Genome-wide identification of mRNAs associated
with the translational regulator PUMILIO
in Drosophila melanogaster
Andre ´ P. Gerber*†‡, Stefan Luschnig†§, Mark A. Krasnow†¶, Patrick O. Brown†‡¶, and Daniel Herschlag†‡
*Institute of Pharmaceutical Sciences, Department of Chemistry and Applied Biosciences, ETH Zurich, 8093 Zurich, Switzerland;†Department of Biochemistry
and¶Howard Hughes Medical Institute, Stanford University School of Medicine, Stanford, CA 94305; and§Department of Genetics, University of Bayreuth,
95440 Bayreuth, Germany
Edited by Christine Guthrie, University of California, San Francisco, CA, and approved February 1, 2006 (received for review October 25, 2005)
Genome-wide identification of RNAs associated with RNA-binding
proteins is crucial for deciphering posttranscriptional regulatory
systems. PUMILIO is a member of the evolutionary conserved
Puf-family of RNA-binding proteins that repress gene expression
posttranscriptionally. We generated transgenic flies expressing
affinity-tagged PUMILIO under the control of an ovary-specific
promoter, and we purified PUMILIO from whole adult flies and
embryos and analyzed associated mRNAs by using DNA microar-
rays. Distinct sets comprising hundreds of mRNAs were associated
with PUMILIO at the two developmental stages. Many of these
mRNAs encode functionally related proteins, supporting a model
for coordinated regulation of posttranscriptional modules by
specific RNA-binding proteins. We identified a characteristic se-
quence motif in the 3?-untranslated regions of mRNAs associated
with PUMILIO, and the sufficiency of this motif for interaction
with PUMILIO was confirmed by RNA pull-down experiments with
biotinylated synthetic RNAs. The RNA motif strikingly resembles
the one previously identified for Puf3p, one of five Saccharomyces
cerevisiae Puf proteins; however, proteins encoded by the associ-
ated mRNAs in yeast and Drosophila do not appear to be related.
The results suggest extensive posttranscriptional regulation by
PUMILIO and uncover evolutionary features of this conserved
family of RNA-binding proteins.
DNA microarray ? posttranscriptional gene regulation ? RNA-binding
often mediated by specific RNA-binding proteins (RBPs) that
bind to elements in the UTRs of mRNAs and regulate the
stability, translation, or localization of the mRNA (1–3).
Whereas many classical studies explored the cellular role of
RBPs with specific mRNA substrates, the recent development
of genome-wide analysis tools enables systematic identification
of mRNA substrates of RBPs and the study of posttranscrip-
tional gene regulation on a global scale (4).
We recently used DNA microarrays to systematically identify
RNAs associated with each of the five Pumilio-Fem 3-binding
protein bound to a large set of mRNAs encoding functionally
and cytotopically related proteins, and characteristic sequence
elements in the 3? UTRs of the target mRNAs were identified
(5). These and other results suggest that there is extensive
coordinate regulation of RNAs by RNA-binding proteins
(4, 6, 7).
Puf proteins are an evolutionary conserved family of RBPs
that are implicated in diverse physiological processes in eu-
karyotes (8, 9). They are defined by the presence of a structurally
conserved RNA-binding domain, termed the Pum-homology
domain (PumHD), composed of eight repeats of 36 aa (10–12).
Puf proteins have been reported to bind to 3? UTR sequences
encompassing a ‘‘UGUR’’ tetranucleotide motif in the RNA,
osttranscriptional regulation of gene expression plays impor-
tant roles in diverse cellular processes. This regulation is
and in concert with other proteins repress gene expression by
affecting mRNA translation or stability (8).
PUMILIO (Pum), the founding Puf-family member and sole
member in the fruit fly Drosophila melanogaster, is implicated in
diverse developmental processes. Nevertheless, only a few
mRNA substrates are known to date. Pum was first identified as
a maternal-effect gene required for posterior patterning in the
embryo (13). Pum, Nanos (Nos), and Brain tumor (Brat) repress
translation of maternally derived hunchback (hb) mRNA in the
posterior of the embryo. This regulation depends on two bipar-
tite nanos-response elements (NREs), each composed of box A
(GUUGU) and box B (AUUGUA) sequences, located in the 3?
UTR of hb mRNA (11, 14, 15). The current model is that Pum
binds to sequences in the NRE and recruits Nos and Brat,
forming a quaternary RNA–protein complex that causes dead-
enylation of hb mRNA and translational repression (15). This
inhibition of expression of the transcription factor Hb in the
posterior of the embryo is essential for formation of the abdo-
men. Pum mutants also show abnormal temporal expression of
bicoid (bcd), the main determinant in the anterior patterning
Pum and Nos have also been implicated in the regulation of
germ cell development and maintenance. During embryogene-
sis, Pum and Nos accumulate in pole cells, the germ-line
progenitor cells. They are required for proper pole-cell migra-
tion to the gonad and are thought to coordinate cell division in
migrating pole cells by repressing translation of maternally
derived cyclin B (cycB) mRNA (17). During oogenesis, Pum
regulates asymmetric cell division of germ-line stem cells
Several recent studies have implicated Pum in neuronal de-
velopment and function. In a genetic screen, several pum alleles
were found to affect long-term memory formation in adult flies
(21). Pum apparently effects morphogenesis of larval peripheral
sensory neurons into dendritic cells (22) and negatively regulates
expression of the cap-binding protein eIF4E at the larval neu-
romuscular junction (23). Notably, the putative regulatory se-
quences in the 3? UTRs of eIF4E and cycB mRNAs lack box A
motifs and hence only partially match NREs present in hb and
bcd, suggesting alternative strategies for regulation of these
Conflict of interest statement: No conflicts declared.
This paper was submitted directly (Track II) to the PNAS office.
Freely available online through the PNAS open access option.
Abbreviations: RBP, RNA-binding protein; PumHD, Pum-homology domain; NRE, nanos-
response element; FDR, false discovery rate; GO, gene ontology; TAP, tandem affinity
(GEO) database, www.ncbi.nlm.nih.gov?geo (accession nos. GSE3580–GSE3582).
‡To whom correspondence may be addressed. E-mail: firstname.lastname@example.org,
email@example.com, or firstname.lastname@example.org.
© 2006 by The National Academy of Sciences of the USA
March 21, 2006 ?
vol. 103 ?
no. 12 ?
mRNAs, perhaps involving combinatorial binding with other
Using DNA microarrays to identify mRNAs associated with
Pum in Drosophila adult ovaries and embryos, we found hun-
dreds of mRNAs that likely represent mRNA targets subject to
regulation by Pum. The sets of mRNAs associated with Pum
differ in adults and embryos and code for subsets of functionally
related proteins. Most of the adult mRNA targets contain a
conserved RNA sequence motif in their 3? UTRs; this motif is
sufficient for interaction with Pum. These results suggest exten-
sive posttranscriptional coordination by Pum in Drosophila,
providing further evidence for a highly organized and multifac-
eted posttranscriptional regulatory system in multicellular
Systematic Identification of RNAs Associated with PUMILIO. To iden-
tify RNAs interacting with Pum in Drosophila, we generated
transgenic flies expressing a tandem-affinity purification (TAP)-
tagged fragment of Pum (24). A TAP-tagged full-length Pum
construct could not be obtained because expression constructs
were toxic to Escherichia coli. Therefore, we used the C-terminal
part of Pum which includes the RNA-binding region (PumHD)
and an additional 24 C-terminal amino acids that likely mediate
interaction with NANOS and BRAT (12, 25). Moreover, this
fragment has been shown to be sufficient for partial rescue of
pum mutant phenotypes (26). TAP-PumHD was expressed in
Drosophila by using the UAS-GAL4 system, which allows tissue-
specific expression of transgenes (27, 28). We used maternal
?-tubulin promoter-regulated GAL4 to drive TAP-PumHD
expression in the female ovary. Maternally synthesized protein is
loaded into the egg, as is endogenous Pum protein (29) (see Fig.
5, which is published as supporting information on the PNAS
web site). Moreover, we estimated the intracellular concentra-
tion of TAP-PumHD in embryos and found it similar to that of
endogenous Pum (?40 nM) (30).
We identified RNAs associated with Pum expressed in ovaries
of adult flies and from embryos by applying a method similar to
one established in Saccharomyces cerevisiae that uses the TAP-
tag to purify RNA-protein complexes (5, 31). We prepared
cell-free extracts from adults and embryos and recovered tagged
PumHD (?5 nM in the extracts) by affinity selection on IgG
was isolated from the eluate after TEV protease cleavage, and
?250 ng of RNA per gram of transgenic adult flies and 1 ?g of
RNA per gram of embryos were obtained. To control for
nonspecifically enriched RNAs, the same experiments were
performed with wild-type flies and embryos lacking a tagged
protein, yielding ?40 ng of RNA per gram of adult flies, and 700
ng of RNA per gram of embryos. Total RNA from extracts and
affinity-purified RNAs were then used to prepare cDNA probes
labeled with different fluorophores, which were comparatively
hybridized to a Drosophila DNA microarray. The ratio of
hybridization signals from the two RNA populations at a given
array element, representing a specific gene, reflects the enrich-
ment of the respective RNA by the Pum affinity purification (5).
To generate a list of mRNAs that were consistently enriched
by Pum affinity purification and, hence, likely targets of Pum, we
compared association of transcripts from the TAP-PumHD
affinity isolations to the mock isolates by two-class Significance
Analysis of Microarrays (SAM) and determined false discovery
rates (FDRs) for each mRNA (see Supporting Text, which is
published as supporting information on the PNAS web site). In
flies, 714 genes were consistently associated with Pum with a
FDR of ?0.1%, and 1,090 genes with a FDR of ?1%. Among
these 1,090 genes are the four previously identified Pum target
mRNAs (hb, bcd, cycB, eIF4E) (Fig. 1A, for a complete list, see
Table 2, which is published as supporting information on the
PNAS web site). The data from embryos yielded no genes with
a FDR of ?0.1% and only 192 with a FDR of ?1%. Although
the smaller number of genes identified at these levels of speci-
ficity may indicate that fewer mRNAs associate with Pum in
embryos, the high amounts of unspecifically bound RNAs evi-
dent in mock control isolates may have hampered our analysis
and hence true mRNA targets may have been missed. Therefore,
we directly compared total RNA obtained from TAP-Pum
affinity isolations from embryos and mock isolates by compar-
ative hybridization to microarrays. To define a list of potential
Pum mRNA targets in embryos, we combined both types of
analyses and selected 165 genes that had a FDR of ?3.2% and
that were at least 3-fold enriched in TAP-Pum affinity purifi-
cations compared to mock control experiments (Fig. 1B, for a
complete list, see Table 3, which is published as supporting
information on the PNAS web site).
Transcripts for only 31 genes were shared between the 714
(4.3%) adult and the 165 (18.8%) embryo Pum targets. Thus, the
vast majority were detected only in adults or in embryos,
two developmental stages (Fig. 1C); this could be due to
different patterns of expression of the target genes at the
different developmental stages, to different Pum protein com-
plexes that lead to altered RNA-binding specificities, or to
different accessibility of the mRNA targets in different tissues.
To estimate the bias from differential gene expression, we
compared mRNA levels in adults and embryos using DNA
microarrays, and then compared this to our lists of Pum mRNA
targets in adults and embryos (Fig. 6, which is published as
supporting information on the PNAS web site). Many mRNAs
are present at both developmental stages, and the mRNAs
selected with Pum from adults were not biased for genes
preferentially expressed during this stage. In contrast, embryo-
selected Pum targets were generally more highly expressed in
and embryos. Rows represent genes (unique DNA elements) and columns
represent individual experiments. The color code indicates the degree of
enrichment (green–red ratio scale). (A) Relative enrichments of mRNA targets
proteins and three mock experiments are shown. Genes were ordered from
top to bottom according to increasing FDRs determined by significance anal-
ysis of microarrays (SAM) analysis. Arrowheads depict enrichment of previ-
ously known Pum targets. (B) Pum RNA targets selected in embryos. A total of
174 transcripts representing 165 genes were clustered based on Euclidean
Specific RNAs selectively associated with Pum in Drosophila adults
www.pnas.org?cgi?doi?10.1073?pnas.0509260103 Gerber et al.
embryos (Fig. 6). This trend could, at least in part, result from
the higher RNA background in the embryo experiments.
Few mRNAs Have Altered Steady-State mRNA Levels in pum Mutant
Flies. We compared mRNA levels of pum13mutant and wild-type
adult female flies with DNA microarrays. Of the ?12,500
transcripts for which data were obtained from two independent
analyses, 48 (130) transcripts were at least 3-fold (2-fold) more
abundant, and 59 (243) transcripts were at least 3-fold (2-fold)
less abundant in pum mutant compared to wild-type flies.
Notably, 13 of the 48 genes present at elevated levels in pum13
mutants (27%) encode proteins involved in the antibacterial?
fungal immune response (P ? 10?17); among them, five (AttB,
AttC, AttD, CecC, CG13422) of the 20 mRNAs encoding
antibacterial peptides (P ? 10?8).
Expression levels of the ‘‘adult’’ Pum mRNA targets were
slightly increased in Pum mutants (median, 1.1 fold; t test
statistics, P ? 10?17; see Fig. 7, which is published as supporting
information on the PNAS web site). This finding suggests the
possibility that Pum might promote accelerated degradation of
mRNA targets as shown for the homologous yeast Puf3 protein
regulating Cox17 mRNA stability (5, 32, 33).
Functionally Related Sets of mRNAs Are Associated with Pum. To
identify functional themes among the mRNAs associated with
Pum, we searched for shared Gene Ontology (GO) annotations
in the lists of Pum target mRNAs (Table 4, which is published
as supporting information on the PNAS web site). The mRNAs
associated with Pum expressed in adult ovaries could be grouped
in two major classes. One contains mRNAs encoding nuclear
proteins involved in nucleotide metabolism and transcriptional
regulation. This set includes two of the previously known targets,
hb and bcd mRNAs. CycB mRNA has also been shown to be
regulated by Pum, and 7 of the 10 cyclin mRNAs (A, B, D, G,
J, K, T) were in this set (P ? 10?6; cyclins C, E, and H were
analyzed but not enriched). The second group contains tran-
scripts encoding proteins localized to organelle membranes. For
example, it includes mRNAs encoding most of the subunits of
the vacuolar proton-translocating V-type ATPase (Fig. 2A). This
complex consists of two domains: Pum associated with mRNAs
encoding six (Vha13, Vha26, Vha44, Vha55, Vha68–1?
Vha68–2, VhaSFD) of the eight subunits comprising the cata-
lytic V1 domain or the head group (P ? 5 ? 10?6) and four
(Vha16, VhaAC39, VhaM9.7–1?7–2, VhaPPA1–1) of the five
subunits of the V0 domain (P ? 2 ? 10?4), which forms the
transmembrane channel (34).
In embryos, some PumHD-bound mRNAs encode proteins
involved in germ cell development or anterior–posterior axis
patterning. Genes required for germ cell development included
four (nos, vas, orb, out) of the eight genes with GO annotations
for germ-line cyst formation (P ? 2 ? 10?6). Eleven of 16
mRNAs (P ? 10?8) that encode proteins acting in the cascade
that mediate anterior-posterior axis patterning were associated
with Pum in adults and embryos (Fig. 2B). These include bcd
mRNA, which encodes the anterior morphogen, the mRNA of
mRNAs, key components of the posterior patterning system. In
addition, we found mRNAs of regulators of this process: bruno-2
and bicaudal-C, which repress oskar mRNA translation, smaug,
which represses translation of nos mRNA, and vasa mRNA,
which encodes an ATP-dependent RNA helicase that promotes
oskar and nanos mRNA translation. The striking identification
of mRNAs that encode many of the proteins required for
posterior patterning suggests a general role for Pum in the
coordination of this process, in addition to its previously iden-
tified role in translational regulation of hb.
Pum also associated with mRNAs encoding regulators of
dorso–ventral axis formation, namely gastrulation defective (gd),
spaetzle, and possibly easter. These genes are also linked to the
antimicrobial immune response in Drosophila, which also in-
volves the Toll signaling pathway. No array data were obtained
for snake, the fourth component of a protease cascade leading to
Toll activation (35).
A Common Sequence Motif in the 3? UTR of Pum mRNA Targets.
Characteristic sequence motifs have been found in the 3? UTRs
Therefore, we examined the sets of mRNAs that associate with
Pum for the presence of common motifs using multiple expec-
tation maximization for motif elicitation (MEME) as a motif
discovery tool (38). The 3? UTR sequence information could be
retrieved (FlyBase) for 113 of the 150 genes most highly enriched
in Pum affinity isolates from adult flies, and MEME analysis
revealed a 16-nt consensus sequence (Fig. 3; see also Table 5,
which is published as supporting information on the PNAS web
site). The consensus includes a highly conserved 8-nt core motif
UGUA(A?U?C)AUA that contains the UGUA tetranucleotide
found in previously identified mRNAs that interact with Puf
proteins (5, 26, 30, 36, 37, 39). The core is flanked by less
conserved AU-rich sequences that could determine the recruit-
ment of other transacting factors or modulate binding affinities,
as shown recently for Fbf-1, a Puf family member from Caeno-
rhabditis elegans (36).
The UGUA tetranucleotide motif was found in 80% of the 3?
UTRs in our list of adult and embryo Pum mRNA targets, a
of the available 3? UTR sequences from D. melanogaster (Table
1). Enrichment of the 8-nt core motif UGUA(A?U?C)AUA was
more striking. This motif was present in 54% of the adult (P ?
10?106) and in 22% of the embryo (P ? 0.02) Pum target 3?
UTRs, compared to its genome-wide occurrence in 3? UTRs
(15%) (Table 1; for a list of all Drosophila 3? UTRs containing
subunits whose mRNAs associated with Pum; gray, subunits whose mRNAs were not enriched with a FDR of ?1%. Subunits of the V1 domain are labeled with
capital letters (34): A, Vha68; B, Vha55; C, Vha44; D, Vha36; E, Vha26; F, Vha14; G, Vha13; H, VhaSFD. Subunits of the V0 domain are indicated by small letters:
a, Vha100; c, Vha16; c?, PPA1; e, VhaM9.7; d, VhaAC39. (B) Components of anterior–posterior pattering systems associated with Pum in embryos and?or adults.
mRNAs associated with Pum are shown in red and their protein products in blue. Proteins whose mRNAs were not found to be associated are in black.
mRNAs associated with Pum encode proteins of specific macromolecular complexes and regulatory pathways. (A) Subunits of the vacuolar ATPase. Red,
Gerber et al.
March 21, 2006 ?
vol. 103 ?
no. 12 ?
this motif, see Table 6, which is published as supporting infor-
mation on the PNAS web site). The reason for the relatively rare
occurrence of the 8-nt motif among the embryo Pum targets is
not known. False positives from Pum affinity purifications,
additional elements in the RNA, or other transacting proteins
that bind RNA may have contributed to the measured RNA
To test putative mRNA targets for interaction with Pum and
to evaluate our predicted RNA recognition motif, we performed
RNA pull-down experiments using synthetic biotinylated tran-
scripts added to Drosophila extracts expressing TAP-PumHD.
We made biotinylated 3? UTR sequences of five Pum targets
(Vha16, caudal, CG1031, CG8414, plutonium) and one negative
control (Rps26) that was not found to be associated with Pum.
All of the five potential target mRNAs bound Pum protein,
whereas the Rps26 control 3? UTR sequence did not (Fig. 4A).
To further map sequences required for Pum-RNA interactions,
we mutated the highly conserved UGU trinucleotide within the
core consensus motif of the 3? UTR of CG8414 to ACA. Assays
with the mutated biotinylated RNA showed no specific interac-
tion with Pum (Fig. 4B). Likewise, addition of a 10-nt competitor
RNA comprising the consensus sequence prevented interaction
of biotinylated Vha16 3? UTR RNA with Pum, but no such
competition was seen with a control RNA in which the con-
served core UGU was mutated to ACA (Fig. 4B). The same
results were obtained in assays performed with extracts prepared
from embryos expressing TAP-PumHD (data not shown). Thus,
RNA sequences encompassing the computationally inferred
core motif are sufficient for association with PumHD in Dro-
Binding Sequence Motif, but Not Gene Function, Conserved Between
Yeast and Drosophila Puf mRNA Targets. The Pum core motif is
strikingly similar to the previously determined RNA recognition
sequence for Puf3 and less similar to sequences recognized by
the Puf4 and Puf5 proteins from S. cerevisiae (Fig. 3) (5). This
finding correlates with amino acid sequence conservation within
the respective PumHD domains of the orthologous proteins:
Puf3p is 46%, Puf4p and Puf5p are 30%, and Puf1 and Puf2
proteins are 20% identical to Drosophila Pum (9). Furthermore,
the three amino acid residues in each of the eight PumHD
repeats that make direct contact with one RNA base (12) are
conserved between Pum and Puf3 but differ at three positions
in the Puf4 and Puf5 protein sequences (for a multiple sequence
alignment of PumHDs from Puf-family members and respective
RNA recognition sequences see Fig. 8, which is published as
supporting information on the PNAS web site). The differences
in these PumHD residues may contribute to the altered RNA
recognition specificities of Puf4 and Puf5 compared to Pum and
Puf3. Further investigation will be required to define precise
rules for RNA recognition by this protein family.
We wondered whether the apparent conservation of protein
structure and RNA recognition sequences extended to the
functional level, i.e., whether the set of proteins encoded by the
between Drosophila and S. cerevisiae. Twenty-eight percent of all
Drosophila proteins match a homologous S. cerevisiae protein
sequence (40). A similar fraction (29%) of proteins encoded by
the Pum mRNA targets in embryos have a yeast homolog (Table
7, which is published as supporting information on the PNAS
web site; for protein BLAST results, see Tables 8 and 9, which are
published as supporting information on the PNAS web site).
However, a significantly greater fraction of proteins encoded by
the adult Pum mRNA targets, 39% (P ? 2 ? 10?8), have a yeast
homolog, which may indicate a role of Pum in regulating
yeast Puf3, Puf4 and Puf5 proteins (5). Height of the letter indicates the
probability of appearing at the position in the motif. Nucleotides with ?10%
appearance were omitted.
RNA consensus motif in 3? UTR sequences associated with Pum and
Table 1. Sequence motifs enriched in Pum targets
Consensus motifSearch option3? UTRs available3? UTRs with motif (%)
Adult (0.1% FDR)
Adult (0.1% FDR)
between biotin-labeled 3? UTRs and extracts of adult Drosophila expressing
the presence of TAP-PumHD by immunoblot analysis with Peroxidase-Anti-
Peroxidase Soluble Complex (Sigma). (A) Biotin-labeled 3? UTR sequences for
indicated genes (lanes 2 to 8) were incubated with Drosophila extract (input,
lane 1). Rps26 3? UTR was used as a negative control probe RNA (lane 8). (B)
Validation of the core sequence motif identified in 3? UTRs sequences of Pum
target mRNAs. Lane 1, input (Drosophila extract); lanes 2–4, biotin-labeled
RNA corresponding to the Vha16 3? UTR was combined with Drosophila
extract (lane 2) and 100-fold excess of competitor RNA (R1; AUUGUAAAUA;
lane 3) or control RNA where the core motif is mutated (R2; AUACAAAAUA;
lane 4). The conserved trinucleotide motif UGU in the 3? UTR of CG8414 was
control probe RNA (lane 5).
Validation of Pum mRNA targets. RNA–protein complexes formed
www.pnas.org?cgi?doi?10.1073?pnas.0509260103Gerber et al.
evolutionarily conserved cellular processes. We also determined
the fraction of the yeast homologs to the Pum targets whose
mRNAs associated with at least one of the five yeast Puf
proteins. Only 15% were targets for a yeast Puf protein (5),
suggesting that the Drosophila Pum targets are not particularly
conserved in yeast.
These differences are further underscored by comparison of
the targets of yeast Puf3 and Drosophila Pum, the most highly
conserved family members. Yeast Puf3p binds almost exclusively
to transcripts of nuclear genes that encode mitochondrial pro-
targets encode mitochondrial proteins. Nevertheless, a large
fraction of those targets (23 mRNAs) contain in their 3? UTRs
our computationally defined 8-nt Pum core motif; this motif is
present in only 31 (8%) of the 400 Drosophila nuclear genes
encoding mitochondrial proteins (for a list of the compiled 400
mitochondrial proteins, see Table 10, which is published as
supporting information on the PNAS web site). Thus, Puf3p and
Pum both appear to interact with nuclear-encoded mRNAs
encoding mitochondrial proteins. Whereas Puf3p interacts pre-
dominantly with these mRNAs (87% of targets), Pum interacts
with only a small fraction of this group (4%). Conversely, most
of the Pum targets are mRNAs encoding for nonmitochondrial
proteins. It is possible that the need for greater complexity in
gene regulation and coordination in more complex, multicellular
organisms has been addressed, in part, by breaking up function-
ally related mRNAs into smaller subgroups. In addition, the
ability to use combinatorial binding of RBPs would allow
coordination of and differentiation between these subgroups.
We have systematically analyzed mRNAs associated with Pum,
an important posttranscriptional regulator of gene expression in
Drosophila by a method that combines RNA copurification with
an affinity-tagged RNA-binding protein and DNA microarray
analysis of the associated RNAs (5, 31, 41–46). We identified
?1,000 distinct Pum-associated mRNAs, many of which encode
functionally related proteins and contain characteristic 3? UTR
sequence elements sufficient for interaction with Pum. This
finding represents a tremendous increase in potential mRNA
targets that may be subject to translational or other posttran-
scriptional regulation by Pum, and highlights the potential
importance of posttranscriptional regulation in multicellular
organisms (8, 9). The roles of Pum in embryonic development,
stem cell biology, and the function of the nervous system were
discovered by classical forward-genetic approaches; although
these approaches uncovered essential functions of Pum and
identified several important target mRNAs, our genomic anal-
ysis points to many hitherto unrecognized targets mRNAs whose
products may be involved in processes less readily accessed by
classical genetic approaches. However, our assay is unlikely to
exclusively and completely uncover target mRNAs that are
associated with Pum in vivo: nontarget mRNAs may associate
with Pum and true mRNA targets may dissociate during the
affinity-isolation procedure (47). Moreover, our assay does not
reveal whether Pum interacts directly with its target mRNA or
indirectly via another protein. Further biochemical and func-
tional experiments are required to verify and dissect regulation
of particular mRNAs by Pum. Nevertheless, the identification of
mRNA targets strongly suggest an underlying biological role for
many of the interactions we have identified.
Not only was the number of mRNAs associated with Pum
remarkably large, but the protein products of these mRNAs
shared functional links, including function in the anterior–
posterior patterning system, most cyclins, and most subunits of
the vacuolar H-ATPase. These functional links add to the
growing evidence for an extensive posttranscriptional regulatory
system and support recent models for functionally related post-
transcriptional modules organized by RBPs (5–7). Perhaps Pum,
in concert with other proteins, coordinates the temporal or
spatial pattern of translation of a large set of mRNAs. For
instance, Pum may help ensure that maternally derived mRNAs,
which are stored in the unfertilized egg, are translated at the
correct developmental stage. Translational repression of mater-
nally derived mRNAs before fertilization is an important mech-
anism to control the onset of expression of anterior–posterior
patterning genes (48). Another role of Pum is to regulate mitosis
of migrating pole cells by inhibition of cycB expression (17). Our
finding that most cyclin mRNAs were associated with Pum
introduces the possibility of a more general role for Pum in the
coordination of the cell-cycle, although cyclins can also have
non-cell-cycle-related functions (49). Regulation of cyclins by
a proposed ancestral function of Puf-family proteins (8).
In this work, as in the previous work on yeast Puf proteins (5),
a consensus RNA-binding element was defined by a genome-
wide unbiased search for common sequence motifs among
mRNAs selected by a biochemical procedure (Fig. 3). The 8-nt
core motif [UGUA(A?C?U)AUA] defined here is remarkably
similar to the sequences in and surrounding box B of the hb
NRE, and it resembles core motifs bound by diverse Puf family
proteins (5, 26, 30, 36, 39). It is also in agreement with a crystal
structure of human PumHD in complex with hb RNA, which
revealed that each of the eight repeats comprising the PumHD
interacts with one of eight bases in the bound RNA and
suggested that RNA recognition is highly modular (12). Inter-
interact with one RNA base are conserved between Pum and
yeast Puf3p (Fig. 8, which is published as supporting information
on the PNAS web site) paralleling the almost identical core
sequences bound by these proteins (Fig. 3). Other Puf proteins
that bear the same critical amino acid residues for RNA base
contacts may bind to highly similar RNA consensus sequences.
The definition of core motifs allows us to search for additional
potential mRNA targets that were not identified in our affinity
isolation procedure. About 10% of all genes in Drosophila
contain our computationally defined 8-nt core motif in their 3?
UTR; a search for GO terms overrepresented among these 1431
annotated genes found that an unexpectedly large fraction
encode proteins involved in morphogenesis or organ develop-
ment (243 genes, P ? 10?29), neurogenesis (153 genes, P ?
10?27), transcriptional regulation (172 genes, P ? 10?17), or
proteins that are localized to membranes (264 genes, P ? 5 ?
10?5), in particular the plasma membrane (116 genes, P ? 2 ?
10?12). Interestingly, many of the mRNAs that have the putative
Pum binding site but were not enriched in our assays encode
proteins with neuronal functions; e.g., Complexin (Cpx;
CG32490), which bears a cluster of 10 core motifs in its 3? UTR.
Because TAP-Pum was specifically expressed in the ovaries of
flies, neuron-specific mRNA targets would not have been ac-
cessible to TAP-Pum in vivo and therefore were not expected to
be identified. It will be important to extend this analysis to other
tissues and organs including the nervous system by the use of
tissue-specific drivers available in Drosophila. In addition to
identification of tissue-specific potential mRNA targets of Pum,
these experiments will also allow the determination of whether
and to what extent exchange of Pum-associated mRNAs occurs
after cell lysis.
Systematic identification of mRNAs associated with homolo-
gous RBPs in various species provides a basis for considering
their evolution. Large sets of target mRNAs can now be com-
pared with respect to their structural and functional common-
alties and differences. In the case of Pum, conservation of amino
acid residues in the PumHDs between homologous Puf proteins
correlates with our identified core motifs in 3? UTR of mRNA
Gerber et al.
March 21, 2006 ?
vol. 103 ?
no. 12 ?
targets. However, the proteins encoded by the mRNA targets
appeared not to be particularly conserved. This discordance
suggests that acquisition or loss of RBP-binding motifs in UTRs
of genes may provide a surprisingly fluid evolutionary mecha-
nism to modify posttranscriptional regulatory connections.
Materials and Methods
Oligonucleotide Sequences. See Supporting Text.
RNA Affinity Isolations. Wild-type (yellow white, y w) or mat-?-
tubGAL4-VP16(V67);UAS-TAP-PumHD(3–1-4) flies were
grown in large food cages (the generation of transgenic TAP-
Pum flies is described in Supporting Text). Adults (0–5 days old)
were collected, frozen in liquid nitrogen, and stored at ?80°C.
Embryos (0–16 h) were collected from apple juice agar plates,
dechorionated, and washed twice with buffer A (20 mM
Tris?HCl, pH 8.0?150 mM NaCl?10 mM EDTA, pH 8.0?0.2%
Nonidet P-40?0.02 mg/ml heparin) before freezing.
Five grams of adult flies or 2.5 g of embryos were used in each
affinity purification. Flies or embryos were suspended in 15 ml of
buffer B [buffer A plus 1.5 mM DTT?1 mM PMSF?0.5 ?g/ml
leupeptin?0.8 ?g/ml pepstatin?20 units/ml DNase I?100 units/ml
transferred to a glass-dounce, thawed, and dounced until the pestle
reached the bottom. The suspension was centrifuged twice at 4°C
and 10,000 ? g for 10 min. The fat layer on top was aspirated off
after each centrifugation. Cleared extract (12.5 ml) was incubated
with 600 ?l of a slurry (50% vol?vol) of IgG-agarose beads (Sigma)
for 90 min at 4°C. The beads were washed once with buffer B for
15 min at 4°C, and three times for 15 min at 4°C with buffer C (20
mM Tris?HCl, pH 8.0?150 mM NaCl?1 mM EDTA, pH 8.0?10%
glycerol?0.01% Nonidet P-40?1 mM DTT?10 units/ml RNasin).
PumHD was released from beads by incubation with 150 units of
AcTEV protease (Invitrogen) for 2 h at room temperature. RNA
was isolated from extracts and from the TEV eluates with TRIzol
reagent (Invitrogen) followed by RNeasy Mini kit (Qiagen) puri-
fication according to the manufacturer’s instructions.
Concentration of tagged transgene products in extracts was
determined by a filter affinity blot analysis using protein A as a
standard for calibration (5). Intracellular protein concentrations
in embryos were estimated with parameters described in ref. 30.
Microarray Analysis and Bioinformatics. Procedures for microarray
analysis, data selection, bioinformatics and gene expression
profiling are described in Supporting Text. Microarray data sets
are available at the Stanford Microarray Database (SMD) (50)
or at the Gene Expression Omnibus at www.ncbi.nlm.nih.gov?
geo (accession nos. GSE3580–GSE3582).
Synthesis of Biotinylated RNAs and Pull-Down Experiment. See Sup-
We thank Dr. Eric Johnson for help with motif searches and Dan Hogan
for discussions. We also thank Stanford Microarray Database and the
Functional Genomics Center Zu ¨rich for providing microarray related
infrastructure. A.P.G. and S.L. were supported by long-term fellowships
from the Human Frontier Science Program Organization. P.O.B. and
M.A.K. are investigators of the Howard Hughes Medical Institute. This
work was supported by the Howard Hughes Medical Institute and
National Institutes of Health Grant GM49243 (to D.H.).
1. Dreyfuss, G., Kim, V. N. & Kataoka, N. (2002) Nat. Rev. Mol. Cell. Biol. 3,
2. Gebauer, F. & Hentze, M. W. (2004) Nat. Rev. Mol. Cell. Biol. 5, 827–835.
3. Lopez de Heredia, M. & Jansen, R. P. (2004) Curr. Opin. Cell. Biol. 16, 80–85.
4. Hieronymus, H. & Silver, P. A. (2004) Genes Dev. 18, 2845–2860.
5. Gerber, A. P., Herschlag, D. & Brown, P. O. (2004) PLoS Biol. 2, E79.
6. Keene, J. D. & Tenenbaum, S. A. (2002) Mol. Cell 9, 1161–1167.
7. Keene, J. D. & Lager, P. J. (2005) Chromosome Res. 13, 327–337.
8. Wickens, M., Bernstein, D. S., Kimble, J. & Parker, R. (2002) Trends Genet. 18,
9. Spassov, D. S. & Jurecic, R. (2003) IUBMB Life 55, 359–366.
10. Zamore, P. D., Williamson, J. R. & Lehmann, R. (1997) RNA 3, 1421–1433.
11. Sonoda, J. & Wharton, R. P. (1999) Genes Dev. 13, 2704–2712.
13. Lehmann, R. & Nusslein-Volhard, A. (1987) Nature 329, 167–170.
14. Wharton, R. P. & Struhl, G. (1991) Cell 67, 955–967.
15. Sonoda, J. & Wharton, R. P. (2001) Genes Dev. 15, 762–773.
16. Gamberi, C., Peterson, D. S., He, L. & Gottlieb, E. (2002) Development
(Cambridge, U.K.) 129, 2699–2710.
17. Asaoka-Taguchi, M., Yamada, M., Nakamura, A., Hanyu, K. & Kobayashi, S.
(1999) Nat. Cell Biol. 1, 431–437.
18. Lin, H. & Spradling, A. C. (1997) Development (Cambridge, U.K.) 124,
19. Forbes, A. & Lehmann, R. (1998) Development (Cambridge, U.K.) 125,
20. Szakmary, A., Cox, D. N., Wang, Z. & Lin, H. (2005) Curr. Biol. 15, 171–178.
21. Dubnau, J., Chiang, A. S., Grady, L., Barditch, J., Gossweiler, S., McNeil, J.,
Smith, P., Buldoc, F., Scott, R., Certa, U., et al. (2003) Curr. Biol. 13, 286–296.
22. Ye, B., Petritsch, C., Clark, I. E., Gavis, E. R., Jan, L. Y. & Jan, Y. N. (2004)
Curr. Biol. 14, 314–321.
23. Menon, K. P., Sanyal, S., Habara, Y., Sanchez, R., Wharton, R. P., Ramaswami,
M. & Zinn, K. (2004) Neuron 44, 663–676.
24. Rigaut, G., Shevchenko, A., Rutz, B., Wilm, M., Mann, M. & Seraphin, B.
(1999) Nat. Biotechnol. 17, 1030–1032.
25. Edwards, T. A., Pyle, S. E., Wharton, R. P. & Aggarwal, A. K. (2001) Cell 105,
26. Wharton, R. P., Sonoda, J., Lee, T., Patterson, M. & Murata, Y. (1998) Mol.
Cell 1, 863–872.
27. Brand, A. H. & Perrimon, N. (1993) Development (Cambridge, U.K.) 118,
28. Rorth, P. (1998) Mech. Dev. 78, 113–118.
29. Macdonald, P. M. (1992) Development (Cambridge, U.K.) 114, 221–232.
30. Zamore, P. D., Bartel, D. P., Lehmann, R. & Williamson, J. R. (1999)
Biochemistry 38, 596–604.
31. Shepard, K. A., Gerber, A. P., Jambhekar, A., Takizawa, P. A., Brown, P. O.,
Herschlag, D., DeRisi, J. L. & Vale, R. D. (2003) Proc. Natl. Acad. Sci. USA
32. Olivas, W. & Parker, R. (2000) EMBO J. 19, 6602–6611.
33. Jackson, J. S., Jr., Houshmandi, S. S., Lopez Leban, F. & Olivas, W. M. (2004)
RNA 10, 1625–1636.
34. Allan, A. K., Du, J., Davies, S. A. & Dow, J. A. (2005) Physiol. Genomics 22,
35. Dissing, M., Giordano, H. & DeLotto, R. (2001) EMBO J. 20, 2387–2393.
36. Bernstein, D., Hook, B., Hajarnavis, A., Opperman, L. & Wickens, M. (2005)
RNA 11, 447–458.
37. Opperman, L., Hook, B., Defino, M., Bernstein, D. S. & Wickens, M. (2005)
Nat. Struct. Mol. Biol. 12, 945–951.
38. Bailey, T. L. & Elkan, C. (1994) Proc. Int. Conf. Intell. Syst. Mol. Biol. 2, 28–36.
39. White, E. K., Moore-Jarrett, T. & Ruley, H. E. (2001) RNA 7, 1855–1866.
40. Rubin, G. M., Yandell, M. D., Wortman, J. R., Gabor Miklos, G. L., Nelson,
C. R., Hariharan, I. K., Fortini, M. E., Li, P. W., Apweiler, R., Fleischmann,
W., et al. (2000) Science 287, 2204–2215.
41. Tenenbaum, S. A., Carson, C. C., Lager, P. J. & Keene, J. D. (2000) Proc. Natl.
Acad. Sci. USA 97, 14085–14090.
42. Takizawa, P. A., DeRisi, J. L., Wilhelm, J. E. & Vale, R. D. (2000) Science 290,
43. Roy, P. J., Stuart, J. M., Lund, J. & Kim, S. K. (2002) Nature 418, 975–979.
44. Hieronymus, H. & Silver, P. A. (2003) Nat. Genet. 33, 155–161.
45. Inada, M. & Guthrie, C. (2004) Proc. Natl. Acad. Sci. USA 101, 434–439.
46. Kim Guisbert, K., Duncan, K., Li, H. & Guthrie, C. (2005) RNA 11, 383–393.
47. Mili, S. & Steitz, J. A. (2004) RNA 10, 1692–1694.
in Translational Control of Gene Expression, eds. Sonenberg, N., Hershey,
J. W. B. & Mathews, M. B. (Cold Spring Harbor Lab. Press, Plainview, New
York), pp. 295–370.
49. Edgar, B. A. & Lehner, C. F. (1996) Science 274, 1646–1652.
50. Ball, C. A., Awad, I. A., Demeter, J., Gollub, J., Hebert, J. M., Hernandez-
Boussard, T., Jin, H., Matese, J. C., Nitzberg, M., Wymore, F., et al. (2005)
Nucleic Acids Res. 33, D580–D582.
www.pnas.org?cgi?doi?10.1073?pnas.0509260103Gerber et al.