Donald M Coen

Harvard University, Cambridge, Massachusetts, United States

Are you Donald M Coen?

Claim your profile

Publications (215)1318.67 Total impact

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: A collection of genomic DNA sequences of herpes simplex virus (HSV) strains has been defined and analyzed, and some information is available about genomic stability upon limited passage of viruses in culture. The nature of genomic change upon extensive laboratory passage remains to be determined. In this report we review the history of the HSV-1 KOS laboratory strain and the related KOS1.1 laboratory sub-strain, also called KOS (M), and determine the complete genomic sequence of an early passage stock of the KOS laboratory sub-strain and a laboratory stock of the KOS1.1 sub-strain. The genomes of the two sub-strains are highly similar with only five coding changes, 20 non-coding changes, and about twenty non-ORF sequence changes. The coding changes could potentially explain the KOS1.1 phenotypic properties of increased replication at high temperature and reduced neuroinvasiveness. The study also provides sequence markers to define the provenance of specific laboratory KOS virus stocks.
    Full-text · Article · Jan 2016 · Virology
  • [Show abstract] [Hide abstract]
    ABSTRACT: Herpesvirus nucleocapsids escape from the nucleus in a process orchestrated by a highly conserved, viral nuclear egress complex. In human cytomegalovirus, the complex consists of two proteins, UL50 and UL53. We solved structures of versions of UL53 and the complex by X-ray crystallography. The UL53 structures, determined at 1.93 and 3.0 Å resolution, contained unexpected features including a Bergerat fold resembling that found in certain nucleotide-binding proteins, and a Cys3His zinc finger. Substitutions of zinc-coordinating residues decreased UL50-UL53 co-localization in transfected cells, and, when incorporated into the HCMV genome, ablated viral replication. The structure of the complex, determined at 2.47 Å resolution, revealed a mechanism of heterodimerization in which UL50 clamps onto helices of UL53 like a vise. Substitutions of particular residues on the interaction interface disrupted UL50-UL53 co-localization in transfected cells and abolished virus production. The structures and the identification of contacts can be harnessed toward the rational design of novel and highly specific antiviral drugs and will aid in the detailed understanding of nuclear egress.
    No preview · Article · Oct 2015 · The EMBO Journal
  • Mayuri Sharma · Jeremy P. Kamil · Donald M. Coen
    [Show abstract] [Hide abstract]
    ABSTRACT: Herpesviruses, like most DNA viruses, replicate their genomes in the host cell nucleus. Their DNA is then packaged and assembled into viral nucleocapsids, which, in most cases, are too large to pass through the nuclear pore complex. Instead, herpesviruses use a complex multistep pathway, termed nuclear egress, to exit the nucleus. Key players in this process include two conserved viral proteins that form the nuclear egress complex (NEC). In human cytomegalovirus, these NEC proteins are UL50, embedded in the inner nuclear membrane, and its nucleoplasmic partner UL53. Both are essential for viral nuclear egress. However, other viral components as well as host nuclear envelope proteins may also participate in nuclear egress. Identifying these viral and cellular factors may provide important insight into the herpesvirus lifecycle and its relationship to the underlying, yet still-mysterious, host nuclear egress pathway. We developed an immunoprecipitation-based protocol, described herein, to identify protein-protein interactions involving the NEC from the nuclear fraction of infected cells that express an epitope-tagged version of NEC subunit UL53.
    No preview · Chapter · Sep 2015
  • [Show abstract] [Hide abstract]
    ABSTRACT: Herpesviruses require a nuclear egress complex (NEC) for efficient transit of nucleocapsids from the nucleus to the cytoplasm. The NEC orchestrates multiple steps during herpesvirus nuclear egress, including disruption of nuclear lamina and particle budding through the inner nuclear membrane. In the important human pathogen human cytomegalovirus (HCMV), this complex consists of nuclear membrane protein UL50, and nucleoplasmic protein UL53, which is recruited to the nuclear membrane through its interaction with UL50. Here, we present an NMR-determined solution-state structure of the murine CMV homolog of UL50 (M50; residues 1-168) with a strikingly intricate protein fold that is matched by no other known protein folds in its entirety. Using NMR methods, we mapped the interaction of M50 with a highly conserved UL53-derived peptide, corresponding to a segment that is required for heterodimerization. The UL53 peptide binding site mapped onto an M50 surface groove, which harbors a large cavity. Point mutations of UL50 residues corresponding to surface residues in the characterized M50 heterodimerization interface substantially decreased UL50-UL53 binding in vitro, eliminated UL50-UL53 colocalization, prevented disruption of nuclear lamina, and halted productive virus replication in HCMV-infected cells. Our results provide detailed structural information on a key protein-protein interaction involved in nuclear egress and suggest that NEC subunit interactions can be an attractive drug target.
    No preview · Article · Jul 2015 · Proceedings of the National Academy of Sciences
  • Source
    Han Chen · G Peter Beardsley · Donald M Coen
    [Show abstract] [Hide abstract]
    ABSTRACT: Many antiviral and anticancer drugs are nucleoside analogs that target polymerases and cause DNA chain termination. Interestingly, ganciclovir (GCV), the first line of therapy for human cytomegalovirus (HCMV) infections, induces chain termination despite containing the equivalent of a 3'-hydroxyl group. Certain HCMV GCV resistance (GCV(r)) mutations, including ones associated with treatment failures, result in substitutions in the 3'-5' exonuclease (Exo) domain of the catalytic subunit of the viral DNA polymerase (Pol). To investigate how these mutations confer resistance, we overexpressed and purified wild-type (WT) HCMV Pol and three GCV(r) Exo mutants. Kinetic studies provided little support for resistance being due to effects on Pol binding or incorporation of GCV-triphosphate. The mutants were defective for Exo activity on all primer templates tested, including those with primers terminating with GCV, arguing against the mutations increasing excision of the incorporated drug. However, although the WT enzyme terminated DNA synthesis after incorporation of GCV-triphosphate and an additional nucleotide (N+1), the Exo mutants could efficiently synthesize DNA to the end of such primer templates. Notably, the Exo activity of WT Pol rapidly and efficiently degraded N+2 primer templates to N+1 products that were not further degraded. On N+1 primer templates, WT Pol, much more than the Exo mutants, converted the incoming deoxynucleoside triphosphate to its monophosphate, indicative of rapid addition and removal of incorporated nucleotides ("idling"). These results explain how GCV induces chain termination and elucidate a previously unidentified mechanism of antiviral drug resistance.
    Preview · Article · Nov 2014 · Proceedings of the National Academy of Sciences
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Unlabelled: Herpesvirus nucleocapsids exit the host cell nucleus in an unusual process known as nuclear egress. The human cytomegalovirus (HCMV) UL97 protein kinase is required for efficient nuclear egress, which can be explained by its phosphorylation of the nuclear lamina component lamin A/C, which disrupts the nuclear lamina. We found that a dominant negative lamin A/C mutant complemented the replication defect of a virus lacking UL97 in dividing cells, validating this explanation. However, as complementation was incomplete, we investigated whether the HCMV nuclear egress complex (NEC) subunits UL50 and UL53, which are required for nuclear egress and recruit UL97 to the nuclear rim, are UL97 substrates. Using mass spectrometry, we detected UL97-dependent phosphorylation of UL50 residue S216 (UL50-S216) and UL53-S19 in infected cells. Moreover, UL53-S19 was specifically phosphorylated by UL97 in vitro. Notably, treatment of infected cells with the UL97 inhibitor maribavir or infection with a UL97 mutant led to a punctate rather than a continuous distribution of the NEC at the nuclear rim. Alanine substitutions in both UL50-S216 and UL53-S19 resulted in a punctate distribution of the NEC in infected cells and also decreased virus production and nuclear egress in the absence of maribavir. These results indicate that UL97 phosphorylates the NEC and suggest that this phosphorylation modulates nuclear egress. Thus, the UL97-NEC interaction appears to recruit UL97 to the nuclear rim both for disruption of the nuclear lamina and phosphorylation of the NEC. Importance: Human cytomegalovirus (HCMV) causes birth defects and it can cause life-threatening diseases in immunocompromised patients. HCMV assembles in the nucleus and then translocates to the cytoplasm in an unusual process termed nuclear egress, an attractive target for antiviral therapy. A viral enzyme, UL97, is important for nuclear egress. It has been proposed that this is due to its role in disruption of the nuclear lamina, which would otherwise impede nuclear egress. In validating this proposal, we showed that independent disruption of the lamina can overcome a loss of UL97, but only partly, suggesting additional roles for UL97 during nuclear egress. We then found that UL97 phosphorylates the viral nuclear egress complex (NEC), which is essential for nuclear egress, and we obtained evidence that this phosphorylation modulates this process. Our results highlight a new role for UL97, the mutual dependence of the viral NEC and UL97 during nuclear egress, and differences among herpesviruses.
    Full-text · Article · Oct 2014 · Journal of Virology
  • Source
    Brian J Bender · Donald M Coen · Blair L Strang
    [Show abstract] [Hide abstract]
    ABSTRACT: Protein-protein and protein-nucleic acid interactions within sub-cellular compartments are required for viral genome replication. To understand the localization of the human cytomegalovirus viral replication factor UL84 relative to other proteins involved in viral DNA synthesis and to replicating viral DNA in infected cells, we created a recombinant virus expressing a FLAG tagged version of UL84 (UL84FLAG) and used this virus in immunofluorescence assays. UL84FLAG localization differed at early and late times of infection, transitioning from diffuse distribution throughout the nucleus to exclusion from the interior of replication compartments with some concentration at the periphery of replication compartments with newly labeled DNA and the viral DNA polymerase subunit UL44. Early in infection, UL84FLAG co-localized with the viral single stranded DNA binding protein UL57, but co-localization became less prominent as infection progressed. A portion of UL84FLAG also co-localized with the host nucleolar protein nucleolin, at the periphery of both replication compartments and nucleoli. siRNA mediated knockdown of nucleolin resulted in a dramatic elimination of UL84FLAG from replication compartments and other parts of the nucleus and its accumulation in the cytoplasm. Reciprocal co-immunoprecipitation of viral proteins from infected cell lysates revealed association of UL84, UL44 and nucleolin. These results indicate that UL84 localization during infection is dynamic, which is likely relevant to its functions, and suggest that its nuclear and subnuclear localization is highly dependent on direct or indirect interactions with nucleolin.
    Preview · Article · Jul 2014 · Journal of Virology
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The catalytic site of the HIV integrase is contained within an RNase H-like fold, and numerous drugs have been developed that bind to this site and inhibit its activity. Herpes simplex virus (HSV) encodes two proteins with potential RNase H-like folds, the infected cell protein 8 (ICP8) DNA-binding protein, which is necessary for viral DNA replication and exhibits recombinase activity in vitro, and the viral terminase, which is essential for viral DNA cleavage and packaging. Therefore, we hypothesized that HIV integrase inhibitors might also inhibit HSV replication by targeting ICP8 and/or the terminase. To test this, we evaluated the effect of 118-D-24, a potent HIV integrase inhibitor, on HSV replication. We found that 118-D-24 inhibited HSV-1 replication in cell culture at submillimolar concentrations. To identify more potent inhibitors of HSV replication, we screened a panel of integrase inhibitors, and one compound with greater anti-HSV-1 activity, XZ45, was chosen for further analysis. XZ45 significantly inhibited HSV-1 and HSV-2 replication in different cell types, with 50% inhibitory concentrations that were approximately 1 µM, but exhibited low cytotoxicity, with a 50% cytotoxic concentration greater than 500 µM. XZ45 blocked HSV viral DNA replication and late gene expression. XZ45 also inhibited viral recombination in infected cells and ICP8 recombinase activity in vitro. Furthermore, XZ45 inhibited human cytomegalovirus replication and induction of Kaposi's sarcoma herpesvirus from latent infection. Our results argue that inhibitors of enzymes with RNase H-like folds may represent a general antiviral strategy, which is useful not only against HIV but also against herpesviruses.
    Full-text · Article · Jul 2014 · mBio
  • Mayuri Sharma · Donald M Coen
    [Show abstract] [Hide abstract]
    ABSTRACT: Human cytomegalovirus (HCMV) kinase UL97 is required for efficient nuclear lamina disruption during nuclear egress. However, cellular protein kinase C (PKC) has been implicated in this process in other systems. Comparing the effects of UL97 and cellular kinase inhibitors on HCMV nuclear egress confirms a role for UL97 in lamina disruption and nuclear egress. A pan-PKC inhibitor did not affect lamina disruption but did reduce the number of cytoplasmic capsids more than the number of nuclear capsids.
    No preview · Article · Jun 2014 · Journal of Virology
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: After infecting peripheral sites, herpes simplex virus (HSV) invades the nervous system and initiates latent infection in sensory neurons. Establishment and maintenance of HSV latency require host survival, and entail repression of productive cycle ("lytic") viral gene expression. We find that a neuron-specific microRNA, miR-138, represses expression of ICP0, a viral transactivator of lytic gene expression. A mutant HSV-1 (M138) with disrupted miR-138 target sites in ICP0 mRNA exhibits enhanced expression of ICP0 and other lytic proteins in infected neuronal cells in culture. Following corneal inoculation, M138-infected mice have higher levels of ICP0 and lytic transcripts in trigeminal ganglia during establishment of latency, and exhibit increased mortality and encephalitis symptoms. After full establishment of latency, the fraction of trigeminal ganglia harboring detectable lytic transcripts is greater in M138-infected mice. Thus, miR-138 is a neuronal factor that represses HSV-1 lytic gene expression, promoting host survival and viral latency.
    Full-text · Article · Apr 2014 · Cell host & microbe
  • Source
    Shariya L Terrell · Jean M Pesola · Donald M Coen
    [Show abstract] [Hide abstract]
    ABSTRACT: The catalytic subunit of the herpes simplex virus 1 DNA polymerase (HSV-1 Pol) is essential for viral DNA synthesis and production of infectious virus in cell culture. While mutations that affect 5-3 polymerase activity have been evaluated in animal models of HSV-1 infection, mutations that affect other functions of HSV-1 Pol have not. In a previous report, we utilized bacterial artificial chromosome technology to generate defined HSV-1 pol mutants with lesions in the previously uncharacterized pre-NH2-terminal domain. We found that the extreme N-terminal 42 residues (deletion mutant polΔN43) were dispensable for replication in cell culture, while residues 44-49 (alanine-substitution mutant polA6) were required for efficient viral DNA synthesis and production of infectious virus. In this study, we sought to address the importance of these conserved elements in viral replication in a mouse corneal infection model. Mutant virus polΔN43 exhibited no meaningful defect in acute or latent infection despite strong conservation of residues 1-42 with HSV-2 Pol. The polA6 mutation caused a modest defect in replication at the site of inoculation, and was severely impaired for ganglionic replication, even at high inocula that permitted efficient corneal replication. Additionally, the polA6 mutation resulted in reduced latency establishment and subsequent reactivation. Moreover, we found that the polA6 replication defect in cultured cells was exacerbated in resting cells as compared to dividing cells. These results reveal an important role for the conserved motif at residues 44-49 of HSV-1 Pol for ganglionic viral replication.
    Preview · Article · Jan 2014 · Journal of General Virology
  • Source
    Anna R Cliffe · Donald M Coen · David M Knipe
    [Show abstract] [Hide abstract]
    ABSTRACT: Unlabelled: The herpes simplex virus (HSV) genome is associated with heterochromatic histone modifications, including trimethylation of the lysine 27 residue of histone H3 (H3K27me3), during latent infection of neurons. Here we have examined the kinetics of general chromatin and H3K27me3 association with the viral genome during establishment of latent infection. Using both wild-type virus and a mutant virus that is unable to undergo replication in neurons, we found that histone H3 associates with viral gene promoters by 7 days postinfection (dpi). Levels of H3K27me3 were low at 7 dpi but increased dramatically by 14 dpi. Hence, general chromatin association and/or other factors may play a key role(s) in the initial silencing of lytic genes, and H3K27me3 may play a role in further suppression of the genome and/or the maintenance of latency. A component of Polycomb repressive complex 2 (PRC2), which mediates the addition of K27me3 to histone H3 (Suz12), was also recruited by 14 dpi. We have shown previously that the levels of H3K27me3 during latent infection are increased in the presence of the latency-associated transcript (LAT). However, the initial targeting of PRC2 was not found to be dependent on the LAT. We found that a component of the PRC1 complex (Bmi1), which binds to H3K27me3, was not enriched at promoters found previously to be enriched for H3K27me3. Our results are consistent with (i) chromatinization of viral DNA or other mechanisms causing the initial silencing of HSV lytic genes and (ii) facultative heterochromatin maintaining that silencing during latent infection of neurons. Importance: The human pathogen herpes simplex virus (HSV) hides for the lifetime of the host in peripheral neurons. The mechanism by which HSV is able to shut off its gene expression and persist in neurons is not known. Here we show that the HSV DNA first associates with histone H3, with later recruitment of Polycomb repressor complex 2 (PRC2) and trimethylation of the lysine 27 residue of histone H3 (H3K27me3), a modification associated with heterochromatin. This work indicates that the initial silencing of HSV gene expression is not correlated with enrichment of H3K27me3 and that PRC2 may be recruited to already-silenced genes to further silence gene expression and/or maintain gene silencing. We demonstrate that recruitment of PRC2 is not dependent upon expression of the noncoding HSV latency-associated transcripts, indicating the presence of unknown triggers for PRC2 recruitment during the establishment of latent infection.
    Full-text · Article · Dec 2013 · mBio
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To facilitate studies of herpes simplex virus 1 latency, cell culture models of quiescent or latent infection have been developed. Using deep sequencing, we analyzed the expression of viral microRNAs (miRNAs) in two models employing human fibroblasts and one using rat neurons. In all cases, the expression patterns differed from that in productively infected cells, with the rat neuron pattern most closely resembling that found in latently infected human or mouse ganglia in vivo.
    Full-text · Article · Dec 2013 · Journal of Virology
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Herpesvirus nucleocapsids traverse the nuclear envelope into the cytoplasm in a process called nuclear egress that includes disruption of the nuclear lamina. In several herpesviruses, a key player in nuclear egress is a complex of two proteins, whose homologs in human cytomegalovirus (HCMV) are UL50 and UL53. However, their roles in nuclear egress during HCMV infection have not been shown. Based largely on transfection studies, UL50 and UL53 have been proposed to facilitate disruption of the nuclear lamina by recruiting cellular protein kinase C (PKC), as occurs with certain other herpesviruses, and/or viral protein kinase UL97 to phosphorylate lamins. To investigate these issues during HCMV infection, we generated viral mutants null for UL50 or UL53. Correlative light electron microscopic analysis of null mutant-infected cells showed the presence of intranuclear nucleocapsids and the absence of cytoplasmic nucleocapsids. Confocal immunofluorescence microscopy revealed that UL50 and UL53 are required for disruption of the nuclear lamina. A subpopulation of UL97 co-localized with the nuclear rim, dependent on UL50 and, to a lesser extent, UL53. However, PKC was not recruited to the nuclear rim, and its localization was not affected by the absence of UL50 or UL53. Immunoprecipitation from cells infected with HCMV expressing tagged UL53, detected UL97, but not PKC. In summary, HCMV UL50 and UL53 are required for nuclear egress and disruption of nuclear lamina during HCMV infection, and recruit UL97, not PKC for these processes. Thus, despite the strong conservation of herpesvirus nuclear egress complexes, a key function can differ among them.
    Full-text · Article · Oct 2013 · Journal of Virology
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We investigated thymidine kinase (tk) mutants isolated during multiple episodes of recurrent bilateral acyclovir resistant herpes simplex keratitis in an immunocompetent patient. From one eye, we found a single guanine insertion, previously shown to greatly reduce TK expression, and from the other, a previously unidentified substitution, which genetic experiments confirmed confers drug resistance. The substitution, although distant from substrate binding sites, reduced thymidine phosphorylation 10–20-fold, and acyclovir phosphorylation >100-fold. This phenotype should permit reactivation from latency to cause recurrent disease. The results may have implications for the prevalence and prevention of acyclovir resistance in patients with herpes simplex keratitis.
    Preview · Article · Aug 2013 · The Journal of Infectious Diseases
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Human cytomegalovirus (HCMV) encodes one conventional protein kinase, UL97. During infection, UL97 phosphorylates the retinoblastoma tumor suppressor protein (pRb) on sites ordinarily phosphorylated by cyclin-dependent kinases (CDK), inactivating the ability of pRb to repress host genes required for cell cycle progression to S-phase. UL97 is important for viral DNA synthesis in quiescent cells, but this function can be replaced by human papillomavirus-16 E7, which targets pRb for degradation. However, viruses in which E7 replaces UL97 are still defective for virus production. UL97 is also required for efficient nuclear egress of viral nucleocapsids, which is associated with disruption of the nuclear lamina during infection, and phosphorylation of lamin A/C on serine 22, which antagonizes lamin polymerization. We investigated whether inactivation of pRb might overcome the requirement of UL97 for these roles, as pRb inactivation induces CDK1, and CDK1 phosphorylates lamin A/C on serine 22. We found that lamin A/C serine 22 phosphorylation during HCMV infection correlated with expression of UL97 and was considerably delayed in UL97-null mutants, even when E7 was expressed. E7 failed to restore gaps in the nuclear lamina seen in wild type but not UL97-null infections. In electron microscopy analyses, a UL97-null virus expressing E7 was as impaired as a UL97-null mutant in cytoplasmic accumulation of viral nucleocapsids. Our results demonstrate that pRb inactivation is insufficient to restore efficient viral nuclear egress of HCMV in the absence of UL97 and instead argue further for a direct role of UL97 in this stage of the infectious cycle.
    Full-text · Article · Feb 2013 · Journal of Virology
  • Source
    Dataset: Figure S5
    Anna R. Cliffe · Donald M. Coen · David M. Knipe
    [Show abstract] [Hide abstract]
    ABSTRACT: Cellular controls for the Suz12 ChIP carried out on TG isolated from mice infected with KdlLAT or KFSLAT. Download
    Preview · Dataset · Jan 2013
  • Source
    Dataset: Figure S6
    Anna R. Cliffe · Donald M. Coen · David M. Knipe
    [Show abstract] [Hide abstract]
    ABSTRACT: Cellular controls for the Bmi1 ChIP carried out on TG isolated from mice infected with WT HSV-1. Download
    Preview · Dataset · Jan 2013
  • Source
    Dataset: Figure S3
    Anna R. Cliffe · Donald M. Coen · David M. Knipe
    [Show abstract] [Hide abstract]
    ABSTRACT: Viral lytic gene expression at 14 days following infection with WT virus. Mice were infected with WT HSV-1, and at 14 dpi, the mice were sacrificed and trigeminal ganglia were collected. The absolute copy numbers of lytic gene transcripts for ICP0, ICP4, and UL48 are shown. Download
    Preview · Dataset · Jan 2013
  • Source
    Dataset: Figure S2
    Anna R. Cliffe · Donald M. Coen · David M. Knipe
    [Show abstract] [Hide abstract]
    ABSTRACT: Cellular controls for ChIP on TG isolated from dlsptk-infected mice. Mice were infected with a dlsptk virus and at 7, 10, and 14 dpi, mice were sacrificed and the TG isolated. ChIP was carried out as described in the materials and methods. Shown are the percentages of cellular DNA immunoprecipitated with IgG control antibody (A), with a histone H3 antibody (B), and with an H3K27me3 antibody (C). (D) The fraction of cellular DNA immunoprecipitated with the H3K27me3 antibody was normalized to the fraction immunoprecipitated with the total H3 antibody. Download
    Preview · Dataset · Jan 2013

Publication Stats

12k Citations
1,318.67 Total Impact Points

Institutions

  • 1990-2016
    • Harvard University
      Cambridge, Massachusetts, United States
    • University of British Columbia - Vancouver
      Vancouver, British Columbia, Canada
  • 1984-2015
    • Harvard Medical School
      • Department of Biological Chemistry and Molecular Pharmacology
      Boston, Massachusetts, United States
  • 2005-2007
    • State University of New York Upstate Medical University
      • Department of Microbiology and Immunology
      Syracuse, NY, United States
  • 1998-2007
    • University of Padova
      • Department of Biomedical Sciences - DSB
      Padua, Veneto, Italy
  • 1995
    • Joslin Diabetes Center
      Boston, Massachusetts, United States
  • 1986-1995
    • University of Cambridge
      • Department of Pathology
      Cambridge, ENG, United Kingdom
  • 1985
    • Fred Hutchinson Cancer Research Center
      • Division of Basic Sciences
      Seattle, Washington, United States
  • 1980-1985
    • Dana-Farber Cancer Institute
      Boston, Massachusetts, United States