mRNA Silencing in Erythroid Differentiation: hnRNP K and hnRNP E1 Regulate 15Lipoxygenase Translation from the 3′ End

Abstract

Although LOX mRNA accumulates early during differentiation, a differentiation control element in its 3′ untranslated region confers translational silencing until late stage erythropoiesis. We have purified two proteins from rabbit reticulocytes that specifically mediate LOX silencing and identified them as hnRNPs K and E1. Transfection of hnRNP K and hnRNP E1 into HeLa cells specifically silenced the translation of reporter mRNAs bearing a differentiation control element in their 3′ untranslated region. Silenced LOX mRNA in rabbit reticulocytes specifically coimmunoprecipitated with hnRNP K. In a reconstituted cell-free translation system, addition of recombinant hnRNP K and hnRNP E1 recapitulates this regulation via a specific inhibition of 80S ribosome assembly on LOX mRNA. Both proteins can control cap-dependent and internal ribosome entry site–mediated translation by binding to differentiation control elements. Our data suggest a specific cytoplasmic function for hnRNPs as translational regulatory proteins.

Full text

Cell, Vol. 89, 597–606, May 16, 1997, Copyright 1997 by Cell Press
mRNA Silencing in Erythroid Differentiation:
hnRNP K and hnRNP E1 Regulate
15-Lipoxygenase Translation from the 39 End
Dirk H. Ostareck,*
§
Antje Ostareck-Lederer,*
§
have been found to operate during early development
in C. elegans, Drosophila, Xenopus, and mammals (Hu-Matthias Wilm,
†
Bernd J. Thiele,
‡
Matthias Mann,
†
and Matthias W. Hentze* arte et al., 1992; Jackson, 1993; Wickens et al., 1993,
1996; Evans et al., 1994; Curtis et al., 1995). In spite*Gene Expression Programme and
†
Protein and Peptide Group of its wide occurrence, the mechanisms underlying 39
UTR–mediatedtranslationalregulationarepoorlyunder-European Molecular Biology Laboratory
Meyerhofstrasse 1 stood.
Erythroid 15-lipoxygenase (LOX) isa keyenzyme dur-D-69117 Heidelberg
Germany ing erythroid cell differentiation. It can attack intact
phospholipids and appears to participate in the break-
‡
Institute of Biochemistry
Humboldt University of Berlin down of internal membranes (such as mitochondrial
membranes) duringthe late stagesof reticulocytematu-D-10115 Berlin
ration (Rapoport and Schewe, 1986). LOX activity must
thus be temporally restricted.LOX mRNA is transcribed
as the most abundant message after globin mRNAs in
Summary
bone marrow erythroid precursor cells, but is transla-
tionally silenced until reticulocytes in the peripheral
Although LOX mRNA accumulates early during differ-
blood undergo thefinalsteps inmaturation (Thiele etal.,
entiation, a differentiation control element in its 39 un-
1981, Ho
¨
hneet al.,1988).ACU-rich, repetitivesequence
translated region confers translational silencing until
motif in the 39 UTR of LOX mRNA (Fleming et al., 1989;
late stage erythropoiesis. We have purified two pro-
Hunt, 1989), which we will refer to as the differentiation
teins from rabbit reticulocytes that specifically medi-
control element (DICE), has been shown to mediate this
ate LOX silencing and identified them as hnRNPs K
translational silencing (Ostareck-Lederer et al., 1994).
and E1. Transfection of hnRNP K and hnRNP E1 into
LOX silencing appears not to require alterations in poly(A)
HeLa cells specifically silenced the translation of re-
tail length, because it can be reconstituted by addition
porter mRNAs bearing a differentiation control ele-
of rabbitreticulocyteproteinseluted fromaDICEaffinity
ment in their 39 untranslated region. Silenced LOX
column to a nonpolyadenylated transcript in a cell-free
mRNA in rabbit reticulocytes specifically coimmuno-
translation system(Ostareck-Lederer et al.,1994). Here,
precipitated with hnRNPK. In a reconstitutedcell-free
we report the identification, cloning, and expression of
translation system, addition of recombinant hnRNP
two proteins that bind to the DICE, and demonstrate
K and hnRNP E1 recapitulates this regulation via a
the specificity and selectivity with which they control
specific inhibition of 80S ribosome assembly on LOX
the translation of LOX mRNA in vivo and in vitro. We
mRNA. Both proteins can control cap-dependent and
show that they are able to silence not only cap-depen-
internal ribosome entry site–mediated translation by
dent but also IRES (internal ribosomal entrysite)-driven
binding to differentiation control elements. Our data
translation via a mechanism that causes the specific
suggest a specific cytoplasmic function for hnRNPs
inhibition of 80S ribosome assembly.
as translational regulatory proteins.
Introduction
Results
Genetic control of cell metabolism and differentiation,
Identification of the LOX Regulatory
the cell cycle, and embryonic development involve
Proteins as hnRNPs K and E1
translational regulatory circuits. Defined regulatory ele-
To purify and clone proteins that bind to and regulate
ments located within the 59 or 39 untranslated regions
LOX mRNA, S100 extract prepared from reticulocytes
(UTRs) areutilized for translational regulationof specific
of anemicrabbitswassubjectedtoaffinitychromatogra-
mRNAs. For example, 59 UTR sequences control amino
phy on two different biotinylated RNAs: the DICE from
acid metabolism in Saccharomyces cerevisiae, where
the 39 UTR of rabbit LOX mRNA consisting of a 10-fold
upstream open reading frames (ORFs) and a regulated
repeat of a CU-rich 19 nt motif (10R), or nonregulatory
eIF2kinase operate atranslation reinitiation mechanism
LOX sequences downstream from the repeat (NR) as a
that controls 80S ribosome assembly onthe GCN4 ORF
specificity control. After extensive washing (Figure 1A,
(Dever et al., 1992; Hinnebusch, 1994). Iron metabolism
lanes 1 and 3, [w]), the retained proteins were eluted
in mammalian cells is regulated by a molecular mecha-
and analyzed by SDS–PAGE. In addition to some minor
nism that controls the binding of the 40S ribosomal
bands, bothcolumns retain thetwo prominent polypep-
subunit to the ferritin mRNAby an iron-sensitive regula-
tides labeled (n.s.) in Figure1. Two sharp bands migrat-
tory protein that recognizes a specific 59 UTR element
ing at z66 kDa and z43 kDa and a diffuse band of z48
(Gray and Hentze, 1994; Hentze and Ku
¨
hn, 1996). More
kDa specificallyelute from theDICEresin (comparelane
commonly, however, the 39 UTR harbors the regulatory
4 with lane2).In contrasttothe reproducible purification
sequences. Most examples of this type of regulation
of the proteins migrating at 66 kDa and 43 kDa, the
appearance ofthe48 kDaband was variable.The eluate
from the DICE resin, but not that from the NR control
§
These authors contributed equally to this work.

Cell
598
Figure 2. HnRNP E1andhnRNP K SpecificallyRegulate LOXmRNA
Translation in Rabbit Reticulocyte Lysate
(A and B) LOX-10R ([A], lanes 1–12) or LOX-NR mRNAs ([B], lanes
1–12) were cotranslated with CAT mRNA as internal control. In (A)
and (B), dialysis buffer (lanes 1 and 12) or protein fractions purified
from NTA or DICE columns were added as indicated, hnRNPs E1/K
in a 1:3 ratio. [
35
S]Met incorporation into translation products was
determined by phosphoimaging (Compaq Phosphor Imager with
Figure 1. Identification of the LOX Regulatory Proteins as hnRNPs
MolecularDynamicsImage Quant software),and thesignals in lanes
K and E1
1 and12weretaken as 100%standardsandaveraged. Thepercent-
Scheme of the LOX mRNA with its 39 UTR–regulatory (10R) and
age of LOX expression normalized for the CAT internal control is
nonregulatory region (NR).
shown below each lane.
(A) Silver-stained gelwith final wash ([w], lanes 1 and 3) and eluates
(C) A
32
P-labeled 10R probe was photocrosslinked to the regulatory
([e], lanes 2 and 4) after protein purification using NR (lanes 1 and
proteins or to 150 mg of rabbit reticulocyte lysate (S100) (lane 1) at
2) or 10R (lanes 3 and 4) affinity resins. Nonspecific peptides are
254 nm. Note that the N-terminal histidine tag adds z3 kDa to the
labeled (n.s.); three bands of 66kDa,48 kDa,and43 kDaspecifically
recombinant proteins.
elute from 10R.
(B) Nanoelectrospray mass spectrum of the 43 kDa band ([A], lane
4); see Experimental Procedures.
resin,represses LOXmRNA translation invitro (datanot
(C) Parent ion scan: detection of ions leading to Ile/Leu–containing
peptides.
shown, but see Figures 2A and 5–7), suggesting that
(D) A sequence tag of a peptide ion identified in (C) was assembled
the 66 kDa and/or the 43 kDa proteins may represent
in the high m/z part of the spectrum. Arrows indicate the sequence
the regulatory polypeptides.
for peptide sequence tag searching. An alignment of the complete
Both proteins were analyzed by nanoelectrospray
C-terminal ion series is shown.
mass spectrometry (Wilm et al., 1996a). As shown in
Figure 1B forthe 43 kDa polypeptide,the affinity prepa-
rations yielded proteinamounts that were not sufficient
for peptides to be distinguished in the mass spectrum.

LOX Regulation from the 39 UTR
599
Parent ionscansof unseparatedpeptidemixtures (Wilm recombinant protein thanDICE affinity-purified proteins
from rabbit reticulocytes (see Figure 2A; compare laneset al., 1996b) allowed peptide ions to be distinguished
fromthebackground(Figure1C).Peptidesequencetags 5–7 with lanes 2–4 and lane 11). Secondary purification
of the recombinant proteins on a DICE column revealedwere derived from the tandem mass spectra. The band
at 66 kDa produced 5 peptides, all of which identified thatonly z5%–10%oftheproteincould bind.Theeluate
from the DICE column displaysspecific repressor activ-the protein as hnRNP K (heterogeneous nuclear ribo-
nucleoprotein K; Matunis et al., 1992; acc. no. S74678). ity similar to the affinity-purified material from rabbit
reticulocytes (Figure 2A, lanes 8–11), which correlatesThe peptide sequence of the 43 kDa band (Figure 1D)
uniquely identifieda 13 amino acidpeptide from hnRNP with its DICE-binding activity (Figure 2C). While these
data donot excludea possible roleofadditional factors,E1 (Aasheim et al., 1994; Kiledjian et al., 1995; Leffers et
al., 1995; acc.no.X78137). Comparison ofthe predicted we conclude that recombinant hnRNPs K and E1 can
fully account for translational silencing of LOX mRNA infragment ion masses against the tandem mass spectrum
verified the complete sequence. The protein amounts a cell-free translation system from rabbit reticulocytes.
To investigate whether the complete CU-rich regionin this small scale isolation were in the low nanogram
range, as estimated from the intensity and number of in rabbit LOX mRNA is required for DICE function, the
10-fold repeat (LOX-10R) was replaced by one (LOX-measured peptide ions. The 48 kDa band yielded six
peptidesthatallmappedtotheC-terminalpartofhnRNP 1R), two (LOX-2R), and four (LOX-4R) copies. While the
translation of LOX-1R mRNA was only marginally af-K, suggesting that the variable appearanceof this band
results frompartial proteolysis ofhnRNP Kduring purifi- fected by recombinant hnRNPs K and E1 or the the
purified reticulocyte proteins,the doseresponse curvescation. Interestingly, a 48 kDa polypeptide was pre-
viously implicated as the LOX mRNA regulatory protein of LOX-10R and LOX-2R mRNAs were similar (data not
shown). We also constructed a negative control mRNALOX-BP (Ostareck-Lederer et al., 1994). In light of the
present data, we believe that LOX-BP corresponds to of identical length (LOX-2Rmut) in which a central 59
CCCUCUU 39 sequence was replaced by 59 AGAGAGAa 48 kDa fragment of hnRNP K.
39. This mutant is not silenced by hnRNPs K and E1 (or
the cellular proteins) in rabbit reticulocyte lysate (dataDICE-Mediated Translational Control by
hnRNP K and hnRNP E1 Is Highly not shown). Next, the role of hnRNPs K and E1 as 39
UTR–regulatory proteins was examined in transfectedSelective in Cell-Free Extracts
and Transfected Cells HeLa cells. Two luciferase reporter constructs, LUC-2R
and LUC-2Rmut, were cotransfected with hnRNP E1To test thefunction of hnRNPsK and E1as LOX regula-
tory proteins, cDNAs were isolated from a HeLa cell and/or hnRNP K expression vectors. A human growth
hormone (hGH) expression vector was cotransfectedaslibrary by PCR, sequenced, and expressed as N-termi-
nally histidine-tagged fusion proteins in E. coli. Photo- an internal specificity control, and an expression vector
for U1A (instead of hnRNPs K or E1) was cotransfectedcrosslinking of recombinant hnRNPs K and E1 demon-
strated their specific binding to the DICE (data not as a negative control RNA-binding protein. Luciferase
expression wasmonitoredby labelingwith [
35
S]-methio-shown). In addition to the DICE, poly(C) competed for
photocrosslinking ofhnRNPsK andE1tothe10Rprobe, nine and immunoprecipitation (Figure 3A) and by enzy-
matic assay (Figure 3B). While luciferase expressionwhich was expected from the known relative affinity of
both proteins for poly(C) (Swanson and Dreyfuss, 1988; from LUC-2R mRNA is repressed by hnRNP K and
hnRNP E1, but not by U1A (Figures 3A and 3B, lanesMatunis et al., 1992; Kiledjian et al., 1995).
Translation ofLOX mRNA in a rabbitreticulocyte cell- 1–6), neither the hGH control (Figure 3C), the endoge-
nous ferritin control (Figure 3D), nor the LUC-2Rmutfree translation system is specifically silenced by affin-
ity-purified DICE-binding proteins (Ostareck-Lederer et control (lanes 7–12) are affected by either of the two
regulatory proteins. Importantly, LUC mRNA levels re-al., 1994). To test whether hnRNP K and/or hnRNP E1
account for this activity, capped LOX mRNA was co- main unchanged when both total RNA (Figure 3E) or
cytoplasmic RNA (data not shown) are analyzed bytranslated with cappedCAT mRNA as aninternal speci-
ficity control (Figure 2A, lanes 1–12). In the absence Northern blotting.Similar results wereobtained intrans-
fection experiments with LUC-10R, using LUC-NR as aof added silencer proteins, both mRNAs are efficiently
translated (lanes 1 and 12), whereas the DICE eluate negative control (data not shown).
HnRNP E1 has recently been identified as a compo-(Figure 1A, lane 4) diminishes LOX translation without
significant effect on the translation of CAT (Figure 2A, nent of an erythroid ribonucleoprotein complex that
binds to a CU-richregion in the 39 UTR of human a-glo-lane 11). Importantly, recombinant hnRNP E1 (lane 2),
hnRNPK(lane3), orthetwocombined(lane 4)allsilence bin mRNA in vitro (a complex) (Kiledjian et al., 1995).
Therefore,weevaluatedthea-globin39 UTRin ourtrans-LOX synthesis. When the DICE is deleted from the 39
UTR, LOX synthesis is unaffected by hnRNPs K and E1 fection assay (Table 1).Whileboth hnRNPK andhnRNP
E1 silencethetranslationofLUC-2RmRNA,thea-globin(Figure 2B, lanes 1–12). Thus, silencing by the recombi-
nant proteins is mediated by the DICE. Extraction/re- sequence does not function as a DICE in spite of its
CU-rich nature. Neither of two negative control RNA-translation experimentsdemonstrated that translational
silencing is reversible and does notaffect the structural binding proteins (U1A and IRP-1) mimicks the effect of
hnRNPs Kand E1.Cotransfection of anexpression vec-or functionalintegrity of the mRNA (datanot shown;see
also Figure6). However,when the recombinantproteins tor for hnRNP A1 affects luciferase activity nonspecifi-
callyfrom LUC-2R,LUC-2Rmut,and LUC-a-globin,pos-were purified viatheir histidine tags byNTA chromatog-
raphy, silencing required approximately 10-fold more sibly due to a general disturbance of RNA metabolism

Cell
600
Figure 3. Translational Silencing by hnRNPs
K and E1 In Vivo
(A–D) HeLa cells were transiently transfected
with 5 mg of LUC reporterconstructs bearing
a 38 nt DICE [LUC-2R] (lanes 1–6) or a mu-
tated version as a negative control [LUC-
2Rmut] (lanes 7–12), and 5 mg of a human
growth hormone (hGH) internal specificity
control plasmid driven by the same promoter
as the LUC reporter constructs (lanes 1–12).
In addition, 10 mg of U1A (lanes 1 and 7), 5
mg (lanes 2 and 8) or 10 mg (lanes 3 and 9) of
hnRNP E1; 5 mg (lanes 4 and 10) or 10 mg
(lanes 5and11)of hnRNPK; or5mgof hnRNP
E1 plus 5 mg of hnRNP K (lanes 6 and 12)
expression plasmids were cotransfected.
Lane 13 shows an analysis of untransfected
cells. Parallel dishes were transfected with
the same precipitate, and used for metabolic
labeling for2 hours with [
35
S]-Metor forprep-
aration of total RNA and cell extract. Labeled
extracts were used for immunoprecipitations
with polyclonal antibodies against luciferase
(A), hGH (C), and ferritin (D). Unlabeled ex-
tracts (50 ng of protein) were used for enzy-
matic analysisof luciferase activity (B). Lucif-
erase activitiesin lanes 1 and7, respectively,
were set at 100%. (E) Northern blot of the
total RNA hybridized with a luciferase probe.
Equal signal intensities were also obtained
with a ferritin H chain probe (for lanes 1–13)
and an hGH probe (for lanes 1–12, data not
shown).
of the transfected cells. We conclude that both hnRNP in wheat germ extract and meets the same specificity
K and E1 regulate translation with high specificity for criteria,including the discrimination against theCU-rich
the DICE in vitro and in vivo. element from a-globin mRNA (Figure 5, lanes 15–20),
and the lack of silencer activity ofthe RNA-binding pro-
Association of LOX mRNA with hnRNP K
teins IRP-1 (lane 7) and hnRNP A1 (lane 6) against the
in Translationally Silent mRNPs
DICE. The lack of LOX suppression by recombinant
of Erythroid Cells
hnRNP A1 in vitro is consistent with our interpretation
To directly addressthe role of hnRNPs Kand E1 as LOX
of the data shown in Table 1. Note that recombinant
mRNA regulatory proteins in immature erythroid cells,
hnRNPsK andE1 purifiedbyDICE affinitychromatogra-
weisolatedtranslationally inactivemRNPsfromreticulo-
phy display specific silencing activities similar to the
cytes of anemic rabbits. To assess the association of
ones displayed by the DICE eluate from rabbit reticulo-
specific mRNAs with specific proteins in these mRNPs,
cyte lysate. These results show that no species- or tis-
hnRNP K–containing complexes were immunoprecipi-
sue-specific accessory factors are required to enable
tated with a polyclonal antibody (Figure 4, lanes 4 and
hnRNPsK andE1tocontrol translation.To delineatethe
5). As controls, preimmune serum (lanes 2 and 3) or
mechanism of translational control, translation initiation
buffer (lanes 6 and 7) was used. RNA was extracted
assayswere performed.Capped LOX-2RorLOX-2Rmut
from the precipitates (p) and the supernatants (s), and
examined by Northern blotting. Almost half of the LOX
Table 1. Functional Discrimination between CU-Rich 39 UTR
mRNA is precipitated by the hnRNP K antibody (ratio
Sequences from LOX and a-globin mRNAs
of supernatant:precipitate [s:p] 5 1.3), whereas barely
Effector Indicator
detectable quantities of LOX mRNA are precipitated
when preimmune serum (s:p 5 26.0)or buffer (s:p 5 38.7)
LUC-2R LUC-2Rm LUC-a-globin
areused. Bycontrast,the hnRNPKantisera(s:p5 76.0),
U1A 100 100 100
thepreimmuneserum(s:p 5 53.4),and thebuffercontrol
IRP 1 95 101 93
(s:p 5 64.2) fail to coprecipitate significant quantities of
E1 53 86 92
b-globinmRNA.Dueto thelackofanappropriateantise-
K408995
E1/K 18 92 93
rum, similar experiments with hnRNP E1 could not be
A1 15 14 12
performed. The specific association of LOX mRNA with
hnRNP K in translationally silenced mRNPs from rabbit
HeLa cells were transiently transfected as described in the legend
reticulocytessupports therole ofthisproteinintheregula-
to Figure 3. Luciferase activity is expressed as a percentage of the
tion of LOX expression during erythroid differentiation.
activity from samples cotransfected with 10 mg of U1A expres-
sion plasmid. Luciferase mRNA levels (measured by Northern blot-
ting) were similar in cells transfected with U1A, IRP-1, hnRNP K,
Mechanism of Translational Control
and hnRNP E1, but strongly diminished in cells transfected with
As observed in rabbit reticulocyte lysate and in HeLa
hnRNP A1.
cells, the regulatory mechanism is also fully operative

LOX Regulation from the 39 UTR
601
Figure 4. In Vivo Association of hnRNP K and LOX mRNA
mRNPs from rabbit reticulocytes were precipitated with hnRNP K
polyclonal antiserum (lanes 4 and 5),preimmune serum (lanes 2 and
3) or buffer (lanes 6 and 7). RNA was extracted from mRNPs (lane
1), orsupernatants (s) andpellets (p)after immunoprecipitation,and
analyzed by Northern blotting with
32
P-dCTP-labeled LOX and b-glo-
bin cDNAs. The signals were quantitated by phosphoimaging and
Figure 6. HnRNPs K and E1 Specifically Inhibit 80S Ribosome As-
expressed as supernatant-to-pellet ratios.
sembly
32
P-labeled RNAs were preincubated with dialysis buffer (open cir-
cles) or 500 ng of recombinant NTA-agarose purified hnRNP E1/K
(1:3 ratio) (plus signs). Translation initiation complexes were subse-
mRNAs were
32
P-labeled, and ribosomes were allowed
quently allowed to assemble on the labeled LOX-2R (left panel) or
to assembleduring a 5min incubation inrabbit reticulo-
LOX-2Rmut (right panel) mRNAs in cycloheximide-treated rabbit
cyte lysate in the presence or absence of hnRNPs
reticulocyte lysate, andresolved bycentrifugationin5%–25% linear
sucrose gradients. After fractionation (lower panels), the labeled
K/E1. The initiation complexes were resolved on 5%–
mRNAswere extractedandanalyzed byautoradiography afterform-
25%linearsucrosegradients andthen fractionated,and
aldehyde/agarose gel electrophoresis (even-numbered fractions),
the association of mRNAs with ribosomal subunits was
or the radioactivity counted in the odd-numbered fractions, ex-
assessedby scintillation countingof the odd-numbered
pressed as the percentage of total counts recovered, and plotted
fractions (Figure 6 graph) and by RNA extraction, gel
against the fraction numbers. The dashed line denotes the A
254
ab-
sorption profile, whichwasidentical for gradients withadded buffer
electrophoresis, and autoradiographyof the even-num-
or recombinant proteins.
bered samples (Figure 6 inset). In the absence of the
silencer proteins (open circles), both LOX-2R and LOX-
2Rmut mRNAs show a biphasic distribution between
encephalomyocarditisvirus(Kaminski etal.,1990).LOX-
ribosome-associated (fractions 1–7) and lighter fractions
10R mRNAwas cotranslatedas an internalpositivecon-
(fractions 10–14). While the hnRNPs K/E1 (indicated by
trol (Figure 7, upper band in all lanes) in rabbit reticulo-
[1]) do not affect ribosome assembly on LOX-2Rmut
cyte lysate together with CAT mRNA (Figure 7, lanes
mRNA (right panel), the association of 80S ribosomes
1–4), CAT-10R mRNA (lanes 5–8), IRES-CAT mRNA
with LOX-2R mRNA is clearly diminished and the tran-
(lanes 9–12),orIRES-CAT-10RmRNA (lanes13–16). Due
script shifted to the lighter fractions (left panel). This
to the construction strategy, both IRES-driven tran-
result demonstrates that the regulatory proteins inhibit
scripts encode an N-terminally extended CAT fusion
translationinitiation bypreventing80Sribosomeassem-
protein (compare lanes 9–16 with lanes 1–8). As ex-
bly, and confirms that the stability of LOX mRNA is not
pected, the translation of LOX-10R (lanes 2, 6, 10, and
affected.
14), CAT(lane 2), andCAT-10R (lane6) mRNAs is dimin-
To further analyze the regulatory mechanism, we ex-
ished by the addition of cap analog, whereas the cap-
amined mRNAs on which ribosomes assemble either in independenttranslation ofIRES-CAT(lane10)andIRES-
CAT-10R (lane 14) is unaffected. Likewise, recombinantthe cap-stimulated mode or mediated by the IRES from
Figure 5. HnRNPs K and E1 Silence LOX
mRNA Translation in Wheat Germ Extract
LOXmRNAs bearingeither tworepetitive ele-
ments(LOX-2R,lanes1–7),amutated version
of the two repeat motif (LOX-2Rmut, lanes
8–14), or the a-globin 39 UTR (LOX-a-globin,
lanes 15–20) were cotranslated with CAT
mRNA (lanes 1–20) as an internal control. Di-
alysis buffer (lanes 1, 8, and 15); 50 ng of
DICE column–purified recombinant proteins
hnRNP E1, hnRNP K, hnRNP E1/K (1:3 ratio);
DICEcolumn–purifiedrabbit reticulocytereg-
ulatory proteins; or recombinant proteins
hnRNP A1 and IRP-1 were added as indi-
cated.Thedata were quantitativelyevaluated
as described in the legend to Figure 2.

Cell
602
Figure 7. HnRNPs K and E1 Specifically Si-
lence Both Cap-Dependent and IRES-Medi-
ated Translation
LOX-10R mRNA as an internal positive con-
trol (lanes 1–16) was cotranslated with the
following CAT reporter mRNAs (schemati-
cally depicted at the top): CAT (lanes 1–4),
CAT-10R (lanes 5–8), IRES-CAT (lanes 9–12),
or IRES-CAT-10R (lanes 13–16). Dialysis
buffer(lanes1,5,9,and13),3.3mMm
7
GpppG
cap-analog (lanes 2, 6, 10, and 14), 500 ng of
NTA-agarose-purified hnRNPs E1/K (1:3 ra-
tio, lanes 3, 7, 11, and 15), or 50 ng of DICE-
purified rabbit reticulocyte regulatory pro-
teins (lanes 4, 8, 12, and 16) were added to
thetranslation reactions inrabbitreticulocyte
lysate. The positions of
35
S-methionine-
labeled LOX,CAT,andIRES-CATfusionpoly-
peptides are indicated. Translation of LOX
and CAT mRNAs was quantitated by phos-
phoimaging and expressed as percentages
of the signal in lanes without the addition of
regulatory proteins.
hnRNPs K and E1 as well as the purified regulatory Dreyfuss, 1993). Our findings appear to establish such
a cytoplasmic function.proteins from rabbit reticulocyte lysate repress LOX-
10R (lanes 3, 4, 7, 8, 11, 12, 15, and 16) and CAT-10R Searching of the EMBL database identified a protein
that is closely related to hnRNP E1: hnRNP E2 (Hahm(lanes 7 and 8) mRNA translation, without perturbing
CAT (lanes 3 and 4) or IRES-CAT (lanes 11 and 12) etal.,1993;Leffersetal.,1995;acc.noX78136).Wehave
also expressed this protein, and preliminary analysismRNA.Interestingly,theregulatory proteinsalsocontrol
the translation of IRES-CAT-10R mRNA (lanes 15 and suggests that it may regulate LOX translation similar to
hnRNPE1(datanotshown).Interestingly,hnRNPs K,E1,16), albeit less efficiently. Thus, hnRNPs K and E1 can
regulate cap-dependent and IRES-mediatedtranslation andE2have previously beendiscussedinthe contextof
gene regulation.The human c-myc gene contains a CT-initiation, at least in vitro.
rich promoter element that binds hnRNP K (Takimoto
et al., 1993) and mediates hnRNP K–induced transcrip-Discussion
tional activation (Michelotti et al., 1996). Furthermore,
hnRNPsE1 and E2 have been shown tobe componentsTranslational regulation mediatedby 39 UTRelements is
a widely usedmechanismofposttranscriptional control. of a ribonucleoprotein complex that binds to a CU-rich
region in the 39 UTR of human a-globin mRNA in vitroWe have cloned and expressed two proteins that bind
to the DICE located in the 39 UTR of LOX mRNA, and (a complex) (Kiledjian et al., 1995). The human a-globin
mutationConstant Spring is characterizedby areducedhave reconstituted regulationwith therecombinant pro-
teins in transfected cells and in vitro. We believe this to stability of the a-globin mRNA and lack of a complex
formation (Wang et al., 1995). The suggested involve-represent the first such functional reconstitution of 39
UTR–mediated translational control in eukaryotes. We ment of hnRNP E1 in regulating both a-globin mRNA
stabilityand LOXtranslationvia CU-rich 39 UTR–bindingalso provide an initial characterization of the underlying
regulatory mechanism. motifs prompted us to directly evaluate the a complex–
binding region for function as a translational regulatory
element.Remarkably, thetranslation ofreporter mRNAsA Cytoplasmic Function for hnRNP K
and hnRNP E1 bearing the CU-rich region of a-globin mRNA is not
affected by the hnRNPs E1 and K in transfected HeLaThe data in Figures 2–7 show that the hnRNPs K and
E1 function as specific cytoplasmic regulatory proteins cells (Table 1), cell-free translation extracts from wheat
germ (Figure5), orrabbit reticulocytes (datanot shown).of LOX mRNA translation. This finding appears to be
surprising, both in terms of the subcellular localization By contrast,the CU-richDICEconfers translational con-
trol underthe same conditions.Thus, while the hnRNPsand thebroad RNA-bindingspecificity thatis character-
istic of hnRNPs. However, previous work on the regula- K and E1 show relatively broad specificity for C(U)-rich
sequences in biochemical binding assays (Matunis ettion of alternative splicing in mammalian cells (Mayeda
and Krainer, 1992; Min et al., 1995) and the control of al., 1992; Kiledjian et al., 1995), they are exquisitely dis-
criminatory in their functional specificity.dorsoventral axis formation in Drosophila (Matunis et
al., 1994) have implicated hnRNP A1, hnRNP F, and
squid (hnRNPs A/B), respectively, in the control of spe-
cific genes. Recentstudies have revealed thata distinct Lipoxygenase Regulation during
Erythroid Differentiationsubset of hnRNPs, including hnRNP K, shuttle between
the nucleus and the cytoplasm, suggesting unexpected LOX mRNA is expressed but translationally silenced in
early erythroid cells in the bone marrowand the periph-functions of these shuttling hnRNPs in nucleocytoplas-
mic transport and/or in the cytoplasm (Pin
˜
ol-Roma and eral blood (Schewe et al., 1986). We have investigated

LOX Regulation from the 39 UTR
603
the association of LOX mRNA with hnRNP K in transla- (cap-dependent), but not the downstream (IRES-medi-
ated), ORF (Dubnau and Struhl, 1996). This result sug-tionally inactive mRNPs of peripheral blood reticulo-
gests that bicoid may interfere with a cap-dependent
cytes. The specific association of LOX mRNA and
step of translation initiation.
hnPNP K in these mRNPs (Figure 4) supports the role
By contrast, both cap-dependent and IRES-driven
of this protein as a translational silencer of LOX mRNA
translation are silenced by the LOX regulatory system
during erythroiddifferentiation, althoughit doesnot for-
(Figure 7), suggesting that hnRNPs K and E1 exert their
mally prove it.
regulatory action at a step downstream of the conver-
Ourfindingsraise thequestionofhowerythroid differ-
gence of the two translation initiation pathways, and
entiation signals alter the function of hnRNPs K and
excluding eIF4E as the molecular target for the LOX
E1 in LOX mRNA silencing. Several scenarios can be
silencing mechanism. Their ubiquitous occurrence and
envisaged. The regulatory proteins might be degraded
their ability to function in different cell types raises the
or sequestered; their RNA-binding activity reduced or
question of whether hnRNPsK andE1 arealso involved
their RNA-bindingspecificity might be changed; or their
in the translational regulation of other mRNAs.
ability to interfere with 80S ribosome assembly could
be switched off. The data shown in Figure 3 and Table
Experimental Procedures
1 indicate that the LOX-2R and LUC-2Rmut mRNAs are
translated equally well in HeLa cells that express only
Plasmids
their endogenous complement of hnRNPs K and E1.
The full-length rabbit reticulocyte 15-LOX cDNA (LOX-10R), the 241
Translational silencing in HeLa cells is only observed
nt fragmentwith the39 UTRLOXregulatoryregion(10R),thenonreg-
ulatory 335 nt downstream fragment (NR), and the CAT construct
after cotransfection of additional hnRNPs K and/or E1.
bearing the LOX regulatory region (CAT-10R) have been described
A relatedobservation applies tothe translationof DICE-
(Ostareck-Lederer et al., 1994). The LOX cDNA lacking the 241 nt
containing mRNAs in rabbit reticulocyte lysates (Figure
CU-rich fragment (LOX-NR) was generated by appending the NR
2). At present, we do not know the level of overexpres-
fragment 39 to the LOX ORF cloned into pBluescript II-SK (LOX).
sion or the subcellular localization of these proteins in
To generate LOX-2R and LOX-2Rmut,respectively, complementary
the transiently transfected cells. The additional copies
oligonucleotides representing two repeat elements of the 10R se-
quence, 59 AGCTTCCCCA CCCTCTTCCCCAAGCCCCA C CCTCTT
maytitrate acellular activitythat modulatesthe function
CCCCAAGAT39 (forLOX-2R)and59 AGCTTCCCCAAGAGAGACCC
of hnRNPs K and E1 as translational silencers. In this
CAAGCCCCAA GAGAGACCCC AAGAT 39 (for LOX-2Rmut, ex-
respect,it is alsointeresting thathnRNPK interacts with
changed nucleotidesare underlined; flanking sequencesforcloning
the SH3 domains of Src (Taylor and Shalloway, 1994;
purposes are inbold ) were cloned into the HindIII/ClaIsites of LOX.
Weng et al., 1994), Fyn, Lyn (Weng et al., 1994), and
LOX mRNA bearing the human a-globin 39 UTR was generated by
insertion of a fragment containing nucleotides 1–107 downstream
Vav (Hobert et al., 1994; Bustelo et al., 1995). Vav is a
of the a-globin termination codon (Wang et al., 1995; a kind gift of
hematopoietic cell–specific, signal-transducing protein
S. Liebhaber) into HindIII/ClaI of LOX. IRES-CAT was generated
that becomes tyrosine phosphorylated after engage-
from pEMC2.1 (a kind gift from M. Howell and R. J. Jackson) and
ment with several cell surface receptors. The hnRNP
contains the IRES of EMCV, followed by the CAT ORF. For the
K/Vav interaction is mediated by two proline-rich se-
construction of IRES-CAT-10R, the complete 39 UTR of rabbit 15-
quences in the central region ofhnRNP K. Interestingly,
LOX mRNA was inserted into IRES-CAT downstream of the CAT
ORF. ForLUCindicatorconstructs, theluciferase cDNAfrompGEM-
p95vav also interacts with a second z45 kDa poly(rC)-
LUC(Promega) was insertedintopSG5 (Greenetal., 1988) to gener-
binding protein, which may represent hnRNP E1 and/or
ate LUC. LUC-2R and LUC-2Rmut were made in analogy to LOX-
E2 (Bustelo et al., 1995), and phosphorylation appears
2R and LOX-2Rmut with synthetic oligonucleotides inserted 39 of
to negatively affect RNA binding by hnRNPs E1 and E2
the ORF of LUC. pSG5-LUC-a-globin was cloned from the a-globin
(Leffers et al., 1995). In addition, hnRNP K binds and
39 UTR plasmid (Wang et al., 1995) by PCR and insertion 39 to
the LUC ORF. pSG5-hnRNP E1 and -hnRNP K were generated by
is phosphorylated by an IL-1-responsive kinase (Van
insertion ofthe cDNAswithoutthehistidine tags . The hGH transfec-
Seuningen etal., 1995). The possible role ofthese inter-
tion control vector waspSG5-hGH.TRS3 (Pantopoulos and Hentze,
actions and of the phosphorylation of hnRNPs K and
1995). The hnRNP A1 cDNA was cloned by PCR from a pT7-7TT-
E1intranslationalcontrol during erythroiddifferentiation
hnRNPA1construct(Izaurraldeetal.,inpress)intopSG5.All plasmid
will warrant further investigation.
constructs were verified by DNA sequencing.
Affinity Purification of LOX Regulatory Proteins
Reticulocytosis in rabbits was induced by venesection on five con-
Mechanism of Translational Regulation
secutive days(10–12mlofblood/kgrabbit perday).Therabbits were
Binding of regulatory proteins to 39 UTR elements can
exsanguinated two days later by cardiac puncture. Reticulocyte
mediate translational activation or silencing. During oo-
extracts (S100) (Thiele et al., 1982) and the biotinylated RNA tran-
cyte maturationinXenopuslaevis, anelement upstream
scripts 10R and NR were prepared exactly as described (Ostareck-
Lederer et al., 1994). All subsequent procedures were performed at
from the polyadenylation signal acts as a cytoplasmic
48C. Aliquots (1 ml) of S100 extract were preincubated for 15 min
polyadenylationelement (CPE).CPEB, a62 kDa protein,
with 100 U RNAguard (Pharmacia)in the presence of 2% 2-mercap-
appearsto activatetranslationby stimulatingpoly(A) tail
toethanol, andthe supernatantwascollectedafter 1 mincentrifuga-
elongation (Hake and Richter, 1994). In Drosophila, the
tion at 3000 rpm in an Eppendorf centrifuge. Packed streptavidine-
anterior pattern determinant bicoid binds to the 39 UTR
agarose beads (300 ml; GIBCO-BRL) were washed four times with
purification buffer (150 mM KCl,1.5 mM MgCl
2
, 10 mM Tris–HCl [pH
of caudal mRNA and spatially restricts caudal protein
7.4], 0.5 mM DTT), and incubated with 10R or NR transcripts (z2
expression by silencing translation (Dubnau and Struhl,
mg) for1 hr. TheS100 extract was tumbled with thetranscript/beads
1996; Rivera-Pomaret al.,1996). Ina hybrid mRNAbear-
affinity resin for 100 min, followed by four washes (10 min each)
ing the antennapedia internal ribosome entry site in the
with 1.0 ml of purification buffer. Proteins bound tothe affinity resin
intercistronic region between two ORFs and the caudal
were collected in three consecutive elutions (10 min each) with 150
ml of buffer (3 M KCl, 1.5 mM MgCl
2
, 10 mM Tris–HCl [pH 7.4], 0.5
39 UTR, bicoid-suppressed translation of the upstream

Cell
604
mM DTT) and pooled, followedby dialysis against multiplechanges UV Cross-Linking Assays
UV cross-linking assays werecarried out exactly asdescribed (Ost-of 10 mM CH
3
COOK, 0.3 mM MgCl
2
, 10 mM Tris–HCl (pH 7.4),1 mM
DTT for 3 hr. Aliquots (10 ml) of the wash and elution fractions were areck-Lederer etal., 1994), usingasinputmaterial S100lysate(120–
150 mg of protein per ml) from rabbit reticulocytes or the indicatedanalyzed by SDS–PAGE and silver staining.
amounts of purified rabbit or recombinant human proteins.
Peptide Purification and Sequencing
Cell-Free Translations
Protein bands were excised from silver-stained gels and digested
The capped, nonpolyadenylated mRNAs were preincubated with
in-gel as described (Shevchenko et al., 1996; Wilm et al., 1996a).
the regulatory proteins or appropriate controls for 10 min on ice.
For nanoelectrospraytandemmassspectrometry (MS)analysis, the
Standard 12 ml reactions contained 8 ml of RRL (Jackson and Hunt,
peptide solution was desalted on a 50 nl microcartridge (Wilm and
1983) adjusted to 100 mM CH
3
COOK, 0.5 mM MgCl
2
, and 0.1 mM
Mann, 1996) of PorosR2 material(Perseptive Biosystems, Framing-
amino acids without methionine; 0.5 U of InhibitAce (5Prime–
ham, MA). A glass capillary with the unseparated peptide mixture
3Prime); 1.0 ml of capped mRNAs (10 ng of LOX [10R] or LOX [NR],
was mounted in the nanoelectrospray ion source for mass spectro-
1ngof CAT);and 0.5mCi/ml[
35
S]Met (Amersham,SJ1015). In Figure
metric analysis on a tandem mass spectrometer API III (PE-Sciex,
7, a 12 ml reaction contained 4.8 ml of RRL (Promega), and the final
Ontario, Canada) equipped with a nanoelectrospray ion source
concentration of the components addedwas 10 mM CH
3
COOK, 0.3
(Wilm andMann,1994,1996).Peptide massspectra(Q1 scans) were
mM MgCl
2
, and 0.06 mM amino acids without methionine; 0.5 U of
recorded with0.1 Dastep width andunit massresolution.Parent ion
InhibitAce; 10 ng of poly(C); 1 ml of capped mRNAs; and 0.7 ml
scans for the immoniumion ofisoleucine andleucine to specifically
of [
35
S] Met (Amersham, SJ1015, 10 mCi/ml). The addition of 10 ng
detect isoleucine/leucine–containing peptides were performedwith
of poly(C) was dispensable when 8 ml (instead of 4.8 ml) of RRL
0.2 Da step width and a lower resolution (2 Da peak width at 50%
was used. Where indicated, RRL was preincubated with 3.3 mM
height) to increase sensitivity (Wilm et al., 1996b). For the tandem
m
7
GpppG cap analog for 3 min at 308C. Translation reactions were
MS investigation, Q1 was set to transmit a 2 Da mass window. The
incubated at 308C for 1 hr and stopped by incubation with RNase
tandem MS spectrawere acquiredwith 0.2Da stepwidth. The mass
A (Jackson and Hunt, 1983). Translation in micrococcal nuclease–
resolution was set such that identifiable Y99 ions (Roepstorff and
treated wheat germ extract (WGE) (Promega) was performed ac-
Fohlman, 1984) could be assigned to better than 0.5 Da. Peptide
cording to the manufacturer’s protocol using 12.5 ng of capped
sequence tags (Mann and Wilm, 1994)were assembled using Apple
LOX-2R, LOX-2Rmut, or LOX-a-globin and 1 ng of capped CAT
Scripts, and a nonredundant database comprising more than
mRNAs.
200,000 sequences was searched. No restrictions on molecular
weight, pI (isoelectric point), or species of origin were applied. The
Translation Initiation Assays and Sucrose
retrieved sequence was verified by comparing it to the entire frag-
Gradient Analysis
ment spectrum using Bio-MultiView (PE-Sciex).
RRLwas preincubatedwith 0.5 mMcycloheximide for3 minat308C.
The followingpeptides wereidentifiedfromtwoindependentanal-
Labeled LOX-2R or LOX-2Rmut mRNAs (4 ng) wasincubated for 10
yses: five peptides for the 66 kDa protein, corresponding to hnRNP
min at 48C with recombinant NTA-agarose-purified hnRNPs E1 (125
K: IITITGTQDQIQNAQYLLQNAQYLLQNSVK, LLIHQSLAGGIIGVK,
ng)andK (375ng), orwithdialysisbuffer.Initiationreactions at308C
IILDLISESPIKGR, NLPLPPPPPPR, and ILSISADIETIGEI; six frag-
werestopped after 5 minby additionof ice-colddilution buffer, and
ments forthe 48kDaband,corresponding tothe C-terminalregionof
initiation complexes resolved on linear 5%–25% sucrose gradients
hnRNP K; IITITGTQDQIQNAQYLLQNSVK, GSYGDLGGPIITTQVTIPK,
(Gray and Hentze, 1994). Sixteen fractions (250 ml each) were col-
DLAGSIIGKGGQRIK, LLIHQSLAGGIIGVK, IILDLISESPIKGR, and
lected from thebottom of the gradientand were analyzed by scintil-
NLPLPPPPPPR; and one peptide that is uniquely found in hnRNP
lationcounting or RNAextractionand electrophoresis throughform-
E1: IITLTGPTNAIFK.
aldehyde agarose gels.
Expression of Recombinant Proteins
In Vivo Analyses
The cDNAs for hnRNP E1 and hnRNP K were PCR amplified from
HeLa cells were transiently transfected by the calcium phosphate
a HeLa cell cDNA library(gift of H. Stunnenberg). Gel-purifiedPCR-
method of Graham and van der Eb (1973) with 5 mg of LUC-2R,
products were cloned into pET16b (Novagene) for transformation
LUC-2Rmut, or LUC-a-globin, 5 mg of hGH control plasmid, plus 5
of E. coli BL21 (DE3). Transformed bacteria were grown in LB me-
or10mg ofpSG5-hnRNP E1 or pSG5-hnRNPK, aloneorin combina-
dium at 378C to a density of 0.6–0.8 OD
600
, pelleted, washed with
tion, 10 mg of pSG5 IRP-1 (Kollmus et al., 1996), or 10 mg of pSG5
minimal medium, and induced for 4 hr in minimal medium with 10
hnRNP A1. The quantity of expression plasmids for RNA-binding
mM isopropyl-1-thio-b-D-galactopyranosideat 238C.Thecells were
proteins was equalized by addition of pSG5-U1A (Stripecke et al.,
lysed (20 mM Tris–HCl [pH 8.0], 250 mM NaCl, 0.25% NP-40) by
1994)wherenecessary.Analysis bymetabolic labelingand immuno-
freeze thawing and subsequent sonication for 1 min with a Branson
precipitation was performed as described (Stripecke et al., 1994)
Sonifier B15 inthe presence of 0.1% PMSFand 10 mg/ml leupeptin.
using saturating quantities of polyclonal anti-hGH (Daco), anti-lucif-
After centrifugation for 10 min at 6000 rpm in a Sorvall SS34 rotor,
erase (a gift of H. Hauser), and anti-ferritin (Boehringer Mannheim)
the His-tagged proteins werepurified from the supernatanton Ni
2
1
-
antibodies.Luciferase assayswere carried out according to Brasier
NTA-Agarose beads (Quiagen). The eluate (200 mM imidazol for
et al. (1989). Total cellular RNA was isolated using the RNA-Clean
hnRNP E1, 500 mM imidazol for hnRNP K) wasdialyzed extensively
System (AGS), and Northern blots were probed with a
32
P-dCTP-
against the bufferused forthe affinity-purifiedproteins.For purifica-
labeled luciferase cDNA.
tion on a DICE column, the procedure described for rabbit reticulo-
Toanalyzethe associationof LOXmRNA withhnRNP Kinreticulo-
cyte lysate was followed. The final recovery was z10% (hnRNP E1)
cyte RNPs, S100 lysate was prepared and mRNPs pelleted after
and z2% (hnRNP K). Recombinant IRP-1 (Gray et al., 1993) and
removal of polysomes at 300000 3 g for 180 min. The RNPs were
hnRNP A1 (Izaurralde et al., in press) were prepared as described.
resuspended in TKM (50 mM Tris [pH 7.6], 50 mM KCl, 5 mM Mg
(CH
3
COO)
2
) to a protein concentration of 15 mg/ml. RNPs (100 ml)
were incubated with 1 ml of preimmune serum, hnRNP K polyclonalIn Vitro Transcription
Uncapped
32
P-labeled RNAs (specific activity 7 3 10
7
cpm/mg) or antibody (a kind gift of K. Bomsztyk), or water for 1.5 hr at 48C,
followed byan addition of50 ml ofprotein A–Sepharose(Pharmacia)unlabeled competitor RNAs for UV cross-linking assays were tran-
scribed with T3 RNA polymerase (Stratagene) and purified as de- for 1.5 hr. The beads were washed three times with buffer (10 mM
HEPES [pH 7.2], 3 mM MgCl
2
, 1 mM DTT), and the pellets resus-scribed (Ostareck-Lederer etal.,1994).CappedmRNAs(.95%cap-
ping efficiency) were generated as described (Gray et al., 1993); pended in 200ml of PK buffer (100mM Tris [pH 7.5], 12.5mM EDTA,
150 mM NaCl, 1% SDS) with 4 ml of proteinase K 20 mg/ml (Sigma)excess cap analog was removed after transcription and phenol/
chloroform extraction bytwo rounds of chromaspin-100 (Clontech). and 20 mg of glycogen (Boehringer Mannheim). The supernatant
(100ml)received 100mlof23PK buffer, proteinaseK, andglycogen.For translation initiation assays, capped
32
P-labeled mRNAs were
transcribed andpurified asabove (specific activity 5 3 10
7
cpm/mg). The sampleswere incubatedat508C for45 min,and pelletsremoved

LOX Regulation from the 39 UTR
605
bycentrifugation, extracted twicewithphenol/chloroform, andetha- iron metabolism: mRNA-based regulatory circuits operated by iron,
nitric oxide and oxidative stress. Proc. Natl. Acad. Sci. USA 93,nol precipitated. The recovered RNAs were analyzed by Northern
blotting. 8175–8182.
Hinnebusch, A.G. (1994). Translational control of GCN4: an in vivo
Acknowledgments
barometer of initiation-factor activity. Trends Biochem. Sci. 19,
409–414.
Correspondence regarding this paper should be addressed to
Hobert, O., Jallal, B., Schlessinger, J., and Ullrich, A. (1994). Novel
M. W. H.(Hentze@EMBL-Heidelberg.de).WethankFatimaGebauer,
signaling pathway suggested by SH3 domain–mediated p95vav/
Thomas Preiss, and Iain Mattaj for constructive comments on the
heterogeneous ribonucleoprotein K interaction. J. Biol. Chem. 269,
manuscript; Richard Jackson, Steve Liebhaber, and Iain Mattaj for
20225–20228.
plasmids; Nina Vagner for recombinant hnRNP A1 protein; Henk
Stunnenberg for the HeLa cell cDNA library; Hansjoerg Hauser for
Ho
¨
hne, M., Thiele, B.J., Prehn, S., Giessmann, E., Nack, B., and
the luciferase antibodies; and Karol Bomsztyk for the hnRNP K
Rapoport, S.M. (1988). Activation of translationally inactive lipoxy-
antibodies. A. O.-L. was supported by fellowships from Boehringer
genase mRNP particles. Biomed. Biochim. Acta 47, 75–78.
Ingelheim and EMBO. M. W. H. acknowledges support from the
Huarte, J., Stutz, A., O’Connell, M.L., Gubler, P., Belin, D., Darrow,
DeutscheForschungsgemeinschaft(He 1442/2-3) andtheEuropean
A., Strickland, S., and Vassalli, J.-D. (1992). Transient translational
Commission Biotechnology Program (BIO4-CT95-0045).
silencing by reversible mRNA deadenylation. Cell 69, 1021–1030.
Hunt, T. (1989). On the translational control of suicide in red cell
Received August 19, 1996; revised February 28, 1997.
development. Trends Biochem. Sci. 14, 393–394.
References
Izaurralde, E., Jarmolowski, A., Beisel, C., Mattaj, I.W., Dreyfuss, G.,
and Fischer, U. (1997). A role for the hnRNP A1 M9-transport signal
Aasheim, H.-C., Loukianova, T., Deggerdal, A., and Smeland, E.B.
in mRNA nuclear export. J. Cell Biol., in press.
(1994). Tissue specific expression and cDNA structure of a human
Jackson, R.J. (1993). Cytoplasmic regulation of mRNA function: the
transcript encoding a nucleic acid [oligo(dC)] protein related to the
importance of the 39 untranslated region. Cell 74, 9–14.
pre-mRNA binding protein K. Nucl. Acids Res. 22, 959–964.
Jackson, R.J., and Hunt, T.(1983). Preparationanduse of nuclease-
Brasier, A.R., Tate, J.E., and Habener, J.F. (1989). Optimized use of
treated rabbit reticulocyte lysate for the translation of eukaryotic
the firefly luciferase assay as a reporter gene in mammalian cell
messenger RNA. Meth. Enzymol. 96, 50–74.
lines. Biotechniques 7, 1116–1122.
Kaminski, A., Howell, M.T., and Jackson, R.J. (1990). Initiation of
Bustelo,X.R., Suen,K.L., Michael,W.M.,Dreyfuss,G., andBarbacid,
encephalomyocarditis virus RNA translation: the authentic initiation
M.(1995). Associationofvav proto-oncogene productwith poly(rC)-
site is not selected by a scanning mechanism. EMBO J. 9, 3753–
specific RNA-binding proteins. Mol. Cell. Biol. 15, 1324–1332.
3759.
Curtis, D.,Lehmann, R., andZamore,P.D. (1995).Translationalregu-
Kiledjian, M., Wang, X., and Liebhaber, S.A. (1995). Identification of
lation in development. Cell 81, 171–178.
two KH domain proteins in the a-globin mRNP stability complex.
Dever, T.E., Feng, L., Wek, R.C., Cigan, R.M., Donahue, T.F., and
EMBO J. 14, 4357–4364.
Hinnebusch, A.G. (1992). Phosphorylation of initiation factor 2a by
protein kinase GCN2 mediates gene-specific translational control
Kollmus, H., Hentze, M.W., and Hauser, H. (1996). Regulated ribo-
of GCN4 in yeast. Cell 68, 585–596.
somal frameshifting by an RNA–proteininteraction. RNA 2,316–323.
Dubnau, J.,and Struhl, G. (1996).RNA recognition and translational
Leffers, H., Dejgaard, K., and Celis, J.E. (1995). Characterisation of
regulation by a homeodomain protein. Nature 379, 694–699.
two majorcellular poly(rC)-bindinghuman proteins, eachcontaining
three K-homologous (KH) domains. Eur. J. Biochem. 230, 447–453.
Evans, T.C., Crittenden, S.L., Kodoyianni, V., and Kimble, J. (1994).
Translational control of maternal glp-1 mRNA establishesan asym-
Mann, M., and Wilm, M.S. (1994). Error tolerant identification of
metry in the C. elegans embryo. Cell 77, 183–194.
peptides in sequence databases by peptide sequence tags. Anal.
Fleming, J., Thiele, B.J.,Chester,J., O’Prey, J., Janetzky,S., Aitken,
Chem. 66, 4390–4399.
A., Anton,I.A., Rapoport,S.M.,and Harrison,P. (1989).The complete
Matunis, M.J., Michael,W.M., and Dreyfuss, G. (1992) Characteriza-
sequence of the rabbit erythroid cell–specific 15-lipoxygenase
tion and primary structure of the poly (C)-binding heterogeneus
mRNA: comparison of the predicted amino acid sequence of the
nuclear ribonucleoprotein complex K protein. Mol. Cell. Biol. 12,
erythroidlipoxygenase withotherlipoxygenases. Gene79, 181–188.
164–171.
Graham, F.L., and Van der Eb, A.J. (1973). A new technique for
Matunis, E., Kelley, R., and Dreyfuss, G. (1994). Essential role for
the assay of infectivity of human adenovirus 5 DNA. Virology 52,
a heterogeneous nuclear ribonucleoprotein (hnRNP) in oogenesis:
456–464.
hrp40 is absentfrom thegermline in thedorsoventral mutant squid.
Gray, N.K, Quick, S., Goossen, B., Constable, A., Hirling, H., Ku
¨
hn,
Proc. Natl. Acad. Sci. USA 91, 2781–2784.
L., and Hentze, M.W. (1993). Recombinant iron-regulatory factor
Mayeda, A., and Krainer, A.R. (1992). Regulation of alternative pre-
functions as an iron-responsive-element binding protein, a transla-
mRNA splicing by hnRNP A1 and splicing factor SF2. Cell 68,
tional repressor and an aconitase. A functional assay for transla-
365–375.
tional repression and direct demonstration of the iron switch. Eur.
J. Biochem. 218, 657–667.
Michelotti, E.F., Michelotti, G.A., Aronsohn, A.I., and Levens, D.
(1996). HeterogeneousnuclearribonucleoproteinKis atranscription
Gray,N.K.,and Hentze,M.W.(1994).Iron regulatoryproteinprevents
factor. Mol. Cell. Biol. 16, 2350–2360.
binding of the 43S translation pre-initiation complex to ferritin and
eALAS mRNAs. EMBO J. 13, 3882–3891.
Min, H.,Chan, R.C., and Black, D.L. (1995). Thegenerally expressed
hnRNP F is involved in a neural-specific pre-mRNA splicing event.
Green, S., Issemann,I., and Sheer, E. (1988). A versatilein vitro and
Genes Dev. 9, 2659–2671.
in vivo eukaryotic expression vector for protein engineering. Nucl.
Acids Res. 16, 369.
Ostareck-Lederer, A., Ostareck, D.H., Standart, N., and Thiele, B.J.
Hahm, K., Kim, G., Turck, C.W., and Smale, S.T. (1993). Isolation of
(1994). Translation of 15-lipoxygenase mRNA is inhibited by a pro-
a murine gene encoding a nucleic acid–binding protein withhomol-
tein that bindsto a repeated sequencein the 39 untranslatedregion.
ogy to hnRNP K. Nucl. Acids Res. 21, 3894.
EMBO J. 13, 1476–1481.
Hake, L.E., and Richter, J.D. (1994). CPEB is a specific factor that
Pantopoulos, K., and Hentze, M.W. (1995). Nitric oxide signaling to
mediatescytoplasmicpolyadenylation during Xenopus oocyte mat-
iron-regulatory protein: direct control of ferritin mRNA translation
uration. Cell 79, 617–627.
and transferrin receptor mRNA stability in transfected fibroblasts.
Proc. Natl. Acad. Sci. USA 92, 1267–1271.Hentze, M.W.,andKu
¨
hn,L.C. (1996). Molecularcontrolof vertebrate

Cell
606
Pin
˜
ol-Roma, S., and Dreyfuss, G. (1993). hnRNP proteins: localiza-
tion and transport between the nucleus and the cytoplasm. Trends
Biochem. Sci. 3, 151–155.
Rapoport, S.M.,andSchewe,T.(1986).Thematurational breakdown
of mitochondria in reticulocytes. Biochem. Biophys. Acta 864,
471–495.
Rivera-Pomar, R., Niessing, D., Schmidt-Ott, U., Gehring, W.J., and
Ja
¨
ckle, H. (1996). RNA binding and translational suppression by
bicoid. Nature 379, 746–749.
Roepstorff, P., and Fohlman, J. (1984). Proposed nomenclature for
sequence ions. Biomed. Mass Spectrom. 11, 601.
Schewe, T., Rapoport, S.M., and Ku
¨
hn, H. (1986). Enzymology and
physiology of reticulocyte lipoxygenase: comparison with other li-
poxygenases. In Advances in Enzymology and Related Areas of
Molecular Biology,Volume 58,A.Meister, ed.(NewYork: JohnWiley
and Sons, Inc.), pp. 191–271.
Shevchenko, A., Wilm, M., Vorm, O., and Mann, M. (1996). Mass
spectrometric sequencing of proteins from silver stained polyacryl-
amide gels. Anal. Chem. 68, 850–858.
Stripecke, R., Oliveira, C.C., McCarthy, J.E.G., and Hentze, M.W.
(1994).Protein binding to 59 untranslatedregionsites: generalmech-
anism for translational regulation of mRNAs in human and yeast
cells. Mol. Cell. Biol. 14, 5898–5909.
Swanson, M.S., and Dreyfuss, G. (1988). Classification and purifica-
tion of proteins of heterogeneous nuclear ribonucleoprotein parti-
cles by RNA-binding specificities. Mol. Cell. Biol. 8, 2237–2241.
Takimoto, M., Tomonaga, T., Matunis, M., Avigan, M., Krutzsch, H.,
Dreyfuss, G., and Levens D. (1993). Specific binding of heteroge-
neous ribonucleoprotein particleprotein K tothe human c-myc pro-
moter in vitro. J. Biol. Chem. 268, 18249–18258.
Taylor, S.J.,and Shalloway, D.(1994).An RNA-binding proteinasso-
ciated with Src through itsSH2 andSH3 domains inmitosis. Nature
368, 867–871.
Thiele,B.J.,Andree,H.,Ho
¨
hne,M., andRapoport,S.M. (1981).Regu-
lation of the synthesis of lipoxygenase in erythroid cells. Acta Biol.
Med. Germanica 40, 597–602.
Thiele, B.J., Andree, H., Ho
¨
hne, M., and Rapoport, S.M. (1982). Li-
poxygenase mRNA in rabbit reticulocytes. Its isolation, character-
ization and translational repression. Eur. J. Biochem. 129, 133–141.
Van Seuningen, I., Ostrowski, J., Bustelo, X.R., Sleath, P.R., and
Bomsztyk, K. (1995). TheK proteindomain that recruitsthe interleu-
kin 1–responsive K protein kinase lies adjacent to a cluster of c-Src
and Vav SH3-binding sites. J. Biol. Chem. 270, 26976–26985.
Wang, X., Kiledjian, M., Weiss, I.M., and Liebhaber, S.A. (1995).
Detection and characterization of a 39 untranslated region ribo-
nucleoprotein complex associated with the human a-globin mRNA
stability. Mol. Cell. Biol. 15, 1769–1777.
Weng, Z., Thomas, S.M., Rickles, R.J., Taylor, J.A., Brauer, A.W.,
Seidel-Dugan, C., Michael, W.M., Dreyfuss, G., and Brugge, J.S.
(1994). Identification of Src, Fyn, and Lyn SH3-binding proteins:
Implicationsfor afunctionof SH3-domains.Mol.Cell. Biol. 14,4509–
4521.
Wickens, M. (1993). Springtime in the desert. Nature 363, 305–306.
Wickens, M.,Kimble,J., andStrickland, S.(1996).Translational con-
trol of developmental decisions. In Translational Control, J.W.B.
Hershey, M.B. Mathews, and N. Sonenberg, eds. (Cold Spring Har-
bor, New York: Cold Spring Harbor Laboratory Press),pp. 411–450.
Wilm, M.S., and Mann, M. (1994). Electrospray and Taylor-Cone
theory, Dole’s beam of macromolecules at last? Int. J. Mass Spec-
trom. Ion Proc. 136, 167–180.
Wilm, M., and Mann, M. (1996). Analytical properties of the nano
electrospray ion source. Anal. Chem. 68, 1–8.
Wilm, M., Shevchenko, A., Houthaeve, T., Breit, S., Schweigerer, L.,
Fotsis, T., and Mann,M.(1996a). Femtomole sequencing of proteins
from polyacrylamide gels by nano electrospray mass spectrometry.
Nature 379, 466–469.
Wilm, M., Neubauer, G., and Mann, M. (1996b). Parent ion scans of
unseparated peptide mixtures. Anal. Chem. 68, 527–533.