To date, all components of the ASH1
mRNP including the core factors connecting mRNA to the myosin motor have been identified 
. However, largely due to the lack of quantitative interaction studies, the molecular mechanisms leading to specific mRNA recognition and mRNP assembly remain ambiguous. To our knowledge, we performed the first in vitro reconstitution of the core of an mRNA-transport complex. We followed the path of mRNP assembly from the nucleus to the cytoplasm and assessed binding affinities, specificities, and synergisms in complex assembly. Key findings were confirmed by complementing in vitro and in vivo experiments.
Since She2p binds ASH1
mRNA co-transcriptionally and escorts it until it is anchored at the bud tip 
, it had been assumed that this protein is responsible for the specific incorporation of zip code containing mRNAs into localizing mRNPs. We found that neither She2p nor Puf6p, which is the other ASH1
-E3 element-interacting protein present in the nucleus, bind zip-code RNAs with high specificity (). Because in a previous study a more specific recognition of ASH1
-E3 RNA by Puf6p had been suggested 
, we performed additional control experiments to ensure that Puf6p is folded and active in our experiments (Figure S1
). Whereas RNA-binding assays with unlabeled competitors are virtually identical in both studies, we also observed strong Puf6p binding to unrelated control RNAs in standard EMSAs. Although we arrive at a somewhat different estimation of the binding specificity of Puf6p, it seems clear from both studies that RNA binding by Puf6p is less specific than by other Pumilio/FBP family members 
. It might be that Puf6p requires an additional co-factor to achieve the specificity reported for its in vivo function.
Previous immunoprecipitation experiments with yeast extracts showed that the Puf6p:She2p complex is RNase-resistant, suggesting that both proteins might interact directly 
. With recombinant proteins, we recapitulated an RNA-mediated interaction of Puf6p and She2p, but not a direct protein-protein binding (). These results suggest that ASH1
mRNPs are rather RNase-insensitive, stable particles that allow for co-purification of both factors even after RNase treatment.
When performing in vitro RNA-binding assays with Puf6p and She2p, we also observed that their combination failed to bind synergistically or with higher specificity to zip-code RNAs (). Together, these experiments suggest that a complex consisting of ASH1 mRNA, She2p, and Puf6p is able to assemble in the nucleus with moderate specificity (). Although unlikely, we cannot exclude the requirement of other factors not yet implicated in the assembly of specific nuclear mRNPs.
Model for the sequential assembly of ASH1-transport complexes in yeast.
In the cytoplasm, She3p is thought to link the RNA:She2p complex to the myosin motor Myo4p 
. The prevailing model proposes that besides Myo4p, the translationally repressed ASH1
-E3 complex harbors She2p, She3p, and Puf6p. To date, it had not been shown that such a complex really assembles on the RNA. In our in vitro assays, we reconstituted this Puf6p-containing complex in an ASH1
-E3 RNA-dependent manner (Figure S4
). Whether Khd1p behaves in a similar manner remains to be shown.
A previous study reported that E. coli
–expressed, crude-purified GST-She3p causes a supershift in an EMSA with GST-She2p and ASH1
-E3 RNA 
. From this single experiment, it remained unclear whether this supershift is based on protein-protein or protein-RNA interactions and whether the presence of She3p has implications on RNA-binding affinity and specificity. Diverging models have been proposed on the function of She3p 
, including a scenario where She2p is not required for cytoplasmic ASH1
-mRNA transport 
When assessing molecular interactions of She3p, we first identified a direct and specific binding to She2p with a Kd of 1.6 µM (). This interaction suffices to explain the previously described supershift by She3p in EMSAs 
. It is also consistent with a previous report where She3p was immunoprecipitated with a She2p mutant that lacks RNA-binding capacity 
. However, the modest affinity and transient nature of the interaction between She2p and She3p and the low RNA-specificity of She2p appears insufficient to explain the specific localization of ASH1
mRNA observed in vivo.
More surprisingly, we identified the myosin adapter She3p as a new, rather unspecific RNA-binding protein () with affinities comparable to She2p and Puf6p. Database searches failed to identify any known RNA-binding motif in She3p. However, the most important finding for a mechanistic understanding of specific cargo recognition and mRNP assembly was our observation that She2p and She3p together form a highly specific ternary complex with all zip-code elements of localizing mRNAs tested in this study. EMSAs demonstrated that the She2p:She3p complex has at least 60-fold higher Kds for zip-code RNAs over unrelated RNA stem-loops ( and S5F–H
). In addition, this RNA interaction stabilizes the rather weak binary She2p:She3p complex () to a similar extent. Since for the control HIV-1 TAR RNA no binding was observed even at the highest experimental concentration (), the true difference in Kd might be even higher. These findings were confirmed by a total of 12 mutations in She2p and She3p, in which for seven of them an impaired complex formation was observed (). Two of these She2p mutants were also tested in vivo, where they showed a total loss of RNA localization (). Interestingly, four of the mutations in She2p and She3p with a defect in synergistic complex formation also showed defects in both binary interactions, RNA and protein binding. This observation suggests that both interactions are spatially and mechanistically intertwined to allow for the synergism described in this study. This conclusion is further supported by UV cross-linking experiments, which demonstrate that She2p and She3p both directly bind to RNA in the ternary complex (, , and S7
No known RNA-binding motif can be identified in She3p. We nevertheless found a C-terminal fragment of 92 amino acids to be sufficient for synergistic RNA binding with She2p. Within this 92 amino acids long fragment, we mapped residues that are of functional importance for the synergism. Our subsequent UV cross-linking/mass-spectrometry experiment with She2p, RNA, and the C-terminal fragment of She3p confirmed that a functionally important subfragment of 20 residues is indeed involved in RNA binding ( and S7
). These cross-linking experiments also showed that the helix E of She2p directly interacts with RNA. The latter finding confirms our observations on the functional importance of the helix E and assigns a direct molecular function, i.e. RNA binding, to this protein region. A surface plot of all She2p-surface regions that are important for RNA binding shows a large continuous RNA-interaction surface (Figure S12
In a recent study, a She3p-dependent ASH1
mRNA-transport system was reported in Candida albicans 
. Although in this yeast species no clear She2p homolog could be identified, ASH1
mRNA is transported in a fashion similar to S. cerevisiae
. Thus, the question arose of how cargo binding is achieved. Our identification of S. cerevisiae
She3p as an RNA-binding protein suggests that C. albicans
She3p could mediate at least part of the RNA binding for mRNP assembly and transport. A sequence alignment of She3p from different yeast species reveals that the C-terminal half of this protein shows great differences between species with and without She2p in their genomes (Figure S13
). As shown in this study, the C-terminal part of S. cerevisiae
She3p binds to She2p () 
as well as to RNA () and is required for synergistic RNA binding with She2p (). It therefore appears likely that the She2p-lacking species have optimized the C-terminal She3p sequence for an interaction with a different RNA-binding protein or even for a more specific RNA-binding by She3p itself.
Previous studies suggested that She3p acts only in the cytoplasm 
. However, this assumption had not been rigorously tested before. We used a nuclear export mutant to show that She3p indeed does not shuttle into the nucleus (). To scrutinize whether a small sub-fraction of She3p might play a role in the nucleus, we also analyzed the recently reported co-transcriptional recruitment of She2p to chromatin by ChIP experiments 
. Since we observed no significant difference of She2p occupancy in wild-type and Δshe3
strains, a nuclear role of She3p could be further excluded ().
Shen et al. also reported a reduction of She2p chromatin binding in ChIP experiments after RNase treatment selectively at open reading frames (ORFs) of localizing mRNAs. The authors concluded that part of the chromatin-associated She2p interacts selectively with localizing mRNAs already during transcription. However, we could not confirm this observation. Although we also detected a reduction of She2p-dependent enrichment after RNase treatment, this effect was observed for all transcripts (). Thus, our ChIP experiments suggest rather unspecific She2p association with nascent transcripts. They further indicate that She3p does not play a functional role in the nucleus.
In summary, our and previous data suggest the following model: First, She2p binds co-transcriptionally to RNA polymerase II and to nascent transcripts (). After transcription, nuclear mRNAs are bound by She2p and Puf6p with only limited specificity (), followed by a nuclear export of both proteins together with mRNAs (). In the cytoplasm, She2p and localizing mRNAs form a highly specific co-complex with myosin-bound She3p (). This transport complex mediates the translocation of cargo mRNAs to the bud cell (), where after anchoring at the bud tip translation is activated.
On one hand, She2p and She3p function together exclusively in the cytoplasm to select zip-code RNAs. On the other hand, the mRNA cargo itself substantially stabilizes the She2p:She3p interaction. Because this interaction brings together the She2p-dependent pre-mRNP with the cytoplasmic motor complex, we conclude that it is the mRNA cargo itself that triggers joining of all components into the mature transport complex. This interpretation is fully consistent with the observation that RNase-treated ASH1
-mRNPs do not have motile activity in vitro 
. We propose that coupling of specific mRNA recognition and assembly of stable transport complexes constitutes a critical quality control step to ensure that only target mRNAs are transported.
Previous publications reported only moderate in vitro selectivity for localizing mRNAs in yeast and Drosophila
(e.g. between 3- and 7-fold higher Kd for localizing RNAs) 
. Thus it remained ambiguous what difference in affinity to localizing and non-localizing RNAs might be required for specific mRNA transport and how highly specific mRNP assembly is achieved in vivo. In comparison, the ternary complex formation described in this study shows an unprecedented selectivity for zip code containing RNAs. In our understanding, this observation gives a more realistic example for the cargo specificity required for mRNA localization. It also demonstrates that co-complexes of transport factors might play a much more important role in the recognition of transcripts than previously assumed. Last but not least, our finding that the cargo RNA itself triggers the incorporation of all protein-core factors into one mature transport complex provides a new mechanistic paradigm for the assembly of RNA-localization complexes.
An important question arising from this study is whether synergistic binding to RNA cargo is a more general feature in eukaryotes. For instance, during the oocyte-to-embryo transition of Drosophila
development, the RNA-binding protein Staufen is involved in the localization of bicoid
mRNA. In vitro, Staufen yielded strong binding to specific as well as to control RNAs with extensive secondary structures 
. In contrast, in vivo injection experiments in Drosophila
embryos showed that Staufen-containing mRNPs only form and localize efficiently when its native target, the bicoid
3′UTR, is injected 
. Thus, it might well be that those Drosophila
complexes also require mRNP assembly for specific mRNA binding and localization.
In the past, in vitro reconstitution of molecular assemblies has been very successful to provide new insights into biological processes as diverse as transcription and membrane fusion 
. By showing that specific mRNA recognition and assembly of stable cytoplasmic transport complexes is coupled, we demonstrate that in vitro reconstitution is also well suited to advance our mechanistic understanding of mRNA-transport.