Ptp10D and Ptp4E are clearly the result of a gene duplication that occurred much more recently than the split between the other Drosophila Rptp genes. The amino acid sequences of their catalytic PTP domains share 89% identity, versus 36–40% identity for pairwise comparisons of Ptp4E with other Drosophila RPTPs. Their XC domains have a very similar organization, containing chains of 11 FN3 repeats in Ptp4E and 12 FN3 repeats in Ptp10D, and are 58% identical in amino acid sequence (Figure 1c). The Ptp4E gene encodes two predicted preproteins, of 1,767 and 1,607 amino acids, while the Ptp10D gene encodes preproteins of 1,931 and 1,631 amino acids. The sequences that differ between the alternative gene products are at the carboxyl terminus in both cases, but there is no sequence similarity between the Ptp4E and Ptp10D proteins within this region. Both genes reside on the X chromosome, and the nine Ptp4E introns within conserved coding sequence all correspond exactly in position to Ptp10D introns. Ptp10D has one additional intron not found in Ptp4E.

The Caenorhabditis elegans gene dep-1 is the ortholog of both Ptp10D and Ptp4E. Humans and mice have five genes encoding type III RPTPs, defined as proteins with XC domains composed of long chains of FN3 domains and a single PTP domain. Among these, the product of the PTPRB gene (PTPβ, not to be confused with RPTPβ, which is a different protein also known as PTPζ) has a somewhat higher alignment score to Ptp10D and Ptp4E than the other four mammalian type III proteins. These are: DEP-1/CD148, encoded by the Ptprj gene; PTPRO; SAP-1, encoded by the Ptprh gene; and PTPRQ. Since the radiation into the five mammalian genes seen today occurred after the split between arthropods and mammals, one cannot define any of the type III mammalian genes as an ortholog of one of the fly genes. A more complete description of the relationships among all the Drosophila, C. elegans, and mammalian RPTPs is found in 5.

The recent availability of genome sequences from twelve different Drosophila species, three mosquito species, two hymenopterans, a beetle, and the silkmoth allowed us to trace the evolution of the Ptp10D/Ptp4E gene pair within insect lineages. Surprisingly, we find that the Ptp4E gene is found only in drosophilid species. Mosquitoes, which are also dipterans, and all other sequenced insects have only a single Rptp gene corresponding to this gene pair. This gene is always much more closely related to Ptp10D than to Ptp4E. In addition, the Ptp4E sequence exhibits more sequence diversity among the drosophilid species than does the Ptp10D sequence. These data indicate that Ptp10D is the ancestral gene and that its sequence has been constrained more by evolution than the Ptp4E sequence since the time of the duplication. Ptp4E has evolved much more rapidly, suggesting that it may have acquired new function(s) since its emergence or was less essential for fitness than Ptp10D. These relationships are displayed in the phylogenetic tree of Figure 1d.

The Ptp10D ortholog found in all insect species always contains the Ptp10D-specific intron, and all Ptp4E orthologs in drosophilids lack this intron. This suggests that the intron may have been lost at the time of the duplication from the copy that evolved into Ptp4E. This would have been between 235 million years ago (the estimated time at which the mosquito and fly lineages diverged from each other) and 40 million years ago (the estimated time at which the radiation among the 12 sequenced drosophilid species occurred) 2223.

We attempted to trace the history of the duplication by examining the genes adjacent to Ptp10D and Ptp4E, but found that the organizations of the Ptp10D and Ptp4E regions in D. melanogaster arose long after the Ptp10D and Ptp4E genes diverged from each other. Ptp10D is flanked by the Rst(1)JH and bifocal genes. Rst(1)JH is found upstream of the Ptp10D ortholog in both the obscura and melanogaster groups, but is separated from it in D. willistoni and all other drosophilids. bifocal orthologs are adjacent to the Ptp10D gene only in the melanogaster group. Similarly, the two genes flanking Ptp4E, SIP3 and CG4068, are located next to the Ptp4E ortholog only within the obscura and melanogaster groups. There are no significant sequence similarities between the genes that flank Ptp10D and Ptp4E. See 24 for a phylogenetic tree displaying the relationships among the sequenced insect species.

We generated a deletion mutation in Ptp4E, denoted Ptp4E¹, by imprecise excision of a P element, EP425, located upstream of the putative transcription start site. This mutation removes the first and second exons, thus deleting the sequences encoding the initiating methionine and the first 67 amino acids of the Ptp4E protein (Figure 1a).

Ptp4E¹/Y males are viable, fertile, and apparently healthy, as are Ptp4E¹/Df(1)ovo4 females. We could not detect any alterations in the CNS or neuromuscular system in embryos or larvae of these genotypes. Df(1)ovo4/Y embryos also had no CNS or neuromuscular system phenotypes that could be detected by antibody staining. To determine if the absence of a loss-of-function phenotype for Ptp4E is due to compensation by the closely related Ptp10D protein, we constructed Ptp4E¹ Ptp10D¹ double mutants. Ptp10D¹ is an excision mutation that removes the amino-terminal coding sequences of Ptp10D 3. Ptp10D¹ animals are viable and fertile, and also have no detectable defects in their embryonic nervous systems. In contrast, Ptp4E¹ Ptp10D¹/Y animals can hatch out into first instar larvae, but die immediately after hatching. To ensure that this lethality is not due to other mutations on these chromosomes that confer lethality when combined in the Ptp4E¹ Ptp10D¹ recombinant chromosome, we also made a double mutant chromosome containing independently isolated insertions just upstream of Ptp4E (Ptp4E^KG2328) and Ptp10D (Ptp10D^EP1172). This chromosome is also lethal in hemizygous males (Ptp4E^KG2328 Ptp10D^EP1172/Y). These data confirm the hypothesis that the viability of animals lacking either Ptp4E or Ptp10D is due to compensation by the other protein.

The published in situ hybridization data suggest that Ptp4E mRNA is ubiquitously expressed in late embryos, although there are some level differences between tissues 25. To further analyze expression, and to ensure that the observed pattern was not affected by cross-hybridization between the closely related Ptp4E and Ptp10D phosphatase domain sequences, we repeated this analysis using a probe from the first four Ptp4E exons, which are not closely related to Ptp10D.

In gastrulating embryos, Ptp4E mRNA is enriched in the invaginating mesoderm (Figure 2a). In germ band extended embryos (stage 10–11), the strongest expression is observed in the posterior midgut primordium. There is also an interesting 'scalloped' pattern of expression observed at the ectodermal border (Figure 2b,c, vm). This has an intriguing correspondence to visceral mesoderm dpERK staining 26, suggesting that Ptp4E may be enriched at sites of RTK activation. Interestingly, thisbe, which encodes a ligand for the fibroblast growth factor receptor Heartless-, is expressed in a similar pattern 27. Figure 2d is a germ-band extended embryo expressing a UAS-linked Ptp4E-green fluorescent protein (GFP) fusion from the engrailed-GAL4 driver. This shows the expected striped pattern of expression, confirming that the probe recognizes Ptp4E and allowing an estimate of the relative levels of driven versus endogenous Ptp4E mRNA.

At stage 14, Ptp4E is widely expressed, with highest levels observed in the midgut. Expression in the CNS can also be seen (Figure 2e). Finally, at stage 17 expression is much higher in the gut than elsewhere, with particularly strong expression observed in the hindgut and at the anterior end of the midgut (Figure 2f).

To examine Ptp4E protein expression and localization, we generated a variety of mouse monoclonal and polyclonal antibodies against Ptp4E. We made a construct that consisted of 6×-His-tagged Ptp4E XC domain fused to human placental alkaline phosphatase (PTP4E:AP; Figure 1b). PTP4E:AP was expressed in insect cells, purified over a Ni-NTA column, and injected into mice. To characterize the resulting antibodies, we first stained embryos that overexpressed Ptp4E-GFP using the engrailed-GAL4 driver. Figure 2g,h shows double staining of a germ-band extended embryo with anti-Ptp4E (red) and anti-GFP (green). The polyclonal antibodies clearly recognize the Ptp4E protein, as anti-Ptp4E (red) signal colocalizes with anti-GFP (green) signal in the expected striped pattern. Figure 2i shows a wild-type embryo at the same stage stained with anti-Ptp4E, where the expression appears to be ubiquitous. Note that the staining between the stripes in Figure 2g is much weaker than within the stripes, suggesting that the endogenous protein is expressed at low levels.

Although the ubiquitous staining observed with the antibody is consistent with the in situ hybridization data, we cannot be sure that antibody staining in wild-type embryos is due to Ptp4E protein, because it is not significantly reduced in Ptp4E mutant embryos (Ptp4E¹ or Df(1)ovo4). This finding could be explained in two ways. First, Ptp4E¹ mutants might continue to make an abnormal Ptp4E protein(s) due to initiation of translation at methionine residues encoded in exons not removed by the excision mutation (the second methionine residue in Ptp4E is at amino acid 377, within an undeleted exon). This protein, if it exists, would lack a signal sequence and may be nonfunctional, because the CNS phenotypes of Ptp4E¹ Ptp10D¹ and Df(1)ovo4 Ptp10D¹/Y embryos are identical (Figure 3). Df(1)ovo4 deletes the entire Ptp4E gene (Figure 1). The presence of antibody staining in Df(1)ovo4 mutants could be due to persistence of Ptp4E protein synthesized from maternal mRNA, since early embryos contain large amounts of Ptp4E mRNA 25. Second, it is possible that our Ptp4E antibodies cross-react with another ubiquitously expressed protein. They do not cross-react with Ptp10D, because the signal does not decrease in Ptp10D¹ mutant embryos.

Although we could not use the antibody to define where Ptp4E is expressed in the embryonic CNS in wild-type embryos, its ability to recognize overexpressed Ptp4E protein (Figure 2g,h) allowed us to ask whether Ptp4E can localize to axons. To do this, we drove Ptp4E-GFP with the pan-neuronal elav-GAL4 driver. In these embryos, bright staining of both CNS and peripheral nervous system (PNS) axons is observed with anti-Ptp4E and anti-GFP antibodies, and the two patterns are superimposable (Figure 2j–o). Interestingly, Ptp4E-GFP also appears to localize to neuronal cell bodies in the PNS and CNS (Figures 2j–l). In contrast, Ptp10D, Ptp69D, Lar, and Ptp99A, which are restricted to axons in wild-type embryos, are also axon-specific when overexpressed (unpublished data).

Expression of Ptp10D protein was detected only in the nervous system in published work. It is selectively expressed on embryonic CNS axons 1617, and in the neuropil of the larval and adult brain 28. Our recent data, however, show that Ptp10D is also expressed by embryonic tracheal cells. These findings suggest that the embryonic/larval lethality of Ptp4E¹ Ptp10D¹ animals might be due to either nervous system or tracheal phenotypes. In fact, we have found that these embryos have severe tracheal defects. These will be described elsewhere (MJ and KZ, manuscript in preparation). Their nervous system defects, however, are relatively mild (see below), and would not be expected to produce early lethality. Consistent with this, we find that GAL4-driven pan-neural expression of a UAS-Ptp4E-GFP fusion, which is capable of rescuing the tracheal phenotype when driven in tracheal cells by breathless-GAL4 (MJ and KZ, manuscript in preparation), does not rescue lethality in the Ptp4E¹ Ptp10D¹ background.

Driving Ptp10D in tracheae with breathless-GAL4 in a Ptp4E¹ Ptp10D^EP1172 background (the EP1172 line is a UAS-containing P element insertion upstream of the gene, so it allows rescue by crossing in GAL4 drivers) rescues lethality, allowing some adults to emerge. These data confirm that lethality in the double mutant is rescuable by Ptp10D expression in tracheae (or in other cells that express breathless-GAL4). We also attempted to rescue lethality by ubiquitous expression of Ptp4E, but found that pancellular overexpression of Ptp4E-GFP driven by tubulin-GAL4 is lethal.

To evaluate CNS defects in multiply mutant embryos, we stained them with a monoclonal antibody (mAb), 1D4, and also crossed a SemaIIB-τMyc reporter 29 into the mutant backgrounds. mAb 1D4, directed against the cytoplasmic domain of Fasciclin II 30, stains three longitudinal bundles on each side of the CNS in stage 17 embryos. Figure 3 shows 1D4 staining of the CNS in stage 17 wild-type, Ptp4E¹, Ptp10D¹, Ptp4E¹ Ptp10D¹, and Df(1)ovo4 Ptp10D¹ embryos. In both of the Ptp4E double mutant genotypes, a mild fraying of the longitudinal bundles is observed. The outer bundle is often incompletely formed, and the outer bundle invades the middle longitudinal bundle. The phenotypes of the two genotypes are of approximately equal severity, suggesting that Ptp4E¹ could be a null mutation. Table 1 shows the quantitative comparison of the CNS phenotypes in double and triple mutant combinations.

The CNS defects are specific to loss of Ptp4E, because they can be rescued by supplying wild-type Ptp4E in neurons. When we drove expression of the UAS-Ptp4E-GFP transgene with elav-GAL4, the break and fraying phenotypes were rescued to near-wild-type levels. The differences between Ptp4E Ptp10D and the rescued or wild-type embryos are both highly statistically significant (p < 0.0001; Table 1). There is also a subtle overexpression phenotype produced by driving Ptp4E-GFP in neurons, in which the longitudinal bundles have a 'wavy' appearance (data not shown).

SemaIIB-τMyc is a useful marker for a specific axon pathway that crosses the midline in the anterior commissure and then extends anteriorly within the longitudinal tract 29. Staining with anti-Myc antibodies in embryos expressing this reporter in the Ptp4E¹ Ptp10D¹ background reveals that the longitudinal portions of the SemaIIB axon pathway exhibit a mild fraying, consistent with the 1D4 results (Figure 3).

Previous work from our group showed that Ptp10D Ptp69D mutants have a specific CNS phenotype in which some 1D4-positive longitudinal axons abnormally cross the midline. Two distinct longitudinal bundles usually remain in these mutants. The Ptp10D Ptp69D combination genetically interacts with robo1, slit, and commissureless mutations, and the data suggest that repulsive signalling through Robo1 in response to engagement of the midline Slit ligand is reduced in the absence of these RPTPs 3.

Since Ptp4E is similar to Ptp10D, we wondered whether Ptp4E might also genetically interact with Ptp69D to produce a synthetic CNS phenotype affecting the 1D4-positive longitudinal axons. This, however, is not the case. Ptp4E Ptp69D double mutants, like Ptp69D single mutants, have wild-type 1D4 and SemaIIB-τMyc patterns (Table 1, Figure 4a,e).

We then asked whether removal of both Ptp4E and Ptp10D together with Ptp69D would generate a new synthetic phenotype. We observed that the Ptp4E Ptp10D; Ptp69D triple mutant has a strong 1D4 phenotype, with extensive ectopic midline crossing (Figure 4f, Table 1). The 1D4 pattern shows midline crossing defects like those observed in Ptp10D Ptp69D mutants 3, in which a 1D4-positive bundle crosses in the posterior commissure, but the crossing bundle often is thicker than in the double mutants (Figure 4f, asterisk).

The longitudinal tracts look different from those in Ptp10D Ptp69D mutants, because they have more distinct bundles. In Ptp10D Ptp69D, two 1D4 bundles are observed that sometimes fuse into one, and the CNS is narrowed, suggesting that the outer bundle is missing (Figure 4b). In the triple mutants, there are always two, and sometimes three, longitudinal bundles. Extensive fusion and breakage of the bundles are observed, however (Figure 4f).

When examined with SemaIIB-τMyc, the Ptp10D Ptp69D double mutant displays occasional ectopic crossing by single axons in the posterior commissure (Figure 4c, arrowhead). Rare embryos have much more severe phenotypes in which the axons stall along the longitudinal pathways (Figure 4d, asterisk), and axons project out of the CNS rather than along the longitudinal tracts (Figure 4d, arrows). The triple mutant is observed to have more ectopic midline crossing in the posterior commissure than does the double mutant, and the crossing bundles are thicker (Figure 4g, arrowhead; Figure 4h). In one segment of Figure 4h, all of the axons that should project anteriorly appear to cross instead (top arrowhead). Other axons start crossing and then stall (Figure 4g, asterisk). We never observed SemaIIB axons that projected out of the CNS in triple mutant embryos (n = 10). It is difficult to say whether removing Ptp4E from a Ptp10D Ptp69D double mutant suppresses or enhances the double mutant phenotype. The phenotype changes, so that the outer 1D4 longitudinal bundle is restored in some segments of triple mutants. However, more ectopic crossing of SemaIIB axons is observed, and the ectopically crossing 1D4 bundles are often thicker.

What does this pattern of phenotypic interactions imply about the roles of these three RPTPs in CNS axon guidance? First, the existence of the synthetic Ptp10D Ptp69D phenotype indicates that Ptp10D and Ptp69D have redundant functions with regard to control of midline crossing, and that the activities of Ptp10D that are compensated for by Ptp69D with respect to midline crossing are not shared with Ptp4E. Second, there could be additional Ptp69D functions in CNS axonal guidance that are redundant with Ptp4E activities that are not shared between Ptp4E and Ptp10D. However, elimination of these activities in Ptp4E Ptp69D double mutants does not produce a strong 1D4 or SemaIIB phenotype. Third, when all functions of the Ptp10D/Ptp4E subfamily of RPTPs are eliminated, the 1D4 and SemaIIB phenotypes are still weak because these functions are mostly redundant with activities of Ptp69D. Finally, when all the Ptp10D/Ptp4E functions are removed together with the Ptp69D functions, the phenotype is subtly altered relative to the Ptp10D Ptp69D phenotype. The lack of a strong synthetic triple mutant phenotype suggests that Ptp4E is a relatively minor player in regulation of CNS axon guidance.

An earlier study from our group examined CNS axons in double, triple, and quadruple mutant combinations of Lar, Ptp10D, Ptp69D, and Ptp99A mutations 4. Strikingly, in this paper it was found that Lar Ptp69D Ptp99A and Ptp10D Lar Ptp99A triple mutants both have an almost normal pattern of 1D4-positive longitudinal axons, but when the fourth RPTP is removed as well (in a Ptp10D Lar Ptp69D Ptp99A quadruple mutant) all of the 1D4-positive longitudinal pathways are converted to commissural pathways that cross the midline. Thus, Lar and Ptp99A can have a strong effect on CNS axon guidance, but only when they are removed together with both Ptp10D and Ptp69D.

We wondered whether analysis of other mutant combinations involving Ptp4E would produce results consistent with the idea that strong synthetic 1D4 CNS phenotypes are unique to combinations in which both Ptp10D and Ptp69D are mutant, as suggested by 4. To examine this, we made Ptp4E Lar, Ptp4E Ptp52F, and Ptp4E Ptp99A double mutants, and Ptp4E Ptp10D Lar and Ptp4E Ptp10D Ptp52F triple mutants. Ptp4E Lar and Ptp4E Ptp99A mutants had no detectable alterations in 1D4-positive CNS longitudinal tracts. Ptp4E Ptp10D Lar mutants had phenotypes like those of Ptp4E Ptp10D double mutants, indicating that there are no important longitudinal axon guidance functions that are redundant between the Ptp10D/Ptp4E protein pair and Lar. Ptp52 mutants are the only Rptp single mutants that have a detectable 1D4 phenotype 5. Ptp4E Ptp52F mutants had 1D4 phenotypes indistinguishable from Ptp52F single mutants, while Ptp4E Ptp10D Ptp52F triple mutants had stronger phenotypes. However, we had already demonstrated that Ptp10D Ptp52F double mutants have a more disorganized pattern of 1D4-positive bundles than Ptp52F single mutants 5. The Ptp4E Ptp10D Ptp52F triple mutant does not show an obvious enhancement of phenotype relative to the Ptp10D Ptp52F double mutant (data not shown). In summary, of the eight mutant combinations involving Ptp4E that we examined, only the triple mutant that lacks both Ptp10D and Ptp69D has a strong CNS phenotype that is detectable with the markers used in this study. This phenotype is much like the Ptp10D Ptp69D double mutant phenotype, but differs from it in some subtle ways.

We also used mAb 1D4 staining to evaluate the motor axon phenotypes in all of these mutant combinations. There are approximately 36 motor neurons in each abdominal hemisegment. Motor axons exit the CNS in two main nerve roots (ISN and SN), then split into five distinct nerve branches. ISNd and SNc innervate ventral muscles, ISNb innervates ventrolateral muscles (VLMs), SNa innervates lateral muscles, and ISN innervates dorsal muscles. The patterns of motor axons in segments A2–A7 are essentially identical, so it is possible to quantitatively score phenotypes in up to 12 hemisegments per embryo.

Earlier work from our group and others showed that Rptp mutations affect every guidance decision made by these motor axon branches (with the exception of SNc branching, which has not been studied). Lar and Ptp52F are the only single mutants that have strong phenotypes. Lar mutations produce a 'parallel bypass' phenotype (approximately 30% penetrance in zygotic nulls), in which the ISNb axons leave the common ISN pathway at the exit junction but then fail to enter the VLM field. ISNd also fails to extend (approximately 80% penetrance) in these mutants 219. Ptp52F regulates bifurcation of the SNa nerve. One of the two SNa branches (anterior/dorsal or posterior/lateral) is missing in approximately 40% of hemisegments in zygotic null mutants 5.

When double, triple, and quadruple combinations of Rptp mutations were analyzed, we observed that other decisions are perturbed in specific patterns. Each combination of mutations has a unique phenotype, and the data suggest that specific RPTPs are required for each pathway choice made by motor axon growth cones. This is described for ISNb guidance in the Background section. Another example is extension of the ISN past branchpoints in the dorsal muscle field. In Lar single mutants, the ISN usually reaches its normal termination point but arborization on muscle 1 is reduced. In the absence of both Lar and Ptp69D, the ISN stops at the second branchpoint near muscle 2, while in Lar Ptp69D Ptp99A triple mutants it stops at the first branchpoint near muscle 3 19.

Not all combinations of Rptp mutations produce phenotypic enhancement. Ptp99A and Lar have opposing activities in controlling ISNb entry into the VLM field. Thus, in the strongest Lar mutant combination only 40% of ISNb branches successfully enter the VLM field, while in Lar Ptp99A double mutants this phenotype is almost completely suppressed, and approximately 96% of ISNbs enter the muscle field in a normal manner 419.

For the first four RPTPs to be analyzed (Lar, Ptp69D, Ptp99A, Ptp10D), it was possible to construct and analyze all 15 possible single, double, triple, and quadruple mutant combinations, and thus obtain a complete picture of the genetic relationships among these signalling molecules 4. When we identified mutations in the fifth gene, Ptp52F, it became impossible to make and analyze every combination, so we examined only double mutants 5. These two studies led to the conclusion that Ptp10D is relatively unimportant, relative to the other RPTPs, for regulation of motor axon guidance. This conclusion is opposite to that reached for CNS axon guidance, where we observe ectopic midline crossing only in genotypes where Ptp10D is missing 4.

Enhancement of Ptp52F phenotypes is observed in Ptp10D Ptp52F double mutants, so that ISNb 'stall' phenotypes, in which the ISNb stops short of its normal termination point at the ventral edge of muscle 12, and ISN truncation phenotypes are observed more frequently 5. These are the only double mutants involving Ptp10D that exhibit phenotypic enhancement relative to the corresponding single mutant. Ptp10D mutations increase the penetrance of some double and triple mutant motor axon phenotypes involving Lar, Ptp69D, and Ptp99A, and decrease the penetrance of others 4.

We wondered whether this apparent lack of a role for Ptp10D in motor axon guidance arises from the fact that Ptp10D is the only RPTP in Drosophila that has a paralog within the same subfamily. Each of the other RPTPs appears to represent its own subfamily. The Lar subfamily has a representative in C. elegans and three members in mammals, while Ptp69D and Ptp99A each have one worm counterpart but no obvious mammalian orthologs 5. Ptp52F has similarities to type III RPTPs, but has a unique amino-terminal sequence and appears to be unique to drosophilids (unpublished results). Because of their close sequence relationship, the loss of Ptp10D function in the motor axon system might be compensated for by the presence of Ptp4E. This model predicts that Ptp4E double mutant combinations also would not exhibit phenotypic enhancement relative to the corresponding single mutants. Indeed, we found this to be the case: no motor axon phenotypes are observed in Ptp4E Ptp99A mutants, and the phenotypes seen in Ptp4E Ptp69D, Ptp4E Lar, and Ptp4E Ptp52F mutants are no stronger than in the corresponding Ptp69D, Lar, and Ptp52F single mutants. We also found that removal of both members of the Ptp10D subfamily, in Ptp4E Ptp10D double mutants, does not produce motor axon phenotypes, suggesting that the subfamily has no essential role in motor axon guidance that is not compensated for by the presence of one of the other RPTPs (Table 2 and data not shown).

We then asked whether strong phenotypic interactions might be observed in triple mutants lacking both Ptp4E and Ptp10D together with one of the other subfamilies. If so, this would indicate that there are motor axon guidance functions that are redundant between the Ptp10D subfamily and another subfamily, so that synthetic phenotypes would be observed only when both subfamilies are eliminated. To examine this question, we analyzed the triple mutants described above for motor axon phenotypes. Our data show that Ptp4E Ptp10D Lar mutants do not have any new phenotypes, and also do not exhibit phenotypic enhancement relative to Lar single mutants (Table 2). Thus, as for CNS axon guidance, there appear to be no important functions of the Ptp10D/Ptp4E subfamily that are uncovered by removal of Lar.

For the other two triple mutant combinations, however, we did observe phenotypic enhancement by removal of both Ptp4E and Ptp10D. Ptp4E Ptp10D Ptp69D triple mutants have a stronger ISNb phenotype than any of the component double mutants. This is a 'clump' phenotype in which 90% of ISNbs terminate in a darkly staining blob at the dorsal border of muscle 6 (Figures 5l,m). In cases where the ISNb passes this point, it is often misrouted (Figure 5k, right hemisegment), or bypasses the muscle field (Figure 5l) by growing along the ISN. As in bypass phenotypes seen in other mutants, the axons often ectopically project interiorly to muscle 12 from the ISN (Figure 5l, right arrow), suggesting that they retain an affinity for this muscle. Ptp10D Ptp69D double mutants have a similar percentage of ISNb branches that fail to reach muscle 12 (stall phenotype); however, these branches maintain a growth cone-like appearance (Figure 5j).

In almost every affected triple mutant hemisegment, the ISNb clump is at the same point. These data suggest that extension of ISNb axons onto muscle 13 might require a specific signalling event(s) that can be mediated by either a type III RPTP or by Ptp69D. In contrast, for most other published Rptp stall phenotypes (for example, Ptp69D Ptp99A 119) defective ISNbs terminate at a variety of sites, suggesting that loss of the RPTPs produces multiple defects in pathway choice and extension that occur in a probabilistic manner. The differences between the triple and double mutant phenotypes are highly statistically significant (p < 0.0001; Table 2).

We also observed a subtle phenotypic enhancement for SNa defects. We previously showed that Ptp10D Ptp69D double mutants have a SNa defect in which one of the branches is missing (13% penetrance of a completely absent branch). For a smaller percentage of SNas (10%), both branches are missing or the SNa fails to even reach the bifurcation point. In Ptp4E Ptp10D Ptp69D triple mutants, the penetrances of the missing branch phenotypes are similar to those in the double mutant, but an additional 9% of SNa nerves are very thin (Figure 5f, arrow) or completely absent; this phenotype is not seen in Ptp10D Ptp69D (p < 0.0001; Table 2) Also, there are SNa nerves that project along abnormal pathways (Figure 5d).

The other major effect on motor axon guidance we observed in response to loss of Ptp4E was enhancement of a Ptp10D Ptp52F phenotype in which the ISN is truncated at the second branchpoint. This phenotype is observed in about 20% of hemisegments in the double mutant. However, the penetrance of second branchpoint truncation increases to 60% in the Ptp4E Ptp10D Ptp52F triple mutant (p < 0.0001; Figure 5b, Table 2). For the other phenotypes of Ptp10D Ptp52F (SNa bifurcation failure and ISNb stalling), we saw only small increases in phenotypic penetrance when Ptp4E is removed.

Ptp4E¹ was generated by imprecise excision of EP425 34 inserted 1,157 nucleotides upstream of the transcription start. PCR was used to map the deletion endpoints. The deletion removes approximately 10.5 kb, starting from 130 bp upstream of the EP425 insertion site to before the start of exon 3. The excision event introduced chromosomal rearrangement, and as a result there are unrecognizable sequences between the endpoints.

The following Rptp mutant flies were used: Ptp10D¹ 3; Ptp69D¹ and Df(3L)8ex25 1; La^r5.5 and Lar^13.2 2; Ptp52F^18.3 5. Lethal mutations were balanced over GFP balancer chromosomes for sorting homozygous mutant embryos. Double and triple mutant lines were checked for the presence of Ptp4E¹ by PCR. Elav-GAL4 was obtained from Bloomington Stock Center (Bloomington, IN, USA). All crosses were carried out at 25°C.

The UAS-Ptp4E-GFP construct was made by PCR amplification of the entire coding region of the long Ptp4E cDNA, Ptp4E-A, and cloning it into the Gateway System vector (DGRC, Bloomington, IN, USA). The vector used placed a GFP sequence downstream of the Ptp4E coding sequence. The entire construct was sequenced, and also expressed in S2 cells to check for GFP expression to verify that PCR errors were not introduced into the clone. Ptp4E-GFP DNA was injected into embryos using standard methods to generate transgenic flies. Multiple lines were generated and examined; UAS-4E-GFP#4/CyO was used for all studies.

To generate antibodies against the native protein, we made a construct 4E-AP that was expressed in insect cells using a baculovirus expression system. The 4E-AP construct was made by dropping in a 2.7 Psi I/Kpn I fragment from Ptp4E-A (extracellular domain amino acids 169–1073) into a modified baculovirus expression vector pAcGP67-A (BD Biosciences, San Jose, CA, USA). The vector contains a pg67 secretion signal sequence followed by a 6x-His tag, placed upstream of the Ptp4E insertion site. The sequence for human placental alkaline phosphatase was placed in frame, carboxy-terminal to the Ptp4E sequence. The protein expressed consists of the gp67 signal peptide-6xHis-Ptp4E-AP fusion protein that was harvested from insect cell media. Protein was concentrated ten-fold, then purified over a Ni-NTA column (Qiagen, Valencia, CA, USA). Conditions used are: binding buffer (10 mM imidazole, 300 mM NaCl, 20 mM, Tris pH8.0), wash buffer (25 mM imidazole, 300 mM NaCl, 20 mM, Tris pH8.0), and elution buffer (250 mM imidazole, 300 mM NaCl, 20 mM, Tris pH8.0). Purified protein was injected at approximately 100 μg per boost. Antibodies were generated at the Caltech Monoclonal Facility.

Embryos were collected overnight, fixed and hybridized with digoxigenin-UTP antisense RNA probe. Probes were generated from approximately 750 bp PCR product amplified from Ptp4E cDNA that corresponds to the first four exons. Conditions were used as previously described in 35 with some modifications.

Whole-mount antibody staining of stage 16–17 embryos was performed using standard procedures 36, with some variations. Embryos were fixed for 20 minutes with 4% paraformaldehyde. Washes were done in 0.1% PT (0.1% Triton X-100 in phosphate-buffered saline), and blocking buffer used was PT with 0.1% bovine serum albumin + 5% normal goat serum. Double and triple mutant embryos were balanced over GFP balancer chromosomes and homozygous mutants were sorted for lack of GFP expression. Anti-GFP and mAb 1D4 were simultaneously incubated on embryos. Alexa-488 conjugated secondary antibodies were used against anti-GFP, and non-green embryos were sorted under an epifluorescent dissecting microscope. These embryos were then treated with horseradish peroxidase (HRP) immunohistochemistry using standard procedures 36 to visualize 1D4 staining pattern.

In cases where RPTP specific antibodies were used for sorting, staining was done in a sequential manner. Embryos were first incubated with RPTP antibodies, sorted, and then incubated with 1D4. For some of the triple mutant combinations, live embryos were sorted under the epifluorescent dissecting microscope, then dissected live on Superfrost/Plus adhesion slides (Fisher Scientific, Pittsburgh, PA, USA) in phosphate-buffered saline. These live dissected embryos were fixed on the slide, and carried through the same procedures as whole-mount embryo staining procedure.

Primary antibodies were used at the following dilutions: anti-GFP (Invitrogen, Carlsbad, CA, USA) at 1:1,000, mAb 1D4 at 1:3, mouse anti-Ptp4E (polyclonal) at 1:500, mouse anti-Ptp10D (8B2) at 1:3; mouse anti-Ptp69D (2C2) at 1:3, mouse anti-Myc (clone 9E10, Sigma, St. Louis, MO, USA) at 1:250. Secondary antibodies used were: Alexa Fluor 488 and Alexa Fluor 568 (Invitrogen) at 1:1,000, HRP-conjugated goat anti-mouse IgG+IgM (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). Samples were mounted on Vectashield (Vector Laboratories, Burlingame, CA, USA) or glycerol. Samples were photographed on the Zeiss Axioplan microscope using DIC optics. Fluorescent images were taken on Zeiss LSM510, and images were processed using Adobe Photoshop.

Phylogenetic tree construction was performed using the software Geneious Pro, available at 37. The accession numbers of sequences are the following. Ptp4E-PA: AAF45998 (melanogaster), CH940655 (virilis), CH902621 (ananassae). Ptp10D-PA: AAS65319 (melanogaster), CH940655 (virilis), CH902630 (ananassae). Aedes aegypti, AAEL012083. Anopheles gambiae, AGAP004246. Apis mellifera, GB19351.