Differential cytoplasmic mRNA localisation adjusts pair-rule transcription factor activity to cytoarchitecture in dipteran evolution

Many cell types use cytoplasmic mRNA localisation to distribute protein products within cells (Kloc et al.,2002). In neurons, various transcripts are selectively sorted to different regions of the cytoplasm (Job and Eberwine, 2001), and in motile fibroblasts localisation ofβ -actin mRNA to the leading edge contributes to remodelling of the cytoskeleton (Kislauskis et al.,1997; Shestakova and Singer,2001). Localisation can also be coupled to cell divisions in which protein determinants must be segregated asymmetrically(Long et al., 1997; Takizawa et al., 1997).

In the syncytial blastoderm embryo of the fruit fly Drosophila melanogaster, transcripts of the pair-rule gene class are localised asymmetrically to the apical side of a layer of peripheral nuclei(Hafen et al., 1984; Ingham et al., 1985; Kilchherr et al., 1986; Macdonald et al., 1986). Pair-rule transcripts encode transcription factors that are expressed in partially overlapping sets of seven circumferential stripes, and act in combination to establish segmental organisation(Pankratz and Jäckle,1993). The segment polarity gene wingless (wg),which encodes an extracellular signalling molecule of the WNT family, also has its transcripts localised apically. Localisation of wg mRNA augments Wg signalling activity, although it is not yet clear by what mechanism this is achieved (Simmonds et al.,2001).

A shared machinery localises pair-rule and wg transcripts during embryogenesis (Wilkie and Davis,2001; Bullock and Ish-Horowicz,2001), and also translocates maternal mRNAs from their sites of synthesis in the nurse cells into the early oocyte(Bullock and Ish-Horowicz,2001). The localisation machinery involves active transport towards the minus-ends of microtubules by the dynein/dynactin motor complex(Wilkie and Davis, 2001). Transport also depends on the proteins Egalitarian (Egl) and Bicaudal-D (BicD)(Bullock and Ish-Horowicz,2001). The machinery may be used widely because BicD and dynein components are highly conserved in metazoans and Egl homologues are found in the nematode C. elegans.

All the transcript cargoes studied to date have been shown to contain localisation signals in their 3′-untranslated regions (3′-UTRs),which must direct interaction with the transport machinery. RNA recognition is dependent on secondary structure - localisation signals in different transcripts consist of double-stranded stem-loops that share no overt primary sequence similarity - although higher-order RNA folding may also be significant (Macdonald and Kerr,1998; Bullock et al.,2003).

Although the mechanisms of pair-rule mRNA transport in the blastoderm embryo are emerging, the developmental and evolutionary significance of this process remains unclear. Pair-rule patterning is used in many insects, but RNA signals for the dynein machinery have only been studied within the genus Drosophila, where signals in the pair-rule transcript hairy(h) (Bullock et al.,2003), as well as those of wg(Simmonds et al., 2001) and the maternal transcript bicoid(Macdonald, 1990), are functionally conserved. The conservation of h localisation in drosophilids is not surprising, however, because these flies share very similar blastoderm types. By contrast, blastoderm morphologies can differ drastically in less closely related dipteran taxa(Anderson, 1972).

In this study, we have analysed transcript localisation of a large number of newly identified homologues of the pair-rule genes even-skipped(eve) and h, and of wg in multiple families of the insect order Diptera (true flies) by in situ hybridisation to endogenous transcripts and by injection of fluorescently labelled transcripts into Drosophila embryos. We show that Egl-dependent localisation signals are conserved in eve and h transcripts over 145 million years of evolution in higher (cyclorrhaphan) flies, indicating that this process is functionally significant, but absent in some, but not all, branches of lower Diptera. By contrast, wg transcript localisation appears to be conserved throughout Diptera. The phylogenetic occurrence of pair-rule transcript localisation suggests a selective advantage of this trait in species with a thickened peripheral cytoplasm and apically residing blastoderm nuclei. Consistent with this, we find that, in Drosophila,localisation of pair-rule mRNAs targets their proteins apically, in close proximity to the nuclei, and that interfering with localisation lowers the activity of pair-rule genes. We provide evidence that RNA localisation augments levels of protein within the nucleus and propose that, by affecting perinuclear translation, this mechanism may be used in a wide variety of organisms to modulate the activity of nuclear factors.

Fly cultures and egg collections of Megaselia abdita (Phoridae;scuttle or humpbacked flies), Coboldia fuscipes (Scatopsidae;scavenger flies) and Clogmia albipunctata (Psychodidae; moth flies)have been described previously (Rohr et al., 1999; Stauber et al.,2002). Empis livida (Empididae; dance flies), Haematopota pluvialis (Tabanidae; horse flies) and Platypeza consobrina (Platypezidae; flat-footed flies) were collected in the surroundings of Göttingen (Germany). Embryos of Episyrphus balteatus (Syrphidae; hover flies) were a gift from Peter Hondelmann(University of Hannover, Germany). Anopheles gambiae (Culicidae;African malaria mosquito, Suakoko strain) genomic DNA was a gift from Hans-Michael Müller (European Molecular Biology Laboratory, Germany). Throughout the text we refer to Cyclorrhapha as `higher Diptera' (including Drosophila, Episyrphus, Megaselia and Platypeza). The term`lower Diptera' is used for the paraphyletic assemblages of orthorrhaphous Brachycera (including Empis and Haematopota) and Nematocera(including Coboldia, Clogmia and Anopheles).

Wild-type Drosophila embryos were of the Oregon-R strain. egl^WU50, eve^1.27, ftz¹³, hⁱ²², kni¹, Kr¹ and wg^CX4 are reported to be strong loss-of-function or null alleles. egl^3e is a partial loss-of-function allele that compromises the interaction of the protein with dynein light chain (Navarro et al., 2004). For Fig. 5B, larvae heterozygous for the pair-rule gene mutation were distinguished from wild types using balancer chromosomes marked with green fluorescent protein (GFP) driven from the actin promoter(Reichhart and Ferrandon,1998). Embryos of similar zygotic genotype but different maternal origin were generated from reciprocal crosses (i.e. heterozygous pair-rule mutant males were mated to egl mutant females and vice versa). Larval cuticle preparations were performed as described(Wieschaus and Nüsslein-Volhard,1986).

3′ regions of Anopheles-eve, -ftz, -h,-odd and -wg homologues were PCR amplified with specific primers (based on sequences from the Anopheles genome project) on genomic DNA of Anopheles gambiae strain Suakoko, a gift from Hans-Michael Müller (EMBL, Heidelberg). Partial sequences of the other eve, h and wg homologues were amplified by PCR on genomic DNA with the degenerate primer pairs GITAYMGIACIRCITTYACNMGNGA/TANACNGCNGCRTANGGCCANGC (eve),AAYAARCCIATHATGGARAARMGNMG/GYTTIACIGTYWTYTCIARDATRTCNGC (h) and GARTGYAARTGYCAYGGNATG/RTAICCICKICCRCARCACAT (wg). For Clogmia-eve1, Platypeza-eve and Platypeza-h, phages were isolated from genomic Lambda FIX II (Stratagene) phage libraries (S. Lemke and U.S.-O., unpublished) and used as templates for PCR-amplification of 3′-regions. 3′-UTRs of all other homologues were amplified by PCR on cDNA prepared with Marathon or SMART RACE cDNA Amplification Kit(Clontech). Table S1lists details of clones and contains information on templates, primers and products.

In situ hybridisation was carried out essentially as described(Lehmann and Tautz, 1994; Stauber et al., 2002) using NBT/BCIP and Fast Red (Roche) for colourimetric and fluorescent detection of transcripts, respectively. Nuclei were stained by a 2 hours room temperature incubation in 5 μg/ml of Alexa 660-wheat germ agglutinin (WGA; Molecular Probes) in PBS. For immunostaining, we used mouse anti-Hairy (S. M. Pinchin, unpublished) and guinea pig anti-Run(Kosman et al., 1998)polyclonal primary antibodies and Alexa-488 or Alexa 594-conjugated secondary antibodies (Molecular Probes).

Fluorescent, capped mRNAs incorporating Alexa-488- (Molecular Probes), Cy3-or Cy5-UTP (Perkin Elmer) were synthesised as described(Bullock and Ish-Horowicz,2001). Briefly, for each transcript, 250 ng/μl solutions were injected into nuclear cycle 14 blastoderm embryos which were fixed ∼8 minutes after injection of the last embryo (∼11 minutes after injection of the first). Anti-Egl antibody (Mach and Lehmann, 1997) was injected 10 minutes before the RNAs.

To guard against potential dominant lethality arising from expression of non-localising transcripts, we used a conditional expression system in which an upstream FRT-stop-FRT cassette terminates transcription and prevents transgene expression unless excised using FLP recombinase(Struhl et al., 1993). We cloned the wild-type h cDNA, or one containing the hΔD deletion that removes 20 nucleotides required for localisation (Bullock et al.,2003), appended to 3′ h genomic sequences containing the polyadenylation signal, into a unique PmeI site within the Drosophila transformation vector P{w^+mC (eve2)2>hsp70 3′>}, a derivative of P{w^+mC(eve2)2 >hsp70 3′> eve3′}(Wu et al., 2001) with the eve 3′UTR removed. h fragments were generated by PCR(details available upon request) and the h-coding regions in the final constructs were sequenced to ensure that no mutations were introduced. The resultant constructs (P{w^+mC (eve2)2 >hsp70 3′> h^wt} and P{w^+mC (eve2)2>hsp70 3′> hΔD^nloc}contain two copies of a minimal eve stripe 2 enhancer that, following excision of the stop signal, drive expression of h in parasegment 3(Kosman and Small, 1997). In the event, removal of the transcriptional terminating sequence in the male germ-line with the β2-tubulin-FLP transgene(Struhl and Basler, 1993)yielded viable transgenic lines which were used in all the experiments described here. To quantitate levels of st2-h expression in different lines, we generated cDNA from 150-300 blastoderm embryos 2.5-3.25 hours after egg laying at 25°C from crosses of homozygous flies (nlocA, B and C; wtA and B) or heterozygous flies when homozygous lines were lethal (nlocD; wtD) or semi-lethal (wtC). The data shown in Fig. 6E are normalised for gene dosage. We used real-time PCR with Taqman™ probes to quantify amounts of activated st2-h cDNA relative to actin 5C cDNA using the comparative C_T method described by the manufacturers (PerkinElmer). The Taqman probes (used at 0.1 μM) and primers(0.4 μM each) were as follows: st2-h 5′ GTGACCGCCGCACAGTC; st2-h 3′ AACTTCAAGATCCCCATTCAAAGT; st2-h Taqman™probe CAACTAACTGCCTTCGTTAATATCCTCTGAATAAGCC; Actin 5′GGTTTATTCCAGTCATTCCTTTCAA; Actin 3′ ACTGTAAACGCAAGTGGCGA; Actin Taqman™ probe CCGTGCGGTCGCTTAGCTCAGC.

The st2-h oligonucleotides amplify sequences between the eve 5′UTR and the FRT site in the transgene. Using cDNA from embryos of the parental yw strain, we showed that these primers do not amplify endogenous cDNAs. Nor do they amplify products when no reverse transcription step is included. We confirmed the validity of our assay using serial dilutions of one of the transgenic cDNA samples.

To investigate the functional significance and phylogenetic occurrence of pair-rule mRNA localisation, we first cloned eve and horthologues from species throughout Diptera(Fig. 1) (see Fig. S1for full sequence alignments). We assayed transcript localisation in four of these species that can be cultured in the laboratory by whole-mount in situ hybridisations on blastoderm embryos. Two of them, Episyrphus(Syrphidae) and Megaselia (Phoridae), are cyclorrhaphan flies (i.e. higher dipterans) but, unlike Drosophila, belong to basal branches of this taxon; the other two, Coboldia (Scatopsidae) and Clogmia (Psychodidae), belong to different branches of lower Diptera(Fig. 2).

In each species, eve and h transcripts are expressed in circumferential stripes (Fig. 2A-E,K-O). In the higher dipterans Drosophila, Episyrphusand Megaselia, seven eve and h stripes are formed at blastoderm stage (Fig. 2A-C,K-M). In the lower dipterans, Clogmia (where eve has been duplicated; Fig. 1 and Fig. S1) and Coboldia, fewer than seven eve and h stripes are present at this stage(Fig. 2D,E,N,O) and posterior pair-rule stripes are added after the onset of gastrulation(Rohr et al., 1999). The striped expression of the eve and h transcripts in blastoderm embryos suggests that they act during segmentation in these species, although, interestingly, pair-rule expression of the putative h orthologue in Clogmia may not be conserved(Fig. 2O).

In Megaselia, eve and h transcripts are tightly localised apically throughout blastoderm stages, much like their counterparts in Drosophila (Fig. 2H,R). In these species blastoderm embryos have a thickened blastoderm with nuclei that reside apically throughout the cellularisation process (Fig. 2F-H). In Episyrphus, transcripts are also enriched apically, but in contrast to Drosophila and Megaselia, a substantial proportion of the mRNA accumulates in the basal cytoplasm(Fig. 2G,Q). Early Episyrphus blastoderm embryos also have apical nuclei. However, in this species the nuclei adopt a more central position during the cellularisation process, at a time when pair-rule genes are active (see Fig. S2). In Clogmia and Coboldia, however, eve and h transcripts are distributed evenly throughout the cytoplasm and these species have a thin blastoderm with little cytoplasm surrounding the nuclei (Fig. 2I,J,S,T). These results suggest that localisation of pair-rule transcripts is not associated with pair-rule patterning per se and that a requirement for the apical localisation of pair-rule genes may be influenced by the cytoarchitecture of the blastoderm. In addition, the apical localisation of eve and h mRNAs in diverse cyclorrhaphan flies that evolved independently for∼145 million years (Grimaldi and Cumming, 1999) indicates that pair-rule transcript localisation is functionally significant.

To test whether the localisation machinery is active in transporting other transcripts in Clogmia and Coboldia, we cloned homologues of wg (Fig. 1, Fig. 3A-C, Fig. S1),which encodes an extracellular signalling protein, from these species. In Drosophila, wg transcripts are localised apically in late blastoderm(Fig. 3D) and cellularised postgastrular embryos (Fig. 3G). We find that wg transcripts are also enriched apically in Coboldia embryos at equivalent stages(Fig. 3E,H). This suggests that the Egl/BicD/dynein localisation machinery is active, and that the failure of Coboldia-eve and -h transcripts to localise reflects a lack of mRNA localisation signals.

In Clogmia, wg transcripts are distributed uniformly in the cytoplasm of blastoderm embryos and become apically enriched only in epithelial cells of postgastrular embryos(Fig. 3F,I). These observations suggest that the Egl/BicD/dynein localisation machinery is present in lower Diptera, but that it can be deployed at different stages in different species.

We wanted to survey localisation of eve and h orthologues in additional species in order to gain further insights into the phylogenetic occurrence of pair-rule mRNA localisation. However, embryos from many phylogenetically informative taxa are not available for in situ hybridisation. We therefore used injection of fluorescently labelled transcripts into Drosophila embryos (Bullock and Ish-Horowicz, 2001; Wilkie and Davis, 2001) to test for localisation signals in eve, hand wg homologues cloned from other dipteran species. Each labelled RNA included a region of coding sequence as well as the full-length 3′UTR (Table S1). The distribution of all 11 injected Megaselia, Episyrphus, Clogmiaand Coboldia transcripts mirrors closely their endogenous distributions observed by in situ hybridisation(Fig. 2, Fig. 3, Fig. 4A,B). In all cases,apical accumulation of localising transcripts is prevented by pre-injecting Drosophila embryos with antibodies that specifically inhibit Egl function (Bullock and Ish-Horowicz,2001) (not shown). Based on these data, we believe that RNA injection into Drosophila blastoderm embryos provides a reliable tool for detecting Egl/BicD/dynein-dependent localisation signals throughout Diptera. The similarities between localisation of endogenous and injected transcripts indicates that the specificity of the RNA recognition factors has changed little during more than 210 million years of dipteran evolution,presumably because it is constrained by the need to recognise multiple cargoes in different cell types.

We therefore used this heterologous assay to probe for Egl-dependent localisation signals in transcripts of several other species where it is not possible to examine transcript localisation in situ(Fig. 4). eve and h RNAs from the cyclorrhaphan Platypeza (Platypezidae)localise apically in Drosophila blastoderm embryos, providing further evidence that localisation signals in pair-rule transcripts are common throughout Cyclorrhapha.

We detected localisation signals in wg and three out of four tested pair-rule transcripts of the Malaria mosquito Anopheles[odd skipped (odd), fushi tarazu (ftz)(not shown) and h (Fig. 4A); a signal was not found in the full-length evetranscript]. The presence of localisation signals in pair-rule genes of this lower dipteran is interesting, because similar blastoderm types appear to have evolved convergently in the cyclorrhaphan and the culicomorphan branch of Diptera to which Anopheles belongs(Ivanova-Kasas, 1949; Anderson, 1972; Monnerat et al., 2002). Of Empis-eve, Haematopota-h and Haematopota-eve - transcripts from species that are more closely related to cyclorrhaphan flies than Anopheles, Clogmia and Coboldia - only the last localises upon injection into Drosophila. Although the injection assay does not allow us to discern the developmental context in which localisation signals are used, the results corroborate our conclusion from the analysis of mRNA localisation in situ (Fig. 2, Fig. 3), that, in Diptera,localisation of wg transcripts is conserved, whereas localisation of pair-rule transcripts is labile.

We could not detect any significant stretches of conserved primary sequence in 3′ UTRs of localising transcripts. This is not surprising, however,because efficient signal recognition by the localisation machinery can be mediated by multiple, partially redundant interactions in which the essential features are contained within short stretches of base-paired RNA(Macdonald and Kerr, 1998; Bullock et al., 2003). Even within the genus Drosophila, the primary sequence of the hlocalisation signal has diverged significantly(Bullock et al., 2003).

The apparent correlation between apical pair-rule transcript localisation and apical-residing nuclei led us to the hypothesis that apical transcript localisation augments nuclear uptake of the transcription factor products of the pair-rule genes by targeting translation apically, in close proximity to the nuclei. We therefore assayed the consequences of disrupting pair-rule mRNA localisation in embryos of Drosophila melanogaster. Because components of the pair-rule mRNA localisation machinery are also required maternally for differentiation of the oocyte(Deng and Lin, 2001), we used a partial loss-of-function allele of egl(Navarro et al., 2004), which provides sufficient Egl function to overcome the earlier block in oogenesis. Females that have one copy of this allele (egl^3e) and one copy of a null allele (egl^WU50) lay eggs, ∼40-60% of which are fertilised and develop to blastoderm stages. Whereas embryos laid by wild-type mothers have an almost exclusively apical distribution of pair-rule mRNAs such as eve, ftz, h and runt (run), those laid by egl^3e/egl^WU50 mothers accumulate a large proportion of these transcripts in the basal cytoplasm(Fig. 5A). A slight apical enrichment of pair-rule mRNAs is still detectable in most stripes in these mutants (Fig. 5A), consistent with their retention of some Egl activity.

The defect in transcript localisation in egl mutant embryos also affects the subcellular distribution of pair-rule proteins(Fig. 5A). For example, in wild-type blastoderm embryos, Run protein is first detected predominantly in the apical cytoplasm (not shown); in slightly older blastoderms, the bulk of this protein has accumulated in the nuclei but a proportion of it is still detected in the apical cytoplasm (Fig. 5A). In egl-deficient embryos, more diffuse cytoplasmic Run protein staining is also detected basally, similar to the distribution of its transcripts (Fig. 5A). We also observed basal accumulation of other pair-rule proteins in eglmutants (not shown), but the width and intensity of protein stripes as well as protein levels are not altered noticeably (not shown). Together, these observations suggest that pair-rule transcript localisation targets protein to the apical cytoplasm prior to import into the nuclei.

In embryos from egl mutant mothers, the apical localisation of wg transcripts is very strongly reduced(Fig. 5A) but, despite the inefficient localisation of these and pair-rule mRNAs, segmentation is only slightly impaired. Some (7.4%; n=136) egl^3e/egl^WU50 blastoderm embryos show variable defects in the pattern of segmental engrailed expression (not shown),whereas only 1.3% of embryos from the reciprocal cross - i.e. wild-type mothers mated to egl^3e/egl^WU50 males - exhibit such defects (n=309; P<0.01; Fisher's exact test). The frequency of mild cuticular patterning defects in first instar larvae from egl mutant females is also increased (3.1%, n=349) compared with wild-type controls (0.7%, n=420; P<0.05; Fig. 5B).

However, egl mutant embryos are acutely sensitive to a reduction in pair-rule gene dose (Fig. 5B,C). For example, 32.4% of hⁱ²²/h⁺ first instar larvae from egl^3e/egl^WU50 mothers have pair-rule defects,compared with 3.2% from wild-type mothers (P<0.001). These genetic interactions do not reflect a general sensitivity of early patterning processes to a reduction in Egl function, however, because phenotypes caused by heterozygosity of gap genes such as Krüppel (Kr) and knirps (kni) - which function upstream of the pair-rule genes in segmentation, and whose transcripts are not localised asymmetrically- and wg are not enhanced significantly by the maternal eglmutant genotype (Fig. 5B).

These experiments suggest that apical localisation of pair-rule mRNAs and proteins enhance their activity, although they do not rule out entirely a role for Egl independent of its function in pair-rule transcript localisation. Nor do they address the consequences of completely blocking localisation of a pair-rule mRNA, because the egl mutants still retain some transport activity.

We therefore assayed the activity of a pair-rule protein encoded either by localising transcripts or by transcripts distributed uniformly in the cytoplasm. Wild-type h transcripts or transcripts lacking 20 nucleotides within the 3′UTR that are essential for RNA transport [the hΔD mutation(Bullock et al., 2003)] were misexpressed in the eve stripe 2 domain, and assayed for their ability to repress stripe 2 of the h target gene ftz, which leads to deletions in the larval mesothorax (segment T2). As the effects in this assay are dose dependent (Wu et al.,2001), it is well suited to probe for subtle differences in activities. As expected, lines bearing transgenes that encode the wild-type 3′-UTR produce apically localised transcripts(st2-h^wt lines; Fig. 6B), whereas those expressing the mutant 3′UTR give rise to transcripts distributed uniformly in the cytoplasm(st2-h^nloc lines; Fig. 6C). We did not observe significant differences in the width of the stripe of localising or nonlocalising ectopic h transcripts using a probe that distinguishes st2-h from endogenous h (not shown).

Lines bearing localising or non-localising transcripts of st2-hcan lead to defects in T2 (Fig. 6D), demonstrating that nonlocalising transcripts encode functional protein. To distinguish effects of expression level and mRNA localisation, we used real-time RT-PCR to quantitate levels of st2-hmRNA relative to those of endogenous actin mRNA levels in each transgenic line (Fig. 6E). Comparing lines expressing similar levels of st2-h transcript indicates that T2 defects are more severe and penetrant when transcripts localise apically (Fig. 6D). Consistent with this, localised st2-h RNA is more efficient at repressing ftz transcription than unlocalised transcripts(Fig. 6I,J; see Table S2). Only in very strongly expressing lines (st2-h^wtD and st2-h^nlocD) do the ectopic transcripts cause equivalent phenotypes (disruption of T2 in over 90% of embryos), indicating that high expression levels can overcome the requirement for apical mRNA localisation. We estimate that in moderately expressing lines, localising st2-htranscripts have similar effects to two- to threefold more nonlocalising transcripts (Fig. 6D,E). Consistent with our analysis of egl mutant embryos, these data indicate that apical mRNA localisation augments pair-rule activity.

To investigate how localising mRNA enhances h activity, we compared the level of Hairy protein in the eve stripe 2 domain of transgenic lines encoding localising and non-localising st2-htranscripts. We found that levels of the protein in nuclei of the evestripe 2 domain are clearly lower in st2-h^nlocC than in st2-h^wtB (Fig. 6K-N), even though the former expresses significantly more transcript. Although the anti-Hairy antibody is not sufficiently sensitive to detect cytoplasmic Hairy protein above background, we observe a diffuse distribution of several other pair-rule proteins in the basal cytoplasm when apical pair-rule RNA localisation is compromised in egl mutants (see above). In addition, when an excess of in vitro synthesised wild-type h RNA is injected, Hairy protein is detected in the apical cytoplasm,whereas when transcripts are injected that contain the same inactivating deletion in the localisation signal carried by the st2-h^nloc lines Hairy protein is detected uniformly throughout the cytoplasm (not shown). Together, these data argue that apical h RNA localisation targets protein apically, in close proximity to the nuclei.

Apical localisation of pair-rule mRNAs in Drosophila syncytial blastoderm embryos was first noted 20 years ago, but the developmental and evolutionary significance of this process has remained unclear. We show that apical pair-rule mRNA localisation is conserved in cyclorrhaphan species that diverged over 145 million years ago, indicating that this process has a significant developmental role under natural conditions. Likewise, the widespread maintenance of wg transcript localisation in Diptera supports the importance of this process on a phylogenetic scale, even though,in Drosophila, wg appears to be less sensitive than pair-rule genes to a reduction in endogenous transcript localisation(Fig. 5B).

Unlike wg transcripts, pair-rule mRNAs do not localise in some branches of lower Diptera, and the phylogenetic occurrence of this process provides interesting insights into its functional significance. Enrichment of pair-rule transcripts in the apical cytoplasm correlates with the position of blastoderm nuclei: efficient apical localisation of pair-rule gene transcripts is found in species which retain an asymmetric apical position of nuclei throughout the blastoderm stage (Drosophila, Megaselia); less efficient localisation is seen when the nuclei move from an apical to a more central position during blastoderm stages (Episyrphus); and no apical enrichment of transcripts is seen in species where blastoderm nuclei are surrounded uniformly by a thin layer of cytoplasm (Coboldia,Clogmia). We also find localisation signals in several pair-rule transcripts of the lower dipteran Anopheles. Like Cyclorrhapha, but unlike many other lower Diptera and most other insects, this culicid species has evolved a thickened blastoderm with apically positioned nuclei, probably to allow rapid development as an adaptation to ephemeral larval habitats(Anderson, 1972; Ferrar, 1987): columnar cells that emerge from thickened blastoderms can enter gastrulation directly,whereas cuboidal cells that emerge from thin blastoderms still have to elongate prior to undergoing the requisite cell shape changes.

In Drosophila, we find that pair-rule proteins are enriched in the apical cytoplasm prior to import into the nuclei in wild-type blastoderms,whereas they are detected basally in egl mutant embryos, in which transcript localisation is inefficient. The apical accumulation of pair-rule proteins under normal circumstances is consistent with the observation that apical RNA targeting restricts diffusion of cytoplasmic β-galactosidase(Davis and Ish-Horowicz, 1991). Apically targeted protein is most likely confined by the cellularisation process, in which the plasma membrane invaginates between the nuclei and encloses the apical compartment first (Fig. 7).

Davis and Ish-Horowicz (Davis and Ish-Horowicz, 1991) speculated that mRNA localisation prevents pair-rule proteins from moving into inter-stripe regions, where they would cause dominant patterning defects. However, when pair-rule mRNA localisation is compromised, either by interfering with the localisation machinery or the RNA signals, we do not observe expansion of RNA or protein stripes or ectopic phenotypic effects. Rather, we see a reduction of pair-rule activity in their domains of expression in these experiments, indicating that transcript localisation augments gene function. Pair-rule mRNA localisation does not appear to be obligatory for protein activity in Drosophila but makes the segmentation process more reliable: egl mutants, in which transcripts localise very inefficiently, have a mild increase in segmentation defects and are acutely sensitive to the reduction of pair-rule gene dose.

By what mechanism does pair-rule mRNA localisation augment the activity of their transcription factor products? We demonstrate for h that suppression of transcript localisation reduces nuclear levels of its protein. Pair-rule proteins could be specifically modified in the apical cytoplasm, or localising transcripts could be translated more efficiently. However, given the diffuse distribution of pair-rule proteins in the basal cytoplasm when RNA localisation is disrupted in egl mutants and the correlation between cytoarchitecture and pair-rule transcript localisation in Diptera, we favour a third possibility, namely that apical mRNA localisation increases nuclear uptake of their proteins by targeting translation in close proximity to the nuclei (Fig. 7). Proteins from non-localising mRNAs would not be available at high levels in the immediate vicinity of the nuclei, which would result in a decreased nuclear uptake. Such a role for apical pair-rule mRNA localisation would be redundant in lower Diptera with only a thin layer of cytoplasm surrounding the nuclei, which provides little room for diffusion of pair-rule proteins prior to nuclear import. A mechanism for perinuclear protein targeting might be particularly significant for nuclear proteins with short half-lives, such as those encoded by pair-rule genes (Edgar et al.,1987). Interestingly, localisation of mRNA in the vicinity of the nucleus to aid import of nuclear proteins has also been reported in cultured mammalian cells (Levadoux et al.,1999) and may be a widespread mechanism to efficiently exploit a limited pool of transcripts in cells that are polarised or have a high cytoplasmic:nuclear ratio.

The relationship between cytoarchitecture and apical pair-rule transcript localisation does not appear to be absolute because we detected a signal in eve, but not h, from Haematopota, which has retained the ancestral, cuboidal blastoderm morphology (U.S.-O., unpublished)and because we did not detect a localisation signal in Anopheles-eve. Although we cannot yet discern the developmental context in which these signals are used (in situ hybridisation is currently not possible in these species because of egg shells that are difficult to remove and because of difficulties in obtaining embryos) these data raise the possibility that,within a single species, the differential ability of transcripts to be recognised by the localisation machinery is used to fine-tune transcriptional control of target genes in the blastoderm by modulating the nuclear concentration of pair-rule proteins.

The ability of eve and h pair-rule transcripts to use the localisation machinery varies in Diptera. We observe a range of localisation efficiencies in situ that are mirrored in all 11 cases upon injection into Drosophila embryos. Thus, differences in localisation efficiency appear to reflect changes in the respective localisation signals, rather than alterations in the specificity of the protein machinery. These findings are consistent with previous studies with artificial variants of the Drosophila-h localisation signal, which suggest that the character of localisation signals modulates the efficiency of localisation by determining the kinetics of both the initiation of transport and the transport process itself (Bullock et al., 2003). Localisation efficiency appears to be determined by multiple RNA:protein interactions, the sum of which affects the stability and/or activity of the RNA:motor complex (Macdonald and Kerr,1998; Chartrand et al.,2002; Bullock et al.,2003). Therefore, the efficiency of the localisation process can be modified gradually during evolution by the addition, loss or modification of individual recognition sites within mRNAs.

It seems that localisation signals in pair-rule genes have emerged multiple times within Diptera. For example, although we cannot rule out the possibility that localisation signals in h have been lost in multiple different lineages of lower Diptera, the most parsimonious explanation for the phylogenetic distribution of signals in this transcript is that they evolved independently in response to changes in cytoarchitecture in the lineages leading to Cyclorrhapha and Culicomorpha. Injection of transcripts from additional species into Drosophila will determine whether eve localisation signals emerged independently in the lineages leading to Haematopota and Cyclorrhapha, or were lost in the lineage leading to Empis.

Work in mammalian cells has provided insights into how localisation signals might initially appear (Fusco et al.,2003). These studies suggest that non-localising mRNAs can also interact with a motor complex, albeit with a comparatively small probability,and undergo short movements on microtubules. Localisation signals appear to augment these interactions and lead to the net translocation of an RNA population along a polarised cytoskeleton by increasing the frequency and duration of directed transport (Fusco et al., 2003). The localisation machinery in Diptera may also have a general, weak affinity for mRNAs because a small proportion of particles of injected non-localising transcripts are transported over short distances in Drosophila embryos (M. Wainwright and S.B., unpublished). Asymmetric accumulation of a population of transcripts may therefore evolve gradually as a result of selection for increased interaction between a specific transcript and the localisation machinery.

Using a combination of functional and phylogenetic analyses, we have provided evidence that the alteration of mRNA localisation signals is an important mechanism by which the activity of pair-rule transcription factors is regulated in flies. Apical localisation of these transcripts appears to augment the nuclear concentration of their protein products and makes the segmentation process less sensitive to perturbation of gene activity. It seems that different species have made use of the localisation machinery to adapt the deployment of specific pools of transcripts to evolutionary changes in blastoderm cytoarchitecture. Thus, the mRNA localisation mechanism may permit networks of patterning genes to tolerate changes in cell morphology, such as those imposed by reproductive adaptations.

In Drosophila, transport of mRNAs by the Egl/BicD/dynein machinery determines the distributions of several different kinds of proteins in diverse cell types such as oocytes, epithelial cells and neuroblasts(Bullock and Ish-Horowicz,2001) (J.H., unpublished). Our studies of pair-rule mRNAs imply that the repertoire of other RNA cargoes for the machinery, and their efficiency of transport, may also be modulated readily in evolution through changes in localisation signals. Therefore, differential mRNA localisation is potentially an important factor in facilitating morphological evolution.

We thank Gordon Dowe for sequencing. For reagents, we thank Ruth Lehmann,Peter Hondelmann, Steffen Lemke, Sheena Pinchin, Steve Small and Hans-Michael Müller. We also thank Herbert Jäckle for his advice and generous support of this project, and acknowledge advice, and/or comments on the manuscript by Klaus Sander, Peter Chandler, Hans Ulrich and Michael Clark. Finally, we thank members of our laboratories and colleagues in the laboratory of Herbert Jäckle for discussions and support. Funding was provided by Cancer Research UK (S.B., J.H. and D.I.-H.) and the Beit Memorial Trust(S.B.), the Max-Planck-Gesellschaft, The University of Chicago and the Deutsche Forschungsgemeinschaft (U.S.-O.).