Tài liệu HPLC for Pharmaceutical Scientists 2007 (Part 11) pdf

Included are significant develop-ments in HPLC/MS to support target selection proteomics, biologicalscreening and assay development, high-throughput compound analysis and characterization

Great efficiencies have been achieved in the drug discovery process as a result

of technological advances in target identification, high-throughput screening,

high-throughput organic synthesis, just-in-time in vitro ADME (absorption,

distribution, metabolism, and excretion), and early pharmacokinetic screening

of drug leads These advances, spanning target selection all the way through

to clinical candidate selection, have placed greater and greater demands onthe analytical community to develop robust high-throughput methods Thisreview highlights the various roles of high-performance liquid chromatogra-phy/mass spectrometry (HPLC/MS) in drug discovery and how the field hasevolved over the past several years since the introduction of myriad high-throughput drug discovery technologies Included are significant develop-ments in HPLC/MS to support target selection (proteomics), biologicalscreening and assay development, high-throughput compound analysis and characterization, UV- and mass-directed fractionation for unattended,

automated compound purification, and high-throughput in vitro ADME

screening

Focus within the pharmaceutical industry has been to increase the hood of successfully developing clinical candidates by optimizing the compo-nents of the discovery process (i.e., spanning target identification → chemical

likeli-535

HPLC for Pharmaceutical Scientists, Edited by Yuri Kazakevich and Rosario LoBrutto

Copyright © 2007 by John Wiley & Sons, Inc.

design→ synthesis → compound analysis and purification → registration →biological and ADME screening By optimizing each step in the iterative dis-covery process, it is expected that the compound attrition rate will be reduceddramatically as compounds advance into preclinical development Both HPLCand LC/MS enjoy important roles throughout the discovery process, as will behighlighted in detail in this review Once considered primarily an enabling toolfor medicinal chemists, HPLC and LC/MS are now key technologies incorpo-rated at just about every stage of the drug discovery process Drug discoveryprograms typically initiate, as follows Assuming that the relevant therapeuticarea (e.g., oncology, metabolic diseases, inflammation, pain, CNS, etc.) has beenselected, the next step is to identify a biological target relevant to the disease.

As will be discussed shortly, numerous technological advances in the field

of analytical chemistry (e.g., nanocolumn HPLC/MS/MS) that have greatlyfacilitated protein/target identification have been made since the humangenome initiative was launched Following on the heels of target selection isthe requirement to establish tools for “just-in-time” high-throughput screen-ing of compound repositories (so-called corporate collections) and syntheticlibraries as a means for identifying initial hits/actives In combination with structure–activity relationship (SAR) data generated from these high-throughput screens, chemists incorporate knowledge of protein three-

dimensional structures and utilize computational tools (i.e., in silico methods

that measure diversity and “drug-likeness” as well two-dimensional and dimensional pharmacophore models [descriptors] that predict biological activ-ity) to support iterative compound design, synthesis, and biological testing.Once the hits or actives have been identified, the process of hit refinement andlead optimization is initiated At this stage, a chemistry team is established andboth parallel synthesis and more traditional medicinal chemistry strategies areincorporated to rapidly converge on qualified leads (so-called hit-to-leadstage) HPLC and LC/MS play an extremely important role in the hit-to-leadstage of discovery, providing key enabling analysis and purification capabili-ties to the medicinal chemist Furthermore, activities that were traditionallyrelegated to drug metabolism and pharmacokinetics departments withindevelopment organizations are now integrated into early discovery so as to

three-provide early measurements and predictions of in vivo properties Again,

LC/MS has played an extremely important role in enhancing the drug opability of these hits and leads All of these advances have helped to stream-line the discovery phase of pharmaceutical drug discovery and developmentand are presented within

devel-11.2 APPLICATIONS OF HPLC/MS FOR PROTEIN

IDENTIFICATION AND CHARACTERIZATION

The human genome initiative that took a stronghold on biotechnology panies in the early 1990s through the first few years of the twenty-first century

com-spawned a completely new field that had analytical chemistry as its stone Specifically, high-resolution capillary and nano-column HPLC coupledwith tandem mass spectrometry became one of the tools of choice for char-acterizing proteins and identifying potential therapeutic protein targets.Although capillary HPLC/MS/MS was applied as early as 1989–1991 to thecharacterization of proteins and for identifying sites of post-translational mod-ification [1–4], the field took off in earnest following the genomics boom andbecame known as proteomics, coined by Wilkins et al [5] In essence, themandate of the proteomics field since its inception has been to identify dif-ferences at the protein level, in cells, tissues, plasma, and so on, between adisease state and control (“normal”) The basic premise is that proteins will

corner-be either up- or down-regulated (i.e., over- or underexpressed) in the diseasestate relative to “normal” state, and these differences can be identified andquantified by mass spectrometry There have been several analytical advancesmade in the field of proteomics since its inception, far too numerous to capture

in this review One noteworthy advance in proteomics is the technique of multidimensional protein identification technology (MUDPIT), developed byYates and co-workers, which has been used widely in place of the more labo-rious, less automated method of 2D-polyacrylamide electrophoresis [6].MUDPIT is a column chromatography method whereby ion-exchange chro-matography is used in the first dimension of chromatography to simplify thecomplexity of the complex mixture of peptides by separating them based oncharge followed by reversed-phase HPLC for the higher-resolution separationbased on molecular weight and hydrophobicity An equally important devel-opment in the field of proteomics has been isotope-coded affinity tags (ICAT)technology, a method whereby isotopic labeling of peptides containing cys-teine residues is performed so as to facilitate peptide quantitation and identi-fication of putative biological targets [7] The reader is directed to thefollowing review in the field of proteomics for more information [8]

A wealth of preclinically validated targets has emerged as a result of mousegenetics [9] and siRNA technology [10] For both techniques, a single geneknockout is performed, and the effect of the deletion is monitored/evaluated.Proteomics, on the other hand, generally takes a shotgun approach to identi-fying the targets that are relevant and specific to the disease Unfortunately,because many diseases are polygenic in origin and because protein pathwaysare extremely complex (e.g., intracellular protein signaling pathways [11]),proteomics has been best at identifying a short list of “candidate” proteintargets rather than a single protein target completely unique to the disease.The challenge has been to sift through all the proteins that have been identi-fied as altered in a disease state relative to healthy state, and this has provedextremely challenging

The focus of proteomics has turned to identifying potential biomarkers ofdisease A biomarker, by definition, is (a) a molecular indicator for a specificbiological property or (b) a feature or facet that can be used to measure theprogress of disease or the effects of treatment As an example, a biomarker

for Type II diabetes is higher fasting blood glucose levels relative to matched controls Another, more definitive biomarker of type II diabetes iselevated HbA1c levels For many diseases, however, the relevant biomarkersare less well understood This is especially true in the fields of oncology andinflammation research Biomarker research is a particularly intense area offocus for many pharmaceutical companies, with new departments beingformed for the purpose of identifying both preclinical and clinical biomarkers

age-to facilitate their drug discovery and development programs Like the field ofproteomics, the field of biomarker research is far too vast to warrant its reviewhere A very nice review article by the late Wayne Colburn, a pioneer in dia-betes biomarker research, describes this maturing field [12]

11.3 APPLICATIONS OF HPLC/MS/MS IN SUPPORT

OF PROTEIN CHEMISTRY

Independent of the tool used to identify the protein target, whether it be mousegenetics, siRNA technology, or proteomics, once a protein has been identified

as a suitable target for drug discovery,the next step in the drug discovery process

is to express and purify the protein (carried out combining molecular biologyand protein chemistry techniques) in sufficient quantities so as to support bio-logical screening, X-ray crystallography, and any other drug discovery studiesrequiring purified protein material.The traditional method for assessing proteinexpression and purification has been to use 1D-polyacrylamide gel elec-trophoresis 1D-PAGE is capable of separating proteins based on molecularweight and charge (pI) However, the technique is unable to provide more than

a crude assessment of protein molecular weight Recently, open-access or

walk-up LC/MS has been incorporated into protein chemistry and molecular biologylabs and has greatly facilitated confirmation of protein expression [13–15].Generic gradient LC-MS methods are used to trap and elute expressed,purified proteins by RP-HPLC/ESI/MS Open-access protein QC is a bit morechallenging than its small-molecule counterpart in that not all proteins “fly” byelectrospray ionization, identifying a “universal” HPLC method for their sep-aration can be challenging, and instrument calibration and mass accuracy are

of paramount importance We developed a fast, 5-minute protein QC methodusing a Poroshell 1-mm-i.d column and found the method to be satisfactory forthe vast majority of protein separations and analyses performed in our labora-tory To achieve adequate mass accuracy for protein molecular weight deter-minations, an external calibration with myoglobin is performed at the beginningand end of each overnight queue of protein samples so as to ensure that theinstrument calibration is maintained over the course of the batch analysis Mol-ecular weights of deconvoluted protein spectra are then compared to the pre-dicted protein molecular weight, and the results are captured graphically (in theform of a microtiter plate view) as well as in tabular format, amenable to data-base uploading, as shown in Figure 11-1

11.4 APPLICATIONS OF HPLC/MS/MS IN SUPPORT OF ASSAY DEVELOPMENT AND SCREENING

The overwhelming majority of biological assays have been developed inmicrotiter plate format (typically 96-well, 384-well, 1536-well) and with paral-lel detection methods such as fluorescence polarization The vast majority

of druggable targets, including enzymes, ligand gated ion channels, and protein-coupled receptors, are all amenable to screening in high-throughputmicrotiter plate format

G-In general, serial-based chromatographic methods, such as HPLC andHPLC/MS, are unable to compete with the high-throughput screening tech-nologies However, a small number of targets, such as those involved in medi-ating protein–protein interactions, are not well-suited to HTS methodologies.For this class of targets, HPLC coupled with mass spectrometry has proved

to be a very reliable, albeit lower throughput, alternative The technique thathas been used most widely for directly assessing protein–small molecule andprotein–protein interactions is affinity chromatography–mass spectrometry.Kassel et al [16] presented one of the first papers coupling affinity chro-matography with mass spectrometry In their work, a two-dimensionalLC/LC/MS method was developed to assess protein–ligand binding Affinitychromatography was used in the first dimension of separation, followed byreversed-phase chromatography coupled with mass spectrometry for the identification of binders Kaur et al [17] showed the power of size exclusion

Figure 11-1 Automated protein AnalysisOpenLynx LC/MS for protein molecular

weight confirmation

chromatography (SEC) coupled with reversed-phase HPLC/MS for ing ligands for a receptor derived from a 576-component combinatorial library.Today, size-exclusion columns are available in microtiter plate format,permitting higher-throughput characterization of protein–protein andprotein–ligand interactions.

identify-Berman et al [18] pioneered one of the earliest applications of HPLC insupport of assay development They showed the power of HPLC for the deter-mining preferred substrates of the enzyme collagenase, a metalloprotease.Complex mixtures (pools of 100 components each) of probe substrates for collagenase were prepared by combinatorial methods Each of the pooledlibraries was incubated with enzyme Substrate disappearance (turnover) andproduct appearance profiles were monitored by HPLC and the optimal sub-strate(s) identified Recently, Lambert et al [19] published a two-dimensionalLC/LC/MS method for the identification and optimization of substrates forTNF convertase Scientists at Nanostream, Inc., a company dedicated to high-throughput HPLC, introduced a parallel capillary LC/fluorescence method tosupport screening for kinase inhibitors Their method complements the moretraditional (and higher-throughput) fluorescence-based screening approachbut offers the advantage of chromatographic separation of phosphorylatedand unphosphorylated products, thereby reducing background interference.Another emerging role of HPLC/MS is in support of cell-based assays for which no direct measures of drug effect are possible and require indirectmethods for detection A recent publication by Clark et al highlights thepower of LC-MS for screening inhibitors of HMG-CoA reductase (a rate-determining enzyme in the cholesterol biosynthesis) [20] In addition,Thibodeaux et al [21] and Xu et al [22] reported on methods for directlyassessing the cell-based activity of inhibitors of the metabolic disease target,11β-hydroxysteroid dehydrogenase-1 (11β-HSD-1) LC-MS was used tomeasure the effect of 11β-HSD-1 inhibitors on the intracelleular conversion

of cortisol and cortisone using LC/MS/MS

Once the assay and assay format have been decided upon, the next step in thediscovery process is to initiate compound screening for the purpose of identi-fying hits or lead compounds The fundamental requirement is that the assayresults identify a collection of actives or “hits.” The definition of “hit” variesbetween organizations, but most accept the definition that the compoundshows a confirmed structure, shows a confirmed dose response, exhibits anIC50≤ 10 µM potency, and is a member of a chemotype that is amenable toanaloging and fast follow-on synthesis

What is the source of these initial actives or hits? There is a wide array ofcompound sources Generally, pharmaceutical and biotechnology organiza-tions initiate screening by accessing their internal compound repositories (so-

called corporate collections or compound archives) Often, the corporate lections are not particularly diverse but are biased to the therapeutic focus(es)

col-of the organization Consequently, the screening libraries are col-often augmented

by addition of commercially available screening libraries that are (a) family focused (e.g., GPCR-targeted libraries, kinase-targeted libraries, etc.)and/or (b) general diversity sets Further augmentation of the initial screeningactivities is to include custom synthesis compound libraries (typically pro-duced by automated high-throughput organic synthesis (HTOS) methodolo-gies, such as those described by Nikolaou et al [23]

gene-One of the challenges with compound collections is that they are historical

by nature For large Pharma, it is not uncommon for corporate collections toinclude compounds that were synthesized more than 25 years ago At the time

of synthesis, it can be presumed that the compounds met the purity criteriafor compound registration However, it can also be presumed that a high like-lihood exists that the compounds have degraded over extended storage time.Another reason for poorer quality of compound collections is attributable tothe fact that most compounds are stored as DMSO stock solutions as opposed

to storage as solid materials Storage of compounds in DMSO is done primarilyfor the reason that (a) DMSO is considered a “universal” solvent and (b) solu-tions are much easier to handle in plate-based high-throughput biologicalscreening systems However, the drawback to DMSO is that it is a very hydro-scopic solvent and unless the compounds are stored under inert conditions,they are prone to hydrolysis Kozikowski et al [24] evaluated the effect offreeze/thaw cycles on stability of compounds stored in DMSO

Until very recently, with the introduction of high-throughput analyticaltechnology, these compound sources were far too large to merit re-analysisand/or re-purification and hence were screened “as is.” The result was (and hasbeen observed frequently) that hits could not be reconfirmed during follow-

on bioassay screening, and subsequent evaluation of the compounds by niques such as HPLC/MS and NMR showed that the expected compound wasnot pure and, in some cases, was completely absent! The adage “garbage in,garbage out” became a mantra of many high-throughput screening laborato-ries and forced companies to take a much more serious look at the quality oftheir compound collections Morand et al [25] from Proctor and Gamble setout to fully assess the quality of their >500,000 compound corporate collec-tion.They achieved this goal through incorporation of a massively parallel flowinjection–mass spectrometry system, capable of analyzing a plate of samples

tech-in less than 2 mtech-inutes.The throughput of their technique was one to two orders

of magnitude faster than typical flow injection–mass spectrometry systemsused for reaction monitoring [26]

In addition to quality control over compound collections, the issue of purity

of synthetic libraries derived using combinatorial chemistry quickly cameunder the microscope In the early to mid-1990s, “combichem” became ahousehold word throughout the pharmaceutical industry and was believed to

be a key technology that would revolutionize drug discovery The basis of

combinatorial chemistry was the ability to perform split-mix synthesis on solidsupport and to take advantage of the combinatorial nature of the process togenerate vast arrays of compounds Combinatorial libraries were purported

to be pure, owing to the fact that they were synthesized on solid support andamenable to extensive washing to remove excess reagents and, therefore,directly amenable to high-throughput screening However, these combinator-ial libraries synthesized on solid support suffered from the same problems thathave long plague solution-phase synthesis—that is, the generation of unex-pected and unwanted by-products Due to the shear size of these compoundlibraries and the relatively small amounts available following resin cleavage,

it was not possible to either characterize or purify the expected products ventional split-mix combinatorial methods, though still popular with somebench chemists, have been replaced largely by the technique of directed par-allel solution and parallel solid-phase organic synthesis

CHARACTERIZATION

Combinatorial chemistry paved the way for high-throughput, parallel organicsynthesis techniques, now mainstream in the pharmaceutical and biotechnol-ogy industries for lead generation activities The ability to synthesize com-pound libraries rapidly using automated solution-phase and solid-phaseparallel synthesis has led to a dramatic increase in the number of compoundsnow available for high-throughput screening The unprecedented rate bywhich compound libraries are now being generated has forced the analyticalcommunity to implement high-throughput methods for their analysis andcharacterization

As early as 1994, groups adopted high-speed, spatially addressable mated parallel solid-phase and solution-phase synthesis of discretes [27–31].Both solution-phase and solid-phase parallel synthesis permits the production

auto-of large numbers as well as large quantities auto-of these discrete compounds,eliminating the need for extensive decoding of mixtures and re-synthesis following identification of “active” compounds in high-throughput screening

of combinatorial libraries Importantly, parallel synthesis is performed readily

in microtiter plate format amenable to direct biological screening, as wastouched upon earlier The relative ease of automation of parallel synthesis led to a tremendous in flux of compounds for lead discovery and lead optimization

Almost all of the analytical characterization tools (e.g., HPLC, NMR, FTIR,and LC/MS) are serial-based techniques, and parallel synthesis is inherentlyparallel Consequently, this led rapidly to a new bottleneck in the discoveryprocess (i.e., the analysis and purification of compound libraries) Parallel syn-thesis suffers from some of the same shortcomings of split and mix synthesis(e.g., the expected compound may not be pure, or even synthesized in suffi-

cient quantities) The analytical community was faced with the decision of how

to analyze these parallel synthesis libraries

The traditional method for assessing compound purity has been to performthe following: Purify the desired product to homogeneity by crystallization orcolumn chromatography (e.g., RP- or NP-HPLC), acquire a 1D-NMR and 2D-NMR spectrum on the isolated product, obtain confirmatory molecular weightinformation by mass spectrometry, perform a C, H, and N combustion analy-sis, generate an exact mass measurement (to within 5 ppm of the expectedmass) by high-resolution mass spectrometry, and determine the amount of iso-lated product by weight—all prior to compound submission and biologicalscreening In the era of high-throughput compound library synthesis, however,this extensive characterization is simply not possible Therefore, groups havefocused principally on a limited number of analytical measurements for compound identity and purity—in particular, LC/MS analysis incorporatingorthogonal detection methods, such as UV and evaporative light scatteringdetection (ELSD) and flow-probe 1D-NMR [32] The most commonlyemployed technique for characterizing compound libraries is to incorporateLC/MS with electrospray or atmospheric pressure chemical ionization with

UV and ELSD and, more recently, photoionization [33]

LC/MS emerged as the method of choice for the quality control assessment

to support parallel synthesis because the technique, unlike flow injection massspectrometry, provides the added measure of purity (and quantity) of the com-pound under investigation In addition, “universal-like” HPLC gradients (e.g.,10% to 90% acetonitrile in water in 5 minutes) have been found to satisfy theseparation requirements for the vast majority of combinatorial and parallelsynthesis libraries Fast HPLC/MS has been found to serve as good surrogate

to conventional HPLC for assessing library quantity and purity [34–37] FastHPLC/MS is simple in concept It involves the use of short columns (typically4.6 mm i.d.× 30 mm in length) operated at elevated flow rates (typically 3–

5 mL/min)

Typically, short columns are used for compound analysis because they allowfor fast separations to be carried out at ultrahigh flow rates Also, thesecolumns tend to be more robust than narrow bore columns (1-mm and 2-mmi.d.) (i.e., less clogging is experienced and longer lifetimes are observed whenthese columns are subjected to unfiltered chemical libraries) A typical LC/MSanalysis consists of injection a small aliquot (10–30µL) of the reaction mixture(total concentration of 0.1–1.0 mg/mL) and performing the separation using a

“universal” gradient of 10–90% Buffer B in 2–5 minutes Buffer A is typicallyH2O containing 0.05% trifluoroacetic acid (or formic acid), and Buffer B istypically acetonitrile containing 0.035% trifluoroacetic (or formic acid) HPLCcolumns are operated typically at flow rates of 3–5 mL/min (depending ontheir dimensions), and the cycle time between injections is 3–5 minutes

An example of a fast LC/MS analysis of a combinatorial library nent is shown in Figure 11-2 Fast LC/MS run times incorporating these shortcolumns is typically between 3 and 5 minutes including re-equilibration

Recent reports by Kyranos et al [38] suggest that “pseudo-chromatography”(in essence, step elution chromatography) provides a more rapid and reliableassessment of the quality of library synthesis than methods such as flow injec-tion mass spectrometry.

11.6.1 Purity Assessment of Compound Libraries

The issue of compound purity has received a great deal of attention over thelast several years as more and more chemists have adopted high-throughputorganic synthetic protocols but are unwilling to compromise the quality of themolecules submitted for biological evaluation The general consensus targetpurity of a compound library compound before it is to be archived or screenedfor biological activity is between 90% and 95% pure This purity criterion ismore stringent than in the past, where 85–90% (based on UV detection) wasconsidered acceptable This may be attributed primarily to a shift towardsmaller, focused (or biased) libraries than larger, diverse collections of com-pounds The majority of mass spectrometry manufacturers now offer softwarepackages that aid in the automatic determination of purity

UV chromatograms are typically used, rather than the total ion currentchromatogram, to assess purity This is because the total ion current chro-matogram is a measure of a compound’s “ionizability,” which is well known tovary dramatically from one compound to the next Orthogonal detectionmethods, such as chemiluminescence nitrogen detection (CLND) [39] andELSD [40, 41], have been proposed to be more universal detection methodsthan UV and hence are being used with increasing frequency to assess reac-tion yields and purity CLND, as indicated from its name, measures the amount

of nitrogen in a sample In this method, a compound is transferred to a

high-Figure 11-2 (A) A 4-minute HPLC/MS separation of a solution-phase parallel

synthesis library The gradient profile for fast HPLC/MS was 10–90% acetonitrile

in H2O in 4 minutes with a 1-minute equilibration time (B) A 1-minute, total cycle timechromatographic separation of the same crude product (Reprinted from reference 42,with permission.)

temperature oxygen reaction chamber (set to 1000°C) whereby the compoundundergoes rapid decomposition to form nitrous oxide (NO) The liberated NOreacts with ozone (O3) to form metastable NO2, which is selectively detected

by release of a photon

CLND has been demonstrated to be a valuable tool for quantifying lowquantities of material and has been shown to be particularly well-suited to NP-HPLC and SFC-MS, for the principal reason that separations are carried outusing solvents that do not contain nitrogen (i.e., CO2 and CH3OH) ELSDmeasures the mass (quantity) of the material directly, is often presented asbeing a molecular-weight-independent detector, and is a tool that has gainedwide-scale acceptance for on-line quantification of compound libraries Anexample of a separation of a four-component library incorporating UV, ELSD,CLND, and MS detection is shown in Figure 11-3 Using these various detec-tors, the chemist is able to obtain measures of purity of their library withgreater confidence than when relying solely upon LC/UV/MS data

An example of automated purity assessment of a compound analyzed

by LC/UV/MS is shown in Figure 11-4 In this example, purity is assessed attwo different wavelengths,λ220andλ254 Excel macros are used for automated

Figure 11-3 Column flow rate was 5 mL/min A portion of the column effluent was

split to each of the three detectors (CLND, 200µL/min; ELSD, 200 µL/min; MS,

100µL/min) A make-up flow of 50/50 MeOH/H2O (300µL/min) was added to the flowstream diverted to the mass spectrometer ion source Mass spectra were acquired usingelectrospray ionization with no special modifications to the ion source (A) Total ioncurrent chromatogram showing two of the four components ionize efficiently underelectrospray ionization conditions (B) ELSD chromatogram of the four components,all showing comparable response (C) UV chromatogram (254 nm) shows some selec-tivity in detection as does (D) CLND detection

post-data acquisition processing with associated graphical representations ofdata to facilitate analysis For libraries generated in microtiter plate format,the results of each individual well may be color-coded to reflect relative

degrees of purity.

More often, as described earlier, compound purity is reported taking intoaccount the purities determined from the UV, ELSD, and CLND detectors Insome instances, purity assessment has been made based on the intensity of theexpected ion in the mass spectrum relative to the sum of the intensities of allions in the spectrum This method, however, is only a very crude estimate ofpurity, because ionization efficiencies for compounds can vary widely withinand between classes of compounds Though LC/MS (with UV and/or ELSDdetection) has been adopted as the method of choice for assessing the qualityand quantity of material prepared by parallel synthesis techniques, a decisionstill needs to be made by each respective organization as to what constitutesacceptable quality before submitting a sample for biological testing

Figure 11-4 Purity assessment is a critical component in the decision process by the

chemist as to whether their isolated compound is of sufficient quality to be submittedfor compound registration and biological testing To facilitate automated and rapidpurity assessment of compound libraries, applescripts and visual basic scripts are used.(A) Total ion current chromatogram shows two components (B) Extracted ion chro-matogram for the expected product identifies its retention time (C) Mass spectrumobserved for the expected product (D) UV 220-nm chromatogram indicates theexpected product is approximately 75% pure (E) UV 254-nm chromatogram indicatesthe expected product is approximately 66% pure

Some groups have evaluated ultra-fast chromatography separations called ballistic, pseudo-chromatography) in order to provide a snapshot of thesample purity [42, 43] The major drawback to the ballistic chromatographytechnique is that column resolution is reduced when operating at these sub-optimal linear velocities Also, the pseudo-chromatography approach is bestsuited to applications where purity assessment is secondary to rapid com-pound profiling.

Historically, it was believed that solid-phase synthesis protocols eliminate theneed for purification because excess reagents are removed readily by exten-sive washing Unfortunately, even for solid phase peptide synthesis, final prod-ucts, acid-cleaved from the resin are found to be far from pure Furthermore,parallel solution-phase synthesis has found greater popularity, because it isreadily automated and extends the “portfolio” of reactions available to thechemist for high-throughput parallel synthesis The limitations with solid-phase synthesis and the movement toward parallel solution phase synthesisare forcing numerous groups to evaluate and implement a variety of purifica-tion strategies

A prevailing assumption is that if the chemistry is sufficiently high-yieldingduring the process development phase of synthesis, then it is reasonable toexpect comparably high yields during the production phase of synthesis Inprocess development, a subset of the total library to be synthesized is rigor-ously optimized to maximize reaction yield During production, it is assumedthat the vast majority of members of the library will behave similarly and thatthe desired product will be the major component in the well The reality is thatfar too often, the biological activity cannot be tracked to a single component

or, in some instances, to the expected product in the well Groups attempting

to elucidate the active component(s) of the well have expended significanteffort, only to find that the activity does not correlate with a single componentwithin the sample Consequently, more and more groups have embraced thevalue of “quality in, quality out” and are now applying the same analyticalrigor to parallel synthesis chemical products as they have for more classicalmedicinal chemistry synthesis These activities have enhanced the quality ofstructure–activity relationships (SAR) and structure–inactivity relationships(SIR) that can be derived from the assaying of these compounds for biologi-cal activity

Numerous techniques are available to the organic chemist to supportlibrary purification Zhao et al [44] published an extensive review on com-pound library purification strategies, including HPLC, liquid–liquid extraction,solid-supported liquid–liquid extraction, solid-phase extraction, ion-exchangechromatography, and countercurrent chromatography, among others This

review focuses exclusively on HPLC- and HPLC/MS-based purificationmethods.

11.7.1 UV-Directed Purification

Both activity and inactivity data are being used increasingly to generate SARand direct subsequent synthetic efforts Consequently, organizations have rec-ognized the importance of confirming the purity of compounds prior to screen-ing, and not only those compounds for which activity is observed In order tominimize false positives and false negatives, it is advantageous to assay onlyhigh-quality compounds Therefore, great effort has been devoted to the devel-opment of automated purification technology designed to keep pace with the

output of high-throughput combinatorial/parallel synthesis.

Automated methods are now available to the chemist to perform throughput purification Although HPLC has long been a method available tothe chemist for product purification, only recently have these systems beendesigned for unattended and high-throughput operation.Weller et al [45] wereone of the first groups to demonstrate “walk-up” high-throughput purification

high-of parallel synthesis libraries based on HPLC and UV detection An architecture software interface enabled chemists to select the appropriate separation method from a pull-down menu and initiate an unattended automated reversed-phase UV-based fraction collection Fractionation wasachieved using a predetermined UV threshold Multiple fraction collectorswere daisy-chained in order to provide a sufficient footprint for fraction col-lection Since the early work of Weller et al., a number of commercial systemshave been introduced for walk-up preparative LC/UV purification (includingGilson, Hitachi, and Shimadzu, to name a few)

open-One of the challenges associated with UV-based purification systems is that multiple fractions are collected for every sample injected Although user-defined adjustable triggering parameters (e.g., UV thresholds for initiating andterminating fraction collection) can be used to reduce the total number of frac-tions, all, to some extent, will contain impurities The exact number of chro-matographic peaks for a given sample will be hard to predict, and thereforethe footprint for fraction collection will be difficult to predict Experience hasshown that it is not uncommon for 5–10 fractions to be collected per injection.When purifying only a small number of samples (<10), it is neither particu-larly challenging to collect the fractions nor challenging to perform post-purification analysis so as to identify the fraction(s) containing the desiredproduct However, when attempting to purify compound libraries (e.g., 96-wellplates of samples), the number of fractions and the time it takes to identifythe relevant fraction(s) fast becomes a bottleneck Schultz et al [46] addressedthe fraction collection issue and streamlined post-fraction collection process-ing (including evaporation, re-constitution and post-purification analysis) bycollecting fractions directly into 48-well microtiter plates Their method was

particularly well-suited to semipreparative purifications (using inner-diameter columns to support low milligram quantities).

smaller-In order to gain further efficiencies into UV-based purification of pound libraries, numerous groups have developed automated high through-put UV-based purification systems coupled with on-line mass spectrometricdetection Kibbey [47] was one of the first scientists to implement a fully automated preparative LC/MS system for combinatorial library purifi-cation His approach was to perform a scouting analytical run prior to purification so as to optimize chromatographic method and fraction collectionparameters Fraction location and molecular weight information were captured through a custom LIMS system The added mass spectrometric information greatly facilitated deconvoluting of collected fractions andstreamlined their purification process Hochlowski [48] describes a service-based purification factory incorporating UV and ELSD detection coupledwith mass spectrometry that supports purification of over 200 compounds perday

com-More recently, intelligent UV-based systems for preparative scale tion of combinatorial libraries have been introduced, utilizing knowledge ofretention time of the expected product based on a pre-analytical evaluationfollowed by preparative HPLC with UV-based fractionation using a narrowtime collection window Yan et al [49] coined the “accelerated retentionwindow” method as a tool for improving high-throughput purification effi-ciency In their method, a high-throughput parallel LC/MS analysis is per-formed prior to preparative purification to confirm that the expected product

purifica-is indeed contained within the well and to identify the approximate retentiontime of the expected synthetic product Only those compounds found to be

≥10% pure based on the analytical run are candidates for final product cation Furthermore, the information from the high-throughput parallel analy-sis was uploaded to a stand-alone preparative LC system for final productpurification Fraction collection was initiated using an accelerated retentiontime window method so as to accelerate the preparative HPLC analysis Addi-tional refinements of UV-based purification strategies have been maderecently, allowing for further simplification of the fraction collection and post-purification analysis step In one embodiment, Karancsi et al [50] implemented

purifi-a “mpurifi-ain component” frpurifi-action collection method bpurifi-ased on UV-triggering thpurifi-atsupports the “holy grail” of high-throughput purification, that being the onecompound/one fraction concept [50]

11.7.2 Mass-Directed Preparative Purification

The technique of preparative LC/MS, introduced in the late 1990s was the first technique to greatly simplify the purification process For the first time,preparative LC/MS (PrepLCMS) methods allowed the concept of one com-pound/one fraction to be realized [51–55] In the Prep LC-MS mode, the mass

spectrometer is used in this mode as a highly selective detector for directed fractionation and isolation This technique provides a means forreducing dramatically the number of HPLC fractions collected per sample andvirtually eliminates the need for post-purification analysis to determine themass of the UV-fractionated compound Preparative LC/MS is now widelyincorporated in the pharmaceutical industry Systems for preparative LC/MSare configured in numerous ways and are operated in numerous ways, includ-ing an expert user mode, walk-up or open access mode, or a project teamsetting, supporting small teams of chemists working on similar chemistries Allcomponents of the system are under computer control and are hence trulyautomated Components of these systems are nearly identical to stand-aloneHPLC systems with the addition of a flow splitter device to divert a smallportion of the column flow to the mass spectrometer for on-line detection andfraction collector triggering Typical systems are configured in an automatedanalytical/preparative mode of operation In this configuration, the chemist isable to select between a variety of column sizes for either analytical, semi-preparative, or preparative separations The HPLC, switching valves, massspectrometer, and fraction collector are under complete computer control, asshown in Figure 11-5 In some instances, a solvent pump is added to deliver amethanol make-up flow to the mass spectrometer The flow splitter and extrasolvent pump serve the primary purpose of reducing the potential for over-loading of sample into the ion source An advantage of the flow splitter andmake-up pump is that it reduces the trifluoroacetic acid (ion pairing) in the

mass-Figure 11-5 Preparative LC/MS systems on the market consist of a binary HPLC

system, a combined autosampler/fraction collector (footprint of a Gilson 215inject/collect liquid handler shown in the figure), and a single quadrupole mass spectrometer

ion source, which can affect the sensitivity of detection for acidic library components.

An example of a mass-guided fractionation of a combinatorial library isshown in Figure 11-6 In this example the crude reaction product is only about30% pure The component of interest shows a prominent single chromato-graphic peak when monitoring specifically for its corresponding mass Post-purification analysis of this singly isolated fraction (based on mass-directedfractionation) demonstrates that the compound of interest was purified togreater than 90% Had a UV-based fractionation system been used in this par-ticular example, at least five individual fractions would have been isolated.Extending this to a 96-component library synthesized in microtiter plateformat (and assuming this compound was representative of the quality of themembers of the library), a UV-based approach would have led to approxi-mately 400–500 fractions requiring reanalysis to pinpoint the desired product.This not only would be a time-consuming reanalysis process but also wouldrequire significant time to transfer the appropriate fractions to a screeningplate for biological assessment

Debate exists as to whether UV-based or mass-based fraction collection isthe more appropriate tool for purifying compound libraries The choice of

Figure 11-6 Fifty milligrams of a crude reaction product was solubilized in 1 mL of

50/50 MeOH/DMSO and injected onto a 20-mm × 50-mm-i.d reversed-phase column.Separation was achieved using a gradient of 10–90% ACN in 7 minutes (A) TIC chro-matogram shows five components well-separated (B) Extracted ion chromatogram(XIC) for expected product shows a single, prominent peak at 6.49 minutes Fractioncollection was initiated and terminated, as indicated by the arrows directly below theXIC peak (C) Post-purification analysis of the isolated component shows that the com-pound was purified to approximately 90% level

technique probably should be governed by the relative importance of anygiven sample and the purification throughput requirements at any moment intime As a simple rule of thumb, during the earlier stages of the discoveryprocess, where larger numbers and smaller quantities (<50 mg) of compoundare needed for biological screening and early ADME/PK screening, a mass-spectrometry-based fraction collection system probably makes the most sense,since the total number of isolated fractions can be reduced to a minimum Atlater stages of a discovery program (e.g., during late stage lead optimization,where a smaller number and larger (≥100 mg) quantities of compounds are

being evaluated for in vivo efficacy), UV-based methods might take priority.

Independent of the debate, it is widely agreed that mass spectrometry serves

as a highly sensitive and selective detector for analysis and purification of pound libraries Mass-triggered fraction collection enables compound libraries

com-to be purified based solely on their expected product mass Libraries can bepurified maintaining a “one compound-one fraction” model, which facilitatessample tracking, registration, and biological testing and enables screeningresults to be readily correlated with synthetic structure

11.8.1 Fluorous Split-Mix Library Synthesis and Preparative

LC/MS De-Mixing

The interest in combinatorial chemistry (split-mix technology) as a means forgenerating large compound collections plummeted over the past several years,due primarily for the reasons of poor quality control over synthesis andcomplex decoding strategies Conventional split-mix technology was replaced

by a number of parallel synthesis strategies that did not require the de-mixing

or de-coding step Recently, a relatively new technique called fluorous thesis, developed by Luo et al [56], offers an interesting twist on the conven-tional split-mix combinatorial chemistry approach In fluorous synthesis, amixture of substrates is paired with a series of perfluoroalkyl (Rf) phase tagsand is taken through multiple synthetic reaction steps culminating in a mixture

syn-of tagged products Demixing is achieved by tag-controlled fluorous HPLCpreparative separation, the fractions are stripped of solvent, and the tag isremoved to yield individually purified and structurally distinct products Arecent publication by Zhang et al [57] describes both the synthetic and ana-lytical HPLC strategy for a 420-component fluorous library Members of Syrrx’analytical team, in collaboration with Zhang and co-workers, recently devel-oped a higher throughput mass-directed purification method to support purifi-cation of their fluorous tagged libraries [58] Figure 11-7 shows the UV andtotal ion current chromatogram (TIC) for a pool of five compounds from the 420-component library Separation of the five-component mixture wasachieved in less than 6 minutes Fraction collection was initiated when the

expected [M + H]+ion for each of the tagged products exceeded the pre-setion intensity threshold, and fraction collection was terminated when the ionintensity for the expected product(s) dropped below a second pre-set ionintensity threshold value Combining split-mix fluorous synthesis with high-speed chromatography provides a means for rapidly generating large numbersand large quantities of highly purified druggable molecules.

11.8.2 Parallel Analysis and Parallel Purification

The synthetic throughput achievable by the medicinal chemist (havingadopted parallel synthesis strategies) has rendered analysis and purificationone of the key (and possibly rate-limiting) steps in the discovery process.Although advances in sample analysis throughput have been clearly demon-strated, there is a limit as to how fast a separation and analysis can be achieved

Figure 11-7 A five-component fluorous split-mix crude reaction mixture was injected

onto a 20-mm × 50-mm-i.d reversed-phase column (A) UV chromatogram and (B)Total ion current chromatogram Compounds were purified using mass-directed fraction collection (peaks highlighted)

while maintaining good separation efficiency and quality analysis Two techniques that have been developed to increase throughput without com-promising column chromatography are (a) rapid column switching and regen-eration systems for enhanced-throughput serial-based analysis and (b) parallelchromatography methods A simple and elegant modification of the LC/MSmethod is to incorporate a set of switching valves and a third pump to reducecycle time between injections, as shown in Figure 11-8 While one column isbeing used to perform the LC/MS analysis, the other column is being regen-erated.An alternative use of 10-port switching valves is to allow for rapid serialsampling between columns This technique works well for samples that areamenable to either isocratic or step elution While one sample is being loadedonto one column, the contents of the other column are eluted into the ionsource.

In order to increase sample throughput while maintaining high-quality analytical data, groups have begun to perform separations in parallel [59–63].Numerous groups have independently developed parallel sample introductiontechniques, although the MUX ion source from Micromass/Waters is the onlyone commercially available By performing analyses in parallel, chromato-graphic integrity can be maintained while effectively addressing samplethroughput Di Biasi et al [59] and Wang et al [60] presented novel ion sourceinterfaces enabling four to eight samples to be processed in parallel, therebyincreasing the sample analysis throughput dramatically over conventional,serial-based LC/MS analyses Commercially available parallel spray interfacesconsist of a multiple spray head assembly and a blocking device (e.g., rotatingplate), enabling individual sprayers to be sampled at specific and defined time

Figure 11-8 Schematic representation of a column switching configuration to support

analysis from one column while the second column is equilibrating