Journal of Chromatography B, 782 (2002) 27–55
`
`www.elsevier.com/locate/ chromb
`
`Review
`A decade of high-resolution liquid chromatography of nucleic acids
`on styrene–divinylbenzene copolymers
`
`*
`, Christian G. Huber
`Peter J. Oefner
`aGenome Technology Center, Stanford University,855 California Avenue, Palo Alto, CA 94304,USA
`b
`¨
`Instrumental Analysis and Bioanalysis,University of the Saarland,66123 Saarbrucken, Germany
`
`a ,
`
`b
`
`Abstract
`
`The introduction of alkylated, nonporous poly-(styrene–divinylbenzene) microparticles in 1992 enabled the subsequent
`development of denaturing HPLC that has emerged as the most sensitive screening method for mutations to date. Denaturing
`HPLC has provided unprecedented insight into human origins and prehistoric migrations, accelerated the cloning of genes
`involved in mono- and polygenic traits, and facilitated the mutational analysis of more than a hundred candidate genes of
`human disease. A significant step toward increased sample-throughput and information content was accomplished by the
`recent introduction of monolithic poly(styrene–divinylbenzene) capillary columns. They have enabled the construction of
`capillary arrays amenable to multiplex analysis of fluorescent dye-labeled nucleic acids by laser-induced fluorescence
`detection. Hyphenation of denaturing HPLC with electrospray ionization mass spectrometry, on the other hand, has allowed
`the direct elucidation of the chemical nature of DNA variation and determination of phase of multiple alleles on a
`chromosome.
` 2002 Elsevier Science B.V. All rights reserved.
`
`Keywords: Reviews; Nucleic acids; Styrene–divinylbenzene copolymers
`
`Contents
`
`1 . Introduction ............................................................................................................................................................................
`2 . Development of a micropellicular polymeric sorbent for ion-pair reversed-phase chromatography of nucleic acids .........................
`3 . Operational variables in ion-pair reversed-phase chromatography of nucleic acids........................................................................
`4 . Applications of non-denaturing ion-pair reversed-phase HPLC ...................................................................................................
`5 . Denaturing high-performance liquid chromatography (DHPLC) .................................................................................................
`5 .1. Behavior of double-stranded DNA at elevated column temperatures ...................................................................................
`5 .2. Partially denaturing HPLC...............................................................................................................................................
`5 .3. Sensitivity and specificity of DHPLC ...............................................................................................................................
`5 .4. DHPLC in comparison to other methods commonly used in mutation analysis ....................................................................
`6 . Completely denaturing HPLC ..................................................................................................................................................
`7 . Hyphenation of ion-pair reversed-phase HPLC and electrospray ionization mass spectrometry......................................................
`8 . Multiparallel DHPLC ..............................................................................................................................................................
`9 . Applications of DHPLC...........................................................................................................................................................
`
`28
`28
`30
`31
`32
`32
`33
`33
`35
`36
`37
`39
`39
`
`*Corresponding author. Tel.: 11-650-812-1926; fax: 11-650-812-1975.
`E-mail address: oefner@genome.stanford.edu (P.J. Oefner).
`
`1570-0232/02/$ – see front matter  2002 Elsevier Science B.V. All rights reserved.
`PII: S1570-0232( 02 )00700-6
`
`1
`
`MTX1012
`
`

`

`28
`
`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`9 .1. Identification of biallelic polymorphisms on the Y-chromosome .........................................................................................
`9 .2. Gene mapping by DHPLC ...............................................................................................................................................
`9 .3. Mutational analysis of genes ............................................................................................................................................
`1 0. Conclusions ..........................................................................................................................................................................
`Acknowledgements......................................................................................................................................................................
`References ..................................................................................................................................................................................
`
`39
`43
`44
`48
`49
`49
`
`1 . Introduction
`
`Ten years ago, it was considered a ‘‘no-brainer’’
`that capillary electrophoresis, due to its separation
`power and amenability to laser-induced fluorescence
`detection, would dominate the analysis of nucleic
`acids. And indeed,
`it was capillary array electro-
`phoresis that enabled the completion of a human
`genome draft sequence ahead of time. During the
`same period of time, high-performance liquid chro-
`matography (HPLC) found slowly but steadily its
`way into clinical and molecular biology laboratories
`mainly for the purpose of mutation screening [1].
`The reason for that has been that individual genomes
`of a species rarely differ
`from each other and,
`therefore,
`the labor and expense associated with
`sequencing would yield nothing more than confirma-
`tion of the identity of two sequences. This has led to
`the development of numerous physical, chemical,
`and enzymatic methods for mutation screening over
`the last
`two decades [2,3]. The majority of the
`techniques provide little more than information about
`the mere presence or absence of sequence differ-
`ences, the nature and location of which have to be
`established by conventional sequencing. But screen-
`ing for mutations pays off only if the technique used
`matches the sensitivity and specificity of sequencing.
`The latter is realized by most methods, but no
`technique until the introduction of denaturing HPLC
`came close to yielding a mutation detection sensitivi-
`ty of 100%.
`review provides a comprehensive
`The present
`account of the development of the stationary phase
`employed in denaturing high-performance liquid
`chromatography and the various applications of the
`technology. It also describes recent developments
`that have led to a significant
`increase in sample
`throughput and information content of HPLC of
`nucleic acids that may challenge the monopoly of
`capillary electrophoresis in DNA sequencing.
`
`2 . Development of a micropellicular polymeric
`sorbent for ion-pair reversed-phase
`chromatography of nucleic acids
`
`Progress builds usually on prior accomplishments.
`This is also true for the considerations and experi-
`ments that led eventually to the synthesis of mi-
`cropellicular
`alkylated
`poly(styrene–dinvinylben-
`zene) (PS–DVB) particles that have remained until
`this very day the standard in HPLC of nucleic acids
`[4–6]. Of all the chromatographic modes utilized for
`the separation of nucleic acids prior to 1992, includ-
`ing size-exclusion [7,8], anion-exchange [9,10], ion-
`pair reversed-phase [11,12], mixed-mode [13], and
`slalom chromatography [14], anion-exchange chro-
`matography was by far the most prominent. Per-
`formed on micropellicular polymeric support materi-
`als that carried such functional groups as diethyl-
`aminoethyl and trimethylammonium, anion-exchange
`chromatography enabled the separation of DNA
`restriction fragments with high resolution within a
`few minutes [15].
`Using sodium chloride gradients to elute the
`adsorbed nucleic acids, retention in anion-exchange
`chromatography was observed to be dependent pri-
`marily on the size of the DNA fragments and,
`secondarily, on their AT-content with those par-
`ticularly rich in AT-base pairs being retained longer
`than fragments of identical size but richer in GC-
`base pairs. More size-dependent separations were
`obtained by applying a gradient of tetraethylam-
`monium chloride [16]. This was attributed to a local
`reduction in surface potential through the preferential
`binding of positively charged tetraethylammonium
`ions to AT-rich sequences that normalized their
`binding to the anion-exchanger.
`It was the very use of tri- and tetraalkylammonium
`salts as ion-pair reagents for ion-pair reversed-phase
`HPLC that allowed the separation of double-stranded
`DNA fragments according to size [11]. The volatile
`
`2
`
`

`

`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`29
`
`nature of triethylammonium acetate and the hydro-
`organic solvents used for elution also carried the
`advantage that
`they could be simply removed by
`evaporation, while removal of the inorganic salts in
`nucleic acid fractions collected by anion-exchange
`chromatography requires more elaborate methods
`such as precipitation, dialysis, ultrafiltration, or size
`exclusion chromatography. Nevertheless,
`ion-pair
`reversed-phase HPLC was not as popular as anion-
`exchange chromatography prior
`to 1992, as the
`porous character of commercially available station-
`ary phases necessitated the use of low flow-rates and
`shallow gradients to compensate for the slow mass
`transport of DNA fragments within the chromato-
`graphic packing material due to their low diffusivity
`[11]. Consequently, separations lasted hours rather
`than minutes.
`it was
`From the aforementioned observations,
`obvious that the use of a micropellicular, reversed-
`phase column packing material should accelerate the
`analysis. But the application of commercially avail-
`able silica-based, alkylated, non-porous stationary
`phases proved less than satisfactory. The reasons for
`that were twofold. One is the contamination of silica
`with metal cations that causes secondary interactions
`leading to significant peak broadening and reduced
`recovery. This can be alleviated by the addition of
`EDTA to the mobile phase, but it does not address a
`second concern stemming from the requirement to
`use a mobile phase of neutral or alkaline pH to allow
`the electrostatic interaction between the dissociated
`phosphodiester groups of the sugar-phosphate back-
`bone of DNA and the positively charged trialkylam-
`monium ions adsorbed at the interface between the
`nonpolar stationary phase and the hydro-organic
`mobile phase. Since silica is chemically unstable at
`elevated pH-values, and even more so at elevated
`temperatures, column lifetime tends to be short with
`resolution deteriorating rapidly over the course of a
`few hundred runs, although the use of a scavenger
`precolumn that saturates the mobile phase entering
`the analytical column with silicic acid, hence protect-
`ing the latter from dissolution, and a protective
`polymer-coating can slow down such deterioration of
`separation efficiency.
`The logic response to those limitations of silica-
`based microspheres was the synthesis of alkylated,
`nonporous PS–DVB particles with a diameter of 2
`
`the PS–DVB particles
`mm [4–6]. Alkylation of
`proved essential for resolving DNA restriction frag-
`ments larger than 100 bp (Fig. 1a) and was believed
`to be a consequence of the effective shielding of the
`aromatic rings of the PS–DVB support material,
`which might otherwise interact with the DNA.
`However, recent observations with monolithic PS–
`DVB supports appear to invalidate that explanation,
`at least in part, as similar to even better separation of
`
`Fig. 1. High-resolution capillary IP-RP-HPLC separation of a
`mixture of double-stranded DNA fragments in capillary columns
`packed with (a) PS–DVB-C
`particles and (b) a PS–DVB
`18
`monolith. Column, (a) PS–DVB-C , 2.1 mm, 6030.20 mm I.D.,
`18
`(b) PS–DVB monolith, 6030.20 mm I.D.; mobile phase, (A) 100
`mM TEAA, pH 7.0, (B) 100 mM TEAA, pH 7.0, 25% acetoni-
`trile; linear gradient, (a) 37–55% B in 6.0 min, 55–63% B in 14.0
`min; flow-rate, 2.0 ml/min; (b) 35–55% B in 6.0 min, 55–78% B
`in 14 min; flow-rate, 2.2 ml /min; temperature, 50 8C; detection,
`UV, 254 nm; sample, |4 ng pBR322 DNA-HaeIII and |4 ng
`pBR322 DNA-MspI digest. Reproduced with permission from
`Refs. [65,253], respectively.
`
`3
`
`

`

`30
`
`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`nucleic acids could be accomplished without alkyla-
`tion (Fig. 1b) [17]. The reason for the difference in
`separation efficiency between non-alkylated mi-
`cropellicular and monolithic PS–DVB remains to be
`elucidated, but it is interesting to note that mono-
`lithic PS–DVB, most
`likely as a consequence of
`differences in polymerization conditions,
`is more
`hydrophilic than particulate PS–DVB as lower con-
`centrations of acetonitrile are required for elution of
`the adsorbed nucleic acids.
`
`3 . Operational variables in ion-pair reversed-
`phase chromatography of nucleic acids
`
`In ion-pair reversed-phase HPLC, separation of
`nucleic acids is achieved by electrostatic interactions
`between the positive surface potential generated by
`the amphiphilic tri- or
`tetraalkylammonium ions
`adsorbed at the stationary phase and the negative
`surface potential generated by the dissociated phos-
`phodiester groups of the sugar-phosphate backbone
`of nucleic acids [18]. Additional solvophobic interac-
`tions are minimal
`in the case of double-stranded
`DNA because of its highly hydrophilic outer surface.
`At column temperatures up to 50–55 8C, the elution
`order is strictly related to the length of the double-
`stranded DNA molecule (Fig. 2a, 53.6 8C), whereas
`chromatographic retention becomes sequence-depen-
`dent at
`temperatures above 55 8C due to partial
`denaturation (Fig. 2a, 62.5 8C). The retention times
`of single-stranded DNA and RNA molecules, on the
`contrary, are strongly sequence-dependent at all
`practicable column temperatures as a result of addi-
`tional hydrophobic interactions of the nucleobases
`with the non-polar stationary phase (Fig. 2b). The
`degree of such hydrophobic interactions can be
`modulated by the choice of ion-pair reagent [19].
`When using tetramethylammonium ion and, al-
`though to a lesser degree, tetraethylammonium ion,
`as ion-pair reagent, phosphorylated oligonucleotides
`were retained shorter than their nonphosphorylated
`analogs. Upon addition of tetrapropyl- and, even
`more so,
`tetrabutylammonium ion,
`the order of
`elution reversed in accordance with the hypothesis
`that retention of oligonucleotides is a direct function
`of the number of negative charges, with the addition-
`al phosphomonoester at the 59 terminus providing
`
`Fig. 2. Dependence of chromatographic retention on molecular
`size of
`(a) double-stranded DNA under non-denaturing and
`denaturing conditions and of (b) single-stranded DNA under
`denaturing conditions. Column, DNA-Sep姠, 5034.6 mm I.D.;
`mobile phase, (A) 0.1 M triethylammonium acetate, pH 7.0, (B)
`0.1 M triethylammonium acetate, 20% acetonitrile, pH 7.0; linear
`gradient from (a) 35–95% B in 30 min, (b) 0–50% B in 30 min;
`flow-rate, (a) 0.75 ml/min, (b) 0.8 ml/min; temperature, (a) 53.6
`and 62.5 8C, (b) 80 8C; sample, (a) 0.41 mg pBR322 DNA MspI
`digest, (b) pd(G)
`, pd(C)
`, pd(A)
`, pd(T)
`, 0.2 mg
`12 – 18
`12 – 18
`12 – 18
`each. Data from Refs. [26,59], respectively.
`
`12 – 18
`
`one or two more negative charges capable of forming
`ion pairs. The observation confirmed a previous
`finding that the distribution coefficients of aliphatic
`alcohols, carboxylic acids, and aldehydes increase
`exponentially, independently of the functional group
`attached, with an increase in the number of methyl-
`ene groups in the alkyl chain, and that the chain
`length must exceed four carbon atoms to obtain
`complete coverage [20]. Hence, one may assume that
`in the case of tetramethyl- and tetraethylammonium
`ion only partial coverage of the stationary phase is
`obtained, which consequently will retain to various
`degrees its hydrophobic properties.
`
`4
`
`

`

`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`31
`
`While the nature of chromatographic retention by
`electrostatic forces in ion-pair reversed-phase HPLC
`is similar to that of anion-exchange HPLC, elution of
`the adsorbed nucleic acids in ion-pair reversed-phase
`chromatography is not obtained by a gradient of
`increasing ionic strength, but by an increase in the
`concentration of organic solvent in the mobile phase
`resulting in the joint desorption of nucleic acids and
`amphiphilic ions. While regeneration and equilibra-
`tion times between injections are generally short, of
`the order of 30–60 s, sufficient equilibration of a
`newly packed column over at least 60 min is crucial
`for obtaining high-performance separations. Continu-
`ous increases in separation efficiency have been
`observed up to 20 h after the beginning of initial
`equilibration [21].
`Triethylammonium acetate has been the preferred
`ion-pairing reagent because it yields the best res-
`olution of double-stranded DNA fragments on
`alkylated PS–DVB particles and it is relatively easy
`to remove by evaporation [5]. The use of eluents
`containing trialkylammonium salts with alkyl groups
`longer than ethyl resulted in longer retention of DNA
`accompanied by deterioration of
`separation ef-
`ficiency and peak shape [18]. Interestingly, in con-
`trast
`to single-stranded nucleic acid [19],
`tetra-
`alkylammonium salts were not applicable to the
`separation of double-stranded DNA fragments.
`More recently, in conjunction with on-line electro-
`spray ionization mass spectrometry (ESI-MS) of
`nucleic acids, butyldimethylammonium bicarbonate
`was found to improve detection sensitivity [22]. This
`correlates well with the observation that mass spec-
`trum quality is enhanced by the use of less basic
`amines:
`the pK value of triethylamine is 10.72,
`a
`while that of butyldimethylamine is 10.06 [23].
`Moreover, the concentration of acetonitrile necessary
`to elute the nucleic acids from the column is about
`threefold higher with butyldimethylammonium than
`triethylammonium ions, a consequence of the higher
`affinity of the former to the stationary phase. The
`concomitant decrease in surface tension and the
`increase in volatility of the electrosprayed eluent are
`responsible for an additional improvement in electro-
`spray ionization efficiency.
`The choice of counter-ion in the ion-pair reagent
`necessary to ensure permanent protonation of the
`amine has been shown to exert some effect on the
`
`12
`
`resolution of short oligonucleotides [24]. The res-
`olution between p(dT)
`and p(dT)
`, for instance,
`13
`decreased
`in
`the
`order
`triethylammonium
`bicarbonate.triethylammonium formate.triethyl-
`ammonium chloride.triethylammonium acetate.
`However, the separation efficiency for longer oligo-
`nucleotides such as p(dT)
`and p(dT) was similar
`30
`29
`with all four ion-pair reagents and the observed
`differences in resolution were all within the ex-
`perimental error.
`Contrary to earlier reports in the literature [25],
`detectability in ESI-MS was not found to improve
`with increasing volatility of the acidic counter-ion.
`To the contrary, the total ion current decreased in the
`order acetate.bicarbonate.formate.chloride. The
`suppression of oligonucleotide signal upon addition
`of the acids is most likely due to the competition of
`anions for ionization. Since ESI generates a roughly
`constant ion current, an increase in the signal intensi-
`ty from added acids will reduce intensity of the
`oligonucleotide ions. Moreover, ions of higher con-
`ductivity will be more efficient in signal suppression.
`Accordingly, chloride is the most efficient, while
`acetate is the least efficient anion in signal suppres-
`sion. Ideally for mass spectrometric detectability, one
`should employ a mobile phase free of acid. However,
`this is not practicable as deprotonation of triethylam-
`monium ion above pH 8.50 leads to a rapidly
`increasing loss of chromatographic retention and
`resolution (Fig. 3).
`Shallower gradients allow higher resolution of
`double-stranded DNA fragments
`in ion-pair
`re-
`versed-phase HPLC. For
`instance, a gradient of
`0.34% acetonitrile per minute yielded a 9.6–21.7%
`higher resolution of DNA restriction fragments in the
`size range of 100–600 bp compared to a gradient of
`0.4% acetonitrile per minute [18]. Increasing column
`temperature also exerts a beneficial effect on res-
`olution [5]. However, at temperatures .50 8C the
`size-dependence of separation is lost due to the onset
`of denaturation (Fig. 2a) [26,27].
`
`4 . Applications of non-denaturing ion-pair
`reversed-phase HPLC
`
`The separation, sizing and preparative fractiona-
`tion of DNA fragments are routine laboratory tasks
`
`5
`
`

`

`32
`
`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`transcripts by slab gel electrophoresis had failed to
`identify the heteroduplices formed between mixed
`strands of native and competitor amplicons, resulting
`in inaccurate gene quantification. Taking account of
`the heteroduplex formed obviated the need for
`titration of known RNA inputs. Subsequently, it was
`demonstrated that differences in secondary structure
`between native and competitor transcripts gave rise
`to reproducible differences in reverse transcription
`efficiency, while they were not found to affect PCR
`efficiency. Although competitive PCR in combina-
`tion with HPLC may be more accurate in measuring
`gene expression in low abundance samples such as
`microdissected nephron segments, it has been com-
`pletely replaced by real-time RT-PCR that can be
`performed in a single step [29]. It was not until the
`successful detection of single-base mismatches with
`a sensitivity close to 100% [27,30] that HPLC of
`DNA became an established technique that is now
`practised worldwide in clinical and genetic laborator-
`ies.
`
`5 . Denaturing high-performance liquid
`chromatography (DHPLC)
`
`5 .1. Behavior of double-stranded DNA at elevated
`column temperatures
`
`The crucial observation that led to the develop-
`ment of partially denaturing HPLC was the loss of
`size-dependent separation of DNA restriction frag-
`ments at temperatures .50 8C. It was noticed that
`fragments rich in AT-base pairs, which form only
`two hydrogen bonds compared to the three formed
`by GC-base pairs, were the first ones to shift toward
`shorter retention times. Hence, a change in base
`composition as small as the replacement of a single
`GC-base pair with an AT-base pair may be detected
`due to a change in retention time. Indeed, this can be
`observed [1], but in most instances, particularly with
`increasing fragment length, the differences in melt-
`ing behavior are too subtle to cause a detectable shift
`in retention.
`Far more pronounced changes in melting and,
`hence, retention behavior can be expected from base
`pair mismatches, i.e. base pairs other than the typical
`Watson–Crick base pairs AT and GC, as they form
`
`Fig. 3. Influence of pH on retention and resolution of oligodeox-
`ythymidylic acids. Column, 2.1 mm PS–DVB-C , 7030.20 mm
`18
`I.D.; mobile phase, (A) 0.1 M triethylammonium acetate, pH
`6.80–10.40, (B) 0.1 M triethylammonium acetate, pH 6.80–10.40,
`20% acetonitrile; linear gradient, 25–60% B in 12 min; flow-rate,
`2.3 ml/min; temperature, 50 8C; detection, UV, 254 nm; sample,
`p(dT)
`, 0.70 ng each.
`
`12 – 30
`
`and have been performed for decades by electro-
`phoresis using either agarose or polyacrylamide slab
`gels. With the exception of
`the purification of
`synthetic oligonucleotides, neither the high resolu-
`tion, with fragments differing only 1–5% in chain
`length up to 500 base pairs being resolved success-
`fully, nor the high accuracy of sizing, or the degree
`of automation have provided enough incentive to
`offset such obvious disadvantages as the cost of
`HPLC instrumentation and the comparatively low
`throughput.
`One of its more promising applications was the
`quantitative analysis of competitive reverse transcrip-
`tase polymerase chain reactions
`(RT-PCR)
`for
`quantification of gene expression [28]. It could be
`demonstrated that previous attempts at quantitatively
`determining the amounts of native and competitive
`
`6
`
`

`

`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`33
`
`less hydrogen bonds which makes them
`no or
`thermally far less stable. Such mismatches are gener-
`ated typically in vitro by PCR amplification of the
`same region of two chromosomes that differ in
`sequence [31]. The amplicons representing the two
`chromosomes are then mixed at equimolar ratio,
`denatured and renatured. During renaturation, not
`only the original homoduplices are formed but also
`heteroduplices between the sense and antisense
`strands of either homoduplex. Using four clones that
`differed in a single base (A, G, C, or T) at the same
`position,
`it was possible to generate all possible
`mismatches. These experiments established quickly
`that DHPLC was capable of detecting all single-base
`mismatches independent of their nature in fragments
`up to 1500 base pairs. This clearly demonstrated the
`superiority of DHPLC over slab-gel electrophoresis
`in the separation of homo- and heteroduplices
`[32,33]. The reasons for the superior performance of
`DHPLC appear to be a result of its greater resolving
`power and differences in affinity between intact
`double-stranded and single-stranded DNA fragments.
`Retention of the latter decreases more rapidly than
`that of the former with increasing column tempera-
`ture [1].
`
`5 .2. Partially denaturing HPLC
`
`Partially denaturing HPLC compares typically two
`or more chromosomes as a mixture of denatured and
`reannealed polymerase chain reaction (PCR) am-
`plicons. In the presence of a mutation in one of the
`two chromosomal fragments, not only the original
`homoduplices are formed again upon reannealing
`but, simultaneously, the sense and anti-sense strands
`of either homoduplex form heteroduplices that are
`thermally less stable. The more extensive but still
`partial denaturation of the heteroduplices at elevated
`temperatures,
`typically in the range of 50–70 8C
`depending on the GC-content of the DNA fragment
`under investigation, results in their reduced retention
`on the chromatographic separation matrix. As a
`consequence, one or more additional peaks appear in
`the chromatogram, with different mutations yielding
`in most but not all instances distinctively different
`peak profiles [1,34,35].
`Temperature is the most important experimental
`parameter affecting mutation detection sensitivity,
`
`and its optimum can be predicted by computation at
`the publicly available web site http:// insertion.stan-
`ford.edu/melt.html [36]. Single-nucleotide substitu-
`tions, deletions and insertions have been detected
`successfully within 2 to 3 min in unpurified am-
`plicons typically 200–1000 bp in length, with sen-
`sitivity and specificity of DHPLC consistently ap-
`proaching 100% [1]. Prerequisite, however, is the
`proper pre-conditioning of the DNA sample. This is
`accomplished by placing a heat exchanger made of
`80 cm of 0.01-in. I.D. PEEK tubing encased into a
`tin alloy block before the sample loop, both of which
`are kept in the oven. Alternatively, the sample can be
`injected at ambient temperature. In that case, the heat
`exchanger has to be placed between the injection
`valve and the column. The concomitant increase in
`extra-column volume will decrease resolution. In
`practice, however, the effect is too small to impact
`the ability of DHPLC to resolve homo- and
`heteroduplices.
`
`5 .3. Sensitivity and specificity of DHPLC
`
`As mentioned above, the choice of column tem-
`perature is critical
`in ensuring high sensitivity of
`DHPLC in mutation detection. Originally, the op-
`timum temperature at which to screen a particular
`DNA sequence was determined empirically by in-
`jecting repeatedly a test sample at gradually increas-
`ing column temperatures until the duplex product
`peak was retained about a minute less than at 50 8C.
`At this point, the presence of a single mismatch will
`be usually detected by the appearance of one to three
`additional peaks. Whether one or two heteroduplex
`and homoduplex peaks, respectively, are observed
`depends on several factors. They include the in-
`fluence of nearest neighbor sequence on the stability
`of base pair mismatches [37] and hydrogen bonding
`between non Watson–Crick base oppositions such as
`G–T and G–A [38]. In addition, temperature may
`affect the chromatographic profile with as little a
`difference in temperature as 2 8C resulting either in
`the separation of all four species, i.e. both the two
`hetero- and the two homoduplices, or
`just
`the
`separation of the two hetero- or the two homodup-
`lices, while the corresponding homo- and heterodup-
`lices elute as one peak. In the presence of more than
`one mismatch,
`the number of heteroduplex peaks
`
`7
`
`

`

`34
`
`P.J. Oefner, C.G. Huber / J. Chromatogr. B 782 (2002) 27–55
`
`observed may be greater than two based on the
`extent of denaturation.
`The empirical approach of determining the appro-
`priate temperature of analysis, however, harbors the
`risk that mismatches in low-melting AT-rich do-
`mains may go undetected due to complete denatura-
`tion. Further,
`it
`impedes automation and sample
`throughput. For these reasons, an algorithm was
`developed that calculates for every site in a known
`sequence the temperature at which 50% of
`the
`fragments are closed. Analysis is routinely per-
`formed at
`the highest of all site temperatures.
`Analysis is repeated in 4 8C decrements if
`the
`melting temperatures predicted span more than 4 8C
`or multitudes thereof [36]. Using such strategy, 165
`out of 166 different polymorphic DNA fragments
`that had been part of a total of 476 fragments
`screened with an average length of 563 bp were
`readily detected [39].
`The ability of DHPLC to detect mismatches
`appeared independent of the number and nature of
`mismatches—one fragment contained as many as 20
`base substitutions and insertions/ deletions—and the
`sequence flanking the mismatch. The only poly-
`morphic fragment that had escaped detection con-
`tained a single mismatch the resultant homo- and
`heteroduplexes of which resolved only over a tem-
`perature range as narrow as 2 8C. This is rather
`unusual based on a study of 103 mutations in 42
`different sequence contexts [36]. The median number
`of temperatures at which heterozygosity could be
`detected was eight, and the range was 4–11 8C. The
`values represent minimum values, as the samples
`were not analyzed beyond 5 8C on either side of the
`temperature recommended by the algorithm.
`The predictive power of the melting algorithm is
`excellent. A blinded analysis of 103 mutations
`showed that all but four could be readily detected at
`the temperatures recommended by the algorithm
`[36]. The mutations missed, however, could be
`detected successfully by increasing column tempera-
`ture by 2 8C. At present, the number of mutations
`whose temperature for detection has been predicted
`incorrectly is still too small to deduce conclusively a
`denominator that could lead to an appropriate adjust-
`ment of the algorithm. Though a recent study of 18
`mutations in exon 1 of the VHL tumor suppressor
`the WAVEMaker姠 soft-
`gene [40] suggested that
`
`ware, which is commercially available from Trans-
`genomic (Omaha, NE, USA), might predict melting
`temperatures more accurately (there were no dis-
`crepancies between the predictions made by the two
`algorithms for two other exons of the gene), this
`software has been never subjected to the same
`rigorous blinded analyses as the publicly available
`software. Hence, it remains to be evaluated whether
`improved prediction for certain domains is accom-
`plished at the expense of less accurate predictions for
`other domains. Finally, for reasons unknown,
`the
`study analyzed the samples at a temperature that was
`2 8C lower than the one actually recommended by
`the publicly available algorithm. In the meantime, it
`is recommended to determine empirically whether a
`higher temperature is required. This is accomplished
`by injecting the sample at a temperature 2 8C higher
`than the recommended one; if the DNA fragment
`elutes only slightly earlier (|0.5 min) than at the
`highest
`temperature
`recommended,
`the
`analysis
`should be repeated at the higher temperature. If the
`shift
`in elution time is greater,
`the analysis was
`already performed at the optimum temperature.
`The temperature displayed on the column oven
`does not necessarily correspond to the actual tem-
`perature in the column compartment. Measurements
`with certified temperature probes have shown devia-
`tions of up to 2 8C particularly with ovens calibrated
`at one temperature only. Moreover, in systems that
`use a mere coil