Drosophila Tempura, a Novel Protein Prenyltransferase α Subunit, Regulates Notch Signaling Via Rab1 and Rab11

A forward genetic screen in Drosophila looking for Notch signaling regulators identifies Tempura, a new and non-redundant protein prenyltransferase of Rab proteins.


Introduction
Notch signaling is an evolutionarily conserved pathway that plays a pivotal role in many developmental processes, including lateral inhibition, binary cell fate determination, and boundary formation [1,2]. Aberrant Notch signaling is implicated in diseases such as Alagille syndrome, spondylocostal dysostosis (SCD), cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), and numerous types of cancer [3,4]. Notch signaling depends on the direct contact between cells: the membrane-bound ligand, Delta (Dl) or Serrate, activates Notch on neighboring cells, resulting in proteolytic cleavages of Notch to generate a Notch intracellular domain (NICD) that activates the transcription of target genes [5].
The developing adult external sensory organs (ESOs) on the Drosophila notum serves as a model system to study lateral inhibition and cell fate determination [6] and have led to the isolation of some Notch signaling components in forward genetic screens [7][8][9][10]. An ESO consists of four cells-shaft, socket, sheath, and neuronwhich are derived from a single mother cell, the sensory organ precursor (SOP or pI) ( Figure 1A and 1A9). Lateral inhibition ensures that only one SOP is selected from a proneural cluster. SOPs undergo several rounds of asymmetric cell division to generate four different cells and Notch signaling activity determines cell fates in each division ( Figure 1A and 1A9). Loss of Notch signaling during lateral inhibition results in a higher density of ESOs, whereas its loss during cell fate determination causes ESO cells to take on neuronal fates, resulting in adult notum balding [11].
Notch signaling activity can be altered by defects in vesicular trafficking [12], coordinated by specific Rabs and their effectors. Rabs are small GTPases belonging to the Ras superfamily of small G proteins, which can switch between GDP-bound inactive and GTP-bound active forms [13][14][15]. Newly synthesized Rabs are prenylated with geranylgeranyl groups on C-terminal cysteines by a protein prenyltransferase (PPT) complex, a process required for their proper membrane localization and hence function [16][17][18][19][20]. PPTs are composed of ab heterodimers and they add prenyl lipids (15-carbon farnesyl or 20-carbon geranylgeranyl groups) to cysteine residues located close to the C-termini of their substrates. Three PPTs have been identified so far: farnesyltransferase (FT), geranylgeranyltransferase I (GGTI), and the Rab geranylgeranyltransferase [RabGGT, also known as geranylgeranyltransferase II (GGTII)] ( Figure S1) [21,22].
Here, we describe the isolation of mutations in a novel PPT a subunit repeat (PPTA) motif containing protein from an unbiased forward mosaic genetic screen for essential genes that affect Notch signaling. Due to its role in adding lipids to its substrates, we name this gene ''tempura'' (temp), based on a Japanese deep-fried dish. Our data show that Temp forms a new PPT to modify a small subset of Rabs, including Rab1, which has not previously been implicated in Notch signaling, and Rab11. Loss of temp leads to aberrant subcellular distribution of Rab1 and Rab11, which in turn leads to Notch signaling defects. In summary, we describe the function of a previously unidentified PPT, provide the first link between Rab1 and Notch signaling, and show that some Rabs are modified by two nonredundant PPTs. These data imply a complex regulation of Rabs by different PPTs that was not previously appreciated.

temp Is Required for Notch Signaling Activity During ESO Development
To identify novel modulators of Notch signaling, we performed a forward genetic screen on the Drosophila X chromosome using ethyl methanesulfonate (EMS) [23]. We induced homozygous mutant clones of essential genes in the notum of otherwise heterozygous mutant animals with the FLP/FRT system [24], and screened for adult notum balding. We identified a novel complementation group named temp, consisting of seven alleles that exhibit a strong balding phenotype ( Figure 1B).
To characterize the ESO development defects in temp mutant clones, we examined lateral inhibition and cell fate determination at different pupal stages. At 12 h after puparium formation (APF), the density of SOPs [marked by Senseless (Sens)] [25] is higher in temp mutant clones than in neighboring wild-type (wt) tissue, indicating a lateral inhibition defect ( Figure 1C). To determine if cell fate specification is impaired, we assessed the expression of Tramtrack (Ttk), a downstream effector of Notch that is upregulated in the signal-receiving pIIa cell but not in the signalsending pIIb cell at 19 h AFP [26,27]. We observed a loss of Ttk in temp mutant pIIa, indicating a loss of Notch signaling during cell fate determination at the two-cell stage ( Figure 1D). At 27 h APF, when the four cells comprising an ESO are specified, many temp mutant ESOs contain multiple neurons (marked by the neuronal marker Embryonic Lethal Abnormal Vision, ELAV, [28] and lack socket cells, marked by Suppressor of Hairless, Su(H)) [29], indicating an alteration in cell fate ( Figure 1A and 1E-F). This phenotype is not fully penetrant as 63% of the temp B mutant ESO cells are ELAV-positive, similar to what we observed previously for dEHBP1 and sec15 [8,9], two players that affect vesicle trafficking and Notch signaling during ESO development. Although we observe some minor apoptosis in some temp A mutant clones (a strong allele; Figure S2A), there is no obvious apoptosis in temp B mutant clones (a less strong allele; Figure S2B). Overexpression of the antiapoptotic protein p35 in temp A mutant clones ( Figure S2C) does not alter the phenotype ( Figure S2D) when compared to temp mutant clones without p35 expression. In addition, there is no decrease in the number of sensory progenitor cells (marked by either Sens or Cut) ( Figure 1C-F). These data indicate that lateral inhibition and cell fate transformation are due to loss-of-Notch signaling. Hence, temp is necessary for proper Notch signaling during the development of ESO lineage.
temp Encodes a Protein with a PPTA Motif Through duplication mapping [30], deficiency mapping [31], and complementation tests with existing lethal P element insertion lines [32], we mapped temp to CG3073 (Figure 2A). This gene encodes a 398 amino acid (a.a.) protein containing a single small 29 a.a. PPTA motif which has only been found to be present in the a subunit of PPTs ( Figure 2B). temp is evolutionarily conserved in most but not all species queried ( Figure 2C), implicating that the function of the temp homolog might be assumed by another protein in species lacking this gene. The vertebrate homolog of temp is named PPTA containing protein 1 (PTAR1), but its biochemical or in vivo function has not yet been characterized.
The lethality and loss-of-Notch signaling phenotypes of temp are rescued by both genomic and cDNA rescue constructs (Figure 2A and 2D). Moreover, the human PTAR1 cDNA can also rescue the ESO developmental defects in temp mutant clones ( Figure 2D). These data indicate that the lethality and balding phenotypes in temp mutants are caused by mutations in CG3073 and that the molecular function of temp is evolutionarily conserved between fly and human. To investigate the endogenous expression pattern of Temp, we attempted to generate several antibodies against Temp, but these were unsuccessful. We therefore examined the expression pattern of Hemagglutinin (HA)-tagged genomic constructs (HA-gtemp) in a temp A homozygous mutant background. We find that Temp is expressed weakly during early ESO development

Author Summary
Notch signaling is an evolutionarily conserved signaling pathway that regulates many developmental processes. Abnormal Notch signaling activity can lead to numerous diseases and developmental defects. To better understand the regulation of this pathway, we performed a forward genetic screen for Notch signaling components that have not been previously identified in Drosophila. Here, we report the identification of an evolutionarily conserved protein, Tempura, which is required for Notch signalingmediated lateral inhibition and cell fate determination of external sensory organs. We show that loss of tempura leads to mistrafficking of Delta and Scabrous, two important Notch signaling components. In addition, Rab1 and Rab11, two major coordinators of vesicular trafficking, are mislocalizaed in tempura mutants. We further show that Tempura functions as a subunit of a previously uncharacterized lipid modification complex to geranylgeranylate (a type of prenylation) Rab1 and Rab11. This post-translational modification is shown to be required for the proper subcellular localization and function of these Rabs. We find that dysfunction of Rab1 causes an accumulation of Delta and Scabrous in the secretory pathway and dysfunction of Rab11 further interferes with the trafficking of Delta. In addition to the known Rab geranylgeranyltransferse, our data indicate the presence of another functionally nonredundant Rab geranylgeranyltransferse, Tempura.
( Figure 2E) and that it is somewhat enriched in the cytoplasm of ESOs at the four-cell stage ( Figure 2E9). HA-gtemp is expressed weakly and dispersed throughout the cytoplasm. Similarly, when we overexpress HA-temp cDNA using dpp-Gal4 in the wing disc, HA-Temp is also diffuse in the cytoplasm ( Figure S3).

temp Is Required for Scabrous Secretion
The elevated SOP density in temp mutant clones ( Figure 1C) is similar to what has been observed in scabrous (sca) mutants ( Figure 3A) [33]. Sca is secreted by the SOP to facilitate Notch signaling in nearby cells to promote lateral inhibition [33][34][35][36]. It is rapidly secreted and degraded upon synthesis [37] and therefore is very difficult to detect in wt tissue on the notum ( Figure 3B; GFPpositive cells). Interestingly, Sca is strongly up-regulated in temp mutant ESOs ( Figure 3B; GFP-negative cells). A similar elevation of Sca in sensory organs is also observed in developing wing and eye imaginal discs ( Figure S4A-B9), indicating that this elevation is not limited to the notum. Given that the lateral inhibition defect in temp mutant clones is similar to that of loss-of-function of sca, we hypothesized that Sca is produced, but that it fails to be secreted, and hence accumulates in temp mutant ESOs. To test this idea, we first determined whether the up-regulation of Sca in temp mutant is transcriptional or posttranscriptional. The expression of the sca-lacZ reporter [36,38], a readout for sca transcription, is similar between temp mutant and wt ESO ( Figures 3C and S4C-D), indicating that the level of Sca in temp mutant ESOs is posttranscriptionally up-regulated. To determine if Sca secretion is impaired, we developed a secretion assay by overexpressing a Sca-GFP fusion protein [39] in wt and temp mutant clones using the mosaic analysis with a repressible cell marker (MARCM) [40]. Sca-GFP can be secreted from wt clones into the neighboring area that does not produce Sca-GFP ( Figure 3D). However, Sca-GFP produced in temp mutant clones fails to be secreted into the neighboring area ( Figure 3E), indicating defective Sca secretion.
To determine where Sca accumulates intracellularly, we performed coimmunostaining of Sca and an array of subcellular organelle markers (Table S1). The Sca-positive puncta mainly colocalize with GM130 [41], a cis-Golgi marker, but not with Syntaxin 16 (Syx16) [42], a trans-Golgi marker ( Figure 3F-G9). Therefore, in temp mutant cells, Sca accumulates in a GM130-positive compartment and cannot be secreted, which in turn contributes to the defect in Notch-signaling-mediated lateral inhibition.

temp Is Required for Proper Localization of dEHBP1 and Dl
Since Sca is primarily involved in the lateral inhibition process, other proteins are likely to contribute to the cell fate specification defect in temp mutant clones ( Figure 1E and 1F). Cell fate determinants, including the adaptor protein Numb [26,43] and the E3 ligase Neuralized [44][45][46], are asymmetrically segregated during the division of the pI cell to bias Notch signaling between the pIIa and pIIb. We did not observe obvious defects in the localization of either protein, suggesting that asymmetric segregation of cell fate determinants is not affected (not shown). Although the expression and localization of Notch is not affected ( Figure  S5A-C), the number of Dl-positive puncta is increased in temp mutant sensory organs ( Figure 4A and 4A9). We previously proposed that recycled Dl travels to the apical actin-rich structure (ARS) localized between the pIIa and pIIb at the two-cell stage ( Figure 4B) [10]. This process is necessary for proper Notch signaling activation and requires the Arp2/3 complex as well as the vesicle trafficking regulators Sec15 and dEHBP1, two binding partners of Rab11 [8,9,47,48]. Mutations in these genes have been shown to exhibit cell fate defects and notum balding, similar to the loss of temp. While the ARS is properly formed (not shown) and apical-basal polarity is not affected in temp mutant clones ( Figure  S5E), we found that dEHBP1 accumulates basally ( Figure 4C), similar to what is observed in sec15 mutants [8]. In dEHBP1 mutant ESOs, the level of Dl is reduced at the cell surface [8], a feature that we also observe in about half (yellow arrows) of the ESOs in temp mutant clones ( Figure S5D). These data indicate that Dl accumulates intracellularly and that the mislocalization of EHBP1 may at least partially contribute to the Dl trafficking defect in temp mutant ESOs. Additionally, we found that many of the Dl and Sca puncta colocalize in temp mutant ESO, indicating that some of them are trapped in the same intracellular compartments ( Figure 4D and 4D9). The majority of colocalized proteins are in the Golgi complex ( Figure 4E, blue arrows). However, some of the Dl-and Sca-positive puncta colocalize with a late endosomal and lysosomal marker LAMP1-GFP ( Figure 4F), suggesting that defects in the secretory pathway may cause some Sca and Dl to be sorted to the endo-lysosomal pathway for degradation. Therefore, mistrafficking of Dl is also likely to contribute to both lateral inhibition and cell fate defects in temp mutants.

Temp Is a New a Subunit of RabGGT Complex
Protein prenylation regulates protein targeting and activity of numerous proteins [21,22]. Given that Temp contains a PPTA motif and that temp mutants affect protein trafficking, Temp may regulate the prenylation of proteins involved in vesicular trafficking. Indeed, RabGGTb was identified as a potential interactor for Drosophila Temp in a high-throughput yeast-2 hybrid screen [49]. RabGGTb forms a complex with RabGGTa [also called PPTA containing protein 3 (PTAR3)] and Rab escort protein (REP) to geranylgeranylate Rabs ( Figure 5A) [16,[50][51][52][53][54], which are major coordinators of vesicle trafficking [13,14]. We therefore tested if Temp may act as an alternative a subunit of the RabGGT complex ( Figure 5B). Indeed, we found that Temp can interact with RabGGTb and REP in coimmunoprecipitation (coIP) experiments in Drosophila S2 cells ( Figure 5C and 5D). In the presence of RabGGTb, the expression level of Temp is increased, suggesting that Temp is stabilized by RabGGTb. To assess whether Temp is an additional subunit of the original RabGGT complex or an alternative a subunit in a new RabGGT complex, we performed a RabGGTb competition assay ( Figure 5E). We pulled down the same amount of RabGGTb and found that its binding to Temp is reduced when the amount of RabGGTa increases, indicating that Temp and RabGGTa compete for RabGGTb. Note that the total level of Temp is also decreased when it cannot bind to RabGGTb. These data suggest that RabGGTb can form two RabGGT complexes: a canonical complex with RabGGTa and a novel complex with Temp.

Impairing Rab1 and Rab11 Functions Phenocopies Loss of temp
Rabs are the only known substrates of the canonical RabGGT complex (RabGGTa-RabGGTb) [16,21,55]. Hence, the targets of the Temp-RabGGTb complex may also be Rabs. To identify the substrate Rabs responsible for the loss-of-Notch signaling phenotypes in temp mutants, we performed a genetic screen and overexpressed a subset of UAS-YFP dominant-negative Rab (DN-Rab) lines [56] in wing imaginal discs and pupal notum to screen The lethality and phenotype of temp mutant is rescued by Dp(1;2;Y)w+ (cytological regions 2C10-3C10) and fails to be complemented by Df(1)ED6574 and Df(1)ED409 narrowing the candidate region to 2E1 and 2F5. P{lacW}l(1)G0144 [72], a P-element insertion (maroon triangle) in CG3073, fails to complement all temp alleles. The orange region indicates the extent of the genomic rescue constructs tagged with mCherry-or HA-tag (red triangle). (B) After sequencing the genomic region to which the temp alleles are mapped, we identified both nonsense mutations (temp A , temp B , temp C , temp D , and temp G ) and missense mutations (temp E and temp F ) within CG3073. This gene encodes a 398 a.a. protein containing a conserved PPTA subunit repeat motif (shown in pink). (C) temp homologs are conserved in many species queried. The percentage in the table is obtained via bl2seq. (D) Both lethality and balding phenotypes in all temp alleles (except for temp E ) can be rescued by genomic and cDNA rescue constructs of CG3073. Because the lethality and balding of temp E can be rescued by the duplication, Dp(1;2;Y)w + (cytological regions 2C10-3C10) but not gtemp, this indicates that temp E carries a second lethal mutation in the 2C10-3C10 interval. Based on lethal phase analysis, the temp A , temp D , and temp G with early nonsense mutations are the most severe alleles. +, rescued; 2, not rescued; ND, not determined. *Overexpression of human homolog of temp, hPTAR1, can rescue the balding phenotype in temp mutant clones on the notum, but its ubiquitous overexpression is toxic in flies even in the wt background. (E-E9) HA-tagged genomic temp (HA-gtemp) is expressed very weakly in the cytoplasm on the pupal notum during early ESO development (12 h APF) (E) and becomes slightly enriched in the sensory organs at the later stage (27 h APF) (E9). Scale bars, 5 mm.
for Sca accumulation and balding phenotypes similar to temp mutant clones, respectively (Table S2). Because Sca secretion defects in temp mutants are not restricted to the notum, we tested Rabs that are broadly expressed [57] and those involved in protein secretion [14].
Among the 15 Rabs tested for Sca accumulation, only overexpression of DN-Rab1 reproduces key features of temp mutant cells. It causes a strong Sca accumulation in ESO in third instar wing discs (Table S2), consistent with what we observed in rab1 homozygous mutant clones ( Figure S6A) [58]. We also observe a co-accumulation of both Dl and Sca in DN-Rab1-expressing ESOs on the pupal notum ( Figure S6B). Similar to temp mutants, large Sca puncta colocalize with enlarged GM130 compartments [59] when we overexpress DN-Rab1 ( Figure 6A-A0). Most importantly, Rab1 accumulates in enlarged GM130-positive compartments in temp mutant clones ( Figure 6B and 6B9), suggesting that Temp regulates the subcellular distribution of Rab1. Together, these data suggest that Rab1 may be a substrate for Temp. Therefore, loss of temp leads to aberrant localization of Rab1 and its effector GM130, which in turn causes an accumulation of Sca and Dl.
Because dysfunction of Rab1 only leads to minor adult balding (Table S2 and Figure S6C), the cell fate transformations observed in temp mutant cells may be contributed by the dysfunction of other Rab(s). Among the 10 broadly expressed Rabs [57], we found that expression of DN-Rab5 and DN-Rab11 lead to strong balding (Table S2). Because we observed aberrant subcellular localization of Rab11 ( Figure 6C), but not Rab5 (not shown), in temp mutant clones, we focused on Rab11, which functions in Dl recycling during ESO development [9,47]. In temp mutant cells, Rab11 is enriched apically ( Figure 6C). This apical clustering is likely due to mislocalization of Rab11 because the protein level of Rab11 is not changed in temp mutant animals ( Figure S6D).. Cells that express DN-Rab11 exhibit an accumulation of Dl but not of Sca ( Figure  S6E and S6F), indicating that Rab11 regulates trafficking of Dl but not of Sca. In addition, dEHBP1, a Rab11 binding partner [8], aberrantly accumulates basally in DN-Rab11-expressing notum ( Figure 6D), similar to temp mutants ( Figure 4C). Therefore, loss of temp causes an aberrant localization of Rab11, which leads to a mislocalization of dEHBP1 and mistrafficking of Dl. These data indicate that Rab11 may also be a substrate of Temp.
Because temp mutant clones exhibit an altered distribution of both Rab1 and Rab11, we tested whether they are misdistributed to the same intracellular compartment. Coimmunostaining of Rab1 and Rab11 reveals that they mostly do not overlap in temp mutant clones ( Figure 6E), suggesting that additional factors regulate their aberrant subcellular distribution in the absence of temp.

Rab1 and Rab11 Are Substrates of the Temp-RabGGTb Prenyltransferase Complex
As shown previously, Temp can form a new RabGGT complex with RabGGTb and interact with REP ( Figure 5B-D). If Rab1 and Rab11 are substrates of this complex, they should physically interact with Temp. Indeed, Temp binds to Rab1 and Rab11 in coIP experiments in S2 cells ( Figure 7A and 7B, last lanes). Temp and these Rabs can also be pulled down without cotransfecting RabGGTb or REP, possibly because of the presence of these proteins in S2 cells. We knocked down the REP and RabGGTa proteins in S2 cells using RNAis that we designed. Unfortunately, these cells grow very slowly and most die, suggesting that REP and RabGGTa are required for the viability of S2 cells. These data are in agreement with our observation that RabGGTa mutant cells are lethal in vivo (unpublished data). To determine whether Temp can prenylate Rab1 and Rab11 with geranylgeranyl groups, we performed an in vitro prenylation assay [60]. We cotransfected REP, RabGGTb, and HA-Temp or HA-RabGGTa into S2 cells and pulled down the enzyme complex with anti-HA beads ( Figure 7C and 7D, left panels). We also purified GST-Rab1 and GST-Rab11 as unmodified substrates from bacteria. The prenylation assays were performed by adding GST-Rab1 or GST-Rab11 to biotin-labeled geranylgeranyl groups and the anti-HA beads loaded with the enzymatic complex (HA-Temp-RabGGTb-REP or HA-RabGGTa-RabGGTb-REP). In the absence of Temp and RabGGTa, we observe weakly prenylated bands, likely due to nonspecific pull-down of endogenous RabGGT complexes ( Figure 7C and 7D). However, in the presence of Temp or RabGGTa (positive control) the prenylated band is obviously enhanced ( Figure 7C and 7D, right panels). These data indicate that the complexes containing Temp can prenylate Rab1 and Rab11. Temp seems to be more efficient in prenylating Rab1, whereas RabGGTa seems to be more efficient in prenylating Rab11, indicating that they may have different substrate preferences. In conclusion, our data show that Temp can form a novel PPT complex to prenylate Rab1 and Rab11.

Discussion
We isolated a novel Notch signaling player, temp, whose loss causes defects in lateral inhibition and cell fate determination of ESO from an unbiased genetic screen. temp encodes an unstudied protein with a 29 a.a. PPTA motif implicated in protein prenylation. Here, we show that Temp forms a complex with RabGGTb and REP to prenylate Rab1 and Rab11. Loss of temp leads to an aberrant subcellular distribution of Rab1 and Rab11. Loss-of-function of Rab1 causes an accumulation of Sca and Dl, whereas loss-of-function of Rab11 further contributes to the Dl trafficking defect. Together, our data indicate that Temp functions as the a subunit of a new RabGGT complex that modulates Notch signaling via Rab1 and Rab11.
Although the role of Rab11 in Notch signaling pathway has been previously documented [47], our data indicate that Rab1 is required for proper trafficking of Sca and Dl. To our knowledge, this is the first time that Rab1 has been linked to Notch signaling. Sca trafficking is mainly affected by Rab1, whereas Dl trafficking is affected by both Rab1 and Rab11. Because null alleles of both Rabs are cell lethal (not shown), but temp mutant clones are not, the loss of Rab1 and Rab11 function can only be partial in the temp mutant clones. This suggests that the other a subunit, RabGGTa, still prenylates a portion of Rab1 and Rab11 that play a role in other cellular processes in the absence of temp, consistent with the prenylation assay results ( Figure 7C and 7D). In addition, overexpression of constitutively active (CA) forms of Rab1 and Rab11 (Rab1CA and Rab11CA) in temp mutant clones does not alter Sca accumulation ( Figure S7A and S7B) nor does it alter cell fate transformation ( Figure S7C and S7D). This suggests that loss of temp is epistatic to the constitutively active forms of Rabs. These  genetic interaction data are consistent with our model as constitutively active forms of Rab must be properly prenylated [18].
In temp mutant clones and notum cells expressing DN-Rab1, Sca mostly accumulates in GM130-positive compartments, traditionally considered as cis-Golgi. However, because Rab1 functions in ER-to-Golgi trafficking [13,14], loss of Rab1 is expected to prevent cargo from entering the Golgi apparatus. This may be because the cis-Golgi is closely associated with the ER exit sites (tER) to form the ''tER-Golgi unit'' [61], which is optically difficult to distinguish from the cis-Golgi compartment. Indeed, a number of Golgi proteins including GM130, GRASP, and p115 localize at tER sites in S2 cells [61][62][63]. For example, GRASP also colocalizes with Sca in temp mutants ( Figure S4E). Therefore, loss of temp or its target, Rab1, causes Sca to accumulate in GM130-positive compartments corresponding to the cis-Golgi, the tER, or an intermediate compartment between the two.
In yeast and cultured mammalian cells, Rabs that lack proper geranylgeranylation diffuse in the cytoplasm [17,18]. Surprisingly, we find that the distribution of Rab1 and Rab11 in temp mutants is restricted to specific subcellular compartments: Rab1 colocalizes with enlarged GM130 compartments ( Figure 6B) whereas the majority of Rab11 is apically enriched ( Figure 6C). Given that Rab1 and Rab11 do not diffuse in the cytoplasm and that they do not redistribute to the same microdomains ( Figure 6E), we propose that the subcellular localization of Rab1 and Rab11 is in part determined by different/other Rab binding partners upon reduction of proper prenylation in temp mutants.
The fly and human genomes both contain three genes that encode proteins containing the PPTA motif: PTAR1/Temp, PTAR2/FTa, and PTAR3/RabGGTa. PTAR2 forms PPTs with two different b subunits, whereas PTAR3 forms a RabGGT with RabGGTb ( Figure S1) [21,22,[52][53][54]64]. We show that PTAR1 and RabGGTb form an alternative RabGGT complex to prenylate Rab1 and Rab11. This raises an interesting question: what is the labor distribution of RabGGTa-RabGGTb and Temp-RabGGTb complexes? Because the PTAR1 (temp) homolog is absent in some species like S. cerevisiae and C. elegans, but is present in Dictyostelium and Arabidopsis ( Figure 2C), it is likely that it was lost in some evolutionary branches and that its function is covered by PTAR3 in these species. While we can obtain large homozygous mutant clones with temp null alleles, we find that homozygous RabGGTa mutant clones in the thorax and eyes are cell lethal (not shown). Hence, we speculate that Temp has evolved to play a more specific role to prenylate a subset of Rabs, whereas RabGGTa is able to modify most, if not all, Rabs [16,21,55] yet is not sufficient for proper trafficking of Scabrous and Delta. Indeed, although both Temp and RabGGTa can prenylate Rab1 and Rab11 in vitro with different efficiency (Figure 7C and 7D), expression of fly RabGGTa fails to rescue the ESO phenotypes in temp mutant clones ( Figure 2D). Moreover, expression of fly temp does not alleviate the cell lethality in RabGGTa mutant clones (data not shown). These data indicate that the functions of the two RabGGT complexes are nonredundant in vivo and that the functions of the two RabGGT complexes towards different Rabs may also be regulated in a tissue-specific manner through unknown interaction partners and/or posttranslational modifications in vivo. This tissue-specific regulation is supported by the gene expression data from FlyAtlas ( Figure S8) [65,66]. RabGGTa mRNA is transcribed ubiquitously at moderate levels and the expression pattern of temp mRNA is much more restricted to the nervous system with high levels in thoracic-abdominal ganglion cells. This suggests that Temp plays a role in the nervous system, including ESO development.
As major coordinators of vesicular trafficking, Rabs are crucial for maintaining normal cellular function and misregulation of some Rabs results in cellular dysfunction. Indeed, dysfunction of some Rabs and their prenylation factors are implicated in several diseases [13,55,67]. For example, Choroideremia, an inherited retinal degenerative disease, is caused by mutations in REP1. On the other hand, Rab1 is hijacked by the pathogen Legionella pneumophila during infection to support the bacterium with the ERto-Golgi secretory system [68]. In various cancers, the expression level of numerous Rabs, including Rab1 and Rab11, is upregulated. Up-regulation of Rab11 family members (Rab11A/ 11B/25) is associated with more aggressive prostate, ovarian, and breast cancers [13,55]. Finally, toxins produced by Bacillus anthracis inhibit Rab11 and Sec15, which in turn reduce Notch signaling activity in both flies and mammalian endothelial cells [69], suggesting a possible role for Temp in aberrant Notch signaling induced by bacterial infection.

Fly Strains, Transgenesis, and Crossing Schemes
The following stocks were used in this study: (1) isogenized y w FRT19A (y w FRT19A iso or iso19A), The forward genetic screen that identified alleles of temp was performed as previously described [23]. A total of 577 stocks have wing notching or notum balding phenotypes and were mapped through X chromosome duplication mapping, deficiency mapping, and complementation tests.
Genomic and cDNA rescue transgenic flies were generated by phiC31-mediated transgenesis at vk33 or vk37 docking sites [74]. Phenotypic rescue was assessed in temp mutant background either in whole animal or by overexpression in temp MARCM clones. UAS-YFP-DN-Rab flies [56] were crossed to C96-Gal4 [75], which and Rab11 (B). The interaction can occur without the transfection of RabGGTb and REP. This is possibly due to the endogenous expression of these two proteins in S2. The very weak pulled down band of Rab1/Rab11 in the absence of Temp is due to nonspecific binding to a-HA beads. (C-D) In vitro prenylation assays: (C, Left) Western blot result of input and IP for the prenylation assay. Because the whole pull-down process is performed in the prenylation buffer, which is not an optimal condition for IP, some nonspecific binding is present in the control. (C, Right) GST-Rab1 can be prenylated with geranylgeranyl groups in the presence of Temp and RabGGTa. The prenylation efficiency for Rab1 is much higher with Temp than with RabGGTa. The weak band in the control condition might be due to endogenous activity. (D, Left) Western blot result of input and IP for the prenylation assay. (D, Right) GST-Rab11 can be prenylated with geranylgeranyl groups in the presence of Temp and RabGGTa. The prenylation efficiency for Rab11 is higher with RabGGTa than with Temp under this substrate concentration. The weak band in the control condition might be due to endogenous activity. can drive ectopic expression of UAS transgenes around the wing margin, and used for immunostaining against Sca in the third instar larval wing disc. We overexpressed UAS-GFPnls as a negative control.
We used the MARCM strategy in the balding screen and to assess Sca secretion and mark subcellular compartments. Because these transgenes encode GFP/YFP/CFP fusion proteins, we crossed out Actin-Gal4, UAS CD8::GFP in MARCM line, which was then combined with the transgene of interest and subsequently crossed to either y w FRT19A iso ; Actin-Gal4/Cyo or y w temp FRT19A; Actin-Gal4/Cyo (temp A and temp D ) for performing experiments in control and mutant genetic backgrounds, respectively. Stocks were maintained at RT and crosses were performed at 25uC.

Genomic and cDNA Constructs
A genomic rescue construct was constructed by PCR amplification of a 4.5 kb amplicon spanning CG3073 locus and cloned into pattB [76]. We added N-terminal HA or mCherry tags to the genomic rescue construct via conventional cloning methods. cDNA of RabGGTb was constructed in the pMT vector, while human PTAR1(hPTAR1), temp, RabGGTa, REP, rab1, and rab11 were cloned into pUASTattB [76] with N-terminal HA and FLAG tags. In addition, rab1 and rab11 were cloned to pGEX vectors (GE Healthcare). Cloning and DNA purification were performed based on standard protocols. Enzymes are from NEB, and DNA purification kits are from Invitrogen and Qiagen. All constructs were sequenced before injection or transfection.
For Sca secretion assays, we assess extracellular Sca-GFP by staining for rabbit a-GFP without any detergents. Then, we permeabilize the notum tissue with 0.1% Triton-PBS and perform a staining for the total GFP using chicken a-GFP antibody. To costain both extracellular and total Dl in temp mosaic notums, we first stain extracellular Dl using mouse a-Dl antibody without detergents. Then we permeabilize the noum with 0.1% Triton-PBS and stain for the total Dl with the guinea pig a-Dl antibody. For Notch extracellular staining, the notum tissue is first stained with mouse a-NECD antibody without any detergents. Then, we permeabilize the notum tissue with 0.1% Triton-PBS and perform the later steps of immunostaining similarly.

Rab Protein Purification and In Vitro Rab Prenylation Assay
GST, GST-Rab1, and GST-Rab11 were purified from BL21 pLys (Invitrogen): 25 ml overnight cultures were diluted to 250 ml with LB medium and kept growing at 37uC until OD 600 reached 0.5,0.7. GST, GST-Rab1, and GST-Rab11 proteins were induced at 37uC for 3 h by adding IPTG at a final concentration of 0.1 mM. Bacteria were then lysed using CelLytic express (Sigma-Aldrich) according to the manufacturer's instructions. Supernatant was incubated with Glutathione Sepharose 4B (GE healthcare) at 4uC for 2 h and then the beads were washed with 50 mM HEPES pH 8.0 buffer by inversion for three times. GST fusion proteins were eluted with 10 mM glutathione in 50 mM HEPES pH 8.0 buffer. Protein concentration was measured using Bradford method (BioRad). We used 0.3 mM GST-Rab1 and 3 mM GST-Rab11 as substrates in the prenylation assays with 0.3 mM and 3 mM GST as negative controls, respectively. On the other hand, all components of the RabGGT complex (His-RabGGTb-V5, FLAG-REP, and HA, HA-Temp, or HA-RabGGTa) were transfected to S2 cells as described in the previous section. Cells were lysed on ice in prenylation buffer [50 mM HEPES, pH 7.2, 50 mM MaCl, 2 mM MgCl 2 , 0.01% Triton X-100, and EDTA-free cocktail complete protease inhibitor (Roche)] [60] using syringes. Supernatant was then incubated with EZ red a-HA beads (Sigma-Aldrich) for 4 h at 4uC, and beads were washed with prenylation buffer. We added HA, HA-Temp, or HA-RabGGTa binding beads together with GST control or GST-Rab substrates, 5 mM biotin-labeled lipid precursor (B-GPP; Jena Bioscience), 2 mM DTE, and 20 mM GDP. The prenylation assays were carried out on beads at 25uC for 1 h. Reactions were stopped by adding 26 Laemmli buffer. Western blot was performed as described in the previous section with 5% BSA as blocking solution and Streptavidin-HRP 1:50,000 (Jackson ImmunoResearch) for Biotin detection.

Supporting Information
Figure S1 Schematic diagram and features of the three well-characterized PPTs. PPTs are heterodimers of a and b subunits. They can be subdivided into three types: FT, GGTI, and GGTII (also called RabGGT) based on the lipid length and target motif present on their substrate proteins. All these genes are conserved from yeast to human. Because the a subunits of PPTs contain PPTA subunit repeat motifs, the FTa is also called PTAR2 and RabGGTa is also called PTAR3. PPTs add prenyl lipids (15-carbon farnesyl or 20-carbon geranylgeranyl group) via a thioether linkage (-C-S-C-) to a cysteine residue of soluble proteins. FT and GGTI share the same a subunit for substrate recognition and have a unique b subunit. They recognize the CaaX motif (C for cysteine residue, a for aliphatic residues, X can be a wide range of residues) in the C-terminal region of various substrates, including lamins as well as many small GTPases such as Ras and Rho. On the other hand, GGTII does not require a specific C-terminal motif for recognition. Because all of its substrates identified to date are Rab proteins, it is also called RabGGT. It requires assistance from REP to recruit Rab substrates. The target motif on most of the Rab proteins contains two Cysteine residues, and both are conjugated with geranylgeranyl groups.  Figure S8 temp mRNA expression is highly enriched in the nervous system. The mRNA expression patterns of temp (CG3073) and RabGGTa (CG12007) are quite different in both larval and adult stages based on FlyAtlas data [65]. RabGGTa mRNA is transcribed at moderate level ubiquitously and temp mRNA is expressed highly in the nervous system, suggesting that Temp plays a role in the nervous system. This figure is adapted and modified from Flybase (http://flybase.org/) [66]. (TIF) Table S1 Sca accumulates in GM130-positive compartments in both temp mutant and Rab1DN-expressing notum clones. We used MARCM to express ER-YFP, Golgi-CFP, FYVE-GFP, and LAMP-1-GFP in temp mutant or wt clones. For other markers, we performed coimmunostaining using specific antibodies (see Materials and Methods). ER, endoplasmic reticulum. +, colocalization; +/-, minor colocalization; -, no obvious colocalization; N/A, not tested. (TIF)