ASSAY and Drug Development TechnologiesVolume 1, Number 5, 2003© Mary Ann Liebert, Inc.

Literature Search and Review

J. Fraser Glickman

In each issue of ASSAY and Drug Development Technologies, our Literature Editor, J. Fraser Glickman, Ph.D.,selects several significant papers covering timely and pertinent topics that will keep our readers up-to-date on theever-changing field of assays and drug development technologies. Dr. Glickman provides relevant commentary oneach of the cited abstracts.

735

Baker JG, Hall IP, Hill SJ: Influence of agonist effi-cacy and receptor phosphorylation on antagonistaffinity measurements: differences between secondmessenger and reporter gene responses. Mol Phar-macol 2003;64:679–688.

Abstract: The ability of an antagonist to bind to a recep-tor is an innate property of that ligand–receptor chemical in-teraction. Provided no change in the antagonist or receptorchemical nature occurs, this affinity should remain constantfor a given antagonist–receptor interaction, regardless of theagonists used. This fundamental assumption underpins theclassification of receptors. Here, measurements of beta2-adrenoceptor-mediated cAMP accumulation and cAMP re-sponse-element (CRE)-mediated reporter-gene transcriptionrevealed differences in antagonist affinity that dependedupon agonist incubation time and the efficacy of the com-peting agonist. In cAMP accumulation studies (10-min ag-onist incubation), antagonist affinities were the same re-gardless of the agonist used. The CRE-reporter gene assay(5 h of incubation) antagonist affinities were 10-fold lowerin the presence of isoprenaline and adrenaline than whensalbutamol or terbutaline were present (e.g., log KD propra-nolol 28.65 6 0.08, n 5 22, and 29.68 6 0.07, n 5 17, forisoprenaline- and salbutamol-induced responses, respec-tively). Isoprenaline and adrenaline were more efficaciousin functional studies, and their ability to internalize GFP-tagged human beta2-adrenoceptors. Longer-term cAMPstudies also showed significant differences in KD valuesmoving toward that seen with gene transcription. Agonist-dependent differences in antagonist affinity were reducedfor reporter-gene responses when a phosphorylation-defi-cient mutant of the beta2-adrenoceptor was used. This studysuggests that high-efficacy agonists induce a chemical mod-

ification in beta2-adrenoceptors (via phosphorylation) thatreduces antagonist affinities. Because reporter-gene assaysare used for high-throughput screening in drug discovery,less efficacious or partial agonists may be more reliable thanhighly efficacious agonists when reporter-gene techniquesare used to estimate antagonist affinity.

Commentary: A solid understanding of the molecularpharmacology of G protein-coupled receptor (GPCR) ag-onists and antagonists is essential for the determination ofstructure-activity relationships of novel lead compoundsfor this target class. Aside from the classic ligand-bindingassays, various assay readouts for GPCR activity are com-monly used, including the measurement of calcium flux,cAMP formation, guanine nucleotide exchange, and tran-scriptional reporter assays. Often, candidate compoundsare tested in a number of such assays in order to confirmpotency, selectivity, and mechanism. “Real-world” phar-macology on lead compounds sometimes can result in dis-parate results between, for example, reporter gene assaysand binding experiments. In this work, Baker et al. havepresented some reasonable explanations for some of thesedisparities—namely, that with long-term readouts, down-stream events, such as receptor phosphorylation, desensi-tization and internalization can result in an altered affin-ity between the receptor and ligand, resulting in a shiftedEC50 curve. One might infer from this that ligand-inducedreceptor phosphorylation induces a conformationalchange that alters its affinity for certain alternative li-gands. Such findings have important ramifications in thedesign of screening assays and in the optimization of leadcompounds. Additionally, the authors present methods formeasuring ligand-induced cAMP formation that demon-strate significant cAMP detected extracellularly.

��

A Lot More than Pharmacology 101

Chao SH, Walker JR, Chanda SK, Gray NS, Cald-well JS: Identification of homeodomain proteins,PBX1 and PREP1, involved in the transcription ofmurine leukemia virus. Mol Cell Biol 2003;23:831–841.

Abstract: Cyclin-dependent kinase inhibitors (CDKIs)have been shown to block human immunodeficiencyvirus and herpes simplex virus. It is hypothesized thatCDKIs block viral replication by inhibiting transcriptionof specific cellular genes. Here we find that three CD-KIs, flavopiridol, purvalanol A, and methoxyroscovitine,block Moloney murine leukemia virus (MLV) transcrip-tion events. Using gene expression microarray technol-ogy to examine the inhibitory effects of CDKIs, we observed a cellular gene, the pre-B-cell leukemia tran-scription factor 1 (Pbx1) gene, down-regulated by CDKItreatment. The PBX consensus element (PCE), TGA-TTGAC, is conserved in the long terminal repeats of sev-eral murine retroviruses, including Moloney MLV. Mu-tations in the PCE completely inhibited viral transcriptionwhereas overexpression of PBX1 and a PBX1-associatedprotein, PREP1, enhanced viral transcription. The inter-

action between the PCE and PBX1-PREP1 proteins wasconfirmed by gel shift experiments. Blocking PBX1 pro-tein synthesis resulted in a significant decrease in viraltranscription. Collectively, our results represent the firstwork demonstrating that the homeodomain proteins PBX1and PREP1 are cellular factors involved in Moloney MLVtranscription regulation.

Commentary: This article gives an example of how aclass of small molecular weight inhibitors of known en-zymes (protein kinases) might be used to identify newpathways. A commonly known set of protein kinase in-hibitors was found to block Moloney murine leukemiavirus-induced transcription. Results of a gene expressionarray experiment showed that treatment of host cells withthese inhibitors resulted in lowered expression of a par-ticular DNA-binding protein (PBX1). Subsequent bio-chemical and mutational experiments showed that thishomeodomain protein is subverted by the virus and is im-portant for viral life cycle. Given that the protein kinaseinhibitors used in this study are not very specific, it is in-teresting that they could still be used as useful tools inthe understanding of viral infection processes.

Literature Search and Review736

Kinase Inhibitors and Homeodomains

It Must Have Been the Mixture?

Borisy AA, Elliott PJ, Hurst NW, Lee MS, Lehar J,Price ER, Serbedzija G, Zimmermann GR, Foley MA,Stockwell BR, Keith CT: Systematic discovery of mul-ticomponent therapeutics. Proc Natl Acad Sci U S A2003;100:7977–7982.

Abstract: Multicomponent therapies, originating throughdeliberate mixing of drugs in a clinical setting, throughhappenstance, and through rational design, have a suc-cessful history in a number of areas of medicine, includ-ing cancer, infectious diseases, and CNS disorders. Wehave developed a high-throughput screening method foridentifying effective combinations of therapeutic com-pounds. We report here that systematic screening of com-binations of small molecules reveals unexpected interac-tions between compounds, presumably due to interactionsbetween the pathways on which they act. Through sys-tematic screening of approximately 120,000 different two-component combinations of reference-listed drugs, weidentified potential multicomponent therapeutics, includ-ing (i) fungistatic and analgesic agents that together gen-erate fungicidal activity in drug-resistant Candida albi-cans, yet do not significantly affect human cells, (ii)glucocorticoid and antiplatelet agents that together sup-

press the production of tumor necrosis factor-alpha in hu-man primary peripheral blood mononuclear cells, and (iii)antipsychotic and antiprotozoal agents that do not exhibitsignificant antitumor activity alone, yet together preventthe growth of tumors in mice. Systematic combinationscreening may ultimately be useful for exploring the con-nectivity of biological pathways and, when performedwith reference-listed drugs, may result in the discovery ofnew combination drug regimens.

Commentary: It appears that the right mixture of twocompounds in an assay can have surprising synergisticeffects. New automated liquid dispensing systems makepossible the rapid testing of ordered combinations ofsmall molecules in compound assays. In this study,known drugs were chosen for screening double combi-nations of compounds against antifungal targets, anti-inflammatory assays, and oncology assays. Previouslyunknown synergistic effects between two unrelated com-pounds were discovered. It is difficult to explain suchresults unless one considers that the side effects of com-mon drugs can become important factors as a multi-component strategy for hitting a particular cellular path-way. For example, weakly inhibiting two branches in a

particular signaling pathway may add up to a potentoverall inhibition in a cell-based assay. The approach,in this case, has the advantage that the compounds iden-tified, because they are drugs, are already orallybioavailable and relatively nontoxic. One should be

aware that toxicity is still a potential problem for com-bination therapies. It will be interesting to see if onecould advance this approach to treat human or veteri-nary diseases—perhaps clinical data already exist tosupport such approaches.

Literature Search and Review 737

Revealing Expressions

Gunther EC, Stone DJ, Gerwien RW, Bento P, HeyesMP: Prediction of clinical drug efficacy by classifica-tion of drug-induced genomic expression profiles invitro. Proc Natl Acad Sci U S A 2003;100:9608–9613.

Abstract: Assays of drug action typically evaluate bio-chemical activity. However, accurately matching thera-peutic efficacy with biochemical activity is a challenge.High-content cellular assays seek to bridge this gap bycapturing broad information about the cellular physiol-ogy of drug action. Here, we present a method of pre-dicting the general therapeutic classes into which variouspsychoactive drugs fall, based on high-content statisticalcategorization of gene expression profiles induced bythese drugs. When we used the classification tree and ran-dom forest supervised classification algorithms to ana-lyze microarray data, we derived general “efficacy pro-files” of biomarker gene expression that correlate withantidepressant, antipsychotic, and opioid drug action onprimary human neurons in vitro. These profiles were usedas predictive models to classify naive in vitro drug treat-ments with 83.3% (random forest) and 88.9% (classifi-cation tree) accuracy. Thus, the detailed information con-tained in genomic expression data is sufficient to matchthe physiological effect of a novel drug at the cellularlevel with its clinical relevance. This capacity to identifytherapeutic efficacy on the basis of gene expression sig-natures in vitro has potential utility in drug discovery anddrug target validation.

Commentary: Expression profiling is becoming a usefultool in drug discovery. For example, one can examine

the effect of a compound on the simultaneous expressionof many RNA transcripts in a cell. The compound can ei-ther increase the expression level, decrease the expres-sion level, or have no effect on the transcript. Sometimesthis information can be useful in making correlationsabout the mechanism of action of the compound or on thecompound’s selectivity for particular pathways. A weak-ness of such experiments is that performing time-courseexperiments is not technically possible without a largeamount of processing time, and thus one cannot easilyexamine the temporal pattern of gene expression changes,which might fluctuate between positive and negative ef-fects over time. Because the experimental methodologiesvary and the expression patterns are complex, the ex-periments are open to multiple interpretations. Thus, theinterpretation of complex expression profiling experi-ments have become reliant upon computer programs.

In this study, the patterns of expression profiles werecategorized by computer algorithms and correlated withparticular drug mechanisms. Psychoactive compoundswith known mechanisms were used in expression profilingexperiments as a training set for a computer program topredict the class of an “unknown” compound. A numberof test compounds with known mechanism of action wereused to see if the computer program could accurately pre-dict the class of compound that resulted in a particular ex-pression profile. The results suggest that such methodscould be used to predict the class of an unknown lead com-pound, if there are sufficient data from training sets ofknown compounds. This approach would seem to have lim-itations when very new targets are being addressed, andthus training sets would be unavailable.

More Model Zebrafish

Milan DJ, Peterson TA, Ruskin JN, Peterson RT,MacRae CA: Drugs that induce repolarization ab-normalities cause bradycardia in zebrafish. Circula-tion 2003;107:1355–1358.

Abstract: Background: Drug-induced QT prolongationand torsades de pointes remain significant and often un-predictable clinical problems. Current in vitro preclinicalassays are limited by biological simplicity, and in vivo

models suffer from expense and low throughput. Meth-ods and Results: During a screen for the effects of 100small molecules on the heart rate of the zebrafish, Daniorerio, we found that drugs that cause QT prolongation inhumans consistently caused bradycardia and AV block inthe zebrafish. Of 23 such drugs tested, 18 were positivein this initial screen. Poor absorption explained 4 of 5false-negative results, as demonstrated by microinjection.Overall, 22 of 23 compounds that cause repolarization ab-

normalities were positive in this assay. Antisense “knock-down” of the zebrafish KCNH2 ortholog yielded brady-cardia in a dose-dependent manner, confirming the effectsof reduction of repolarizing potassium current in thismodel. Classical drug–drug interactions between erythro-mycin and cisapride, as well as cimetidine and terfenadine,were also reproduced. Conclusion: This simple high-throughput assay is a promising addition to the repertoireof preclinical tests for drug-induced repolarization abnor-malities. The genetic tractability of the zebrafish will al-low the exploration of heritable modifiers of such drug ef-fects.

Commentary: A common mechanism among experimen-tal drugs for lethal side effects is through blocking of var-ious potassium channels in the heart, thus ultimately lead-ing to QT prolongation. Prescreening of lead compoundsfor such activity could eliminate such toxic compoundsearly in the discovery process. Biochemical approaches to

prescreening are useful because they provide rapid meansfor testing new compounds, but also have various limita-tions in predicting the side effects in vivo. The false pos-itives and negatives in the biochemical assays are proba-bly a result of the sensitivity of various assays and to thenumber of receptors and channels that are involved. Thus,more physiological assays performed in animals would beuseful, provided that they are relatively quick. Methods fortesting potential for cardiotoxic side effects in zebrafishpresented in this article have such advantages. The methodrequires the photographing of the animals over time witha microscope. The heart is visible and the heartbeat canbe quantified using algorithms that measure pixel density.The study surveyed 100 drugs with 23 known QT pro-longers. The study found that the QT prolongers causedbradycardia and AV block that was easily detectable inthe zebrafish. The false negatives could be explained bypoor tissue penetration. Such an approach should be fea-sible as a standard lead profiling tool.

Literature Search and Review738

Partial for Ion Channels

Jin R, Banke TG, Mayer ML, Traynelis SF, GouauxE: Structural basis for partial agonist action ationotropic glutamate receptors. Nat Neurosci 2003;6:803–810.

Abstract: An unresolved problem in understanding neu-rotransmitter receptor function concerns the mecha-nism(s) by which full and partial agonists elicit differ-ent amplitude responses at equal receptor occupancy.The widely held view of “partial agonism” posits thatresting and active states of the receptor are in equilib-rium, and partial agonists simply do not shift the equi-librium toward the active state as efficaciously as fullagonists. Here we report findings from crystallographicand electrophysiological studies of the mechanism of ac-tivation of an AMPA-subtype glutamate receptor ionchannel. In these experiments, we used 5-substitutedwillardiines, a series of partial agonists that differ byonly a single atom. Our results show that the GluR2 li-gand-binding core can adopt a range of ligand-depen-dent conformational states, which in turn control theopen probability of discrete subconductance states of theintact ion channel. Our findings thus provide a structure-based model of partial agonism.

Commentary: This study presents an excellent demon-stration of how a very general question in biology canbe answered with the correct set of tools and assays.In this case, a mechanistic understanding of how par-tial agonists (agonists that elicit only a partial func-tional response despite full receptor occupancy) work,was achieved by “gathering together” a well-charac-terized functional assay (AMPA-type glutamate recep-tors), the ability to perform single channel current mea-surements, a high-quality crystal structure, and a seriesof single element substituted analogs (willardiines).For example, x-ray structures of willardiine–receptorcomplexes showed that increasing the size of the 5-sub-stituent of the agonist (willardiines) gave a graded levelof cleft opening between two domains of the receptor.Full agonists caused complete closure, whereas partialagonists only gave partial closure. The graded cleftopenings were correlated with degrees of subconduc-tance ion-channel opening. The study provides us witha better understanding of the molecular mechanisms ofthe phenomenon of partial agonism, and we can con-clude that this type of agonism can be caused by spe-cific and different conformational effects on the recep-tor (See figure).

Literature Search and Review 739

0.7Coupling efficiency (e)

FWBrW

HW

IW

Glutamatee = 0.67

Dom

ain

Clo

sure

(de

gree

)

10

20

0.4

C

0.7

A

A

A

AA

Coupling efficiency (e)

Glutamatee = 0.67

p(17.9 pS) = e44 gates open17.9 pS

e indicates coupling efficiencybetween domain closure andsingle subunit gating HW e = 0.59

FW e = 0.58

BrW e = 0.52 IW e = 0.41

I/I(G

luta

mat

e)

0.3

1.0

0.4

AA

A

p(11.2 pS) = 4e3(1-e)3 gates open11.2 pS

A

A

p(6.5 pS) = 6e2(1-e)22 gates open*6.5 pS

Ap(low pS) = 4e(1-e)31 gate open

Low conduct

p(closed) = (1-e)40 gates open0 pS

A

A

A

A

AA

AA

AA

Receptor Interactions

Fiorentini C, Gardoni F, Spano P, Di Luca M, MissaleC: Regulation of dopamine D1 receptor traffickingand desensitization by oligomerization with glutamateN-methyl-D-aspartate receptors. J Biol Chem2003;278:20196–20202.

Abstract: Activation of dopamine D1 receptors is crit-ical for the generation of glutamate-induced long-termpotentiation at corticostriatal synapses. In this study, wereport that, in striatal neurons, D1 receptors are co-lo-calized with N-methyl-D-aspartate (NMDA) receptors inthe postsynaptic density and that they co-immunopre-cipitate with NMDA receptor subunits from postsynap-tic density preparations. Using modified biolumines-cence resonance energy transfer, we demonstrate that D1and NMDA receptor clustering reflects the existence ofdirect interactions. The tagged D1 receptor and NR1 sub-unit co-transfected in COS-7 cells generated a signifi-cant bioluminescence resonance energy transfer signalthat was insensitive to agonist stimulation and that did

not change in the presence of the NR2B subunit, sug-gesting that the D1 receptor constitutively and selectivelyinteracts with the NR1 subunit of the NMDA channel.Oligomerization with the NR1 subunit substantiallymodified D1 receptor trafficking. In individually trans-fected HEK293 cells, NR1 was localized in the endo-plasmic reticulum, whereas the D1 receptor was targetedto the plasma membrane. In co-transfected cells, both theD1 receptor and NR1 subunit were retained in cytoplas-mic compartments. In the presence of the NR2B subunit,the NR1–D1 receptor complex was translocated to theplasma membrane. These data suggest that D1 and NMDAreceptors are assembled within intracellular compartmentsas constitutive heteromeric complexes that are deliveredto functional sites. Co-expression with NR1 and NR2Bsubunits also abolished agonist-induced D1 receptor cy-toplasmic sequestration, indicating that oligomerizationwith the NMDA receptor could represent a novel regu-latory mechanism modulating D1 receptor desensitiza-tion and cellular trafficking.

Subunit-linked channel opening is correlated with degree of domain closure. (A) We modeled the function of homomeric GluR2-L483Y receptors by assuming that each of four agonist-bound subunits undergoes a conformational change that influences thepermeation properties of the channel. The coupling efficiency « describes the probability that an independent subunit-dependentchange that influences permeation will occur when an agonist is bound. We assigned the shared IW- and glutamate-activatedconductance levels to channel conformations with 2, 3 or 4 subunit-dependent gates open; we assumed the conductance for onesubunit-dependent gate open was too low to measure. The probabilities that 0, 1, 2, 3 and 4 independent subunit gates were openwere determined using the binomial expansion; *indicates the existence of three different conformations for dimeric channels(only one is shown). (B) The response amplitude to the maximal concentration of agonist was calculated from the probability ofopenings to different subconductance levels as a function of « (Methods). The smooth line shows the ratio of the response am-plitude for « between 0 and 1 to the glutamate response calculated using « 5 0.67. The symbols show the value for « that cor-responds to the ratio of the response to maximal concentrations of partial agonists to the maximal glutamate response from oocyteor HEK cell data (filled and open symbols, respectively). (C) Graphical representation of the correlation between coupling effi-ciency and ligand binding domain closure. Reprinted from Nat. Neurosci 6:803–810. © 2003.

A B

C

Commentary: The molecular basis for the pharmaco-logic relationship between the NMDA (glutamate-gatedion channels) receptors and dopamine D1 (G protein-coupled) receptors has recently been expanded upon. Thetwo receptors have been demonstrated to co-localize tosimilar neurons and subcellular compartments. Also, ithas been known that the two receptors have direct in-teractions in cells. This article goes on to show that thedirect interaction between these two types of receptors

in living cells [as measured using several techniques, including bioluminescent resonance energy transfer(BRET)] is part of a system in which the NR1 subunit ofNMDA receptor influences the D1 dopamine receptor’slocalization to the membrane, and also prevents the ag-onist-induced internalization of the receptor that causesdesensitization. Thus, these heteroreceptor complexesare a very important mechanism for regulation of re-ceptor function through subcellular transport.

Literature Search and Review740