both sexes to reach their fitness optimum. There is evidence that sexual selection drives the evolution of male-biased gene expression in particular. However, previous research in this area in animals comes from either theoretical models or comparative studies of already old sex chromosomes. We therefore investigated changes in gene expression under three different selection regimes for the simultaneous hermaphrodite Macrostomum lignano subjected to sex-limited experimental evolution (i.e., selection for fitness via eggs, via sperm, or a control regime allowing both). After 21 and 22 generations of selection for male-specific or female-specific fitness, we characterized changes in whole-organism gene expression. We found that female-selected lines had changed the most in their gene expression. Although annotation for this species is limited, GO-term and KEGG pathway analysis suggests that metabolic changes (e.g., biosynthesis of amino acids and carbon metabolism) are an important adaptive component. As predicted, we found that expression of genes previously identified as testis-biased candidates tended to be downregulated in the female-selected lines. We did not find any significant expression differences for previously identified candidates of other sex-specific organs, but this may simply reflect that few transcripts have been characterized in this way. In conclusion, our experiment suggests that changes in testis-biased gene expression are important in the early evolution of sex chromosomes and gonochorism. Overall design: We carried out sex-limited experimental evolution in a hermaphrodite using a GFP marker as a proxy for sex. F lines were selected for female fitness (fecundity), M lines were selected for male fitness (fertilization success), and in the C lines half of the marked individuals were selected for female fitness and the other half for male fitness. There were 4 replicate lines within each selection treatment (i.e. F1, F2, F3, F4, M1, M2, M3, M4, C1, C2, C3, C4). RNAseq data was obtained from two pooled samples of 40 individuals each from all selection lines (one sample in generation 21 and the other in generation 22). We also seequenced 3 samples of the ancestral GFP stock. 3 samples were excluded prior to sequencing due to low amounts of RNA (C1gen22, F1gen21, and M3gen22), for a total of 24 samples.