110 Brief Communication
Self assembly of the transmembrane domain promotes signal
transduction through the erythropoietin receptor
Katharina F. Kubatzky*, Weiming Ruan†, Rolf Gurezka†, Jacob Cohen‡,
Robin Ketteler*, Stephanie S. Watowich§, Drorit Neumann‡,
Dieter Langosch† and Ursula Klingmu¨ller*
Hematopoietic cytokine receptors, such as the Results and discussion
erythropoietin receptor (EpoR), are single To test the dimerization potential of the erythropoietin
membrane-spanning proteins. Signal transduction receptor (EpoR) transmembrane domain (TM), we used
through EpoR is crucial for the formation of mature the ToxR reporter system [5]. This assay is based on mem-
erythrocytes [1]. Structural evidence shows that in brane-spanning chimeric ToxR proteins containing heter-
the unliganded form EpoR exists as a preformed ologous TMs that trigger the transcription of a b-galactosi-
homodimer in an open scissor–like conformation [2] dase reporter gene upon TM-mediated dimerization of
precluding the activation of signaling. In contrast the protein [6]. In agreement with our previous observa-
to the extracellular domain of the growth hormone tion [7], expression of the leucine-rich segments of the
receptor (GHR) [3], the structure of the agonist- murine or human EpoR TM in the context of the ToxR
bound EpoR extracellular region shows only minimal protein elicited significant b-galactosidase expression.
contacts between the membrane-proximal regions The signal was comparable to that achieved with a self-
[4]. This evidence suggests that the domains associating oligo-leucine helix (Leu 16), but it was slightly
facilitating receptor dimerization may differ weaker than the strongly dimerizing TM of glycophorin
between cytokine receptors. We show that the EpoR A (GPA) [6, 8] (Figure 1b). In comparison, an equivalent
transmembrane domain (TM) has a strong part of the GHR TM (Figure 1a) did not significantly
potential to self interact in a bacterial reporter self interact (Figure 1b). This finding supports the notion
system. Abolishing self assembly of the EpoR TM that the dimerization mechanisms may differ for GHR
by a double point mutation (Leu 240–Leu 241 and EpoR. To disrupt self assembly of the murine EpoR
mutated to Gly-Pro) impairs signal transduction by TM, we introduced several point mutations. Whereas
EpoR in hematopoietic cells and the formation of multiple exchanges to alanine residues or the mutation
erythroid colonies upon reconstitution in erythroid of Leu 241 in the center of the TM to proline (L!P)
progenitor cells from EpoR2/2 mice. Interestingly, did not appreciably affect the ability to self interact, the
inhibiting TM self assembly in the constitutively simultaneous mutation of two conserved leucine residues
active mutant EpoR R129C abrogates formation of (LL!GP; Leu 240 to glycine and Leu 241 to proline)
disulfide-linked receptor homodimers and strongly reduced self assembly, presumably by introduc-
consequently results in the loss of ligand- ing a kink into the TM helix (Figure 1b). Control experi-
independent signal transduction. Thus, efficient ments confirmed that the ToxR chimeric proteins were
signal transduction through EpoR and possibly other expressed at comparable levels (Figure 1c) and function-
preformed receptor oligomers may be determined ally complemented maltose binding proteins (MalE) defi-
by the dynamics of TM self assembly. ciency, which indicated their correct membrane insertion
(for details, see Supplementary material).
*Max-Planck-Institute of Immunobiology, Freiburg 79108, Germany.
†Department of Neurobiology, University of Heidelberg, Heidelberg To determine whether reduced self assembly of the TM
69120, Germany. ‡Sackler Faculty of Medicine, Department of Cell
influenced signal transduction through EpoR, we intro-Biology and Histology, Tel-Aviv University, Ramat-Aviv 69978, Israel.
duced the LL!GP mutation into the TM of the wild-§University of Texas MD Anderson Cancer Center, Houston, Texas
77030, USA. type EpoR and the constitutively active mutant EpoR
R129C [9] to generate EpoR-GP and R129C-GP, respec-
Correspondence: Ursula Klingmu¨ller
tively. Receptors harbouring either the wild-type or theE-mail: klingmueller@immunbio.mpg.de
mutant TM were expressed in the interleukin (IL)-3 de-
Received: 16 October 2000 pendent cell line BaF3. Binding of [125I]Epo demonstrated
Revised: 27 November 2000 that EpoR cell surface expression was not affected by the
Accepted: 28 November 2000
mutations in the TM (for details see the Supplementary
material available with the electronic version of this arti-Published: 23 January 2001
cle). To compare the signaling capacity of the wild-type
Current Biology 2001, 11:110–115 EpoR and the EpoR-GP, we analyzed ligand-induced
tyrosine phosphorylation of EpoR, JAK2, and STAT5.0960-9822/01/$ – see front matter
Ó 2001 Elsevier Science Ltd. All rights reserved. EpoR immunoprecipitations and anti-phosphotyrosine
Brief Communication 111
Figure 1 Figure 2
Self assembly of cytokine receptor TMs. (a) Comparison of the TMs
Importance of the TM for efficient EpoR signal transduction andof the murine (m) EpoR, the human (h) EpoR and the human (h)
receptor dimerization. Epo-induced tyrosine phosphorylation of theGHR. The comparison is based on an alignment performed by Clustal
(a) EpoR, (b) JAK2, and (c) STAT5 was determined in BaF3 cellsW that uses the conserved box1 sequence present in the
showing comparable surface expression of the receptors (for detailscytoplasmic domain of cytokine receptors as an anchor point (not
see Supplementary material). The BaF3 cell lines were left unstimulatedshown). The mutants derived from the murine EpoR TM are indicated,
or were stimulated with 0.1–10 U/ml Epo. Immunoprecipitations (IP)and dots represent wild-type residues. (b) Self assembly of cytokine
with anti-EpoR (Santa Cruz), anti-JAK2 (UBI), or anti-STAT5b (Santareceptor TMs as compared to self-interacting oligo-leucine (L16) or
Cruz) antisera were analyzed by immunoblotting (IB) using the anti-GPA TMs determined by b-galactosidase activity (mean 6 standard
phosphotyrosine antibody 4G10 (UBI). Equal protein loading wasdeviation, n 5 20) upon the expression of the ToxR chimeric proteins
ensured by reprobing the immunoblots with antibodies recognizingin the reporter strain FHK12. The activity elicited by an oligo-alanine
the respective proteins.(A16) reflects the background of the system (dashed line). (c)
Comparable expression of the ToxR chimeric proteins was confirmed
by immunoblotting, for which a MalE antiserum was used. Whereas
A16 is consistently overexpressed, as noted previously [7], in some
An important mediator of signal transduction through cy-experiments the mutant LL!GP was less efficiently expressed.
tokine receptors such as EpoR is the JAK-STAT pathway
[10]. Therefore, we asked whether the LL!GP muta-
tions in the EpoR TM impaired ligand-induced JAK2immunoblot analysis revealed that the receptors analyzed
activation and consequently resulted in less tyrosine-were functional and showed an increase in receptor tyro-
phosphorylated STAT5. As shown in Figure 2b, tyrosinesine phosphorylation proportional to the amount of Epo
phosphorylation of JAK2 is reduced in cells expressingused for stimulation (Figure 2a). In cells expressing the
EpoR-GP. In fact, in saturating Epo concentrations, thewild-type EpoR, a stimulus of 1 U/ml Epo elicited low
level of JAK2 tyrosine phosphorylation in EpoR-GP cellslevels of receptor tyrosine phosphorylation, and maximal
is only 48% of the level observed in wild-type EpoRphosphorylation was achieved at 5 U/ml Epo (Figure 2a,
cells (Figure 2b; LumiAnalyst quantification not shown).upper left panel). In contrast, EpoR-GP was less effi-
Interestingly, ligand-induced tyrosine phosphorylation ofciently tyrosine phosphorylated at all Epo concentrations
STAT5 was reduced in EpoR-GP cells at low Epo concen-tested. Even saturating concentrations of Epo (10 U/ml)
trations (e.g., 1 U/ml Epo), whereas STAT5 tyrosine phos-resulted in the stimulation of only 63% of receptor tyrosine
phorylation levels were comparable to those achieved withphosphorylation relative to the wild-type EpoR (Figure
the wild-type EpoR at high doses of Epo (Figure 2c).2a, LumiAnalyst quantification not shown). These results
Therefore, the LL!GP mutation in the EpoR TM abro-demonstrate that self assembly of the TMs plays a critical
role in tyrosine phosphorylation of EpoR. gates the efficient activation of EpoR-dependent signal-
112 Current Biology Vol 11 No 2
Figure 3ing pathways, particularly in subsaturating concentrations
of ligand that approximate serum levels of Epo in vivo.
In summary, inactivating the self interaction of the EpoR
TM as determined by the ToxR assay correlates with the
loss of efficient signaling through EpoR, which is not
compensated by doses of Epo exceeding 10 U/ml Epo
(data not shown).
Constitutive activation of EpoR R129C is due to ligand-
independent formation of disulfide-linked receptor ho-
modimers [9]. To assess whether self assembly of the
EpoR TM is required for constitutive receptor activity, we
compared tyrosine phosphorylation of R129C and R129C-
GP. As expected, R129C stimulated a low level of Epo-
independent receptor tyrosine phosphorylation (Figure 3a),
which was increased upon Epo stimulation. Strikingly,
ligand-independent tyrosine phosphorylation was not ob-
served in cells expressing R129C-GP (Figure 3a). Further-
more, in BaF3-R129C-GP cells stimulated with 10 U/ml
Epo, receptor tyrosine phosphorylation was reduced by
67% compared to the level obtained in BaF3-R129C cells.
To determine whether the LL!GP mutation in the TM
of R129C arrests the receptor in a conformation pre-
venting signal transduction or whether it impairs forma-
tion of disulfide-linked receptor homodimers, we per-
formed metabolic pulse-chase labeling experiments to
assess dimerization (Figure 3b). As expected, R129C
formed disulfide-linked receptor dimers (Figure 3b, lane
3) as evidenced by approximately 160 kDa receptor com-
plexes detected in nonreducing gel electrophoresis [9].
In contrast, R129C-GP did not form disulfide-linked di-
mers at detectable levels (Figure 3b, lane 6). These results The LL!GP mutation in the TM abrogates formation of the
show that formation of a membrane-proximal disulfide constitutively active EpoR R129C disulfide-linked homodimers.
BaF3 cells expressing the mutant EpoRs R129C or R129C-GP werebond in R129C is facilitated by self assembly of the EpoR
pulse labeled (P) with [35S]methionine/cysteine for 10 min at 378CTM and indicate that EpoR dimerization may be medi-
and then chased for 2 hr at 378C (37) or 188C (18).ated through the receptor TM region.
Immunoprecipitated EpoRs were separated by nonreducing gel
electrophoresis and detected by autoradiography as described
previously [9]. Arrows indicate the positions of the monomeric EpoR.Next we asked whether reduced signaling by EpoRs har-
The position of the homodimeric EpoR is marked with an asterisk.
boring the LL!GP mutation is translated into impaired
biological responses. First, we determined whether EpoR-
GP and R129C-GP could stimulate proliferation of BaF3
cells (Figure 4a). As expected, parental (e.g., untrans- was unable to stimulate Epo-independent cell prolifera-
tion (Figure 4a), and this finding is consistent with theduced) BaF3 cells did not respond to Epo, while the
proliferation of cells expressing the wild-type EpoR was inability of R129C-GP to undergo constitutive disulfide-
linked dimerization. Collectively, these data demonstratedetected in Epo concentrations as low as 0.1 U/ml Epo.
Maximal proliferation of wild-type EpoR cells was ob- the importance of TM self interaction for both Epo-depen-
dent proliferation through the wild-type EpoR and ligand-tained at Epo concentrations ranging from 5 to 10 U/ml
Epo; cell growth at these Epo concentrations was equiva- independent proliferation mediated by R129C.
lent to 95% of the growth obtained in IL-3–conditioned
medium (Figure 4a). In contrast, BaF3 cells expressing To investigate if the EpoR TM mutations affect erythroid
development, we analyzed Epo-dependent and Epo-inde-EpoR-GP showed reduced proliferative responses at all
Epo concentrations tested. Even at the highest dose of Epo pendent CFU-E colony formation. We retrovirally trans-
duced erythroid progenitor cells from EpoR2/2 animals(10 U/ml Epo), the levels of EpoR-GP cell proliferation
were only 51% of those observed for wild-type EpoR to reconstitute EpoR expression and cultivated them in
methylcellulose supplemented with either Epo and stemcells. As expected, R129C supported cell proliferation in
the absence of Epo (Figure 4a). In contrast, R129C-GP cell factor (SCF) or SCF only for 2 days to allow for
Brief Communication 113
Figure 4 expression of R129C stimulated Epo-independent CFU-E
formation to levels similar to those achieved with wild-
type EpoR-expressing cells cultured in Epo and SCF.
Most strikingly, the TM mutations present in R129C-GP
impaired Epo-independent CFU-E development and re-
sulted in a 69% (p , 0.001) reduction of the number of
CFU-E colonies formed by R129C-GP–expressing cells
relative to R129C-expressing cells. Thus, self assembly of
the EpoR TM facilitates efficient development of erythro-
cytes in the presence of Epo and critically supports ligand-
independent erythroid maturation mediated by R129C.
In summary, our data demonstrate that decreasing self
assembly of the EpoR TM by the LL!GP mutation
correlates with the loss of disulfide-linked receptor homo-
dimers in the mutant EpoR R129C and results in impaired
signal transduction that is required for cell proliferation
and differentiation.
Previous crystallization of the extracellular domain sug-
gests that the unliganded EpoR exists as a preformed
dimer in which the TMs are kept at a distance of 73 A˚
[11]. Agonist binding to EpoR was reported to reduce the
inferred TM-TM distance to 39 A˚ [12]. Considering these
data and the results we have presented, we propose that
in the absence of ligand, the homodimeric EpoR exists in
a conformational equilibrium between (1) a configuration
positioning the TMs in a signaling-incompetent distance
due to preassociation of the extracellular domains and (2)
juxtaposition of the TMs mediated by their self interac-
tion (Figure 5). This latter conformation is transient unless
agonist binding stabilizes the juxtaposition of the TMs
and thus ensures efficient signaling. Therefore, the dy-
namics of TM self interaction may facilitate sufficient
Self interaction of the EpoR TM modulates proliferation and
approximation of the unliganded extracellular domains indifferentiation mediated by EpoR. (a) Proliferation of parental BaF3
the constitutively active EpoR R129C to support forma-cells or BaF3 cells expressing the wild-type EpoR or mutant EpoRs
in response to Epo. Cell numbers were determined with a Coulter tion of the membrane-proximal disulfide bond. Apart from
counter. Growth is displayed as the mean percentage 6 standard EpoR, ligand-independent oligomers have been reported
deviation of the cell numbers obtained in WEHI conditioned medium for other receptors, including those for interleukin-2, epi-for three independent cell pools. The experiment was performed four
dermal growth factor, FAS, and tumor necrosis factor [13–times with similar results. (b) Formation of CFU-E colonies upon
expression of the wild-type or mutant EpoRs in fetal liver cells from 16]. It is conceivable that in these receptors self assembly
EpoR2/2 mice. CFU-E colonies were counted upon benzidine staining of the TMs also promotes a conformational switch re-
of hemoglobinized cells. The values plotted (mean 6 standard quired for efficient signal transduction.deviation, n 5 4) are given as the percentage of CFU-E colonies
achieved upon transduction with the wild-type EpoR (on average 52
CFU-E colonies per 5 3 103 fetal liver cells). P-values indicated by Materials and methods
double asterisks (p , 0.01) and triple asterisks (p , 0.001) were
Plasmids and constructsdetermined with a two-sided Student’s T test comparing receptors
Mutations in the murine EpoR [17] were generated by PCR mutagenesiswith wild type TM with the respective GP mutant. Similar results were
and were subcloned in frame into the retroviral expression vector pMXobtained in three independent experiments. Measuring GFP expression
puro (for further details see Supplementary material). For transductionin transduced wild-type fetal liver cells by flow cytometry confirmed
of fetal liver cells, the MoMLV-based bicistronic retroviral expressioncomparable gene transfer rates of the transducing supernatants [20].
vector pOS-EpoR-IRES-GFP (I. Swameye, S. Esser, R. Lee, M. Wickert,
J. K. Offe, R. K., J. C., D. N., H. Wu, and U. K., unpublished data) was
used.
CFU-E colony formation. As expected, CFU-E formation
Cell lines and cultureswas strictly Epo-dependent in cells transduced with the Transducing supernatants were generated and used to transduce the
wild-type EpoR (Figure 4b) and was reduced by 21% (p , IL-3 dependent pro-B cell line BaF3 or fetal liver cells as described
(R. K., S. Glaser, O. Sandra, U. M. Martens, and U. K., submitted). BaF30.01) upon the expression of EpoR-GP. In contrast, the
114 Current Biology Vol 11 No 2
Figure 5
Self assembly of the TM promotes a
conformational switch that upon ligand binding
permits the activation of signaling. In the
absence of a ligand, EpoR oscillates
between two conformations mediated by self
interaction of the extracellular domains or the
TMs. In the mutant EpoR R129C,
approximation of the TMs supports the
formation of the membrane-proximal disulfide
bond. The position of R129 is indicated (R).
Epo binding stabilizes the conformation
induced by TM self assembly, and this results
in the initiation of signal transduction (black
arrows) through EpoR as indicated by JAK2
activation and tyrosine phosphorylation (P) of
cellular proteins, including EpoR.
cell pools were selected, maintained, and tested in proliferation assays Acknowledgements
as described [18]. We thank Dr. O. Sandra for his assistance with fetal liver cell experiments,
Dr. I. Swameye for critical discussions, and Dr. W. Huttner for continuous
support. We also thank Dr. H. Wu for generously providing the EpoR2/2
Fetal liver cells of 13-day-old embryos from EpoR2/2 mice [1] were mice and Dr. C. Galanos for frequent use of his Coulter counter. We acknowl-
edge Dr. K. Eichmann for critically reading the manuscript. This work wasprepared and transduced as described (R. K., S. Glaser, O. Sandra,
supported by grants from the Deutsche Forschungsgemeinschaft (SFB 364)U. M. Martens, and U. K., submitted). The transduced cells were plated
to R. K. and U. K., the Deutsche Forschungsgemeinschaft (grant La699/4-in 0.8% methylcellulose (CellSystem) supplemented with 4 U/ml Epo
3, Graduiertenkolleg and Heisenberg Programm) and the Fonds der Chem-(Cilag-Jansen) and 20 ng/ml SCF. CFU-E colony formation was moni-
ischen Industrie to D.L., by grants from the Israeli Science Foundation adminis-tored by benzidine staining of hemoglobinized cells. To ensure compara- tered by the Israeli Academy of Sciences and Humanities and the Israeli
ble transduction rates, we assessed GFP expression by FACS analysis Cancer Research Fund to D. N., and by grants from the National Cancer
(FACScan, Becton Dickinson) in transduced fetal liver cells of wild- Institute, National Institutes of Health (CA 77447), the Texas Higher Educa-
type mice after 20 hr of cultivation in IMEM FCS/b-mercaptoethanol tion Coordinating Board (15-120), and the Gillson Longenbaugh Foundation
supplemented with 0.4 U/ml Epo (R. K., S. Glaser, O. Sandra, U. M. to S. S. W.
Martens, and U. K., submitted).
References
1. Wu H, Liu X, Jaenisch R., Lodish HF: Generation of committedImmunoprecipitation and immunoblottin