HUMANMUTATION 27(5),444^452,2006

RESEARCH ARTICLE

Denaturing Temperature Selection MayUnderestimate Keratin Mutation Detectionby DHPLC

Pavel Strnad,1,2 Tim Christian Lienau,1–3 Guo-Zhong Tao,1,2 Nam-On Ku,1,2 Thomas M. Magin,4

and M. Bishr Omary1,2�

1Department of Medicine, Palo Alto Veterans Affairs Medical Center, Palo Alto, California; 2Digestive Disease Center, School of Medicine,Stanford University, Stanford, California; 3Medizinische Klinik mit Schwerpunkt Hepatologie und Gastroenterologie, Charite, UniversitatsmedizinBerlin, Campus-Virchow-Klinikum, Berlin, Germany; 4University of Bonn, Institute of Physiological Chemistry, Bonn, Germany

Communicated by Peter Oefner

Keratins 8 and 18 (KRT8 and KRT18 genes; K8 and K18 proteins) variants are risk factors for developingend-stage liver disease and may be associated with inflammatory bowel disease and chronic pancreatitis. Thefrequency of K8/K18 variants in American, British, German, and Italian populations differs. For example,one study showed no amino acid–altering K8/K18 mutations in 256 German patients with liver disorders,while another found 58 out of 467 American liver disease patients with K8/K18 mutations. Both studies usedthe WaveSystemTM, which utilizes DHPLC. We hypothesized that experimental conditions contribute to thediscrepancy, and we tested this hypothesis using previously described K8/K18 variants and a novel KRT18c.1057C4G variant (K18 p.R353G) to optimize the DHPLC conditions in 10 examined exons under a range ofdenaturing temperatures. Six of 16 tested variants in three of the 10 exons, including the frequent KRT8c.184G4T (K8 p.G62C), KRT8 c.187A4G (K8 p.I63V), and KRT8 c.1022G4A (K8 p.R341H), could notbe reliably detected when using temperatures suggested by the prediction software, but all these variants werereadily detectable at 21C higher denaturing temperatures. Using optimized temperatures, we then testedavailable genomic DNA from 151 out of the 256 German liver disease patients for the presence of K8 variantsin exons 1 and 6, where most of the American cohort K8 variants occur. We identified 12 exonic and twointronic K8 variants: one KRT8 c.184G4T (K8 p.G62C), two KRT8 c.187A4G (K8 p.I63V), seven KRT8c.1022G4A (K8 p.R341H), one KRT8 c.1128G4A (K8 p.E376E), two intronic KRT8 c.1202146 A4T, andone hitherto undescribed KRT8 c.1138G4A (K8 p.V380I). Therefore, although DHPLC offers a robust andhigh throughput means for mutation analysis, assessment of denaturing temperature ranges, and possible inclusionof control mutants should be considered. Hum Mutat 27(5), 444–452, 2006. Published 2006 Wiley-Liss, Inc.y

KEY WORDS: KRT8; KRT18; chronic liver disease; cirrhosis

INTRODUCTION

It is widely accepted that the course of human diseases isinfluenced by the genetic background of their carriers [Kleebergerand Peden, 2005; Schreiber et al., 2005]. Therefore, analysis ofDNA sequence variations and their influence on the predisposi-tion to different pathologic conditions is an area of activeinvestigation. For this, numerous techniques are utilized includingdirect sequencing, denaturing gradient gel electrophoresis, single-strand conformational analysis, and DHPLC. Of these mutationdetection methods, DHPLC began to be applied in the mid-1990s[Xiao and Oefner, 2001] and quickly became a widely used tool fordiscovery of new genetic variants. The advantages of DHPLCinclude automation that allows for high-throughput analysis, andexcellent sensitivity and specificity that are reported to be greaterthan 96% [Xiao and Oefner, 2001]. DHPLC is based on theobservation that partial denaturation of a DNA duplex leads to asignificant decrease in its retention in ion-pair reversed-phasechromatography in comparison to nondenatured double-strandedDNA [Xiao and Oefner, 2001]. This principle can be employed to

identify heterozygous sequence variations, since reannealing of twononidentical DNA strands leads to formation of heteroduplexes,which are denatured earlier than their homozygote counterparts.

Published online 30 March 2006 in Wiley InterScience (www.interscience.wiley.com).yThis article is a US Government work and, as such, is in the publicdomain in the United States of America.

DOI10.1002/humu.20311

The Supplementary Material referred to in this article can beaccessed at http://www.interscience.wiley.com/jpages/1059-7794/suppmat.

Received 12 November 2005; accepted revised manuscript 5January 2006.

Grant sponsor:This work was supported by a Department ofVeter-ans A¡airs Merit Award and National Institutes of Health grantDK52951 (to M.B.O.), and National Institutes of Health DigestiveDisease Center grant; Grant number: DK56339.

�Correspondence to: Bishr Omary, Palo Alto VA Medical Center,Mail code154J,3801Miranda Avenue, PaloAlto, CA 94304.E-mail: mbishr@stanford.edu

PUBLISHED 2006 WILEY-LISS, INC.

Such samples are therefore retained less and form an additionalpeak when analyzed near the denaturing temperature.

In this study, we used DHPLC to detect sequence variations inthe human keratin 8 and 18 genes (KRT8 and KRT18 genes; K8and K18 proteins; MIM]s 148060 and 148070, respectively) ingenomic DNA isolated from patients with liver disease. Keratinsare intermediate filament (IF) cytoskeletal proteins that arespecifically expressed in epithelial cells [Moll et al., 1982;Coulombe and Omary, 2002]. All IF proteins consist of a centralcoil–coil a-helical ‘‘rod’’ domain that is flanked by non-a-helicalN-terminal ‘‘head’’ and C-terminal ‘‘tail’’ domains [Fuchs andCleveland, 1998; Omary et al., 2004]. Within the IF family,keratins are divided into type I (K9–K20) and type II (K1–K8).Epithelial cells express at least one type I and one type II keratinpair as obligate noncovalent heteropolymers in an epithelial cell-type-preferential manner (e.g., K3/K12 in corneal epithelia, K5/K14 and K1/K10 in keratinocytes depending on their differentia-tion state, and K8/K18 in glandular ‘‘simple-type’’ epithelia).Keratin mutations cause, or predispose to, a broad range ofepithelial diseases that reflect the cell-specific expression of a givenkeratin [Fuchs and Cleveland 1998; Irvine and McLean, 1999;Omary et al., 2004]. One feature that distinguishes human keratinmutations of simple epithelia (i.e., involving K8 and K18) fromkeratin mutations of non simple epithelia (e.g., involving K1–K6,K9, K10, K12–K14) is that the former predispose to but do notdirectly cause end-stage liver disease [Ku et al., 1997, 2001, 2003,2005] (MIM] 215600) or possibly inflammatory bowel disease[Owens et al., 2004] (MIM]s 601458 and 266600), while thelatter directly cause a variety of skin, oral, or ocular diseasesdepending on the involved keratin. One potential explanation forthis finding is that the location of simple epithelial keratinmutations is typically within the head or tail domains and does notinvolve the highly conserved beginning and end of the rod, whichare the most common mutation sites in nonsimple epithelialkeratins and are hypothesized to be lethal if found in K8/K18[Omary et al., 2002; Porter and Lane, 2003].

In humans, the overall frequency of K8/K18 variants in patientswith end-stage liver disease is 12.4% when studied in an Americancohort of patients and upon analysis of all exonic K8/K18 regions[Ku et al., 2005]. This variant frequency excludes polymorphismsfelt to be nonpathogenic since they were readily detectable in acontrol group derived from blood-bank volunteers [Ku et al.,2005]. However, analysis of a German cohort of patients withchronic liver disease (many of the patients had chronic hepatitiswith an unknown extent of liver disease) showed no K8 or K18amino acid–altering variants [Hesse et al., 2004]. A significantportion of the exons in the American patient cohort and all theexons in the German cohort were analyzed for the presence of K8/K18 variants by DHPLC [Hesse et al., 2004; Ku et al., 2001, 2003,2005]. Other European patients were reported to have K8/K18variants though the methods of analysis differed. For example, K8and/or K18 variants were identified in an English cohort by DNAsequencing of the entire coding regions of K8 and K18 [Owenset al., 2004] or by specifically searching for the presence of KRT8c.160T4C (K8 p.Y54H) and KRT8 c.184G4T (K8 p.G62C)SNPs in a German cohort with liver disease [Halangk et al., 2004]or inflammatory bowel disease [Buning et al., 2004], or in anItalian cohort with pancreatitis [Cavestro et al., 2003]. Wehypothesized that experimental conditions likely contributed tothe differences in K8/K18 variant frequency detection in theAmerican [Ku et al., 2005] and German [Hesse et al., 2004]cohorts, and undertook a detailed analysis to test this hypothesis.We used as controls previously described K8 and K18 variants that

were found in the American cohort with liver disease or that wereintroduced as transgenes into mice to test DHPLC conditionsunder a range of denaturing temperatures. We then used optimizedconditions to reexamine the frequency of variants in two relevantK8 exons in a subgroup of patients that were analyzed in thepreviously described German cohort [Hesse et al., 2004].

MATERIALSANDMETHODSGenomic DNA Samples

Genomic DNA from human liver explants, mouse tails, orperipheral blood was prepared using a DNeasy tissue kit (Qiagen,Valencia, CA; www.qiagen.com). The heterozygous KRT18c.1057C4G variant (K18 p.R353G) (Supplementary Fig. S2;available online at http://www.interscience.wiley.com/jpages/1059-7794/suppmat) was identified in one out of three children of aparent who was homozygous for the HFE H63D (MIM] 235200)variant (the affected father did not carry the K18 variant andDNA was not available from the mother). Genomic material forhuman KRT18 c.[100A4G;101G4C] (K18 p.S34A), KRT18c.[157A4G;158G4C] (K18 p.S53A), and KRT18 c.268C4T(K18 p.R90C) was obtained from transgenic mice that overexpressthe corresponding human transgenes [Ku et al., 1995, 1998, 2002].Genomic sequences (M34482.1 and AF179904.1 for K8 and K18,respectively) were used to design primers and analyze meltingprofiles of the amplified fragments, while mRNA sequences(NM_002273.2 and NM_000224.2, for K8 and K18, respectively)were employed to assign the location of the identified variants.

PCR Analysis

DNA fragments (201–481 bp; Table 1) were amplified with‘‘hot-start’’ Amplitaq Gold DNA Polymerase (Applied Biosystems;

TABLE 1. Sequences of Primers Used forAmpli¢cation, andLength andGCContent of theAmpli¢edProducts

Exon Primer Size (bp) %GC

KRT81F TGCCTCTACCATGTCCATCA 392 63.81R CGGGACTACCAGGAGAAAGG

4F TGGCAACTAGAAAGTCCTGTG 279 57.74R AGCCTCTGGTTGAGTCTCAGG

5F CACTTGCCCTCTTCCCCACAG 333 58.35R CACCCCCAACCCGGCCCATAC

6F CATACCCAACCTGACCTACTTACC 369 63.76R AGAACAACAGGACCCCAAGTC

8F TACCTCTGTCCCTCACCAGG 300 64.78R CTCCTGTTCCCAGTGCTACC

KRT181F CAAAGCCTGAGTCCTGTCCTTT 481 63.61R AGTTGAGGTCCCTCCTACCCCTTAC

2F CTGGCTTTCTATTCATGGAAC 201 51.92R AACTACCCAGCCTGGGGAGCA

4F CACTTTTGCCCCTGTCACCTTTAG 237 57.84R GTCTGCCTCCCTCCCACACCTT

5F CTGCCAAGGTGTGGGAGGGAG 211 57.35R AGGTGATGTGAAGGCACTCAC

6F CAGAAGGCCAGCTTGGAGAAC 270 61.96R ATCTCCTGATCCCAGCACGTG

HUMAN MUTATION 27(5),444^452,2006 445

Human Mutation DOI 10.1002/humu

Foster City, CA; www.appliedbiosystems.com). High amplificationspecificity of the PCR reactions were achieved by a touchdownPCR protocol and titration of the Mg21 concentration, andconfirmed by the presence of a single amplified product whentested by agarose gel electrophoresis (not shown). After the PCRamplification, samples were heated to 941C and gradually cooleddown to 651C over 25 min to ensure random recombinationof DNA strands. In some cases, DNA samples were mixed inequimolar ratio, heated to 941C and cooled down slowly to allowformation of heteroduplexes.

SoftwareAnalysis

Two software systems (WavemakerTM 4.1.44 and NavigatorTM;Transgenomic, Omaha, NE; www.transgenomic.com) were used,which provided similar temperature predictions when comparedside by side (not shown). For uniformity, all the temperaturesreported in this study were derived using the Wavemaker 4.1.44software.

DHPLC Analysis

DHPLC screening was performed using a WAVETM DNAFragment Analysis System (Transgenomic), with a nonporouspoly(styrene/divinyl-benzene) DnaSepTM column (Transgenomic).This system was previously used to screen for K8/K18 variants inthe American [Ku et al., 2005] and German [Hesse et al., 2004]cohorts. The composition of buffers was 0.1 M triethyl ammoniumacetate for buffer A, and 0.1 M triethyl ammonium acetatetogether with 25% acetonitrile for buffer B. Elution of bound DNAwas carried out at a flow rate of 0.9 ml/min and a buffer B gradientincrease of 2% per min. DNA concentration was estimated bymeasuring the absorbance at 254 nm. For sample screening, a2-min quick-time detection mode was used, whereas a prolongedelution gradient was chosen to monitor the elution patternthroughout the temperature spectrum and to assemble experi-mental melting profiles. The appropriate oven calibration formelting temperatures was done regularly with each use by applyingcommercially available mutation standards (Transgenomic). Thesestandards confirmed the correct oven calibration at 561C and701C. Samples with ‘‘shifted’’ elution patterns were purified with aQiaquick PCR purification kit (Qiagen) then sequenced in theforward and reverse directions using an ABI 377 sequencer(Applied Biosystems). All samples with a predicted mutationby DHPLC were reanalyzed by repeated PCR amplificationand DHPLC of the relevant amplicon to confirm validity ofthe findings.

RESULTSComparison of Predicted and ExperimentalTemperatures for DHPLCMutation Detection

In order to obtain optimal resolution of heterozygous DNAvariants by DHPLC, samples are typically analyzed at temperaturesleading to the formation of �75% helical fragments [Xiao andOefner, 2001]. All amino acid–altering K8 and K18 variantsidentified to date are primarily heterozygous or, in a few cases,compound heterozygous [Owens et al., 2004; Ku et al., 2005]. Wefocused on examining five K8 and five K18 exons (Fig. 1), which iswhere most of the K8/K18 variants reported to date are located[Owens et al., 2004; Ku et al., 2005]. For each of the 10 analyzedexons, we used previously reported genomic material from liverexplants of patients that are known to harbor the mutations shownin Figure 1 [Ku et al., 2005]. In the case of K18 exon 1, we used asvariant controls three previously studied human mutations that

were introduced into transgenic mice (K18 p.S34A, p.S53A, orp.R90C; these variants have not been described in humans). Inthe case of K18 exon 6, we used as a variant control a hithertounreported mutation (K18 p.R353G) that we identified in anapparently healthy donor (see Materials and Methods andSupplementary Fig. S2). R353 resides in subdomain 2B of theK18 rod domain (Fig. 1) and is highly conserved in all reportedK18 sequences from human to Xenopus (Supplementary Fig. S3)but not in other type I keratins (not shown).

The DHPLC system is equipped with WavemakerTM 4.1.44software that provides an estimate of the appropriate analysistemperature. It predicts the helical fraction of an entire DNAamplicon over a temperature spectrum or at specific temperaturesas exemplified for the K8 exon 1 amplified product (Fig. 2A) or forthe K8 exon 5 amplified product (Fig. 3A). Detectability of avariant is most likely when the helical fraction is between�40–90%. Although the predicted overall melting profiletemperature for K8 exons 1 and 5 is identical (i.e., 62.61C;Figs. 2A and 3A, upper panels), the presence of a significantpredicted helical fraction varies widely among different regionsof the K8 exon 1 and exon 5 amplicons (Figs. 2A and 3A, lowerpanels). For example, the software-generated profiles of K8 exon 5at 61.5 or 62.51C (i.e., at temperatures close to the overallpredicted temperature of 62.61C) show more than 50% helicalfraction over nearly the entire amplicon (Fig. 3A, lower panel),while the higher temperature of 65.71C provides a predictedadequate helical fraction only for nucleotides 50–210 of K8 exon 1(Fig. 2A, lower panel). When the variant sequences of KRT8c.187A4G (K8 p.I63V) and K8 p.Y54H are analyzed by thesoftware, the predicted temperatures for 75% helical fraction are64.61C and 65.71C, respectively (Fig. 2B), while the predictedtemperatures for 75% helical fraction for the KRT8 c.955G4Tvariant (K8 p.A319S) is 62.51C (Fig. 3B). Experimental analysis ofthe overall melting profiles of wild-type K8 exon 1 demonstrates amelting temperature of 65.71C (Fig. 2C).

The above two examples of exons 1 and 5 of K8 highlight thepotential discrepancy between software and experimental predic-tions of the helical fraction temperature profile of a given amplifiedamplicon. Using the above approaches, we tested a range oftemperatures to compare the profile of wild-type K8 with theprofiles of K8 p.Y54H and p.I63V (exon 1) and K8 p.A319S (exon5). We examined the elution time and profile of DNA samplesover a range of temperatures and generated melting profiles, sinceelution time is proportional to the denaturation level; specimensthat are 100% helical at 501C become completely denatured at751C. The K8 p.A319S variant can be seen as a clear double peak(vs. a peak with a shoulder for WT K8) at the predictedtemperature of 62.51C (Fig. 3C), though a shifted single peak forsome amplicons (not shown) and detection temperatures can beequally informative. In contrast, the K8 p.Y54H variant wasclearly detectable at 67.5–691C (Fig. 2D) vs. the predicted 65.71C(Fig. 2B,C). The temperature discrepancy is not due to incorrectoven calibration since the oven temperature is assessed regularlyusing commercially available mutation standards (not shown, seealso Materials and Methods). Complete analysis of the profiles of16 K8 and K18 variants (which include the most commonpolymorphisms and the liver disease–related variants identified todate [Ku et al., 2005]) that are found in 10 exons are summarizedin Table 2. Of the 10 analyzed exons, K8 exons 1 and 6 and K18exon 1 had the largest gap (1.7–3.41C) between the predicted andexperimentally observed evidence for the presence of a variantbased on DHPLC profiles, while the remaining seven exons hadgood concordance (Table 2).

446 HUMAN MUTATION 27(5),444^452,2006

Human Mutation DOI 10.1002/humu

Retesting for the Presence of K8 Variantsin a German Cohort

Three of the 10 analyzed exons (K8 exons 1 and 6, and K18exon 1) have a wide discrepancy between the predicted versusexperimental melting temperature profiles (Table 2), which canhave a dramatic effect on the ability of DHPLC to identify a

potential keratin variant (Fig. 2). Based on this observation, wereanalyzed available genomic DNA from 151 of the 256 Germanpatients with liver disease who (together with 100 controls) werereported to lack K8 or K18 amino acid–altering variants [Hesseet al., 2004]. For this analysis, we focused on K8 exons 1 and 6,since these exons harbor the three most common K8 and K18variants identified to date. This includes K8 p.G62C and p.Y54H

FIGURE 1. Distribution of keratin variants within the K8 and K18 exons selected for analysis. A schematic of K8 and K18 proteindomains (Head, Rod,Tail) and location of the variants (based on sequences NM_002273.2 and NM_000224.2 for K8/K18, respec-tively) that were used to compare the predicted vs. the mutation detection temperatures.The rod domain is divided into subdomains(1A,1B,2A,2B) that are connected by short nonhelical linker sequences (L1, L12, L2). Also shown are the ampli¢ed exons and theircorresponding amino acid spans.The K18 variants enclosed in boxes (K18 p.S34A, p.S53A, and p.R90C) have not been found inhumansbutwereampli¢ed fromhumanK18mutantgenes thatwereused togenerate transgenicmice for functional studies [Kuet al.,1995,1998,2002].The K18 p.R353G variant (highlighted by an asterisk) is a new variant.The schematic lists codon numbers inclu-sive of that of the ¢rst methionine. ForDNA nomenclature of tested variants seeTable 2.

HUMANMUTATION 27(5),444^452,2006 447

Human Mutation DOI 10.1002/humu

in exon 1, which together were found in 11 out of 467 (2.4%)patients with liver disease, and KRT8 c.1022G4A (K8 p.R341H)in exon 6, which was found in 6.4% of the patients [Ku et al., 2005].

Using similar analysis conditions to those described in Figure 2and Table 2, we identified 12 exonic and two intronic K8 variants:one K8 p.G62C, two K8 p.I63V, seven K8 p.R341H, one KRT8c.1128G4A (K8 p.E376E), two intronic KRT8 c.1202146 A4T,and one novel KRT8 c.1138G4A (K8 p.V380I), which has notbeen previously described (Fig. 4). The biologic impact of K8p.V380I is unknown but K8 V380 is highly conserved across allspecies and in all type II keratins (except for K1 and K5) and inother cytoplasmic IF proteins (Supplementary Fig. S3). We alsotested whether additional differences in experimental conditionsmay have been responsible for the initial lack of detection of K8/K18 variants in the German cohort. For example, we used primersidentical to those previously used [Hesse et al., 2004] to amplifythe region coding for K8 exon 6 (which was only 5–10 basesremoved from the regions amplified by the primers we used in

Table 1) and found identical results to those in Figure 4 (notshown). Furthermore, to mimic the previously used conditions, wemixed heterozygous variants with wild-type samples and allowedthem to hybridize prior to analysis, and found that the K8 variantscould be reliably detected (not shown). Therefore, keratin variantsalso occur in German patients with liver disease, and the choiceof analysis temperatures can be critical for variant detectionby DHPLC.

DISCUSSION

The three major conclusions of this study are as follows: 1) fullassessment should be exercised in choosing detection temperaturesfor DHPLC-based analysis of potential mutations, to ensure thatvariants are not overlooked; 2) keratin variants are indeed presentin a German cohort of patients with chronic liver disease,including the two most common variants K8 p.R341H and

FIGURE 2. Analysis of variantofK8 exon1.A:Thehelical fractionofK8 exon1throughout a temperature spectrum (upper histogram)and at three chosen temperatures (lower histogram) are shown, as predicted by the software. Arrowheads in the lower histogramrepresent the K8 variants p.Y54H (left arrowhead at base position 170) and p.I63V (right arrowhead at base position 197). B:Thehelical fractionofK8 p.Y54H (uppergraph) andp.I63Vvariants (lowergraph) plotted as a functionof temperature aspredictedusingthe software. C: Amelting curve is generated experimentally after DHPLC analysis of ampli¢edwild type humanK8 exon1. Elutiontime at 501C represents100% helical DNA,whereas elution time at 751C represents fully denaturedDNA (0% helical fraction). Each¢lled square along the curve represents an increment of 11C (58^751C).This experimental melting curve predicts an approximately31C higher melting curve than the predicted curve shown in A. D: Genomic DNA was isolated from liver explants that expresswild-type, p.Y54H, or p.I63V heterozygous K8, followed by ampli¢cation of K8 exon 1, then analysis by DHPLC at the indicatedtemperatures (50^761C). Note that both the p.Y54H and p.I63V K8 variants (and the p.G62C variant, not shown but see Table 2)cannot be reliably detected at 65.7 and 64.61C, respectively, as suggested by the software, but formed a characteristic double-peakwhen analyzed at 2^41C higher temperatures. Position of K8 variants is based on sequence NM_002273.2. [Color ¢gure can beviewed in the online issue,which is available at www.interscience.wiley.com.]

448 HUMAN MUTATION 27(5),444^452,2006

Human Mutation DOI 10.1002/humu

p.G62C; and 3) two new variants, K8 p.V380I and K18 p.R353G,were identified as part of our analysis.

DHPLC is clearly a robust technique that allows rapid mutationscreening, but some caveats deserve highlighting. For example,amplicon length and GC content can be confounding variables. Inthis regard, K8 exons 1 and 6 and K18 exon 1, which had the mostsignificant difference in predicted vs. detection temperature, werethe longest and tended to have the highest GC content (Table 1)in contrast with the remaining seven analyzed exons that hadcomparably reliable predicted and experimental temperatures(Table 2). The determination of analysis temperature is also

known to be challenging in isolated high melting domainsembedded in low melting DNA regions, as is the case for theK18 p.S53A variant (see also Breton et al. [2003]). However, allvariants could be reliably detected by adding a second analysistemperature 21C higher than those recommended by softwareprediction, which may be necessary [Jones et al., 1999; Ravnik-Glavac et al., 2002]. Since the vast majority of variants can bedetected over a temperature window of at least 21C [Jones et al.,1999], adding an additional analysis temperature is an effectiveway to increase DHPLC sensitivity. Therefore, DHPLC analysis atmultiple temperatures is necessary for amplicons with hetero-

FIGURE 3. Analysis of variants of K8 exon 5. A:The helical fraction of K8 exon1 throughout a temperature spectrum (upper histo-gram) and at three chosen temperatures (lower histogram) are shown, as predicted by theWavemakerTM 4.1.44 software.The arrow-head in the lower histogram highlights base position 286 of K8 p.A319S. B:The software-predicted helical fraction of K8 p.A319S(based on sequence NM_002273.2) is plotted as a function of temperature. C: Genomic DNA was isolated from liver explants thatexpresswild-typeor p.A319SK8, followed by ampli¢cationofK8 exon 5, thenexaminationbyDHPLCat the indicated temperatures(50^751C). Note that the p.A319S K8 variant can be reliably detected at the temperature suggested byWavemaker 4.1.44 software(62.5^63.51C). [Color ¢gure can be viewed in the online issue,which is available at www.interscience.wiley.com.]

TABLE 2. Comparison of Predicted andDetectionTemperatures in SelectedK8/K18 Variantsy

Variants

Gene Exon] Amino acid Nucleotide Predicted temp. (1C) Detection temp. (1C)

K8 E1 p.Y54H c.160T4C 65.7� 68p.G62C c.184G4T 64.8� 68p.163V c.187A4G 64.6� 68

E4 p.L227L c.681A4G 62.2 61.5E5 p.A319S c.955G4T 62.5 63.5E6 p.R341H c.1022G4A 64� 65.7

p.E376E c.1128G4A 64.3 65.7E8 p.G434S c.1300G4A 64.8 64.8

p.V480I c.1438G4A 64.6 64.8K18 E1 p.S34A c.[100A4G;101G4C] 66.9 69

p.S53A c.[157A4G;158G4C] 65.7� 69p.R90C c.268C4T 62.6� 64

E2 p.I150V c.448A4G 61.7 61.5E4 p.S230T c.689G4C 62.5 62.5E5 p.T297I c.890C4T 63.8 62.6E6 p.R353G c.1057C4G 65.1 66

yTheWavemaker softwarewas used to predict the detection temperatures of16 exonic variants from ¢veK8 and ¢veK18 exons (see Fig.1 for display ofthe exons).The position of K8/K18 variants is based on sequences NM_ 002273.2 and NM_ 000224.2, respectively, with 11 as A of ATG start codon.Genomic DNAwas then used to test the optimal temperature for mutation detection as carried out in Figures 2 and 3.�These variants could not be reliably recognized using the predicted temperatures. Detection temperature is the temperature that allowed unequivocaldiscrimination betweenwild-type and variant samples.

HUMANMUTATION 27(5),444^452,2006 449

Human Mutation DOI 10.1002/humu

geneous melting profiles, which are also the most challenging to beanalyzed correctly (Figs. 2A and 3A; Supplementary Fig. S1).In fact, we are currently using four temperatures in increments of21C to ensure our confidence of complete analysis of K18 exon 1.In addition, generation of an experimental melting curve tocomplement the software data is strongly recommended. For K8,the experimental melting curve for exon 1 suggests that the DNAmelting may occur at approximately 21C higher temperature than

predicted by the software analysis. When feasible, the use of invitro generated mutants, as done in this report for the K18 headdomain, offers a significant advantage to control for theappropriateness of selected analysis temperatures.

The second major finding of this study involved testing thehypothesis that differences between predicted and experimentaltemperature melting conditions were a major reason why K8/K18variants were not initially identified in a German cohort of patients

FIGURE 4. K8 variants identi¢ed in theGerman patient cohort. A: A summary of the intronic and exonic variants found upon ampli-¢cation of genomic DNA regions inclusive of exons 1 and 6, isolated from 151 patients of the German cohort that was previouslyanalyzed [Hesse et al., 2004]. For comparison, results from theAmerican cohort [Ku et al., 2005] are included. �5all the variants inK8 exons 1 and 6, some of which were not observed in the German cohort; ��5a newly identi¢ed variant (see also SupplementaryFigs. S2 and S3). B,C: Anexample of theDHPLC elution pattern (B) and sequencing results (C) of ampli¢ed genomicmaterial fromtwoGermanpatients,onewithwild-typeK8 exon 6 andonewith aheterozygousK8 p.V380I variant.ThepositionofK8/K18 variantsis based on sequencesNM_002273.2 andNM_000224.2, respectively.

450 HUMANMUTATION 27(5),444^452,2006

Human Mutation DOI 10.1002/humu

with chronic liver disease [Hesse et al., 2004]. It is possible thatadditional variants in other exons are present in that cohort butwe focused on the two K8 exons in which most of the variantshave been reported [Ku et al., 2005]. We were not able to carryout an exhaustive study of all the exons due to the lack of availablegenomic DNA, and since this aspect of the study was more of aproof-of-principle. Although population differences may indeedexist in terms of the frequency of specific K8 or K18 variants,it was the absence of any variants that led us to explore thediscrepancy which can now be readily explained since thetemperature range used by Hesse et al. [2004] was below theoptimal detection temperature for all the major mutants found inK8 exon 1 and 6. Other variables that may be related to PCRamplification conditions could contribute to differences inmutation detection via masking of a small shoulder or peakbroadening that may be indicative of a heterozygous variant.In addition, mixing of a wild-type sample (determined by completesequencing) with the test samples, as used by Hesse et al. [2004]to potentially optimize identification of homozygous variants mayrender heterozygous variant identification at nonoptimal analysistemperature more difficult, although DHPLC should detectvariants at frequencies lower than 5% [Wolford et al., 2000].In addition to our findings, population differences for somevariants may also occur as noted for K8 p.Y54H and KRT8c.1300G4A (K8 p.G434S), which were found at a higherfrequency in those of African origin in the American cohort[Ku et al., 2005], though further confirmatory studies are stillrequired.

The disease-related and biologic significance of the two newlyidentified variants remains to be defined. K8 p.V380I is highlyconserved in K8 across species and also in multiple cytoplasmic IFproteins; while K18 p.R353G is conserved in K18 across speciesbut not among other type I keratins (Supplementary Fig. S3).These variants are rare, in that they were not identified in 467liver disease patients or in 349 controls [Ku et al., 2005], nor in acohort of 140 patients with inflammatory bowel disease and 100controls [Owens et al., 2004]. As more disease and control groupsare analyzed for K8 and K18 variants, the status of and need topursue these rare variants should become apparent. Searching forkeratin and nonkeratin mutations by DHPLC remains an effectivetool, provided that assessment of denaturing temperature rangesand possible inclusion of control mutants are considered.

ACKNOWLEDGMENTS

We thank Nahid Madani for assistance with DNA sequencing,Michael Hesse for helpful suggestions, Kris Morrow for expertassistance with figure preparation, and all the patients and theirfamilies who participated and made this work possible. P.S. issupported by a European Molecular Biology Organization long-term postdoctoral fellowship. G-Z.T. is supported in part by aCrohn’s and Colitis Foundation of America Research Award.T.C.L. is supported by a scholarship from Humboldt University ofBerlin, Germany; and part of this work will be contained in hisdissertation.

REFERENCES

Breton J, Sichel F, Abbas A, Marnay J, Arsene D, Lechevrel M.2003. Simultaneous use of DGGE and DHPLC to screen TP53mutations in cancers of the esophagus and cardia from aEuropean high incidence area (Lower Normandy, France).Mutagenesis 18:299–306.

Buning C, Halangk J, Dignass A, Ockenga J, Deindl P, Nickel R,Genschel J, Landt O, Lochs H, Schmidt H, Witt H. 2004.Keratin 8 Y54H and G62C mutations are not associated withinflammatory bowel disease. Dig Liver Dis 36:388–391.

Cavestro GM, Frulloni L, Nouvenne A, Neri TM, Calore B, FerriB, Bovo P, Okolicsanyi L, Di Mario F, Cavallini G. 2003.Association of keratin 8 gene mutation with chronic pancrea-titis. Dig Liver Dis 35:416–420.

Coulombe PA, Omary MB. 2002. ‘Hard’ and ‘soft’ principlesdefining the structure, function and regulation of keratinintermediate filaments. Curr Opin Cell Biol 14:110–122.

Fuchs E, Cleveland DW. 1998. A structural scaffolding ofintermediate filaments in health and disease. Science 279:514–519.

Halangk J, Berg T, Puhl G, Mueller T, Nickel R, Kage A, Landt O,Luck W, Wiedenmann B, Neuhaus P, Witt H. 2004. Keratin 8Y54H and G62C mutations are not associated with liver disease.J Med Genet 41:e92.

Hesse M, Berg T, Wiedenmann B, Spengler U, Woitas RP, MaginTM. 2004. A frequent keratin 8 p.L227L polymorphism, but nopoint mutations in keratin 8 and 18 genes, in patients withvarious liver disorders. J Med Genet 41:e42.

Irvine AD, McLean WHI. 1999. Human keratin diseases: theincreasing spectrum of disease and subtlety of the phenotype-genotype correlation. Br J Dermatol 140:815–828.

Jones AC, Austin A, Hansen N, Hoogendoorn B, Oefner PJ,Cheadle JP, O’Donovan MC. 1999. Optimal temperatureselection for mutation detection by denaturing HPLC andcomparison to single-stranded conformation polymorphism andheteroduplex analysis. Clin Chem 45:1133–1140.

Kleeberger SR, Peden D. 2005. Gene-environment interactions inasthma and other respiratory diseases. Annu Rev Med 56:383–400.

Ku NO, Michie S, Oshima RG, Omary MB. 1995. Chronichepatitis, hepatocyte fragility, and increased soluble phospho-glycokeratins in transgenic mice expressing a keratin 18conserved arginine mutant. J Cell Biol 131:1303–1314.

Ku NO, Wright TL, Terrault NA, Gish R, Omary MB. 1997.Mutation of human keratin 18 in association with cryptogeniccirrhosis. J Clin Invest 99:19–23.

Ku NO, Michie SA, Soetikno RM, Resurreccion EZ, Broome RL,Omary MB. 1998. Mutation of a major keratin phosphorylationsite predisposes to hepatotoxic injury in transgenic mice. J CellBiol 143:2023–2032.

Ku NO, Gish R, Wright TL, Omary MB. 2001. Keratin 8mutations in patients with cryptogenic liver disease. N Engl JMed 344:1580–1587.

Ku NO, Michie S, Resurreccion EZ, Broome RL, Omary MB.2002. Keratin binding to 14-3-3 proteins modulates keratinfilaments and hepatocyte mitotic progression. Proc Natl AcadSci USA 99:4373–4378.

Ku NO, Darling JM, Krams SM, Esquivel CO, Keeffe EB,Sibley RK, Lee YM, Wright TL, Omary MB. 2003. Keratin 8and 18 mutations are risk factors for developing liver diseaseof multiple etiologies. Proc Natl Acad Sci USA 100:6063–6068.

Ku NO, Lim JK, Krams SM, Esquivel CO, Keeffe EB, Wright TL,Parry DAD, Omary MB. 2005. Keratins as susceptibility genesfor end stage liver disease. Gastroenterology 129:885–893.

Moll R, Franke WW, Schiller DL, Geiger B, Krepler R. 1982.The catalog of human cytokeratins: patterns of expressionin normal epithelia, tumors and cultured cells. Cell 31:11–24.

Omary MB, Ku NO, Toivola DM. 2002. Keratins: guardians of theliver. Hepatology 35:251–257.

HUMANMUTATION 27(5),444^452,2006 451

Human Mutation DOI 10.1002/humu

Omary MB, Coulombe PA, McLean WH. 2004. Intermediatefilament proteins and their associated diseases. N Engl J Med351:2087–2100.

Owens DW, Wilson NJ, Hill AJ, Rugg EL, Porter RM, HutchesonAM, Quinlan RA, Van Heel D, Parkes M, Jewell DP,Campbell SS, Ghosh S, Satsangi J, Lane EB. 2004. Humankeratin 8 mutations that disturb filament assembly observedin inflammatory bowel disease patients. J Cell Sci 117:1989–1999.

Porter RM, Lane EB. 2003. Phenotypes, genotypes and theircontribution to understanding keratin function. Trends Genet19:278–285.

Ravnik-Glavac M, Atkinson A, Glavac D, Dean M. 2002. DHPLCscreening of cystic fibrosis gene mutations. Hum Mutat19:374–383.

Schreiber S, Rosenstiel P, Albrecht M, Hampe J, Krawczak M.2005. Genetics of Crohn disease, an archetypal inflammatorybarrier disease. Nat Rev Genet 6:376–388.

Wolford JK, Blunt D, Ballecer C, Prochazka M. 2000. High-throughput SNP detection by using DNA pooling anddenaturing high performance liquid chromatography (DHPLC).Hum Genet 107:483–487.

Xiao W, Oefner PJ. 2001. Denaturing high-performance liquidchromatography. A review. Hum Mutat 17:439–474.

452 HUMAN MUTATION 27(5),444^452,2006

Human Mutation DOI 10.1002/humu