We have analyzed the structures and locations of small nuclear RNA (snRNA) precursors at various stages in their synthesis and maturation. In the nuclei of pulse-labeled Xenopus laevis oocytes, we detected snRNAs that were longer than their mature forms at their 3' ends by up to 10 nucleotides. Analysis of the 5' caps of these RNAs and pulse-chase experiments showed that these nuclear snRNAs were precursors of the cytoplasmic pre-snRNAs that have been observed in the past. Synthesis of pre-snRNAs was not abolished by wheat germ agglutinin, which inhibits export of the pre-snRNAs from the nucleus, indicating that synthesis of these RNAs is not obligatorily coupled to their export. Newly synthesized Ul RNAs could be exported from the nucleus regardless of the length of the 3' extension, but pre-Ul RNAs that were elongated at their 3' ends by more than about 10 nucleotides were poor substrates for trimming in the cytoplasm. The structure at the 3' end was critical for subsequent transport of the RNA back to the nucleus. This requirement ensures that truncated and incompletely processed Ul RNAs are excluded from the nucleus.Small nuclear RNAs (snRNAs) are essential components of the small nuclear ribonucleoprotein particles (snRNPs) involved in processing mRNA precursors (reviewed in references 6, 53, and 81). Synthesis and processing of snRNAs and their assembly into snRNPs involve many steps in both the nucleus and the cytoplasm (reviewed in references 55 and 65). Precursors of snRNAs Ul to U5 (pre-snRNAs) are transcribed by RNA polymerase II from families of multiple genes (reviewed in reference 15). The 3' ends of these presnRNAs are made in a promoter-specific manner in response to a 3'-end signal downstream from each RNA coding region (11,32,63). Initially, they have 7-methylguanosine (m7G) caps at their 5' ends (32, 49, 54, 80; this paper) and 1 to 10 extra nucleotides at their 3' ends compared with the mature snRNAs that accumulate in the nucleus (19,50,51).Although experiments with isolated nuclei have confirmed the nucleus as the site of their transcription (39,44,49), the pre-snRNAs have been detected in vivo only in the cytoplasm (18,19,30,50,51,94, 96). One possible explanation for the difficulty in finding pre-snRNAs in the nucleus might be leakage while the nuclei and cytoplasms were being separated (18, 50). Other explanations involve the coupling of pre-snRNA export to transcription (15). This situation contrasts with that of mRNA and tRNA export, which takes place only after processing of precursors has occurred in the nucleus (83; reviewed in references 1, 12, and 77). Several lines of evidence indicate that RNAs are exported from the nucleus through nuclear pores. For example, the lectin wheat germ agglutinin (WGA), which binds to N-acetylglucosamine residues found in the glycoproteins of nuclear pores, has been shown to inhibit the export of 5S RNA, tRNA, mRNA, and pre-snRNAs (4, 78, 79; E. Lund and J. E. Dahlberg, unpublished data; J. G. Koster and M. Zasloff, personal communication).Free prote...