The SUMO protease Verloren regulates dendrite and axon targeting in olfactory projection neurons.
ABSTRACT Sumoylation is a post-translational modification regulating numerous biological processes. Small ubiquitin-like modifier (SUMO) proteases are required for the maturation and deconjugation of SUMO proteins, thereby either promoting or reverting sumoylation to modify protein function. Here, we show a novel role for a predicted SUMO protease, Verloren (Velo), during projection neuron (PN) target selection in the Drosophila olfactory system. PNs target their dendrites to specific glomeruli within the antennal lobe (AL) and their axons stereotypically into higher brain centers. We uncovered mutations in velo that disrupt PN targeting specificity. PN dendrites that normally target to a particular dorsolateral glomerulus instead mistarget to incorrect glomeruli within the AL or to brain regions outside the AL. velo mutant axons also display defects in arborization. These phenotypes are rescued by postmitotic expression of Velo in PNs but not by a catalytic domain mutant of Velo. Two other SUMO proteases, DmUlp1 and CG12717, can partially compensate for the function of Velo in PN dendrite targeting. Additionally, mutations in SUMO and lesswright (which encodes a SUMO conjugating enzyme) similarly disrupt PN targeting, confirming that sumoylation is required for neuronal target selection. Finally, genetic interaction studies suggest that Velo acts in SUMO deconjugation rather than in maturation. Our study provides the first in vivo evidence for a specific role of a SUMO protease during neuronal target selection that can be dissociated from its functions in neuronal proliferation and survival.
proteases are required for the maturation and deconjugation of SUMO proteins, thereby either promoting or reverting sumoyla-
tion to modify protein function. Here, we show a novel role for a predicted SUMO protease, Verloren (Velo), during projection
lobe (AL) and their axons stereotypically into higher brain centers. We uncovered mutations in velo that disrupt PN targeting
specificity. PN dendrites that normally target to a particular dorsolateral glomerulus instead mistarget to incorrect glomeruli
within the AL or to brain regions outside the AL. velo mutant axons also display defects in arborization. These phenotypes are
rescued by postmitotic expression of Velo in PNs but not by a catalytic domain mutant of Velo. Two other SUMO proteases,
DmUlp1 and CG12717, can partially compensate for the function of Velo in PN dendrite targeting. Additionally, mutations in
required for neuronal target selection. Finally, genetic interaction studies suggest that Velo acts in SUMO deconjugation rather
than in maturation. Our study provides the first in vivo evidence for a specific role of a SUMO protease during neuronal target
selection that can be dissociated from its functions in neuronal proliferation and survival.
tion neuron (PN) targets its dendrites specifically to one of 50
a specific class of olfactory receptor neurons (Jefferis et al., 2001).
centers (Marin et al., 2002; Wong et al., 2002; Jefferis et al., 2007).
Previous studies have identified several molecules, including
transmembrane receptors and transcription factors, acting cell-
autonomously to regulate PN dendrite targeting specificity
al., 2009). However, it has been suggested that misregulation of
post-translational modifications may also lead to defects in PN
targeting (Tea et al., 2010). In a genetic screen for molecules that
regulate PN dendrite wiring specificity, we discovered a previ-
ously uncharacterized Small Ubiquitin-like Modifier (SUMO)
protease that we named Verloren (Velo).
Sumoylation is a reversible post-translational modification that
occurs in numerous cellular proteins to inhibit, modify or enable
protein-protein interactions, thereby modulating protein localiza-
tion and function. The functional consequences of SUMO attach-
involved in the nuclear trafficking of cytosolic proteins (Geiss-
Friedlander and Melchior, 2007). SUMO conjugation is a highly
dynamic process and can be rapidly reversed by SUMO proteases.
newly synthesized SUMO precursors. SUMO proteases are highly
conserved across species and are evolutionary divided into two
and Ulp2-related proteases (Hay, 2007). Yeast Ulp1 possesses a hy-
activity essential for SUMO deconjugation (Li and Hochstrasser,
1999). Ulp2 possesses only the isopeptidase activity (Li and
Although sumoylation controls many basic cellular processes
like transcriptional regulation, chromatin organization, replica-
tion and repair, it is also a key determinant in many neuronal
processes (Martin et al., 2007; Scheschonka et al., 2007). Sumoy-
lation plays a role during neuronal development and function.
For example, desumoylation of the Monocyte Enhancer Factor
2A (MEF2A) transcription factor leads to MEF2A activation and
the inhibition of synapse formation (Shalizi et al., 2006). In ad-
TheJournalofNeuroscience,June13,2012 • 32(24):8331–8340 • 8331
ling neuronal activity (Feligioni et al., 2009). Sumoylation is not
restricted to the nucleus but can also occur in the cytoplasm and
al., 2004; Dorval and Fraser, 2007).
vivo function for the SUMO protease Velo in controlling neuro-
nal targeting and morphogenesis. Furthermore, our genetic in-
teraction studies suggest that Velo regulates PN target selection
via the deconjugation of SUMO protein(s).
Fly stocks. The insertions LL05207, LL05209 and e01260 in velo originate
from two piggyBac collections (Thibault et al., 2004; Schuldiner et al.,
2008). Information for all other mutant alleles used can be found in
Flybase (http://flybase.bio.indiana.edu). Mosaic analysis with a repressi-
ble cell marker (MARCM) was performed as previously described using
Gal4-GH146 (Wu and Luo, 2006a).
the relationship between budding yeast (Sc), human (Hs), and Drosophila
(Dm) homologs of SUMO protease proteins. The number shown at each
Its corresponding sequence was isolated, aligned using Clustal W multiple
protein sequence alignment, and compared. PHYLIP (PHYLogeny Infer-
ence Package) v3.69 was used to generate the tree through bootstrapping
analysis. This requires using the SEQBOOT program to generate boot-
strapped dataset based on the parsimony method, and the CONSENSE
gram to root the tree with HsSENP8/NEDP1 yielding the final tree.
All programs in the PHYLIP package are freely available online
Plasmid and transgene construction. To generate UAS-velo-HA (long), a
5502 base pair (bp) fragment was amplified from a full-length cDNA
To generate UAS-DmUlp1-HA, a 4540 bp fragment was amplified
using a full-length cDNA (GH15225) as a template and the following
primers (5?-3?): CACCATGTCGCTGCCTCCCGAGGAC and CTGC-
2044 bp fragment was amplified using genomic DNA from w1118flies as
a template (no introns are present in CG12717) and the following prim-
ers (5?-3?): CACCATGGATCGCAAAGAAACTG and TTTGAGTG-
TATTCCTTCTCCTCGG. We found that the published sequence of the
and recombined into pTWH (Gateway Collection, Drosophila Genom-
ics Resource Center) using the Gateway LR clonase II enzyme mix (In-
vitrogen). The UAS-velo-HA #108L (long), UAS-DmUlp1-HA #5.1, and
UAS-CG12717-HA #4 transgenes, all inserted on the second chromo-
some and generated by standard injection techniques, were used for
For Velo expression in Drosophila S2 cells, VeloWT and VeloC?S
cDNAs were recombined from pENTR-D/TOPO into pAWH (Gateway
Collection, Drosophila Genomics Resource Center) by LR reaction.
In vitro mutagenesis. In vitro mutagenesis was performed according to
manufacturer’s instructions (QuikChange Site-Directed Mutagenesis Kit,
Stratagene) to generate a point mutation in the velo transgene to generate
UAS-veloC?S-HA using the primers: GAACAACTTCACCGATAGCGGC
CTGTATCTGCTGC and GCAGCAGATACAGGCCGCTATCGGTGAA
amino acid 1624 of the long velo transcript resulting in a cysteine to serine
change, disrupting the catalytic domain of Velo. A UAS-veloC?S-HA #1.2
tant CG12717ex#6was generated by FLP-mediated excision described by
Parks et al. (2004) using the flanking P-element P(XP)d05069 and the
piggyBac insertion pBac(RB)e01706a. 40 single white females were
screened by PCR to confirm the deletion and determine its extent using
the primers: AATGATTCGCAGTGGAAGGCT (XP 5?plus, left) and TG
CAAATTTCAAAGCGGACTATCG and AGATCCATGTGTTTGCGC
SUMO imprecise P-element excision. The P-element P(lacW)2(l)SH0182
SUMO gene and used to perform an imprecise excision. 100 single white
males were screened by PCR to determine the extent of genomic deletions
CAGAAGTTCAAGGTGG or TATGGTGCTGAGTCATGGTGAGAC.
that removes the entire open reading frame.
Immunostainings. Confocal images were taken on a LSM510 confocal
microscope (Zeiss). Fly brains were dissected, fixed and stained as de-
rat-anti-mCD8 1:100 (Invitrogen Caltag MCD0800), mouse nc82 1:40
(DSHB #nc82; E. Buchner, University of Wu ¨rzburg, Wu ¨rzburg, Ger-
many), rabbit anti-HA 1:100 (Abcam) and mouse anti-HA 1:500 (gift
from K. Wehner, Stanford University, Stanford, CA).
mented with 10% heat-inactivated fetal bovine serum (Invitrogen).
Transfections were performed on cells growing in 6-well plates, using the
Effectene transfection reagent (QIAGEN). 72 h after transfection a mix on
itor Cocktail (Roche) were added to the medium. Cells were harvested and
lysed in 2? SDS sample buffer. For Western blot analysis, the lysates were
nitrocellulose membrane (Protran). Blocking and antibody incubations
used: mouse anti-?-Tubulin (Sigma-Aldrich, DM1A, 1:1000), mouse
anti-HA (12CA5, 1:1000), and mouse anti-GFP (Clontech, #632375,
1:5000). HRP-conjugated secondary antibodies (Jackson Immu-
noResearch) were used at 1:10,000. Blots were developed using the
SuperSignal West Femto Maximum Sensitivity Substrate (Thermo-
Scientific) and imaged by ChemiDoc XRS? System and Image Lab
Version 2.0.1 software (Bio-Rad).
To identify genes essential for dendrite targeting in Drosophila
olfactory projection neurons (PNs), we performed a MARCM-
collection (Schuldiner et al., 2008). We found that the insertions
LL05207 and LL05209, integrated in the first and fifth intron of
the long transcript of CG10107, respectively (Fig. 1A), exhibited
PN dendrite targeting defects. Additionally, a third allele,
pBac(RB)e01260 (Thibault et al., 2004) which is inserted in the
drite targeting defects. All three alleles were homozygous lethal
and fail to complement each other (data not shown).
The MARCM technique (Lee and Luo, 1999) allows the visual-
ization and genetic manipulation of PNs in neuroblast and single-
cell clones in an otherwise heterozygous animal. We used the
8332 • J.Neurosci.,June13,2012 • 32(24):8331–8340 Berdniketal.•SUMOProteaseRegulatesNeuronTargeting
to label PNs from two neuroblast lineages, anterodorsal (ad) and
targeted stereotyped sets of glomeruli in neuroblast clones (Fig.
defects (Fig. 1C,E). First, the number of neurons was reduced from
an average of 35 adPNs in WT (Jefferis et al., 2001) to 5 neurons in
velo mutant clones (quantified in Fig. 4G). Second, the overall den-
dritic mass was reduced and disorganized. Third, dendrites inner-
(Thibault et al., 2004) and is inserted 33 bp upstream from the second exon in the first intron of the short transcript. Exons are shown as black, UTRs as gray bars, and introns as lines. B, WT
single neurons arranged into five classes according to their dendrite phenotypes. Class 1 veloLL05209mutant single neurons innervate the DL1 glomerulus often more sparsely and additionally
yw hsFlp122UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/?; Gal80 FRT2A/ FRT2A, FRT82B, y? (B, D, F, M), yw hsFlp122UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/?; Gal80 FRT2A/
Berdniketal.•SUMOProteaseRegulatesNeuronTargetingJ.Neurosci.,June13,2012 • 32(24):8331–8340 • 8333
To study dendrite targeting in greater
detail, we used Gal4-GH146 and MARCM
to label single cell PN clones produced in
newly hatched larva. In WT, they always
projected their dendrites to the posterior,
dorsolateral glomerulus DL1 in the AL,
feris et al., 2001). This specific innervation
pattern was formed during early pupa de-
velo mutant DL1 PNs, we observed various
dendrite mistargeting defects that were al-
be grouped into five distinct phenotypic
cent areas (Fig. 1G). In class 2, dendrites
AL but excluded the DL1 glomerulus (Fig.
1H). Class 3 mutant DL1 PNs projected to
ventromedial regions in the AL, preferably
but not exclusively to the VM6 or VC3
mistargeted their dendrites outside the AL
mostly into the subesophageal ganglion
(Fig. 1J). Class 5 mutant dendrites inner-
1K). The severity of dendrite mistargeting
class 1 being the mildest and closest to the
DL1 glomerulus. We quantified the pen-
etrance of the different phenotypic classes
for these 3 velo alleles. We determined that
e01260 was the strongest allele, lacking PNs
of class 1 that targeted to or close to DL1.
LL05207 was the weakest allele lacking class
4 and 5 phenotypes, and LL05209 repre-
sented the intermediate allele with all 5
verity of the phenotypes across different al-
leles, as measured by DL1 PN mistargeting,
and the presence of mistargeting pheno-
In addition to dendrite mistargeting, we also observed defects in
axon morphologies of velo mutant DL1 PNs. WT DL1 PN axons
project stereotypically into the lateral horn (LH) after passing
through the mushroom body calyx (MBC), where they form ?5
collateral branches. After entering the LH, DL1 PN axons always
form one characteristic dorsal branch (arrowhead in Fig. 2B)
while the main branch terminates at the lateral edge of the LH
In velo mutant DL1 PNs, axons extended along the normal path
and always reached the end of the LH, but formed only 0–2
collaterals in the MBC, and had often a missing or shorter dorsal
branch in the LH (Fig. 2D,F,G). These axonal phenotypes were
independent of the phenotypic class of dendrite mistargeting de-
scribed earlier (Fig. 2D,F) and occurred in multiple velo alleles,
including LL05209 (Fig. 2D,F) and e01260 (Fig. 2H). However,
to the VM6 glomerulus) often exhibited extensive, but nonste-
reotyped axon branching within the LH (Fig. 2E–H).
Transposon insertions can cause effects in genes distant to their
insertion site. To determine that the loss of Velo function was
indeed the cause for the mutant phenotypes, we used two ap-
ond, we generated UAS-velo-HA rescue transgenes using the
coding sequence of the long transcript (Fig. 1A). MARCM
overexpression of Velo-HA in WT DL1 single PNs did not
result in detectable phenotypes in their dendrites and axons
(data not shown), but rescued the velo mutant dendrite and
single neurons generated by MARCM are shown in green stained with antibodies to CD8, red represents staining using the
MBC and LH are within the dotted lines, an arrowhead points to the dorsal branch (B). C–F, Representative images for two
veloLL05209single PNs. One exhibits the mildest dendrite mistargeting phenotype illustrated by dendrites spilling into adjacent
8334 • J.Neurosci.,June13,2012 • 32(24):8331–8340 Berdniketal.•SUMOProteaseRegulatesNeuronTargeting
axon phenotypes (Fig. 3A,B). Twenty of 22 (91%) examined
velo mutant DL1 PNs normally innervated the DL1 glomeru-
lus when one copy of UAS-velo-HA was simultaneously ex-
pressed in this neuron (compare Figs. 2C,E,G, 3A).
To examine the axon morphogenesis defects, we quantified the
velo mutant axon phenotypes by counting the number of collateral
only 3% of examined DL1 axons had 2 or fewer collaterals in the
of velo mutant axons the dorsal branch was
missing, a phenotype reverted partially
upon rescue with the velo transgene (Fig.
innervation and axon branch formation of
DL1 PNs. Moreover, because Gal4-GH146
expression is restricted to postmitotic neu-
rons (Spletter et al., 2007), these rescue ex-
periments also demonstrated that velo
SUMO protease. The catalytic domain was
ity in human homologs of Velo (Lima and
Reverter, 2008; Shen et al., 2009). To test
whether Velo requires its catalytic domain
ated a rescue transgene bearing a mutation
in the catalytic domain of the protease. We
the core catalytic triad Cys-His-Asp pre-
dicted to yield a catalytically dead variant.
We found that WT and mutant Velo pro-
pressed in cell culture (Fig. 3E), indicating
stabilization of the Velo protein. We intro-
duced one copy of the modified transgene,
UAS-veloC?S-HA, into velo mutant DL1
PNs. We did not detect any rescue in den-
was not due to low or absent expression of
the transgene, as we can readily detect high
the nucleus (Fig. 3F, inset). The nuclear lo-
calization was consistent with the expres-
sion of WT UAS-velo-HA transgene (Fig.
3A, inset) and suggests that Velo likely acts
predominantly in the nucleus. In addition
geting, expression of VeloC?S-HA in velo
mutant DL1 PNs also did not rescue the
exhibited the same axon morphology
defects regarding MBC branching and
dorsal branch formation; furthermore, 3/8 DL1 PN axons failed
to innervate the LH (Fig. 3G). In summary, we conclude that the
catalytic domain of Velo is essential for its role during dendrite
indeed might act as SUMO protease, likely in the nucleus.
Velo is highly conserved across species from yeast to human.
Sentrin-specific protease 7 is the closest human homolog (Fig.
fusion protein to the nucleus (A, inset), and rescues veloe01260mutant dendrite phenotypes in 20/22 cases (A) and axon (B)
in the LH. Numbers on top of the bars indicate the number of single PNs analyzed. E, VeloWT-HA, VeloC?S-HA or GFP were
Berdniketal.•SUMOProteaseRegulatesNeuronTargetingJ.Neurosci.,June13,2012 • 32(24):8331–8340 • 8335
teine proteases responsible for the process-
ing and deconjugation of SUMO proteins.
SUMO proteases are abundant in eu-
karyotes including two members in bud-
ding yeast, the Ubiquitin-like protein-
specific proteases (Ulp), and seven in
mammals, called Sentrin-specific Pro-
teases (SENPs). SUMO proteases diverge
into two major branches, one related to
Ulp1 and the other Ulp2 (Hay, 2007; Muk-
hopadhyay and Dasso, 2007). Velo has a
split C48 catalytic domain characteristic
of SENP6 and SENP7. Our phylogenetic
analysis of sequence similarities classified it
Protein sequence comparisons predict
five SUMO proteases in the Drosophila ge-
nome: three are Ulp1-related (DmUlp1,
CG11023 and CG32110) and two are more
similar to Ulp2 (velo, CG12717) related.
in vivo, nor were reagents for mutant
and overexpression analysis available.
To assess whether Velo is uniquely
required in PN targeting, we generated
mutants for the closest velo ortholog,
CG12717, using FLP recombinase-mediated
excision (Parks et al., 2004). The flanking
P-element P(XP)d05069 and piggyBac line
pBAC(RB)e01706a allowed us to delete the
entire open reading frame of CG12717,
while not interfering with adjacent genes
mutant CG12717ex#6with FRT19A, per-
formed MARCM experiments, and found
that dendrites and axons targeted normally
in CG12717ex#6neurons (Fig. 4C and data
not shown). PN cell numbers were normal
in neuroblast clones (Fig. 4D), and
CG12717 mutant single DL1 PNs normally
innervated the DL1 glomerulus (Fig. 4C,
compare to Fig. 1F). Therefore, we con-
To address the question of whether
CG12717 or another SUMO protease
from the Ulp1 family, DmUlp1, can
functionally compensate for Velo in PN
targeting, we generated HA-tagged trans-
genesfor these two SUMO proteases and
expressed them in velo mutant PNs. We
observed a significant reversion of the
DL1 mistargeting phenotype in velo mutant PNs supple-
mented with DmUlp1-HA; 12 of 14 (86%) examined PN den-
drites target to the DL1 glomerulus (Fig. 4F). Similar
experiments using UAS-CG12717-HA also yielded a signifi-
cant rescue in DL1 targeting, although to a lesser extent (5 of
13 innervate DL1; data not shown). However, velo mutant
axon phenotypes could not be reverted by expressing either
DmUlp1-HA or CG12717-HA in DL1 PNs (data not shown).
This suggests that other SUMO proteases of both families can
compensate for the function of Velo in PN dendrite targeting
but not axon arborization.
Aberrant sumoylation can lead to many biological malfunc-
tions, including defects during cell division (Di Bacco et al.,
2006). In velo mutant neuroblast clones PN cell numbers were
to an average of 5.5 neurons per ad neuroblast clone (Figs. 1C,
4G,H). By driving UAS-Velo-HA expression in velo mutant ad
PNs, this defect was partially rescued as the number of neurons
ScUlp2-related members in green letters. B, Schematic of the CG12717 gene locus indicating the insertions of the P-element
DmUlp1 fusion protein. Asterisks in C, E, F denote single cell bodies. H–J, veloe01260adPN cell numbers are reduced (H), a
clones and after introducing various UAS rescue transgenes. Numbers on top of the bars indicate the number of analyzed an-
hsFlp122UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/ ?; Gal80 FRT2A/ veloe01260[w?], FRT2A, FRT82B (E, H, J), yw hsFlp122
UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/ UAS-DmUlp1-HA#5.1; Gal80 FRT2A/ veloe01260[w?], FRT2A, FRT82B (F, G, I, J), yw
hsFlp122UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/ UAS-velo-HA#108L; Gal80 FRT2A/ veloe01260[w?], FRT2A, FRT82B and yw
8336 • J.Neurosci.,June13,2012 • 32(24):8331–8340Berdniketal.•SUMOProteaseRegulatesNeuronTargeting
was doubled (?10.6 neurons; Fig. 4G). The number of adPNs
also increased upon expressing UAS-DmUlp1-HA (?12.5 neu-
numbers could reflect different expression levels of the distinct
the postmitotic expression of various SUMO proteases in velo
However, the fact that cell numbers in the rescued adPNs were
markedly less than WT suggests that Velo also functions during
the three SUMO protease transgenes. While Velo-HA was found
exclusively in the nucleus (Fig. 3A, inset), DmUlp1-HA and
CG12717-HA localized both to the nucleus and cytoplasm, al-
though neither could be detected in dendrites or axons (Fig. 4J,
inset in 4F, and data not shown). This difference in subcellular
SUMO protease. Based on these expression experiments, the
three proteases appeared to be equally potent in the partial
rescue of cell numbers (Fig. 4G). Moreover, DmUlp1-HA and
CG12717-HA (to a lesser extent) can functionally substitute for
Velo in PN dendrite targeting. However, axonal morphogenesis
appears to specifically require Velo.
Sumoylation is a reversible post-translational modification and
tion (Johnson, 2004; Geiss-Friedlander and Melchior, 2007).
C-terminal di-glycine motif that is subsequently conjugated to a
activating enzyme, an E2 conjugating enzyme, and an E3 ligase.
SUMO proteases deconjugate SUMO(s) from the substrate or
from SUMO chains (chain editing; Fig. 5A).
The involvement of a SUMO protease in PN target selection
prompted us to examine other components of the SUMO path-
way in the same process. To this end, we tested mutants for
SUMO and the E2 conjugating enzyme lesswright (lwr) for their
role in dendrite target selection in DL1 PNs using MARCM. We
identified a P-element, P(lacW)l(2)SH0182[w?], which is in-
serted in the promoter region of the SUMO gene (Fig. 5G, red
triangle). PN targeting to the DL1 glomerulus was normal when
DL1 PNs (n ? 10) were homozygous for this insertion (compare
Fig. 5B,C). As this insertion might not impair SUMO func-
tion, we generated SUMO mutants via imprecise excision of
P(lacW)l(2)SH0182[w?] by screening for white eyes and map-
in Drosophila. SUMO deconjugation requires cleavage of the amide bond between mature SUMO and the target lysine within the substrate (green). Chain editing is chemically identical to
DNA ladder is loaded in right lane. Genotypes: yw hsFlp122UAS-mCD8GFP; FRT40A Gal4-GH146 UAS-mCD8GFP/ FRT40A (B), yw hsFlp122UAS-mCD8GFP; FRT40A Gal4-GH146 UAS-mCD8GFP/
Berdniketal.•SUMOProteaseRegulatesNeuronTargeting J.Neurosci.,June13,2012 • 32(24):8331–8340 • 8337
deletion, SUMOex77, that covered the en-
tire SUMO gene but left neighboring
genes unaffected (Fig. 5G,H). After re-
combining the SUMOex77-null allele onto
FRT40A, we performed MARCM experi-
ments. This yielded a very low clonal fre-
quency and often resulted in lethality in
both larvae and pupae, suggesting that
SUMO is required for many essential
functions in cells and even the generation
of homozygous mutant clones cannot
prevent animal death. However, from the
four single cell clones we obtained from
?500 dissected brains, we found dendrite
mistargeting in each case (Fig. 5D; data
not shown). As such, we conclude that
SUMO is essential for PN target selection.
PN target selection by inducing MARCM
clones of the LOF allele lwr13(Sun et al.,
2003). Dendrites of lwr13mutant DL1 PN
single cell clones innervated large areas
ulus (Fig. 5E); 5/24 (21%) innervated the
DL1 glomerulus but spill into several adja-
cent glomeruli, and the remaining 5/24
(21%) innervate simultaneously dorsolat-
eral and ventromedial portions of the AL
phenotypes were fully rescued by supple-
We conclude from our data examin-
ing the two LOF alleles, SUMOex77and
lwr13, that sumoylation in PNs is cell-
autonomously required for the correct tar-
get selection of their dendrites within the
AL. The rescue experiment for lwr further
neurons to regulate PN dendrite targeting. The neuronal targeting
defects of two members of the sumoylation pathway and Velo sup-
SUMO proteases can act at two distinct steps in the sumoylation
pathway. First, they promote sumoylation by cleaving immature
SUMO into its active form. Second, they revert sumoylation by
cleaving off one or more SUMOs from their substrates (Fig. 5A).
We performed genetic interaction experiments to distinguish
whether Velo promotes (Fig. 6A, model 1) or reverts (Fig. 6A,
model 2) sumoylation. We took advantage of the fact that the
veloLL05209allele produced an intermediate dendrite mistargeting
phenotype in DL1 PNs that might be sensitive to both enhance-
ment and suppression of the phenotype (Figs. 1L, 6E). A typical
example (38%) for a veloLL05209DL1 PN exhibiting ventromedi-
ally mistargeted dendrites is presented in Figure 6B. We gener-
ated veloLL05209DL1 PNs using MARCM and reduced SUMO
or Lwr protein levels by half using flies heterozygous for the
SUMOex77or lwr13LOF alleles, respectively. If Velo functions in
SUMO maturation, we would expect an enhancement of the
mild class 1 phenotype (Fig. 1G). However, if velo acts in SUMO
deconjugation, we would expect a suppression of the dendrite
mistargeting phenotype causing an increase in neurons exhibit-
ing the mild class 1 phenotype.
We observed suppression of dendrite mistargeting in
veloLL05209mutant DL1 PNs with either SUMO or Lwr protein
levels reduced by half in the same mutant neuron (Fig. 6C–E).
Quantifications of the various phenotypic classes representing
dendrite targeting defects for single PNs of the three distinct ge-
notypes (Fig. 6E) suggested that Velo acts predominantly in
SUMO deconjugation rather than in SUMO maturation. Our
finding is consistent with previous results that the closest human
homolog of Velo, SENP7, is unable to process SUMO precursors
for SUMO maturation but rather contributes to SUMO decon-
jugation from poly-SUMO chains (Shen et al., 2009).
Protein sumoylation plays an important role in a wide range of
cellular processes, including transcription, chromosome organi-
duction, and cell cycle progression. Since its discovery, several
anti-CD8 (green). Scale bar: 20 ?m. Genotypes: yw hsFlp122UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/ ?; Gal80 FRT2A/
veloLL05209[DsRed], FRT2A, FRT82B (B, E), yw hsFlp122UAS-mCD8GFP; Gal4-GH146 UAS-mCD8GFP/ SUMOex77; Gal80 FRT2A/
8338 • J.Neurosci.,June13,2012 • 32(24):8331–8340 Berdniketal.•SUMOProteaseRegulatesNeuronTargeting
hundred sumoylation substrates have been identified, including
proteins localized to the nucleus, cytoplasm, or at the plasma
studies have shown that many of these sumoylated substrates are
2010). However, because the major components of the sumoyla-
tion pathway are essential for cell viability, it is challenging to
effects of sumoylation in vivo.
Here, from a forward genetic screen using a powerful mosaic
analysis technique, we identified a predicted SUMO protease,
Velo, that regulates dendrite and axon targeting in postmitotic
neurons in vivo. Several lines of evidence indicate that Velo con-
trols neuronal morphogenesis by regulating protein sumoyla-
tion. First, the catalytic domain of the protease is required for its
function in neurons. Second, the dendrite targeting phenotypes
can partially be rescued by two other predicted SUMO proteases
from two evolutionarily separable branches. SUMO proteases
tion and deconjugation of SUMO from mono-sumoylated sub-
strates, while Ulp2-like proteases deconjugate SUMO protein(s)
from poly-SUMO chains. Interestingly, overexpression of both
Ulp1 and Ulp2 family proteases was able to rescue the velo den-
drite phenotypes. Third, two other components of the sumoyla-
tion pathway, SUMO itself and the unique E2 conjugating
enzyme Lesswright (Lwr), are also required cell-autonomously
for PN dendrite targeting. Fourth, SUMO and Lwr exhibit
dosage-sensitive interactions with Velo; velo mutant dendrite
age by half. Indeed, the nature of these genetic interactions sug-
gests that Velo acts primarily to reverse sumoylation via SUMO
deconjugation rather than to promote sumoylation via SUMO
It has previously been shown that the knockdown of the Dro-
sophila SUMO protease Ulp1 and overexpression of human
SENP7 result in a change of total SUMO conjugates in cultured
cells (Smith et al., 2004; Shen et al., 2009). We performed similar
experiments to test the biochemical activity of Velo as a SUMO
protease by overexpressing Velo in cultured cells. We have not
been able to detect significant changes in the overall spectrum of
SUMO conjugates upon Velo overexpression (data not shown).
cultured cells, or that Velo’s substrates are absent in cultured
cells. For these reasons and other technical hurdles, such as the
Velo protein in bacteria, biochemical evidence for Velo acting as
a SUMO protease is still missing. We also cannot rule out the
possibility that some of the effects of Velo on PN dendrite and
axon targeting are caused by its action on substrates unrelated to
the SUMO pathway.
Although velo, SUMO and lwr mutant PNs exhibit aberrant
dendrite targeting, their phenotypes are not identical. One pos-
sibility for the phenotypic differences could be due to the redun-
dant action of SUMO proteases either between members of the
same branch or even the two distinct branches. For example,
CG12717, the closest homolog of Velo, or the Ulp1-related
tion. This is consistent with the fact that the overexpression of
transgenes for both proteases can partially revert velo mutant
enzyme. Therefore, their loss-of-function phenotypes are more
severe. Another possibility for the phenotypic differences we ob-
serve in velo, SUMO and lwr mutant PNs could be attributed to
the differential perdurance of these proteins in single neurons
generated by MARCM. Finally, the two members of the sumoy-
lation pathway we examined act in opposite ways with Velo:
SUMO and Lwr promote, whereas Velo reverts, sumoylation.
This feature implies that the dynamics of sumoylation are
essential for dendrite and axon targeting: too much or not
enough sumoylation are both harmful to PNs and cause neu-
ronal mistargeting. Although all three possibilities can con-
tribute, the last one might contribute most to the observed
The closest human homolog to Velo is SENP7 (Fig. 4A).
regarding Velo protein distribution. The catalytic domain of
SENP7 is essential for its protease activity. Biochemical assays
revealed that this protease functions preferably during deconju-
gation of poly-SUMO chains (Lima and Reverter, 2008; Shen et
unknown in eukaryotes and few substrates have been identified.
SUMO chain formation requires internal lysines within sumoy-
yeast during vegetative growth (Bylebyl et al., 2003). However,
SUMO polymers play a structural role during meiosis in yeast
and mitosis in mammalian cells (Cheng et al., 2006; Zhang et al.,
strate can promote its subsequent ubiquitylation and degrada-
tion, thereby acting as ubiquitylation signals in the turnover of
SUMO targets (Ulrich, 2008; Geoffroy and Hay, 2009). We spec-
ulate that Velo acts likely in the deconjugation of poly-SUMO
chains because of the sequence similarities to SENP7. However,
roles for poly-SUMO chains in neurons and crosstalks between
sumoylation and ubiquitination pathways during neuronal tar-
get selection remain to be determined.
PN dendrite and axon targeting requires identification of its tar-
get substrate(s). Because Velo-HA localizes to the nucleus, the
potential substrate is likely a nuclear protein. Numerous studies
have demonstrated a role for sumoylation regulating transcrip-
tion (Seeler and Dejean, 2003; Verger et al., 2003). For example,
the E3 SUMO ligase and transcriptional coregulator Protein In-
hibitor of Activated Stat3 (Pias3) controls rod photoreceptor de-
velopment and differentiation in the mouse retina by regulating
transcription factors via sumoylation (Onishi et al., 2009, 2010).
Furthermore, several transcription factors have been shown to
regulate PN dendrite target selection when misexpressed or mu-
et al., 2007). Another likely set of substrates for Velo includes
factors involved in chromosome organization and function. In-
deed, it has recently been shown that SMC1, a cohesin subunit
required for sister chromatid cohesion during mitosis and meio-
sis, and the chromatin remodeling factor Rpd3, a class 1 histone
deacetylase (HDAC1) involved in chromatin integrity, play roles
during PN targeting (Schuldiner et al., 2008; Tea et al., 2010).
Future studies on candidate genes that exhibit similar neuronal
targeting errors, together with biochemical and proteomic ap-
proaches, might uncover potential Velo substrates, and provide
further insight into how sumoylation participates in the precise
wiring of the olfactory circuit.
BylebylGR,BelichenkoI,JohnsonES (2003) TheSUMOisopeptidaseUlp2
prevents accumulation of SUMO chains in yeast. J Biol Chem
Berdniketal.•SUMOProteaseRegulatesNeuronTargetingJ.Neurosci.,June13,2012 • 32(24):8331–8340 • 8339
Cheng CH, Lo YH, Liang SS, Ti SC, Lin FM, Yeh CH, Huang HY, Wang TF
and polycomplex in meiosis of Saccharomyces cerevisiae. Genes Dev
DiBaccoA,OuyangJ,LeeHY,CaticA,PloeghH,GillG (2006) TheSUMO-
specific protease SENP5 is required for cell division. Mol Cell Biol
Dorval V, Fraser PE (2007) SUMO on the road to neurodegeneration.
Biochim Biophys Acta 1773:694–706.
Feligioni M, Nishimune A, Henley JM (2009) Protein SUMOylation mod-
ulates calcium influx and glutamate release from presynaptic terminals.
Eur J Neurosci 29:1348–1356.
Geiss-Friedlander R, Melchior F (2007) Concepts in sumoylation: a decade
on. Nat Rev Mol Cell Biol 8:947–956.
Geoffroy MC, Hay RT (2009) An additional role for SUMO in ubiquitin-
mediated proteolysis. Nat Rev Mol Cell Biol 10:564–568.
HayRT (2007) SUMO-specificproteases:atwistinthetail.TrendsCellBiol
Hong W, Zhu H, Potter CJ, Barsh G, Kurusu M, Zinn K, Luo L (2009)
Leucine-rich repeat transmembrane proteins instruct discrete dendrite
targeting in an olfactory map. Nat Neurosci 12:1542–1550.
Jefferis GS, Marin EC, Stocker RF, Luo L (2001) Target neuron prespecifi-
cation in the olfactory map of Drosophila. Nature 414:204–208.
LuoL (2004) Developmentaloriginofwiringspecificityintheolfactory
system of Drosophila. Development 131:117–130.
Jefferis GS, Potter CJ, Chan AM, Marin EC, Rohlfing T, Maurer CR Jr, Luo L
Johnson ES (2004) Protein modification by SUMO. Annu Rev Biochem
Komiyama T, Luo L (2007) Intrinsic control of precise dendritic targeting
by an ensemble of transcription factors. Curr Biol 17:278–285.
KomiyamaT,JohnsonWA,LuoL,JefferisGS (2003) Fromlineagetowiring
specificity: POU domain transcription factors control precise connec-
tions of Drosophila olfactory projection neurons. Cell 112:157–167.
Komiyama T, Sweeney LB, Schuldiner O, Garcia KC, Luo L (2007) Graded
expression of semaphorin-1a cell-autonomously directs dendritic target-
ing of olfactory projection neurons. Cell 128:399–410.
LeeT,LuoL (1999) Mosaicanalysiswitharepressiblecellmarkerforstudies
of gene function in neuronal morphogenesis. Neuron 22:451–461.
LiSJ,HochstrasserM (1999) Anewproteaserequiredforcell-cycleprogres-
sion in yeast. Nature 398:246–251.
Li SJ, Hochstrasser M (2000) The yeast ULP2 (SMT4) gene encodes a novel
protease specific for the ubiquitin-like Smt3 protein. Mol Cell Biol
Lima CD, Reverter D (2008) Structure of the human SENP7 catalytic do-
main and poly-SUMO deconjugation activities for SENP6 and SENP7.
J Biol Chem 283:32045–32055.
MarinEC,JefferisGS,KomiyamaT,ZhuH,LuoL (2002) Representationof
the glomerular olfactory map in the Drosophila brain. Cell 109:243–255.
Martin S, Wilkinson KA, Nishimune A, Henley JM (2007) Emerging ex-
tranuclear roles of protein SUMOylation in neuronal function and dys-
function. Nat Rev Neurosci 8:948–959.
Mukhopadhyay D, Dasso M (2007) Modification in reverse: the SUMO
proteases. Trends Biochem Sci 32:286–295.
Onishi A, Peng GH, Hsu C, Alexis U, Chen S, Blackshaw S (2009) Pias3-
dependent SUMOylation directs rod photoreceptor development. Neu-
OnishiA,PengGH,ChenS,BlackshawS (2010) Pias3-dependentSUMOy-
Parks AL, Cook KR, Belvin M, Dompe NA, Fawcett R, Huppert K, Tan LR,
Winter CG, Bogart KP, Deal JE, Deal-Herr ME, Grant D, Marcinko M,
Miyazaki WY, Robertson S, Shaw KJ, Tabios M, Vysotskaia V, Zhao L,
Andrade RS, et al. (2004) Systematic generation of high-resolution de-
letion coverage of the Drosophila melanogaster genome. Nat Genet
Scheschonka A, Tang Z, Betz H (2007) Sumoylation in neurons: nuclear
and synaptic roles? Trends Neurosci 30:85–91.
Schuldiner O, Berdnik D, Levy JM, Wu JS, Luginbuhl D, Gontang AC, Luo L
(2008) piggyBac-based mosaic screen identifies a postmitotic function for
SeelerJS,DejeanA (2003) NuclearandunclearfunctionsofSUMO.NatRev
Mol Cell Biol 4:690–699.
ShaliziA,Gaudillie `reB,YuanZ,Stegmu ¨llerJ,ShiroganeT,GeQ,TanY,Schul-
man B, Harper JW, Bonni A (2006) A calcium-regulated MEF2 sumoyla-
Shen LN, Geoffroy MC, Jaffray EG, Hay RT (2009) Characterization of
SENP7, a SUMO-2/3-specific isopeptidase. Biochem J 421:223–230.
Smith M, Bhaskar V, Fernandez J, Courey AJ (2004) Drosophila Ulp1, a
nuclear pore-associated SUMO protease, prevents accumulation of cyto-
plasmic SUMO conjugates. J Biol Chem 279:43805–43814.
Spletter ML, Liu J, Su H, Giniger E, Komiyama T, Quake S, Luo L (2007)
ing specificity. Neural Develop 2:14.
Lukacsovich T, Zhu YZ, Cattaneo E, Pandolfi PP, Thompson LM, Marsh
JL (2004) SUMO modification of Huntingtin and Huntington’s disease
pathology. Science 304:100–104.
Stocker RF, Heimbeck G, Gendre N, de Belle JS (1997) Neuroblast ablation
in Drosophila P[GAL4] lines reveals origins of olfactory interneurons.
J Neurobiol 32:443–456.
Sun X, Huang L, Van Doren MR, Tanda S (2003) Isolation of amorphic
alleles of the lesswright gene by P-element mediated male recombination
in Drosophila melanogaster. Dros Inf Serv 86:79–83.
Tea JS, Chihara T, Luo L (2010) Histone deacetylase Rpd3 regulates olfac-
tory projection neuron dendrite targeting via the transcription factor
Prospero. J Neurosci 30:9939–9946.
Thibault ST, Singer MA, Miyazaki WY, Milash B, Dompe NA, Singh CM,
Buchholz R, Demsky M, Fawcett R, Francis-Lang HL, Ryner L, Cheung
LM, Chong A, Erickson C, Fisher WW, Greer K, Hartouni SR, Howie E,
Jakkula L, Joo D, et al. (2004) A complementary transposon tool kit for
Drosophila melanogaster using P and piggyBac. Nat Genet 36:283–287.
Ulrich HD (2008) The fast-growing business of SUMO chains. Mol Cell
Verger A, Perdomo J, Crossley M (2003) Modification with SUMO. A role
in transcriptional regulation. EMBO Rep 4:137–142.
Wilkinson KA, Henley JM (2010) Mechanisms, regulation and conse-
quences of protein SUMOylation. Biochem J 428:133–145.
Wilkinson KA, Nakamura Y, Henley JM (2010) Targets and consequences
of protein SUMOylation in neurons. Brain Res Rev 64:195–212.
Wong AM, Wang JW, Axel R (2002) Spatial representation of the glomeru-
lar map in the Drosophila protocerebrum. Cell 109:229–241.
Wu JS, Luo L (2006a) A protocol for mosaic analysis with a repressible cell
marker (MARCM) in Drosophila. Nat Protoc 1:2583–2589.
Wu JS, Luo L (2006b) A protocol for dissecting Drosophila melanogaster
brains for live imaging or immunostaining. Nat Protoc 1:2110–2115.
Zhang XD, Goeres J, Zhang H, Yen TJ, Porter AC, Matunis MJ (2008)
with kinetochores and progression through mitosis. Mol Cell 29:729–
8340 • J.Neurosci.,June13,2012 • 32(24):8331–8340 Berdniketal.•SUMOProteaseRegulatesNeuronTargeting