Article A Krebs Cycle Component Limits Caspase Activation Rate through Mitochondrial Surface Restriction of CRL Activation Graphical Abstract Highlights d Caspase-mediated Drosophila spermatid cytoplasmic content extrusion is restricted d SCSb of the Krebs cycle activates a Ub ligase on the spermatid mitochondrial surface d The Ub ligase limits the source of caspase activation to the mitochondrial vicinity d This moonlighting function of SCSb reduces the potential rate of caspase activation Authors Lior Aram, Tslil Braun, Carmel Braverman, Yosef Kaplan, Liat Ravid, Smadar Levin-Zaidman, Eli Arama Correspondence [email protected]In Brief How is caspase activity restricted in the cell when removing large cellular structures? Aram et al. uncover a moonlighting function of the Krebs cycle component SCSb during sperm individualization in limiting activation of the Ub ligase complex required for caspase activity to the mitochondrial surface and reducing caspase activation potential rate. Aram et al., 2016, Developmental Cell 37, 15–33 April 4, 2016 ª2016 Elsevier Inc. http://dx.doi.org/10.1016/j.devcel.2016.02.025
47
Embed
AKrebsCycleComponentLimitsCaspaseActivation Rate through ... · Developmental Cell Article A Krebs Cycle Component Limits Caspase Activation Rate through Mitochondrial Surface Restriction
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Article
AKrebs Cycle Component Limits Caspase Activation
Rate through Mitochondrial Surface Restriction ofCRL Activation
Graphical Abstract
Highlights
d Caspase-mediated Drosophila spermatid cytoplasmic
content extrusion is restricted
d SCSb of the Krebs cycle activates a Ub ligase on the
spermatid mitochondrial surface
d The Ub ligase limits the source of caspase activation to the
mitochondrial vicinity
d Thismoonlighting function of SCSb reduces the potential rate
A Krebs Cycle Component Limits Caspase ActivationRate through Mitochondrial SurfaceRestriction of CRL ActivationLior Aram,1 Tslil Braun,1 Carmel Braverman,1 Yosef Kaplan,1 Liat Ravid,1 Smadar Levin-Zaidman,2 and Eli Arama1,*1Department of Molecular Genetics, Weizmann Institute of Science, Rehovot 76100, Israel2Electron Microscopy Unit, Weizmann Institute of Science, Rehovot 76100, Israel
How cells avoid excessive caspase activity andunwanted cell death during apoptotic caspase-mediated removal of large cellular structures ispoorly understood. We investigate caspase-medi-ated extrusion of spermatid cytoplasmic contentsin Drosophila during spermatid individualization.We show that a Krebs cycle component, the ATP-specific form of the succinyl-CoA synthetase b sub-unit (A-Sb), binds to and activates the Cullin-3-basedubiquitin ligase (CRL3) complex required for caspaseactivation in spermatids. In vitro and in vivo evidencesuggests that this interaction occurs on the mito-chondrial surface, thereby limiting the source ofCRL3 complex activation to the vicinity of this organ-elle and reducing the potential rate of caspase acti-vation by at least 60%. Domain swapping betweenA-Sb and the GTP-specific SCSb (G-Sb), which func-tions redundantly in the Krebs cycle, show that themetabolic and structural roles of A-Sb in spermatidscan be uncoupled, highlighting a moonlighting func-tion of this Krebs cycle component in CRL activation.
INTRODUCTION
Regulated cellular destruction processes (CDPs) are instru-
mental for the development, function, and homeostasis of a
variety of tissues and organs in virtually all organisms. At the
cellular level, these processes range from complete demolition
of the cell by programmed cell death (PCD) (Lettre and Hengart-
ner, 2006; Fuchs and Steller, 2011), partial cellular destruction by
specialized cell remodeling (Raff et al., 2002; Feinstein-Rotkopf
and Arama, 2009), to the selective turnover of damaged organ-
elles (Kirkin et al., 2009; Youle and Narendra, 2011). Whereas
some CDPs are mediated by distinct mechanisms designed to
cope with the different levels of destruction required for each
process, other CDPs may utilize similar pathways to bring about
distinct mechanistic outcomes. For example, activation of cas-
pases, the executioners of the apoptotic program, usually leads
to cell death, but caspases can also mediate partial CDPs
(Abraham and Shaham, 2004; Kuranaga and Miura, 2007; Fein-
stein-Rotkopf and Arama, 2009; Yi and Yuan, 2009). How a
demolition pathway, normally used to destroy an entire cell,
can be harnessed to remodel a cell has been poorly understood.
Furthermore, it is still amystery why some cells prefer a particular
CDP pathway to another. While it is relatively easy to trigger
‘‘macro’’ CDPs such as apoptosis, inducing the more precise
or selective CDPs in ways that may recapitulate the physiological
conditions, is much more complicated. One way to overcome
some of these obstacles is to use genetically malleable model
organisms, however thus far, only very few physiological para-
digms have been available to investigate partial or selective
CDPs (Arama et al., 2003; Politi et al., 2014).
Apoptosis, the most common form of PCD, is executed by
the action of unique proteases called caspases (Yuan et al.,
1993; Salvesen, 2002). Executioner (or effector) caspases are
regulated by both activating and inhibitory proteins, such as
the apoptosome components (cytochrome c, Apaf-1, and
caspase-9) (Shi, 2002; Jiang and Wang, 2004) and members
of the inhibitor of apoptosis protein (IAP) family, respectively
(Salvesen and Abrams, 2004; Steller, 2008). Although caspase
activation has become almost synonymous with apoptosis,
accumulating evidence, mostly in the past decade, suggests
that active caspases also mediate a variety of vital cellular
processes (Abraham and Shaham, 2004; Launay et al., 2005;
Lamkanfi et al., 2007; Kuranaga and Miura, 2007; Yi and Yuan,
2009; Feinstein-Rotkopf and Arama, 2009; Crawford and Wells,
2011). To date, more than 50 vital cellular processes have been
described in diverse tissues and organisms, such as neurite
pruning and synapse remodeling in neurons (Williams et al.,
2006; Kuo et al., 2006; Huesmann and Clayton, 2006; Nikolaev
et al., 2009; Li et al., 2010) and the removal of many organelles
(including the nuclei) during terminal differentiation of lens
epithelial cells and red blood cells in mammals (Ishizaki et al.,
1998; Zermati et al., 2001; Carlile et al., 2004; Bassnett, 2009).
Likewise, apoptotic proteins, including active caspases, the
apoptosome, and the giant IAP, dBruce, are all involved in the
process of spermatid individualization in Drosophila (Arama
et al., 2003, 2006; Huh et al., 2004; Muro et al., 2006; Arama
et al., 2007; Kaplan et al., 2010; D’Brot et al., 2013). During this
partial CDP, spermatids separate from each other and acquire
the typical streamlined morphology by losing their bulk cyto-
plasmic contents and organelles. This process is driven by the
caudal movement of an actin-based individualization complex
(IC), which collects the extruded material into a membrane-
enclosed structure called the cystic bulge (CB), and is eventually
Developmental Cell 37, 15–33, April 4, 2016 ª2016 Elsevier Inc. 15
(A) Left, schematic structures of Cul3T and Cul3S protein sequences and the TeNC domain polypeptide from Cul3T. Middle and right, two Y2H assays: (I) a
b-galactosidase filter assay and (II) a nutrient-omitted medium assay testing the interactions between the three proteins indicated at the left (baits) and the
A-Sb(ckr) polypeptide (prey).
(B) Schematic representation of the A-SbT protein and the relative locations of its major domains. The predicted mitochondrial targeting sequence (MTS) may be
partial, as its deletion could not abrogate mitochondrial localization in S2 cells (data not shown).
(C) Schematic illustration of CRL1 and CRL3 complexes and their regions of interactions with A-Sb, respectively based on Hughes et al. (2008) and the current
study. SIM, substrate interacting motif.
(D) Yeast growth assay on selective media plates. Yeast were introduced with a Klhl10 bait plasmid and either of the following prey plasmids: A-Sb(ckr), an empty
vector (control), or Klhl10 (which serves as a positive control due to its ability to homodimerize). Whereas yeast colonies containing both vectors can grow on a
medium lacking tryptophan and leucine (�2), only colonies with protein-protein interaction events grow on a medium that also lacks histidine and adenine (�4).
(legend continued on next page)
16 Developmental Cell 37, 15–33, April 4, 2016
pinching off the spermatids as awaste bag (WB) (Tokuyasu et al.,
1972; Fuller, 1993; Fabrizio et al., 1998; Noguchi and Miller,
2003). Whereas this bulk of evidence counters the old dogma
that all cells expressing active caspases are doomed to die, it re-
mains largely unclear how some cells avoid excessive caspase
activity that leads to apoptosis.
Cullin-RING ubiquitin ligases (CRLs) constitute eight groups of
plexes, implicated in the ubiquitination of hundreds of proteins,
and thus have impact on essentially all basic cellular processes
(Willems et al., 2004; Petroski and Deshaies, 2005). The unique
properties of the different CRL groups are determined by the
type of Cullin protein they contain. The Cullins function as scaf-
fold for two protein units: a catalytic unit (at the C terminus) and a
substrate recruitment unit (at the N terminus). The catalytic unit is
almost invariable in all CRL groups and is composed of a small
RING domain protein, which in turn recruits the ubiquitin-conju-
gating E2 enzyme. In contrast, hundreds of different proteins are
believed to serve as substrate recruitment units, thus facilitating
CRL-substrate specificity. The composition of the substrate
recruitment units varies between the different CRL groups; for
example, Cullin-1 binds to the adapter protein Skp1 (also known
as SkpA), which in turn recruits a substrate receptor from the
F-box protein family, whereas the substrate recruitment module
in Cullin-3-based CRLs (or CRL3s) is composed of single BTB-
domain-containing proteins (Pintard et al., 2004; Willems et al.,
2004; Petroski and Deshaies, 2005; Genschik et al., 2013).
CRLs are generally activated by a reversible covalent modifica-
tion of the Cullin subunit with the ubiquitin-like protein, Nedd8,
whereas at least for CRL1 complexes (and possibly also other
CRL groups), the inactive fraction of the Cullin is associated
with Cand1, a protein that disrupts both the Cullin association
with Skp1 and neddylation (Wu et al., 2006; Bosu and Kipreos,
2008; Merlet et al., 2009; Deshaies et al., 2010; Duda et al.,
2011). However, as opposed to the dramatic progress in our
understanding of global CRL regulatory mechanisms, very little
is known about the more specific mechanisms that differen-
tially regulate distinct CRL complexes to prevent improper and
deleterious activation of hundreds of CRLs, all at the same
time (Deshaies et al., 2010).
We use theDrosophila sperm as amodel system to investigate
how apoptotic caspases are regulated during the vital process of
spermatid individualization. Previously, we discovered a CRL3
complex composed of a testis-specific Cullin-3 isoform, Cul3T,
and the substrate recruitment protein Klhl10, which is required
for caspase activation in this system (Arama et al., 2007). In a
subsequent work, we identified a pseudosubstrate inhibitor of
this complex, called Soti, which directs graded activity of the
CRL3 complex, thus helping to restrict caspase activity in time
(Kaplan et al., 2010). Although the Soti-mediated mechanism ex-
plains how the more distal regions of spermatids are protected
against excessive caspase activation and cell death, it does
not explain how spermatids, particularly the more proximal re-
gions, avoid the rapid activation of caspases. This question is
(E and F) S2 cells were transfected with one or more of the indicated constructs. (
Klhl10, Cul3T, and GFP was confirmed using the indicated antibodies. Transfecte
slower migrating, possibly neddylated, Cul3T form (upper arrow in lane 5).
See also Figure S1.
particularly puzzling, since the process time of individualization,
about 12 hr (Noguchi and Miller, 2003), is much longer than the
time it usually takes to accumulate a level of caspase activity
that induces cell death (less than 4 hr in Drosophila epithelial
cells; Florentin and Arama, 2012). Furthermore, while cytoplasm
removal is impaired in soti mutant spermatids, the IC still con-
tinues to progress about three-quarters of the spermatid length
(Kaplan et al., 2010), implying that even after 9 hr of individualiza-
tion, the caspase activity levels do not reach the threshold that
triggers apoptosis.
Here, we uncover a mitochondrial-based mechanism that
limits the rate of caspase activation in spermatids. We identified
a Krebs cycle component, the ATP-specific form of the succinyl-
CoA synthetase (SCS) b subunit (A-Sb) that specifically binds
to Cul3T and Klhl10. A-Sb is localized to mitochondria during
all stages of spermatogenesis, but its levels are significantly
increased at the onset of spermatid individualization. We
demonstrate that knockdown or mutation in A-Sb abrogates
caspase activation and spermatid individualization, a phenotype
reminiscent of mutations in the CRL3 complex. Spermatid mito-
chondrial fractionation analyses reveal that a relatively large
portion of the CRL3 complex is also localized to the mitochon-
dria, and that this depends, at least in part, on A-Sb. Moreover,
the mitochondrial portion of the complex was preferentially acti-
vated, and A-Sb could significantly enhance/induce the activity
of the CRL3 complex in a Drosophila transgenic system and in
cultured cells. We then identify two major isoforms of A-Sb,
one of which is testis specific and localized to the surface of
the mitochondria. We demonstrate that the activating arm
(i.e., A-Sb) competes with the inhibitory arm (mediated by Soti)
for binding to the CRL3 complex, suggesting an antagonistic
interplay. Finally, ubiquitination assays in cell culture followed
by in situ domain swapping experiments between the two func-
tionally redundant b subunits of SCS, A-Sb and the GTP-specific
SCSb (G-Sb), uncouple the metabolic role (Krebs cycle) and the
structural role (CRL3 activation) of A-Sb in spermatids. Collec-
tively, this work uncovers a mode of CRL regulation by a moon-
lighting function of a Krebs cycle component, which ultimately
reduces the caspase activation rate by at least 60% through
restriction of the source of caspase activation to the vicinity of
the mitochondria.
RESULTS
The Krebs Cycle Enzyme Subunit A-Sb SpecificallyInteracts with Cul3T and Klhl10The Drosophila cullin-3 locus encodes for two isoforms, somatic
(Cul3S) and testis-specific (Cul3T), both of which share the bulk of
the protein except for a unique 181 amino acid (aa) region at the
N terminus of Cul3T (also called the TeNC domain) (Figure 1A;
Arama et al., 2007). Whereas the precise function of the TeNC
domain is unknown, transgenic rescue analyses in spermatids
suggested an important role of this region in the proper activa-
tion of caspases in spermatid individualization. Interestingly,
E) Klhl10 and (F) Cul3T were immunoprecipitated and the presence of A-Sb(ckr),
d CD8-GFP construct controlled for transfection efficiency. (F) Note the slightly
Developmental Cell 37, 15–33, April 4, 2016 17
Figure 2. A-SbT Is Localized to the Mitochondria during All Stages of Spermatogenesis, and Its Expression Level Significantly Increases at
the Onset of Individualization
(A) Schematic structure of the A-SbT genomic region within a BAC clone, which was engineered to include GFP fused in-frame to the C-terminal end of A-SbT,
therefore specifically labeling A-SbT and not A-SbS. Exons are depicted by bars, while introns are indicated by thin lines. Black bars indicate coding sequences;
gray bars indicate UTRs; green bar indicates the GFP insertion.
(B–K) A-SbT expression in the testis (green) revealed by either (B–E, I–K) the GFP insertion described in (A) or (F–H) an anti-A-Sb antibody. Nuclei are in blue (DAPI)
and F-actin (phalloidin), which marks the individualization complex (IC), is in red.
(B–D) A-SbT is expressed in patterns reminiscent of mitochondria during early spermatogenesis stages. Shown are (B) pre-meiotic spermatocytes (white arrows)
and post-meiotic round spermatids (yellow arrows pointing at the nebenkerns). (C) Dividing germ cells during the anaphase/telophase stages of meiosis I (a white
arrow pointing at mitochondria between two dividing germ cells), and (D) early elongating spermatids (very early and more advanced elongating spermatid cysts
are indicated by a yellow asterisk and white arrowheads, respectively).
(legend continued on next page)
18 Developmental Cell 37, 15–33, April 4, 2016
although the TeNC domain is not sufficient to bind to Klhl10 by
itself (and Cul3S can still bind to Klhl10 but with very low effi-
ciency), this domain significantly increases the binding efficiency
of Cul3T to Klhl10 (Arama et al., 2007). Therefore, we hypothe-
sized that the activity of the CRL3 complex might be modulated
through the regulation of the TeNC domain of Cul3T in sperma-
tids. To explore this idea, we used the yeast two-hybrid (Y2H)
approach and tested the ability of an isolated TeNC domain to
bind to candidate proteins, originally identified as binding part-
ners of Cul3T in a Y2H screen (Arama et al., 2007). We identified
a polypeptide encompassing about a third of the A-Sb protein
encoded by the CG11963 gene. This polypeptide includes the
CoA-ligase domain and was termed the Cul3/Klhl10 interacting
region or ckr (Figures 1A and 1B).
The mitochondrial enzyme succinyl-CoA synthetase or SCS is
best known for its key functions in the Krebs cycle, ketone meta-
bolism and heme synthesis (Berg et al., 2002). In the Krebs cycle,
it catalyzes the reversible conversion of succinyl-CoA and ADP
or GDP to succinate and ATP or GTP (Johnson et al., 1998).
Active SCS enzymes are heterodimers composed of a and b
subunits, in which the b subunit may be switched between
A-SCSb and G-SCSb, resulting in an ADP-forming SCS and a
GDP-forming SCS, respectively (Fraser et al., 1999, 2000).
A-Sb and G-Sb are the respective Drosophila homologs of
A-SCSb and G-SCSb, but more recent studies suggested
that A-Sbmay also have other unconventional functions outside
of the mitochondria, including at the plasma membrane and at
centrosomal microtubules in Drosophila (Gao et al., 2008;
Hughes et al., 2008). Interestingly, the latter function, which is
required for proper centrosome duplication, involves the associ-
ation of A-Sb (also dubbed Skap for SkpA-associated protein)
with Skp1, the adapter protein in CRL1 complexes (Hughes
et al., 2008). Since Skp1 is known to assume a three-dimensional
fold highly similar to that of the BTB domain (Figure 1C; Stogios
et al., 2005), we tested whether A-Sb may also interact with the
BTB domain of Klhl10. Using both Y2H and co-immunoprecipi-
tation (coIP) assays in Drosophila S2 cells, we show that the
ckr domain of A-Sb (A-Sb(ckr)) can also bind to Klhl10 in a BTB-
domain-dependent fashion (Figures 1D and 1E and data not
shown). Of note, the BTB domain is also required for the binding
of Klhl10 to Cul3T, and a single point mutation in this domain ab-
rogates the function of the CRL3 complex in spermatids (Arama
et al., 2007).
Interestingly, when we validated the binding between A-Sb
and Cul3T using a similar coIP experiment (Figure 1F), the
interaction between A-Sb and Cul3T appeared to be much
weaker than that between A-Sb and Klhl10 (namely much more
A-Sb was co-immunoprecipitated with Klhl10 than with Cul3T;
compare Figures 1E and 1F). We therefore asked whether the
(E and F) A-SbT levels dramatically increase in individualizing spermatids, as reve
are indicated by white arrows.
(G, H, J, and K) A-SbT is not extruded with the cytoplasmic contents to the WB
arrowheads point at post-individualized regions, which still contain A-SbT; white
mature sperm, which still express A-SbT), and (K) a post-individualized mature c
yellow arrow in E).
(I) A-SbT level specifically increases at the onset of individualization. Whereas jus
dramatic increase is recorded at the onset of individualization when the IC (red; wh
10 mm; (B, D, and I), 20 mm; (K), 50 mm; (E–H), 100 mm.
See also Figure S2.
binding of A-Sb to the CRL3 complex might be stronger than
to each of the single components of the complex alone. For
this, a reciprocal coIP experiment was carried out in which
A-Sb was immunoprecipitated in the presence of either Flag-
tagged Cul3T, Flag-tagged Klhl10, or both. Whereas the levels
of bound Klhl10 were similar in both setups, the Cul3T levels
dramatically increased by 70% in the context of the complex
(Figures S1A and S1B). Given that, regardless of A-Sb, Cul3Tand Klhl10 can physically interact, this result implies a role of
A-Sb as part of a tertiary complex rather than as an interactor
of each of the single components alone.
To conclude our physical interaction analyses, we explored
the ability of a transgenically expressed and functionally active
GFP fusion of A-Sb (A-SbT-GFP; see also Figure 2A) to generate
a complex with endogenous Cul3T and Klhl10 in spermatids. To
control for non-specific binding, we also tested a transgenic line
carrying the testis-specific inner mitochondrial protein (Don
Juan) DJ-GFP (Santel et al., 1998; Bazinet andRollins, 2003). Us-
ing a gel mobility retardation assay, dissected testes from these
fly lines were subjected to chemical cross-linking followed by
electrophoretic mobility analysis of the protein extracts. In this
setup, the mobility of a protein complex is slower than that of
either protein alone. Significantly, a strong mobility shift of both
Cul3T and Klhl10 was detected in the A-SbT-GFP extract but
not the DJ-GFP extract (Figure S1C). Furthermore, A-SbT-GFP
also exhibited a similar mobility shift and appeared at the same
molecular weight as Cul3T and Klhl10, indicating that A-Sb can
form a complex with Cul3T and Klhl10 in spermatids.
Expression of a Testis-Specific Isoform of A-SbSignificantly Increases at the Onset of IndividualizationTo investigate the physiological relevance of the interaction be-
tween A-Sb and the CRL3 complex, we first asked whether the
expression pattern of A-Sbmay support a role during spermatid
individualization. Gene annotation data (FlyBase) for A-Sb sug-
gest that this locus encodes for twomain, long and short, protein
isoforms (hereinafter respectively referred to as A-SbT, for testis-
specific, and A-SbS, for somatic; see also in the next paragraph).
While sharing most of the coding region, the long isoform (A-SbT)
contains a unique exon at the 30 end, encompassing extra coding
sequences at the C terminus and an alternative 30 UTR (Figures
S2A and S2B). In addition, sequencing analyses of several ex-
pressed sequence tags and RT-PCR products from different tis-
sues, revealed that A-SbT contains an additional tiny exon with
only one codon for the amino acid cysteine (Cys-253) (Figures
S2A and S2B). Both the testis-specific C-term region (tail) and
the Cys-253 are not covered by the ckr domain, implying that
they are probably not directly involved in the A-Sb/CRL3 com-
plex interaction (Figure S2B).
aled both by (E) the A-SbT-GFP expression and (F) the anti-A-Sb antibody. CBs
s during individualization. Shown are (G, J) individualizing spermatids (white
arrows point at CBs), (H) a torn seminal vesicle (SV; a yellow asterisk labels
yst on its way to the seminal vesicle (an enlargement of the cyst marked by a
t prior to individualization A-SbT-GFP level is relatively low (white arrowhead),
ite arrow) is assembled in the vicinity of the nuclei (blue). Scale bars in (C and J),
Developmental Cell 37, 15–33, April 4, 2016 19
To examine which of the A-Sb isoforms is expressed in the
male germ cells, we performed comparative RT-PCR experi-
ments using specific primer pairs for the long and short isoforms
(arrows in Figure S2A), and with RNA from wild-type testis, a
germ-cell-less line (male progeny of oskar mutant females), a
control knockdown line (cont_IR), and a germ-cell-specific
A-Sb knockdown line (A-Sb_IR). Note that the wild-type RNA
was PCR amplified for saturation (35 cycles) to generate the
two reference bands (WT testes; Figures S2C and S2D). While
A-SbT was highly expressed in control testes (black asterisk),
A-SbS was only weakly expressed there (red asterisk; Figures
S2C and S2D). In accordance, knockdown of A-Sb in the pre-
meiotic germ cells specifically targeted the A-SbT isoform, as
in these testes the A-SbS band prevailed (Figures S2C and
S2D). Finally, A-SbT (but not A-SbS) was absent from the germ-
cell-less male reproductive tracts, indicating that the expression
of the long isoform is restricted to the germ cells while the short
isoform expression is mainly confined to the soma (Figure S2C).
It is noteworthy that in addition to germ cells, the testis also con-
tains some somatic cells, which may explain the (low) presence
of A-SbS in this tissue.
We next explored the protein distribution of the long A-Sb
isoform in the male germ cells. For this, we first used the recom-
bineering technique (Venken et al., 2009) to introduce the GFP
coding region at the C-terminal end of A-SbT (just prior to the
stop codon), all within a large genomic BAC clone encompassing
the entire A-Sb locus, as well as some flanking genes (this
construct is termed A-SbT-GFPR-BAC; Figure 2A). As a control,
GFP was also introduced at the A-SbS C terminus within the
same BAC clone (A-SbS-GFPR-BAC; data not shown). Transgenic
flies carrying these large clones (inserted at the same genomic
point) were generated and their testes were analyzed using
confocal microscopy. Consistent with the idea that A-SbS is
the somatic isoform, flies carrying the A-SbS-GFPR-BAC trans-
gene displayed mitochondrial GFP expression in essentially all
examined tissues (data not shown). On the other hand, GFP
expression in flies carrying the A-SbT-GFPR-BAC transgene was
restricted to the male germ cell mitochondria throughout sper-
matogenesis (Figures 2B–2D). Significantly, GFP expression in
these flies dramatically increased in elongated spermatids
undergoing individualization compared with earlier elongating
spermatids (Figure 2E). Closer examination revealed that this
elevation in GFP expression specifically occurs at the onset of
(A-Sbcc/Df; a P element insertion in the second intron in trans to
a small chromosomal deletion of the gene locus; Figure S2A).
Of note, whereas in the knockdown, the A-SbT isoform was spe-
cifically targeted (see the reduction in the mRNA and protein
levels in Figures S2C–S2E, 3A, and 3B), the protein (but not
mRNA) levels of both the long and short isoforms were reduced
in the mutant (Figures S2D and S2E). Importantly, an almost
complete block of caspase activation was detected in sperma-
tids from both the knockdown and mutant A-Sb flies (Figures
3C–3F). Consistently, although the assembly of ICs was readily
detected in A-Sb-deficient spermatids, these ICs were unable
to move and collect the spermatids’ cytoplasmic contents, lead-
ing to individualization failure and male sterility (Figures 3C–3F
and data not shown). Finally, re-introduction of the A-SbT-
GFPR-BAC transgene to the A-Sbmutant males restored caspase
activation, spermatid individualization, and fertility of these flies
(Figure 3G). Taken together, these findings are consistent with
a role of A-SbT in the regulation of the CRL3 complex and cas-
pase activation in spermatids.
Similar to mammals, the effector caspases in Drosophila
display DEVD cleaving activity (DEVDase; Fraser et al., 1997;
Song et al., 1997; Florentin and Arama, 2012). Furthermore, the
adult testis also contains a pronounced DEVDase activity, and
this activity is significantly reducedwhen caspase activation dur-
ing spermatid individualization is blocked (i.e., in cytochrome c
and cul3T mutant testes (Arama et al., 2006, 2007)), suggesting
that individualizing spermatids are the major source of caspase
activity during spermatogenesis. To provide independent evi-
dence for a requirement of A-Sb in caspase activation in sperma-
tids, we measured DEVDase activity in extracts from A-Sb_IR
knockdown testes. Wild-type testes and cul3T homozygous
mutant (cul3mds1) testes served as positive and negative con-
trols, respectively. Portions of the testis extracts were used as
controls to determine the relative amounts of total protein in
each extract using the anti-b-tubulin antibody (Figure 3I).
Whereas lysates of wild-type testes displayed relatively high
levels of DEVDase activity, activity level in A-Sb_IR knockdown
testes was significantly reduced to a level reminiscent of that in
the cul3mds1 mutant testes, indicating compromised caspase
activation during spermatid individualization (Figure 3H). Consis-
tent with their specific role in spermatid individualization, addi-
tion of the potent caspase inhibitor Z-VAD-FMK to the lysates
further reduced the residual (low) level of caspase activity in
these mutants, suggesting that other caspase activity-associ-
ated spermatogenesis processes (e.g., apoptosis of some testis
somatic cells) are not affected by A-Sb and the CRL3 complex
(Figure 3H).
A-Sb Activates the CRL3 ComplexDuring spermatogenesis, the mitochondria undergo extensive
structural organization, beginning at the round spermatid
stage with the aggregation of all the mitochondria and subse-
quent fusion to a giant sphere called the Nebenkern. When the
Figure 3. Inactivation of A-Sb Abrogates
Caspase Activation during Spermatid Indi-
vidualization
(A–D) Germ-cell-specific knockdown of A-Sb
using the bam-Gal4 line to drive expression of
(A and C) a control RNAi transgene (cont_IR) and
(B and D) an A-Sb RNAi transgene (A-Sb_IR).
(A–D) Testes were stained to visualize the nuclei
(blue), individualization complex (IC; red), and
either (A and B) A-Sb (green) or (C and D) cleaved/
activated effector caspases (cCasp.3; green).
(E–G) Testes from (E) wild-type flies, (F) hypomor-
phic A-Sb mutants, and (G) hypomorphic A-Sb
mutants carrying the A-SbT-GFPR-BAC transgene
(described in Figure 2A), were stained to visualize
activated effector caspases (red), the nuclei (blue),
and A-SbT-GFP expression (G; green). Scale bars,
100 mm.
(A–G) To obtain high-resolution picture of the entire
testis, serial confocal images were captured and
stitched together. The reconstructed pictures were
rotated to position the elongating spermatids in
a top (nuclei)-to-bottom orientation. Gaps in the
reconstructed images (due to rectangle-shape
cropping) were filled with gray background.
(H) Thediagramdepictsacaspase-3-like (DEVDase)
activity for testes from wild-type (wt), cul3T null
mutant (cul3mds1), and A-SbT knockdown (A-Sb_IR)
flies. Note the suppression of the activity after
treatment with the caspase-3 inhibitor Z-VAD-FMK.
DEVDase activity is presented as relative activity
with respect to the activity in thewt testis. Each time
point represents an average (mean ± SEM) of three
independent experiments.
(I) Western blot analysis to control for the protein
levels in the testis extract used in the DEVDase
activity assay.
flagellum starts growing, the Nebenkern unfolds into two mito-
chondrial masses that elongate down the side of the axoneme,
giving rise to two, major and minor, �2 mm long, mitochondrial
derivatives (Fuller, 1993). To test the idea that A-Sbmay regulate
the CRL3 complex in spermatids, a modified mitochondrial frac-
tionation protocol was used to isolate the spermatid mitochon-
drial derivatives. Indeed, the efficiency of this protocol was
confirmed by western blot analysis of mitochondrial and cyto-
solic testis fractions using antibodies against the mitochondrial
inner-membrane protein, ATP synthase-a (ATPsyn-a), the endo-
plasmic reticulum (ER) protein, Sec16, and the cytosolic protein,
GAPDH. Whereas the mitochondrial fraction was highly specific,
displayingminor or no contamination by cytosolic or ER proteins,
the cytosolic fraction was somewhat less pure, containing some
contamination from the mitochondria (Figure 4A). In agreement
Dev
with the immunofluorescence data, A-Sb
was mainly present in the mitochondrial
fraction (60%; Figure 4A, lanes 1 and 4).
Critically however, relatively large por-
tions of Cul3T and Klhl10 were also local-
ized at the mitochondria (45%), and their
presence in this fraction was partially
abrogated upon knockdown of A-Sb (Fig-
ures 4A, 4B, S3A, and S3B).
Since A-SbT is required for caspase activation in spermatids
and can bind to the CRL3 complex, a plausible hypothesis is
that the CRL3 complex may get activated upon binding to
A-SbT at the mitochondria. This idea is further supported by
the finding that transfection of A-Sb(ckr) with Cul3T and Klhl10,
but not of the latter two proteins alone, resulted in an additional
(shifted) Cul3T band of slightly higher molecular mass, suggest-
ing that A-Sb binding may promote neddylation/activation of this
complex in cultured cells (the upper arrow in Figure 1F, lane 5).
More importantly, a similar phenomenon was also apparent
in spermatids, where a significant enrichment (more than
3.5-fold) of the neddylated (activated) form compared with the
unneddylated form of Cul3T was detected in the testis mito-
chondrial fraction but not in the cytosolic fraction (lanes 1 and
4 in Figure 4A and quantifications of the relative levels of the
elopmental Cell 37, 15–33, April 4, 2016 21
Figure 4. A-Sb Activates the CRL3 Complex
(A) Testis mitochondrial and cytosolic fractions from flies expressing a control RNAi or an A-Sb RNAi, and from cul3T mutant flies (cul3mds1). The blot was
sequentially exposed to several antibodies to reveal the endogenous proteins indicated at the left. ATPsyn-a, an inner mitochondrial membrane protein; Sec16,
an ER membrane protein; GAPDH, a cytosolic protein.
(B) Quantification of protein levels in the mitochondrial fractions in (A) relative to the ATPsyn-a levels. Top, the neddylated (Nedd) and unneddylated (Unnedd)
forms of Cul3T. Bottom, Klhl10 protein levels.
(legend continued on next page)
22 Developmental Cell 37, 15–33, April 4, 2016
two forms in the mitochondrial fraction in Figure 4B, upper
graph). Note that the upper band corresponds to the neddylated
form of Cul3T, as confirmed using an anti-Nedd8 antibody (Fig-
ures 4A and S3B). Consistently, in A-Sb knockdown, the same
bands were similarly reduced using both anti-Cul3 and anti-
Nedd8 antibodies (lanes 1 and 2 in Figure 4A). In addition, in
testes from the Klhl10 mutants where no active CRL3 complex
is formed, the upper (neddylated) band of Cul3T is missing (Fig-
ure S3B), as was reported for other CRL complexes (Merlet et al.,
2009). Altogether, these data clearly indicate that the upper band
of Cul3T indeed represents the neddylated form of this protein,
which is enriched on the mitochondria. Another remarkable
piece of evidence along these lines is the stabilization (�2-fold)
of Klhl10 in the mitochondrial, but not the cytosolic, fraction of
the cul3T mutant (cul3mds1 in Figure 4A, lanes 3 and 6, and the
quantification in Figure 4B, bottom graph). This stabilization is
attributed to constant auto-ubiquitination and degradation of
the substrate recruitment proteins in CRL complexes (Bosu
and Kipreos, 2008), which we previously also demonstrated for
Klhl10 (Kaplan et al., 2010), and which therefore also reflects
the activity state of the CRL3 complex in spermatids. Consis-
tently, the accumulation of Klhl10 on the mitochondria of cul3Tmutant spermatids depends on the presence of A-SbT, as
no accumulation of Klhl10 was detected in the mitochondrial
fraction when A-Sb was further inactivated (lanes 3 and 4 in
Figure S3A). Altogether, these findings indicate that the CRL3
complex is highly activated at the mitochondria, but whether
the binding to A-SbT is sufficient for this activation or additional
factors are required was still unresolved.
To address this question, we first tested the ability of A-SbT to
activate the CRL3 complex in an ectopic system, the Drosophila
compound eye. It is noteworthy that although this system was
previously shown to reliably reflect the activity of the CRL3
complex by virtue of the eye size (Kaplan et al., 2010), eye photo-
receptor cells are distinct from spermatids and may thus reflect
induction of different molecular pathways downstream of the
CRL3 complex activity (for instance, the CRL3 complex may
engage different substrates in the different cells, but the readout,
the eye size, shall still reflect its activity state). Using the eye-
specific GMR-Gal4 driver line, several UAS-dependent trans-
genes were tested. Whereas transgenic expression of GFP and
A-SbT did not cause any gross eye defect, a highly reduced
size to complete elimination of the eye was induced following
expression of Klhl10 alone or together with Cul3T, respectively
(compare Figures 4C, I, II with 4C, III, V). The pronounced effect
obtained by individually expressing Klhl10 is attributed to its abil-
ity to also bind the endogenous somatic Cul3 isoform, albeit with
less efficiency than the binding to the Cul3T isoform (Kaplan
et al., 2010), as demonstrated by suppression of the eye effect
in the cul3S heterozygous mutant background (cul3gft06430, a
specific mutation in the cul3S but not the cul3T isoform, in Fig-
(C) Transgenesis experiments using the Drosophila adult compound eye. The tr
imaginal discs. Representative eyes from newly eclosed females are shown. Typic
did not eclose and died in their pupal stages.
(D) Ubiquitination experiment in S2 cells using Cyt-c-d as substrate. Cells were tr
ubiquitin construct, and increasing concentrations A-Sb(ckr). A CD8-GFP construc
inhibitor, Z-VAD-FMK, and the proteasome inhibitors, Velcade and MG132. Cyt
revealed by anti-ubiquitin antibody.
ure S3C). Importantly, transgenic expression of A-SbT enhanced
ing that A-SbT is a positive regulator of the CRL3 complex (Fig-
ure 4C). Furthermore, the facts that Cul3S lacks the TeNC
domain that is required for binding to A-SbT, and that expression
of A-SbT was sufficient to enhance the Klhl10-induced eye
phenotype, imply that binding of A-SbT to Klhl10 is sufficient
to trigger at least some activation of the CRL3 complex, even
without the binding of A-SbT to Cul3.
To ultimately test how binding to A-SbT may affect the activity
of the CRL3 complex toward its substrates, we performed ubiq-
uitination assays by transfecting S2 cells with the CRL3 complex
components (i.e., Cul3T, Klhl10, and the somatic RING domain
protein Roc1a), the testis-specific cytochrome c, Cyt-c-d, which
is required for caspase activation in spermatids (Arama et al.,
2006) and was recently found to be a bona fide target of this
complex in vivo (L.R., L.A., and E.A., unpublished data), and
either with or without A-Sb(ckr). Critically, western blot analysis
demonstrated that, whereas the CRL3 complex components
could not induce ubiquitination of Cyt-c-d (above background
levels), neither individually nor combined (Figure 4D, lane 4),
the addition of A-Sb(ckr) resulted in marked ubiquitination of
this substrate (Figure 4D, lane 5). It also appears that the stoichi-
ometry of A-Sb(ckr) and the CRL3 complex components is crit-
ical, as increased concentrations of the former failed to activate
this complex (Figure 4D, lanes 6–8). We conclude that A-Sb is an
activator of the CRL3 complex and that this may be independent
of other mitochondrial-associated factors, as A-Sb(ckr), which
can bind to the CRL3 complex but reside in the cytosol (Figure 1
and data not shown), and be sufficient for this activation.
A-Sb Mediates Localization of the CRL3 Complex to theSurface of the MitochondriaAn intriguing question regarding these findings is where, at the
subcellular level, the binding between A-SbT and the CRL3 com-
plex occurs, since A-Sb is normally localized at the matrix of the
mitochondria as part of its role in the Krebs cycle, whereas
caspase activation by the CRL3 complex occurs in the cytosol.
A reconciling hypothesis is that the interaction occurs at the sur-
face of the mitochondria, such that A-SbT is docked on the outer
membrane of the mitochondria facing the cytosol. To test this
idea, we first examined the effect of A-SbT expression on the
mitochondrial localization of the CRL3 complex in cell culture.
For this, S2 cells were transfected with Cul3T and Klhl10, with
or without A-SbT. Mitochondrial and cytosolic fractions were
then examined by western blot for the presence of these three
proteins. In agreement with the idea that A-SbT mediates mito-
chondrial localization and activation of the CRL3 complex in
spermatids (Figure 4A), increased mitochondrial levels of Cul3Tand Klhl10 were detected in cells transfected with A-SbT, and
this was accompanied by a reciprocal decrease in their levels
ansgenes indicated below each panel were ectopically expressed in the eye
al of mutants with extremely underdeveloped eyes, flies from the genotype in VI
ansfected with one or more of the indicated constructs (on the left), Drosophila
t controlled for transfection efficiency. Cells were also treated with the caspase
-c-d was immunoprecipitated and the presence of ubiquitinated Cyt-c-d was
Developmental Cell 37, 15–33, April 4, 2016 23
in the cytosolic fractions (Figure 5A and the quantification at the
bottom).
We next explored the idea that the interaction between A-SbTand the CRL3 complex may occur at the mitochondrial surface.
Testis mitochondrial fractions were subjected to a Proteinase K
(PK) protection assay, which helps define compartmental locali-
zations of mitochondrial proteins based on their sensitivity to PK
proteolysis (i.e., proteins residing in internal mitochondrial com-
partments are better protected against PKproteolysis than those
localized on the mitochondrial surface; illustrated in Figure 5B).
Critically, whereas the mitochondrial inner-membrane protein
(ATPsyn-a) and the somatic A-Sb isoform (A-SbS) were highly
resistant to cleavage by this promiscuous protease, A-SbT was
highly sensitive to this treatment (Figure 5C; A-SbT and A-SbSare inblackand red text, respectively, andcorrespond to the illus-
tration in Figure 5B). Therefore, at least someportion of the A-SbTprotein is localized at an external mitochondrial compartment,
while A-SbS is found in the inner compartments. Furthermore,
the mitochondrial portions of Cul3T and Klhl10 were even more
sensitive to PK treatment than A-SbT (Figure 5C), suggesting
that their mitochondrial localization may be unstable and thus
may likely reflect localization by protein-protein interaction rather
than direct docking at the mitochondrial surface.
Because of the importance of this point to the overall model of
CRL3 complex activation by A-SbT in spermatids, we wanted to
confirm these findings using an independent approach and
obtain an estimate of the level of A-SbT at the spermatid mito-
chondrial surface. For this, we prepared testes for immunocyto-
logical localization at the electron microscopy (EM) level, a tech-
nique commonly known as immuno-EM. Using an anti-GFP
antibody, we performed immunogold labeling of A-SbT-GFP
on ultrathin testis sections from transgenic flies carrying the
A-SbT-GFPR-BAC construct (Figure 2A). To control for the
inherent inaccuracy in the position of the gold particles, which
are usually 15–30 nm away from the site to which the primary
antibody is bound (Hermann et al., 1996), we also labeled the in-
ner mitochondrial protein DJ, using flies expressing the DJ-GFP
transgene (Santel et al., 1998; Bazinet and Rollins, 2003). In addi-
tion, for the purpose of quantification, this technical limitation
was further buffered by considering an annular ring area as the
surface area (encompassing 24% of the total mitochondrial
area) and not merely themitochondrial circumference. We there-
fore hypothesized that a mitochondrial surface protein shall
display an average number of gold particles, which is signifi-
cantly above the chance level of random particle distribution
(i.e., 24%) in this ring area. Using transmission EM (TEM), we de-
tected specific labeling of gold particles over the mitochondria in
cross-sections of elongated spermatids expressing A-SbT-GFP
and DJ-GFP, but not in wild-type controls (Figure 5D and data
not shown). Importantly, quantifications of gold particle numbers
in more than 100 spermatids from each genotype revealed that
whereas more than 80% of the DJ-GFP particles were found at
the inner compartments (setting the inherent inaccuracy level
to less than 20%), half of the A-SbT-GFP was localized at the
mitochondrial surface (Figure 5E). Given the significant bias of
A-SbT-GFP at the surface area, which is highly above both the
inaccuracy level and the chance level, we conclude that about
50% ± 20% of the total A-SbT protein in elongated spermatids
is found at the mitochondrial surface.
24 Developmental Cell 37, 15–33, April 4, 2016
A-SbT Antagonizes the Binding of Soti to the CRL3ComplexThe findings that the CRL3 complex is tightly regulated both by
an activating arm (i.e., A-SbT) and an inhibitory arm (mediated
by the pseudosubstrate inhibitor Soti; Kaplan et al., 2010), raised
the possibility that the binding of A-SbT to the CRL3 complex
may antagonize Soti binding to this complex. To test this idea,
we performed competition assays by transfecting S2 cells
with GFP-tagged Soti, Cul3T, Klhl10, and increasing levels of
A-Sb(ckr), followed by coIP experiments with anti-GFP anti-
bodies. Whereas both Cul3T and Klhl10 were readily present in
the Soti-GFP immunoprecipitates, their levels were dramatically
reduced upon A-Sb(ckr) transfection (Figure S4, lanes 5 and 6;
also see the quantifications at the bottom). Notably, A-Sb(ckr)caused a 50% drop in the levels of Cul3T and Klhl10 that bind
to Soti at a transfection level of 0.15 mg, which is about four times
less than the transfection levels of all the other components
(0.66 mg) (Figure S4). It is also interesting to note that, despite
the progressive reduction in the bound Cul3T and Klhl10 levels
when more A-Sb(ckr) was transfected, the bound A-Sb(ckr) levels
were also increased (Figure S4, third panel, lanes 6–9). Whereas
this increase may be partially attributed to direct binding be-
tween A-SbT and Soti (Figure S4, lane 3), it is more likely that
the increase in A-Sb(ckr) concentration may lead to an increase
in the bonding affinity between A-Sb(ckr) and the remaining
bound Cul3T and Klhl10. Collectively, these findings suggest
that A-SbT is a potent antagonist of Soti-Klhl10 interaction,
revealing amode of CRL complex regulation. Furthermore, these
results also suggest that the binding of A-SbT to the complex
may change its state from an inactive conformation (where it
prefers binding to Soti) to an active conformation (where it may
prefer binding to its true substrates), which further supports an
active role for A-Sb in promoting the activation state of the
CRL3 complex rather than merely stabilizing the complex on
the mitochondria.
A-Sb but Not the Alternative b Subunit of SCS, G-Sb,Restores Full Caspase Activation and SpermatidIndividualizationWe next wanted to further uncouple between the metabolic role
of A-Sb in the Krebs cycle and its function in the activation of the
CRL3 complex and caspases in spermatids. In the Krebs cycle,
the two SCS enzymes, which are differentiated by virtue of their b
subunits (i.e., A-Sb or G-Sb), may alternate to generate succinate
and ATP or GTP, respectively (Bridger et al., 1987; Johnson
et al., 1998; Lambeth et al., 2004). We therefore reasoned that
if the activation of the CRL3 complex by A-Sb merely occurs
through its metabolic role in the Krebs cycle, transgenic G-Sb
might be able to substitute for the loss of A-Sb in mutant sperma-
tids; whereas in the case A-Sb activates this complex in a
manner distinct from its metabolic function, no rescue of these
mutant spermatids is expected. To directly test this idea, we
generated transgenes that cover the entire coding regions of
either A-SbT or G-Sb (both tagged with a C-terminal HA) under
the control of the A-Sb promoter and 50 and 30 UTRs, and in-
serted them at a defined site of the Drosophila genome (Figures
6A, III and IV). These transgenic flies were then crossed to the
hypomorphic A-Sb mutant, A-Sbcc/Df, tested for protein expres-
sion (Figure 6B), and analyzed for their ability to rescue the
Figure 5. A-SbT Is Localized to the Spermatid’s Mitochondrial Surface
(A) Mitochondrial and cytosolic fractions of S2 cells transfected with one or more of the indicated constructs. To reveal the proteins indicated at the left, the blot
was sequentially exposed to the relevant antibodies. The quantification at the bottom represents the Cul3T and Klhl10 protein levels in the mitochondrial and
cytosolic fractions relative to the levels of the ATPsyn-a and GAPDH proteins, respectively. C, Cul3T; K, Klhl10; S, A-SbT.
(B) Schematic description of the Proteinase K (PK) protection assay used in (C).
(C) Mitochondrial testis fractions from wild-type flies were incubated with the indicated concentrations of PK. To reveal the endogenous proteins indicated at the
left, the blot was sequentially exposed to the relevant antibodies. The respective black and red colors of A-SbT and A-SbS correspond to the schematic protein
colors in (B).
(D) Representative immuno-electron (EM) micrographs of cross-sections through elongated spermatids expressing the A-SbT-GFP or DJ-GFP transgenes,
displaying specific GFP labeling with gold particles (black dots) over the mitochondria. MD, major mitochondrial derivative; Axo, axoneme. Each gold particle
represents a single protein either located at the mitochondrial surface (green arrowheads) or at the inner mitochondrial area (red arrowheads). Scale bar, 200 nm.
(E) Statistical analysis of the gold particle distributions over the mitochondrial surface and inner areas (shown in D) in more than 100 spermatids of each of the
indicated genotypes. For comparison, the chance level of random particle distribution for each of these areas is also indicated in a separate column (24% and
76% surface versus inner areas, respectively).
Developmental Cell 37, 15–33, April 4, 2016 25
Figure 6. A-SbT and G-Sb Domain Swapping and Rescue Studies Reveal Both Interchangeable and Specific Domains Essential for the
Non-metabolic Function of A-SbT in Spermatids
(A) Schematic structures of full-length A-SbT (metal blue) and G-Sb (metal green) constructs (III, IV) and four additional A-SbT/G-Sb hybrid constructs (V–VIII).
Identifier Roman numerals on the left correspond to the experiments presented in (B–E).
(B) Validation of transgenic expression by western analysis of protein extracts from testes inserted with the indicated rescue constructs using the anti-HA
antibody. b-Tubulin levels served as loading controls. The numbers at the bottom represent expression levels of each transgene relative to the b-tubulin levels.
(C) Representative wild-type (I) and A-SbT mutant (II–VIII) testes were stained with anti-cleaved caspase-3 antibody (green), phalloidin (red; ICs), and DAPI (blue,
nuclei). The schematic structures of the inserted transgenes are indicated at the bottom of each image (III–VIII). Individualizing cysts are detected by the presence
of CBs and WBs (arrowheads). Scale bar, 100 mm.
(D) Quantification of cleaved caspase-3 staining levels (divided into four color-coded categories) in testes with the indicated constructs. The number of testes
examined is indicated above thecorrespondingbars,whereas the fertility status of thedifferent genotypes is indicatedbelow thebars either as fertile (F) or sterile (S).
(legend continued on next page)
26 Developmental Cell 37, 15–33, April 4, 2016
sterility phenotypes associated with this mutant. As expected,
the A-SbT transgene fully restored caspase activation, sper-
matid individualization, and male fertility (Figures 6C and 6D).
In contrast, no rescue of spermatid individualization and male
fertility was obtained with the G-Sb transgene, albeit some
mild increase in caspase activation was observed (Figures 6C
and 6D; compare genotypes II and IV). These results are consis-
tent with our model of a Krebs-cycle-independent role of A-SbTin the activation of the CRL3 complex and consequent caspase
activation in spermatids.
The ckr Domain of G-Sb Can Bind to and Activate theCRL3 Complex in Cell Culture, but Its N-TerminalDomain Precludes Appropriate Complex Activation inSpermatidsThe fact that G-Sb cannot substitute for the loss of A-Sb sug-
gests that A-Sb might have acquired some unique sequence
properties specialized for its role in spermatids. These se-
quences were evolved to either directly affect the binding to
and activation of the CRL3 complex, or indirectly facilitate
appropriate complex activation through the acquisition of other
important sperm-specific traits (such as improved mitochondrial
surface localization, folding, binding to additional components,
etc.). In Drosophila, A-SbS and G-Sb share 46% amino acid
sequence identity (60% similarity) over the entire protein, with
55% identity and 67% similarity over the ckr domain (i.e., the
C-terminal region of the protein), and 42% identity and 57% sim-
ilarity over the N-terminal region. In addition, as aforementioned,
A-SbT also contains a unique 51 aa C-terminal tail, absent in both
A-SbS andG-Sb. To start exploring the function and specificity of
the different A-SbT domains for CRL3 complex activation, we
first tested whether the G-Sb ckr domain equivalent (G-Sb(ckr))
may also bind to and activate this complex in cell culture. Inter-
estingly, coIP assays in S2 cells revealed that G-Sb(ckr) can also
bind to Klhl10 and Cul3T (Figures S5A and S5B). Moreover,
similar to A-Sb(ckr), G-Sb(ckr) was also sufficient to activate the
CRL3 complex in S2 cells, inducing Cyt-c-d ubiquitination under
concentrations of G-Sb(ckr) which are reminiscent of those of
A-Sb(ckr) (Figure S5C). Given the inability of the full-length G-Sb
to substitute for the loss A-Sb in spermatids, these findings sug-
gest that other sequences outside of the A-Sb ckr domain may
also be essential for CRL3 complex activation in vivo.
The A-SbT N-Terminal Domain, but Not Its UniqueC-Terminal Tail or the G-Sb N-Terminal Domain, Is AlsoRequired for CRL3-Induced Caspase Activation andSpermatid IndividualizationTo investigate the structural basis for the functional differences
between A-SbT and G-Sb in spermatids, we performed domain
swapping between these two proteins followed by functional
rescue analysis in A-Sbcc/Df mutant background. The first two
rescue constructs were designed to identify which of the two
main domains of A-SbT is responsible for its unique function
(E) PK protection assay with mitochondrial testis fractions from transgenic flies e
mutant. An anti-HA antibody was used to reveal the expressed transgenic prot
mitochondrial and cytosolic fractions, respectively. The corresponding quantifica
the loading controls).
See also Figure S5.
in vivo. For this, the N-terminal domain of A-Sb was fused to
the ckr domain of G-Sb, and vice versa, and both hybrid proteins
also contained the unique A-SbT C-terminal tail (Figures 6A, V
and VI). On the other hand, to examine the significance of the
unique A-SbT C-terminal tail, termed A-Sb(C-term), for the function
of this protein in spermatids, two additional constructs were
designed. One construct was solely composed of the A-SbSprotein sequence, which lacks the A-Sb(C-term), whereas the
other construct contained the G-Sb protein sequence fused to
the A-Sb(C-term) (Figures 6A, VII and VIII). The resulting protein se-
quences were placed under the control of the A-Sb promoter
and 50 and 30 UTRs, and four transgenic lines were generated.
To equalize other possible parameters that may strongly affect
protein expression levels, these transgenes were generated
similarly to the parental transgenes (i.e., the full-length A-SbTandG-Sb; Figures 6A, III and IV) and inserted at the same defined
site of the Drosophila genome. Since all the transgenes con-
tained an HA-tag at the C terminus, the relative expression levels
were determined using an anti-HA antibody, revealing only mild
variations in the expression levels of these transgenic proteins in
spermatids (Figure 6B).
Analyzing the ability of these transgenes to restore caspase
activation, spermatid individualization, and fertility in A-Sbcc/Df
mutant males, revealed that only constructs III, V, and VII were
able to restore all these three phenotypic traits, indicating that
the N-terminal domain of A-SbT is uniquely important for its spe-
cial role in spermatids (Figures 6C and 6D). Furthermore, consis-
tent with the cell culture binding assays and ubiquitination
studies, the ckr domains of A-Sb and G-Sb are also functionally
interchangeable in vivo, whereas the unique A-SbT C-terminal
tail appears to be dispensable for the function of the protein in
spermatids (Figures 6C and 6D). Similarly, the unique A-SbTcysteine was also dispensable for this testis-specific function,
as construct VII, which restored fertility in the A-SbT mutant,
does not contain this residue. Of note, amild increase in caspase
activation levels was recorded in mutant spermatids expressing
constructs IV, VI, and VIII, but this was not sufficient to trigger
individualization and restore fertility in these flies (Figure 6D).
Collectively, these results indicate that although the ckr domain
is sufficient to bind to and activate the CRL3 complex in cell cul-
ture, additional sequences residing at the N-terminal domain are
equally important to assure the appropriate Krebs-cycle-inde-
pendent function of A-Sb in spermatids.
The finding that the A-Sb N-terminal domain, which is not
required for binding to the CRL3 complex, still has a crucial
role in promoting the non-metabolic function of A-SbT in sperma-
tids, prompted us to explore possible effects on the subcellular
localization of the protein. For this, testis mitochondrial fractions
from transgenic flies expressing the A-SbT and G-Sb parental
rescue constructs (Figures 6A, III and IV) were subjected to a
PK protection assay. Similar to the endogenous A-SbT (Fig-
ure 5C), transgenic A-SbT was also abundant in the mitochon-
drial fraction and highly sensitive to PK proteolysis (Figure 6E).
xpressing the A-SbT and G-Sb rescue constructs in the background of A-SbTeins, while the ATPsyn-a and the b-tubulin served as loading controls for the
tions are presented in graphs, showing the transgenic protein levels (relative to
Developmental Cell 37, 15–33, April 4, 2016 27
Figure 7. An Integrated Model of Spatio-
temporal Restriction of the CRL3-Induced
Caspase Activation in Spermatids
(A) An illustration of a single spermatid during the
process of individualization. Different spermatidal
components (i.e., organelles and proteins) are
color coded and indicated at the bottom left.
Note that the tail end-to-sperm head descending
gradient of the CRL3’s pseudosubstrate inhibitor,
Soti (red), also reflects the dBruce protein gradient,
whereas the complementary gradient (sperm
head-to-tail end) of activated caspases (green)
also represents the active CRL3 complex gradient.
The inset provides an enlarged illustration of the
small white rectangular region, depicting the
axoneme (purple) attached to the two mitochon-
drial derivatives (sky blue), the Soti protein (red
speckles), as well as the caspase activation source
(green speckles) at the vicinity of the mitochondria.
(B and C) Illustrations of the CRL3 complex
OFF/ON states, respectively.
See also Figure S4.
In contrast, not only was transgenic G-Sb less abundant in the
mitochondria and more abundant in the cytosol compared with
transgenic A-SbT, it was also almost insensitive to PK treatment
(Figure 6E). Importantly, after PK treatment, the transgenic A-SbTmitochondrial levels were similar to the transgenic G-Sb mito-
chondrial levels with and without treatment, suggesting that
both proteins accumulate to similar levels in the inner mitochon-
drial compartment (i.e., the matrix), while A-SbT, but not G-Sb,
has the ability to also accumulate at the mitochondrial surface
(Figure 6E). Finally, note that the level of the PK-sensitive trans-
genic A-SbT portion is about 45%, which is in high agreement
with the levels of the mitochondrial surface portion of A-SbT re-
vealed by the immuno-EM (compare Figures 6E and 5E). Taken
together, these results indicate a crucial requirement of mito-
chondrial surface localization of A-SbT in spermatids, which is
likely embodied in the N-terminal domain of this protein.
DISCUSSION
An Integrated Model for Restricted Caspase Activationin SpermatidsIn this work, we uncovered a mode of CRL3 complex regulation
during the caspase-dependent process of spermatid individual-
ization in Drosophila (Figure 7A). According to this model, A-Sb
binds to and activates a fraction of the CRL3 complex at the sur-
face of the (two) spermatids’ mitochondrial derivatives, thereby
restricting activation of this complex and the consequent activa-
tion of caspases to the vicinity of the mitochondria (see inset in
Figure 7A). When overexpressed in the Drosophila eye, A-Sb
enhanced the CRL3 complex-induced small eye phenotype,
suggesting that the photoreceptor cells may be more sensitive
than spermatids to apoptosis induction through CRL3 com-
plex-mediated caspase activation. This may be due to the natu-
ral anatomical differences between these two cell types, such as
length (2-mm-long spermatids versus photoreceptor cells that
reach to the length of 80 mm) and/or molecular differences,
such as differences in the expression levels of caspase inhibitory
proteins. Quantification of the levels of the CRL3 complex in the
mitochondrial fraction suggests that approximately 40% of this
28 Developmental Cell 37, 15–33, April 4, 2016
complex is associated with the mitochondria at any given time,
implying that this mechanism may limit the rate of caspase
activation in spermatids at least by 60%. Given our current find-
ings that the stoichiometry of the CRL3 complex components is
important for its activation, and the notion that CRL complexes
form dimers andmultimers in vivo (Chew et al., 2007; Wimuttisuk
and Singer, 2007; Tang et al., 2007), aswell as the finding that the
level of a caspase activator in spermatids, Cyt-c-d, is restricted
by this complex (L.R., L.A., and E.A., unpublished data), we
believe that the efficiency of this mechanism in limiting the
source of caspase activation is probably even higher.
Previous work from our lab demonstrated that activation of
the CRL3 complex and consequent caspase activation is also
restricted in time (Kaplan et al., 2010). This is primarily mediated
by a tail end-to-sperm head gradient of Soti, a pseudosubstrate
inhibitor of the CRL3 complex, which competes with true sub-
strates for binding to the complex (red in Figure 7A). Importantly,
Soti and A-Sb display an antagonistic interplay, in which binding
of A-Sb to the CRL3 complex leads to the detachment of Soti
from Klhl10, possibly as a result of conformational change in
the CRL3 complex, thus allowing binding of true substrates,
such as the IAP, dBruce, to the CRL3 complex (Figures 7B and
7C). Upon ubiquitination by the complex, dBruce is redistributed
in a tail end-to-sperm head descending gradient, which leads to
activation of caspases in an opposite graded fashion (green in
Figure 7A). The fact that individualization is a dynamic process
in which activated caspases are extruded together with the cyto-
plasmic contents in a sperm head-to-tail end direction indicates
that spermatidal regions that are the first to individualize (low
Soti/A-Sb ratio) encounter the highest levels of caspase activity
but for the shortest time, whereas the regions that are the last to
individualize (high Soti/A-Sb ratio) experience the lowest levels of
caspase activity but for the longest time. Ultimately, this spatio-
temporal restriction mechanism ensures that each spermatidal
region encounters relatively low levels of caspase activity insuf-
ficient to trigger apoptosis but sufficient to facilitate proper
removal of the cytoplasmic contents. This fine balance of cas-
pase activity levels is crucial for male fertility, as either too low
levels (CRL3 inactivation) or too high levels (Soti and dBruce
inactivation) of caspase activity can lead to individualization
failure, unwanted cell death in the latter case, and male sterility
(Arama et al., 2003, 2006, 2007; Kaplan et al., 2010).
It is noteworthy that several observations in different experi-
mental systems support the notion that caspase activity may
be restricted in space and/or time during vital non-apoptotic
cellular processes, albeit the underlying mechanisms remain
largely vague (Feinstein-Rotkopf andArama, 2009). For example,
transient caspase activation has been observed during differen-
tiation of erythrocytes (Droin et al., 2008), lens fiber cells (Bass-
nett, 2009), and PC12 cells (Rohn et al., 2004) and is required
for LTD and AMPA receptor internalization in hippocampal
neurons (Li et al., 2010), while compartmentalized expression of
active caspases has been noted during platelet formation in
megakaryocytes (De Botton et al., 2002) and for active cas-
pase-8 during T cell activation (Koenig et al., 2008). Likewise, in
addition to restriction in space and time, compartmentalized acti-
vation of caspases was also reported in the Drosophila sperm
individualization system. In particular, it was shown that the
caspase activator, Hid, and the apoptosome components, Ark
and Dronc, as well as the apoptosome activator, Tango7, are
all localized in the vicinity of the advancing IC (Huh et al., 2004;
D’Brot et al., 2013). This may indicate another layer of caspase
regulation in thisprocess, inwhich spermatids utilize lowcaspase
activity throughout the cell during the individualization process
through the CRL3 complex mechanism, whereas in order to
activation factors but in a compartmentalized manner (i.e., Hid
and Tango7). Interestingly, Cullin-3 was also shown to further
increase caspase activation in mammalian cells by promoting
polyubiquitination of caspase-8, which leads to its aggregation,
processing, and full protease activation (Jin et al., 2009). Taken
together, this growingbodyof evidence suggests that harnessing
caspase activity for vital cellular processes may be much more
abundant than has been previously appreciated, and that cells
can tolerate low to moderate levels of caspase activity, as long
as this activity is below the threshold required for cell-death in-
duction (Florentin and Arama, 2012).
The Role of the Mitochondria for Caspase Activation inSpermatids and a Possible Link to Cell MetabolismBeing at the core of the intrinsic apoptotic pathway, mitochon-
dria have emerged as central regulators of apoptosis, providing
a reservoir for pro-apoptotic protein factors, such as cyto-
chrome c (Wang, 2001; Soriano and Scorrano, 2010; Martinou
and Youle, 2011), which is also required for caspase activation
during spermatid individualization (Arama et al., 2003, 2006;
Huh et al., 2004). However, the current study uncovers another
caspase regulatory role of the spermatid mitochondria in limiting
the rate of caspase activation by compartmentalizing the source
of activation to its surface. Interestingly, protein sequestration to
mitochondrial surfaces was also suggested for the C. elegans
Bcl-2-like protein, CED-9, and the Apaf-1-like cell-death acti-
vator, CED-4. In living cells, full activation of the caspase
(CED-3) through binding to CED-4 is prevented via the seques-
tration of CED-4 to themitochondrial surface by CED-9, whereas
following apoptosis induction, CED-4 assumes a perinuclear
localization and local activation (Chen et al., 2000; Pourkarimi
et al., 2012).
It is intriguing to hypothesize that the use of a bona fide mito-
chondrial metabolic enzyme subunit such as A-Sb to promote
key molecular activities (e.g., CRL3 complex activation), may
link between the metabolic state of the cells and the induction
of fundamental cellular processes (e.g., spermatid terminal
differentiation). According to this model, A-Sb may function as
a sensor of the cell metabolic state, allowing progression to
the terminal differentiation stage only when the spermatids fulfill
the metabolic requirements. How this sensing feature may work
requires a more comprehensive study of additional metabolic
enzymes, such as other components of the Krebs cycle and
the respiratory chain. It is interesting to note that a mitochondrial
surface tethering of another CRL3 complex substrate adaptor
protein, Keap1, and its substrate, the transcription factor Nrf2,
has been also reported in human cells. Tethering of Keap1 is
mediated by binding to phosphoglycerate mutase 5 (PGAM5),
which belongs to a family of metabolic enzymes usually involved
in glucose homeostasis, and is thought to facilitate coordination
between mitochondrial function and regulation of Nrf2-depen-
dent anti-oxidant gene expression by allowing proper ubiquiti-
nation and degradation of Nrf2 (Lo and Hannink, 2008). On the
other hand, it is of high interest to note that regulation of the
CRL3 complex is not the only non-traditional function suggested
for A-Sb inDrosophila. Previous reports proposed two additional
distinct roles of A-Sb outside of the mitochondria, albeit the
exact mechanisms of these actions remained vague. In cultured
cells, A-Sb was shown to bind to and modulate the activity of a
voltage-gated KCNQ potassium channel on the cell plasma
membrane (Gao et al., 2008). Furthermore, as mentioned in the
Results section, A-Sb was also found to bind to Skp1 on centro-
somal microtubules and regulate centrosome duplication during
mitosis in embryos (Hughes et al., 2008). Significantly, this may
suggest that regulation of CRL complexes may be a more gen-
eral function of A-Sb, as Skp1 and BTB domains, which assume
similar 3D folds (Stogios et al., 2005), are found in hundreds
of different CRL complexes. Therefore, A-Sb may represent a
multifunctional protein, which could mediate several different
cellular processes in eukaryotic cells, perhaps providing a link-
age between these processes and cellular metabolism.
The Moonlighting Function of A-SbMoonlighting proteins are special multifunctional proteins that
performmultiple autonomous, often unrelated, functions without
partitioning these functions into different protein domains (Hu-
berts and van der Klei, 2010). This group contains enzymes, re-
ceptors, transmembrane channels, chaperones, and ribosomal
proteins that have multiple functions. Some of these proteins
are mitochondrial enzymes that have mitochondrial roles and
additional unrelated roles, such as the electron transport chain
protein and the apoptosome assembly inducer, cytochrome c
(Jeffery, 2003; Copley, 2003).
Our findings suggest that A-Sb also constitutes amoonlighting
protein, functioning both in the Krebs cycle and as an activator of
the CRL3 complex. At this point, it is still unclear how A-Sb alter-
nates between the two mitochondrial compartments (i.e., the
matrix and outer membrane surface), as both the somatic and
rabbit anti-GAPDH (IMGENEX, 1:500), rabbit anti-Nedd8 ((Chan et al., 2008);
1:5000), rabbit anti-Sec16 ((Ivan et al., 2008), 1:2000), mouse anti-Ub (P4D1, Santa
Cruz, 1:500). All secondary antibodies were used in a dilution of 1:10000 (Jackson
Immuno-Research).
Immunogold labeling and electron microscopy analysis
Dissected testes were fixed in 4% paraformaldehyde with 0.1% glutaraldehyde in 0.1
M cacodylate buffer (pH�=�7.4) for 1 hour at room temperature, and kept at 4°C
ON. The samples were soaked ON in 2.3 M sucrose and rapidly frozen in liquid
nitrogen. Frozen ultrathin (70-90 nm) sections were cut with a diamond knife at -
120°C on a Leica EM UC6 ultramicrotome. The sections were collected on 200-mesh
Formvar coated nickel grids and blocked with a solution containing 1% BSA, 0.1%
glycine, 0.1% gelatin, and 1% Tween 20. Immuno-labeling was performed ON at 4°C
using affinity purified anti-GFP antibody (1:50, ab6556, Abcam), followed by washes
in PBS with 0.1% glycine, and exposure to goat anti-Rabbit IgG coupled to 10-nm
gold particles (1∶20, Jackson Immuno-Research) for 30 min at RT. Contrast staining
and embedding were performed as previously described in (Tokuyasu, 1986). The
embedded sections were viewed and photographed with a FEI Tecnai SPIRIT (FEI,
Eidhoven, Netherlands) transmission electron microscope operated at 120 kV, and
equipped with an EAGLE CCD Camera. These studies were conducted at the Irving
and Cherna Moskowitz Center for Nano and Bio-Nano Imaging at the Weizmann
Institute of Science.
Generation of the anti-Klhl10 antibody
A DNA fragment encoding the C-terminal portion of the Klhl10 protein (W373-Stop)
was PCR amplified from BDGP’s EST clone AT19737 (forward primer
CCCGAATTCTGGGTAACCATTAACGCAGAGG, reverse primer
GCGCGGCCGCCTATGTACGACGACGAATTTC, with added EcoRI and NotI
sites, respectively), and cloned into the corresponding sites of the pGEX-4T-1 vector,
in frame to the N-terminal GST tag. This C-terminal portion of Klhl10, which spans
most of the Kelch domain, is completely absent in the klhl103 mutant due to a
premature stop codon (Arama et al., 2007). The expression of the GST-Klhl10
(W373-stop) polypeptide was induced by IPTG induction for 1 hour in BL21(DE3).
To raise polyclonal antibodies against this polypeptide, the protein was purified under
native conditions using Glutathione-agarose (Sigma) and injected into rats.
Gel mobility retardation assay
270 testes were dissected out from young adult flies and chemically cross-linked in 1
% Formaldehyde in PBS for 10 minutes in RT. The cross-linker was quenched by the
addition of 1.25 mM Glycine for 5 minutes in RT. Samples were homogenized and
centrifuged at 13,000 rpm for 15 minutes at 4oC. The pellet was suspended in SDS
buffer (60 mM Tris [pH 6.8], 2 % SDS) and incubated for 3 minutes at 65oC. 950 μl
of NP-40 buffer (10 mM HEPES [pH 7.4], 1 mM MgCl2, 100 mM NaCl, 1% Igepal
CA-360, and protease inhibitor mix) were added to the solution and the samples were
shaken for 2 hours at 4oC. Sample buffer was added and the samples were boiled for 8
minutes and loaded on an SDS-PAGE gradient gel.
Supplemental References
Arama,E., Bader,M., Rieckhof,G.E., and Steller,H. (2007). A Ubiquitin Ligase Complex Regulates Caspase Activation During Sperm Differentiation in Drosophila. PLoS Biol 5, e251.
Bazinet,C. and Rollins,J.E. (2003). Rickettsia-like mitochondrial motility in Drosophila spermiogenesis. Evol. Dev. 5, 379-385.
Chan,Y., Yoon,J., Wu,J.T., Kim,H.J., Pan,K.T., Yim,J., and Chien,C.T. (2008). DEN1 deneddylates non-cullin proteins in vivo. J. Cell Sci. 121, 3218-3223.
Gao,L., Fei,H., Connors,N.C., Zhang,J., and Levitan,I.B. (2008). Drosophila ortholog of succinyl-CoA synthetase {beta} subunit: a novel modulator of Drosophila KCNQ channels. J Neurophysiol. 99, 2736-2740.
Hirose,F., Ohshima,N., Shiraki,M., Inoue,Y.H., Taguchi,O., Nishi,Y., Matsukage,A., and Yamaguchi,M. (2001). Ectopic expression of DREF induces DNA synthesis, apoptosis, and unusual morphogenesis in the Drosophila eye imaginal disc: possible interaction with Polycomb and trithorax group proteins. Mol. Cell Biol. 21, 7231-7242.
Hughes,J.R., Meireles,A.M., Fisher,K.H., Garcia,A., Antrobus,P.R., Wainman,A., Zitzmann,N., Deane,C., Ohkura,H., and Wakefield,J.G. (2008). A microtubule interactome: complexes with roles in cell cycle and mitosis. PLoS. Biol 6, e98.
Ivan,V., de,V.G., Xanthakis,D., Spoorendonk,K.M., Kondylis,V., and Rabouille,C. (2008). Drosophila Sec16 mediates the biogenesis of tER sites upstream of Sar1 through an arginine-rich motif. Mol. Biol. Cell. 19, 4352-4365.
Kaplan,Y., Gibbs-Bar,L., Kalifa,Y., Feinstein-Rotkopf,Y., and Arama,E. (2010). Gradients of a Ubiquitin E3 Ligase Inhibitor and a Caspase Inhibitor Determine Differentiation or Death in Spermatids. Dev. Cell 19, 160-173.
Morin,X., Daneman,R., Zavortink,M., and Chia,W. (2001). A protein trap strategy to detect GFP-tagged proteins expressed from their endogenous loci in Drosophila. Proc. Natl. Acad. Sci. U. S. A. 98, 15050-15055.
Reveal,B., Yan,N., Snee,M.J., Pai,C.I., Gim,Y., and Macdonald,P.M. (2010). BREs mediate both repression and activation of oskar mRNA translation and act in trans. Dev. Cell 18, 496-502.
Tokuyasu,K.T. (1986). Application of cryoultramicrotomy to immunocytochemistry. J. Microsc. 143, 139-149.
Venken,K.J., Carlson,J.W., Schulze,K.L., Pan,H., He,Y., Spokony,R., Wan,K.H., Koriabine,M., de Jong,P.J., White,K.P., Bellen,H.J., and Hoskins,R.A. (2009). Versatile P[acman] BAC libraries for transgenesis studies in Drosophila melanogaster. Nat. Methods 6, 431-434.
Venken,K.J., He,Y., Hoskins,R.A., and Bellen,H.J. (2006). P[acman]: a BAC transgenic platform for targeted insertion of large DNA fragments in D. melanogaster. Science 314, 1747-1751.