Report

RNA Helicase DDX3 Is a Regulatory Subunit of Casein Kinase 1 in Wnt–β-Catenin Signaling

See allHide authors and affiliations

Science  22 Mar 2013:
Vol. 339, Issue 6126, pp. 1436-1441
DOI: 10.1126/science.1231499

Three Tales of Wnt Signaling

The Wnt signaling pathway has important roles in regulating many biological processes during development and is also implicated in the behavior of some cancer cells (see the Perspective by Berndt and Moon). Cruciat et al. (p. 1436, published online 14 February) describe the mechanism of action of a protein found in a screen for proteins that influence Wnt signaling. DDX3, a DEAD-box RNA helicase, is required for proper Wnt signaling in Xenopus and Caenorhabditis elegans. It appears to act not through its action as an RNA helicase or through adenosine triphosphate binding, but rather by interacting with the protein kinase, casein kinase 1, and promoting its activation. Huang et al. (p. 1441, published online 31 January) investigated the function of receptor-interacting protein kinase 4 (RIPK4), the product a gene whose mutation causes severe developmental defects in mice and humans. Over-expression of the protein in cultured human cells activated transcription of genes regulated by the Wnt signaling pathway, and loss of RIPK4 function inhibited Wnt signaling in Xenopus embryos. At the molecular level, RIPK4 interacted with the Wnt co-receptor LRP6 and the Wnt signaling adaptor protein DVL2 and promoted phosphorylation of DVL2. Habib et al. (p. 1445) used Wnt-immobilized beads to understand how external cues direct asymmetrical stem cell divisions. Spatially restricted Wnt signals oriented the plane of mitotic division and lead to pluripotency gene expression in the Wnt-proximal daughter cell while the more distal daughter cell acquired hallmarks of differentiation. Thus, asymmetric gene expression patterns can arise as a consequence of orientation by a short-range signal.

Abstract

Casein kinase 1 (CK1) members play key roles in numerous biological processes. They are considered “rogue” kinases, because their enzymatic activity appears unregulated. Contrary to this notion, we have identified the DEAD-box RNA helicase DDX3 as a regulator of the Wnt–β-catenin network, where it acts as a regulatory subunit of CK1ε: In a Wnt-dependent manner, DDX3 binds CK1ε and directly stimulates its kinase activity, and promotes phosphorylation of the scaffold protein dishevelled. DDX3 is required for Wnt–β-catenin signaling in mammalian cells and during Xenopus and Caenorhabditis elegans development. The results also suggest that the kinase-stimulatory function extends to other DDX and CK1 members, opening fresh perspectives for one of the longest-studied protein kinase families.

Wnt–β-catenin signaling plays a pivotal role in the development of multicellular organisms and in disease, notably cancer (1, 2). One important question in Wnt–β-catenin signaling concerns the casein kinase 1 (CK1) family, members of which have both negative and positive functions in Wnt signaling (3). CK1 kinases play diverse cellular roles from yeast to human—e.g., in membrane transport, cell division, DNA repair, circadian rhythms, and nuclear localization, as well as Wnt and Hedgehog signaling. Yet, despite their importance, whether and how these serine-threonine kinases are regulated is poorly understood, and the prevailing view is that CK1 family members are constitutively active (35). Among CK1 isoforms, CK1ε plays a critical and evolutionarily conserved role in Wnt–β-catenin signaling (68). As a primary target, CK1ε phosphorylates and activates the scaffold protein dishevelled (Dvl) (9, 10), which binds components at the interface of Wnt receptors and the β-catenin destruction complex (11, 12). Dvl phosphorylation by CK1ε promotes binding of the coeffector Frat, dissociation of PP2A from the β-catenin degradation complex, and stabilization of β-catenin (13, 14).

We identified DDX3 in a genome-wide small interfering RNA (siRNA) screen for previously undescribed Wnt–β-catenin regulators in human embryonic kidney 239T (HEK293T) cells (15, 16) (information on materials and methods is available on Science online). DDX3 belongs to the family of adenosine 5′-triphosphate (ATP)–dependent DEAD-box RNA helicases, for which cellular functions and biological roles remain incompletely understood (17). DDX3 is a multifunctional protein involved in mRNA biogenesis processes and is implicated in cell cycle control, apoptosis, tumorigenesis, and viral infection (18). Knockdown of DDX3 by four independent siRNAs inhibited signaling in TOPFLASH reporter assays stimulated by Wnt3a but not by β-catenin (fig. S1, A and B). Cotransfection of Xenopus DDX3 rescued siDDX3 in a dose-dependent manner (fig. S1C), excluding siRNA off-target effects. Wnt1, Wnt3a, and Dvl1 signaling were inhibited by siDDX3, whereas constitutively active LRP6 (LRP6ΔE1-4) and β-catenin were unaffected (Fig. 1A). This epistasis is consistent with DDX3 acting at the level of Dvl, which is required for LRP6 activation. Knockdown of DDX3 and of all three Dvls by siRNA did not reduce cytoplasmic β-catenin abundance, but inhibited nuclear β-catenin accumulation (Fig. 1B), indicating that loss of DDX3 phenocopies loss of Dvl in Wnt signaling. Overexpressed DDX3 synergized with Wnt3a in Xenopus embryonic axis duplication assays (Fig. 1C), as well as in siamois induction (Fig. 1D). These results suggest that DDX3 functions in Wnt–β-catenin signaling also in Xenopus.

Fig. 1

DDX3 is required for Wnt–β-catenin signaling in mammalian cells and during a-p neural patterning in Xenopus. (A) Wnt luciferase reporter assay in HEK293T cells stimulated with Wnt3a-conditioned medium or by transfection with the indicated constructs, in the presence of the indicated siRNAs. RLA, relative luciferase activity. Error bars indicate SDs; n = 3, biological triplicates of one representative assay. (B) Western blot analysis of endogenous β-catenin from cytosolic and nuclear fractions of HEK293T cells stimulated with control or Wnt3a-conditioned medium in the presence of the indicated siRNAs. (C) Xenopus axis duplication assay by injection of the indicated mRNAs into the ventral blastomeres of four-cell-stage embryos. (D) Quantitative polymerase chain reaction (QPCR) analysis of siamois using VMZ explants from Xenopus embryos injected with the indicated mRNAs. Explants were excised and analyzed from gastrula-stage embryos. Error bars indicate SDs; n = 2 assays. VMZ, ventral marginal zone; sia, siamois. (E) Tadpole-stage Xenopus embryos that were injected at the two-cell stage in the animal hemisphere with DDX3 or LRP6 antisense Mo oligonucleotides in the absence or presence of human DDX3 mRNA (WT, wild type) as indicated. (F) Whole-mount in situ hybridization of neurula-stage embryos injected at four-cell stage in animal blastomeres with the indicated Mo plus β-galactosidase mRNA lineage tracer (red; arrow marks injected side).

DDX3 was ubiquitously expressed throughout Xenopus embryogenesis (fig. S2). Injection of DDX3 antisense morpholino (Mo) oligonucleotides elicited anteriorized embryos with enlarged heads and eyes, shortened tails, and defective melanocyte and eye pigmentation (Fig. 1E). This is the characteristic phenotype caused by inhibition of zygotic Wnt signaling (19, 20). Indeed, knockdown of the Wnt co-receptor LRP6 with established morpholinos (21) phenocopied DDX3 morphants (Fig. 1E). The DDX3 Mo phenotype was specific because it was very efficiently rescued by coinjection of human DDX3 mRNA (Fig. 1E).

Anteriorization of DDX3 morphants was confirmed by expansion of the forebrain markers bf1 and otx2, as well as down-regulation of the mid-hindbrain boundary marker en2 and the hindbrain marker krox20, whereas the pan-neural marker sox3 was unaffected (Fig. 1F and fig. S4A). This expression pattern was similar to that observed in LRP6 morphants (Fig. 1F). Moreover, DDX3 morphants showed no apoptosis or changes in cell proliferation in the central nervous system (CNS) (fig. S3), indicating that the altered neural marker gene expression represents a defect of antero-posterior (a-p) neural patterning, wherein Wnt–β-catenin signaling acts as a morphogen (22). This conclusion was corroborated by gain-of-function experiments. When injected as plasmid DNA in Xenopus embryos, overexpressed XWnt8 induces posteriorized embryos, which lack anterior neural structures. DDX3 phenocopied this defect, because embryos injected with either DDX3 or XWnt8 DNA showed reduced expression of otx2, en2, and sox3 and moderately increased expression of krox20 (fig. S4B).

In Xenopus animal cap assays, DDX3 Mo inhibited Wnt3a-stimulated induction of the direct Wnt targets siamois, xnr3, and myoD, as did LRP6 Mo (fig. S4C). Expression of siamois was rescued by coinjection of human DDX3 as well as of β-catenin mRNA (fig. S4D), indicating that DDX3 is specifically required for Wnt signaling and that it acts upstream of β-catenin. Moreover, in Xenopus Wnt-reporter assays, DDX3 Mo inhibited both endogenous and Wnt3a-stimulated luciferase expression (fig. S4E). The requirement of DDX3 was specific for Wnt, because signaling by various other growth factors (bone morphogenetic protein, nodal, transforming growth factor–β, and fibroblast growth factor) was unaffected by DDX3 depletion in Xenopus (fig. S5, A to E) and HEK293T cells (fig. S5, F to I). We conclude that DDX3 regulates a-p CNS patterning in Xenopus embryos through Wnt–β-catenin signaling.

To investigate whether the function of DDX3 in canonical Wnt–β-catenin signaling is evolutionarily conserved beyond vertebrates, we knocked down F58E10.3, a DDX3 homolog in Caenorhabditis elegans. We assayed migration of Q neuroblast descendants, which depends on canonical EGL-20–Wnt signaling (fig. S6, A to D) (23). To sensitize the assay, we included a mutation in the Wnt secretion factor vps-29 to reduce EGL-20–Wnt ligand concentration (24). Ubiquitous knockdown of F58E10.3 resulted in a twofold enhancement of the Wnt loss-of-function phenotype in vps-29 mutants [QL migration: 35% Co RNA interference (RNAi); n = 187; 70% F58E10.3 RNAi; n = 60; data not shown]. This suggests that F58E10.3 is required for EGL-20–Wnt signaling. Next, we used the egl-17 promoter to specifically express F58E10.3 double-stranded RNA in the QL neuroblasts (25). This also resulted in an increased Wnt signaling defect (fig. S6B), demonstrating that F58E10.3 functions cell autonomously. To position F58E10.3 in the canonical EGL-20–Wnt pathway, we performed epistasis experiments with transgenic lines that overexpress EGL-20–Wnt (26) or a constitutively active, N-terminally truncated version of BAR-1–β-catenin (ΔN-BAR-1) (27). Overexpression of EGL-20–Wnt induces posterior migration of QR descendants (fig. S6A bottom) (26), and we found that this response was reduced in animals treated with F58E10.3 RNAi (fig. S6C). By contrast, the response to ΔN-BAR-1 overexpression was similar in control and F58E10.3 RNAi–treated animals (fig. S6D), indicating that F58E10.3 acts downstream of EGL-20–Wnt but upstream of BAR-1–β-catenin. In support of this, we found that F58E10.3 RNAi enhances the Wnt signaling defect of mig-5–Dvl RNAi (fig. S6B′). The results show that a DDX3-related helicase is required for Wnt signaling in C. elegans, supporting its evolutionarily conserved function.

Because Dvl is positively regulated by DDX3 and CK1ε phosphorylation (9, 10), we analyzed whether DDX3 may affect phosphorylation of Dvl2. Phospho-Dvl2 is typically monitored by its lower electrophoretic mobility, and both siDDX3 and siCK1ε inhibited Dvl2 phosphorylation (Fig. 2A). Conversely, DDX3 cotransfected with a limiting dose of CK1ε induced Dvl2 phosphorylation (Fig. 2B). Moreover, in Xenopus animal cap assays, DDX3 synergized in a dose-dependent manner with limiting doses of Dvl2 and CK1ε, but not with β-catenin, to increase expression of siamois (Fig. 2C). Thus, DDX3 is necessary and sufficient for Dvl2 phosphorylation by CK1ε and functionally interacts with both in Wnt signaling.

Fig. 2

DDX3 is required for Dvl2 phosphorylation and signalosome formation. (A) Western blot of endogenous Dvl2, DDX3, and CK1ε from lysates of HEK293T cells treated with the indicated siRNAs. (B) Western blot analysis of Dvl2 phosphorylation status in lysates from HEK293T cells transfected as indicated. (C) QPCR analysis of siamois (sia) in animal cap explants from Xenopus embryos. Injected mRNA: hDDX3 (0.5 and 1 ng). Ornithine decarboxylase was used for normalization. Dvl2, CK1ε, and β-catenin injected samples are set to 1. (D) Western blot of the indicated endogenous proteins from membrane lysates of NTERA2 cells treated with the indicated siRNAs and stimulated for 1 hour with Wnt3a-conditioned medium or control medium. tot. LRP6, total LRP6. (E) Confocal microscopy of HeLa cells transfected with Dvl2-HA or V5-CK1ε in the absence or presence of DDX3-Myc. A low dose of CK1ε was used, as high CK1ε leads to nonvesicular Dvl due to decreased polymerization (41). (F) CoIP of V5-CK1ε with Dvl2-Flag from lysates of transfected HEK293T cells in the presence or absence of DDX3-Myc.

Wnt signal transduction involves LRP6 receptor clustering in signalosomes and phosphorylation at Thr1479 (T1479), a process that depends on Dvl (28). Thus, LRP6 T1479 phosphorylation is also a proxy for Dvl activation. Indeed, Wnt-induced LRP6 T1479 phosphorylation was inhibited not only by CK1ε and Dvls siRNAs, but also by depletion of DDX3 (Fig. 2D), showing its requirement for signalosome formation.

Dvl localizes in characteristic cytoplasmic punctae, which reflect its property to oligomerize (28, 29). By immunofluorescence microscopy of cells transfected with Dvl2 and a low dose of CK1ε, we observed the typical Dvl2 punctae, whereas CK1ε showed diffuse staining. Cotransfection of DDX3 induced recruitment of CK1ε in Dvl2 punctae (Fig. 2E). Likewise, in coimmunoprecipitation (CoIP) experiments, transfected DDX3 enhanced recruitment of CK1ε to Dvl2 (Fig. 2F). Dvl2-CK1ε-DDX3 punctae were positive for the multivesicular body (MVB) marker TSG101, and transfected Dvl2 alone also partially localized in MVBs (fig. S7, A and B). MVBs represent a late-stage compartment of endocytic Wnt signalosomes and they sequester glycogen synthase kinase GSK3β from the cytoplasm, thereby protecting β-catenin from degradation (30). These gain- and loss-of-function experiments suggest that DDX3 induces Dvl phosphorylation via CK1ε and thereby stimulates signalosome formation.

In CoIP experiments with overexpressed proteins, Dvl2 and CK1ε bound to DDX3 (fig. S7C). However, in in vitro binding assays only CK1ε bound to DDX3, but not Dvl2 (Fig. 3A), suggesting that the interaction with CK1ε is direct, whereas Dvl2 interaction with DDX3 may be indirect and, e.g., via CK1ε. The interaction of DDX3 with CK1ε was also confirmed by CoIP of endogenous proteins (Fig. 3B and fig. S8A), and this binding was insensitive to ribonuclease treatment (fig. S8, B and C). We examined whether the DDX3-CK1ε interaction is regulated by Wnt, because CK1ε activity is Wnt inducible (31). In CoIPs with endogenous proteins, Wnt3a stimulation enhanced binding of CK1ε to DDX3 (Fig. 3C). Wnt3a stimulation also promoted recruitment of DDX3 to the plasma membrane (fig. S7D). The results indicate that DDX3 binds directly to CK1ε in a Wnt-dependent manner.

Fig. 3

DDX3 binds CK1ε and Wnt promotes this interaction. (A) In vitro binding assays with the indicated recombinant proteins. IPs were performed with antibody against Myc (anti-Myc) and analyzed by Western blotting with anti-Dvl2 and anti-CK1ε. Asterisk (*) denotes nonspecific band. (B) CoIPs of endogenous proteins from lysates of HEK293T cells. Anti-Dvl3 serves as negative control. (*) Nonspecific band. (C) CoIPs of endogenous proteins from lysates of HEK293T cells stimulated with control or Wnt3a-conditioned medium. (D) Schematic representation of C-terminally Myc-tagged DDX3 constructs used. Q and I to VI represent the conserved motifs of DEAD-box helicases, which are organized in domain 1 and 2 (17). Motifs Q, I, II, and VI bind ATP and are required for ATP hydrolysis. Motifs Ia, Ib, IV, and V are involved in RNA binding. Motif III is responsible for the communication between ATP-binding and RNA-binding sites. The Wnt signaling ability of DDX3 deletion mutants (from figs. S9 and S10) is summarized on the right. (E) CoIPs from lysates of HEK293T cells transfected with the indicated tagged constructs. DDX3 constructs as in (D).

DDX3 consists of an N- and a C-terminal region, which form independently folding subdomains (32). Both domains contribute to two enzymatic activities of DDX3, ATP hydrolysis and RNA unwinding (Fig. 3D). Structure-function analysis revealed that for DDX3 binding to CK1ε (Fig. 3E) and for Wnt signaling (figs. S9, A to D, and S10), neither of these enzymatic activities is essential and that most of the C-terminal subdomain of DDX3 is required, whereas the N-terminal subdomain is dispensable. Deletion of the N terminus up to amino acid residue 456, including domains essential for ATP binding and hydrolysis and RNA helicase, had little effect on CK1ε binding (Fig. 3E, lanes 3 and 4) or on the ability of the protein to rescue Wnt signaling in siDDX3-treated cells (fig. S9, A to C). By contrast, DDX3508-662 failed in both binding to CK1ε (Fig. 3E, lane 5) and Wnt signaling (fig. S9D). Failure in Wnt signaling was also observed for C-terminal deletions DDX3456-608, DDX3456-573, and DDX31-573 (fig. S10). In addition, we analyzed two mutants with impaired enzymatic activities: an RNA helicase–deficient mutant (AAA) and a mutant in which both helicase and ATPase activity are inhibited (DQAD) (33). When injected in Xenopus animal caps as mRNA, both mutants cooperated with Wnt3a in siamois induction (fig. S8, D and E), again demonstrating that DDX3 enzymatic activity is dispensable for Wnt–β-catenin signaling. These mutations also had no effect on the ability of the protein to colocalize with Dvl2 in transfected cells (fig. S8, F and G).

Because the data indicated that DDX3 promotes phosphorylation of Dvl2 by CK1ε, we asked if this effect is direct. In in vitro kinase assays with recombinant proteins, DDX3 greatly stimulated CK1ε phosphorylation of Dvl2, and DDX3 itself also became phosphorylated (Fig. 4A). Moreover, DDX3 strongly enhanced CK1ε activity toward a generic CK1-specific peptide substrate (Fig. 4B).

Fig. 4

DDX3 and other DEAD-box RNA helicases directly activate CK1 family members. (A) CK1ε in vitro kinase assay with recombinant Dvl2 as substrate in the absence or presence of recombinant DDX3. Autoradiography after SDS–page polyacrylamide gel electrophoresis showing 32P incorporation. (B to D) In vitro kinase assays with the indicated recombinant kinases and CK1- or GSK3-specific peptide substrate in the absence or presence of DDX3. (E and F) In vitro kinase assays of endogenous CK1ε (E) and CK1α (F) coimmunoprecipitated from lysates of HEK293T that were treated with the indicated siRNAs and stimulated with control or Wnt3a-conditioned medium. Kinase activity was normalized to total kinase (Western blot, lower panel). The experiment was performed three times with similar results. (G) CK1ε in vitro kinase assay with CK1 peptide substrate in the absence or presence of the indicated recombinant proteins. Error bars indicate SDs; n = 3 assays. cpm, counts per minute.

CK1ε is an autorepressed enzyme because its autophosphorylated C-terminal domain interacts with and inhibits the kinase (34, 35). This autoinhibition can be relieved in vitro by protein phosphatases (36). However, the phosphatase inhibitor okadaic acid did not change the stimulatory effect of DDX3 on CK1ε (fig. S11A). Moreover, a CK1ε deletion mutant lacking the autoinhibitory tail (Flag-CK1εΔC) had increased kinase activity, as expected, but was even more stimulated by DDX3 (Fig. 4C) and still bound to DDX3 (fig. S11B). In addition, DDX3 had no influence on the autophosphorylation of full-length or mutant CK1ε (fig. S11, C and D). Moreover, other CK1 isoforms, which are not autorepressed, were also stimulated by DDX3 (see below). Hence, DDX3 activates CK1ε independent of its C-terminal autophosphorylation.

Instead, kinetic analysis revealed a novel mechanism of CK1 activation: Although DDX3 had little effect on the Km (apparent Michaelis constant) of CK1ε toward its peptide substrate, it greatly improved the Km of the kinase toward ATP, from 108 to 24 μM (fig. S12). This was accompanied by a ~15-fold increased Vmax. Thus, DDX3 acts as a V-type allosteric activator of CK1ε, a regulatory mode typically found in protein kinases and small GTPases involved in signal transduction (37).

To delineate the domains required for CK1ε stimulation, we tested immunopurified DDX3 deletion mutants in kinase assays. Whereas DDX3425-662 and DDX3456-662 stimulated CK1ε activity, DDX3508-662 showed no activity (fig. S9E), reflecting their CK1ε binding and Wnt signaling activities.

Because CK1 family members share a highly conserved kinase domain, we examined whether DDX3 activates other CK1 isoforms as well. Indeed, DDX3 stimulated not only CK1ε but also the three CK1 isoforms α, δ, and γ2, whereas GSK3β activity was unaffected (Fig. 4D). To corroborate that DDX3 regulates CK1 activity in vivo, we carried out kinase assays with CK1ε, CK1α, and CK1γ immunoisolated from cells after DDX3 siRNA knockdown. Wnt treatment was reported to stimulate CK1ε kinase activity (31), and we could confirm this (Fig. 4E). Notably, siDDX3 reduced CK1ε kinase activity in unstimulated as well as in Wnt3a-stimulated cells, indicating that DDX3 regulates CK1ε activity in vivo. siDDX3 also reduced activity of CK1α, which negatively acts in Wnt signaling (3), but upon Wnt stimulation, siDDX3 ceased to reduce CK1α activity (Fig. 4F), suggesting that Wnt signaling uncouples CK1α regulation by DDX3. Unlike CK1ε and CK1α, CK1γ activity was not inhibited by siDDX3 (fig. S13A). To corroborate that DDX3 promotes Wnt signaling primarily by activation of CK1ε instead of CK1γ, we analyzed constitutively active LRP6ΔE1-4, for which T1479 phosphorylation bypasses the need for Wnt and Dvl (28). Although LRP6ΔE1-4 T1479 phosphorylation was, as expected, reduced by dominant negative CK1γ, it remained unaffected by siDDX3, siDvl, or siCK1ε (fig. S13B).

Finally, we tested other DDX family members in CK1ε kinase assays (Fig. 4G). Notably, DDX4 and DDX56 also potently stimulated CK1ε kinase, whereas other DDX proteins had little or no effect. No stimulation was observed also with DHX58 and -32 RNA helicases, DNA helicases XRCC5 and 6 and Fubp1, as well as the GTP-ase ARFRP1. Together the results raise the possibility that other DDX members also function as regulatory subunits of the CK1 family.

Our study reveals an unexpected biological and cellular activity of the RNA helicase DDX3 in Wnt–β-catenin signaling and shows that this multifunctional protein is a regulatory subunit of CK1ε. We propose a model wherein Wnt activation promotes recruitment of DDX3 to CK1ε, which stimulates the kinase to phosphorylate and regulate target proteins, including Dvl. This in turn promotes clustering of Wnt-receptor complexes on Dvl aggregates and formation of signalosomes. Intriguingly, DDX3 is recurrently mutated in medulloblastomas belonging to a “Wnt subgroup” (3840), which may be rationalized by our findings. Although DDX3 regulates Wnt–β-catenin signaling in Xenopus development, C. elegans, and mammalian cells, it may not universally activate CK1.

That DDX3 regulates CK1ε is surprising given the long-held view that CK1 family members are constitutively active and that specificity is regulated by substrate priming and enzyme-substrate proximity. Our results indicate that the kinase-stimulatory function may well extend to other DDX members and likewise not be limited to CK1ε, opening fresh perspectives for one of the longest-studied protein kinase families.

Supplementary Materials

www.sciencemag.org/cgi/content/full/science.1231499/DC1

Materials and Methods

Figs. S1 to S13

References

References and Notes

  1. Acknowledgments: We thank M. Bienz, H. Clevers, X. He, R. Moon, R. Nusse, Y. H. Wu Lee, and D. Wu for reagents; D. Ingelfinger and M. Boutros for assistance with siRNA screening; and R. Voit for help with kinase assays. We thank G. Roth and Aska Pharmaceuticals Tokyo for generous supply of human chorionic gonadotropin. This work was supported by the Deutsche Forschungsgemeinschaft.
View Abstract

Navigate This Article