Encyclopedia of Signaling Molecules

2018 Edition
| Editors: Sangdun Choi


  • Chotirat Rattanasinchai
  • Jian Chen
  • Kathleen A. Gallo
Reference work entry
DOI: https://doi.org/10.1007/978-3-319-67199-4_247


Historical Background

The mixed-lineage kinases are so named for their sequence similarity to both tyrosine kinases and serine/threonine kinases. However, based on biochemical assays, only serine/threonine kinase activity has been demonstrated for the MLKs. The three subfamilies of MLKs reside within the “tyrosine kinase-like” branch of the human kinome. Members of the MLK subfamily, comprised of MLK1-4, share a conserved domain arrangement and 75% identity within their catalytic domains (Fig. 1). The dual leucine zipper–bearing kinase (DLK) subgroup of MLKs, which includes DLK/ZPK/MUK and LZK, is characterized by a kinase catalytic domain followed by two leucine zipper motifs. A third subgroup of MLKs represented by ZAK/MLTK contains both a leucine zipper motif and a sterile-alpha motif. MLK3 has emerged as the paradigm for the MLK subfamily. There is no ortholog for MLK3 in yeast but in Drosophila, the MLK1-4 ortholog, Slipper, is critical for the cell sheet movement during dorsal closure in the fly embryo, which involves activation of the JNK pathway. MLK3-deficient mice are fully viable but have a reduced thickness of the dorsal epidermal tissue, which parallels the effects of disruption of Slipper in Drosophila (reviewed in Gallo and Johnson 2002).
MLK3, Fig. 1

Conserved domains in the MLK subfamily. The domain arrangement within the MLK subfamily, depicting the relative positions of the Src-homology-3 (SH3) kinase, leucine-zipper (LZ), and Cdc42/Rac1 interactive binding (CRIB) motifs. The number of amino acids in each kinase is shown (See text for details)

Regulation of MLK3 Activity

MLK3 contains several protein interaction domains that are important in regulating its activity. A short glycine-rich region is followed sequentially by a Src homology 3 (SH3) domain, a kinase catalytic domain, a leucine zipper region, and a Cdc42/Rac Interactive Binding (CRIB) motif (Fig. 1). The carboxyl terminal region of MLK3 is rich in proline, serine, and threonine residues (Gallo and Johnson 2002). Like many protein kinases, MLK3 activation involves phosphorylation within the activation loop of the kinase domain. Based on site-directed mutagenesis studies, Thr 277 and Ser 281 in the activation loop act as positive regulatory phosphorylation sites; and phospho-specific antibodies directed against these sites are widely used to monitor MLK3 activity. Notably these potential phosphorylation sites are conserved within the activation loops of MLK1-4, suggesting that they may serve a similar function. Leucine zippers form coiled coil dimers that are stabilized by the interaction of leucine or other nonaromatic aliphatic residues at the interface of the helices. Deletion of the entire zipper or introduction of a helix disrupting Pro residue for one of the conserved Leu residues prevents dimerization and activation loop phosphorylation of MLK3. Thus, leucine zipper–mediated dimerization can lead to MLK3 activation. These findings are consistent with the high MLK3 activity observed upon overexpression in many cultured cell lines. Presumably when MLK3 is expressed at high levels, a portion of MLK3 is dimerized in the absence of a physiologically appropriate stimulus, leading to elevated MLK3 activity.

MLK3 contains a CRIB motif, a short conserved sequence required for binding the Rho family GTPases, Cdc42 and Rac. Subsequent work demonstrated that the small GTPases, Cdc42 and Rac, are indeed capable of activating MLK3. Like other GTPase effector proteins, MLK3 interacts with the active, GTP-bound GTPase, but not the inactive, GDP-bound form. Binding of activated Cdc42 (or Rac) promotes MLK3 dimerization, activates loop phosphorylation, increases MLK3 catalytic activity, and translocates MLK3 to the cell periphery (reviewed in Rattanasinchai and Gallo 2016). Posttranslational COOH-terminal prenylation (geranylgeranylation) of Cdc42 and Rac allows for membrane targeting. A prenylation-defective site-directed mutant of activated Cdc42 retains the ability to bind MLK3 and promotes activation loop phosphorylation of MLK3, but fails to translocate MLK3 to membranes. Thus the physical interaction between the activated GTPase and MLK3 is the key mechanism by which it activates MLK3. However, membrane targeting by activated, prenylated Cdc42 is accompanied by additional phosphorylation events on MLK3 and further enhances MLK3 in vitro kinase and cellular signaling activities.

SH3 domains are modular domains of about 60 amino acids that typically bind proline-rich sequences to interact with intracellular signaling partners. The N-terminal SH3 domain of MLK3 functions as an autoinhibitory domain (reviewed in Gallo and Johnson 2002). A single Pro residue located between the leucine zipper and CRIB motif is required for the interaction with its SH3 domain. Though not formally shown, this autoinhibitory interaction is presumed to be intramolecular. Since this Pro residue is conserved in MLK1-4, it is likely that these MLKs are also regulated by SH3-mediated autoinhibition.

An integrated model for MLK3 activation by Cdc42 and Rac consistent with available data is shown in Fig. 2. Binding of the active, GTP-bound GTPase to MLK3 in a region containing the CRIB motif disrupts the SH3 autoinhibitory interaction, promoting leucine zipper–mediated dimerization and transautophosphorylation within the activation loop, yielding the active kinase at cell membranes (reviewed in Gallo and Johnson 2002).
MLK3, Fig. 2

Model for MLK3 activation by small GTPases, Cdc42 and Rac. Autoinhibition of MLK3 is maintained by an interaction between SH3 domain of MLK3 and a proline-containing sequence located between the LZ region and CRIB motif. The binding of GTP-bound GTPase(s) through the CRIB motif disrupts the autoinhibition, promotes dimerization, and transautophosphorylation at Thr 277 and Ser 281 ensues

MLK3 stability is regulated by protein-protein interactions. Heat shock protein (HSP) 90, a molecular chaperone that facilitates cancer progression, along with its cochaperone p50cdc37, binds MLK3 and is required for MLK3 stability and function (reviewed in Rattanasinchai and Gallo 2016), whereas interaction with Merlin, a tumor suppressor protein encoded by Neurofibromatosis type 2 (NF2) gene, inhibits MLK3 activity, apparently by disrupting the interaction between MLK3 and Cdc42. While MLK3 is stabilized by binding HSP90/cdc37, MLK3 also interacts with an E3 ligase of Hsc-70-interacting protein (CHIP) together with E2 ubiquitin-conjugating enzyme UbcH5 which leads to ubiquitination and subsequent targeting of MLK3 to proteasomal degradation (reviewed in Rattanasinchai and Gallo 2016).

In addition to phosphorylation within the activation loop, numerous MLK3 phosphorylation sites have been identified through mass spectrometry, the majority of which are Ser or Thr residues followed immediately by Pro residues, conforming to the consensus sequence for proline-directed kinases. The downstream MAPK, JNK, has been shown to phosphorylate MLK3 at multiple proline-directed kinase sites within the COOH terminal region in a positive feedback loop that stabilizes MLK3 and/or redistributes MLK3 into Triton-soluble cellular fractions. Ser 674 has been identified as an Akt(PKB)-mediated phosphorylation site on MLK3 that inhibits MLK3-mediated apoptosis (reviewed in Handley et al. 2007). In contrast, Ser 789 and Ser 793 have been identified as sites for GSK-3 beta-mediated phosphorylation on MLK3 that promote neuronal cell death (Mishra et al. 2007).

MLK3 Signaling Triggered by Cell-Surface Receptors

MLK3 signals through multiple mitogen-activated protein kinase (MAPK) pathways, including JNK, p38 MAPK, and ERK. MAPK pathways are three-tiered cascades wherein a MAP3K phosphorylates and activates a MAP2K which, in turn, phosphorylates and activates the terminal MAPK. Activated MAPKs have both nuclear and cytosolic substrates. MLK3, also known as MAP3K11, activates the JNK pathway by phosphorylation-mediated activation of MKK4/MKK7 and activates the p38 MAPK pathway through phosphorylation-mediated activation of MKK3/MKK6. Whereas catalytic activity of MLK3 is required for activation of the JNK and p38 MAPK pathways, MLK3 acts as an indispensable scaffold required for B-Raf-mediated ERK activation (reviewed in Kyriakis 2007). Scaffold proteins, like the JIPs and POSH, are able to bind MLK3 as well as downstream MKKs and MAPKs, and may serve to localize, organize, or assemble MLK3 pathway complexes at specific subcellular locations (reviewed in Dhanasekaran et al. 2007).

Substantial evidence indicates that MLK3 can signal through multiple receptor types, including receptor tyrosine kinases, cytokine receptors, and G-protein-coupled receptors (GPCRs) (Fig. 3), but the details of these signaling pathways are largely unknown. The cytokine, tumor necrosis factor-alpha (TNF), is the best-described stimulus for activating MLK3 signaling to JNK. The TNF receptor-associated factor 2 (TRAF2) complexes with both the TNF receptor and MLK3 and is critical for TNF-induced JNK activation (reviewed in Rattanasinchai and Gallo 2016). It should be noted that other MAP3Ks, such as ASK, have also been implicated in TNF-induced JNK activation. Receptor tyrosine kinase signaling also involves MLK3, since silencing of MLK3 prevents epidermal growth factor (EGF)-induced activation of JNK, p38 MAPK, and ERK (reviewed in Kyriakis 2007). MLK3 can be activated by stimulation of cells with carbachol, a chemical ligand for G-protein-coupled acetylcholine receptors (Swenson-Fields et al. 2008). In this context, MLK3 complexes with the Rho guanine nucleotide exchange factor (GEF), p63RhoGEF, to inhibit RhoA activation.
MLK3, Fig. 3

Role of MLK3 in signaling through cell surface receptors to MAPKs. MLK3-mediated signaling pathways downstream of various receptors are shown. EGFR epidermal growth factor receptor, TNFR tumor necrosis factor receptor, GPCR G-protein-coupled receptor. Activated MAPKs can phosphorylate cytosolic substrates or enter the nucleus to phosphorylate nuclear substrates, including transcription factors that regulate gene expression (See text for details)

MLK3 in Cancer

Deregulation of signal transduction pathways drives the development of human malignancies. MLK3 has been shown to be overexpressed in breast cancer cell lines compared with nontumorigenic mammary epithelial cells, suggesting that MLK3 might contribute to acquisition of malignant phenotypes in breast cancer (Chen et al. 2010). Ectopic expression of wild-type MLK3 causes cellular transformation of immortalized fibroblasts (reviewed in Kyriakis 2007) and promotes a malignant phenotype of mammary epithelial spheroids in 3D culture (Chen et al. 2010). MLK3 silencing or inhibition can inhibit proliferation in some, but not all, tumor cell lines, perhaps depending upon the oncogenic signaling signature in those cells ( reviewed in Kyriakis 2007).

Several lines of evidence demonstrate a critical role of MLK3 in migration and invasion of cancer cells of epithelial origin (reviewed in Chadee 2013). Induced expression of MLK3 promotes migration of poorly invasive breast cancer cells and invasion of mammary epithelial cells (Chen et al. 2010). Small interfering RNA-mediated silencing of MLK3 blocks migration of highly invasive breast, lung, and gastric carcinoma cells, indicating an essential function for MLK3 in migration and/or invasion of a broad array of epithelial-derived tumor cells (Chen et al. 2010; Swenson-Fields et al. 2008). A major mechanism by which MLK3 controls migration and invasion is through activation of JNK and its downstream transcription factor AP-1, leading to the expression of genes that promote invasion or epithelial-to-mesenchymal transition, such as MMP-7, fra-1, vimentin, and N-cadherin (Chen et al. 2010; Shintani et al. 2008). A kinase-independent role for MLK3 in blocking activation of RhoA, through sequestration of the RhoA-specific guanine nucleotide exchange factor (GEF), p63RhoGEF, has been identified in lung carcinoma cells (Swenson-Fields et al. 2008). More recent studies demonstrated that MLK3 limits RhoA activity in breast cancer cells in a kinase-dependent manner, through JNK-mediated phosphorylation of paxillin (Chen and Gallo 2012). The inactivation of the neurofibromatosis-2 (NF-2) tumor suppressor gene is associated with the formation of benign brain tumors. NF-2/merlin can interact with MLK3 to inhibit its activity, preventing proliferation and invasion of Schwann cells (reviewed in Chadee 2013).

Missense mutations in mlk3 have been identified in human gastrointestinal cancers and are significantly associated with microsatellite instability (MSI) phenotype in mismatch repair-deficient gastrointestinal carcinomas. The identified mutations are found in different functional domains of MLK3, including the SH3 domain (Y99C), the kinase domain (A165S; P252H), and the COOH-terminal proline-rich region (R799C; P840L). When ectopically expressed in fibroblasts, these mutant forms of MLK3 are more transforming and tumorigenic than wild-type MLK3. Microarray analysis reveals that ectopic expression of MLK3 P252H affects several signaling pathways associated with colorectal cancer including WNT, MAPK, NOTCH, TGF-β, and P53 (Velho et al. 2014).

MLK1-4 have been linked to acquired B-Raf inhibitor resistance of melanomas harboring B-RAF V600E mutation. In this context, MLK1-4 are able to directly phosphorylate MEK and reactivate ERK signaling in a B-RAF-independent manner, thus resulting in survival of melanoma cells in the presence of B-RAF inhibitors (reviewed in Rattanasinchai and Gallo 2016).

MLK3 in Neurodegenerative Diseases

Numerous neurotoxic insults can induce JNK activation and mitochondria-mediated apoptosis. MLK3 has been implicated as an upstream mediator of JNK in neuronal cell death in several experimental systems, primarily based on the use of dominant negative forms of MLK3 and the pan-MLK inhibitors, CEP-1347 and CEP-11004. For example, deprivation of nerve growth factor (NGF) induces JNK activation and apoptosis in PC12 cells and in cultured superior cervical rat ganglia, which can be attenuated by blocking MLK activity (reviewed in Wang et al. 2004). Studies in rodents support a role for MLK3 in apoptosis in response to cerebral ischemia-reperfusion, a model for stroke (Zhang et al. 2009). MLK3 has also been linked to JNK activation in kainate-induced neurotoxicity, with the scaffolding protein PSD-95 interacting with both the GluR6 receptor and MLK3 (reviewed in Gallo and Johnson 2002). Finally, substantial evidence has accumulated for MLK3 in JNK activation and neuronal death in cell-based and in vivo models of Parkinson’s disease that use the dopaminergic selective neurotoxin MPTP or its derivative MPP+, respectively (reviewed in Wang et al. 2004). In response to treatment with MPTP/MPP+, the MLK inhibitor CEP-1347 was able to suppress JNK activation and increase survival of dopaminergic neurons. These studies suggested that CEP-1347 might be a promising therapeutic for treating patients with Parkinson’s disease. CEP-1347 ultimately progressed to Stage II/III clinical trials, but failed to delay progression of patients with early stage Parkinson’s disease (Parkinson Study Group PRECEPT Investigators 2007).

MLK3 in Metabolic Dysfunction and Inflammation

Obesity is associated with metabolic dysfunction. Mouse models of high fat diet-induced obesity have revealed that saturated free fatty acids (FFAs) induce metabolic stress which eventually leads to insulin resistance. MLK3 has been implicated in diet-induced metabolic dysfunction (reviewed in Craige et al. 2016). Mechanistically, FFAs from a high-fat diet stimulate JNK1 activation resulting in a cascade of metabolic stress response. This process is mediated by MLK3 as loss of MLK3 or its downstream effectors MKK4/7 prevents FFA-stimulated JNK activation. Depletion of MLK2 and MLK3, which are both ubiquitously expressed in peripheral insulin-responsive tissues, protects against high fat diet-induced insulin resistance in this animal model.

In addition to insulin resistance, MLK3 deletion is associated with the progression of liver injury in diet-induced nonalcoholic steatohepatitis (NASH) (reviewed in Craige et al. 2016). In a mouse model of high fat, high carbohydrate (HFHC) diet-induced NASH, MLK3-deficient mice fed with HFHC show reduction of markers for liver injury including hepatocyte apoptosis, inflammation, and fibrogenesis. Thus, MLK3-JNK signaling appears to be crucial for diet-induced metabolic dysfunctions and MLK3 could also be a potential target for the treatment of diet-induced insulin resistance and NASH.

MLK3 signaling is important in the production of proinflammatory proteins. In the context of interferon-gamma-activated macrophages, MLK signals to p38 MAPK increasing the mRNA levels of TNF-α and interferon inducible protein 10, presumably through increased mRNA stability. In microglia activated by the bacterial endotoxin, lipopolysaccharide (LPS), an MLK or JNK inhibitor reduces AP-1-mediated transcription of TNF-α (reviewed in Handley et al. 2007). In primary cortical astrocytes, activated by a mixture of proinflammatory cytokines, MLK signaling to both p38 MAPK and JNK has been implicated in the induction of inflammation-responsive genes (Falsig et al. 2004).

Thus, there is considerable interest in MLK inhibitors as anti-inflammatory drugs, particularly in attenuating neuroinflammation associated with HIV infection. Recently, the MLK inhibitor, CEP-1347, was shown to prevent the production of cytokines and chemokines in HIV-infected human macrophages and to elicit anti-inflammatory and neuroprotective effects in mouse models of HIV-1 encephalitis (Eggert et al. 2010). A newly developed MLK3 inhibitor URMC-099 showed efficacy in preclinical studies as a neuroprotective therapeutic agent in rodent models of HIV-associated neurocognitive disorders (HAND) (Marker et al. 2013). Furthermore, a combination of nanoformulated antiviral therapy with URMC-099 demonstrated a better outcome than either drug alone in reducing viral load and the number of HIV-1 infected CD4+ T-cells in lymphoid tissues in infected humanized mice (Zhang et al. 2016).


MLK3 is an intracellular serine/threonine kinase, which belongs to a larger family of related kinases that are evolutionarily conserved in metazoans. The kinase activity of MLK3 is autoinhibited through its SH3 domain. The small GTPases, Cdc42 and Rac, can bind to MLK3, disrupting autoinhibition and promoting zipper-mediated dimerization and subsequent transphosphorylation within the kinase domain to yield active MLK3. MLK3 signals through multiple cell surface receptors, including receptor tyrosine kinases, cytokine receptors, and heterotrimeric G-protein-coupled receptors. MLK3 contributes to the activation of multiple MAPK pathways, functioning as a MAP3K to activate the JNK and p38 MAPK pathways, and as a scaffold in B-Raf-mediated ERK activation. Although soon after its identification, MLK3 was implicated in neuronal apoptosis, emerging data indicates that MLK3 is a critical player in many pathophysiological processes, including cancer metastasis, metabolic dysfunction, and inflammation. Thus MLK inhibitors may prove valuable in multiple therapeutic arenas.


  1. Chadee DN. Involvement of mixed lineage kinase 3 in cancer. Can J Physiol Pharmacol. 2013;91:268–74.PubMedCrossRefGoogle Scholar
  2. Chen J, Miller EM, Gallo KA. MLK3 is critical for breast cancer cell migration and promotes a malignant phenotype in mammary epithelial cells. Oncogene. 2010;29:4399–411.PubMedCrossRefGoogle Scholar
  3. Chen J, Gallo KA. MLK3 regulates paxillin phosphorylation in chemokine-mediated breast cancer cell migration and invasion to drive metastasis. Cancer Res. 2012;72:4130–40.PubMedCrossRefGoogle Scholar
  4. Craige SM, Reif MM, Kant S. Mixed-lineage kinases (MLKs) in inflammation, metabolism, and other disease states. Biochim Biophs Acta – Mol Basis Dis. 2016;1862:1581–6.CrossRefGoogle Scholar
  5. Dhanasekaran DN, Kashef K, Lee CM, Xu H, Reddy EP. Scaffold proteins of MAP-kinase modules. Oncogene. 2007;26:3185–202.PubMedCrossRefGoogle Scholar
  6. Eggert D, Dash PK, Gorantla S, Dou H, Schifitto G, Maggirwar SB, Dewhurst S, Poluektova L, Gelbard HA, Gendelman HE. Neuroprotective activities of CEP-1347 in models of neuroAIDS. J Immunol. 2010;184:746–56.PubMedCrossRefGoogle Scholar
  7. Falsig J, Pörzgen P, Lotharius J, Leist M. Specific modulation of astrocyte inflammation by inhibition of mixed lineage kinases with CEP-1347. J Immunol. 2004;173:2762–70.PubMedCrossRefGoogle Scholar
  8. Gallo KA, Johnson GL. Mixed-lineage kinase control of JNK and p38 MAPK pathways. Nat Rev Mol Cell Biol. 2002;3:663–72.PubMedCrossRefGoogle Scholar
  9. Handley ME, Rasaiyaah J, Chain BM, Katz DR. Mixed lineage kinases (MLKs): a role in dendritic cells, inflammation and immunity? Int J Exp Pathol. 2007;88:111–26.PubMedPubMedCentralCrossRefGoogle Scholar
  10. Kyriakis JM. The integration of signaling by multiprotein complexes containing Raf kinases. Biochim Biophys Acta. 2007;1773:1238–47.PubMedCrossRefGoogle Scholar
  11. Marker DF, Tremblay ME, Puccini JM, Barbieri J, Gantz Marker MA, Loweth CJ, Muly EC, Lu SM, Goodfellow VS, Dewhurst S, Gelbard HA. The new small-molecule mixed-lineage kinase 3 inhibitor URMC-099 is neuroprotective and anti-inflammatory in models of human immunodeficiency virus-associated neurocognitive disorders. J Neurosci. 2013;33:9998–10010.PubMedPubMedCentralCrossRefGoogle Scholar
  12. Mishra R, Barthwal MK, Sondarva G, Rana B, Wong L, Chatterjee M, Woodgett JR, Rana A. Glycogen synthase kinase-3beta induces neuronal cell death via direct phosphorylation of mixed lineage kinase 3. J Biol Chem. 2007;282:30393–405.PubMedPubMedCentralCrossRefGoogle Scholar
  13. Parkinson Study Group PRECEPT Investigators. Mixed lineage kinase inhibitor CEP-1347 fails to delay disability in early Parkinson disease. Neurology. 2007;69:1480–90.CrossRefGoogle Scholar
  14. Rattanasinchai C, Gallo KA. MLK3 signaling in cancer invasion. Cancers. 2016;8(5):51. Review.PubMedCentralCrossRefGoogle Scholar
  15. Shintani Y, Fukumoto Y, Chaika N, Svoboda R, Wheelock MJ, Johnson KR. Collagen I-mediated up-regulation of N-cadherin requires cooperative signals from integrins and discoidin domain receptor 1. J Cell Biol. 2008;180:1277–89.PubMedPubMedCentralCrossRefGoogle Scholar
  16. Swenson-Fields KI, Sandquist JC, Rossol-Allison J, Blat IC, Wennerberg K, Burridge K, Means AR. MLK3 limits activated Galphaq signaling to Rho by binding to p63RhoGEF. Mol Cell. 2008;32:43–56.PubMedPubMedCentralCrossRefGoogle Scholar
  17. Velho S, Pinto A, Licasto D, Oliveira MJ, Sousa F, Stupka E, Seruca R. Dissecting the signaling pathways associated with the oncogenic activity of MLK3 P252H mutation. BMC Cancer. 2014;14:182.PubMedPubMedCentralCrossRefGoogle Scholar
  18. Wang LH, Besirli CG, Johnson Jr EM. Mixed-lineage kinases: a target for the prevention of neurodegeneration. Annu Rev Pharmacol Toxicol. 2004;44:451–74.PubMedCrossRefGoogle Scholar
  19. Zhang Q-G, Wang R, Hana D, Dong Y, Branna DW. Role of Rac1 GTPase in JNK signaling and delayed neuronal cell death following global cerebral ischemia. Brain Res. 2009;1265:138–47.PubMedPubMedCentralCrossRefGoogle Scholar
  20. Zhang G, Guo D, Dash PK, Arainga M, Wiederin JL, Haverland NA, Knibbe-Hollinger J, Martinez-Skinner A, Ciborowski P, Goodfellow VS, Wysocki TA, Wysocki BJ, Poluektova LY, Liu XM, McMillan JM, Gorantla S, Gelbard HA, Gendelman HE. The mixed lineage kinase-3 inhibitor URMC-099 improves therapeutic outcomes for long-acting antiretroviral therapy. Nanomedicine. 2016;12:109–22.PubMedCrossRefGoogle Scholar

Copyright information

© Springer International Publishing AG 2018

Authors and Affiliations

  • Chotirat Rattanasinchai
    • 1
    • 2
  • Jian Chen
    • 3
    • 4
  • Kathleen A. Gallo
    • 1
    • 2
  1. 1.Cell and Molecular Biology ProgramMichigan State UniversityEast LansingUSA
  2. 2.Department of PhysiologyMichigan State UniversityEast LansingUSA
  3. 3.Department of Biochemistry and Molecular BiologyMichigan State UniversityEast LansingUSA
  4. 4.Department of MedicineDuke UniversityDurhamUSA