and ZAM transposons and is the primary, if not the sole, source of rasiRNAs against these selfish elements. Most rasiRNAs correspond to the anti- sense strand of transposons and can therefore bind to, and presumably destroy, the RNA transcripts of transposons 1,8 . These antisense rasiRNAs bind to Piwi and Aubergine. The two new papers 1,2 report that the small RNAs bound to Ago3 are nearly all of the sense orientation. Of the 353 Ago3-associated rasiRNAs identi- fied by Siomi’s group 2 , the first 10 nucleotides of 16 sequences could be paired with rasiRNAs bound to Aubergine. And more than 11,200 (48%) of the Ago3-associated sense rasiRNAs identified by Hannon’s group 1 could form an offset couple with at least one antisense rasiRNA. A 10-nucleotide offset between the begin- ning of a small RNA guide and a second RNA molecule has a special meaning. Argonaute proteins cut target RNAs by measuring 10 nucleotides from the beginning of their RNA guide to the site of cleavage on their target. Such a pairing scheme suggests that the starting nucleotide of each antisense rasiRNA is defined by a cut that is guided by a corresponding sense rasiRNA. Reinforcing this view, nearly all the antisense rasiRNAs begin with the nucleotide U, whereas the sense rasiRNAs show no bias for beginning with U, A, C or G. Instead, the tenth nucleotide of the sense rasiRNAs was almost always A, which would allow it to pair with the first nucleotide — U — of an Aubergine- or Piwi-bound antisense rasiRNA (Fig. 1b). Imagine, then, that a mother fly protects her offspring by providing her developing eggs with some of her Aubergine- and Piwi- bound antisense rasiRNAs. These could then generate sense rasiRNAs by cleaving the RNA transcripts of transposons, thereby simultane- ously silencing them and initiating a cycle of rasiRNA amplification. The sense rasiRNAs bound to Ago3 would then cleave the long, antisense transcripts produced by master reg- ulatory loci such as flamenco, producing new antisense rasiRNAs that would bind to Auber- gine or Piwi. If antisense transcripts from mas- ter regulatory loci are generally more abundant than sense transcripts from transposons — a Figure 1 | piRNA-mediated silencing of transposons. a, A few trigger loci generate piRNAs, which target transposons (yellow) elsewhere in the genome and prevent their transcription. b, Complementary binding of Aubergine- and Ago3-associated piRNA sequences results in their amplification, ensuring efficient silencing of transposons. piRNA cluster 1 2 3 4 5 6 7 8 9 10 25 1 2 3 4 5 6 7 8 9 10 23 3’ 3’ 5’ Ago3-associated piRNA Aubergine-associated piRNA 5’ a b A U reasonable assumption as transposons are nor- mally silenced — then the pool of rasiRNAs would, as observed, be disproportionately antisense. Of course, the model proposed by these authors 1,2 only explains how the start of each rasiRNA is defined. How the 3ʹ end of the rasiRNA is made remains to be discovered. And what of piRNAs in humans? piRNAs were recently discovered in immature mouse, rat and human sperm cells 10 , and in zebrafish testes and ovaries 11 , although they were mostly not associated with transposon sequences. What piRNAs do in mammalian sperm is unknown, but, like fly piRNAs, they derive from large genomic clusters. And like fly piRNAs 8 , they do not seem to be made by Dicer. Perhaps all piRNAs are made and amplified by reciprocal cycles of Piwi-catalysed slicing of sense and antisense transcripts. Stay tuned for further detailed sequence analyses. ■ Phillip D. Zamore is in the Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School, Worcester, Massachusetts 01605, USA. e-mail: phillip.zamore@umassmed.edu 1. Brennecke, J. et al. Cell 128, 1089–1103 (2007). 2. Gunawardane, L. S. et al. Science 315, 1587–1590 (2007). 3. Fire, A. et al. Nature 391, 806–811 (1998). 4. Hamilton, A. J. & Baulcombe, D. C. Science 286, 950–952 (1999). 5. Ketting, R. F., Haverkamp, T. H., van Luenen, H. G. & Plasterk, R. H. Cell 99, 133–141 (1999). 6. Tabara, H. et al. Cell 99, 123–132 (1999). 7. Djikeng, A., Shi, H., Tschudi, C. & Ullu, E. RNA 7, 1522–1530 (2001). 8. Vagin, V. V. et al. Science 313, 320–324 (2006). 9. Prud’homme, N., Gans, M., Masson, M., Terzian, C. & Bucheton, A. Genetics 139, 697–711 (1995). 10. Kim, V. N. Genes Dev. 20, 1993–1997 (2006). 11. Houwing, S. et al. Cell 129, 69–82 (2007). NEUROBIOLOGY Feeling right about doing right Deborah Talmi and Chris Frith Reason and emotion come into conflict in making all kinds of judgements. Results of work with brain-damaged patients constitute one line of evidence that the emotional component is not to be dismissed. In resolving moral dilemmas, should emotion be our guide? This is a question prompted by various research avenues, including work described in the paper by Koenigs et al. 1 on page 908 of this issue. In a typical moral dilemma, we have to choose between the lesser of two evils. Caus- ing the death of one person is bad, but causing the death of five people is even worse. So, if you are on a runaway trolley with no other options, many people say that it is better to switch to the left fork in the track, resulting in the death of one person, than to carry on along the right fork and kill five. But what if there was no fork in the track and the only way to stop the trolley killing five people was to throw a large person, who happens to be standing next to you, under the wheels? From a utilitarian point of view the dilemma is the same: should we sacrifice one person for the sake of five? But, given this ver- sion of the dilemma, most people will choose not to throw their companion to his death. Why the difference? There is increasing evidence that there is a strong emotional component to our moral intuitions, and that this determines, to a large degree, how we make moral judgements 2 . Thus the benefit from sacrificing a single life for the greater good must be pitted against the emotional aversion associated with the taking of life, particularly when we are face-to-face with our victim. Measurement of brain activity while people are presented with these dilem- mas confirm this intuition: the moral dilemma involving throwing our companion onto the track elicits more activity in emotion-process- ing regions of the brain than the standard run- away-trolley problem (see ref. 3 for a review). The implication of these ideas is that people with impaired emotional responses will have altered moral intuitions. Koenigs and his col- leagues 1 have tested this hypothesis with a group of patients with damage to part of the brain called the ventral medial prefrontal cor- tex (VMPFC). As is typical after such damage, the autonomic nervous system in these patients showed reduced responses to emotionally charged pictures and, according to their spouses, the patients showed reduced feelings of empa- thy and guilt. When confronted with moral 865 NATURE|Vol 446|19 April 2007 NEWS & VIEWS